d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
7f86780195
freebsd-skq/sys/dev/ray/if_ray.c

3831 lines
106 KiB
C
Raw Normal View History

Move the $NetBSD$ tag to the top of the file, per standard NetBSD practice (which we seem to mostly follow in the tree). Move the $FreeBSD$ tag to its more proper place after all copyright and license notices. Add '-' to the copyright notice for Christian E. Hopps so my copyright script picks it up.
2005-02-06 21:15:19 +00:00
/* $NetBSD: if_ray.c,v 1.12 2000/02/07 09:36:27 augustss Exp $ */
Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
* Copyright (C) 2000
* Dr. Duncan McLennan Barclay, dmlb@ragnet.demon.co.uk.
*
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY DUNCAN BARCLAY AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL DUNCAN BARCLAY OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*
*/
Move the $NetBSD$ tag to the top of the file, per standard NetBSD practice (which we seem to mostly follow in the tree). Move the $FreeBSD$ tag to its more proper place after all copyright and license notices. Add '-' to the copyright notice for Christian E. Hopps so my copyright script picks it up.
2005-02-06 21:15:19 +00:00
/*-
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
* Copyright (c) 2000 Christian E. Hopps
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
Move the $NetBSD$ tag to the top of the file, per standard NetBSD practice (which we seem to mostly follow in the tree). Move the $FreeBSD$ tag to its more proper place after all copyright and license notices. Add '-' to the copyright notice for Christian E. Hopps so my copyright script picks it up.
2005-02-06 21:15:19 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
/*
* Card configuration
* ==================
*
* This card is unusual in that it uses both common and attribute
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* memory whilst working. It should use common memory and an IO port.
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
* The bus resource allocations need to work around the brain deadness
* of pccardd (where it reads the CIS for common memory, sets it all
* up and then throws it all away assuming the card is an ed
* driver...). Note that this could be dangerous (because it doesn't
* interact with pccardd) if you use other memory mapped cards in the
* same pccard slot as currently old mappings are not cleaned up very well
* by the bus_release_resource methods or pccardd.
tweak comments about OLDCARD workarounds
2006-10-19 05:25:53 +00:00
* XXX Are these workarounds necessary with NEWCARD?
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
* Ad-hoc and infra-structure modes
* ================================
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* The driver supports ad-hoc mode for V4 firmware and infrastructure
* mode for V5 firmware. V5 firmware in ad-hoc mode is untested and should
* work.
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
* The Linux driver also seems to have the capability to act as an AP.
* I wonder what facilities the "AP" can provide within a driver? We can
* probably use the BRIDGE code to form an ESS but I don't think
* power saving etc. is easy.
*
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Packet framing/encapsulation/translation
* ========================================
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Currently we support the Webgear encapsulation:
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
* 802.11 header <net/if_ieee80211.h>struct ieee80211_frame
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
* 802.3 header <net/ethernet.h>struct ether_header
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* IP/ARP payload
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* and RFC1042 encapsulation of IP datagrams (translation):
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
* 802.11 header <net/if_ieee80211.h>struct ieee80211_frame
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
* 802.2 LLC header
* 802.2 SNAP header
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* 802.3 Ethertype
* IP/ARP payload
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Framing should be selected via if_media stuff or link types but
* is currently hardcoded to:
* V4 encapsulation
* V5 translation
MFRELENG_3
2000-05-21 21:20:18 +00:00
*
*
* Authentication
* ==============
*
* 802.11 provides two authentication mechanisms. The first is a very
* simple host based mechanism (like xhost) called Open System and the
* second is a more complex challenge/response called Shared Key built
* ontop of WEP.
*
* This driver only supports Open System and does not implement any
* host based control lists. In otherwords authentication is always
* granted to hosts wanting to authenticate with this station. This is
* the only sensible behaviour as the Open System mechanism uses MAC
* addresses to identify hosts. Send me patches if you need it!
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*/
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* ***check all XXX_INFRA code - reassoc not done well at all!
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* ***watchdog to catch screwed up removals?
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* ***error handling of RAY_COM_RUNQ
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* ***error handling of ECF command completions
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* ***can't seem to create a n/w that Win95 wants to see.
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* ***remove panic in ray_com_ecf by re-quing or timeout
* ***use new ioctl stuff - probably need to change RAY_COM_FCHKRUNNING things?
* consider user doing:
* ifconfig ray0 192.168.200.38 -bssid "freed"
* ifconfig ray0 192.168.200.38 -bssid "fred"
* here the second one would be missed in this code
* check that v5 needs timeouts on ecf commands
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
* write up driver structure in comments above
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
* UPDATE_PARAMS seems to return via an interrupt - maybe the timeout
* is needed for wrong values?
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
* proper setting of mib_hop_seq_len with country code for v4 firmware
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* best done with raycontrol?
* countrycode setting is broken I think
* userupdate should trap and do via startjoin etc.
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
* fragmentation when rx level drops?
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* v5 might not need download
* defaults are as documented apart from hop_seq_length
* settings are sane for ad-hoc not infra
*
* driver state
* most state is implied by the sequence of commands in the runq
* but in fact any of the rx and tx path that uses variables
* in the sc_c are potentially going to get screwed?
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
*
* infra mode stuff
* proper handling of the basic rate set - see the manual
* all ray_sj, ray_assoc sequencues need a "nicer" solution as we
* remember association and authentication
* need to consider WEP
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* acting as ap - should be able to get working from the manual
* need to finish RAY_ECMD_REJOIN_DONE
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* finish authenitcation code, it doesn't handle errors/timeouts/
* REJOIN etc.
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* ray_nw_param
Suck out all of the current and desired n/w parameters.
2000-06-11 13:57:59 +00:00
* promisc in here too? - done
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* should be able to update the parameters before we download to the
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* device. This means we must attach a desired struct to the
* runq entry and maybe have another big case statement to
* move these desired into current when not running.
* init must then use the current settings (pre-loaded
* in attach now!) and pass to download. But we can't access
* current nw params outside of the runq - ahhh
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* differeniate between parameters set in attach and init
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* sc_station_addr in here too (for changing mac address)
* move desired into the command structure?
* take downloaded MIB from a complete nw_param?
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* longer term need to attach a desired nw params to the runq entry
*
*
* RAY_COM_RUNQ errors
*
* if sleeping in ccs_alloc with eintr/erestart/enxio/enodev
* erestart try again from the top
* XXX do not malloc more comqs
* XXX ccs allocation hard
* eintr clean up and return
* enxio clean up and return - done in macro
*
* if sleeping in runq_arr itself with eintr/erestart/enxio/enodev
* erestart try again from the top
* XXX do not malloc more comqs
* XXX ccs allocation hard
* XXX reinsert comqs at head of list
* eintr clean up and return
* enxio clean up and return - done in macro
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*/
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
#define XXX 0
MFRELENG_3
2000-05-21 21:20:18 +00:00
#define XXX_ACTING_AP 0
#define XXX_INFRA 0
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
#define RAY_DEBUG ( \
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* RAY_DBG_AUTH | */ \
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* RAY_DBG_SUBR | */ \
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
/* RAY_DBG_BOOTPARAM | */ \
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/* RAY_DBG_STARTJOIN | */ \
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* RAY_DBG_CCS | */ \
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/* RAY_DBG_IOCTL | */ \
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* RAY_DBG_MBUF | */ \
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
/* RAY_DBG_RX | */ \
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* RAY_DBG_CM | */ \
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
/* RAY_DBG_COM | */ \
/* RAY_DBG_STOP | */ \
MFRELENG_3
2000-05-21 21:20:18 +00:00
/* RAY_DBG_CTL | */ \
/* RAY_DBG_MGT | */ \
/* RAY_DBG_TX | */ \
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
/* RAY_DBG_DCOM | */ \
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
0 \
)
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* XXX build options - move to LINT
*/
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
#define RAY_CM_RID 0 /* pccardd abuses windows 0 and 1 */
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
#define RAY_AM_RID 3 /* pccardd abuses windows 0 and 1 */
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
#define RAY_COM_TIMEOUT (hz/2) /* Timeout for CCS commands */
#define RAY_TX_TIMEOUT (hz/2) /* Timeout for rescheduling TX */
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
#define RAY_ECF_SPIN_DELAY 1000 /* Wait 1ms before checking ECF ready */
#define RAY_ECF_SPIN_TRIES 10 /* Wait this many times for ECF ready */
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
/*
* XXX build options - move to LINT
*/
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#ifndef RAY_DEBUG
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
#define RAY_DEBUG 0x0000
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#endif /* RAY_DEBUG */
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#include <sys/param.h>
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
#include <sys/systm.h>
Deprecate machine/limits.h in favor of new sys/limits.h. Change all in-tree consumers to include <sys/limits.h> Discussed on: standards@ Partially submitted by: Craig Rodrigues <rodrigc@attbi.com>
2003-04-29 13:36:06 +00:00
#include <sys/limits.h>
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#include <sys/malloc.h>
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
#include <sys/kernel.h>
Add missing <sys/module.h> includes
2004-05-30 20:08:47 +00:00
#include <sys/module.h>
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/sockio.h>
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
#include <net/bpf.h>
#include <net/ethernet.h>
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
update to new 802.11 support
2003-06-28 17:57:30 +00:00
#include <net80211/ieee80211.h>
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
#include <net/if_llc.h>
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
#include <net/if_types.h>
MFRELENG_3
2000-05-21 21:20:18 +00:00
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
#include <dev/pccard/pccardvar.h>
#include "card_if.h"
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
#include <dev/ray/if_rayreg.h>
#include <dev/ray/if_raymib.h>
#include <dev/ray/if_raydbg.h>
#include <dev/ray/if_rayvar.h>
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
add defs just removed from ieee80211.h Approved by: re (blanket wireless)
2007-09-05 21:25:58 +00:00
typedef uint8_t *ieee80211_mgt_beacon_t;
typedef uint8_t *ieee80211_mgt_auth_t;
#define IEEE80211_AUTH_ALGORITHM(auth) \
((auth)[0] | ((auth)[1] << 8))
#define IEEE80211_AUTH_TRANSACTION(auth) \
((auth)[2] | ((auth)[3] << 8))
#define IEEE80211_AUTH_STATUS(auth) \
((auth)[4] | ((auth)[5] << 8))
Make a bunch of malloc types static. Found by: src/tools/tools/kernxref
2005-02-10 12:02:37 +00:00
static MALLOC_DEFINE(M_RAYCOM, "raycom", "Raylink command queue entry");
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
/*
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
* Prototyping
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static int ray_attach (device_t);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static int ray_ccs_alloc (struct ray_softc *sc, size_t *ccsp, char *wmesg);
static void ray_ccs_fill (struct ray_softc *sc, size_t ccs, u_int cmd);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
static void ray_ccs_free (struct ray_softc *sc, size_t ccs);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static int ray_ccs_tx (struct ray_softc *sc, size_t *ccsp, size_t *bufpp);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_com_ecf (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_com_ecf_done (struct ray_softc *sc);
static void ray_com_ecf_timo (void *xsc);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
static struct ray_comq_entry *
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_init (struct ray_comq_entry *com, ray_comqfn_t function, int flags, char *mesg);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
static struct ray_comq_entry *
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_malloc (ray_comqfn_t function, int flags, char *mesg);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_com_runq (struct ray_softc *sc);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
static int ray_com_runq_add (struct ray_softc *sc, struct ray_comq_entry *com[], int ncom, char *wmesg);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_com_runq_done (struct ray_softc *sc);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static int ray_detach (device_t);
Tidy up if_init routine so that it does not return an error.
2001-01-02 20:29:25 +00:00
static void ray_init (void *xsc);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static int ray_init_user (struct ray_softc *sc);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_init_assoc (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_init_assoc_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
static void ray_init_auth (struct ray_softc *sc, struct ray_comq_entry *com);
static int ray_init_auth_send (struct ray_softc *sc, u_int8_t *dst, int sequence);
static void ray_init_auth_done (struct ray_softc *sc, u_int8_t status);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_init_download (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_init_download_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
static void ray_init_download_v4 (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_init_download_v5 (struct ray_softc *sc, struct ray_comq_entry *com);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
static void ray_init_mcast (struct ray_softc *sc, struct ray_comq_entry *com);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_init_sj (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_init_sj_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static void ray_intr (void *xsc);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
static void ray_intr_ccs (struct ray_softc *sc, u_int8_t cmd, u_int8_t status, size_t ccs);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_intr_rcs (struct ray_softc *sc, u_int8_t cmd, size_t ccs);
static void ray_intr_updt_errcntrs (struct ray_softc *sc);
static int ray_ioctl (struct ifnet *ifp, u_long command, caddr_t data);
static void ray_mcast (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_mcast_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static int ray_mcast_user (struct ray_softc *sc);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static int ray_probe (device_t);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_promisc (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_repparams (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_repparams_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static int ray_repparams_user (struct ray_softc *sc, struct ray_param_req *pr);
static int ray_repstats_user (struct ray_softc *sc, struct ray_stats_req *sr);
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
static int ray_res_alloc_am (struct ray_softc *sc);
static int ray_res_alloc_cm (struct ray_softc *sc);
static int ray_res_alloc_irq (struct ray_softc *sc);
static void ray_res_release (struct ray_softc *sc);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_rx (struct ray_softc *sc, size_t rcs);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static void ray_rx_ctl (struct ray_softc *sc, struct mbuf *m0);
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
static void ray_rx_data (struct ray_softc *sc, struct mbuf *m0, u_int8_t siglev, u_int8_t antenna);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static void ray_rx_mgt (struct ray_softc *sc, struct mbuf *m0);
static void ray_rx_mgt_auth (struct ray_softc *sc, struct mbuf *m0);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_rx_mgt_beacon (struct ray_softc *sc, struct mbuf *m0);
Update for net80211 changes.
2004-12-08 17:36:51 +00:00
static void ray_rx_mgt_info (struct ray_softc *sc, struct mbuf *m0, union ieee80211_information *elements);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_rx_update_cache (struct ray_softc *sc, u_int8_t *src, u_int8_t siglev, u_int8_t antenna);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
static void ray_stop (struct ray_softc *sc, struct ray_comq_entry *com);
static int ray_stop_user (struct ray_softc *sc);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_tx (struct ifnet *ifp);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_tx_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_tx_timo (void *xsc);
Subtle Tx bugs - I wonder why the cast wans't picked up...
2000-06-21 21:37:27 +00:00
static int ray_tx_send (struct ray_softc *sc, size_t ccs, int pktlen, u_int8_t *dst);
MFRELENG_3
2000-05-21 21:20:18 +00:00
static size_t ray_tx_wrhdr (struct ray_softc *sc, size_t bufp, u_int8_t type, u_int8_t fc1, u_int8_t *addr1, u_int8_t *addr2, u_int8_t *addr3);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_upparams (struct ray_softc *sc, struct ray_comq_entry *com);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
static void ray_upparams_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static int ray_upparams_user (struct ray_softc *sc, struct ray_param_req *pr);
static void ray_watchdog (struct ifnet *ifp);
static u_int8_t ray_tx_best_antenna (struct ray_softc *sc, u_int8_t *dst);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
#if RAY_DEBUG & RAY_DBG_COM
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_com_ecf_check (struct ray_softc *sc, size_t ccs, char *mesg);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
#endif /* RAY_DEBUG & RAY_DBG_COM */
#if RAY_DEBUG & RAY_DBG_MBUF
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
static void ray_dump_mbuf (struct ray_softc *sc, struct mbuf *m, char *s);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
#endif /* RAY_DEBUG & RAY_DBG_MBUF */
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
'PC Card' instead of other variants
2005-09-22 06:01:44 +00:00
* PC Card (PCMCIA) driver definition
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static device_method_t ray_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, ray_probe),
DEVMETHOD(device_attach, ray_attach),
DEVMETHOD(device_detach, ray_detach),
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
{ 0, 0 }
};
static driver_t ray_driver = {
"ray",
ray_methods,
sizeof(struct ray_softc)
};
static devclass_t ray_devclass;
DRIVER_MODULE(ray, pccard, ray_driver, ray_devclass, 0, 0);
/*
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
* Probe for the card by checking its startup results.
*
* Fixup any bugs/quirks for different firmware.
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
*/
static int
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
ray_probe(device_t dev)
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
struct ray_softc *sc = device_get_softc(dev);
struct ray_ecf_startup_v5 *ep = &sc->sc_ecf_startup;
int error;
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
sc->dev = dev;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
/*
* Read startup results from the card.
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
error = ray_res_alloc_cm(sc);
if (error)
return (error);
Remove if_ray_oldcard.h because pccard support multiple windows now. Setup attribute memory resource in ray_probe so that it is added to the print out of the resource list on card insertion.
2001-05-07 13:11:08 +00:00
error = ray_res_alloc_am(sc);
if (error) {
ray_res_release(sc);
return (error);
}
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
RAY_MAP_CM(sc);
SRAM_READ_REGION(sc, RAY_ECF_TO_HOST_BASE, ep,
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
sizeof(sc->sc_ecf_startup));
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
ray_res_release(sc);
/*
* Check the card is okay and work out what version we are using.
*/
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
if (ep->e_status != RAY_ECFS_CARD_OK) {
RAY_PRINTF(sc, "card failed self test 0x%b",
ep->e_status, RAY_ECFS_PRINTFB);
return (ENXIO);
}
if (sc->sc_version != RAY_ECFS_BUILD_4 &&
sc->sc_version != RAY_ECFS_BUILD_5) {
RAY_PRINTF(sc, "unsupported firmware version 0x%0x",
ep->e_fw_build_string);
return (ENXIO);
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
RAY_DPRINTF(sc, RAY_DBG_BOOTPARAM, "found a card");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
sc->sc_gone = 0;
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
* Fixup tib size to be correct - on build 4 it is garbage
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4 && sc->sc_tibsize == 0x55)
sc->sc_tibsize = sizeof(struct ray_tx_tib);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
return (0);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
}
/*
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
* Attach the card into the kernel
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
*/
static int
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
ray_attach(device_t dev)
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
struct ray_softc *sc = device_get_softc(dev);
struct ray_ecf_startup_v5 *ep = &sc->sc_ecf_startup;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
size_t ccs;
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
int i, error;
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ifp = if_alloc(IFT_ETHER);
if (ifp == NULL)
return (ENOSPC);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
if ((sc == NULL) || (sc->sc_gone)) {
if_free(ifp);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
return (ENXIO);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
/*
* Grab the resources I need
*/
error = ray_res_alloc_cm(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
if (error) {
if_free(ifp);
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
return (error);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
}
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
error = ray_res_alloc_am(sc);
if (error) {
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
if_free(ifp);
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
ray_res_release(sc);
return (error);
}
error = ray_res_alloc_irq(sc);
if (error) {
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
if_free(ifp);
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
ray_res_release(sc);
return (error);
}
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
* Reset any pending interrupts
*/
RAY_HCS_CLEAR_INTR(sc);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
* Set the parameters that will survive stop/init and
* reset a few things on the card.
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Do not update these in ray_init_download's parameter setup
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
RAY_MAP_CM(sc);
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
bzero(&sc->sc_d, sizeof(struct ray_nw_param));
bzero(&sc->sc_c, sizeof(struct ray_nw_param));
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
/* Clear statistics counters */
sc->sc_rxoverflow = 0;
sc->sc_rxcksum = 0;
sc->sc_rxhcksum = 0;
sc->sc_rxnoise = 0;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
/* Clear signal and antenna cache */
bzero(sc->sc_siglevs, sizeof(sc->sc_siglevs));
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
/* Set all ccs to be free */
bzero(sc->sc_ccsinuse, sizeof(sc->sc_ccsinuse));
ccs = RAY_CCS_ADDRESS(0);
for (i = 0; i < RAY_CCS_LAST; ccs += RAY_CCS_SIZE, i++)
RAY_CCS_FREE(sc, ccs);
/*
* Initialise the network interface structure
*/
Fix probe/attach a little. We don't have to call bpfattach() anymore since ether_ifattach() does it for us. We do need to call ether_ifdetach() instead of if_detach(). And we don't have to check for ifp->if_name already being initialized because it never will be in FreeBSD 4.x and above. Reviewed by: Warner Losh, Duncan Barclay
2000-12-08 21:35:11 +00:00
ifp->if_softc = sc;
Replace the if_name and if_unit members of struct ifnet with new members if_xname, if_dname, and if_dunit. if_xname is the name of the interface and if_dname/unit are the driver name and instance. This change paves the way for interface renaming and enhanced pseudo device creation and configuration symantics. Approved By: re (in principle) Reviewed By: njl, imp Tested On: i386, amd64, sparc64 Obtained From: NetBSD (if_xname)
2003-10-31 18:32:15 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Fix probe/attach a little. We don't have to call bpfattach() anymore since ether_ifattach() does it for us. We do need to call ether_ifdetach() instead of if_detach(). And we don't have to check for ifp->if_name already being initialized because it never will be in FreeBSD 4.x and above. Reviewed by: Warner Losh, Duncan Barclay
2000-12-08 21:35:11 +00:00
ifp->if_timer = 0;
Since if_ray doesn't contain locking or run with INTR_MPSAFE, mark the interface as IFF_NEEDSGIANT so if_start is run holding Giant.
2004-08-13 23:39:07 +00:00
ifp->if_flags = (IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST |
IFF_NEEDSGIANT);
Fix probe/attach a little. We don't have to call bpfattach() anymore since ether_ifattach() does it for us. We do need to call ether_ifdetach() instead of if_detach(). And we don't have to check for ifp->if_name already being initialized because it never will be in FreeBSD 4.x and above. Reviewed by: Warner Losh, Duncan Barclay
2000-12-08 21:35:11 +00:00
ifp->if_hdrlen = sizeof(struct ieee80211_frame) +
sizeof(struct ether_header);
ifp->if_baudrate = 1000000; /* Is this baud or bps ;-) */
ifp->if_start = ray_tx;
ifp->if_ioctl = ray_ioctl;
ifp->if_watchdog = ray_watchdog;
Tidy up if_init routine so that it does not return an error.
2001-01-02 20:29:25 +00:00
ifp->if_init = ray_init;
Fix probe/attach a little. We don't have to call bpfattach() anymore since ether_ifattach() does it for us. We do need to call ether_ifdetach() instead of if_detach(). And we don't have to check for ifp->if_name already being initialized because it never will be in FreeBSD 4.x and above. Reviewed by: Warner Losh, Duncan Barclay
2000-12-08 21:35:11 +00:00
ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
ether_ifattach(ifp, ep->e_station_addr);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Initialise the timers and driver
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
*/
callout_handle_init(&sc->com_timerh);
callout_handle_init(&sc->tx_timerh);
TAILQ_INIT(&sc->sc_comq);
/*
* Print out some useful information
*/
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (bootverbose || (RAY_DEBUG & RAY_DBG_BOOTPARAM)) {
RAY_PRINTF(sc, "start up results");
if (sc->sc_version == RAY_ECFS_BUILD_4)
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Firmware version 4\n");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
else
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Firmware version 5\n");
printf(". Status 0x%b\n", ep->e_status, RAY_ECFS_PRINTFB);
printf(". Ether address %6D\n", ep->e_station_addr, ":");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4) {
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Program checksum %0x\n", ep->e_resv0);
printf(". CIS checksum %0x\n", ep->e_rates[0]);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
} else {
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". (reserved word) %0x\n", ep->e_resv0);
printf(". Supported rates %8D\n", ep->e_rates, ":");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
}
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Japan call sign %12D\n", ep->e_japan_callsign, ":");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_5) {
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Program checksum %0x\n", ep->e_prg_cksum);
printf(". CIS checksum %0x\n", ep->e_cis_cksum);
printf(". Firmware version %0x\n",
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ep->e_fw_build_string);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
printf(". Firmware revision %0x\n", ep->e_fw_build);
printf(". (reserved word) %0x\n", ep->e_fw_resv);
printf(". ASIC version %0x\n", ep->e_asic_version);
printf(". TIB size %0x\n", ep->e_tibsize);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
}
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
return (0);
}
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
/*
* Detach the card
*
* This is usually called when the card is ejected, but
* can be caused by a modunload of a controller driver.
* The idea is to reset the driver's view of the device
* and ensure that any driver entry points such as
* read and write do not hang.
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static int
ray_detach(device_t dev)
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
struct ray_softc *sc = device_get_softc(dev);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
struct ray_comq_entry *com;
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int s;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
s = splimp();
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STOP, "");
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
return (0);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* Mark as not running and detach the interface.
*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* N.B. if_detach can trigger ioctls so we do it first and
* then clean the runq.
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
sc->sc_gone = 1;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sc->sc_c.np_havenet = 0;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
ether_ifdetach(ifp);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* Stop the runq and wake up anyone sleeping for us.
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
untimeout(ray_com_ecf_timo, sc, sc->com_timerh);
untimeout(ray_tx_timo, sc, sc->tx_timerh);
com = TAILQ_FIRST(&sc->sc_comq);
Another round of the <sys/queue.h> FOREACH transmogriffer. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 16:08:18 +00:00
TAILQ_FOREACH(com, &sc->sc_comq, c_chain) {
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
com->c_flags |= RAY_COM_FDETACHED;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
com->c_retval = 0;
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_STOP, "looking at com %p %b",
com, com->c_flags, RAY_COM_FLAGS_PRINTFB);
if (com->c_flags & RAY_COM_FWOK) {
RAY_DPRINTF(sc, RAY_DBG_STOP, "waking com %p", com);
wakeup(com->c_wakeup);
}
}
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
/*
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* Release resources
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
ray_res_release(sc);
Make sure that we call if_free(ifp) after bus_teardown_intr. Since we could get an interrupt after we free the ifp, and the interrupt handler depended on the ifp being still alive, this could, in theory, cause a crash. Eliminate this possibility by moving the if_free to after the bus_teardown_intr() call.
2005-09-19 03:10:21 +00:00
if_free(ifp);
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_STOP, "unloading complete");
splx(s);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
return (0);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Network ioctl request.
*/
static int
ray_ioctl(register struct ifnet *ifp, u_long command, caddr_t data)
{
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
struct ray_softc *sc = ifp->if_softc;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_param_req pr;
struct ray_stats_req sr;
Move _detach to be underneath _attach. Move the code that checks the card into _probe and out of _attach and shuffle a couple of things around in attach. Set ifp in the variable block - catches unused variables.
2000-05-07 16:03:36 +00:00
struct ifreq *ifr = (struct ifreq *)data;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
int s, error, error2;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_IOCTL, "");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
return (ENXIO);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
error = error2 = 0;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
s = splimp();
switch (command) {
case SIOCSIFFLAGS:
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "SIFFLAGS 0x%0x", ifp->if_flags);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
Remove unused ioctl locking
2000-05-28 23:23:24 +00:00
* If the interface is marked up we call ray_init_user.
* This will deal with mcast and promisc flags as well as
* initialising the hardware if it needs it.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
if (ifp->if_flags & IFF_UP)
error = ray_init_user(sc);
else
error = ray_stop_user(sc);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "ADDMULTI/DELMULTI");
error = ray_mcast_user(sc);
break;
case SIOCSRAYPARAM:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "SRAYPARAM");
if ((error = copyin(ifr->ifr_data, &pr, sizeof(pr))))
break;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
error = ray_upparams_user(sc, &pr);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
error2 = copyout(&pr, ifr->ifr_data, sizeof(pr));
error = error2 ? error2 : error;
break;
case SIOCGRAYPARAM:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GRAYPARAM");
if ((error = copyin(ifr->ifr_data, &pr, sizeof(pr))))
break;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
error = ray_repparams_user(sc, &pr);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
error2 = copyout(&pr, ifr->ifr_data, sizeof(pr));
error = error2 ? error2 : error;
break;
case SIOCGRAYSTATS:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GRAYSTATS");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
error = ray_repstats_user(sc, &sr);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
error2 = copyout(&sr, ifr->ifr_data, sizeof(sr));
error = error2 ? error2 : error;
break;
case SIOCGRAYSIGLEV:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GRAYSIGLEV");
error = copyout(sc->sc_siglevs, ifr->ifr_data,
sizeof(sc->sc_siglevs));
break;
case SIOCGIFFLAGS:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GIFFLAGS");
error = EINVAL;
break;
case SIOCGIFMETRIC:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GIFMETRIC");
error = EINVAL;
break;
case SIOCGIFMTU:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GIFMTU");
error = EINVAL;
break;
case SIOCGIFPHYS:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GIFPYHS");
error = EINVAL;
break;
case SIOCSIFMEDIA:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "SIFMEDIA");
error = EINVAL;
break;
case SIOCGIFMEDIA:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "GIFMEDIA");
error = EINVAL;
break;
default:
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "OTHER (pass to ether)");
error = ether_ioctl(ifp, command, data);
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
splx(s);
return (error);
}
Tidy up if_init routine so that it does not return an error.
2001-01-02 20:29:25 +00:00
/*
* Ethernet layer entry to ray_init - discard errors
*/
static void
ray_init(void *xsc)
{
struct ray_softc *sc = (struct ray_softc *)xsc;
ray_init_user(sc);
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
/*
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
* User land entry to network initialisation and changes in interface flags.
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* We do a very little work here, just creating runq entries to
* processes the actions needed to cope with interface flags. We do it
* this way in case there are runq entries outstanding from earlier
* ioctls that modify the interface flags.
MFRELENG_3
2000-05-21 21:20:18 +00:00
*
* Returns values are either 0 for success, a varity of resource allocation
* failures or errors in the command sent to the card.
Remove unused ioctl locking
2000-05-28 23:23:24 +00:00
*
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* Note, IFF_DRV_RUNNING is eventually set by init_sj_done or init_assoc_done
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
static int
ray_init_user(struct ray_softc *sc)
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
{
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
struct ray_comq_entry *com[6];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int error, ncom;
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* Create the following runq entries to bring the card up.
*
* init_download - download the network to the card
* init_mcast - reset multicast list
* init_sj - find or start a BSS
Correct typos, mostly s/ a / an / where appropriate. Some whitespace cleanup, especially in troff files.
2003-01-01 18:49:04 +00:00
* init_auth - authenticate with an ESSID if needed
* init_assoc - associate with an ESSID if needed
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* They are only actually executed if the card is not running.
* We may enter this routine from a simple change of IP
* address and do not need to get the card to do these things.
* However, we cannot perform the check here as there may be
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* commands in the runq that change the IFF_DRV_RUNNING state of
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* the interface.
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
*/
ncom = 0;
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_download, RAY_COM_FCHKRUNNING);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_mcast, RAY_COM_FCHKRUNNING);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_sj, RAY_COM_FCHKRUNNING);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_auth, RAY_COM_FCHKRUNNING);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_assoc, RAY_COM_FCHKRUNNING);
/*
* Create runq entries to process flags
*
* promisc - set/reset PROMISC and ALLMULTI flags
*
* They are only actually executed if the card is running
*/
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_promisc, 0);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_RUNQ(sc, com, ncom, "rayinit", error);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/* XXX no real error processing from anything yet! */
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_FREE(com, ncom);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
return (error);
}
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
* Runq entry for resetting driver and downloading start up structures to card
*/
static void
ray_init_download(struct ray_softc *sc, struct ray_comq_entry *com)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* If the card already running we might not need to download */
RAY_COM_CHKRUNNING(sc, com, ifp);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* Reset instance variables
*
* The first set are network parameters that are read back when
* the card starts or joins the network.
*
* The second set are network parameters that are downloaded to
* the card.
*
* The third set are driver parameters.
*
* All of the variables in these sets can be updated by the
* card or ioctls.
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
sc->sc_d.np_upd_param = 0;
bzero(sc->sc_d.np_bss_id, ETHER_ADDR_LEN);
sc->sc_d.np_inited = 0;
sc->sc_d.np_def_txrate = RAY_MIB_BASIC_RATE_SET_DEFAULT;
sc->sc_d.np_encrypt = 0;
bzero(sc->sc_d.np_ssid, IEEE80211_NWID_LEN);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4) {
sc->sc_d.np_net_type = RAY_MIB_NET_TYPE_V4;
strncpy(sc->sc_d.np_ssid, RAY_MIB_SSID_V4, IEEE80211_NWID_LEN);
sc->sc_d.np_ap_status = RAY_MIB_AP_STATUS_V4;
sc->sc_d.np_framing = RAY_FRAMING_ENCAPSULATION;
} else {
sc->sc_d.np_net_type = RAY_MIB_NET_TYPE_V5;
strncpy(sc->sc_d.np_ssid, RAY_MIB_SSID_V5, IEEE80211_NWID_LEN);
sc->sc_d.np_ap_status = RAY_MIB_AP_STATUS_V5;
sc->sc_d.np_framing = RAY_FRAMING_TRANSLATION;
}
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
sc->sc_d.np_priv_start = RAY_MIB_PRIVACY_MUST_START_DEFAULT;
sc->sc_d.np_priv_join = RAY_MIB_PRIVACY_CAN_JOIN_DEFAULT;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
sc->sc_d.np_promisc = !!(ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI));
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/* XXX this is a hack whilst I transition the code. The instance
* XXX variables above should be set somewhere else. This is needed for
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
* XXX start_join */
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
bcopy(&sc->sc_d, &com->c_desired, sizeof(struct ray_nw_param));
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Download the right firmware defaults
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4)
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
ray_init_download_v4(sc, com);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
else
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
ray_init_download_v5(sc, com);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Kick the card
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_DOWNLOAD_PARAMS);
ray_com_ecf(sc, com);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#define PUT2(p, v) \
do { (p)[0] = ((v >> 8) & 0xff); (p)[1] = (v & 0xff); } while(0)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Firmware version 4 defaults - see if_raymib.h for details
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
static void
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
ray_init_download_v4(struct ray_softc *sc, struct ray_comq_entry *com)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_mib_4 ray_mib_4_default;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Fix warnings. There's no need to use concatination when you have 'something.macro_arg'. Comment out comment following #endif.
2002-06-01 18:50:35 +00:00
#define MIB4(m) ray_mib_4_default.m
Mostly notes and minor style changes whilst I was travelling.
2000-03-03 17:07:42 +00:00
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MIB4(mib_net_type) = com->c_desired.np_net_type;
MIB4(mib_ap_status) = com->c_desired.np_ap_status;
bcopy(com->c_desired.np_ssid, MIB4(mib_ssid), IEEE80211_NWID_LEN);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB4(mib_scan_mode) = RAY_MIB_SCAN_MODE_V4;
MIB4(mib_apm_mode) = RAY_MIB_APM_MODE_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
bcopy(sc->sc_station_addr, MIB4(mib_mac_addr), ETHER_ADDR_LEN);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB4(mib_frag_thresh), RAY_MIB_FRAG_THRESH_V4);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
PUT2(MIB4(mib_dwell_time), RAY_MIB_DWELL_TIME_V4);
PUT2(MIB4(mib_beacon_period), RAY_MIB_BEACON_PERIOD_V4);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB4(mib_dtim_interval) = RAY_MIB_DTIM_INTERVAL_V4;
MIB4(mib_max_retry) = RAY_MIB_MAX_RETRY_V4;
MIB4(mib_ack_timo) = RAY_MIB_ACK_TIMO_V4;
MIB4(mib_sifs) = RAY_MIB_SIFS_V4;
MIB4(mib_difs) = RAY_MIB_DIFS_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_pifs) = RAY_MIB_PIFS_V4;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB4(mib_rts_thresh), RAY_MIB_RTS_THRESH_V4);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
PUT2(MIB4(mib_scan_dwell), RAY_MIB_SCAN_DWELL_V4);
PUT2(MIB4(mib_scan_max_dwell), RAY_MIB_SCAN_MAX_DWELL_V4);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB4(mib_assoc_timo) = RAY_MIB_ASSOC_TIMO_V4;
MIB4(mib_adhoc_scan_cycle) = RAY_MIB_ADHOC_SCAN_CYCLE_V4;
MIB4(mib_infra_scan_cycle) = RAY_MIB_INFRA_SCAN_CYCLE_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_infra_super_scan_cycle)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_INFRA_SUPER_SCAN_CYCLE_V4;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MIB4(mib_promisc) = com->c_desired.np_promisc;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB4(mib_uniq_word), RAY_MIB_UNIQ_WORD_V4);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_slot_time) = RAY_MIB_SLOT_TIME_V4;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB4(mib_roam_low_snr_thresh) = RAY_MIB_ROAM_LOW_SNR_THRESH_V4;
MIB4(mib_low_snr_count) = RAY_MIB_LOW_SNR_COUNT_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_infra_missed_beacon_count)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_INFRA_MISSED_BEACON_COUNT_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_adhoc_missed_beacon_count)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_ADHOC_MISSED_BEACON_COUNT_V4;
MIB4(mib_country_code) = RAY_MIB_COUNTRY_CODE_V4;
MIB4(mib_hop_seq) = RAY_MIB_HOP_SEQ_V4;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB4(mib_hop_seq_len) = RAY_MIB_HOP_SEQ_LEN_V4;
MIB4(mib_cw_max) = RAY_MIB_CW_MAX_V4;
MIB4(mib_cw_min) = RAY_MIB_CW_MIN_V4;
MIB4(mib_noise_filter_gain) = RAY_MIB_NOISE_FILTER_GAIN_DEFAULT;
MIB4(mib_noise_limit_offset) = RAY_MIB_NOISE_LIMIT_OFFSET_DEFAULT;
MIB4(mib_rssi_thresh_offset) = RAY_MIB_RSSI_THRESH_OFFSET_DEFAULT;
MIB4(mib_busy_thresh_offset) = RAY_MIB_BUSY_THRESH_OFFSET_DEFAULT;
MIB4(mib_sync_thresh) = RAY_MIB_SYNC_THRESH_DEFAULT;
MIB4(mib_test_mode) = RAY_MIB_TEST_MODE_DEFAULT;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB4(mib_test_min_chan) = RAY_MIB_TEST_MIN_CHAN_DEFAULT;
MIB4(mib_test_max_chan) = RAY_MIB_TEST_MAX_CHAN_DEFAULT;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#undef MIB4
SRAM_WRITE_REGION(sc, RAY_HOST_TO_ECF_BASE,
&ray_mib_4_default, sizeof(ray_mib_4_default));
}
/*
* Firmware version 5 defaults - see if_raymib.h for details
*/
static void
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
ray_init_download_v5(struct ray_softc *sc, struct ray_comq_entry *com)
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
{
struct ray_mib_5 ray_mib_5_default;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
Fix warnings. There's no need to use concatination when you have 'something.macro_arg'. Comment out comment following #endif.
2002-06-01 18:50:35 +00:00
#define MIB5(m) ray_mib_5_default.m
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MIB5(mib_net_type) = com->c_desired.np_net_type;
MIB5(mib_ap_status) = com->c_desired.np_ap_status;
bcopy(com->c_desired.np_ssid, MIB5(mib_ssid), IEEE80211_NWID_LEN);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB5(mib_scan_mode) = RAY_MIB_SCAN_MODE_V5;
MIB5(mib_apm_mode) = RAY_MIB_APM_MODE_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
bcopy(sc->sc_station_addr, MIB5(mib_mac_addr), ETHER_ADDR_LEN);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB5(mib_frag_thresh), RAY_MIB_FRAG_THRESH_V5);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
PUT2(MIB5(mib_dwell_time), RAY_MIB_DWELL_TIME_V5);
PUT2(MIB5(mib_beacon_period), RAY_MIB_BEACON_PERIOD_V5);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB5(mib_dtim_interval) = RAY_MIB_DTIM_INTERVAL_V5;
MIB5(mib_max_retry) = RAY_MIB_MAX_RETRY_V5;
MIB5(mib_ack_timo) = RAY_MIB_ACK_TIMO_V5;
MIB5(mib_sifs) = RAY_MIB_SIFS_V5;
MIB5(mib_difs) = RAY_MIB_DIFS_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_pifs) = RAY_MIB_PIFS_V5;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB5(mib_rts_thresh), RAY_MIB_RTS_THRESH_V5);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
PUT2(MIB5(mib_scan_dwell), RAY_MIB_SCAN_DWELL_V5);
PUT2(MIB5(mib_scan_max_dwell), RAY_MIB_SCAN_MAX_DWELL_V5);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB5(mib_assoc_timo) = RAY_MIB_ASSOC_TIMO_V5;
MIB5(mib_adhoc_scan_cycle) = RAY_MIB_ADHOC_SCAN_CYCLE_V5;
MIB5(mib_infra_scan_cycle) = RAY_MIB_INFRA_SCAN_CYCLE_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_infra_super_scan_cycle)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_INFRA_SUPER_SCAN_CYCLE_V5;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MIB5(mib_promisc) = com->c_desired.np_promisc;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
PUT2(MIB5(mib_uniq_word), RAY_MIB_UNIQ_WORD_V5);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_slot_time) = RAY_MIB_SLOT_TIME_V5;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB5(mib_roam_low_snr_thresh) = RAY_MIB_ROAM_LOW_SNR_THRESH_V5;
MIB5(mib_low_snr_count) = RAY_MIB_LOW_SNR_COUNT_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_infra_missed_beacon_count)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_INFRA_MISSED_BEACON_COUNT_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_adhoc_missed_beacon_count)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
= RAY_MIB_ADHOC_MISSED_BEACON_COUNT_V5;
MIB5(mib_country_code) = RAY_MIB_COUNTRY_CODE_V5;
MIB5(mib_hop_seq) = RAY_MIB_HOP_SEQ_V5;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_hop_seq_len) = RAY_MIB_HOP_SEQ_LEN_V5;
PUT2(MIB5(mib_cw_max), RAY_MIB_CW_MAX_V5);
PUT2(MIB5(mib_cw_min), RAY_MIB_CW_MIN_V5);
MIB5(mib_noise_filter_gain) = RAY_MIB_NOISE_FILTER_GAIN_DEFAULT;
MIB5(mib_noise_limit_offset) = RAY_MIB_NOISE_LIMIT_OFFSET_DEFAULT;
MIB5(mib_rssi_thresh_offset) = RAY_MIB_RSSI_THRESH_OFFSET_DEFAULT;
MIB5(mib_busy_thresh_offset) = RAY_MIB_BUSY_THRESH_OFFSET_DEFAULT;
MIB5(mib_sync_thresh) = RAY_MIB_SYNC_THRESH_DEFAULT;
MIB5(mib_test_mode) = RAY_MIB_TEST_MODE_DEFAULT;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MIB5(mib_test_min_chan) = RAY_MIB_TEST_MIN_CHAN_DEFAULT;
MIB5(mib_test_max_chan) = RAY_MIB_TEST_MAX_CHAN_DEFAULT;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
MIB5(mib_allow_probe_resp) = RAY_MIB_ALLOW_PROBE_RESP_DEFAULT;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MIB5(mib_privacy_must_start) = com->c_desired.np_priv_start;
MIB5(mib_privacy_can_join) = com->c_desired.np_priv_join;
MIB5(mib_basic_rate_set[0]) = com->c_desired.np_def_txrate;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#undef MIB5
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
SRAM_WRITE_REGION(sc, RAY_HOST_TO_ECF_BASE,
&ray_mib_5_default, sizeof(ray_mib_5_default));
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#undef PUT2
Mostly notes and minor style changes whilst I was travelling.
2000-03-03 17:07:42 +00:00
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Download completion routine
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_download_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_COM_CHECK(sc, ccs);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_ecf_done(sc);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
}
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
/*
* Runq entry to empty the multicast filter list
*/
static void
ray_init_mcast(struct ray_softc *sc, struct ray_comq_entry *com)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* If the card already running we might not need to reset the list */
RAY_COM_CHKRUNNING(sc, com, ifp);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
/*
* Kick the card
*/
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_UPDATE_MCAST);
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_update_mcast, c_nmcast, 0);
ray_com_ecf(sc, com);
}
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Runq entry to starting or joining a network
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
*/
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_init_sj(struct ray_softc *sc, struct ray_comq_entry *com)
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_net_params np;
int update;
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* If the card already running we might not need to start the n/w */
RAY_COM_CHKRUNNING(sc, com, ifp);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Set up the right start or join command and determine
* whether we should tell the card about a change in operating
* parameters.
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sc->sc_c.np_havenet = 0;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (sc->sc_d.np_net_type == RAY_MIB_NET_TYPE_ADHOC)
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_START_NET);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
else
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_JOIN_NET);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
update = 0;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (sc->sc_c.np_net_type != sc->sc_d.np_net_type)
update++;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (bcmp(sc->sc_c.np_ssid, sc->sc_d.np_ssid, IEEE80211_NWID_LEN))
update++;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (sc->sc_c.np_priv_join != sc->sc_d.np_priv_join)
update++;
if (sc->sc_c.np_priv_start != sc->sc_d.np_priv_start)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
update++;
RAY_DPRINTF(sc, RAY_DBG_STARTJOIN,
"%s updating nw params", update?"is":"not");
if (update) {
bzero(&np, sizeof(np));
np.p_net_type = sc->sc_d.np_net_type;
bcopy(sc->sc_d.np_ssid, np.p_ssid, IEEE80211_NWID_LEN);
np.p_privacy_must_start = sc->sc_d.np_priv_start;
np.p_privacy_can_join = sc->sc_d.np_priv_join;
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_WRITE_REGION(sc, RAY_HOST_TO_ECF_BASE, &np, sizeof(np));
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_net, c_upd_param, 1);
} else
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_net, c_upd_param, 0);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
* Kick the card
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_ecf(sc, com);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Complete start command or intermediate step in assoc command
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_sj_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_COM_CHECK(sc, ccs);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Read back network parameters that the ECF sets
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_READ_REGION(sc, ccs, &sc->sc_c.p_1, sizeof(struct ray_cmd_net));
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* Adjust values for buggy firmware */
if (sc->sc_c.np_inited == 0x55)
sc->sc_c.np_inited = 0;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (sc->sc_c.np_def_txrate == 0x55)
sc->sc_c.np_def_txrate = sc->sc_d.np_def_txrate;
if (sc->sc_c.np_encrypt == 0x55)
sc->sc_c.np_encrypt = sc->sc_d.np_encrypt;
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* Update our local state if we updated the network parameters
* when the START_NET or JOIN_NET was issued.
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (sc->sc_c.np_upd_param) {
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_STARTJOIN, "updated parameters");
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_READ_REGION(sc, RAY_HOST_TO_ECF_BASE,
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
&sc->sc_c.p_2, sizeof(struct ray_net_params));
}
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Hurrah! The network is now active.
*
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* Clearing IFF_DRV_OACTIVE will ensure that the system will send us
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* packets. Just before we return from the interrupt context
* we check to see if packets have been queued.
*/
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (SRAM_READ_FIELD_1(sc, ccs, ray_cmd, c_cmd) == RAY_CMD_START_NET) {
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sc->sc_c.np_havenet = 1;
sc->sc_c.np_framing = sc->sc_d.np_framing;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_ecf_done(sc);
Moved a lot of my inline comments to head of code and documented card configuration hassles. Added a TODO list so I don't forget to finish stuff. Tidyed up a lot of XXX. Tidy'd and documented debugging - all DPRINTF have a debug level associated. RAY_DEBUG = 1, will log packet errors. #if protect common memory hacking Don't rely on pccardd for common memory settings. Added a simple transmit mode - reasonable performance but not great. Will do a version of the NetBSD chained buffers soon. Yup we have two way transmisson! Fix a stupid bug in the common memory code - the ioctl call was hidden behind a debug test!
2000-03-05 14:39:23 +00:00
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Runq entry to authenticate with an access point or another station
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_auth(struct ray_softc *sc, struct ray_comq_entry *com)
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN | RAY_DBG_AUTH, "");
/* If card already running we might not need to authenticate */
RAY_COM_CHKRUNNING(sc, com, ifp);
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
/*
Retire some misleading comments and explain why we need to keep a copy of parameters written to the card.
2003-03-05 18:13:19 +00:00
* Don't do anything if we are not in a managed network
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
*
* XXX V4 adhoc does not need this, V5 adhoc unknown
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
*/
if (sc->sc_c.np_net_type != RAY_MIB_NET_TYPE_INFRA) {
ray_com_runq_done(sc);
return;
}
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/*
* XXX_AUTH need to think of run queue when doing auths from request i.e. would
* XXX_AUTH need to have auth at top of runq?
* XXX_AUTH ditto for sending any auth response packets...what about timeouts?
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Kick the card
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/* XXX_AUTH check exit status and retry or fail as we can't associate without this */
ray_init_auth_send(sc, sc->sc_c.np_bss_id, IEEE80211_AUTH_OPEN_REQUEST);
}
/*
* Build and send an authentication packet
*
* If an error occurs, returns 1 else returns 0.
*/
static int
ray_init_auth_send(struct ray_softc *sc, u_int8_t *dst, int sequence)
{
size_t ccs, bufp;
int pktlen = 0;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN | RAY_DBG_AUTH, "");
/* Get a control block */
if (ray_ccs_tx(sc, &ccs, &bufp)) {
RAY_RECERR(sc, "could not obtain a ccs");
return (1);
}
/* Fill the header in */
bufp = ray_tx_wrhdr(sc, bufp,
IEEE80211_FC0_TYPE_MGT | IEEE80211_FC0_SUBTYPE_AUTH,
IEEE80211_FC1_DIR_NODS,
dst,
- Store pointer to the link-level address right in "struct ifnet" rather than in ifindex_table[]; all (except one) accesses are through ifp anyway. IF_LLADDR() works faster, and all (except one) ifaddr_byindex() users were converted to use ifp->if_addr. - Stop storing a (pointer to) Ethernet address in "struct arpcom", and drop the IFP2ENADDR() macro; all users have been converted to use IF_LLADDR() instead.
2005-11-11 16:04:59 +00:00
IF_LLADDR(sc->ifp),
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sc->sc_c.np_bss_id);
/* Add algorithm number */
SRAM_WRITE_1(sc, bufp + pktlen++, IEEE80211_AUTH_ALG_OPEN);
SRAM_WRITE_1(sc, bufp + pktlen++, 0);
/* Add sequence number */
SRAM_WRITE_1(sc, bufp + pktlen++, sequence);
SRAM_WRITE_1(sc, bufp + pktlen++, 0);
/* Add status code */
SRAM_WRITE_1(sc, bufp + pktlen++, 0);
SRAM_WRITE_1(sc, bufp + pktlen++, 0);
pktlen += sizeof(struct ieee80211_frame);
return (ray_tx_send(sc, ccs, pktlen, dst));
}
/*
* Complete authentication runq
*/
static void
ray_init_auth_done(struct ray_softc *sc, u_int8_t status)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN | RAY_DBG_AUTH, "");
if (status != IEEE80211_STATUS_SUCCESS)
RAY_RECERR(sc, "authentication failed with status %d", status);
/*
* XXX_AUTH retry? if not just recall ray_init_auth_send and dont clear runq?
* XXX_AUTH association requires that authenitcation is successful
* XXX_AUTH before we associate, and the runq is the only way to halt the
* XXX_AUTH progress of associate.
* XXX_AUTH In this case I might not need the RAY_AUTH_NEEDED state
*/
ray_com_runq_done(sc);
}
/*
* Runq entry to starting an association with an access point
*/
static void
ray_init_assoc(struct ray_softc *sc, struct ray_comq_entry *com)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
/* If the card already running we might not need to associate */
RAY_COM_CHKRUNNING(sc, com, ifp);
/*
* Don't do anything if we are not in a managed network
*/
if (sc->sc_c.np_net_type != RAY_MIB_NET_TYPE_INFRA) {
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_runq_done(sc);
return;
}
/*
* Kick the card
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_START_ASSOC);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_com_ecf(sc, com);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Complete association
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_assoc_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_COM_CHECK(sc, ccs);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Hurrah! The network is now active.
*
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* Clearing IFF_DRV_OACTIVE will ensure that the system will send us
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* packets. Just before we return from the interrupt context
* we check to see if packets have been queued.
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sc->sc_c.np_havenet = 1;
sc->sc_c.np_framing = sc->sc_d.np_framing;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_ecf_done(sc);
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Network stop.
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Inhibit card - if we can't prevent reception then do not worry;
* stopping a NIC only guarantees no TX.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* The change to the interface flags is done via the runq so that any
* existing commands can execute normally.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
static int
ray_stop_user(struct ray_softc *sc)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
{
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
struct ray_comq_entry *com[1];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int error, ncom;
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STOP, "");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Schedule the real stop routine
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
ncom = 0;
com[ncom++] = RAY_COM_MALLOC(ray_stop, 0);
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_RUNQ(sc, com, ncom, "raystop", error);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/* XXX no real error processing from anything yet! */
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_FREE(com, ncom);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
return (error);
}
/*
* Runq entry for stopping the interface activity
*/
static void
ray_stop(struct ray_softc *sc, struct ray_comq_entry *com)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
struct mbuf *m;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STOP, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* Mark as not running and drain output queue
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ifp->if_timer = 0;
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
for (;;) {
IF_DEQUEUE(&ifp->if_snd, m);
if (m == NULL)
break;
m_freem(m);
}
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_runq_done(sc);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_watchdog(struct ifnet *ifp)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
{
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
struct ray_softc *sc = ifp->if_softc;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
RAY_MAP_CM(sc);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
return;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_PRINTF(sc, "watchdog timeout");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
/*
* Transmit packet handling
*/
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Send a packet.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
* We make two assumptions here:
* 1) That the current priority is set to splimp _before_ this code
* is called *and* is returned to the appropriate priority after
* return
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* 2) That the IFF_DRV_OACTIVE flag is checked before this code is called
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* (i.e. that the output part of the interface is idle)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*
* A simple one packet at a time TX routine is used - we don't bother
* chaining TX buffers. Performance is sufficient to max out the
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* wireless link on a P75.
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*
* AST J30 Windows 95A (100MHz Pentium) to
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* Libretto 50CT FreeBSD-3.1 (75MHz Pentium) 167.37kB/s
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Nonname box FreeBSD-3.4 (233MHz AMD K6) 161.82kB/s
*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* Libretto 50CT FreeBSD-3.1 (75MHz Pentium) to
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* AST J30 Windows 95A (100MHz Pentium) 167.37kB/s
* Nonname box FreeBSD-3.4 (233MHz AMD K6) 161.38kB/s
*
* Given that 160kB/s is saturating the 2Mb/s wireless link we
* are about there.
*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* In short I'm happy that the added complexity of chaining TX
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* packets together isn't worth it for my machines.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_tx(struct ifnet *ifp)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
{
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
struct ray_softc *sc = ifp->if_softc;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct mbuf *m0, *m;
struct ether_header *eh;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
struct llc *llc;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
size_t ccs, bufp;
MFRELENG_3
2000-05-21 21:20:18 +00:00
int pktlen, len;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_MAP_CM(sc);
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Some simple checks first - some are overkill
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
return;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_RECERR(sc, "cannot transmit - not running");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
return;
}
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
if (!sc->sc_c.np_havenet) {
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_RECERR(sc, "cannot transmit - no network");
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
return;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
if (!RAY_ECF_READY(sc)) {
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* Can't assume that the ECF is busy because of this driver */
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
if ((sc->tx_timerh.callout == NULL) ||
(!callout_active(sc->tx_timerh.callout))) {
sc->tx_timerh =
timeout(ray_tx_timo, sc, RAY_TX_TIMEOUT);
return;
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
} else
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
untimeout(ray_tx_timo, sc, sc->tx_timerh);
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
* We find a ccs before we process the mbuf so that we are sure it
* is worthwhile processing the packet. All errors in the mbuf
* processing are either errors in the mbuf or gross configuration
* errors and the packet wouldn't get through anyway.
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
if (ray_ccs_tx(sc, &ccs, &bufp)) {
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
return;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
* Get the mbuf and process it - we have to remember to free the
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* ccs if there are any errors.
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*/
IF_DEQUEUE(&ifp->if_snd, m0);
if (m0 == NULL) {
RAY_CCS_FREE(sc, ccs);
return;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
pktlen = m0->m_pkthdr.len;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
if (pktlen > ETHER_MAX_LEN - ETHER_CRC_LEN) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "mbuf too long %d", pktlen);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
m0 = m_pullup(m0, sizeof(struct ether_header));
if (m0 == NULL) {
RAY_RECERR(sc, "could not pullup ether");
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
return;
}
eh = mtod(m0, struct ether_header *);
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Write the 802.11 header according to network type etc.
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
if (sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_ADHOC)
bufp = ray_tx_wrhdr(sc, bufp,
IEEE80211_FC0_TYPE_DATA,
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
IEEE80211_FC1_DIR_NODS,
MFRELENG_3
2000-05-21 21:20:18 +00:00
eh->ether_dhost,
eh->ether_shost,
sc->sc_c.np_bss_id);
else
if (sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_TERMINAL)
bufp = ray_tx_wrhdr(sc, bufp,
IEEE80211_FC0_TYPE_DATA,
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
IEEE80211_FC1_DIR_TODS,
MFRELENG_3
2000-05-21 21:20:18 +00:00
sc->sc_c.np_bss_id,
eh->ether_shost,
eh->ether_dhost);
else
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
bufp = ray_tx_wrhdr(sc, bufp,
IEEE80211_FC0_TYPE_DATA,
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
IEEE80211_FC1_DIR_FROMDS,
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
eh->ether_dhost,
sc->sc_c.np_bss_id,
eh->ether_shost);
MFRELENG_3
2000-05-21 21:20:18 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Framing
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Add to the mbuf.
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
switch (sc->sc_c.np_framing) {
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
case RAY_FRAMING_ENCAPSULATION:
MFRELENG_3
2000-05-21 21:20:18 +00:00
/* Nice and easy - nothing! (just add an 802.11 header) */
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
case RAY_FRAMING_TRANSLATION:
/*
* Drop the first address in the ethernet header and
* write an LLC and SNAP header over the second.
*/
m_adj(m0, ETHER_ADDR_LEN);
if (m0 == NULL) {
RAY_RECERR(sc, "could not get space for 802.2 header");
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
return;
}
llc = mtod(m0, struct llc *);
llc->llc_dsap = LLC_SNAP_LSAP;
llc->llc_ssap = LLC_SNAP_LSAP;
llc->llc_control = LLC_UI;
llc->llc_un.type_snap.org_code[0] = 0;
llc->llc_un.type_snap.org_code[1] = 0;
llc->llc_un.type_snap.org_code[2] = 0;
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
default:
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_RECERR(sc, "unknown framing type %d", sc->sc_c.np_framing);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
}
if (m0 == NULL) {
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_RECERR(sc, "could not frame packet");
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
return;
}
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_MBUF_DUMP(sc, RAY_DBG_TX, m0, "framed packet");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Copy the mbuf to the buffer in common memory
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
* We drop and don't bother wrapping as Ethernet packets are 1518
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* bytes, we checked the mbuf earlier, and our TX buffers are 2048
* bytes. We don't have 530 bytes of headers etc. so something
* must be fubar.
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
pktlen = sizeof(struct ieee80211_frame);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
for (m = m0; m != NULL; m = m->m_next) {
pktlen += m->m_len;
if ((len = m->m_len) == 0)
continue;
if ((bufp + len) < RAY_TX_END)
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_WRITE_REGION(sc, bufp, mtod(m, u_int8_t *), len);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
else {
RAY_RECERR(sc, "tx buffer overflow");
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
bufp += len;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Send it off
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
if (ray_tx_send(sc, ccs, pktlen, eh->ether_dhost))
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
ifp->if_oerrors++;
MFRELENG_3
2000-05-21 21:20:18 +00:00
else
ifp->if_opackets++;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
m_freem(m0);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
/*
* Start timeout routine.
*
* Used when card was busy but we needed to send a packet.
*/
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_tx_timo(void *xsc)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
struct ray_softc *sc = (struct ray_softc *)xsc;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
int s;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_OACTIVE) && (ifp->if_snd.ifq_head != NULL)) {
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
s = splimp();
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_tx(ifp);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
splx(s);
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Write an 802.11 header into the Tx buffer space and return the
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
* adjusted buffer pointer.
*/
static size_t
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_tx_wrhdr(struct ray_softc *sc, size_t bufp, u_int8_t type, u_int8_t fc1, u_int8_t *addr1, u_int8_t *addr2, u_int8_t *addr3)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame header;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
RAY_MAP_CM(sc);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
bzero(&header, sizeof(struct ieee80211_frame));
MFRELENG_3
2000-05-21 21:20:18 +00:00
header.i_fc[0] = (IEEE80211_FC0_VERSION_0 | type);
header.i_fc[1] = fc1;
bcopy(addr1, header.i_addr1, ETHER_ADDR_LEN);
bcopy(addr2, header.i_addr2, ETHER_ADDR_LEN);
bcopy(addr3, header.i_addr3, ETHER_ADDR_LEN);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
SRAM_WRITE_REGION(sc, bufp, (u_int8_t *)&header,
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
sizeof(struct ieee80211_frame));
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
return (bufp + sizeof(struct ieee80211_frame));
MFRELENG_3
2000-05-21 21:20:18 +00:00
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
* Fill in a few loose ends and kick the card to send the packet
*
* Returns 0 on success, 1 on failure
*/
static int
Subtle Tx bugs - I wonder why the cast wans't picked up...
2000-06-21 21:37:27 +00:00
ray_tx_send(struct ray_softc *sc, size_t ccs, int pktlen, u_int8_t *dst)
MFRELENG_3
2000-05-21 21:20:18 +00:00
{
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
int i = 0;
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
RAY_MAP_CM(sc);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
while (!RAY_ECF_READY(sc)) {
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
DELAY(RAY_ECF_SPIN_DELAY);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if (++i > RAY_ECF_SPIN_TRIES) {
RAY_RECERR(sc, "ECF busy, dropping packet");
RAY_CCS_FREE(sc, ccs);
return (1);
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
}
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if (i != 0)
RAY_RECERR(sc, "spun %d times", i);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
SRAM_WRITE_FIELD_2(sc, ccs, ray_cmd_tx, c_len, pktlen);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_antenna,
ray_tx_best_antenna(sc, dst));
SRAM_WRITE_1(sc, RAY_SCB_CCSI, RAY_CCS_INDEX(ccs));
RAY_ECF_START_CMD(sc);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
return (0);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
}
/*
* Determine best antenna to use from rx level and antenna cache
*/
static u_int8_t
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_tx_best_antenna(struct ray_softc *sc, u_int8_t *dst)
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
{
struct ray_siglev *sl;
int i;
u_int8_t antenna;
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4)
return (0);
/* try to find host */
for (i = 0; i < RAY_NSIGLEVRECS; i++) {
sl = &sc->sc_siglevs[i];
if (bcmp(sl->rsl_host, dst, ETHER_ADDR_LEN) == 0)
goto found;
}
/* not found, return default setting */
return (0);
found:
Removed a few RAY_MAP_CMs that were not needed. Made checking sc->gone a lot safer by checking to see if sc is NULL first. Made return from tsleep in the ccs allocator detach safe.
2000-06-10 21:24:36 +00:00
/* This is a simple thresholding scheme that takes the mean
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
* of the best antenna history. This is okay but as it is a
* filter, it adds a bit of lag in situations where the
* best antenna swaps from one side to the other slowly. Don't know
* how likely this is given the horrible fading though.
*/
antenna = 0;
for (i = 0; i < RAY_NANTENNA; i++) {
antenna += sl->rsl_antennas[i];
}
return (antenna > (RAY_NANTENNA >> 1));
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
/*
* Transmit now complete so clear ccs and network flags.
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_tx_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors);
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
RAY_CCS_FREE(sc, ccs);
ifp->if_timer = 0;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_OACTIVE)
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Receiver packet handling
*/
/*
* Receive a packet from the card
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
static void
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ray_rx(struct ray_softc *sc, size_t rcs)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame *header;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
struct mbuf *m0;
size_t pktlen, fraglen, readlen, tmplen;
size_t bufp, ebufp;
u_int8_t siglev, antenna;
u_int first, ni, i;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
u_int8_t *mp;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
RAY_MAP_CM(sc);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "using rcs 0x%x", rcs);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
m0 = NULL;
readlen = 0;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
* Get first part of packet and the length. Do some sanity checks
* and get a mbuf.
*/
first = RAY_CCS_INDEX(rcs);
pktlen = SRAM_READ_FIELD_2(sc, rcs, ray_cmd_rx, c_pktlen);
siglev = SRAM_READ_FIELD_1(sc, rcs, ray_cmd_rx, c_siglev);
antenna = SRAM_READ_FIELD_1(sc, rcs, ray_cmd_rx, c_antenna);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
if ((pktlen > MCLBYTES) || (pktlen < sizeof(struct ieee80211_frame))) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "packet too big or too small");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
goto skip_read;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
MGETHDR(m0, M_DONTWAIT, MT_DATA);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
if (m0 == NULL) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "MGETHDR failed");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
goto skip_read;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
if (pktlen > MHLEN) {
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
MCLGET(m0, M_DONTWAIT);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (!(m0->m_flags & M_EXT)) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "MCLGET failed");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
m0->m_pkthdr.rcvif = ifp;
m0->m_pkthdr.len = pktlen;
m0->m_len = pktlen;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
mp = mtod(m0, u_int8_t *);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
* Walk the fragment chain to build the complete packet.
*
* The use of two index variables removes a race with the
* hardware. If one index were used the clearing of the CCS would
* happen before reading the next pointer and the hardware can get in.
* Not my idea but verbatim from the NetBSD driver.
*/
i = ni = first;
while ((i = ni) && (i != RAY_CCS_LINK_NULL)) {
rcs = RAY_CCS_ADDRESS(i);
ni = SRAM_READ_FIELD_1(sc, rcs, ray_cmd_rx, c_nextfrag);
bufp = SRAM_READ_FIELD_2(sc, rcs, ray_cmd_rx, c_bufp);
fraglen = SRAM_READ_FIELD_2(sc, rcs, ray_cmd_rx, c_len);
if (fraglen + readlen > pktlen) {
64-bit clean fixes: Use %zx in preference to %x to print size_t items. Cast a byte to uint8_t before printing.
2005-01-27 01:49:23 +00:00
RAY_RECERR(sc, "bad length current 0x%zx pktlen 0x%zx",
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
fraglen + readlen, pktlen);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
if ((i < RAY_RCS_FIRST) || (i > RAY_RCS_LAST)) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "bad rcs index 0x%x", i);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
ebufp = bufp + fraglen;
if (ebufp <= RAY_RX_END)
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
SRAM_READ_REGION(sc, bufp, mp, fraglen);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
else {
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
SRAM_READ_REGION(sc, bufp, mp,
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
(tmplen = RAY_RX_END - bufp));
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
SRAM_READ_REGION(sc, RAY_RX_BASE, mp + tmplen,
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ebufp - RAY_RX_END);
}
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
mp += fraglen;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
readlen += fraglen;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
skip_read:
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
* Walk the chain again to free the rcss.
*/
i = ni = first;
while ((i = ni) && (i != RAY_CCS_LINK_NULL)) {
rcs = RAY_CCS_ADDRESS(i);
ni = SRAM_READ_FIELD_1(sc, rcs, ray_cmd_rx, c_nextfrag);
RAY_CCS_FREE(sc, rcs);
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
if (m0 == NULL)
return;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* Check the 802.11 packet type and hand off to
* appropriate functions.
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
*/
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
header = mtod(m0, struct ieee80211_frame *);
MFRELENG_3
2000-05-21 21:20:18 +00:00
if ((header->i_fc[0] & IEEE80211_FC0_VERSION_MASK)
!= IEEE80211_FC0_VERSION_0) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "header not version 0 fc0 0x%x",
header->i_fc[0]);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
m_freem(m0);
return;
}
MFRELENG_3
2000-05-21 21:20:18 +00:00
switch (header->i_fc[0] & IEEE80211_FC0_TYPE_MASK) {
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
case IEEE80211_FC0_TYPE_DATA:
ray_rx_data(sc, m0, siglev, antenna);
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case IEEE80211_FC0_TYPE_MGT:
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_rx_mgt(sc, m0);
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case IEEE80211_FC0_TYPE_CTL:
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_rx_ctl(sc, m0);
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
break;
default:
RAY_RECERR(sc, "unknown packet fc0 0x%x", header->i_fc[0]);
ifp->if_ierrors++;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
m_freem(m0);
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
}
}
/*
* Deal with DATA packet types
*/
static void
ray_rx_data(struct ray_softc *sc, struct mbuf *m0, u_int8_t siglev, u_int8_t antenna)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
struct llc *llc;
u_int8_t *sa = NULL, *da = NULL, *ra = NULL, *ta = NULL;
int trim = 0;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_RX, "");
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
* Check the the data packet subtype, some packets have
* nothing in them so we will drop them here.
*/
switch (header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK) {
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_DATA:
case IEEE80211_FC0_SUBTYPE_CF_ACK:
case IEEE80211_FC0_SUBTYPE_CF_POLL:
case IEEE80211_FC0_SUBTYPE_CF_ACPL:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "DATA packet");
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_NODATA:
case IEEE80211_FC0_SUBTYPE_CFACK:
case IEEE80211_FC0_SUBTYPE_CFPOLL:
case IEEE80211_FC0_SUBTYPE_CF_ACK_CF_ACK:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "NULL packet");
m_freem(m0);
return;
break;
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "reserved DATA packet subtype 0x%x",
MFRELENG_3
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
m_freem(m0);
return;
}
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Parse the To DS and From DS fields to determine the length
* of the 802.11 header for use later on.
*
* Additionally, furtle out the right destination and
* source MAC addresses for the packet. Packets may come via
* APs so the MAC addresses of the immediate node may be
* different from the node that actually sent us the packet.
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* da destination address of final recipient
* sa source address of orginator
* ra receiver address of immediate recipient
* ta transmitter address of immediate orginator
*
* Address matching is performed on da or sa with the AP or
* BSSID in ra and ta.
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(1) packet before framing");
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
switch (header->i_fc[1] & IEEE80211_FC1_DIR_MASK) {
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC1_DIR_NODS:
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
da = ra = header->i_addr1;
sa = ta = header->i_addr2;
trim = sizeof(struct ieee80211_frame);
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D",
sa, ":", da, ":");
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC1_DIR_FROMDS:
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
da = ra = header->i_addr1;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
ta = header->i_addr2;
sa = header->i_addr3;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
trim = sizeof(struct ieee80211_frame);
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "ap %6D from %6D to %6D",
ta, ":", sa, ":", da, ":");
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC1_DIR_TODS:
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
ra = header->i_addr1;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
sa = ta = header->i_addr2;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
da = header->i_addr3;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
trim = sizeof(struct ieee80211_frame);
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D ap %6D",
sa, ":", da, ":", ra, ":");
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC1_DIR_DSTODS:
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
ra = header->i_addr1;
ta = header->i_addr2;
da = header->i_addr3;
sa = (u_int8_t *)header+1;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
trim = sizeof(struct ieee80211_frame) + ETHER_ADDR_LEN;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D ap %6D to %6D",
sa, ":", da, ":", ta, ":", ra, ":");
break;
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Framing
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Each case must leave an Ethernet header and adjust trim.
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
switch (sc->sc_c.np_framing) {
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
case RAY_FRAMING_ENCAPSULATION:
/* A NOP as the Ethernet header is in the packet */
break;
case RAY_FRAMING_TRANSLATION:
/* Check that we have an LLC and SNAP sequence */
llc = (struct llc *)((u_int8_t *)header + trim);
if (llc->llc_dsap == LLC_SNAP_LSAP &&
llc->llc_ssap == LLC_SNAP_LSAP &&
llc->llc_control == LLC_UI &&
llc->llc_un.type_snap.org_code[0] == 0 &&
llc->llc_un.type_snap.org_code[1] == 0 &&
llc->llc_un.type_snap.org_code[2] == 0) {
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
struct ether_header *eh;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
/*
* This is not magic. RFC1042 header is 8
* bytes, with the last two bytes being the
* ether type. So all we need is another
* ETHER_ADDR_LEN bytes to write the
* destination into.
*/
trim -= ETHER_ADDR_LEN;
eh = (struct ether_header *)((u_int8_t *)header + trim);
/*
* Copy carefully to avoid mashing the MAC
* addresses. The address layout in the .11 header
* does make sense, honest, but it is a pain.
*
* NODS da sa no risk
* FROMDS da ta sa sa then da
* DSTODS ra ta da sa sa then da
* TODS ra sa da da then sa
*/
if (sa > da) {
/* Copy sa first */
bcopy(sa, eh->ether_shost, ETHER_ADDR_LEN);
bcopy(da, eh->ether_dhost, ETHER_ADDR_LEN);
} else {
/* Copy da first */
bcopy(da, eh->ether_dhost, ETHER_ADDR_LEN);
bcopy(sa, eh->ether_shost, ETHER_ADDR_LEN);
}
} else {
/* Assume RAY_FRAMING_ENCAPSULATION */
RAY_RECERR(sc,
"got encapsulated packet but in translation mode");
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
default:
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_RECERR(sc, "unknown framing type %d", sc->sc_c.np_framing);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ierrors++;
m_freem(m0);
return;
}
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(2) packet after framing");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
* Finally, do a bit of house keeping before sending the packet
* up the stack.
*/
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
m_adj(m0, trim);
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(3) packet after trimming");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
ifp->if_ipackets++;
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
ray_rx_update_cache(sc, header->i_addr2, siglev, antenna);
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
(*ifp->if_input)(ifp, m0);
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
}
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
* Deal with MGT packet types
*/
static void
ray_rx_mgt(struct ray_softc *sc, struct mbuf *m0)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
if ((header->i_fc[1] & IEEE80211_FC1_DIR_MASK) !=
IEEE80211_FC1_DIR_NODS) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "MGT TODS/FROMDS wrong fc1 0x%x",
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
header->i_fc[1] & IEEE80211_FC1_DIR_MASK);
MFRELENG_3
2000-05-21 21:20:18 +00:00
ifp->if_ierrors++;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
m_freem(m0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
return;
}
/*
* Check the the mgt packet subtype, some packets should be
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* dropped depending on the mode the station is in. See pg
* 52(60) of docs
MFRELENG_3
2000-05-21 21:20:18 +00:00
*
* P - proccess, J - Junk, E - ECF deals with, I - Illegal
* ECF Proccesses
* AHDOC procces or junk
* INFRA STA process or junk
* INFRA AP process or jumk
*
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
* +PPP IEEE80211_FC0_SUBTYPE_BEACON
* +EEE IEEE80211_FC0_SUBTYPE_PROBE_REQ
* +EEE IEEE80211_FC0_SUBTYPE_PROBE_RESP
* PPP IEEE80211_FC0_SUBTYPE_AUTH
* PPP IEEE80211_FC0_SUBTYPE_DEAUTH
* JJP IEEE80211_FC0_SUBTYPE_ASSOC_REQ
* JPJ IEEE80211_FC0_SUBTYPE_ASSOC_RESP
* JPP IEEE80211_FC0_SUBTYPE_DISASSOC
* JJP IEEE80211_FC0_SUBTYPE_REASSOC_REQ
* JPJ IEEE80211_FC0_SUBTYPE_REASSOC_RESP
* +EEE IEEE80211_FC0_SUBTYPE_ATIM
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_MBUF_DUMP(sc, RAY_DBG_MGT, m0, "MGT packet");
MFRELENG_3
2000-05-21 21:20:18 +00:00
switch (header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK) {
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_BEACON:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "BEACON MGT packet");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_rx_mgt_beacon(sc, m0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_AUTH:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "AUTH MGT packet");
ray_rx_mgt_auth(sc, m0);
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_DEAUTH:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "DEAUTH MGT packet");
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/* XXX ray_rx_mgt_deauth(sc, m0); */
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_ASSOC_REQ:
case IEEE80211_FC0_SUBTYPE_REASSOC_REQ:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "(RE)ASSOC_REQ MGT packet");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_AP))
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_RECERR(sc, "can't be an AP yet"); /* XXX_ACTING_AP */
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_ASSOC_RESP:
case IEEE80211_FC0_SUBTYPE_REASSOC_RESP:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "(RE)ASSOC_RESP MGT packet");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_TERMINAL))
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_RECERR(sc, "can't be in INFRA yet"); /* XXX_INFRA */
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_DISASSOC:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "DISASSOC MGT packet");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if (sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA)
RAY_RECERR(sc, "can't be in INFRA yet"); /* XXX_INFRA */
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_PROBE_REQ:
case IEEE80211_FC0_SUBTYPE_PROBE_RESP:
case IEEE80211_FC0_SUBTYPE_ATIM:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected MGT packet subtype 0x%0x",
MFRELENG_3
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
break;
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "reserved MGT packet subtype 0x%x",
MFRELENG_3
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
}
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
m_freem(m0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
}
/*
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
* Deal with BEACON management packet types
* XXX furtle anything interesting out
* XXX Note that there are rules governing what beacons to read
* XXX see 8802 S7.2.3, S11.1.2.3
* XXX is this actually useful?
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_rx_mgt_beacon(struct ray_softc *sc, struct mbuf *m0)
MFRELENG_3
2000-05-21 21:20:18 +00:00
{
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ieee80211_mgt_beacon_t beacon = (u_int8_t *)(header+1);
Update for net80211 changes.
2004-12-08 17:36:51 +00:00
union ieee80211_information elements;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
u_int64_t *timestamp;
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
timestamp = (u_int64_t *)beacon;
Removed a few RAY_MAP_CMs that were not needed. Made checking sc->gone a lot safer by checking to see if sc is NULL first. Made return from tsleep in the ccs allocator detach safe.
2000-06-10 21:24:36 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "timestamp\t0x%x", *timestamp);
RAY_DPRINTF(sc, RAY_DBG_MGT, "interval\t\t0x%x", IEEE80211_BEACON_INTERVAL(beacon));
RAY_DPRINTF(sc, RAY_DBG_MGT, "capability\t0x%x", IEEE80211_BEACON_CAPABILITY(beacon));
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_rx_mgt_info(sc, m0, &elements);
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
}
static void
Update for net80211 changes.
2004-12-08 17:36:51 +00:00
ray_rx_mgt_info(struct ray_softc *sc, struct mbuf *m0, union ieee80211_information *elements)
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
ieee80211_mgt_beacon_t beacon = (u_int8_t *)(header+1);
ieee80211_mgt_beacon_t bp, be;
int len;
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
bp = beacon + 12;
be = mtod(m0, u_int8_t *) + m0->m_len;
while (bp < be) {
len = *(bp + 1);
RAY_DPRINTF(sc, RAY_DBG_MGT, "id 0x%02x length %d", *bp, len);
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
switch (*bp) {
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
case IEEE80211_ELEMID_SSID:
if (len > IEEE80211_NWID_LEN) {
RAY_RECERR(sc, "bad SSD length: %d from %6D",
len, header->i_addr2, ":");
}
strncpy(elements->ssid, bp + 2, len);
elements->ssid[len] = 0;
RAY_DPRINTF(sc, RAY_DBG_MGT,
"beacon ssid %s", elements->ssid);
break;
case IEEE80211_ELEMID_RATES:
RAY_DPRINTF(sc, RAY_DBG_MGT, "rates");
break;
case IEEE80211_ELEMID_FHPARMS:
elements->fh.dwell = bp[2] + (bp[3] << 8);
elements->fh.set = bp[4];
elements->fh.pattern = bp[5];
elements->fh.index = bp[6];
RAY_DPRINTF(sc, RAY_DBG_MGT,
"fhparams dwell\t0x%04x", elements->fh.dwell);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"fhparams set\t0x%02x", elements->fh.set);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"fhparams pattern\t0x%02x", elements->fh.pattern);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"fhparams index\t0x%02x", elements->fh.index);
break;
case IEEE80211_ELEMID_DSPARMS:
RAY_RECERR(sc, "got direct sequence params!");
break;
case IEEE80211_ELEMID_CFPARMS:
RAY_DPRINTF(sc, RAY_DBG_MGT, "cfparams");
break;
case IEEE80211_ELEMID_TIM:
elements->tim.count = bp[2];
elements->tim.period = bp[3];
elements->tim.bitctl = bp[4];
RAY_DPRINTF(sc, RAY_DBG_MGT,
"tim count\t0x%02x", elements->tim.count);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"tim period\t0x%02x", elements->tim.period);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"tim bitctl\t0x%02x", elements->tim.bitctl);
#if RAY_DEBUG & RAY_DBG_MGT
{
int i;
for (i = 5; i < len + 1; i++)
RAY_DPRINTF(sc, RAY_DBG_MGT,
"tim pvt[%03d]\t0x%02x", i-5, bp[i]);
}
Fix warnings. There's no need to use concatination when you have 'something.macro_arg'. Comment out comment following #endif.
2002-06-01 18:50:35 +00:00
#endif /* (RAY_DEBUG & RAY_DBG_MGT) */
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
break;
case IEEE80211_ELEMID_IBSSPARMS:
elements->ibss.atim = bp[2] + (bp[3] << 8);
RAY_DPRINTF(sc, RAY_DBG_MGT,
"ibssparams atim\t0x%02x", elements->ibss.atim);
break;
case IEEE80211_ELEMID_CHALLENGE:
RAY_DPRINTF(sc, RAY_DBG_MGT, "challenge");
break;
MFRELENG_3
2000-05-21 21:20:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
default:
RAY_RECERR(sc, "reserved MGT element id 0x%x", *bp);
ifp->if_ierrors++;break;
MFRELENG_3
2000-05-21 21:20:18 +00:00
}
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
bp += bp[1] + 2;
}
}
/*
* Deal with AUTH management packet types
*/
static void
ray_rx_mgt_auth(struct ray_softc *sc, struct mbuf *m0)
{
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
ieee80211_mgt_auth_t auth = (u_int8_t *)(header+1);
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_AUTH, "");
switch (IEEE80211_AUTH_ALGORITHM(auth)) {
case IEEE80211_AUTH_ALG_OPEN:
RAY_DPRINTF(sc, RAY_DBG_AUTH,
"open system authentication sequence number %d",
IEEE80211_AUTH_TRANSACTION(auth));
if (IEEE80211_AUTH_TRANSACTION(auth) ==
IEEE80211_AUTH_OPEN_REQUEST) {
/* XXX_AUTH use ray_init_auth_send */
} else if (IEEE80211_AUTH_TRANSACTION(auth) ==
IEEE80211_AUTH_OPEN_RESPONSE)
ray_init_auth_done(sc, IEEE80211_AUTH_STATUS(auth));
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_AUTH_ALG_SHARED:
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_RECERR(sc,
"shared key authentication sequence number %d",
IEEE80211_AUTH_TRANSACTION(auth));
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc,
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
"reserved authentication subtype 0x%04hx",
MFRELENG_3
2000-05-21 21:20:18 +00:00
IEEE80211_AUTH_ALGORITHM(auth));
break;
}
}
/*
* Deal with CTL packet types
*/
static void
ray_rx_ctl(struct ray_softc *sc, struct mbuf *m0)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CTL, "");
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
if ((header->i_fc[1] & IEEE80211_FC1_DIR_MASK) !=
IEEE80211_FC1_DIR_NODS) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "CTL TODS/FROMDS wrong fc1 0x%x",
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
header->i_fc[1] & IEEE80211_FC1_DIR_MASK);
MFRELENG_3
2000-05-21 21:20:18 +00:00
ifp->if_ierrors++;
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
m_freem(m0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
return;
}
/*
* Check the the ctl packet subtype, some packets should be
* dropped depending on the mode the station is in. The ECF
* should deal with everything but the power save poll to an
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* AP. See pg 52(60) of docs.
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
RAY_MBUF_DUMP(sc, RAY_DBG_CTL, m0, "CTL packet");
switch (header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK) {
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_PS_POLL:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CTL, "PS_POLL CTL packet");
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
if ((sc->sc_d.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_AP))
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_RECERR(sc, "can't be an AP yet"); /* XXX_ACTING_AP */
MFRELENG_3
2000-05-21 21:20:18 +00:00
break;
Changes necessary to make this work. The prior version in the tree was repo-copied from Duncan Barclay's cvs tree. Also add $FreeBSD$ Submitted by: Duncan Barclay Committed-via: raylan link with two webgear cards.
2000-11-12 21:43:52 +00:00
case IEEE80211_FC0_SUBTYPE_RTS:
case IEEE80211_FC0_SUBTYPE_CTS:
case IEEE80211_FC0_SUBTYPE_ACK:
case IEEE80211_FC0_SUBTYPE_CF_END:
case IEEE80211_FC0_SUBTYPE_CF_END_ACK:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected CTL packet subtype 0x%0x",
MFRELENG_3
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
break;
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "reserved CTL packet subtype 0x%x",
MFRELENG_3
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
}
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
m_freem(m0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
}
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
/*
* Update rx level and antenna cache
*/
static void
ray_rx_update_cache(struct ray_softc *sc, u_int8_t *src, u_int8_t siglev, u_int8_t antenna)
{
struct timeval mint;
struct ray_siglev *sl;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
int i, mini;
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* Try to find host */
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
for (i = 0; i < RAY_NSIGLEVRECS; i++) {
sl = &sc->sc_siglevs[i];
if (bcmp(sl->rsl_host, src, ETHER_ADDR_LEN) == 0)
goto found;
}
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* Not found, find oldest slot */
Use the right value for IFQ_MAXLEN and fix brokeness with gatewaying http from windows. Add signal strength cache and antenna selection (untested as firmware v4 doesn't return the antenna). KNF
2000-03-31 20:12:30 +00:00
mini = 0;
mint.tv_sec = LONG_MAX;
mint.tv_usec = 0;
for (i = 0; i < RAY_NSIGLEVRECS; i++) {
sl = &sc->sc_siglevs[i];
if (timevalcmp(&sl->rsl_time, &mint, <)) {
mini = i;
mint = sl->rsl_time;
}
}
sl = &sc->sc_siglevs[mini];
bzero(sl->rsl_siglevs, RAY_NSIGLEV);
bzero(sl->rsl_antennas, RAY_NANTENNA);
bcopy(src, sl->rsl_host, ETHER_ADDR_LEN);
found:
microtime(&sl->rsl_time);
bcopy(sl->rsl_siglevs, &sl->rsl_siglevs[1], RAY_NSIGLEV-1);
sl->rsl_siglevs[0] = siglev;
if (sc->sc_version != RAY_ECFS_BUILD_4) {
bcopy(sl->rsl_antennas, &sl->rsl_antennas[1], RAY_NANTENNA-1);
sl->rsl_antennas[0] = antenna;
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
* Interrupt handling
*/
Added reset code and converted many panics into call to reset. Finished ray_stop and ray_unload. A little more macro tidying.
2000-03-05 22:24:30 +00:00
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Process an interrupt
Added reset code and converted many panics into call to reset. Finished ray_stop and ray_unload. A little more macro tidying.
2000-03-05 22:24:30 +00:00
*/
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
static void
ray_intr(void *xsc)
Added reset code and converted many panics into call to reset. Finished ray_stop and ray_unload. A little more macro tidying.
2000-03-05 22:24:30 +00:00
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
struct ray_softc *sc = (struct ray_softc *)xsc;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
size_t ccs;
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
u_int8_t cmd, status;
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
int ccsi;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
RAY_MAP_CM(sc);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
return;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
* Check that the interrupt was for us, if so get the rcs/ccs
* and vector on the command contained within it.
*/
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
if (RAY_HCS_INTR(sc)) {
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ccsi = SRAM_READ_1(sc, RAY_SCB_RCSI);
ccs = RAY_CCS_ADDRESS(ccsi);
cmd = SRAM_READ_FIELD_1(sc, ccs, ray_cmd, c_cmd);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
status = SRAM_READ_FIELD_1(sc, ccs, ray_cmd, c_status);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (ccsi <= RAY_CCS_LAST)
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
ray_intr_ccs(sc, cmd, status, ccs);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
else if (ccsi <= RAY_RCS_LAST)
ray_intr_rcs(sc, cmd, ccs);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
else
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "bad ccs index 0x%x", ccsi);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_HCS_CLEAR_INTR(sc);
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/* Send any packets lying around and update error counters */
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_OACTIVE) && (ifp->if_snd.ifq_head != NULL))
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_tx(ifp);
if ((++sc->sc_checkcounters % 32) == 0)
ray_intr_updt_errcntrs(sc);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Added reset code and converted many panics into call to reset. Finished ray_stop and ray_unload. A little more macro tidying.
2000-03-05 22:24:30 +00:00
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Read the error counters.
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
static void
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
ray_intr_updt_errcntrs(struct ray_softc *sc)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
size_t csc;
Added reset code and converted many panics into call to reset. Finished ray_stop and ray_unload. A little more macro tidying.
2000-03-05 22:24:30 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
RAY_MAP_CM(sc);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* The card implements the following protocol to keep the
* values from being changed while read: It checks the `own'
* bit and if zero writes the current internal counter value,
* it then sets the `own' bit to 1. If the `own' bit was 1 it
* incremenets its internal counter. The user thus reads the
* counter if the `own' bit is one and then sets the own bit
* to 0.
*/
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
csc = RAY_STATUS_BASE;
if (SRAM_READ_FIELD_1(sc, csc, ray_csc, csc_mrxo_own)) {
sc->sc_rxoverflow +=
SRAM_READ_FIELD_2(sc, csc, ray_csc, csc_mrx_overflow);
SRAM_WRITE_FIELD_1(sc, csc, ray_csc, csc_mrxo_own, 0);
}
if (SRAM_READ_FIELD_1(sc, csc, ray_csc, csc_mrxc_own)) {
sc->sc_rxcksum +=
SRAM_READ_FIELD_2(sc, csc, ray_csc, csc_mrx_overflow);
SRAM_WRITE_FIELD_1(sc, csc, ray_csc, csc_mrxc_own, 0);
}
if (SRAM_READ_FIELD_1(sc, csc, ray_csc, csc_rxhc_own)) {
sc->sc_rxhcksum +=
SRAM_READ_FIELD_2(sc, csc, ray_csc, csc_rx_hcksum);
SRAM_WRITE_FIELD_1(sc, csc, ray_csc, csc_rxhc_own, 0);
}
sc->sc_rxnoise = SRAM_READ_FIELD_1(sc, csc, ray_csc, csc_rx_noise);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Process CCS command completion
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
static void
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
ray_intr_ccs(struct ray_softc *sc, u_int8_t cmd, u_int8_t status, size_t ccs)
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
{
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
switch (cmd) {
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
case RAY_CMD_DOWNLOAD_PARAMS:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "START_PARAMS");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_download_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_UPDATE_PARAMS:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "UPDATE_PARAMS");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_upparams_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_REPORT_PARAMS:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "REPORT_PARAMS");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_repparams_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_UPDATE_MCAST:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "UPDATE_MCAST");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_mcast_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_START_NET:
case RAY_CMD_JOIN_NET:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "START|JOIN_NET");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_sj_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_CMD_TX_REQ:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "TX_REQ");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_tx_done(sc, status, ccs);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_START_ASSOC:
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "START_ASSOC");
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_init_assoc_done(sc, status, ccs);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_CMD_UPDATE_APM:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected UPDATE_APM");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_TEST_MEM:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected TEST_MEM");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_SHUTDOWN:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected SHUTDOWN");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_DUMP_MEM:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected DUMP_MEM");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
case RAY_CMD_START_TIMER:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected START_TIMER");
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unknown command 0x%x", cmd);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
break;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
/*
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
* Process ECF command request
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_intr_rcs(struct ray_softc *sc, u_int8_t cmd, size_t rcs)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
switch (cmd) {
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_ECMD_RX_DONE:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "RX_DONE");
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
ray_rx(sc, rcs);
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_ECMD_REJOIN_DONE:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "REJOIN_DONE");
Retire some misleading comments and explain why we need to keep a copy of parameters written to the card.
2003-03-05 18:13:19 +00:00
sc->sc_c.np_havenet = 1;
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_ECMD_ROAM_START:
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "ROAM_START");
Retire some misleading comments and explain why we need to keep a copy of parameters written to the card.
2003-03-05 18:13:19 +00:00
sc->sc_c.np_havenet = 0;
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
case RAY_ECMD_JAPAN_CALL_SIGNAL:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unexpected JAPAN_CALL_SIGNAL");
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
default:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_RECERR(sc, "unknown command 0x%x", cmd);
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
break;
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
RAY_CCS_FREE(sc, rcs);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* User land entry to multicast list changes
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
static int
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_mcast_user(struct ray_softc *sc)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_comq_entry *com[2];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int error, ncom;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
* Do all checking in the runq to preserve ordering.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
* We run promisc to pick up changes to the ALL_MULTI
* interface flag.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
ncom = 0;
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_mcast, 0);
com[ncom++] = RAY_COM_MALLOC(ray_promisc, 0);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_RUNQ(sc, com, ncom, "raymcast", error);
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
/* XXX no real error processing from anything yet! */
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_FREE(com, ncom);
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
return (error);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Runq entry to setting the multicast filter list
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
*
* MUST always be followed by a call to ray_promisc to pick up changes
* to promisc flag
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
*/
static void
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_mcast(struct ray_softc *sc, struct ray_comq_entry *com)
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ifmultiaddr *ifma;
size_t bufp;
pkh's TAILQ changes to if_mutliaddrs missed a counting loop.
2001-03-04 20:56:45 +00:00
int count = 0;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
RAY_MAP_CM(sc);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
/*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* If card is not running we don't need to update this.
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
*/
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "not running");
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
ray_com_runq_done(sc);
return;
}
/*
* The multicast list is only 16 items long so use promiscuous
* mode and don't bother updating the multicast list.
*/
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_LOCK(ifp);
Convert if_multiaddrs from LIST to TAILQ so that it can be traversed backwards in the three drivers which want to do that. Reviewed by: mikeh
2001-02-06 10:12:15 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
pkh's TAILQ changes to if_mutliaddrs missed a counting loop.
2001-03-04 20:56:45 +00:00
count++;
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
if (count == 0) {
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_UNLOCK(ifp);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
ray_com_runq_done(sc);
return;
} else if (count > 16) {
ifp->if_flags |= IFF_ALLMULTI;
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_UNLOCK(ifp);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
ray_com_runq_done(sc);
return;
} else if (ifp->if_flags & IFF_ALLMULTI)
ifp->if_flags &= ~IFF_ALLMULTI;
/*
* Kick the card
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_UPDATE_MCAST);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs,
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_cmd_update_mcast, c_nmcast, count);
bufp = RAY_HOST_TO_ECF_BASE;
Convert if_multiaddrs from LIST to TAILQ so that it can be traversed backwards in the three drivers which want to do that. Reviewed by: mikeh
2001-02-06 10:12:15 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_WRITE_REGION(
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
sc,
bufp,
LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
ETHER_ADDR_LEN
);
bufp += ETHER_ADDR_LEN;
}
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_UNLOCK(ifp);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
ray_com_ecf(sc, com);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* Complete the multicast filter list update
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_mcast_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
{
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_COM_CHECK(sc, ccs);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
ray_com_ecf_done(sc);
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Runq entry to set/reset promiscuous mode
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
static void
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_promisc(struct ray_softc *sc, struct ray_comq_entry *com)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Remove old hand crafted version of start_join_net for download.
2000-03-11 13:34:13 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
RAY_MAP_CM(sc);
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
/*
* If card not running or we already have the right flags
* we don't need to update this
*/
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
sc->sc_d.np_promisc = !!(ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI));
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING) ||
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
(sc->sc_c.np_promisc == sc->sc_d.np_promisc)) {
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
ray_com_runq_done(sc);
return;
}
/*
* Kick the card
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_UPDATE_PARAMS);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs,
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_cmd_update, c_paramid, RAY_MIB_PROMISC);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_update, c_nparam, 1);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
SRAM_WRITE_1(sc, RAY_HOST_TO_ECF_BASE, sc->sc_d.np_promisc);
Start to use ray_start_join_net (my old code removed on next checkin) requires current and desired network parameters, so I have created a network parameter structure and used two on them in the softc. This gives a bit of room of expansion. Re-wrote the update check in _net to work with this. Tidyed up softc due to above. Removed startccs and startcmd as they were only used by NetBSDs startup timeout.
2000-03-11 13:25:33 +00:00
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
ray_com_ecf(sc, com);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* User land entry to parameter reporting
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
*
* As we by pass the runq to report current parameters this function
* only provides a snap shot of the driver's state.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
static int
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_repparams_user(struct ray_softc *sc, struct ray_param_req *pr)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
struct ray_comq_entry *com[1];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int error, ncom;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* Test for illegal values or immediate responses
*/
if (pr->r_paramid > RAY_MIB_MAX)
return (EINVAL);
if ((sc->sc_version == RAY_ECFS_BUILD_4) &&
!(mib_info[pr->r_paramid][0] & RAY_V4))
return (EINVAL);
if ((sc->sc_version == RAY_ECFS_BUILD_5) &&
!(mib_info[pr->r_paramid][0] & RAY_V5))
return (EINVAL);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (pr->r_paramid > RAY_MIB_LASTUSER) {
switch (pr->r_paramid) {
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
case RAY_MIB_VERSION:
if (sc->sc_version == RAY_ECFS_BUILD_4)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*pr->r_data = RAY_V4;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
else
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*pr->r_data = RAY_V5;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
break;
case RAY_MIB_CUR_BSSID:
bcopy(sc->sc_c.np_bss_id, pr->r_data, ETHER_ADDR_LEN);
break;
case RAY_MIB_CUR_INITED:
*pr->r_data = sc->sc_c.np_inited;
break;
case RAY_MIB_CUR_DEF_TXRATE:
*pr->r_data = sc->sc_c.np_def_txrate;
break;
case RAY_MIB_CUR_ENCRYPT:
*pr->r_data = sc->sc_c.np_encrypt;
break;
case RAY_MIB_CUR_NET_TYPE:
*pr->r_data = sc->sc_c.np_net_type;
break;
case RAY_MIB_CUR_SSID:
bcopy(sc->sc_c.np_ssid, pr->r_data, IEEE80211_NWID_LEN);
break;
case RAY_MIB_CUR_PRIV_START:
*pr->r_data = sc->sc_c.np_priv_start;
break;
case RAY_MIB_CUR_PRIV_JOIN:
*pr->r_data = sc->sc_c.np_priv_join;
break;
case RAY_MIB_DES_BSSID:
bcopy(sc->sc_d.np_bss_id, pr->r_data, ETHER_ADDR_LEN);
break;
case RAY_MIB_DES_INITED:
*pr->r_data = sc->sc_d.np_inited;
break;
case RAY_MIB_DES_DEF_TXRATE:
*pr->r_data = sc->sc_d.np_def_txrate;
break;
case RAY_MIB_DES_ENCRYPT:
*pr->r_data = sc->sc_d.np_encrypt;
break;
case RAY_MIB_DES_NET_TYPE:
*pr->r_data = sc->sc_d.np_net_type;
break;
case RAY_MIB_DES_SSID:
bcopy(sc->sc_d.np_ssid, pr->r_data, IEEE80211_NWID_LEN);
break;
case RAY_MIB_DES_PRIV_START:
*pr->r_data = sc->sc_d.np_priv_start;
break;
case RAY_MIB_DES_PRIV_JOIN:
*pr->r_data = sc->sc_d.np_priv_join;
break;
Suck out all of the current and desired n/w parameters.
2000-06-11 13:57:59 +00:00
case RAY_MIB_CUR_AP_STATUS:
*pr->r_data = sc->sc_c.np_ap_status;
break;
case RAY_MIB_CUR_PROMISC:
*pr->r_data = sc->sc_c.np_promisc;
break;
case RAY_MIB_DES_AP_STATUS:
*pr->r_data = sc->sc_d.np_ap_status;
break;
case RAY_MIB_DES_PROMISC:
*pr->r_data = sc->sc_d.np_promisc;
break;
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
case RAY_MIB_CUR_FRAMING:
*pr->r_data = sc->sc_c.np_framing;
break;
case RAY_MIB_DES_FRAMING:
*pr->r_data = sc->sc_d.np_framing;
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
default:
return (EINVAL);
break;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
pr->r_failcause = 0;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
if (sc->sc_version == RAY_ECFS_BUILD_4)
pr->r_len = mib_info[pr->r_paramid][RAY_MIB_INFO_SIZ4];
else if (sc->sc_version == RAY_ECFS_BUILD_5)
pr->r_len = mib_info[pr->r_paramid][RAY_MIB_INFO_SIZ5];
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
return (0);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
pr->r_failcause = 0;
ncom = 0;
com[ncom++] = RAY_COM_MALLOC(ray_repparams, RAY_COM_FWOK);
com[ncom-1]->c_pr = pr;
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_RUNQ(sc, com, ncom, "rayrparm", error);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
/* XXX no real error processing from anything yet! */
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
if (!com[0]->c_retval && pr->r_failcause)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
error = EINVAL;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_FREE(com, ncom);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
return (error);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Runq entry to read the required parameter
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
*
* The card and driver are happy for parameters to be read
* whenever the card is plugged in
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_repparams(struct ray_softc *sc, struct ray_comq_entry *com)
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
{
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_MAP_CM(sc);
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
/*
* Kick the card
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_REPORT_PARAMS);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs,
ray_cmd_report, c_paramid, com->c_pr->r_paramid);
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_report, c_nparam, 1);
ray_com_ecf(sc, com);
}
/*
* Complete the parameter reporting
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_repparams_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
{
struct ray_comq_entry *com;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
RAY_MAP_CM(sc);
RAY_COM_CHECK(sc, ccs);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com = TAILQ_FIRST(&sc->sc_comq);
com->c_pr->r_failcause =
SRAM_READ_FIELD_1(sc, ccs, ray_cmd_report, c_failcause);
com->c_pr->r_len =
SRAM_READ_FIELD_1(sc, ccs, ray_cmd_report, c_len);
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_READ_REGION(sc, RAY_ECF_TO_HOST_BASE,
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com->c_pr->r_data, com->c_pr->r_len);
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_com_ecf_done(sc);
Mostly notes and minor style changes whilst I was travelling.
2000-03-03 17:07:42 +00:00
}
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* User land entry (and exit) to the error counters
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static int
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_repstats_user(struct ray_softc *sc, struct ray_stats_req *sr)
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
{
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
sr->rxoverflow = sc->sc_rxoverflow;
sr->rxcksum = sc->sc_rxcksum;
sr->rxhcksum = sc->sc_rxhcksum;
sr->rxnoise = sc->sc_rxnoise;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
return (0);
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
}
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* User land entry to parameter update changes
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* As a parameter change can cause the network parameters to be
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* invalid we have to re-start/join.
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static int
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_upparams_user(struct ray_softc *sc, struct ray_param_req *pr)
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
{
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
struct ray_comq_entry *com[4];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
int error, ncom, todo;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
#define RAY_UPP_SJ 0x1
#define RAY_UPP_PARAMS 0x2
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
* Check that the parameter is available based on firmware version
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
pr->r_failcause = 0;
if (pr->r_paramid > RAY_MIB_LASTUSER)
return (EINVAL);
if ((sc->sc_version == RAY_ECFS_BUILD_4) &&
!(mib_info[pr->r_paramid][0] & RAY_V4))
return (EINVAL);
if ((sc->sc_version == RAY_ECFS_BUILD_5) &&
!(mib_info[pr->r_paramid][0] & RAY_V5))
return (EINVAL);
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/*
* Handle certain parameters specially
*/
todo = 0;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
switch (pr->r_paramid) {
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
case RAY_MIB_NET_TYPE: /* Updated via START_NET JOIN_NET */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
sc->sc_d.np_net_type = *pr->r_data;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
todo |= RAY_UPP_SJ;
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
case RAY_MIB_SSID: /* Updated via START_NET JOIN_NET */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
bcopy(pr->r_data, sc->sc_d.np_ssid, IEEE80211_NWID_LEN);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
todo |= RAY_UPP_SJ;
break;
case RAY_MIB_PRIVACY_MUST_START:/* Updated via START_NET */
if (sc->sc_c.np_net_type != RAY_MIB_NET_TYPE_ADHOC)
return (EINVAL);
sc->sc_d.np_priv_start = *pr->r_data;
todo |= RAY_UPP_SJ;
break;
case RAY_MIB_PRIVACY_CAN_JOIN: /* Updated via START_NET JOIN_NET */
sc->sc_d.np_priv_join = *pr->r_data;
todo |= RAY_UPP_SJ;
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
case RAY_MIB_BASIC_RATE_SET:
sc->sc_d.np_def_txrate = *pr->r_data;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
todo |= RAY_UPP_PARAMS;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
break;
case RAY_MIB_AP_STATUS: /* Unsupported */
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
case RAY_MIB_MAC_ADDR: /* XXX Need interface up but could be done */
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
case RAY_MIB_PROMISC: /* BPF */
return (EINVAL);
break;
default:
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
todo |= RAY_UPP_PARAMS;
todo |= RAY_UPP_SJ;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
break;
}
General tidy up and plough through TODO list. Rewrote intro at top of file to reflect my better understanding of how it the memory mapping works. Clear the DONE list and move some thoughts into the TODO list. Remove RECERR from RAY_DEBUG Start to use a desired network parameter structure, only used in download code as I've realised that there are some problems with the idea. Break up ray_rx, and move the data packet handler into a seperate function. This meant some knock on changes in ray_rx_mgt/ray_rx_ctl to do with mbuf freeing. Remove some debug code/XXX comments that are out of date.
2000-06-18 21:40:46 +00:00
/*
* Generate the runq entries as needed
*/
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ncom = 0;
if (todo & RAY_UPP_PARAMS) {
com[ncom++] = RAY_COM_MALLOC(ray_upparams, 0);
com[ncom-1]->c_pr = pr;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
if (todo & RAY_UPP_SJ) {
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_sj, 0);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_auth, 0);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
com[ncom++] = RAY_COM_MALLOC(ray_init_assoc, 0);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
}
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_RUNQ(sc, com, ncom, "rayuparam", error);
MFRELENG_3
2000-05-21 21:20:18 +00:00
/* XXX no real error processing from anything yet! */
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
if (!com[0]->c_retval && pr->r_failcause)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
error = EINVAL;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_COM_FREE(com, ncom);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
return (error);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
/*
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
* Runq entry to update a parameter
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
*
Retire some misleading comments and explain why we need to keep a copy of parameters written to the card.
2003-03-05 18:13:19 +00:00
* The card and driver are basically happy for parameters to be updated
* whenever the card is plugged in. However, there may be a couple of
* network hangs whilst the update is performed. Reading parameters back
* straight away may give the wrong answer and some parameters cannot be
* read at all. Local copies should be kept.
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
static void
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_upparams(struct ray_softc *sc, struct ray_comq_entry *com)
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
{
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
RAY_MAP_CM(sc);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_UPDATE_PARAMS);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
SRAM_WRITE_FIELD_1(sc, com->c_ccs,
ray_cmd_update, c_paramid, com->c_pr->r_paramid);
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_update, c_nparam, 1);
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
SRAM_WRITE_REGION(sc, RAY_HOST_TO_ECF_BASE,
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com->c_pr->r_data, com->c_pr->r_len);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_com_ecf(sc, com);
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
}
/*
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
* Complete the parameter update, note that promisc finishes up here too
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
*/
static void
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
ray_upparams_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
{
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
struct ray_comq_entry *com;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
RAY_MAP_CM(sc);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_COM_CHECK(sc, ccs);
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com = TAILQ_FIRST(&sc->sc_comq);
switch (SRAM_READ_FIELD_1(sc, ccs, ray_cmd_update, c_paramid)) {
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
case RAY_MIB_PROMISC:
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
sc->sc_c.np_promisc = SRAM_READ_1(sc, RAY_HOST_TO_ECF_BASE);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
RAY_DPRINTF(sc, RAY_DBG_IOCTL,
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
"promisc value %d", sc->sc_c.np_promisc);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
break;
default:
com->c_pr->r_failcause =
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
SRAM_READ_FIELD_1(sc, ccs, ray_cmd_update, c_failcause);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
break;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
}
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_com_ecf_done(sc);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
* Command queuing and execution
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Set up a comq entry struct
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
*/
static struct ray_comq_entry *
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_init(struct ray_comq_entry *com, ray_comqfn_t function, int flags, char *mesg)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
{
com->c_function = function;
com->c_flags = flags;
com->c_retval = 0;
fix warnings. NULL -> 0
2003-12-24 19:00:49 +00:00
com->c_ccs = 0;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
com->c_wakeup = NULL;
com->c_pr = NULL;
com->c_mesg = mesg;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
return (com);
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
/*
* Malloc and set up a comq entry struct
*/
static struct ray_comq_entry *
ray_com_malloc(ray_comqfn_t function, int flags, char *mesg)
{
struct ray_comq_entry *com;
MALLOC(com, struct ray_comq_entry *,
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
sizeof(struct ray_comq_entry), M_RAYCOM, M_WAITOK);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
return (ray_com_init(com, function, flags, mesg));
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Add an array of commands to the runq, get some ccs's for them and
* then run, waiting on the last command.
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* We add the commands to the queue first to preserve ioctl ordering.
*
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* On recoverable errors, this routine removes the entries from the
* runq. A caller can requeue the commands (and still preserve its own
* processes ioctl ordering) but doesn't have to. When the card is
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* detached we get out quickly to prevent panics and don't bother
* about the runq.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
MFRELENG_3
2000-05-21 21:20:18 +00:00
static int
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_runq_add(struct ray_softc *sc, struct ray_comq_entry *com[], int ncom, char *wmesg)
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
{
MFRELENG_3
2000-05-21 21:20:18 +00:00
int i, error;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
error = 0;
/*
* Add the commands to the runq but don't let it run until
* the ccs's are allocated successfully
*/
com[0]->c_flags |= RAY_COM_FWAIT;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
for (i = 0; i < ncom; i++) {
com[i]->c_wakeup = com[ncom-1];
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "adding %p", com[i]);
RAY_DCOM(sc, RAY_DBG_DCOM, com[i], "adding");
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
TAILQ_INSERT_TAIL(&sc->sc_comq, com[i], c_chain);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
}
Main changes are resurection of mutlicast code (jgibbs moved the stuff I needed to add into ether_input) and finally sorting IFF_RUNNING through whole driver. As part of the IFF_RUNNING stuff, we've added an extra flag so callers can request that runq routines should check IFF_RUNNING before executing. Remove BPF taps as this is now done by ether_input. Resurrect multicast code, moving the multicast list stuff to the runq routine. Dump ray_promisc_user as all flag changes are now handled by ray_init, and add a couple of checks to ray_promisc. In uppparams_user, allow changes before the card is running (need to fix some breakage with _download here later). In addition, don't assume that the current n/w parameters are valid - they are only valid in the runq. Fix a nasty flag bug - runq_add cleared all the flags on the last command! Remove the hacks for setting the memory flags - problems were down to buggy versions of pccardd. For some reason pccardd only dtrt with the "right" debug_level.
2000-06-04 17:51:36 +00:00
com[ncom-1]->c_flags |= RAY_COM_FWOK;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* Allocate ccs's for each command.
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
for (i = 0; i < ncom; i++) {
error = ray_ccs_alloc(sc, &com[i]->c_ccs, wmesg);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
if (error == ENXIO)
return (ENXIO);
else if (error)
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
goto cleanup;
MFRELENG_3
2000-05-21 21:20:18 +00:00
}
/*
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* Allow the queue to run and sleep if needed.
*
* Iff the FDETACHED flag is set in the com entry we waited on
* the driver is in a zombie state! The softc structure has been
* freed by the generic bus detach methods - eek. We tread very
* carefully!
MFRELENG_3
2000-05-21 21:20:18 +00:00
*/
com[0]->c_flags &= ~RAY_COM_FWAIT;
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
ray_com_runq(sc);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
if (TAILQ_FIRST(&sc->sc_comq) != NULL) {
RAY_DPRINTF(sc, RAY_DBG_COM, "sleeping");
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
error = tsleep(com[ncom-1], PCATCH | PRIBIO, wmesg, 0);
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
if (com[ncom-1]->c_flags & RAY_COM_FDETACHED)
return (ENXIO);
RAY_DPRINTF(sc, RAY_DBG_COM,
"awakened, tsleep returned 0x%x", error);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
} else
error = 0;
MFRELENG_3
2000-05-21 21:20:18 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
cleanup:
/*
* Only clean the queue on real errors - we don't care about it
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
* when we detach as the queue entries are freed by the callers.
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
*/
if (error && (error != ENXIO))
for (i = 0; i < ncom; i++)
if (!(com[i]->c_flags & RAY_COM_FCOMPLETED)) {
RAY_DPRINTF(sc, RAY_DBG_COM, "removing %p",
com[i]);
RAY_DCOM(sc, RAY_DBG_DCOM, com[i], "removing");
TAILQ_REMOVE(&sc->sc_comq, com[i], c_chain);
ray_ccs_free(sc, com[i]->c_ccs);
fix warnings. NULL -> 0
2003-12-24 19:00:49 +00:00
com[i]->c_ccs = 0;
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
}
MFRELENG_3
2000-05-21 21:20:18 +00:00
return (error);
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Run the command at the head of the queue (if not already running)
*/
static void
ray_com_runq(struct ray_softc *sc)
{
struct ray_comq_entry *com;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
com = TAILQ_FIRST(&sc->sc_comq);
MFRELENG_3
2000-05-21 21:20:18 +00:00
if ((com == NULL) ||
(com->c_flags & RAY_COM_FRUNNING) ||
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
(com->c_flags & RAY_COM_FWAIT) ||
(com->c_flags & RAY_COM_FDETACHED))
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
return;
com->c_flags |= RAY_COM_FRUNNING;
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "running %p", com);
RAY_DCOM(sc, RAY_DBG_DCOM, com, "running");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
com->c_function(sc, com);
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
/*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Remove run command, free ccs and wakeup caller.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* Minimal checks are done here as we ensure that the com and command
* handler were matched up earlier. Must be called at splnet or higher
* so that entries on the command queue are correctly removed.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
* Remove the com from the comq, and wakeup the caller if it requested
* to be woken. This is used for ensuring a sequence of commands
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* completes. Finally, re-run the queue.
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
*/
static void
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_runq_done(struct ray_softc *sc)
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_comq_entry *com;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved a bit of resource allocation from probe to attach based on feedback in -hackers on the correct driver structure. Started to tidy up _detach and _stop. Added the comq array stuff into mcast.
2000-05-23 16:38:12 +00:00
com = TAILQ_FIRST(&sc->sc_comq); /* XXX shall we check this as below */
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "removing %p", com);
RAY_DCOM(sc, RAY_DBG_DCOM, com, "removing");
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
TAILQ_REMOVE(&sc->sc_comq, com, c_chain);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
com->c_flags &= ~RAY_COM_FRUNNING;
com->c_flags |= RAY_COM_FCOMPLETED;
com->c_retval = 0;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_ccs_free(sc, com->c_ccs);
fix warnings. NULL -> 0
2003-12-24 19:00:49 +00:00
com->c_ccs = 0;
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (com->c_flags & RAY_COM_FWOK)
wakeup(com->c_wakeup);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
ray_com_runq(sc);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/* XXX what about error on completion then? deal with when i fix
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* XXX the status checking
*
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
* XXX all the runq_done calls from IFF_DRV_RUNNING checks in runq
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* XXX routines should return EIO but shouldn't abort the runq
*/
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
}
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* Send a command to the ECF.
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
*/
static void
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_ecf(struct ray_softc *sc, struct ray_comq_entry *com)
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
{
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
int i = 0;
A messy commit that checkpoints the driver (not known to actually work) before I rip out the scheduler - whilst v. nice 'n all, it is doing the wrong job. We need something that sends commands to the card atomically so dhcp etc. works right. I've renamed and moved a lot of the scheduler code so that it is all in one place and all starts with ray_cmd_ ray_stop has some debugging crap left in - to be deleted rsn
2000-04-15 19:51:18 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
RAY_MAP_CM(sc);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
while (!RAY_ECF_READY(sc)) {
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
DELAY(RAY_ECF_SPIN_DELAY);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if (++i > RAY_ECF_SPIN_TRIES)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_PANIC(sc, "spun too long");
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
}
if (i != 0)
RAY_RECERR(sc, "spun %d times", i);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "sending %p", com);
RAY_DCOM(sc, RAY_DBG_DCOM, com, "sending");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
SRAM_WRITE_1(sc, RAY_SCB_CCSI, RAY_CCS_INDEX(com->c_ccs));
RAY_ECF_START_CMD(sc);
if (RAY_COM_NEEDS_TIMO(
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
SRAM_READ_FIELD_1(sc, com->c_ccs, ray_cmd, c_cmd))) {
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_COM, "adding timeout");
sc->com_timerh = timeout(ray_com_ecf_timo, sc, RAY_COM_TIMEOUT);
}
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* Deal with commands that require a timeout to test completion.
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* This routine is coded to only expect one outstanding request for the
* timed out requests at a time, but thats all that can be outstanding
* per hardware limitations and all that we issue anyway.
*
* We don't do any fancy testing of the command currently issued as we
* know it must be a timeout based one...unless I've got this wrong!
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static void
ray_com_ecf_timo(void *xsc)
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
{
Sync with an early version of the -current code. Mainly a tidy up of some function names, dropping the ISA bits that were never used and removing __P We've renamed the _pccard_ stuff as nothing else probes or attaches. In the next commit we will move some routines around and split out some of the structure definitions to another file. I've dropped the SYSCTL as I never got it working. APM is gone too as it is broken. Note, that the attr_read/write will be in both this and NEWCARD/NEWBUS so I've moved the prototypes to the right place. Add a few casts.
2000-05-07 15:00:06 +00:00
struct ray_softc *sc = (struct ray_softc *)xsc;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_comq_entry *com;
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
u_int8_t cmd, status;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
int s;
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
s = splnet();
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
RAY_MAP_CM(sc);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
com = TAILQ_FIRST(&sc->sc_comq);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved debugging definitions to a header file. Removed NETBSDTX spares. Renamed a load of functions - specifcally start -> tx init has download, sj and assoc as sub-functions report_params and update params move to repparams/upparams Tidied up old #define's Got rid of old DPRINTF, made printf's RAY_PRINTF and panic's RAY_PANIC All code KNF Removed checking the ECF_TO_HOST area in ray_init - this gets hosed when someone updates/read a parameter from the card so causes unneeded grief. This required moving the tib check into ray_attach. Changed handling of interface flags in ioctl so that promiscous mode changes are only done as needed. Sequences of comq entries can be added to an array and automatically dealt with - used in ray_init_user and others. Moved IFF_RUNNING checks from the comq commands to com_runq - still not sure what to do so we PANIC - will be fixed. ray_sj now checks to see if any of the parameters it can update can be updated. ray_sj_done now updates parameters to the current n/w set if we changed them earlier. I was being a bit thick in earlier comments as to why this check was done - the ECF never changes the parameters. Assocication with APs is handled outside of ray_sj - need to add WEP stuff if I ever get my hands on an AP. ray_stop, ray_unload and ray_reset are currently broken - reset isn't used anyway. ray_tx_XXX and ray_rx only have cosmetic changes. Interrupt handler now gets command out of ccs. This is so ray_intr_ccs and ray_intr_rcs don't need to remap CM (I'm trying to roll things up a bit as it must be slowing us down). ray_intr_ccs just vectors commands to functions or error checks - a jump table could replace it in ray_intr. Ditto for ray_intr_rcs. mcast is currently broken - I need to do more work for ALLMULTI etc. This is reasonably easy to fix becuase of the comq array stuff. Added a few more checks into repparams so that it is firmware version aware. It also uses the new comq framework. Parameter updating is also much better with the update parameter return routine re-worked. Added a couple of runq entry helper functions - one that malloc's a new entry and fills it in with sensible defaults and another that adds and runs a seqencue of entries.
2000-04-24 15:19:40 +00:00
cmd = SRAM_READ_FIELD_1(sc, com->c_ccs, ray_cmd, c_cmd);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
status = SRAM_READ_FIELD_1(sc, com->c_ccs, ray_cmd, c_status);
switch (status) {
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
case RAY_CCS_STATUS_COMPLETE:
case RAY_CCS_STATUS_FREE: /* Buggy firmware */
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
ray_intr_ccs(sc, cmd, status, com->c_ccs);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
case RAY_CCS_STATUS_BUSY:
sc->com_timerh = timeout(ray_com_ecf_timo, sc, RAY_COM_TIMEOUT);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
break;
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
default: /* Replicates NetBSD */
if (sc->sc_ccsinuse[RAY_CCS_INDEX(com->c_ccs)] == 1) {
/* give a chance for the interrupt to occur */
sc->sc_ccsinuse[RAY_CCS_INDEX(com->c_ccs)] = 2;
sc->com_timerh = timeout(ray_com_ecf_timo, sc,
RAY_COM_TIMEOUT);
} else
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
ray_intr_ccs(sc, cmd, status, com->c_ccs);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
break;
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
splx(s);
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
* Called when interrupt handler for the command has done all it
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
* needs to. Will be called at splnet.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
static void
ray_com_ecf_done(struct ray_softc *sc)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
untimeout(ray_com_ecf_timo, sc, sc->com_timerh);
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_com_runq_done(sc);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
#if RAY_DEBUG & RAY_DBG_COM
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* Process completed ECF commands that probably came from the command queue
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* This routine is called after vectoring the completed ECF command
* to the appropriate _done routine. It helps check everything is okay.
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
*/
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
static void
ray_com_ecf_check(struct ray_softc *sc, size_t ccs, char *mesg)
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
struct ray_comq_entry *com;
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "%s", mesg);
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
com = TAILQ_FIRST(&sc->sc_comq);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (com == NULL)
RAY_PANIC(sc, "no command queue");
if (com->c_ccs != ccs)
RAY_PANIC(sc, "ccs's don't match");
}
#endif /* RAY_DEBUG & RAY_DBG_COM */
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* CCS allocators
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
/*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Obtain a ccs for a commmand
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*
MFRELENG_3
2000-05-21 21:20:18 +00:00
* Returns 0 and in `ccsp' the bus offset of the free ccs. Will block
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
* awaiting free ccs if needed - if the sleep is interrupted
* EINTR/ERESTART is returned, if the card is ejected we return ENXIO.
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
*/
static int
MFRELENG_3
2000-05-21 21:20:18 +00:00
ray_ccs_alloc(struct ray_softc *sc, size_t *ccsp, char *wmesg)
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
{
size_t ccs;
u_int i;
MFRELENG_3
2000-05-21 21:20:18 +00:00
int error;
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(sc);
Added user level updates and reading of parameters.
2000-03-21 14:27:46 +00:00
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
for (;;) {
for (i = RAY_CCS_CMD_FIRST; i <= RAY_CCS_CMD_LAST; i++) {
/* we probe here to make the card go */
(void)SRAM_READ_FIELD_1(sc, RAY_CCS_ADDRESS(i), ray_cmd,
c_status);
if (!sc->sc_ccsinuse[i])
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
if (i > RAY_CCS_CMD_LAST) {
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "sleeping");
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
error = tsleep(ray_ccs_alloc, PCATCH | PRIBIO,
wmesg, 0);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if ((sc == NULL) || (sc->sc_gone))
Removed a few RAY_MAP_CMs that were not needed. Made checking sc->gone a lot safer by checking to see if sc is NULL first. Made return from tsleep in the ccs allocator detach safe.
2000-06-10 21:24:36 +00:00
return (ENXIO);
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS,
"awakened, tsleep returned 0x%x", error);
MFRELENG_3
2000-05-21 21:20:18 +00:00
if (error)
return (error);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
} else
break;
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
}
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "allocated 0x%02x", i);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
sc->sc_ccsinuse[i] = 1;
ccs = RAY_CCS_ADDRESS(i);
MFRELENG_3
2000-05-21 21:20:18 +00:00
*ccsp = ccs;
return (0);
}
/*
* Fill the easy bits in of a pre-allocated CCS
*/
static void
ray_ccs_fill(struct ray_softc *sc, size_t ccs, u_int cmd)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(sc);
fix warnings. NULL -> 0
2003-12-24 19:00:49 +00:00
if (ccs == 0)
MFRELENG_3
2000-05-21 21:20:18 +00:00
RAY_PANIC(sc, "ccs not allocated");
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd, c_status, RAY_CCS_STATUS_BUSY);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd, c_cmd, cmd);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd, c_link, RAY_CCS_LINK_NULL);
Completed start/join code (well only infrastructure mode) Added receive code and support for Webgear encapsulation. More debugging macros/functions. conditionalised timeout for start/join network conditonalised attribute/common memory hacks identified tracking code with XXX_TRACK sorted out initialistion of instance structure to some extent finished docuementing the start/join sequence
2000-02-27 19:52:29 +00:00
}
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
/*
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
* Free up a ccs allocated via ray_ccs_alloc
*
* Return the old status. This routine is only used for ccs allocated via
* ray_ccs_alloc (not tx, rx or ECF command requests).
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
*/
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
static void
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
ray_ccs_free(struct ray_softc *sc, size_t ccs)
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
{
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(sc);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#if 1 | (RAY_DEBUG & RAY_DBG_CCS)
if (!sc->sc_ccsinuse[RAY_CCS_INDEX(ccs)])
ccs is a size_t. RAY_CCS_INDEX takes ccs and does math on it, resulting in a size_t due to C's rules of arithmetic. Rather than bogusly cast the result to a uint8_t, fix the printf format specifier to have a 'z' modifier which tells the compiler that the sizes really do match. It turns out that change 1.75 was incorrect to assume that this 'really' was a 8bit quantity. It isn't. Although the hardware appears to limit things to < 256, it would be a bug that should be caught by debug printf it it were. Casting it to uint8_t would have lost this useful information. Aslo add 'z' to a nearby debug statement that's never compiled in.
2005-02-06 21:12:22 +00:00
RAY_RECERR(sc, "freeing free ccs 0x%02zx", RAY_CCS_INDEX(ccs));
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
#endif /* RAY_DEBUG & RAY_DBG_CCS */
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
if (!sc->sc_gone)
Remove promisc_user, ray_reset and ray_reset_timo as they are not used. Incidental remove of a timer too. Remove the runq_abort code. Get ray_detach working correctly. This is a very simple routine as it just wakes up sleeping processes. Note that anything woken has NO softc structure available! runq_add is suitably modified to detect a detach and return straight away. Due to ray_detach and its implications use a macro for adding things to the runq in user land.
2000-06-10 13:50:57 +00:00
RAY_CCS_FREE(sc, ccs);
Moved functions around so that they are grouped a little more sensibly.
2000-04-22 14:22:51 +00:00
sc->sc_ccsinuse[RAY_CCS_INDEX(ccs)] = 0;
ccs is a size_t. RAY_CCS_INDEX takes ccs and does math on it, resulting in a size_t due to C's rules of arithmetic. Rather than bogusly cast the result to a uint8_t, fix the printf format specifier to have a 'z' modifier which tells the compiler that the sizes really do match. It turns out that change 1.75 was incorrect to assume that this 'really' was a 8bit quantity. It isn't. Although the hardware appears to limit things to < 256, it would be a bug that should be caught by debug printf it it were. Casting it to uint8_t would have lost this useful information. Aslo add 'z' to a nearby debug statement that's never compiled in.
2005-02-06 21:12:22 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "freed 0x%02zx", RAY_CCS_INDEX(ccs));
MFRELENG_3
2000-05-21 21:20:18 +00:00
wakeup(ray_ccs_alloc);
* add the start_join_net - done * i needed it anyway * remove startccs and startcmd - done * as those were used for the NetBSD start timeout * multicast - done but UNTESTED * I don't have the ability/facilty to test this * rxlevel - done * stats reported via raycontrol * getparams ioctl - done * reported via raycontrol * start_join_done needs a restart in download_done - done * now use netbsd style start up removed spare timerh implemented start_assoc in ray_ccs_done bit more knfing
2000-03-12 22:54:29 +00:00
}
Functional driver with new command queue in place. Working download and start/join network. What's nice is that I didn't touch TX or RX code at all. Not fully done yet - stubs for assoc, cast, promisc and user param updating. I can't work out the "nicest" way to get the mcast and promisc interactions sorted without needing to abort runq commands half way through. And, as mcast/promisc used elsewhere I want to get this right. Other stuff Fancy new debugging with __FUNCTION__ and __LINE__. Lots of panic's inserted until I check some stuff out. Removed all of the old scheduler. Removed start join timeout as I don't need it. Removed the promiscuous broadcast code that waas ifdef'd out. New CCS allocation
2000-04-21 15:01:49 +00:00
MFRELENG_3
2000-05-21 21:20:18 +00:00
/*
* Obtain a ccs and tx buffer to transmit with and fill them in.
*
* Returns 0 and in `ccsp' the bus offset of the free ccs. Will not block
* and if none available and will returns EAGAIN.
*
* The caller must fill in the length later.
* The caller must clear the ccs on errors.
*/
static int
ray_ccs_tx(struct ray_softc *sc, size_t *ccsp, size_t *bufpp)
{
size_t ccs, bufp;
int i;
u_int8_t status;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(sc);
i = RAY_CCS_TX_FIRST;
do {
status = SRAM_READ_FIELD_1(sc, RAY_CCS_ADDRESS(i),
ray_cmd, c_status);
if (status == RAY_CCS_STATUS_FREE)
break;
i++;
} while (i <= RAY_CCS_TX_LAST);
if (i > RAY_CCS_TX_LAST) {
return (EAGAIN);
}
RAY_DPRINTF(sc, RAY_DBG_CCS, "allocated 0x%02x", i);
/*
* Reserve and fill the ccs - must do the length later.
*
* Even though build 4 and build 5 have different fields all these
* are common apart from tx_rate. Neither the NetBSD driver or Linux
* driver bother to overwrite this for build 4 cards.
*
* The start of the buffer must be aligned to a 256 byte boundary
* (least significant byte of address = 0x00).
*/
ccs = RAY_CCS_ADDRESS(i);
bufp = RAY_TX_BASE + i * RAY_TX_BUF_SIZE;
bufp += sc->sc_tibsize;
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_status, RAY_CCS_STATUS_BUSY);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_cmd, RAY_CMD_TX_REQ);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_link, RAY_CCS_LINK_NULL);
SRAM_WRITE_FIELD_2(sc, ccs, ray_cmd_tx, c_bufp, bufp);
SRAM_WRITE_FIELD_1(sc,
ccs, ray_cmd_tx, c_tx_rate, sc->sc_c.np_def_txrate);
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_apm_mode, 0);
MFRELENG_3
2000-05-21 21:20:18 +00:00
bufp += sizeof(struct ray_tx_phy_header);
*ccsp = ccs;
*bufpp = bufp;
return (0);
}
Skeleton framework for network driver and initial probing of device.
2000-02-20 14:56:17 +00:00
/*
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
* Routines to obtain resources for the card
*/
/*
* Allocate the attribute memory on the card
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
*
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
* The attribute memory space is abused by these devices as IO space. As such
* the OS card services don't have a chance of knowing that they need to keep
* the attribute space mapped. We have to do it manually.
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
*/
static int
ray_res_alloc_am(struct ray_softc *sc)
{
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
int error;
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CM, "");
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
sc->am_rid = RAY_AM_RID;
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
sc->am_res = bus_alloc_resource(sc->dev, SYS_RES_MEMORY,
&sc->am_rid, 0UL, ~0UL, 0x1000, RF_ACTIVE);
if (!sc->am_res) {
RAY_PRINTF(sc, "Cannot allocate attribute memory");
return (ENOMEM);
}
error = CARD_SET_MEMORY_OFFSET(device_get_parent(sc->dev), sc->dev,
sc->am_rid, 0, NULL);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
if (error) {
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_PRINTF(sc, "CARD_SET_MEMORY_OFFSET returned 0x%0x", error);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
return (error);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
}
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->am_rid, PCCARD_A_MEM_ATTR);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
if (error) {
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
A bunch of misc. tidy ups really. Generated a new macor, RAY_RECERR for reporting errors with. Verbosity set with IFF_DEBUG (recommended at present). Add PRIBIO to tsleeps. Catch detach on ray_ccs_alloc a little better. Move sc_promisc into desired and current n/w parameters. Remove IFQ_PEEK, we know the driver runs okay without it. Drain the output queue in ray_stop. Only use ray_mcast for ADD/DEL multi ioctls. ray_init_multi resets the multicast list on startup. Simplifies ray_init a little. Tidy some old comments. ray_download_done now copies the whole desired n/w parameter set into the current set. This is because I was missing soem parameters - like the net type!
2000-06-11 13:32:07 +00:00
return (error);
}
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->am_rid, PCCARD_A_MEM_8BIT);
if (error) {
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
return (error);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
sc->am_bsh = rman_get_bushandle(sc->am_res);
sc->am_bst = rman_get_bustag(sc->am_res);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
#if RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM)
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
{
Primary purpose of this commit is to enable support for the Aviator Pro and Raylink cards with version 5 firmware. Only infra-structure mode has been tested. Specific changes for this feature are: o Add RFC1042 encapsulation of IP datagrams o Add authentication and association o Decode of the beacon (although not used) Other changes have been made: o Pass command completion status to *_done (in place for adding proper error recovery) o Move a couple of state variables into the current network parameters structure. This is in prep. for dealing with roaming. MFC after: 1 week
2001-05-17 22:23:49 +00:00
u_long flags;
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
u_int32_t offset;
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
CARD_GET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->am_rid, &flags);
CARD_GET_MEMORY_OFFSET(device_get_parent(sc->dev), sc->dev,
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
sc->am_rid, &offset);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
RAY_PRINTF(sc, "allocated attribute memory:\n"
". start 0x%0lx count 0x%0lx flags 0x%0lx offset 0x%0x",
bus_get_resource_start(sc->dev, SYS_RES_MEMORY, sc->am_rid),
bus_get_resource_count(sc->dev, SYS_RES_MEMORY, sc->am_rid),
flags, offset);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
#endif /* RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM) */
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
return (0);
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
/*
* Allocate the common memory on the card
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
*
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
* As this memory is described in the CIS, the OS card services should
* have set the map up okay, but the card uses 8 bit RAM. This is not
* described in the CIS.
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
*/
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
static int
ray_res_alloc_cm(struct ray_softc *sc)
{
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
u_long start, count, end;
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
int error;
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CM, "");
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_DPRINTF(sc,RAY_DBG_CM | RAY_DBG_BOOTPARAM,
"cm start 0x%0lx count 0x%0lx",
bus_get_resource_start(sc->dev, SYS_RES_MEMORY, RAY_CM_RID),
bus_get_resource_count(sc->dev, SYS_RES_MEMORY, RAY_CM_RID));
Take advantage of the fixes to the pcic code that allows multiple active memory maps. This removes the need to change the memory map from common to attribute every time a packet is sent/received. This increases performance and decreases cpu load (ping times on slow machines improve by about 1.5ms). Move out the old common memory/attrbiute memory hack functions to a new header file to tidy up the main code. I want to keep them available for a while.
2001-01-17 17:55:00 +00:00
sc->cm_rid = RAY_CM_RID;
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
start = bus_get_resource_start(sc->dev, SYS_RES_MEMORY, sc->cm_rid);
count = bus_get_resource_count(sc->dev, SYS_RES_MEMORY, sc->cm_rid);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
end = start + count - 1;
sc->cm_res = bus_alloc_resource(sc->dev, SYS_RES_MEMORY,
&sc->cm_rid, start, end, count, RF_ACTIVE);
if (!sc->cm_res) {
RAY_PRINTF(sc, "Cannot allocate common memory");
return (ENOMEM);
}
error = CARD_SET_MEMORY_OFFSET(device_get_parent(sc->dev), sc->dev,
sc->cm_rid, 0, NULL);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
if (error) {
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_PRINTF(sc, "CARD_SET_MEMORY_OFFSET returned 0x%0x", error);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
return (error);
}
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->cm_rid, PCCARD_A_MEM_COM);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
if (error) {
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
return (error);
}
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->cm_rid, PCCARD_A_MEM_8BIT);
if (error) {
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
return (error);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
}
sc->cm_bsh = rman_get_bushandle(sc->cm_res);
sc->cm_bst = rman_get_bustag(sc->cm_res);
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
#if RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM)
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
{
u_long flags;
u_int32_t offset;
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
CARD_GET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
SYS_RES_MEMORY, sc->cm_rid, &flags);
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
CARD_GET_MEMORY_OFFSET(device_get_parent(sc->dev), sc->dev,
Add attribute memory setup warnings
2000-07-11 21:31:59 +00:00
sc->cm_rid, &offset);
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
RAY_PRINTF(sc, "allocated common memory:\n"
". start 0x%0lx count 0x%0lx flags 0x%0lx offset 0x%0x",
bus_get_resource_start(sc->dev, SYS_RES_MEMORY, sc->cm_rid),
bus_get_resource_count(sc->dev, SYS_RES_MEMORY, sc->cm_rid),
flags, offset);
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
#endif /* RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM) */
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
return (0);
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
/*
* Get an irq and attach it to the bus
*/
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
static int
ray_res_alloc_irq(struct ray_softc *sc)
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
int error;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
Move newbus detection code to alloc routines. Work around pccard nasties.
2000-07-07 19:13:11 +00:00
RAY_DPRINTF(sc,RAY_DBG_CM | RAY_DBG_BOOTPARAM,
"irq start 0x%0lx count 0x%0lx",
bus_get_resource_start(sc->dev, SYS_RES_IRQ, 0),
bus_get_resource_count(sc->dev, SYS_RES_IRQ, 0));
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
sc->irq_rid = 0;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->irq_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, &sc->irq_rid,
RF_ACTIVE);
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
if (!sc->irq_res) {
Minor tweaks to error messages (after writing man page). Renamed varible dst in ray_rx to mp as it is a pointer to an mbuf. Correctly grok addresses in data packets. Promte a couple of RECERRs to real errors.
2000-06-20 20:14:29 +00:00
RAY_PRINTF(sc, "Cannot allocate irq");
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
return (ENOMEM);
}
if ((error = bus_setup_intr(sc->dev, sc->irq_res, INTR_TYPE_NET,
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
NULL, ray_intr, sc, &sc->irq_handle)) != 0) {
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
RAY_PRINTF(sc, "Failed to setup irq");
return (error);
}
RAY_DPRINTF(sc, RAY_DBG_CM | RAY_DBG_BOOTPARAM, "allocated irq:\n"
A bit of a messy monster. Simple stuff ------------ Split _download up so that the MIB settings are in their own functions. Made "tx completed but status is ..." a recoverable error Cut down verbosity of "unloaded" messages Moved ccs_free and com_runq from intr_ccs to ecf_done and runq_done to avoid embarasing mistakes and waits. Merged runq_add and runq_arr into one and called it runq_add Made RAY_COM_DUMP a real debug called RAY_DCOM Consistnet debugging around tsleeps. Use bus_activate_resource for attr/cm mapping, and set the flags correctly in the allocation routines (needs more hacks to sys/pccard/pcic.c) com_malloc is now seperated from the comq initialization. This was done whilst trying to set automatic variables for the comqs. Harder Stuff ------------ As part of the IFF_RUNNING fixes, remove the panic in runq if we are not running. Change, again, runq_add. This time we don't do any cleaning up if there were errors. This is so that callers get the chance to re-try (not that I ever see it being used). In runq_add, only sleep when there is something to sleep on! ioctl locking routines, stolen from awi.c but not used Hardest Stuff ------------- Dealing with serialing ioctls correctly means that we must QUEUE changes to IFF_RUNNING and check it in the QUEUED commands, not in the user commands. Whilst simple to state, it took a few hours of head scratching to get it right. The realisation was that I only have to guarantee that sub-commands from a single process are serialised and "atomic", and that they check the status of the interface flags when invoked and not when they are queued. Another way of looking at it, is that the driver's state is stored in the runq and the IFF_RUNNING flag. These must be changed together. What this means practically, is that IFF_RUNNING is set after we have started/joined/associated with a network. And it is cleared by ray_stop via the runq so that unfinsished commands are not distrupted. I still have to fix up promisc, upp/repparams and mcast. Oh yeah, stop is essentially a noop in that it only changes IFF_RUNNING
2000-05-28 23:10:12 +00:00
". start 0x%0lx count 0x%0lx",
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
bus_get_resource_start(sc->dev, SYS_RES_IRQ, sc->irq_rid),
bus_get_resource_count(sc->dev, SYS_RES_IRQ, sc->irq_rid));
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
return (0);
}
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
/*
* Release all of the card's resources
*/
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
static void
ray_res_release(struct ray_softc *sc)
{
Newbusify resource allocation. Have OLDCARD version of the remapping. Remove BPF conditionals. Remove ISA/3.x stuff and add -current stuff.
2000-05-11 18:55:38 +00:00
if (sc->irq_res != 0) {
bus_teardown_intr(sc->dev, sc->irq_res, sc->irq_handle);
bus_release_resource(sc->dev, SYS_RES_IRQ,
sc->irq_rid, sc->irq_res);
sc->irq_res = 0;
}
if (sc->am_res != 0) {
bus_release_resource(sc->dev, SYS_RES_MEMORY,
sc->am_rid, sc->am_res);
sc->am_res = 0;
}
if (sc->cm_res != 0) {
bus_release_resource(sc->dev, SYS_RES_MEMORY,
sc->cm_rid, sc->cm_res);
sc->cm_res = 0;
}
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
}
/*
* mbuf dump
*/
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
#if RAY_DEBUG & RAY_DBG_MBUF
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
static void
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
ray_dump_mbuf(struct ray_softc *sc, struct mbuf *m, char *s)
Mega additions of NetBSD code - most of the NetBSD code is now in here, if not actually used yet. This created a lot of prototypes. I moved code around so that the functions are in the same order as NetBSD - this helps diff a lot. Things that are missing are the TX routine and ifmedia stuff and ioctls Not all of the stuff is enabled - we are missing calls in ccs_done and ioctl. Promiscious mode is working. Am almost ready to use the NetBSD start up routine - essentially all I need to do is not use a seperate timer to call download_timo. Other misc. things: callout_stop is a newish feature for cancelling a timer without argument checking it. tried to add a sysctl knob but it doesn't work in the module enabled infrastructure code to call netbsd bits.
2000-03-08 23:28:06 +00:00
{
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
u_int8_t *d, *ed;
u_int i;
char p[17];
Eliminate some panics for errors we can recover from. Reduce the verbose memory map setup reports and work with pccardd to set the common memory map up. Use enumeration values for CARD_SET_RES_FLAGS. Use DELAY when spinning waiting for the card to come free instead of a loop. MFC: after 1 week
2001-05-09 00:03:19 +00:00
RAY_PRINTF(sc, "%s", s);
Remove if_ray_oldcard.h because pccard support multiple windows now. Setup attribute memory resource in ray_probe so that it is added to the print out of the resource list on card insertion.
2001-05-07 13:11:08 +00:00
RAY_PRINTF(sc, "\nm0->data\t0x%p\nm_pkthdr.len\t%d\nm_len\t%d",
mtod(m, u_int8_t *), m->m_pkthdr.len, m->m_len);
Move a couple fo things around to be nicer for -current Move to using /sys/dev/ray for driver Backport -current ray_res code - most are stubs. Make ray_attr stuff look like currents layout. Make attribute memory access use a macro and ripple through. Drop FIXUP as we always need it
2000-05-10 21:19:38 +00:00
i = 0;
bzero(p, 17);
for (; m; m = m->m_next) {
d = mtod(m, u_int8_t *);
ed = d + m->m_len;
for (; d < ed; i++, d++) {
if ((i % 16) == 0) {
printf(" %s\n\t", p);
} else if ((i % 8) == 0)
printf(" ");
printf(" %02x", *d);
p[i % 16] = ((*d >= 0x20) && (*d < 0x80)) ? *d : '.';
}
}
if ((i - 1) % 16)
MFRELENG_3
2000-05-21 21:20:18 +00:00
printf(" %s\n", p);
Snap shot check in. The code doesn't compile (I'll do that in a minute), but functionally it gets to receveing the interrupt from the start or join net command. There is a lot of code to deal with the limitations of the PCCard memory mapping code - I'm changing some of it hence the check in to preserve something that worked.
2000-02-26 14:49:24 +00:00
}
Most of this commit are minor changes whilst trying to make the driver robust enough to work with dhcp. Essentially the driver returns to userland before certain commands have been completed by the card. I'm trying to work out the best way to fix this. There are very minor functional changes, most of this is debugging. Rework of debugging so a bit mask is used and re-allocation around the place. Tried to protect ray_init with a tsleep until the network has been joined. I think this is only a partial fix as we also update the mcast and promisc values.
2000-04-09 21:07:43 +00:00
#endif /* RAY_DEBUG & RAY_DBG_MBUF */
Reference in New Issue Copy Permalink