freebsd-dev/sys/dev/ray/if_ray.c

3805 lines
105 KiB
C
Raw Normal View History

/*
* 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.
*
* $FreeBSD$
*
*/
/* $NetBSD: if_ray.c,v 1.12 2000/02/07 09:36:27 augustss Exp $ */
/*
* 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.
*/
/*
* Card configuration
* ==================
*
* This card is unusual in that it uses both common and attribute
* memory whilst working. It should use common memory and an IO port.
*
* 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.
*
* There is no support for running this driver on 4.0.
*
* Ad-hoc and infra-structure modes
* ================================
*
* 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.
*
* 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.
*
*
* Packet framing/encapsulation/translation
* ========================================
*
* Currently we support the Webgear encapsulation:
* 802.11 header <net/if_ieee80211.h>struct ieee80211_frame
* 802.3 header <net/ethernet.h>struct ether_header
* IP/ARP payload
*
* and RFC1042 encapsulation of IP datagrams (translation):
* 802.11 header <net/if_ieee80211.h>struct ieee80211_frame
* 802.2 LLC header
* 802.2 SNAP header
* 802.3 Ethertype
* IP/ARP payload
*
* Framing should be selected via if_media stuff or link types but
* is currently hardcoded to:
* V4 encapsulation
* V5 translation
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!
*/
/*
* ***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?
* ***error handling of RAY_COM_RUNQ
* ***error handling of ECF command completions
* ***can't seem to create a n/w that Win95 wants to see.
* ***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
* write up driver structure in comments above
* UPDATE_PARAMS seems to return via an interrupt - maybe the timeout
* is needed for wrong values?
* proper setting of mib_hop_seq_len with country code for v4 firmware
* best done with raycontrol?
* countrycode setting is broken I think
* userupdate should trap and do via startjoin etc.
* fragmentation when rx level drops?
* 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?
*
* 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
* acting as ap - should be able to get working from the manual
* need to finish RAY_ECMD_REJOIN_DONE
* finish authenitcation code, it doesn't handle errors/timeouts/
* REJOIN etc.
*
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
* promisc in here too? - done
* should be able to update the parameters before we download to the
* 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
* 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?
* 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
*/
#define XXX 0
2000-05-21 21:20:18 +00:00
#define XXX_ACTING_AP 0
#define XXX_INFRA 0
#define RAY_DEBUG ( \
/* 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 | */ \
/* RAY_DBG_BOOTPARAM | */ \
/* 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 | */ \
/* 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 | */ \
/* RAY_DBG_RX | */ \
/* RAY_DBG_CM | */ \
/* RAY_DBG_COM | */ \
/* RAY_DBG_STOP | */ \
2000-05-21 21:20:18 +00:00
/* RAY_DBG_CTL | */ \
/* RAY_DBG_MGT | */ \
/* RAY_DBG_TX | */ \
/* RAY_DBG_DCOM | */ \
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
/*
* XXX build options - move to LINT
*/
#define RAY_CM_RID 0 /* pccardd abuses windows 0 and 1 */
#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 */
#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 */
/*
* XXX build options - move to LINT
*/
#ifndef RAY_DEBUG
#define RAY_DEBUG 0x0000
#endif /* RAY_DEBUG */
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/limits.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <net/bpf.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
2003-06-28 17:57:30 +00:00
#include <net80211/ieee80211.h>
#include <net/if_llc.h>
2000-05-21 21:20:18 +00:00
#include <dev/pccard/pccardvar.h>
#include "card_if.h"
#include <dev/ray/if_rayreg.h>
#include <dev/ray/if_raymib.h>
#include <dev/ray/if_raydbg.h>
#include <dev/ray/if_rayvar.h>
/*
* Prototyping
*/
static int ray_attach (device_t);
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);
2000-05-21 21:20:18 +00:00
static int ray_ccs_tx (struct ray_softc *sc, size_t *ccsp, size_t *bufpp);
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);
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);
static void ray_com_runq_done (struct ray_softc *sc);
static int ray_detach (device_t);
static void ray_init (void *xsc);
2000-05-21 21:20:18 +00:00
static int ray_init_user (struct ray_softc *sc);
static void ray_init_assoc (struct ray_softc *sc, struct ray_comq_entry *com);
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);
static void ray_init_download (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_init_download_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
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);
static void ray_init_mcast (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_init_sj (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_init_sj_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
static void ray_intr (void *xsc);
static void ray_intr_ccs (struct ray_softc *sc, u_int8_t cmd, u_int8_t status, size_t ccs);
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);
static void ray_mcast_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
static int ray_mcast_user (struct ray_softc *sc);
static int ray_probe (device_t);
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);
static void ray_repparams_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
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);
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);
static void ray_rx (struct ray_softc *sc, size_t rcs);
2000-05-21 21:20:18 +00:00
static void ray_rx_ctl (struct ray_softc *sc, struct mbuf *m0);
static void ray_rx_data (struct ray_softc *sc, struct mbuf *m0, u_int8_t siglev, u_int8_t antenna);
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);
static void ray_rx_mgt_beacon (struct ray_softc *sc, struct mbuf *m0);
static void ray_rx_mgt_info (struct ray_softc *sc, struct mbuf *m0, struct ieee80211_information *elements);
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);
static void ray_tx (struct ifnet *ifp);
static void ray_tx_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
static void ray_tx_timo (void *xsc);
static int ray_tx_send (struct ray_softc *sc, size_t ccs, int pktlen, u_int8_t *dst);
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);
static void ray_upparams (struct ray_softc *sc, struct ray_comq_entry *com);
static void ray_upparams_done (struct ray_softc *sc, u_int8_t status, size_t ccs);
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
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
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 */
/*
* PC-Card (PCMCIA) driver definition
*/
static device_method_t ray_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, ray_probe),
DEVMETHOD(device_attach, ray_attach),
DEVMETHOD(device_detach, ray_detach),
{ 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);
/*
* Probe for the card by checking its startup results.
*
* Fixup any bugs/quirks for different firmware.
*/
static int
ray_probe(device_t dev)
{
struct ray_softc *sc = device_get_softc(dev);
struct ray_ecf_startup_v5 *ep = &sc->sc_ecf_startup;
int error;
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, "");
/*
* Read startup results from the card.
*/
error = ray_res_alloc_cm(sc);
if (error)
return (error);
error = ray_res_alloc_am(sc);
if (error) {
ray_res_release(sc);
return (error);
}
RAY_MAP_CM(sc);
SRAM_READ_REGION(sc, RAY_ECF_TO_HOST_BASE, ep,
sizeof(sc->sc_ecf_startup));
ray_res_release(sc);
/*
* Check the card is okay and work out what version we are using.
*/
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);
}
RAY_DPRINTF(sc, RAY_DBG_BOOTPARAM, "found a card");
sc->sc_gone = 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
/*
* 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
*/
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
return (0);
}
/*
* Attach the card into the kernel
*/
static int
ray_attach(device_t dev)
{
struct ray_softc *sc = device_get_softc(dev);
struct ray_ecf_startup_v5 *ep = &sc->sc_ecf_startup;
struct ifnet *ifp = &sc->arpcom.ac_if;
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;
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
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
if ((sc == NULL) || (sc->sc_gone))
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);
/*
* Grab the resources I need
*/
error = ray_res_alloc_cm(sc);
if (error)
return (error);
error = ray_res_alloc_am(sc);
if (error) {
ray_res_release(sc);
return (error);
}
error = ray_res_alloc_irq(sc);
if (error) {
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
/*
* 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
*
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_MAP_CM(sc);
bzero(&sc->sc_d, sizeof(struct ray_nw_param));
bzero(&sc->sc_c, sizeof(struct ray_nw_param));
/* Clear statistics counters */
sc->sc_rxoverflow = 0;
sc->sc_rxcksum = 0;
sc->sc_rxhcksum = 0;
sc->sc_rxnoise = 0;
/* Clear signal and antenna cache */
bzero(sc->sc_siglevs, sizeof(sc->sc_siglevs));
/* 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
*/
bcopy((char *)&ep->e_station_addr,
(char *)&sc->arpcom.ac_enaddr, ETHER_ADDR_LEN);
ifp->if_softc = sc;
ifp->if_name = "ray";
ifp->if_unit = device_get_unit(dev);
ifp->if_timer = 0;
ifp->if_flags = (IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
ifp->if_hdrlen = sizeof(struct ieee80211_frame) +
sizeof(struct ether_header);
ifp->if_baudrate = 1000000; /* Is this baud or bps ;-) */
ifp->if_output = ether_output;
ifp->if_start = ray_tx;
ifp->if_ioctl = ray_ioctl;
ifp->if_watchdog = ray_watchdog;
ifp->if_init = ray_init;
ifp->if_snd.ifq_maxlen = IFQ_MAXLEN;
ether_ifattach(ifp, ep->e_station_addr);
/*
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
*/
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
}
}
return (0);
}
/*
* 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.
*/
static int
ray_detach(device_t dev)
{
struct ray_softc *sc = device_get_softc(dev);
struct ifnet *ifp = &sc->arpcom.ac_if;
struct ray_comq_entry *com;
int s;
s = splimp();
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STOP, "");
if ((sc == NULL) || (sc->sc_gone))
return (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
/*
* Mark as not running and detach the interface.
*
* 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
*/
sc->sc_gone = 1;
sc->sc_c.np_havenet = 0;
2000-05-21 21:20:18 +00:00
ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
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
/*
* 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
*/
untimeout(ray_com_ecf_timo, sc, sc->com_timerh);
untimeout(ray_tx_timo, sc, sc->tx_timerh);
com = TAILQ_FIRST(&sc->sc_comq);
TAILQ_FOREACH(com, &sc->sc_comq, c_chain) {
com->c_flags |= RAY_COM_FDETACHED;
com->c_retval = 0;
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);
}
}
/*
* Release resources
*/
ray_res_release(sc);
RAY_DPRINTF(sc, RAY_DBG_STOP, "unloading complete");
splx(s);
return (0);
}
/*
* Network ioctl request.
*/
static int
ray_ioctl(register struct ifnet *ifp, u_long command, caddr_t data)
{
struct ray_softc *sc = ifp->if_softc;
struct ray_param_req pr;
struct ray_stats_req sr;
struct ifreq *ifr = (struct ifreq *)data;
int s, error, error2;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_IOCTL, "");
if ((sc == NULL) || (sc->sc_gone))
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;
s = splimp();
switch (command) {
case SIOCSIFFLAGS:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "SIFFLAGS 0x%0x", ifp->if_flags);
/*
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.
*/
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);
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);
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);
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);
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:
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "OTHER (pass to ether)");
error = ether_ioctl(ifp, command, data);
break;
}
splx(s);
return (error);
}
/*
* 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);
}
/*
* User land entry to network initialisation and changes in interface flags.
*
* 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.
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.
2000-05-28 23:23:24 +00:00
*
* Note, IFF_RUNNING is eventually set by init_sj_done or init_assoc_done
*/
2000-05-21 21:20:18 +00:00
static int
ray_init_user(struct ray_softc *sc)
{
struct ray_comq_entry *com[6];
int error, ncom;
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
/*
* 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
* 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
*
* 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
* commands in the runq that change the IFF_RUNNING state of
* 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;
com[ncom++] = RAY_COM_MALLOC(ray_init_download, RAY_COM_FCHKRUNNING);
com[ncom++] = RAY_COM_MALLOC(ray_init_mcast, RAY_COM_FCHKRUNNING);
com[ncom++] = RAY_COM_MALLOC(ray_init_sj, RAY_COM_FCHKRUNNING);
com[ncom++] = RAY_COM_MALLOC(ray_init_auth, RAY_COM_FCHKRUNNING);
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
*/
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
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! */
RAY_COM_FREE(com, ncom);
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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
/* 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.
*
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);
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;
sc->sc_d.np_promisc = !!(ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI));
/* XXX this is a hack whilst I transition the code. The instance
* XXX variables above should be set somewhere else. This is needed for
* XXX start_join */
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)
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
ray_init_download_v5(sc, 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
/*
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);
}
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)
/*
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
*/
static void
ray_init_download_v4(struct ray_softc *sc, struct ray_comq_entry *com)
{
struct ray_mib_4 ray_mib_4_default;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
#define MIB4(m) ray_mib_4_default.m
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);
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);
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);
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;
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);
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)
= RAY_MIB_INFRA_SUPER_SCAN_CYCLE_V4;
MIB4(mib_promisc) = com->c_desired.np_promisc;
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;
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)
= 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)
= 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;
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
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);
#define MIB5(m) ray_mib_5_default.m
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);
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);
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);
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;
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);
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)
= RAY_MIB_INFRA_SUPER_SCAN_CYCLE_V5;
MIB5(mib_promisc) = com->c_desired.np_promisc;
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;
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)
= 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)
= 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;
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;
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
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));
}
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
/*
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
*/
static void
ray_init_download_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
{
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);
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
ray_com_ecf_done(sc);
}
/*
* Runq entry to empty the multicast filter list
*/
static void
ray_init_mcast(struct ray_softc *sc, struct ray_comq_entry *com)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
/* If the card already running we might not need to reset the list */
RAY_COM_CHKRUNNING(sc, com, ifp);
/*
* 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 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
*/
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)
{
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 ifnet *ifp = &sc->arpcom.ac_if;
struct ray_net_params np;
int update;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_STARTJOIN, "");
RAY_MAP_CM(sc);
/* 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
/*
* 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
*/
sc->sc_c.np_havenet = 0;
if (sc->sc_d.np_net_type == RAY_MIB_NET_TYPE_ADHOC)
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_START_NET);
else
2000-05-21 21:20:18 +00:00
ray_ccs_fill(sc, com->c_ccs, RAY_CMD_JOIN_NET);
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++;
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)
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;
SRAM_WRITE_REGION(sc, RAY_HOST_TO_ECF_BASE, &np, sizeof(np));
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
*/
ray_com_ecf(sc, 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
* Complete start command or intermediate step in assoc command
*/
static void
ray_init_sj_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
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);
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
* Read back network parameters that the ECF sets
*/
SRAM_READ_REGION(sc, ccs, &sc->sc_c.p_1, sizeof(struct ray_cmd_net));
/* Adjust values for buggy firmware */
if (sc->sc_c.np_inited == 0x55)
sc->sc_c.np_inited = 0;
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 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.
*/
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");
SRAM_READ_REGION(sc, RAY_HOST_TO_ECF_BASE,
&sc->sc_c.p_2, sizeof(struct ray_net_params));
}
/*
* Hurrah! The network is now active.
*
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
* Clearing IFF_OACTIVE will ensure that the system will send us
* 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) {
sc->sc_c.np_havenet = 1;
sc->sc_c.np_framing = sc->sc_d.np_framing;
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_flags |= IFF_RUNNING;
ifp->if_flags &= ~IFF_OACTIVE;
}
ray_com_ecf_done(sc);
}
/*
* Runq entry to authenticate with an access point or another station
*/
static void
ray_init_auth(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 ifnet *ifp = &sc->arpcom.ac_if;
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);
/*
* Don't do anything if we are not in a managed network
*
* XXX V4 adhoc does not need this, V5 adhoc unknown
*/
if (sc->sc_c.np_net_type != RAY_MIB_NET_TYPE_INFRA) {
ray_com_runq_done(sc);
return;
}
/*
* 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
/*
* 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
*/
/* 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,
sc->arpcom.ac_enaddr,
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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
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
*/
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);
}
/*
* Complete association
*/
static void
ray_init_assoc_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
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);
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
/*
* Hurrah! The network is now active.
*
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
* Clearing IFF_OACTIVE will ensure that the system will send us
* packets. Just before we return from the interrupt context
* we check to see if packets have been queued.
*/
sc->sc_c.np_havenet = 1;
sc->sc_c.np_framing = sc->sc_d.np_framing;
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_flags |= IFF_RUNNING;
ifp->if_flags &= ~IFF_OACTIVE;
ray_com_ecf_done(sc);
}
/*
* 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.
*
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.
*/
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)
{
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];
int error, ncom;
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 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
*/
ncom = 0;
com[ncom++] = RAY_COM_MALLOC(ray_stop, 0);
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! */
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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
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
/*
* 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
*/
ifp->if_flags &= ~(IFF_RUNNING | IFF_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;
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);
}
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)
{
struct ray_softc *sc = ifp->if_softc;
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);
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 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");
}
/*
* Transmit packet handling
*/
/*
2000-05-21 21:20:18 +00:00
* Send a packet.
*
* 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
* 2) That the IFF_OACTIVE flag is checked before this code is called
* (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
* 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
* 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
*
* 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.
*
* 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.
*/
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)
{
struct ray_softc *sc = ifp->if_softc;
struct mbuf *m0, *m;
struct ether_header *eh;
struct llc *llc;
size_t ccs, bufp;
2000-05-21 21:20:18 +00:00
int pktlen, len;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
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
*/
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;
if (!(ifp->if_flags & IFF_RUNNING)) {
RAY_RECERR(sc, "cannot transmit - not running");
return;
}
if (!sc->sc_c.np_havenet) {
RAY_RECERR(sc, "cannot transmit - no network");
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
}
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 */
if ((sc->tx_timerh.callout == NULL) ||
(!callout_active(sc->tx_timerh.callout))) {
sc->tx_timerh =
timeout(ray_tx_timo, sc, RAY_TX_TIMEOUT);
return;
}
} 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);
/*
* 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.
*/
2000-05-21 21:20:18 +00:00
if (ray_ccs_tx(sc, &ccs, &bufp)) {
ifp->if_flags |= IFF_OACTIVE;
return;
}
/*
* Get the mbuf and process it - we have to remember to free the
* ccs if there are any errors.
*/
IF_DEQUEUE(&ifp->if_snd, m0);
if (m0 == NULL) {
RAY_CCS_FREE(sc, ccs);
return;
}
pktlen = m0->m_pkthdr.len;
if (pktlen > ETHER_MAX_LEN - ETHER_CRC_LEN) {
RAY_RECERR(sc, "mbuf too long %d", pktlen);
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
}
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 *);
2000-05-21 21:20:18 +00:00
/*
* Write the 802.11 header according to network type etc.
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,
IEEE80211_FC1_DIR_NODS,
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,
IEEE80211_FC1_DIR_TODS,
2000-05-21 21:20:18 +00:00
sc->sc_c.np_bss_id,
eh->ether_shost,
eh->ether_dhost);
else
bufp = ray_tx_wrhdr(sc, bufp,
IEEE80211_FC0_TYPE_DATA,
IEEE80211_FC1_DIR_FROMDS,
eh->ether_dhost,
sc->sc_c.np_bss_id,
eh->ether_shost);
2000-05-21 21:20:18 +00:00
/*
* Framing
*
* Add to the mbuf.
*/
switch (sc->sc_c.np_framing) {
case RAY_FRAMING_ENCAPSULATION:
2000-05-21 21:20:18 +00:00
/* Nice and easy - nothing! (just add an 802.11 header) */
break;
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;
default:
RAY_RECERR(sc, "unknown framing type %d", sc->sc_c.np_framing);
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
}
if (m0 == NULL) {
RAY_RECERR(sc, "could not frame packet");
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
return;
}
RAY_MBUF_DUMP(sc, RAY_DBG_TX, m0, "framed packet");
/*
* Copy the mbuf to the buffer in common memory
*
* We drop and don't bother wrapping as Ethernet packets are 1518
* 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.
*/
pktlen = sizeof(struct ieee80211_frame);
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)
SRAM_WRITE_REGION(sc, bufp, mtod(m, u_int8_t *), len);
else {
RAY_RECERR(sc, "tx buffer overflow");
RAY_CCS_FREE(sc, ccs);
ifp->if_oerrors++;
m_freem(m0);
return;
}
bufp += len;
}
/*
2000-05-21 21:20:18 +00:00
* Send it off
*/
2000-05-21 21:20:18 +00:00
if (ray_tx_send(sc, ccs, pktlen, eh->ether_dhost))
ifp->if_oerrors++;
2000-05-21 21:20:18 +00:00
else
ifp->if_opackets++;
m_freem(m0);
}
/*
* 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)
{
struct ray_softc *sc = (struct ray_softc *)xsc;
struct ifnet *ifp = &sc->arpcom.ac_if;
int 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
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
if (!(ifp->if_flags & IFF_OACTIVE) && (ifp->if_snd.ifq_head != NULL)) {
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);
splx(s);
}
}
/*
2000-05-21 21:20:18 +00:00
* Write an 802.11 header into the Tx buffer space and return the
* adjusted buffer pointer.
*/
static size_t
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)
{
struct ieee80211_frame header;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
RAY_MAP_CM(sc);
bzero(&header, sizeof(struct ieee80211_frame));
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);
2000-05-21 21:20:18 +00:00
SRAM_WRITE_REGION(sc, bufp, (u_int8_t *)&header,
sizeof(struct ieee80211_frame));
return (bufp + sizeof(struct ieee80211_frame));
2000-05-21 21:20:18 +00:00
}
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
ray_tx_send(struct ray_softc *sc, size_t ccs, int pktlen, u_int8_t *dst)
2000-05-21 21:20:18 +00:00
{
int i = 0;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
RAY_MAP_CM(sc);
while (!RAY_ECF_READY(sc)) {
DELAY(RAY_ECF_SPIN_DELAY);
if (++i > RAY_ECF_SPIN_TRIES) {
RAY_RECERR(sc, "ECF busy, dropping packet");
RAY_CCS_FREE(sc, ccs);
return (1);
}
}
if (i != 0)
RAY_RECERR(sc, "spun %d times", i);
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);
2000-05-21 21:20:18 +00:00
return (0);
}
/*
* 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)
{
struct ray_siglev *sl;
int i;
u_int8_t antenna;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
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:
/* This is a simple thresholding scheme that takes the mean
* 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));
}
/*
* Transmit now complete so clear ccs and network flags.
*/
static void
ray_tx_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_TX, "");
RAY_CCSERR(sc, status, if_oerrors);
RAY_CCS_FREE(sc, ccs);
ifp->if_timer = 0;
if (ifp->if_flags & IFF_OACTIVE)
ifp->if_flags &= ~IFF_OACTIVE;
}
/*
* Receiver packet handling
*/
/*
* Receive a packet from the card
*/
static void
ray_rx(struct ray_softc *sc, size_t rcs)
{
struct ieee80211_frame *header;
struct ifnet *ifp = &sc->arpcom.ac_if;
struct mbuf *m0;
size_t pktlen, fraglen, readlen, tmplen;
size_t bufp, ebufp;
u_int8_t siglev, antenna;
u_int first, ni, i;
u_int8_t *mp;
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 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);
m0 = NULL;
readlen = 0;
/*
* 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);
if ((pktlen > MCLBYTES) || (pktlen < sizeof(struct ieee80211_frame))) {
RAY_RECERR(sc, "packet too big or too small");
ifp->if_ierrors++;
goto skip_read;
}
MGETHDR(m0, M_DONTWAIT, MT_DATA);
if (m0 == NULL) {
RAY_RECERR(sc, "MGETHDR failed");
ifp->if_ierrors++;
goto skip_read;
}
if (pktlen > MHLEN) {
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)) {
RAY_RECERR(sc, "MCLGET failed");
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
}
m0->m_pkthdr.rcvif = ifp;
m0->m_pkthdr.len = pktlen;
m0->m_len = pktlen;
mp = mtod(m0, u_int8_t *);
/*
* 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) {
RAY_RECERR(sc, "bad length current 0x%x pktlen 0x%x",
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);
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
if ((i < RAY_RCS_FIRST) || (i > RAY_RCS_LAST)) {
RAY_RECERR(sc, "bad rcs index 0x%x", i);
ifp->if_ierrors++;
m_freem(m0);
m0 = NULL;
goto skip_read;
}
ebufp = bufp + fraglen;
if (ebufp <= RAY_RX_END)
SRAM_READ_REGION(sc, bufp, mp, fraglen);
else {
SRAM_READ_REGION(sc, bufp, mp,
(tmplen = RAY_RX_END - bufp));
SRAM_READ_REGION(sc, RAY_RX_BASE, mp + tmplen,
ebufp - RAY_RX_END);
}
mp += fraglen;
readlen += fraglen;
}
skip_read:
/*
* 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);
}
if (m0 == NULL)
return;
/*
* Check the 802.11 packet type and hand off to
* appropriate functions.
*/
header = mtod(m0, struct ieee80211_frame *);
2000-05-21 21:20:18 +00:00
if ((header->i_fc[0] & IEEE80211_FC0_VERSION_MASK)
!= IEEE80211_FC0_VERSION_0) {
RAY_RECERR(sc, "header not version 0 fc0 0x%x",
header->i_fc[0]);
ifp->if_ierrors++;
m_freem(m0);
return;
}
2000-05-21 21:20:18 +00:00
switch (header->i_fc[0] & IEEE80211_FC0_TYPE_MASK) {
case IEEE80211_FC0_TYPE_DATA:
ray_rx_data(sc, m0, siglev, antenna);
break;
case IEEE80211_FC0_TYPE_MGT:
2000-05-21 21:20:18 +00:00
ray_rx_mgt(sc, m0);
break;
case IEEE80211_FC0_TYPE_CTL:
2000-05-21 21:20:18 +00:00
ray_rx_ctl(sc, m0);
break;
default:
RAY_RECERR(sc, "unknown packet fc0 0x%x", header->i_fc[0]);
ifp->if_ierrors++;
m_freem(m0);
}
}
/*
* 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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
struct llc *llc;
u_int8_t *sa = NULL, *da = NULL, *ra = NULL, *ta = NULL;
int trim = 0;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_RX, "");
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) {
case IEEE80211_FC0_SUBTYPE_DATA:
case IEEE80211_FC0_SUBTYPE_CF_ACK:
case IEEE80211_FC0_SUBTYPE_CF_POLL:
case IEEE80211_FC0_SUBTYPE_CF_ACPL:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "DATA packet");
break;
case IEEE80211_FC0_SUBTYPE_NODATA:
case IEEE80211_FC0_SUBTYPE_CFACK:
case IEEE80211_FC0_SUBTYPE_CFPOLL:
case IEEE80211_FC0_SUBTYPE_CF_ACK_CF_ACK:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "NULL packet");
m_freem(m0);
return;
break;
default:
RAY_RECERR(sc, "reserved DATA packet subtype 0x%x",
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
m_freem(m0);
return;
}
/*
* 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.
*
* 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.
*/
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(1) packet before framing");
switch (header->i_fc[1] & IEEE80211_FC1_DIR_MASK) {
case IEEE80211_FC1_DIR_NODS:
da = ra = header->i_addr1;
sa = ta = header->i_addr2;
trim = sizeof(struct ieee80211_frame);
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D",
sa, ":", da, ":");
break;
case IEEE80211_FC1_DIR_FROMDS:
da = ra = header->i_addr1;
ta = header->i_addr2;
sa = header->i_addr3;
trim = sizeof(struct ieee80211_frame);
RAY_DPRINTF(sc, RAY_DBG_RX, "ap %6D from %6D to %6D",
ta, ":", sa, ":", da, ":");
break;
case IEEE80211_FC1_DIR_TODS:
ra = header->i_addr1;
sa = ta = header->i_addr2;
da = header->i_addr3;
trim = sizeof(struct ieee80211_frame);
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D ap %6D",
sa, ":", da, ":", ra, ":");
break;
case IEEE80211_FC1_DIR_DSTODS:
ra = header->i_addr1;
ta = header->i_addr2;
da = header->i_addr3;
sa = (u_int8_t *)header+1;
trim = sizeof(struct ieee80211_frame) + ETHER_ADDR_LEN;
RAY_DPRINTF(sc, RAY_DBG_RX, "from %6D to %6D ap %6D to %6D",
sa, ":", da, ":", ta, ":", ra, ":");
break;
}
/*
* Framing
*
* Each case must leave an Ethernet header and adjust trim.
*/
switch (sc->sc_c.np_framing) {
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) {
struct ether_header *eh;
/*
* 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");
}
break;
default:
RAY_RECERR(sc, "unknown framing type %d", sc->sc_c.np_framing);
ifp->if_ierrors++;
m_freem(m0);
return;
}
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(2) packet after framing");
/*
* Finally, do a bit of house keeping before sending the packet
* up the stack.
*/
m_adj(m0, trim);
RAY_MBUF_DUMP(sc, RAY_DBG_RX, m0, "(3) packet after trimming");
ifp->if_ipackets++;
ray_rx_update_cache(sc, header->i_addr2, siglev, antenna);
(*ifp->if_input)(ifp, m0);
}
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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
if ((header->i_fc[1] & IEEE80211_FC1_DIR_MASK) !=
IEEE80211_FC1_DIR_NODS) {
RAY_RECERR(sc, "MGT TODS/FROMDS wrong fc1 0x%x",
header->i_fc[1] & IEEE80211_FC1_DIR_MASK);
2000-05-21 21:20:18 +00:00
ifp->if_ierrors++;
m_freem(m0);
2000-05-21 21:20:18 +00:00
return;
}
/*
* Check the the mgt packet subtype, some packets should be
* dropped depending on the mode the station is in. See pg
* 52(60) of docs
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
*
* +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
2000-05-21 21:20:18 +00:00
*/
RAY_MBUF_DUMP(sc, RAY_DBG_MGT, m0, "MGT packet");
2000-05-21 21:20:18 +00:00
switch (header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK) {
case IEEE80211_FC0_SUBTYPE_BEACON:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "BEACON MGT packet");
ray_rx_mgt_beacon(sc, m0);
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_FC0_SUBTYPE_AUTH:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "AUTH MGT packet");
ray_rx_mgt_auth(sc, m0);
break;
case IEEE80211_FC0_SUBTYPE_DEAUTH:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "DEAUTH MGT packet");
/* XXX ray_rx_mgt_deauth(sc, m0); */
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_FC0_SUBTYPE_ASSOC_REQ:
case IEEE80211_FC0_SUBTYPE_REASSOC_REQ:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "(RE)ASSOC_REQ MGT packet");
if ((sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_AP))
RAY_RECERR(sc, "can't be an AP yet"); /* XXX_ACTING_AP */
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_FC0_SUBTYPE_ASSOC_RESP:
case IEEE80211_FC0_SUBTYPE_REASSOC_RESP:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "(RE)ASSOC_RESP MGT packet");
if ((sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_TERMINAL))
RAY_RECERR(sc, "can't be in INFRA yet"); /* XXX_INFRA */
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_FC0_SUBTYPE_DISASSOC:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_MGT, "DISASSOC MGT packet");
if (sc->sc_c.np_net_type == RAY_MIB_NET_TYPE_INFRA)
RAY_RECERR(sc, "can't be in INFRA yet"); /* XXX_INFRA */
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_FC0_SUBTYPE_PROBE_REQ:
case IEEE80211_FC0_SUBTYPE_PROBE_RESP:
case IEEE80211_FC0_SUBTYPE_ATIM:
RAY_RECERR(sc, "unexpected MGT packet subtype 0x%0x",
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
break;
default:
RAY_RECERR(sc, "reserved MGT packet subtype 0x%x",
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
}
m_freem(m0);
2000-05-21 21:20:18 +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?
2000-05-21 21:20:18 +00:00
*/
static void
ray_rx_mgt_beacon(struct ray_softc *sc, struct mbuf *m0)
2000-05-21 21:20:18 +00:00
{
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
ieee80211_mgt_beacon_t beacon = (u_int8_t *)(header+1);
struct ieee80211_information elements;
u_int64_t *timestamp;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
timestamp = (u_int64_t *)beacon;
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));
2000-05-21 21:20:18 +00:00
ray_rx_mgt_info(sc, m0, &elements);
2000-05-21 21:20:18 +00:00
}
static void
ray_rx_mgt_info(struct ray_softc *sc, struct mbuf *m0, struct ieee80211_information *elements)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
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;
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_MGT, "");
2000-05-21 21:20:18 +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);
2000-05-21 21:20:18 +00:00
switch (*bp) {
2000-05-21 21:20:18 +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]);
}
#endif /* (RAY_DEBUG & RAY_DBG_MGT) */
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;
2000-05-21 21:20:18 +00:00
default:
RAY_RECERR(sc, "reserved MGT element id 0x%x", *bp);
ifp->if_ierrors++;break;
2000-05-21 21:20:18 +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));
2000-05-21 21:20:18 +00:00
break;
case IEEE80211_AUTH_ALG_SHARED:
RAY_RECERR(sc,
"shared key authentication sequence number %d",
IEEE80211_AUTH_TRANSACTION(auth));
2000-05-21 21:20:18 +00:00
break;
default:
RAY_RECERR(sc,
"reserved authentication subtype 0x%04hx",
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)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
struct ieee80211_frame *header = mtod(m0, struct ieee80211_frame *);
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CTL, "");
if ((header->i_fc[1] & IEEE80211_FC1_DIR_MASK) !=
IEEE80211_FC1_DIR_NODS) {
RAY_RECERR(sc, "CTL TODS/FROMDS wrong fc1 0x%x",
header->i_fc[1] & IEEE80211_FC1_DIR_MASK);
2000-05-21 21:20:18 +00:00
ifp->if_ierrors++;
m_freem(m0);
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
* AP. See pg 52(60) of docs.
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) {
case IEEE80211_FC0_SUBTYPE_PS_POLL:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CTL, "PS_POLL CTL packet");
if ((sc->sc_d.np_net_type == RAY_MIB_NET_TYPE_INFRA) &&
(sc->sc_c.np_ap_status == RAY_MIB_AP_STATUS_AP))
RAY_RECERR(sc, "can't be an AP yet"); /* XXX_ACTING_AP */
2000-05-21 21:20:18 +00:00
break;
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:
RAY_RECERR(sc, "unexpected CTL packet subtype 0x%0x",
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
break;
default:
RAY_RECERR(sc, "reserved CTL packet subtype 0x%x",
2000-05-21 21:20:18 +00:00
header->i_fc[0] & IEEE80211_FC0_SUBTYPE_MASK);
ifp->if_ierrors++;
}
m_freem(m0);
2000-05-21 21:20:18 +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;
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 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 */
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 */
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;
}
}
/*
* Interrupt handling
*/
/*
* Process an interrupt
*/
static void
ray_intr(void *xsc)
{
struct ray_softc *sc = (struct ray_softc *)xsc;
struct ifnet *ifp = &sc->arpcom.ac_if;
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;
u_int8_t cmd, status;
int ccsi;
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);
if ((sc == NULL) || (sc->sc_gone))
return;
/*
* Check that the interrupt was for us, if so get the rcs/ccs
* and vector on the command contained within it.
*/
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);
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)
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);
else
RAY_RECERR(sc, "bad ccs index 0x%x", ccsi);
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
/* Send any packets lying around and update error counters */
if (!(ifp->if_flags & IFF_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);
}
/*
* Read the error counters.
*/
static void
ray_intr_updt_errcntrs(struct ray_softc *sc)
{
size_t csc;
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 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.
*/
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);
}
/*
* Process CCS command completion
*/
static void
ray_intr_ccs(struct ray_softc *sc, u_int8_t cmd, u_int8_t status, size_t ccs)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
switch (cmd) {
case RAY_CMD_DOWNLOAD_PARAMS:
RAY_DPRINTF(sc, RAY_DBG_COM, "START_PARAMS");
ray_init_download_done(sc, status, ccs);
break;
case RAY_CMD_UPDATE_PARAMS:
RAY_DPRINTF(sc, RAY_DBG_COM, "UPDATE_PARAMS");
ray_upparams_done(sc, status, ccs);
break;
case RAY_CMD_REPORT_PARAMS:
RAY_DPRINTF(sc, RAY_DBG_COM, "REPORT_PARAMS");
ray_repparams_done(sc, status, ccs);
break;
case RAY_CMD_UPDATE_MCAST:
RAY_DPRINTF(sc, RAY_DBG_COM, "UPDATE_MCAST");
ray_mcast_done(sc, status, ccs);
break;
case RAY_CMD_START_NET:
case RAY_CMD_JOIN_NET:
RAY_DPRINTF(sc, RAY_DBG_COM, "START|JOIN_NET");
ray_init_sj_done(sc, status, ccs);
break;
case RAY_CMD_TX_REQ:
RAY_DPRINTF(sc, RAY_DBG_COM, "TX_REQ");
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;
case RAY_CMD_START_ASSOC:
RAY_DPRINTF(sc, RAY_DBG_COM, "START_ASSOC");
ray_init_assoc_done(sc, status, ccs);
break;
case RAY_CMD_UPDATE_APM:
RAY_RECERR(sc, "unexpected UPDATE_APM");
break;
case RAY_CMD_TEST_MEM:
RAY_RECERR(sc, "unexpected TEST_MEM");
break;
case RAY_CMD_SHUTDOWN:
RAY_RECERR(sc, "unexpected SHUTDOWN");
break;
case RAY_CMD_DUMP_MEM:
RAY_RECERR(sc, "unexpected DUMP_MEM");
break;
case RAY_CMD_START_TIMER:
RAY_RECERR(sc, "unexpected START_TIMER");
break;
default:
RAY_RECERR(sc, "unknown command 0x%x", cmd);
break;
}
}
/*
* Process ECF command request
*/
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)
{
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, "");
switch (cmd) {
case RAY_ECMD_RX_DONE:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "RX_DONE");
ray_rx(sc, rcs);
break;
case RAY_ECMD_REJOIN_DONE:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "REJOIN_DONE");
sc->sc_c.np_havenet = 1;
break;
case RAY_ECMD_ROAM_START:
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_RX, "ROAM_START");
sc->sc_c.np_havenet = 0;
break;
case RAY_ECMD_JAPAN_CALL_SIGNAL:
RAY_RECERR(sc, "unexpected JAPAN_CALL_SIGNAL");
break;
default:
RAY_RECERR(sc, "unknown command 0x%x", cmd);
break;
}
RAY_CCS_FREE(sc, rcs);
}
/*
* User land entry to multicast list changes
*/
static int
ray_mcast_user(struct ray_softc *sc)
{
struct ray_comq_entry *com[2];
int error, ncom;
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
/*
* Do all checking in the runq to preserve ordering.
*
* We run promisc to pick up changes to the ALL_MULTI
* interface flag.
*/
ncom = 0;
com[ncom++] = RAY_COM_MALLOC(ray_mcast, 0);
com[ncom++] = RAY_COM_MALLOC(ray_promisc, 0);
RAY_COM_RUNQ(sc, com, ncom, "raymcast", error);
/* XXX no real error processing from anything yet! */
RAY_COM_FREE(com, ncom);
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
* Runq entry to setting the multicast filter list
*
* MUST always be followed by a call to ray_promisc to pick up changes
* to promisc flag
*/
static void
ray_mcast(struct ray_softc *sc, struct ray_comq_entry *com)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
struct ifmultiaddr *ifma;
size_t bufp;
int count = 0;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
RAY_MAP_CM(sc);
/*
* If card is not running we don't need to update this.
*/
if (!(ifp->if_flags & IFF_RUNNING)) {
RAY_DPRINTF(sc, RAY_DBG_IOCTL, "not running");
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.
*/
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
count++;
if (count == 0) {
ray_com_runq_done(sc);
return;
} else if (count > 16) {
ifp->if_flags |= IFF_ALLMULTI;
ray_com_runq_done(sc);
return;
} else if (ifp->if_flags & IFF_ALLMULTI)
ifp->if_flags &= ~IFF_ALLMULTI;
/*
* Kick the card
*/
2000-05-21 21:20:18 +00:00
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, count);
bufp = RAY_HOST_TO_ECF_BASE;
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
SRAM_WRITE_REGION(
sc,
bufp,
LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
ETHER_ADDR_LEN
);
bufp += ETHER_ADDR_LEN;
}
ray_com_ecf(sc, com);
}
/*
* Complete the multicast filter list update
*/
static void
ray_mcast_done(struct ray_softc *sc, u_int8_t status, size_t ccs)
{
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);
RAY_CCSERR(sc, status, if_oerrors); /* XXX error counter */
ray_com_ecf_done(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
* Runq entry to set/reset promiscuous mode
*/
static void
ray_promisc(struct ray_softc *sc, struct ray_comq_entry *com)
{
struct ifnet *ifp = &sc->arpcom.ac_if;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
RAY_MAP_CM(sc);
/*
* If card not running or we already have the right flags
* we don't need to update this
*/
sc->sc_d.np_promisc = !!(ifp->if_flags & (IFF_PROMISC | IFF_ALLMULTI));
if (!(ifp->if_flags & IFF_RUNNING) ||
(sc->sc_c.np_promisc == sc->sc_d.np_promisc)) {
ray_com_runq_done(sc);
return;
}
/*
* Kick the card
*/
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,
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);
SRAM_WRITE_1(sc, RAY_HOST_TO_ECF_BASE, sc->sc_d.np_promisc);
ray_com_ecf(sc, 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
* User land entry to parameter reporting
*
* As we by pass the runq to report current parameters this function
* only provides a snap shot of the driver's state.
*/
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)
{
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];
int error, ncom;
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 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);
if (pr->r_paramid > RAY_MIB_LASTUSER) {
switch (pr->r_paramid) {
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;
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;
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;
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;
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;
default:
return (EINVAL);
break;
}
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];
return (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
pr->r_failcause = 0;
ncom = 0;
com[ncom++] = RAY_COM_MALLOC(ray_repparams, RAY_COM_FWOK);
com[ncom-1]->c_pr = pr;
RAY_COM_RUNQ(sc, com, ncom, "rayrparm", error);
2000-05-21 21:20:18 +00:00
/* XXX no real error processing from anything yet! */
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;
RAY_COM_FREE(com, ncom);
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);
}
/*
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
*
* The card and driver are happy for parameters to be read
* whenever the card is plugged in
*/
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)
{
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);
/*
* Kick the card
*/
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
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);
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);
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);
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);
}
/*
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
*/
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)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
sr->rxoverflow = sc->sc_rxoverflow;
sr->rxcksum = sc->sc_rxcksum;
sr->rxhcksum = sc->sc_rxhcksum;
sr->rxnoise = sc->sc_rxnoise;
return (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
* User land entry to parameter update changes
*
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
* invalid we have to re-start/join.
*/
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)
{
struct ray_comq_entry *com[4];
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
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, "");
/*
* Check that the parameter is available based on firmware version
*/
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);
/*
* Handle certain parameters specially
*/
todo = 0;
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 */
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 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 */
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;
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;
break;
case RAY_MIB_AP_STATUS: /* Unsupported */
case RAY_MIB_MAC_ADDR: /* XXX Need interface up but could be done */
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;
break;
}
/*
* 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;
}
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);
com[ncom++] = RAY_COM_MALLOC(ray_init_auth, 0);
com[ncom++] = RAY_COM_MALLOC(ray_init_assoc, 0);
}
RAY_COM_RUNQ(sc, com, ncom, "rayuparam", error);
2000-05-21 21:20:18 +00:00
/* XXX no real error processing from anything yet! */
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;
RAY_COM_FREE(com, ncom);
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);
}
/*
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
*
* 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.
*/
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)
{
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);
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,
ray_cmd_update, c_paramid, com->c_pr->r_paramid);
SRAM_WRITE_FIELD_1(sc, com->c_ccs, ray_cmd_update, c_nparam, 1);
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);
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);
}
/*
* Complete the parameter update, note that promisc finishes up here too
*/
static void
ray_upparams_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);
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);
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)) {
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:
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,
"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 =
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;
}
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);
}
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 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;
com->c_ccs = NULL;
com->c_wakeup = NULL;
com->c_pr = NULL;
com->c_mesg = 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
return (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
/*
* 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 *,
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));
}
/*
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.
*
* 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
* detached we get out quickly to prevent panics and don't bother
* about the runq.
*/
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
{
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
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];
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
}
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
2000-05-21 21:20:18 +00:00
/*
* Allocate ccs's for each command.
2000-05-21 21:20:18 +00:00
*/
for (i = 0; i < ncom; i++) {
error = ray_ccs_alloc(sc, &com[i]->c_ccs, wmesg);
if (error == ENXIO)
return (ENXIO);
else if (error)
goto cleanup;
2000-05-21 21:20:18 +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!
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");
error = tsleep(com[ncom-1], PCATCH | PRIBIO, wmesg, 0);
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;
2000-05-21 21:20:18 +00:00
cleanup:
/*
* Only clean the queue on real errors - we don't care about it
* when we detach as the queue entries are freed by the callers.
*/
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);
com[i]->c_ccs = NULL;
}
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
}
/*
* 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);
2000-05-21 21:20:18 +00:00
if ((com == NULL) ||
(com->c_flags & RAY_COM_FRUNNING) ||
(com->c_flags & RAY_COM_FWAIT) ||
(com->c_flags & RAY_COM_FDETACHED))
return;
com->c_flags |= RAY_COM_FRUNNING;
RAY_DPRINTF(sc, RAY_DBG_COM, "running %p", com);
RAY_DCOM(sc, RAY_DBG_DCOM, com, "running");
com->c_function(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
* Remove run command, free ccs and wakeup caller.
*
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.
*
* 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.
*/
static void
ray_com_runq_done(struct ray_softc *sc)
{
struct ray_comq_entry *com;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
com = TAILQ_FIRST(&sc->sc_comq); /* XXX shall we check this as below */
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);
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);
com->c_ccs = NULL;
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);
/* 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
*
* XXX all the runq_done calls from IFF_RUNNING checks in runq
* XXX routines should return EIO but shouldn't abort the runq
*/
}
/*
* Send a command to the ECF.
*/
static void
ray_com_ecf(struct ray_softc *sc, struct ray_comq_entry *com)
{
int i = 0;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
RAY_MAP_CM(sc);
while (!RAY_ECF_READY(sc)) {
DELAY(RAY_ECF_SPIN_DELAY);
if (++i > RAY_ECF_SPIN_TRIES)
RAY_PANIC(sc, "spun too long");
}
if (i != 0)
RAY_RECERR(sc, "spun %d times", i);
RAY_DPRINTF(sc, RAY_DBG_COM, "sending %p", com);
RAY_DCOM(sc, RAY_DBG_DCOM, com, "sending");
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))) {
RAY_DPRINTF(sc, RAY_DBG_COM, "adding timeout");
sc->com_timerh = timeout(ray_com_ecf_timo, sc, RAY_COM_TIMEOUT);
}
}
/*
* Deal with commands that require a timeout to test completion.
*
* 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!
*/
static void
ray_com_ecf_timo(void *xsc)
{
struct ray_softc *sc = (struct ray_softc *)xsc;
struct ray_comq_entry *com;
u_int8_t cmd, status;
int s;
s = splnet();
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
RAY_MAP_CM(sc);
com = TAILQ_FIRST(&sc->sc_comq);
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);
status = SRAM_READ_FIELD_1(sc, com->c_ccs, ray_cmd, c_status);
switch (status) {
case RAY_CCS_STATUS_COMPLETE:
case RAY_CCS_STATUS_FREE: /* Buggy firmware */
ray_intr_ccs(sc, cmd, status, com->c_ccs);
break;
case RAY_CCS_STATUS_BUSY:
sc->com_timerh = timeout(ray_com_ecf_timo, sc, RAY_COM_TIMEOUT);
break;
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
ray_intr_ccs(sc, cmd, status, com->c_ccs);
break;
}
splx(s);
}
/*
* 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.
*/
static void
ray_com_ecf_done(struct ray_softc *sc)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "");
untimeout(ray_com_ecf_timo, sc, sc->com_timerh);
ray_com_runq_done(sc);
}
#if RAY_DEBUG & RAY_DBG_COM
/*
* Process completed ECF commands that probably came from the command queue
*
* This routine is called after vectoring the completed ECF command
* to the appropriate _done routine. It helps check everything is okay.
*/
static void
ray_com_ecf_check(struct ray_softc *sc, size_t ccs, char *mesg)
{
struct ray_comq_entry *com;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_COM, "%s", mesg);
com = TAILQ_FIRST(&sc->sc_comq);
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 */
/*
2000-05-21 21:20:18 +00:00
* CCS allocators
*/
/*
2000-05-21 21:20:18 +00:00
* Obtain a ccs for a commmand
*
2000-05-21 21:20:18 +00:00
* Returns 0 and in `ccsp' the bus offset of the free ccs. Will block
* awaiting free ccs if needed - if the sleep is interrupted
* EINTR/ERESTART is returned, if the card is ejected we return ENXIO.
*/
static int
2000-05-21 21:20:18 +00:00
ray_ccs_alloc(struct ray_softc *sc, size_t *ccsp, char *wmesg)
{
size_t ccs;
u_int i;
2000-05-21 21:20:18 +00:00
int error;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(sc);
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;
}
if (i > RAY_CCS_CMD_LAST) {
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "sleeping");
error = tsleep(ray_ccs_alloc, PCATCH | PRIBIO,
wmesg, 0);
if ((sc == NULL) || (sc->sc_gone))
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);
2000-05-21 21:20:18 +00:00
if (error)
return (error);
} else
break;
}
2000-05-21 21:20:18 +00:00
RAY_DPRINTF(sc, RAY_DBG_CCS, "allocated 0x%02x", i);
sc->sc_ccsinuse[i] = 1;
ccs = RAY_CCS_ADDRESS(i);
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);
if (ccs == NULL)
RAY_PANIC(sc, "ccs not allocated");
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);
}
/*
* 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).
*/
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)
{
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CCS, "");
RAY_MAP_CM(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 1 | (RAY_DEBUG & RAY_DBG_CCS)
if (!sc->sc_ccsinuse[RAY_CCS_INDEX(ccs)])
RAY_RECERR(sc, "freeing free ccs 0x%02x", 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 */
if (!sc->sc_gone)
RAY_CCS_FREE(sc, ccs);
sc->sc_ccsinuse[RAY_CCS_INDEX(ccs)] = 0;
RAY_DPRINTF(sc, RAY_DBG_CCS, "freed 0x%02x", RAY_CCS_INDEX(ccs));
2000-05-21 21:20:18 +00:00
wakeup(ray_ccs_alloc);
}
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);
SRAM_WRITE_FIELD_1(sc, ccs, ray_cmd_tx, c_apm_mode, 0);
2000-05-21 21:20:18 +00:00
bufp += sizeof(struct ray_tx_phy_header);
*ccsp = ccs;
*bufpp = bufp;
return (0);
}
/*
* Routines to obtain resources for the card
*/
/*
* Allocate the attribute memory on the card
2000-07-11 21:31:59 +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.
*/
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;
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CM, "");
sc->am_rid = RAY_AM_RID;
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);
2000-07-11 21:31:59 +00:00
if (error) {
RAY_PRINTF(sc, "CARD_SET_MEMORY_OFFSET returned 0x%0x", error);
2000-07-11 21:31:59 +00:00
return (error);
}
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->am_rid, PCCARD_A_MEM_ATTR);
if (error) {
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
return (error);
}
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);
2000-07-11 21:31:59 +00:00
}
sc->am_bsh = rman_get_bushandle(sc->am_res);
sc->am_bst = rman_get_bustag(sc->am_res);
2000-07-11 21:31:59 +00:00
#if RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM)
{
u_long flags;
u_int32_t offset;
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,
sc->am_rid, &offset);
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);
}
#endif /* RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM) */
return (0);
}
/*
* Allocate the common memory on the card
*
* 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.
*/
static int
ray_res_alloc_cm(struct ray_softc *sc)
{
u_long start, count, end;
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
RAY_DPRINTF(sc, RAY_DBG_SUBR | RAY_DBG_CM, "");
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));
sc->cm_rid = RAY_CM_RID;
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);
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);
2000-07-11 21:31:59 +00:00
if (error) {
RAY_PRINTF(sc, "CARD_SET_MEMORY_OFFSET returned 0x%0x", error);
2000-07-11 21:31:59 +00:00
return (error);
}
error = CARD_SET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
SYS_RES_MEMORY, sc->cm_rid, PCCARD_A_MEM_COM);
2000-07-11 21:31:59 +00:00
if (error) {
RAY_PRINTF(sc, "CARD_SET_RES_FLAGS returned 0x%0x", error);
2000-07-11 21:31:59 +00:00
return (error);
}
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);
}
sc->cm_bsh = rman_get_bushandle(sc->cm_res);
sc->cm_bst = rman_get_bustag(sc->cm_res);
#if RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM)
{
u_long flags;
u_int32_t offset;
CARD_GET_RES_FLAGS(device_get_parent(sc->dev), sc->dev,
2000-07-11 21:31:59 +00:00
SYS_RES_MEMORY, sc->cm_rid, &flags);
CARD_GET_MEMORY_OFFSET(device_get_parent(sc->dev), sc->dev,
2000-07-11 21:31:59 +00:00
sc->cm_rid, &offset);
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);
}
#endif /* RAY_DEBUG & (RAY_DBG_CM | RAY_DBG_BOOTPARAM) */
return (0);
}
/*
* Get an irq and attach it to the bus
*/
static int
ray_res_alloc_irq(struct ray_softc *sc)
{
int error;
RAY_DPRINTF(sc, RAY_DBG_SUBR, "");
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));
sc->irq_rid = 0;
sc->irq_res = bus_alloc_resource(sc->dev, SYS_RES_IRQ, &sc->irq_rid,
0, ~0, 1, RF_ACTIVE);
if (!sc->irq_res) {
RAY_PRINTF(sc, "Cannot allocate irq");
return (ENOMEM);
}
if ((error = bus_setup_intr(sc->dev, sc->irq_res, INTR_TYPE_NET,
ray_intr, sc, &sc->irq_handle)) != 0) {
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",
bus_get_resource_start(sc->dev, SYS_RES_IRQ, sc->irq_rid),
bus_get_resource_count(sc->dev, SYS_RES_IRQ, sc->irq_rid));
return (0);
}
/*
* Release all of the card's resources
*/
static void
ray_res_release(struct ray_softc *sc)
{
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;
}
}
/*
* mbuf dump
*/
#if RAY_DEBUG & RAY_DBG_MBUF
static void
ray_dump_mbuf(struct ray_softc *sc, struct mbuf *m, char *s)
{
u_int8_t *d, *ed;
u_int i;
char p[17];
RAY_PRINTF(sc, "%s", s);
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);
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)
2000-05-21 21:20:18 +00:00
printf(" %s\n", p);
}
#endif /* RAY_DEBUG & RAY_DBG_MBUF */