Commit Graph

14484 Commits

Author SHA1 Message Date
Sam Leffler
0bbf544158 use a private task queue thread
MFC after:	2 weeks
2006-02-09 21:48:51 +00:00
Sam Leffler
7a04dc27c1 add adhoc demo mode support
MFC after:	2 weeks
2006-02-09 21:42:53 +00:00
Sam Leffler
17f3f177e4 make regdomain sysctl r/w in case it's possible to do this in the future
MFC after:	2 weeks
2006-02-09 21:36:00 +00:00
Sam Leffler
6c4612b94d cleanup rate setup
MFC after:	2 weeks
2006-02-09 21:31:48 +00:00
Sam Leffler
86e07743eb add tx99 hooks
MFC after:	2 weeks
2006-02-09 21:28:11 +00:00
Sam Leffler
ffa2cab632 move hal statistics to softc; the per-node stats are overkill, they're
only used when operating in station mode

MFC after:	2 weeks
2006-02-09 21:23:44 +00:00
Sam Leffler
913a1ba103 lookup the protection tx rate index in the rate tables
instead of using a known value

MFC after:	2 weeks
2006-02-09 21:17:28 +00:00
Sam Leffler
8b5341de05 honor net80211 mcast tx rate
MFC after:	2 weeks
2006-02-09 21:15:36 +00:00
Sam Leffler
6f890873fb craft unique names for tx q + buffer mtx's to help with interpreting ktr data
MFC after:	2 weeks
2006-02-09 21:09:26 +00:00
Sam Leffler
e2d787fa61 allow the size of tx+rx buffer pools to be tuned
MFC after:	2 weeks
2006-02-09 21:03:25 +00:00
Sam Leffler
be61348037 lower try count on mgt (and ctl) frames to avoid clogging the tx queue
and loading the bss when operating in ap mode under load; adjust recognition
of multi-rate retry to match

MFC after:	2 weeks
2006-02-09 20:57:48 +00:00
Søren Schmidt
466be09c2a Unbreak Promise SATAII/150 controllers caused by the DMA dump changes. 2006-02-09 20:54:42 +00:00
Søren Schmidt
0a3a1935b3 Unbreak dump on legacy systems. Needs a more clean solution but that is
more intrusive and can wait 'til after 6.1
2006-02-09 20:53:32 +00:00
Sam Leffler
55f63772e8 move mgt frame tx rate responsibility from the rate control modules
to the driver; this avoids redundant logic and will be necessary
for future additions

MFC after:	2 weeks
2006-02-09 20:49:55 +00:00
Sam Leffler
b91bf513f3 sync with latest code in madwifi
Obtained from:	madwifi
MFC after:	2 weeks
2006-02-09 20:40:28 +00:00
Gleb Smirnoff
4f4a363562 Fix two important typos in watchdog handling:
- Restart watchdog if we *did* processed any descriptors. [1]
- Log the watchdog event if the link is *up*. [2]

PR:		kern/92948 [1]
Submitted by:	Mihail Balikov <mihail.balikov interbgc.com> [1]
PR:		kern/92895 [2]
Submitted by:	Vladimir Ivanov <wawa yandex-team.ru> [2]
2006-02-09 12:57:17 +00:00
Matt Jacob
c68a0f14fd Garbage collect a file that no longe is used (replaced by mpi_log_fc.h). 2006-02-08 05:58:25 +00:00
Warner Losh
22293c3afb Detach the children before we delete them. This is a little cleaner
than just deleting them.  Also add comments about why we do this.
Given the current behavior of delete_child, I don't think this changes
anything.  It just feels cleaner.
2006-02-07 18:38:51 +00:00
Gleb Smirnoff
d640400bc5 Since em(4) taskqueue is a new network context, we need to conditionally
lock Giant here.

Submitted by:	Andrey V. Elsukov <bu7cher yandex.ru>
2006-02-07 13:11:13 +00:00
Søren Schmidt
52c7e322b2 Unbreak DMA dumo on Intel 31224. 2006-02-06 19:17:48 +00:00
Alexander Leidinger
9d978cc757 Convert NULL checks into KASSERT (and move them before the first
dereferencing) since a NULL value would be a bug here.

Note: Both affected functions look very similar. A refactoring may
be beneficial.

CID:		483, 485
Found with:	Coverity Prevent(tm)
Discussed with:	ariff
MFC after:	5 days
2006-02-05 17:47:26 +00:00
Alexander Leidinger
cd9de7ee61 A pointer was checked for NULL after dereferencing it. The check is not
needed here, except there's a bug which results in detaching the device
twice.

Move the NULL pointer check to the beginning of the function and convert
it into a KASSERT.

CID:		420
Found with:	Coverity Prevent(tm)
Discussed with:	ariff
MFC after:	5 days
2006-02-05 17:33:18 +00:00
Alexander Leidinger
f31eef8b22 Fix memory leak in some failure cases.
CID:		420
Found with:	Coverity Prevent(tm)
2006-02-05 17:10:52 +00:00
Scott Long
5d9299772e Squash a couple more invalid long casts. 2006-02-05 05:05:27 +00:00
Marius Strobl
ea227b498b - Add support for using LOM (Lights Out Management) and RSC (Remote System
Control) devices as console. These are microcontrollers which are either
  on-board or part of an add-on card and provide terminal server, remote
  power switch and monitoring functionality. For console usage these are
  connected to the rest of the system via a SCC or an UART. This commit adds
  support for the following variants (corresponds to what 'input-device' and
  'output-device' have to be set to):
  rsc		found on-board in E250 and supposedly some Netra, connected
		via a SAB82532, com. parameters can be determined via OFW
  rsc-console	RSC card found in E280R, Fire V4x0, Fire V8x0, connected
		via a NS16550, hardwired to 115200 8N1
  lom-console	LOMlite2 card found in Netra 20/T4, connected via a NS16550,
		hardwired to 9600 8N1

- Add my copyright to uart_cpu_sparc64.c as I've rewritten about one third
  of that file over time.

Tested on:	E250, E280R
Thanks to:	dwhite@ for providing access to an E280R
OK'ed by:	marcel
MFC after:	1 week
2006-02-04 23:27:16 +00:00
Warner Losh
ccc0f2f0af Silence the strict-alias warnings. Make a trip through (void *) when
casting a structure to a uint32_t *.  Many drivers in the tree do this, but
I'll not update them until these changes can be reviewed by the pedantic
standards folks.
2006-02-04 22:51:03 +00:00
Warner Losh
53920ff8d9 Type of overrun_buf doesn't matter to this code, but does to gcc.
Make it a void *.
2006-02-04 22:33:08 +00:00
Warner Losh
2805b31bdf Use void * for pointer rather than u_int8_t *, since it doesn't
matter.  Well it does for type punning warnings.
2006-02-04 22:14:20 +00:00
Warner Losh
44dc92eb94 Now that TUNABLE_ULONG is in the tree, and has been for a while, use it.
This also avoids typepunning.
2006-02-04 21:56:14 +00:00
Warner Losh
ed3eb6ebe2 Don't type pun accidentally. Instead, be explicit that we're type
punning with an union so that the compiler knows.
2006-02-04 21:37:39 +00:00
Bill Paul
55ad415824 When ndis_attach() runs, it has to very briefly initialize the card
in order to query the underlying Windows driver for the station address
and some other properties. There is a slim chance that the card may
receive a packet and indicate it up to us before ndis_attach() can call
ndis_halt_nic(). This is bad, because both the softc structure and
the ifnet structure aren't fully initialized yet: many pointers are
still NULL, so if we make it into ndis_rxeof(), we will panic.

To fix this, we need to do the following:

- Move the calls to IoAllocateWorkItem() to before the call to ndis_init_nic().
- Move the initialization of the RX DPC and status callback function pointers
  to before ndis_init_nic() as well.
- Modify ndis_rxeof() to check if the IFF_DRV_RUNNING flag is set. If it
  isn't, we return any supplied NDIS_PACKETs to the NIC without processing
  them.

This fixes a crash than can occur when activating a wireless NIC in
close proximity to a very busy wireless network, reported by Ryan
Beasley (ryan%^$!ATgoddamnbastard-****!!!DOTorg.

MFC after:	3 days
2006-02-04 19:42:49 +00:00
Scott Long
1107d45356 Squash another use of vtophys. Instead of creating separate busdma objects
for doing static memory transfers, start collecting them into a single
object.
2006-02-04 17:56:17 +00:00
Scott Long
73450b30c8 Now that the em driver no longer needs to directly touch the scheduler, remove some
unneeded headers.
2006-02-04 16:50:14 +00:00
Scott Long
b0cb1aaf7c Add the start of busdma infrastructure to this driver. Convert the
ASR_resetIOPR function to use it.
2006-02-04 08:45:19 +00:00
Matt Jacob
51effc8cd5 Actually, no, I had it wrong in 1.109. The arguments to bus_dma_create_tag
are bus_addr_t, not bus_size_t.

In any case, turn off DAC support entirely until it is revamped to actually
work *correctly* for 64 bit platforms (not using a PAE definition and for
both initiator and target mode).
2006-02-04 08:39:02 +00:00
Scott Long
b32f2b9a08 Now that the U32 type is a really 32-bits wide, eliminate a bunch of other
bad assumptions and long values.
2006-02-04 08:20:23 +00:00
Warner Losh
d68da9fe4f Fix mismerge after last cvs update for the IFQ_DRV_DEQUEUE changes. 2006-02-04 08:19:00 +00:00
Warner Losh
e0c1d667ec Remove ifdef notyet SIOCGHWADDR vestige. 2006-02-04 08:16:41 +00:00
Warner Losh
70346a6275 Remove ifdef notdef SIOCHWADDR vestige. 2006-02-04 08:16:07 +00:00
Warner Losh
28383e51e2 Remove ifdef notyet for SIOCGHWADDR
Treat SIOCADDMULTI and SIOCDELMULTI the same, since they had the same code
Remove redundant assignment to error
Convert to using the altq interface completely.
2006-02-04 08:15:29 +00:00
Scott Long
90acf956d3 Give correct definitions to some basic types. 2006-02-04 08:01:49 +00:00
Scott Long
04ab3022a9 Fix a possible memory leak in asr_attach. 2006-02-04 06:08:19 +00:00
Scott Long
4b2dc3c447 i386/PAE defines bus_size_t to be 32-bits when it likely should be 64-bits.
Fixing it is left for another day, so just hack around it for now.
2006-02-04 03:41:48 +00:00
Ruslan Ermilov
bbe9814a5b Fix compilation with -Wundef (NBPF is undefined on FreeBSD >4). 2006-02-03 20:55:30 +00:00
Marius Strobl
8d727356c6 - Don't shift the clock frequency in MHz left by 8 before assigning it
to sbus_mdvec.dv_clock as sbus_mdvec.dv_clock is meant to be specified
  in MHz. While this was a bug it shouldn't have affected FreeBSD/sparc64
  as sbus_mdvec.dv_clock is used to limit the clock rate of chips when
  a machine isn't able to support them at maximum speed which isn't the
  case for sun4u machines.
- Remove the code that checks whether the clock frequency returned by
  sbus_get_clockfreq() is 0 and falls back to 25MHz if it is as that's
  already done in sbus(4).

Approved by:	mjacob
MFC after:	3 days
2006-02-03 12:35:42 +00:00
Warner Losh
6117d67f88 If the device has a PNPID, don't bother to attach. There were some
instances where the probe that was here would falsely grab a device.
2006-02-02 23:57:31 +00:00
Matt Jacob
53af7d226e Remove use of inlines and use the functions as a library.
Larger code space, possibly performance hit, but more portable.
Certainly less questionable use of inlining.

Suggested by: des
2006-02-02 21:31:34 +00:00
Paul Saab
0e74562fbc Enable 64bit SGL's on PERC 4/DC 2006-02-02 17:51:47 +00:00
Paul Saab
447c943f68 - Move the command setup from amr_start1 into the card specific submit
routines.
- Add or replace cpu_spinwait() with DELAY(1) to a few of the busy
  loops when reading from the controller to work around firmware bugs
  which can crash the controller.
2006-02-02 17:50:59 +00:00
Oleg Bulyzhin
ee7ef91c36 Enable 'complete' rx checksum offloading (i.e. let chip calculate checksums
with pseudo header for tcp/udp packets). This could save one in_pseudo() call
per incoming tcp/udp packet.

Approved by:	glebius (mentor)
MFC after:	3 weeks
2006-02-02 09:58:31 +00:00