AR5416 MAC based ath

[Ain Tepp]

Hi,

Seems like enabling interface (like ifconfig ath0 up) causes trap. Other
operations, like setting operational mode works fine.
If i use "ifconfig ath0 up" i get trap for current ifconfig process pid.
If i use
# ifconfig wlan0 up

I get all time pid 0 trap:

Trap cause = 7 (bus error (load or store) - kernel mode)
[ thread pid 0 tid 100028 ]
Stopped at      platform_start+0x1b10:  srl     v0,a1,8
db> bt
Tracing pid 0 tid 100028 td 0xc0c7b000
db_trace_thread+30 (?,?,?,?) ra 80055a48 sp c74657b0 sz 24
8005592c+11c (0,?,ffffffff,?) ra 8005537c sp c74657c8 sz 32
80054fe8+394 (?,?,?,?) ra 8005550c sp c74657e8 sz 168
db_command_loop+78 (?,?,?,?) ra 80057ca8 sp c7465890 sz 24
80057ba0+108 (?,?,?,?) ra 801ace78 sp c74658a8 sz 424
kdb_trap+108 (?,?,?,?) ra 8037d730 sp c7465a50 sz 32
trap+1620 (?,?,?,?) ra 803733e0 sp c7465a70 sz 176
MipsKernGenException+134 (0,deadc0de,a258,c7465c74) ra 80391428 sp
c7465b20 sz0
80391420+8 (?,?,?,?) ra 0 sp c7465be8 sz 0
pid 0
db>


Wbr,

Ain

> In message: <3f993635fc9c54a4d44aa360e26a9be0.squirrel at webmail.equix.ee>
>             "Ain" <freebsd at luftivennad.com> writes:
> : Hello!
> :
> : Last month was here a discussion about AR92XX (with subject line "Trap
> : when setting up ath0"). Now i have same kind trap, my chip is ar9220. If
> i
> : understand correctly, patch for 71xx
> : (http://people.freebsd.org/~imp/ar71xx_ath_war.diff) contains magic
> number
> : AR_RXCFG, what is not defined in ar5416reg.h. Can i use same magic to
> get
> : my  AR92XX working? If not, maybe some other ideas?
>
> The RXCFG bug is reported to only be with the older chipset.  Atheros
> didn't fix that bug in the current generation of processors.  The
> newer, N chipsets don't have this bug.  If you're seeing issues, then
> chances are very good that it is a different problem.  My work around
> likely won't help you much.  I'd start by building a kernel with
> symbols and using the debugger to help you track down the issue.  With
> the WAR I put in, I'd always get a different trap address.  Maybe this
> is something as simple as a driver bug (deref of NULL, say)....
>
> Warner
>