2011-11-11 05:02:15

by Thomas Pedersen

Subject: regression in "mac80211: support adding IV-room in the skb for CCMP keys"

Hello,

Any kernel debugging tips involving stack corruption? :)

When testing a secure mesh topology in qemu using mac80211_hwsim with
master-2011-11-09 I get a hard lockup, but no OOPS. gdb was able to
show:

0xffffffff8129df9a in delay_tsc (loops=2660152) at arch/x86/lib/delay.c:80
80 if (unlikely(cpu != smp_processor_id())) {
(gdb) bt
#0 0xffffffff8129df9a in delay_tsc (loops=2660152) at arch/x86/lib/delay.c:80
#1 0xffffffff8129df0d in __delay (loops=<optimized out>) at
arch/x86/lib/delay.c:112
#2 __const_udelay (xloops=<optimized out>) at arch/x86/lib/delay.c:126
#3 0xffffffff816eeb7e in panic (fmt=<optimized out>) at kernel/panic.c:154
#4 0xffffffff816fbe32 in oops_end (flags=582, regs=0xffff880017c03698, signr=9)
at arch/x86/kernel/dumpstack.c:243
#5 0xffffffff816ee301 in no_context (regs=0xffff880017c03698,
error_code=<optimized out>, address=7)
at arch/x86/mm/fault.c:680
#6 0xffffffff816ee4db in __bad_area_nosemaphore
(regs=0xffff880017c03698, error_code=0, address=7,
si_code=196609) at arch/x86/mm/fault.c:758
#7 0xffffffff816ee50d in bad_area_nosemaphore (regs=<optimized out>,
error_code=<optimized out>,
address=<optimized out>) at arch/x86/mm/fault.c:765
#8 0xffffffff816fe714 in do_page_fault (regs=0xffff880017c03698, error_code=0)
at arch/x86/mm/fault.c:1111
#9 <signal handler called>
#10 0xffffffffa0005825 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
(gdb) file net/mac80211/mac
mac80211.ko mac80211.mod.c mac80211.mod.o mac80211.o
(gdb) file net/mac80211/mac80211.ko

(gdb) bt
#0 0xffffffff8129df9a in ?? ()
#1 0xffff880017c033f8 in ?? ()
#2 0xffffffff8129df0d in ?? ()
#3 0xffff880017c03478 in ?? ()
#4 0xffffffff816eeb7e in ?? ()
#5 0xffffffff810479a3 in ?? ()
#6 0xffffffff00000008 in ?? ()
#7 0xffff880017c03488 in ?? ()
#8 0xffff880017c03428 in ?? ()
#9 0xffff880017c03698 in ?? ()
#10 0x0000000000000001 in ieee80211_napi_poll (napi=0x289738, budget=-623469717)
at net/mac80211/main.c:1099
#11 0x0000000000000002 in ieee80211_napi_poll (napi=0x289738, budget=-623469717)
at net/mac80211/main.c:1099
#12 0x0000000000000000 in ?? ()
(gdb)

I've narrowed this down to 077a9154898b374f20555adc3f620cccd02581d6 "
mac80211: support adding IV-room in the skb for CCMP keys". Not sure
how this will affect encryption under mac80211_hwsim, but hopefully
this is useful to someone more knowledgeable.

Thanks!
Thomas

2011-11-11 06:13:20

by Arik Nemtsov

Subject: Re: regression in "mac80211: support adding IV-room in the skb for CCMP keys"

On Fri, Nov 11, 2011 at 07:01, Thomas Pedersen <[email protected]> wrote:
> Hello,
>
> Any kernel debugging tips involving stack corruption? :)
>
> When testing a secure mesh topology in qemu using mac80211_hwsim with
> master-2011-11-09 I get a hard lockup, but no OOPS. gdb was able to
> show:
>

Did you try applying this
http://www.spinics.net/lists/linux-wireless/msg79976.html ?

Arik

2011-11-11 18:41:40

by Thomas Pedersen

Subject: Re: regression in "mac80211: support adding IV-room in the skb for CCMP keys"

On Thu, Nov 10, 2011 at 10:13 PM, Arik Nemtsov <[email protected]> wrote:
> On Fri, Nov 11, 2011 at 07:01, Thomas Pedersen <[email protected]> wrote:
>> Hello,
>>
>> Any kernel debugging tips involving stack corruption? :)
>>
>> When testing a secure mesh topology in qemu using mac80211_hwsim with
>> master-2011-11-09 I get a hard lockup, but no OOPS. gdb was able to
>> show:
>>
>
> Did you try applying this
> http://www.spinics.net/lists/linux-wireless/msg79976.html ?

That fixed it, thanks!

Thomas