gcc coredump with 2.6.12+ kernels

gcc coredump with 2.6.12+ kernels

[Johnny Stenback]

Hey all,

I just attempted to upgrade my kernel to 2.6.13. The kernel appears to 
boot and run just fine, but when I try to build any larger projects like 
Mozilla or the Linux kernel I constantly get segfaults from gcc. All 
other apps *seem* to work fine. I remember seeing this with 2.6.12 too 
when I tried to upgrade to it too but I didn't have the time to 
investigate at all then, but now I see the same problem with 2.6.13. The 
last version I've used that didn't show this problem is 2.6.11.3, and 
that's running with no problems here.

When gcc segfaults I get the following messages in the messages log:

cc1[16775]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
00007fffffaaf0a0 error 4
cc1[17086]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
00007fffffc4dfc0 error 4
cc1[17788]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
00007fffffd777e0 error 4
cc1[17823]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
00007fffffc4d630 error 4
cc1[17895]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
00007ffffffd2330 error 4

I'm on a dual AMD Opteron system, running x86_64 code. Using Fedora Core 
2 (yeah, old, I know...) and gcc 3.3.3 20040412.

I've ran memtest etc on this system, and with 2.6.11.3 and older kernels 
I've ran this system under moderate load for months w/o rebooting and 
I've never had any problems like this, so I'm fairly certain that I 
don't have bad hardware.

Anyone have any clues as to what could be causing this problem?

-- 
jst

Re: gcc coredump with 2.6.12+ kernels

[Alexander Nyberg]

On Sat, Sep 03, 2005 at 10:25:37AM -0700 Johnny Stenback wrote:

> Hey all,
> 
> I just attempted to upgrade my kernel to 2.6.13. The kernel appears to 
> boot and run just fine, but when I try to build any larger projects like 
> Mozilla or the Linux kernel I constantly get segfaults from gcc. All 
> other apps *seem* to work fine. I remember seeing this with 2.6.12 too 
> when I tried to upgrade to it too but I didn't have the time to 
> investigate at all then, but now I see the same problem with 2.6.13. The 
> last version I've used that didn't show this problem is 2.6.11.3, and 
> that's running with no problems here.
> 
> When gcc segfaults I get the following messages in the messages log:
> 
> cc1[16775]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
> 00007fffffaaf0a0 error 4
> cc1[17086]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
> 00007fffffc4dfc0 error 4
> cc1[17788]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
> 00007fffffd777e0 error 4
> cc1[17823]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
> 00007fffffc4d630 error 4
> cc1[17895]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
> 00007ffffffd2330 error 4
> 
> I'm on a dual AMD Opteron system, running x86_64 code. Using Fedora Core 
> 2 (yeah, old, I know...) and gcc 3.3.3 20040412.

Does it still happen if you run:

echo 0 > /proc/sys/kernel/randomize_va_space

Re: gcc coredump with 2.6.12+ kernels

[Johnny Stenback]

Alexander Nyberg wrote:
> On Sat, Sep 03, 2005 at 10:25:37AM -0700 Johnny Stenback wrote:
> 
>> Hey all,
>>
>> I just attempted to upgrade my kernel to 2.6.13. The kernel appears to 
>> boot and run just fine, but when I try to build any larger projects like 
>> Mozilla or the Linux kernel I constantly get segfaults from gcc. All 
>> other apps *seem* to work fine. I remember seeing this with 2.6.12 too 
>> when I tried to upgrade to it too but I didn't have the time to 
>> investigate at all then, but now I see the same problem with 2.6.13. The 
>> last version I've used that didn't show this problem is 2.6.11.3, and 
>> that's running with no problems here.
>>
>> When gcc segfaults I get the following messages in the messages log:
>>
>> cc1[16775]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffaaf0a0 error 4
>> cc1[17086]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffc4dfc0 error 4
>> cc1[17788]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffd777e0 error 4
>> cc1[17823]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffc4d630 error 4
>> cc1[17895]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007ffffffd2330 error 4
>>
>> I'm on a dual AMD Opteron system, running x86_64 code. Using Fedora Core 
>> 2 (yeah, old, I know...) and gcc 3.3.3 20040412.
> 
> Does it still happen if you run:
> 
> echo 0 > /proc/sys/kernel/randomize_va_space

Just tried that, and I still get the same error, and the same error in 
the log too (just a different address):

cc1plus[2961]: segfault at 0000000000000000 rip
00000036f2b0119e rsp 00007fffffffdbb0 error 4

Anything else I can try?

> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

-- 
jst

Re: gcc coredump with 2.6.12+ kernels

[Johnny Stenback]

Hello again,

Got some more info that may be of value here. After I saw gcc coredump 
on me with 2.6.13 I tried halting my system and I saw a kernel panic 
while shutting down. The stack trace it showed me was:

   flush_tlb_mm+201
   exit_mmap+249
   mmput+49
   do_exit+434
   sys_reboot+264
   __group_send_sig_info+184
   group_send_info+118
   thread_return
   thread_return+82
   filp_close+110
   system_call+126

All typed down by hand since it never made it into a log, so there can 
be omissions and typos in there (and I didn't feel like typing down all 
those 64-bit hex numbers, sorry).

Johnny Stenback wrote:
> Alexander Nyberg wrote:
>> On Sat, Sep 03, 2005 at 10:25:37AM -0700 Johnny Stenback wrote:
>>
>>> Hey all,
>>>
>>> I just attempted to upgrade my kernel to 2.6.13. The kernel appears 
>>> to boot and run just fine, but when I try to build any larger 
>>> projects like Mozilla or the Linux kernel I constantly get segfaults 
>>> from gcc. All other apps *seem* to work fine. I remember seeing this 
>>> with 2.6.12 too when I tried to upgrade to it too but I didn't have 
>>> the time to investigate at all then, but now I see the same problem 
>>> with 2.6.13. The last version I've used that didn't show this problem 
>>> is 2.6.11.3, and that's running with no problems here.
>>>
>>> When gcc segfaults I get the following messages in the messages log:
>>>
>>> cc1[16775]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>>> 00007fffffaaf0a0 error 4
>>> cc1[17086]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>>> 00007fffffc4dfc0 error 4
>>> cc1[17788]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>>> 00007fffffd777e0 error 4
>>> cc1[17823]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>>> 00007fffffc4d630 error 4
>>> cc1[17895]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>>> 00007ffffffd2330 error 4
>>>
>>> I'm on a dual AMD Opteron system, running x86_64 code. Using Fedora 
>>> Core 2 (yeah, old, I know...) and gcc 3.3.3 20040412.
>>
>> Does it still happen if you run:
>>
>> echo 0 > /proc/sys/kernel/randomize_va_space
> 
> Just tried that, and I still get the same error, and the same error in 
> the log too (just a different address):
> 
> cc1plus[2961]: segfault at 0000000000000000 rip
> 00000036f2b0119e rsp 00007fffffffdbb0 error 4
> 
> Anything else I can try?
> 
>> -
>> To unsubscribe from this list: send the line "unsubscribe 
>> linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
>>
> 

-- 
jst

Re: gcc coredump with 2.6.12+ kernels

[Johnny Stenback]

FWIW, this issue was resolved in 2.6.14. I don't know off hand which 
patch did it, but I don't see this problem now after upgrading to 2.6.14.

Thanks to whoever figured this out :)

Alexander Nyberg wrote:
> On Sat, Sep 03, 2005 at 10:25:37AM -0700 Johnny Stenback wrote:
> 
>> Hey all,
>>
>> I just attempted to upgrade my kernel to 2.6.13. The kernel appears to 
>> boot and run just fine, but when I try to build any larger projects like 
>> Mozilla or the Linux kernel I constantly get segfaults from gcc. All 
>> other apps *seem* to work fine. I remember seeing this with 2.6.12 too 
>> when I tried to upgrade to it too but I didn't have the time to 
>> investigate at all then, but now I see the same problem with 2.6.13. The 
>> last version I've used that didn't show this problem is 2.6.11.3, and 
>> that's running with no problems here.
>>
>> When gcc segfaults I get the following messages in the messages log:
>>
>> cc1[16775]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffaaf0a0 error 4
>> cc1[17086]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffc4dfc0 error 4
>> cc1[17788]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffd777e0 error 4
>> cc1[17823]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007fffffc4d630 error 4
>> cc1[17895]: segfault at 0000000000000000 rip 00000036f2b0119e rsp 
>> 00007ffffffd2330 error 4
>>
>> I'm on a dual AMD Opteron system, running x86_64 code. Using Fedora Core 
>> 2 (yeah, old, I know...) and gcc 3.3.3 20040412.
> 
> Does it still happen if you run:
> 
> echo 0 > /proc/sys/kernel/randomize_va_space
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

-- 
jst