[vox-tech] Mystery Segmentation Faults: grep and sed

[Dylan Beaudette]

> On Tue 28 Dec 04, 11:59 AM, Dylan Beaudette <dylan at iici.no-ip.org> said:
>> Good morning!
>>
>> Last night all was well with my G4 powerbook, even after a rather large
>> number of updates from and apt-get update && apt-get upgrade.
>>
>> However, this morning, I noticed some strange segmentation faults when
>> the
>> machine was starting. They all seemed to be happening after a sed
>> command
>> was issued. I tried manually using sed and grep, and both now seem to be
>> giving segementation faults!
>>
>> here is a quick example with the output of strace included:
>>
>> --------------------snip----------------------
>> dylan at pbg4:~$ echo "food" | grep -i food
>> Segmentation fault
>>
>> dylan at pbg4:~$ strace echo "food" | grep -i food
>> execve("/bin/echo", ["echo", "food"], [/* 27 vars */]) = 0
>> uname({sys="Linux", node="pbg4", ...})  = 0
>> brk(0)                                  = 0x10013680
>> mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0)
>> =
>> 0x30016000
>> access("/etc/ld.so.nohwcap", F_OK)      = -1 ENOENT (No such file or
>> directory)
>> open("/etc/ld.so.preload", O_RDONLY)    = -1 ENOENT (No such file or
>> directory)
>> open("/etc/ld.so.cache", O_RDONLY)      = 3
>> fstat64(3, {st_mode=S_IFREG|0644, st_size=72216, ...}) = 0
>> mmap(NULL, 72216, PROT_READ, MAP_PRIVATE, 3, 0) = 0x30027000
>> close(3)                                = 0
>> access("/etc/ld.so.nohwcap", F_OK)      = -1 ENOENT (No such file or
>> directory)
>> open("/lib/libc.so.6", O_RDONLY)        = 3
>> read(3, "\177ELF\1\2\1\0\0\0\0\0\0\0\0\0\0\3\0\24\0\0\0\1\0\1\322"...,
>> 512) = 512
>> fstat64(3, {st_mode=S_IFREG|0644, st_size=1298544, ...}) = 0
>> mmap(0xfea1000, 1370024, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) =
>> 0xfea1000
>> mprotect(0xffd3000, 116648, PROT_NONE)  = 0
>> mmap(0xffe1000, 49152, PROT_READ|PROT_WRITE|PROT_EXEC,
>> MAP_PRIVATE|MAP_FIXED, 3, 0x130000) = 0xffe1000
>> mmap(0xffed000, 10152, PROT_READ|PROT_WRITE|PROT_EXEC,
>> MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0xffed000
>> close(3)                                = 0
>> munmap(0x30027000, 72216)               = 0
>> brk(0)                                  = 0x10013680
>> brk(0x10034680)                         = 0x10034680
>> brk(0)                                  = 0x10034680
>> brk(0x10035000)                         = 0x10035000
>> open("/usr/lib/locale/locale-archive", O_RDONLY|O_LARGEFILE) = 3
>> fstat64(3, {st_mode=S_IFREG|0644, st_size=3202928, ...}) = 0
>> mmap2(NULL, 2097152, PROT_READ, MAP_PRIVATE, 3, 0) = 0x30027000
>> close(3)                                = 0
>> fstat64(1, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
>> mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0)
>> =
>> 0x30017000
>> write(1, "food\n", 5)                   = -1 EPIPE (Broken pipe)
>> --- SIGPIPE (Broken pipe) @ 0 (0) ---
>> +++ killed by SIGPIPE +++
>> Segmentation fault
>> -----------------------snip-----------------------
>
> Hi Dylan,
>
> I don't believe this isn't the actual segfault.  I think you want to do:
>
>    strace -ff -s"99" foo | bar
>
> What you show here is simply writing to a pipe.  The pipe broke because
> the
> thing on the other end segfaulted (I *think*).  This shows echo.  You want
> to show grep.  Again, "I think".  ;)
>
> What kernel are you running?  Someone on the Opera Linux mailing list
> complained of strange segfaults after upgrading his kernel.  Sounds a tad
> similar.  I don't recall his kernel version, but I believe it was 2.6.8
> and
> he said that it also happens with 2.6.9 (I'm sure the Opera launcher, a
> shell script, uses sed and/or grep among other things).  Come to think of
> it, he was also a Debian user.
>
> Hmmmmm.......
>
>> can anyone make any sense of all of this? for some reason my sed and
>> grep
>> broke between yesterday night and this morning.... and i imagine that it
>> had something to do with the update last night.
>>
>> any ideas would be greatly appreciated!
>
> If you can stomach the traffic, debian-user is, IMHO the best general
> technical mailing list on the net.  In its heyday, I thought vox-tech was
> nearly as good (but I'm biased...)  :)
>
> Pete

Thanks for the info Pete! I haven't actually used strace before, and only
thought about it from reading the vox-tech list!

on a whim i restarted my machine, and now all is well! i can't really
explain why that would fix something like that, but i sure am happy to
have a functioning system back.

i cross-posted to the debian-ppc list, and i will report back on any
findings!

thanks,

Dylan