Re: Signal 11 when using 'ps'
>>>>> Ed Cogburn writes:
ec> Thanks guys for the help.
ec> Ok first, my CPU is not overclocked. My hardware is
ec> all fairly new though, so I'll list the major components:
ec> Athlon "Thunderbird" CPU, 1Ghz
You lucky bastard.
ec> Fourth, the sig11 errors are occuring at the same
ec> place every time. 100% consistent. (Although ps's output has
ec> changed, see below.)
Good, that makes it look like a real bug.
ec> Ok, I've been playing around with BIOS settings and
ec> noticed something has changed, ps is now sending its output to
ec> the screen twice. ??? The following is a fragment of an
ec> strace of ps, followed by a gdb session showing the duplicated
ec> output and a backtrace, if that helps.
ec> **************************************
ec> open("/boot/System.map-2.2.17", O_RDONLY|O_NONBLOCK|O_NOCTTY) = -1
ec> ENOENT (No such file or directory)
ec> open("/boot/System.map", O_RDONLY|O_NONBLOCK|O_NOCTTY) = 6
ec> fstat(6, {st_mode=S_IFREG|0664, st_size=141296, ...}) = 0
ec> old_mmap(NULL, 141297, PROT_READ|PROT_WRITE, MAP_PRIVATE, 6, 0) =
ec> 0x401df000
ec> close(6) = 0
ec> mremap(0x401be000, 135168, 12288, MREMAP_MAYMOVE) = 0x401be000
ec> --- SIGSEGV (Segmentation fault) ---
This is exactly the same as last time.
ec> Starting program: /bin/ps
ec> (no debugging symbols found)...
ec> (no debugging symbols found)...
ec> (no debugging symbols found)...
This means that the output could be totally useless. You need to
recompile with debugging symbols.
ec> #0 0x40022d1f in tty_to_dev () from /lib/libproc.so.2.0.6
ec> #1 0x4002300a in open_psdb () from /lib/libproc.so.2.0.6
ec> #2 0x804fcbf in strcpy ()
ec> #3 0x40047cfc in __libc_start_main () from /lib/libc.so.6
ec> (gdb)
ec> **************************************
ec> This is an strace of console-apt which is also giving
ec> me sig11 errors (same place, same time, 100% consistent):
Console-apt has bugs. This is when you are resizing the window,
right? It's probably been fixed in CVS for a long time now and just
not released.
ec> What should I do? Would compiling ps with debug info
ec> and stepping through it help? Should I try something else?
Please do compile ps with debugging symbols and send a gdb backtrace.
I'll have a look at the source and see if I can find the bug. Also,
send a bug report to the maintainers of procps, both upstream and
Debian using the "bug" command.
--
Chris
Reply to: