Greetings, and thanks again! Just a quick note to say my last hunch was
wrong. The reexec after personality set is failing because there is not
enough memory to copy the arguments and environment. This has to be on
the heap and not the stack as we change the stack location with the
personality call. I'll be putting in a fix with the next upload.
Take care,
Robert Jacobs <debbts...@eamp.org> writes:
> Package: maxima
> Version: 5.31.0-4
> Severity: important
>
> Dear Maintainer,
>
> I've had bug #678404 recur on ia32. (Sorry if I should have reopened that one
> instead)
>
> $ gdb /usr/lib/maxima/5.31.0/binary-gcl/maxima
> Program received signal SIGSEGV, Segmentation fault.
> 0x00068179 in ?? ()
> (gdb) bt
> #0 0x00068179 in ?? ()
> #1 0x00069955 in ?? ()
> #2 0x000b4c94 in malloc ()
> #3 0xb7c4fa51 in __GI___strdup (s=0xbffffdb4 "QT_IM_MODULE=xim") at
> strdup.c:42
> #4 0x00021a0e in ?? ()
> #5 0xb7beb8c5 in __libc_start_main (main=0x20bf0, argc=1, ubp_av=0xbffff594,
> init=0x31ccd0, fini=0x31cd40, rtld_fini=0xb7fee0c0 <_dl_fini>,
> stack_end=0xbffff58c) at libc-start.c:260
> #6 0x00021e78 in ?? ()
> (gdb) disas $eip,+20
> Dump of assembler code from 0x68179 to 0x6818d:
> => 0x00068179: mov DWORD PTR [eax+0x4],edi
> 0x0006817c: add esp,0x6c
> 0x0006817f: mov eax,0x1
> 0x00068184: pop ebx
> 0x00068185: pop esi
> 0x00068186: pop edi
> 0x00068187: pop ebp
> 0x00068188: ret
> 0x00068189: fstp st(0)
> 0x0006818b: fld1
> (gdb) info registers
> eax 0x8a8cb0 9079984
> ecx 0x438000 4423680
> edx 0x7 7
> ebx 0x2335 9013
> esp 0xbffff280 0xbffff280
> ebp 0x32d 0x32d
> esi 0x1 1
> edi 0x5dc740 6145856
> eip 0x68179 0x68179
> eflags 0x10246 [ PF ZF IF RF ]
> cs 0x73 115
> ss 0x7b 123
> ds 0x7b 123
> es 0x7b 123
> fs 0x0 0
> gs 0x33 51
> (gdb) x (int *)0x8a8cb4
> 0x8a8cb4: 0x005dc740
> [...]
> cat /proc/[pid]/maps
> [...]
> 00885000-008a9000 r-xp 0086c000 09:00 805313494
> /usr/lib/maxima/5.31.0/binary-gcl/maxima
> 008a9000-008d6000 rwxp 00890000 09:00 805313494
> /usr/lib/maxima/5.31.0/binary-gcl/maxima
> [...]
>
> So it's trying to update a pointer that's (barely) not in a writeable page.
>
> strace:
> [...]
> personality(0xffffffff /* PER_??? */) = 0
> personality(0x8240000 /* PER_??? */) = 0
> personality(0xffffffff /* PER_??? */) = 0x8240000
> mprotect(0x7ae000, 26652672, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
> rt_sigaction(SIGSEGV, {0x25f30, [SEGV], SA_STACK|SA_RESTART|SA_SIGINFO},
> NULL, 8) = 0
> brk(0) = 0x345d000
> brk(0x3679000) = 0x3679000
> --- SIGSEGV (Segmentation fault) @ 0 (0) ---
>
> ltrace:
> [...]
> [pid 16410] memset(0x01389608, '\000', 26112)
> = 0x01389608
> [pid 16410] sbrk(0)
> = 0x02adf000
> [pid 16410] sbrk(0)
> = 0x02adf000
> [pid 16410] sbrk(2211840)
> = 0x02adf000
> [pid 16410] --- SIGSEGV (Segmentation fault) ---
>
>
> I looked into the process of getting debugging symbols but it looks a
> bit daunting.
>
> As far as I know, neither SELinux nor any other thing should be
> messing up permissions; sbrk and brk (obviously) don't look like
> they're randomized, and the personality() call seemed to have returned
> success.
>
> Bisecting using snapshot.debian.org shows this started happening
> between 5.31.0-1 and 5.31.0-3, and only on ia32. (My amd64 machine
> works fine)
>
> Is there anything else I should try?
>
> Thanks-
> - Robert Jacobs
>
>
> -- System Information:
> Debian Release: jessie/sid
> APT prefers unstable
> APT policy: (500, 'unstable'), (500, 'stable')
> Architecture: i386 (i686)
>
> Kernel: Linux 3.10-1-686-pae (SMP w/1 CPU core)
> Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
> Shell: /bin/sh linked to /bin/dash
>
> Versions of packages maxima depends on:
> ii gnuplot-x11 4.6.4-1
> ii libc6 2.17-93
> ii libgmp10 2:5.1.2+dfsg-3
> ii libreadline6 6.2+dfsg-0.1
> ii libx11-6 2:1.6.2-1
>
> Versions of packages maxima recommends:
> ii maxima-share 5.31.0-4
>
> Versions of packages maxima suggests:
> ii maxima-doc 5.31.0-4
> pn maxima-emacs <none>
> pn texmacs <none>
> ii tk8.4 [wish] 8.4.20-1
> ii tk8.5 [wish] 8.5.14-2
> ii xmaxima 5.31.0-4
>
> -- no debconf information
>
>
>
>
--
Camm Maguire c...@maguirefamily.org
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
