[OpenAFS] SIGSEGV on aklog, pts or vos commands at process.c:213

[Michal Svamberg]

Hello,
I have a problem on some computers (Debian Etch + etch-backports).
Older versions of openafs (1.4.2, 1.4.4 and 1.4.6) don't have this problem.

Installed version:
# dpkg -l |grep openafs | awk '{print $2 "\t\t\t" $3}'
libopenafs-dev                  1.4.7~pre3.dfsg1-1~bpo40+1
openafs-client                  1.4.7~pre3.dfsg1-1~bpo40+1
openafs-dbg                     1.4.7~pre3.dfsg1-1~bpo40+1
openafs-doc                     1.4.7~pre3.dfsg1-1~bpo40+1
openafs-fileserver                      1.4.7~pre3.dfsg1-1~bpo40+1
openafs-krb5                    1.4.7~pre3.dfsg1-1~bpo40+1
openafs-modules-2.6.22-4-686
1.4.7~pre3.dfsg1-1~bpo40+1+2.6.22-6~bpo40+2
openafs-modules-source                  1.4.7~pre3.dfsg1-1~bpo40+1

# uname -a
Linux listik.zcu.cz 2.6.22-4-686 #1 SMP Tue Feb 12 16:29:32 UTC 2008
i686 GNU/Linux

$ gdb --quiet
(gdb) file /usr/bin/pts
Reading symbols from /usr/bin/pts...Reading symbols from /usr/lib/debug/usr/bin/
pts...done.
Using host libthread_db library "/usr/lib/debug/libthread_db.so.1".
done.
(gdb) run mem svamberg
Starting program: /usr/bin/pts mem svamberg

Program received signal SIGSEGV, Segmentation fault.
savecontext (ep=0x8076140 <Create_Process_Part2>, savearea=0x80c59c4,
    sp=0xb7cf700c "üýţ˙") at ./process.c:213
213                     (*EP) ();
(gdb) l
208                     jmpBuffer[LWP_FP] = ptr_mangle((jmp_buf_type) sp);
209     #endif
210                     longjmp(jmp_tmp, 1);
211                     break;
212                 case 1:
213                     (*EP) ();
214                     assert(0);      /* never returns */
215                     break;
216                 default:
217                     perror("Error in setjmp1\n");
(gdb) bt
#0  savecontext (ep=0x80754a0 <Create_Process_Part2>, savearea=0x80bde44,
    sp=0xb7caa00c "üýţ˙") at ./process.c:213
#1  0x080757e7 in LWP_CreateProcess (ep=0x80766f0 <IOMGR>,
    stacksize=<value optimized out>, priority=0, parm=0x0,
    name=0x807f01d "IO MANAGER", pid=0x80921c8) at ./lwp.c:409
#2  0x080766e6 in IOMGR_Initialize () at ./iomgr.c:820
#3  0x08074ae4 in rxi_InitializeThreadSupport () at rx_lwp.c:117
#4  0x0806d791 in rx_InitHost (host=0, port=0) at rx.c:403
#5  0x0806d9d9 in rx_Init (port=0) at rx.c:540
#6  0x0804dcf4 in pr_Initialize (secLevel=0, confDir=0x8083040 "/etc/openafs",
    cell=0xbfb25696 "zcu.cz") at ptuser.c:166
#7  0x0804b1aa in auth_to_cell (context=0x809a058, cell=<value optimized out>,
    realm=0x0) at aklog_main.c:720
#8  0x0804c472 in aklog (argc=1, argv=0xbfb2fbe4) at aklog_main.c:1381
#9  0x0804a0c2 in main (argc=Cannot access memory at address 0xf951e550
) at aklog.c:18
#10 0xb7db6450 in __libc_start_main (main=0x804a0a0 <main>, argc=1,
    ubp_av=0xbfb2fbe4, init=0x807b5b0 <__libc_csu_init>,
    fini=0x807b560 <__libc_csu_fini>, rtld_fini=0xb7fb0dc0 <_dl_fini>,
    stack_end=0xbfb2fbdc) at libc-start.c:222

The same SIGSEGV exist with running aklog or vos. What's wrongs?

Thanks.
Michal Svamberg