Netscape or kernel bug, you be the judge...

Ricky Beam (root@defiant.interpath.net)
Thu, 14 Nov 1996 13:59:35 -0500 (EST)


I skipped most of the previous discussion about netscape on the list, so
forgive me if I'm repeating a known problem...

After several days, netscape will just SEGV out of the blue and continually
fail to even start from then on. Rebooting the machine will get everything
back to normal for a few more days...

It happened today after I tiggered 10 or so FTP's (RealAudio stuff.) It
downloaded everything just fine and then existed with a bus error. I've
captured the syscall trace (strace -ftvx) before and after reboot. I'm not
gonna waste the bandwidth to post the entire traces, but here's the small
of it:
(before)
13:25:23 select(7, [6], NULL, NULL, {0, 0}) = 0 (Timeout)
13:25:23 sigprocmask(SIG_BLOCK, NULL, []) = 0
13:25:23 --- SIGSEGV (Segmentation fault) ---
13:25:23 getpid() = 11146
13:25:23 kill(11146, SIGUNUSED) = 0
13:25:23 --- SIGUNUSED (Bus error) ---
13:25:23 +++ killed by SIGUNUSED +++

(after)
226 13:39:18 select(8, [7], NULL, NULL, {0, 0}) = 0 (Timeout)
226 13:39:18 sigprocmask(SIG_BLOCK, NULL, []) = 0
226 13:39:18 sigprocmask(SIG_BLOCK, [ALRM], []) = 0
226 13:39:18 select(8, [7], NULL, NULL, {86400, 0}) = 1 (in [7], left {86399,
440000})
226 13:39:18 gettimeofday({847996758, 836152}, NULL) = 0
226 13:39:18 sigprocmask(SIG_SETMASK, [], NULL) = 0
226 13:39:18 --- SIGALRM (Alarm clock) ---
226 13:39:18 sigreturn() = ? (mask now [])

Those are the coresponding spots of the first select call after starting to
talk to the X server (as best I can tell.) It would appear that netscape is
being killed by a problem with the signalling functions (or some other
non-sycall events between the sigprocmask's. I don't expect Netscape to
be sending out a -ggdb version anytime soon.)
[Actually, that is the first select call - period.]

Anyone interested in the full strace's can get them at:
http://lacota.interpath.net/~jfbeam/Netscape/Strace/

On a side note, I'd just like to know why netscape is executing 'netstat -ni':
227 13:39:15 execve("/bin/netstat", ["netstat", "-ni"]["ignoreeof=10", "HOSTNA
ME=lacota.interpath.net", "MINICOM=-c on", "LD_LIBRARY_PATH=/lib:/usr/lib:/u", "
TERM=xterm", "HOSTTYPE=i386-linux", "PATH=/bin:/usr/bin:/sbin:/usr/sb", "HOME=/h
ome/jfbeam", "SHELL=/bin/sh", "LS_OPTIONS=--8bit --color=tty -F", "PS1=\\h:\\w\\
$ ", "PS2=> ", "USER=jfbeam", "MANPATH=/usr/man/preformat:/usr/", "LESS=-C", "DI
SPLAY=:0.0", ...] <unfinished ...>

--Ricky