Package: dump
Version: 0.4b43-1
Severity: important
i've got two sparcs (a enterprise 250 and a sunfire v240) which were recently
upgraded from lenny to squeeze. they've got locally built kernels (one ancient,
one 2.6.35.12).
on both machines dump locks up straight after writing out the
"estimated xxx blocks" message. strace of the main dump process shows it
waiting for its child process, which is as expected. stracing the child
process shows that it is in a tight loop, waiting for (nonexistent)
children.
a quick check of the source shows that the only relevant wait calls are
in startnewtape() which does a clone(). apparently that doesn't work
properly in this case, or the child wouldn't mistakenly assume it's
the parent.
dump from lenny works fine on both machines, with the very same kernels.
here are some relevant snippets from strace (which unfortunately doesn't
include the cloned process even with -f given...):
execve("/sbin/dump", ["dump", "-0f", "/stuff/dumpfile", "/dev/md0"], [/* 23
vars */]) = 0
...usual startup...
write(2, " DUMP: ", 8 DUMP: ) = 8
write(2, "mapping (Pass II) [directories]\n", 32mapping (Pass II) [directories]
) = 32
write(2, " DUMP: ", 8 DUMP: ) = 8
write(2, "estimated 42588 blocks.\n", 24estimated 42588 blocks.
) = 24
rt_sigprocmask(SIG_BLOCK, [INT], NULL, 8) = 0
time(NULL) = 1303265797
rt_sigprocmask(SIG_UNBLOCK, [INT], NULL, 8) = 0
clone(Process 2099 attached
child_stack=0, flags=0x80000000|SIGCHLD) = 2099
[pid 2096] rt_sigprocmask(SIG_BLOCK, [INT], NULL, 8) = 0
[pid 2096] wait4(-1, Process 2096 suspended
<unfinished ...>
the manual strace on 2099 simply gives a never-ending list of
wait4(-1, 0xffa6e57c, 0, NULL) = -1 ECHILD (No child processes)
wait4(-1, 0xffa6e57c, 0, NULL) = -1 ECHILD (No child processes)
wait4(-1, 0xffa6e57c, 0, NULL) = -1 ECHILD (No child processes)
regards
az
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
