* Hanging process on SMP machines? @ 2004-09-09 20:13 ` H. J. Lu 0 siblings, 0 replies; 6+ messages in thread From: H. J. Lu @ 2004-09-09 20:13 UTC (permalink / raw) To: linux kernel, linux ia64 kernel I notice that a process may hang on SMP machines at random: http://bugzilla.kernel.org/show_bug.cgi?id332 I can reliably trigger it within 15 minutes under SMP kernel on P4 HT and 4-way ia64 machines. Has anyone else seen it? H.J. ^ permalink raw reply [flat|nested] 6+ messages in thread
* Hanging process on SMP machines? @ 2004-09-09 20:13 ` H. J. Lu 0 siblings, 0 replies; 6+ messages in thread From: H. J. Lu @ 2004-09-09 20:13 UTC (permalink / raw) To: linux kernel, linux ia64 kernel I notice that a process may hang on SMP machines at random: http://bugzilla.kernel.org/show_bug.cgi?id=3332 I can reliably trigger it within 15 minutes under SMP kernel on P4 HT and 4-way ia64 machines. Has anyone else seen it? H.J. ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Hanging process on SMP machines? 2004-09-09 20:13 ` H. J. Lu @ 2004-09-10 7:45 ` Andrew Morton -1 siblings, 0 replies; 6+ messages in thread From: Andrew Morton @ 2004-09-10 7:45 UTC (permalink / raw) To: H. J. Lu; +Cc: linux-kernel, linux-ia64 "H. J. Lu" <hjl@lucon.org> wrote: > > I notice that a process may hang on SMP machines at random: > > http://bugzilla.kernel.org/show_bug.cgi?id332 > > I can reliably trigger it within 15 minutes under SMP kernel on P4 HT > and 4-way ia64 machines. Has anyone else seen it? It's easy to reproduce on 2-way x86, however it doesn't look like a kernel bug. akpm 2503 0.0 0.2 3892 640 pts/0 S 00:34 0:00 | \_ make akpm 2504 0.0 0.0 1368 192 pts/0 S 00:34 0:00 | \_ time expect test.exp akpm 2505 0.0 0.5 4776 1276 pts/0 S 00:34 0:00 | \_ expect test.exp akpm 4726 0.0 0.0 0 0 ? Z 00:35 0:00 | \_ [true] <defunct> process 4726 is sleeping at the end of do_exit(): schedule(); <<- here BUG(); /* Avoid "noreturn function does return". */ for (;;) ; } So it has completely exitted and is waiting for someone to reap its exit code and stack slot via wait4(). So what is its parent up to? (gdb) thread 69 [Switching to thread 69 (Thread 2505)]#0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 257 __timeout = schedule_timeout(__timeout); (gdb) bt #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 #1 0xc016579c in sys_select (n=5, inp=0x804b444, outp=0x804b4c4, exp=0x804b544, tvp=0xbfffe1d0) at fs/select.c:354 #2 0xc0105e39 in sysenter_past_esp () at arch/i386/kernel/semaphore.c:177 The parent is sleeping in select() rather than wait()ing for children. (gdb) f 0 #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 257 __timeout = schedule_timeout(__timeout); (gdb) p __timeout $1 = 3000000 For 3,000 seconds. I changed your 3000 to 30 and lo, the script hangs for 30 seconds every now and then, and then resumes. ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Hanging process on SMP machines? @ 2004-09-10 7:45 ` Andrew Morton 0 siblings, 0 replies; 6+ messages in thread From: Andrew Morton @ 2004-09-10 7:45 UTC (permalink / raw) To: H. J. Lu; +Cc: linux-kernel, linux-ia64 "H. J. Lu" <hjl@lucon.org> wrote: > > I notice that a process may hang on SMP machines at random: > > http://bugzilla.kernel.org/show_bug.cgi?id=3332 > > I can reliably trigger it within 15 minutes under SMP kernel on P4 HT > and 4-way ia64 machines. Has anyone else seen it? It's easy to reproduce on 2-way x86, however it doesn't look like a kernel bug. akpm 2503 0.0 0.2 3892 640 pts/0 S 00:34 0:00 | \_ make akpm 2504 0.0 0.0 1368 192 pts/0 S 00:34 0:00 | \_ time expect test.exp akpm 2505 0.0 0.5 4776 1276 pts/0 S 00:34 0:00 | \_ expect test.exp akpm 4726 0.0 0.0 0 0 ? Z 00:35 0:00 | \_ [true] <defunct> process 4726 is sleeping at the end of do_exit(): schedule(); <<- here BUG(); /* Avoid "noreturn function does return". */ for (;;) ; } So it has completely exitted and is waiting for someone to reap its exit code and stack slot via wait4(). So what is its parent up to? (gdb) thread 69 [Switching to thread 69 (Thread 2505)]#0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 257 __timeout = schedule_timeout(__timeout); (gdb) bt #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 #1 0xc016579c in sys_select (n=5, inp=0x804b444, outp=0x804b4c4, exp=0x804b544, tvp=0xbfffe1d0) at fs/select.c:354 #2 0xc0105e39 in sysenter_past_esp () at arch/i386/kernel/semaphore.c:177 The parent is sleeping in select() rather than wait()ing for children. (gdb) f 0 #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 257 __timeout = schedule_timeout(__timeout); (gdb) p __timeout $1 = 3000000 For 3,000 seconds. I changed your 3000 to 30 and lo, the script hangs for 30 seconds every now and then, and then resumes. ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Hanging process on SMP machines? 2004-09-10 7:45 ` Andrew Morton @ 2004-09-10 16:39 ` H. J. Lu -1 siblings, 0 replies; 6+ messages in thread From: H. J. Lu @ 2004-09-10 16:39 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-ia64 On Fri, Sep 10, 2004 at 12:45:49AM -0700, Andrew Morton wrote: > "H. J. Lu" <hjl@lucon.org> wrote: > > > > I notice that a process may hang on SMP machines at random: > > > > http://bugzilla.kernel.org/show_bug.cgi?id332 > > > > I can reliably trigger it within 15 minutes under SMP kernel on P4 HT > > and 4-way ia64 machines. Has anyone else seen it? > > It's easy to reproduce on 2-way x86, however it doesn't look like a kernel > bug. > > > akpm 2503 0.0 0.2 3892 640 pts/0 S 00:34 0:00 | \_ make > akpm 2504 0.0 0.0 1368 192 pts/0 S 00:34 0:00 | \_ time expect test.exp > akpm 2505 0.0 0.5 4776 1276 pts/0 S 00:34 0:00 | \_ expect test.exp > akpm 4726 0.0 0.0 0 0 ? Z 00:35 0:00 | \_ [true] <defunct> > > process 4726 is sleeping at the end of do_exit(): > > > schedule(); <<- here > BUG(); > /* Avoid "noreturn function does return". */ > for (;;) ; > } > > So it has completely exitted and is waiting for someone to reap its exit > code and stack slot via wait4(). So what is its parent up to? > > > (gdb) thread 69 > [Switching to thread 69 (Thread 2505)]#0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, > timeout=0xcc8b5fa0) at fs/select.c:257 > 257 __timeout = schedule_timeout(__timeout); > (gdb) bt > #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 > #1 0xc016579c in sys_select (n=5, inp=0x804b444, outp=0x804b4c4, exp=0x804b544, tvp=0xbfffe1d0) > at fs/select.c:354 > #2 0xc0105e39 in sysenter_past_esp () at arch/i386/kernel/semaphore.c:177 > > The parent is sleeping in select() rather than wait()ing for children. > I don't think SIGCHLD is blocked. Shouldn't SIGCHLD interrupt select? UP kernel doesn't have this problem. H.J. ^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Hanging process on SMP machines? @ 2004-09-10 16:39 ` H. J. Lu 0 siblings, 0 replies; 6+ messages in thread From: H. J. Lu @ 2004-09-10 16:39 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-ia64 On Fri, Sep 10, 2004 at 12:45:49AM -0700, Andrew Morton wrote: > "H. J. Lu" <hjl@lucon.org> wrote: > > > > I notice that a process may hang on SMP machines at random: > > > > http://bugzilla.kernel.org/show_bug.cgi?id=3332 > > > > I can reliably trigger it within 15 minutes under SMP kernel on P4 HT > > and 4-way ia64 machines. Has anyone else seen it? > > It's easy to reproduce on 2-way x86, however it doesn't look like a kernel > bug. > > > akpm 2503 0.0 0.2 3892 640 pts/0 S 00:34 0:00 | \_ make > akpm 2504 0.0 0.0 1368 192 pts/0 S 00:34 0:00 | \_ time expect test.exp > akpm 2505 0.0 0.5 4776 1276 pts/0 S 00:34 0:00 | \_ expect test.exp > akpm 4726 0.0 0.0 0 0 ? Z 00:35 0:00 | \_ [true] <defunct> > > process 4726 is sleeping at the end of do_exit(): > > > schedule(); <<- here > BUG(); > /* Avoid "noreturn function does return". */ > for (;;) ; > } > > So it has completely exitted and is waiting for someone to reap its exit > code and stack slot via wait4(). So what is its parent up to? > > > (gdb) thread 69 > [Switching to thread 69 (Thread 2505)]#0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, > timeout=0xcc8b5fa0) at fs/select.c:257 > 257 __timeout = schedule_timeout(__timeout); > (gdb) bt > #0 0xc01653e9 in do_select (n=5, fds=0xcc8b5fa4, timeout=0xcc8b5fa0) at fs/select.c:257 > #1 0xc016579c in sys_select (n=5, inp=0x804b444, outp=0x804b4c4, exp=0x804b544, tvp=0xbfffe1d0) > at fs/select.c:354 > #2 0xc0105e39 in sysenter_past_esp () at arch/i386/kernel/semaphore.c:177 > > The parent is sleeping in select() rather than wait()ing for children. > I don't think SIGCHLD is blocked. Shouldn't SIGCHLD interrupt select? UP kernel doesn't have this problem. H.J. ^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2004-09-10 16:46 UTC | newest] Thread overview: 6+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2004-09-09 20:13 Hanging process on SMP machines? H. J. Lu 2004-09-09 20:13 ` H. J. Lu 2004-09-10 7:45 ` Andrew Morton 2004-09-10 7:45 ` Andrew Morton 2004-09-10 16:39 ` H. J. Lu 2004-09-10 16:39 ` H. J. Lu
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.