Messages in this thread | | | Date | Mon, 13 Jul 2009 14:38:10 -0700 | From | Andrew Morton <> | Subject | Re: 2.6.31-rc1-mmotm0702 - ps command hangs inside kernel |
| |
On Mon, 13 Jul 2009 16:54:51 -0400 Valdis.Kletnieks@vt.edu wrote:
> Several times recently, I've had the 'ps' command hang inside the kernel > for extended periods of time - usually around 1100 seconds, but today I > had one that hung there for 2351 seconds. > > Info: It's always reading the same file - /proc/<pid>/status, where the > pid is the 'pcscd' daemon. And eventually, it manages to exit on its own. > Whatever is getting horked up resets itself - subsequent 'ps' commands work > just fine. > > My best guess is that it's getting onto a squirrely vma in > get_stack_usage_in_bytes() in the for/follow_page loop. Possibly highly > relevant - /usr/sbin/pcscd is one of the few 32-bit binaries running in > a mostly 64-bit runtime.
OK, thanks for the analysis.
> Here's the traceback of the ps, as reported by 2 alt-sysrq-t several > minutes apart: > > ps R running task 3936 26646 26580 0x00000080 > ffff88005a599be8 ffffffff81499842 ffff88005a599b78 ffffffff8103589b > ffffffff81035805 ffff88007e71ea40 ffff880002120f80 ffff8800657d8fe0 > 000000000000df78 ffff8800657d8fe0 000000000000df78 ffff8800657d8fe8 > Call Trace: > [<ffffffff81499842>] ? thread_return+0xb6/0xfa > [<ffffffff8103589b>] ? finish_task_switch+0xd1/0xf4 > [<ffffffff81035805>] ? finish_task_switch+0x3b/0xf4 > [<ffffffff81065f3f>] ? trace_hardirqs_on_caller+0x1f/0x145 > [<ffffffff8149b2a6>] ? trace_hardirqs_on_thunk+0x3a/0x3f > [<ffffffff8101f76b>] ? smp_apic_timer_interrupt+0x81/0x8f > [<ffffffff81046231>] ? irq_exit+0xaf/0xb4 > [<ffffffff8100bd80>] restore_args+0x0/0x30 > [<ffffffff81065f3f>] ? trace_hardirqs_on_caller+0x1f/0x145 > [<ffffffff8149b2a6>] ? trace_hardirqs_on_thunk+0x3a/0x3f > [<ffffffff8101f76b>] ? smp_apic_timer_interrupt+0x81/0x8f > [<ffffffff810b8222>] ? IS_ERR+0x25/0x2c > [<ffffffff810b8f22>] ? follow_page+0x28/0x2e3 > [<ffffffff81121a04>] ? proc_pid_status+0x60c/0x694 > [<ffffffff81066072>] ? trace_hardirqs_on+0xd/0xf > [<ffffffff8111e42c>] ? proc_single_show+0x57/0x74 > [<ffffffff810e90b4>] ? seq_read+0x249/0x49b > [<ffffffff8116c840>] ? security_file_permission+0x11/0x13 > [<ffffffff810d05e1>] ? vfs_read+0xe0/0x141 > [<ffffffff810d8191>] ? path_put+0x1d/0x21 > [<ffffffff810d06f8>] ? sys_read+0x45/0x69 > [<ffffffff8100b2ab>] ? system_call_fastpath+0x16/0x1b > > ps R running task 3936 26646 26580 0x00000080 > ffff88005a599bd8 ffffffff81499842 ffff88000213bf80 0000000000000000 > ffff88005a599b78 ffffffff8103589b ffffffff81035805 ffff88007e71ea40 > ffff88000213bf80 ffff8800657d8fe0 000000000000df78 ffff8800657d8fe8 > Call Trace: > [<ffffffff81499842>] ? thread_return+0xb6/0xfa > [<ffffffff8103589b>] ? finish_task_switch+0xd1/0xf4 > [<ffffffff81035805>] ? finish_task_switch+0x3b/0xf4 > [<ffffffff81065f3f>] ? trace_hardirqs_on_caller+0x1f/0x145 > [<ffffffff8149b2a6>] ? trace_hardirqs_on_thunk+0x3a/0x3f > [<ffffffff8101f76b>] smp_apic_timer_interrupt+0x81/0x8f > [<ffffffff81046231>] ? irq_exit+0xaf/0xb4 > [<ffffffff8100bd80>] ? restore_args+0x0/0x30 > [<ffffffff810b8222>] ? IS_ERR+0x25/0x2c > [<ffffffff810b8222>] ? IS_ERR+0x25/0x2c > [<ffffffff810b8f22>] ? follow_page+0x28/0x2e3 > [<ffffffff810b91d9>] ? follow_page+0x2df/0x2e3 > [<ffffffff811219d8>] ? proc_pid_status+0x5e0/0x694 > [<ffffffff81066072>] ? trace_hardirqs_on+0xd/0xf > [<ffffffff8111e42c>] ? proc_single_show+0x57/0x74 > [<ffffffff810e90b4>] ? seq_read+0x249/0x49b > [<ffffffff8116c840>] ? security_file_permission+0x11/0x13 > [<ffffffff810d05e1>] ? vfs_read+0xe0/0x141 > [<ffffffff810d8191>] ? path_put+0x1d/0x21 > [<ffffffff810d06f8>] ? sys_read+0x45/0x69 > [<ffffffff8100b2ab>] ? system_call_fastpath+0x16/0x1b
Another possibility is that pcscd has gone off on a long in-kernel sulk (polling hardware?) while holding some lock which ps needs (eg, mmap_sem).
It would be useful if you can grab a pcscd backtrace during the stall.
I have a new version of those patches in my inbox to look at - I'm a bit swamped in backlog at present. I could drop the old version from my tree for a while, but it's unclear how that would help us find the cause of the problem.
> Here's the output of 'cat /proc/2031/maps' currently - I didn't think to > get one during the last event: > > # cat /proc/2031/maps > 08047000-0805e000 r-xp 00000000 fe:05 51682 /usr/sbin/pcscd > 0805e000-0805f000 rw-p 00016000 fe:05 51682 /usr/sbin/pcscd > 0805f000-080e8000 rw-p 00000000 00:00 0 > 09990000-099b1000 rw-p 00000000 00:00 0 [heap] > 44368000-4436f000 r-xp 00000000 fe:05 20682 /usr/lib/libusb-0.1.so.4.4.4 > 4436f000-44371000 rw-p 00006000 fe:05 20682 /usr/lib/libusb-0.1.so.4.4.4 > 45c81000-45cab000 r-xp 00000000 fe:00 8715 /lib/libgcc_s-4.4.0-20090708.so.1 > 45cab000-45cac000 rw-p 0002a000 fe:00 8715 /lib/libgcc_s-4.4.0-20090708.so.1 > 47c00000-47c20000 r-xp 00000000 fe:00 8266 /lib/ld-2.10.1.so > 47c20000-47c21000 r--p 0001f000 fe:00 8266 /lib/ld-2.10.1.so > 47c21000-47c22000 rw-p 00020000 fe:00 8266 /lib/ld-2.10.1.so > 47c24000-47d8f000 r-xp 00000000 fe:00 8360 /lib/libc-2.10.1.so > 47d8f000-47d90000 ---p 0016b000 fe:00 8360 /lib/libc-2.10.1.so > 47d90000-47d92000 r--p 0016b000 fe:00 8360 /lib/libc-2.10.1.so > 47d92000-47d93000 rw-p 0016d000 fe:00 8360 /lib/libc-2.10.1.so > 47d93000-47d96000 rw-p 00000000 00:00 0 > 47d98000-47d9b000 r-xp 00000000 fe:00 8388 /lib/libdl-2.10.1.so > 47d9b000-47d9c000 r--p 00002000 fe:00 8388 /lib/libdl-2.10.1.so > 47d9c000-47d9d000 rw-p 00003000 fe:00 8388 /lib/libdl-2.10.1.so > 47d9f000-47db5000 r-xp 00000000 fe:00 8675 /lib/libpthread-2.10.1.so > 47db5000-47db6000 ---p 00016000 fe:00 8675 /lib/libpthread-2.10.1.so > 47db6000-47db7000 r--p 00016000 fe:00 8675 /lib/libpthread-2.10.1.so > 47db7000-47db8000 rw-p 00017000 fe:00 8675 /lib/libpthread-2.10.1.so > 47db8000-47dba000 rw-p 00000000 00:00 0 > f6c00000-f6c21000 rw-p 00000000 00:00 0 > f6c21000-f6d00000 ---p 00000000 00:00 0 > f6dff000-f6e00000 ---p 00000000 00:00 0 > f6e00000-f7600000 rw-p 00000000 00:00 0 > f7600000-f7621000 rw-p 00000000 00:00 0 > f7621000-f7700000 ---p 00000000 00:00 0 > f773b000-f773c000 ---p 00000000 00:00 0 > f773c000-f7f3f000 rw-p 00000000 00:00 0 > f7f50000-f7f51000 rw-s 0000f000 fe:09 86049 /var/run/pcscd.pub > f7f51000-f7f52000 rw-s 0000e000 fe:09 86049 /var/run/pcscd.pub > f7f52000-f7f53000 rw-s 0000d000 fe:09 86049 /var/run/pcscd.pub > f7f53000-f7f54000 rw-s 0000c000 fe:09 86049 /var/run/pcscd.pub > f7f54000-f7f55000 rw-s 0000b000 fe:09 86049 /var/run/pcscd.pub > f7f55000-f7f56000 rw-s 0000a000 fe:09 86049 /var/run/pcscd.pub > f7f56000-f7f57000 rw-s 00009000 fe:09 86049 /var/run/pcscd.pub > f7f57000-f7f58000 rw-s 00008000 fe:09 86049 /var/run/pcscd.pub > f7f58000-f7f59000 rw-s 00007000 fe:09 86049 /var/run/pcscd.pub > f7f59000-f7f5a000 rw-s 00006000 fe:09 86049 /var/run/pcscd.pub > f7f5a000-f7f5b000 rw-s 00005000 fe:09 86049 /var/run/pcscd.pub > f7f5b000-f7f5c000 rw-s 00004000 fe:09 86049 /var/run/pcscd.pub > f7f5c000-f7f5d000 rw-s 00003000 fe:09 86049 /var/run/pcscd.pub > f7f5d000-f7f5e000 rw-s 00002000 fe:09 86049 /var/run/pcscd.pub > f7f5e000-f7f5f000 rw-s 00001000 fe:09 86049 /var/run/pcscd.pub > f7f5f000-f7f60000 rw-s 00000000 fe:09 86049 /var/run/pcscd.pub > f7f60000-f7f61000 r-xp 00000000 00:00 0 [vdso] > fffe8000-ffffd000 rw-p 00000000 00:00 0 [stack] > >
| |