From: Vladislav B. <vs...@vl...> - 2012-08-24 20:37:30
|
Hi Richard, The essential questions here are: 1. Is the system idle during that or busy? 2. What action did you do to trigger it? 3. If the system recovers itself after some time or hangs forever? If yes, after which time? It might be related to a single threaded nature of workqueues on that ancient kernel, but not yet clear how. Vlad Richard Sharpe, on 08/24/2012 04:21 PM wrote: > Hi folks, > > I am seeing this message sometimes on system shutdown: > > INFO: task scst_usr_releas:10172 blocked for more than 120 seconds. > "echo 0> /proc/sys/kernel/hung_task_timeout_secs" disables this message. > scst_usr_rele D ffff81002101d7a0 0 10172 87 10175 5100 (L-TLB) > ffff8101265e3e00 0000000000000046 ffffffffffffffff 0000000000000000 > 0000000000000000 0000000000000009 ffff810112854100 ffff810107fc8080 > 0000410f9fa0328f 000000000000070a ffff8101128542e8 000000033be00c00 > Call Trace: > [<ffffffff80062ff2>] thread_return+0x62/0xfe > [<ffffffff800a2c56>] keventd_create_kthread+0x0/0xc4 > [<ffffffff80063c53>] __mutex_lock_slowpath+0x60/0x9b > [<ffffffff8003d859>] lock_timer_base+0x1b/0x3c > [<ffffffff80063c9d>] .text.lock.mutex+0xf/0x14 > [<ffffffff8009ff32>] flush_workqueue+0x3f/0x87 > [<ffffffff8886373a>] :scst:sgv_pool_destroy+0x6f/0x13c > [<ffffffff88863aeb>] :scst:sgv_pool_del+0x56/0xa2 > [<ffffffff888ba364>] :scst_user:dev_user_exit_dev+0x1c2/0x28b > [<ffffffff888ba42d>] :scst_user:__dev_user_release+0x0/0x15 > [<ffffffff888ba436>] :scst_user:__dev_user_release+0x9/0x15 > [<ffffffff80032720>] kthread+0xfe/0x132 > [<ffffffff8005dfb1>] child_rip+0xa/0x11 > [<ffffffff800a2c56>] keventd_create_kthread+0x0/0xc4 > [<ffffffff80032622>] kthread+0x0/0x132 > [<ffffffff8005dfa7>] child_rip+0x0/0x11 > > I think I am shutting things down correctly. Is there a quick fix or > something obvious that I am doing wrong? > > This is with SCST 2.1 or there abouts and Linux 2.6.18-274 (with one > patch to export set_cpus_allowed_ptr) on a CentOS 4.7 base. Apart from > that, things seem to work fine. > |