Messages in this thread | | | Date | Wed, 27 Apr 2011 16:03:33 -0700 | From | "Paul E. McKenney" <> | Subject | Re: mcelog stalls on 2.6.39-rc5 |
| |
On Wed, Apr 27, 2011 at 01:59:01PM -0600, Jim Schutt wrote: > Hi, > > Testing 2.6.39-rc5 is giving me the following stall: > > [ 5767.731001] INFO: rcu_sched_state detected stall on CPU 1 (t=60001 jiffies) > [ 5767.732001] INFO: rcu_sched_state detected stalls on CPUs/tasks: { 1} (detected by 0, t=60002 jiffies) > [ 5947.763001] INFO: rcu_sched_state detected stall on CPU 1 (t=240032 jiffies) > [ 5947.764001] INFO: rcu_sched_state detected stalls on CPUs/tasks: { 1} (detected by 0, t=240034 jiffies) > [ 6024.489362] libceph: mon0 172.17.40.34:6789 socket closed > [ 6121.281139] INFO: task mcelog:6513 blocked for more than 120 seconds. > [ 6121.287575] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. > [ 6121.295397] ffff880177aefce8 0000000000000082 ffffffff810339b6 ffff880226d35a40 > [ 6121.302840] ffff88018226c3b0 ffff88018226c3b0 0000000000011e80 ffff880226d35a40 > [ 6121.310284] ffff88018226c760 ffff880177aefe80 ffff880177aefd18 ffffffff813af047 > [ 6121.317725] Call Trace: > [ 6121.320176] [<ffffffff810339b6>] ? calc_load_account_idle+0xe/0x1d > [ 6121.326437] [<ffffffff813af047>] schedule+0x159/0x193 > [ 6121.331569] [<ffffffff813af449>] schedule_timeout+0x36/0xe2 > [ 6121.337223] [<ffffffff810ad9eb>] ? trace_hardirqs_on+0x9/0x20 > [ 6121.343047] [<ffffffff813aee2b>] do_wait_for_common+0x97/0xe3 > [ 6121.348967] [<ffffffff8103e8fa>] ? try_to_wake_up+0x200/0x200 > [ 6121.354794] [<ffffffff8107a4bf>] ? __raw_spin_lock_irq+0x17/0x2f > [ 6121.360878] [<ffffffff813af2a9>] wait_for_common+0x36/0x4d > [ 6121.366441] [<ffffffff813af378>] wait_for_completion+0x1d/0x1f > [ 6121.372356] [<ffffffff8109ae6d>] synchronize_sched+0x40/0x49 > [ 6121.378096] [<ffffffff810635b8>] ? find_get_pid+0x1b/0x1b > [ 6121.383574] [<ffffffff81015fe5>] mce_read+0x17f/0x25d > [ 6121.388707] [<ffffffff81111af5>] ? rw_verify_area+0xac/0xdb > [ 6121.394358] [<ffffffff811121f1>] vfs_read+0xa9/0xe1 > [ 6121.399317] [<ffffffff8111248d>] sys_read+0x4c/0x70 > [ 6121.404278] [<ffffffff813b6deb>] system_call_fastpath+0x16/0x1b > [ 6127.795001] INFO: rcu_sched_state detected stall on CPU 1 (t=420064 jiffies) > [ 6127.796001] INFO: rcu_sched_state detected stalls on CPUs/tasks: { 1} (detected by 0, t=420066 jiffies) > [ 6241.410171] INFO: task mcelog:6513 blocked for more than 120 seconds. > > Reverting commit a4dd99250dc makes the stalls go away: > > rcu: create new rcu_access_index() and use in mce > > The MCE subsystem needs to sample an RCU-protected index outside of > any protection for that index. If this was a pointer, we would use > rcu_access_pointer(), but there is no corresponding rcu_access_index(). > This commit therefore creates an rcu_access_index() and applies it > to MCE. > > Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> > Tested-by: Zdenek Kabelac <zkabelac@redhat.com>
Wow! This is just instructions, just wrapping the access in ACCESS_ONCE().
Was the original RCU CPU stall repeatable?
Thanx, Paul
> Here's some bits from my config that might be relevant: > > # RCU Subsystem > CONFIG_TREE_RCU=y > # CONFIG_PREEMPT_RCU is not set > # CONFIG_RCU_TRACE is not set > CONFIG_RCU_FANOUT=64 > # CONFIG_RCU_FANOUT_EXACT is not set > # CONFIG_RCU_FAST_NO_HZ is not set > # CONFIG_TREE_RCU_TRACE is not set > CONFIG_PREEMPT_NOTIFIERS=y > # CONFIG_PREEMPT_NONE is not set > CONFIG_PREEMPT_VOLUNTARY=y > # CONFIG_PREEMPT is not set > CONFIG_X86_MCE=y > CONFIG_X86_MCE_INTEL=y > CONFIG_X86_MCE_AMD=y > CONFIG_X86_MCE_THRESHOLD=y > # CONFIG_X86_MCE_INJECT is not set > CONFIG_EDAC_DECODE_MCE=y > # CONFIG_EDAC_MCE_INJ is not set > CONFIG_EDAC_MCE=y > # CONFIG_SPARSE_RCU_POINTER is not set > # CONFIG_RCU_TORTURE_TEST is not set > CONFIG_RCU_CPU_STALL_DETECTOR=y > CONFIG_RCU_CPU_STALL_TIMEOUT=60 > CONFIG_RCU_CPU_STALL_DETECTOR_RUNNABLE=y > > Please let me know if there is anything I can do to help sort this out. > > -- Jim >
| |