Skip to contentSkip to conversation

panic at rb_next when do pick_next_task_fair

DISCUSSION 蒋雄伟(蒋冲) 2 participants 2 messages Nov 8, 2016

lore
Share & tools
lore
#0 lore raw

Hi,all

We encountered the following panic in 2.6.32 and 3.10.0:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: [] rb_next+0x1/0x50

Call Trace:
 ? pick_next_task_fair+0x129/0x1d0
  [] __schedule+0x12a/0x910
  [] schedule+0x29/0x70
  [] futex_wait_queue_me+0xc4/0x120
  [] futex_wait+0x179/0x280
  [] ? hrtimer_get_res+0x50/0x50
  [] ? futex_wait_queue_me+0xa2/0x120
  [] do_futex+0xfe/0x5b0
  [] ? blk_finish_plug+0x14/0x40
  [] ? SyS_madvise+0x3c5/0x820
  [] SyS_futex+0x80/0x180
  [] system_call_fastpath+0x16/0x1b

It seems that after passing the check "if (!cfs_rq->nr_running)" , __pick_first_entity sometimes still return NULL, then leads to panic.

Would you please give any advise?
Thanks a lot.

Best regards.

#1 lore raw
Quoted 3 lines · 蒋雄伟(蒋冲)

Hi,all

We encountered the following panic in 2.6.32 and 3.10.0:

Can you reproduce using anything recent? I've really no idea what any of those kernels looked like. That's ancient software.