public inbox for [email protected]
 help / color / mirror / Atom feed
From: Pavel Begunkov <[email protected]>
To: Dan Carpenter <[email protected]>,
	Hillf Danton <[email protected]>
Cc: syzbot <[email protected]>,
	[email protected], [email protected],
	[email protected], [email protected],
	[email protected], [email protected]
Subject: Re: INFO: rcu detected stall in io_uring_release
Date: Mon, 20 Apr 2020 15:57:11 +0300	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <20200420114719.GA2659@kadam>

On 4/20/2020 2:47 PM, Dan Carpenter wrote:
> On Sun, Apr 19, 2020 at 12:06:26PM +0800, Hillf Danton wrote:
>>
>> Sat, 18 Apr 2020 11:59:13 -0700
>>>
>>> syzbot found the following crash on:
>>>
>>> HEAD commit:    8f3d9f35 Linux 5.7-rc1
>>> git tree:       upstream
>>> console output: https://syzkaller.appspot.com/x/log.txt?x=115720c3e00000
>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=5d351a1019ed81a2
>>> dashboard link: https://syzkaller.appspot.com/bug?extid=66243bb7126c410cefe6
>>> compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
>>>
>>> Unfortunately, I don't have any reproducer for this crash yet.
>>>
>>> IMPORTANT: if you fix the bug, please add the following tag to the commit:
>>> Reported-by: [email protected]
>>>
>>> rcu: INFO: rcu_preempt self-detected stall on CPU
>>> rcu: 	0-....: (10500 ticks this GP) idle=57e/1/0x4000000000000002 softirq=44329/44329 fqs=5245 
>>> 	(t=10502 jiffies g=79401 q=2096)
>>> NMI backtrace for cpu 0
>>> CPU: 0 PID: 23184 Comm: syz-executor.5 Not tainted 5.7.0-rc1-syzkaller #0
>>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
>>> Call Trace:
>>>  <IRQ>
>>>  __dump_stack lib/dump_stack.c:77 [inline]
>>>  dump_stack+0x188/0x20d lib/dump_stack.c:118
>>>  nmi_cpu_backtrace.cold+0x70/0xb1 lib/nmi_backtrace.c:101
>>>  nmi_trigger_cpumask_backtrace+0x231/0x27e lib/nmi_backtrace.c:62
>>>  trigger_single_cpu_backtrace include/linux/nmi.h:164 [inline]
>>>  rcu_dump_cpu_stacks+0x19b/0x1e5 kernel/rcu/tree_stall.h:254
>>>  print_cpu_stall kernel/rcu/tree_stall.h:475 [inline]
>>>  check_cpu_stall kernel/rcu/tree_stall.h:549 [inline]
>>>  rcu_pending kernel/rcu/tree.c:3225 [inline]
>>>  rcu_sched_clock_irq.cold+0x55d/0xcfa kernel/rcu/tree.c:2296
>>>  update_process_times+0x25/0x60 kernel/time/timer.c:1727
>>>  tick_sched_handle+0x9b/0x180 kernel/time/tick-sched.c:176
>>>  tick_sched_timer+0x4e/0x140 kernel/time/tick-sched.c:1320
>>>  __run_hrtimer kernel/time/hrtimer.c:1520 [inline]
>>>  __hrtimer_run_queues+0x5ca/0xed0 kernel/time/hrtimer.c:1584
>>>  hrtimer_interrupt+0x312/0x770 kernel/time/hrtimer.c:1646
>>>  local_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1113 [inline]
>>>  smp_apic_timer_interrupt+0x15b/0x600 arch/x86/kernel/apic/apic.c:1138
>>>  apic_timer_interrupt+0xf/0x20 arch/x86/entry/entry_64.S:829
>>>  </IRQ>
>>> RIP: 0010:io_ring_ctx_wait_and_kill+0x98/0x5a0 fs/io_uring.c:7301
>>> Code: 01 00 00 4d 89 f4 48 b8 00 00 00 00 00 fc ff df 4c 89 ed 49 c1 ec 03 48 c1 ed 03 49 01 c4 48 01 c5 eb 1c e8 3a ea 9d ff f3 90 <41> 80 3c 24 00 0f 85 53 04 00 00 48 83 bb 10 01 00 00 00 74 21 e8
>>> RSP: 0018:ffffc9000897fdf0 EFLAGS: 00000293 ORIG_RAX: ffffffffffffff13
>>> RAX: ffff888024082080 RBX: ffff88808df8e000 RCX: 1ffff9200112ffab
>>> RDX: 0000000000000000 RSI: ffffffff81d549c6 RDI: ffff88808df8e300
>>> RBP: ffffed1011bf1c2c R08: 0000000000000001 R09: ffffed1011bf1c61
>>> R10: ffff88808df8e307 R11: ffffed1011bf1c60 R12: ffffed1011bf1c22
>>> R13: ffff88808df8e160 R14: ffff88808df8e110 R15: ffffffff81d54ed0
>>>  io_uring_release+0x3e/0x50 fs/io_uring.c:7324
>>>  __fput+0x33e/0x880 fs/file_table.c:280
>>>  task_work_run+0xf4/0x1b0 kernel/task_work.c:123
>>>  tracehook_notify_resume include/linux/tracehook.h:188 [inline]
>>>  exit_to_usermode_loop+0x2fa/0x360 arch/x86/entry/common.c:165
>>>  prepare_exit_to_usermode arch/x86/entry/common.c:196 [inline]
>>>  syscall_return_slowpath arch/x86/entry/common.c:279 [inline]
>>>  do_syscall_64+0x6b1/0x7d0 arch/x86/entry/common.c:305
>>>  entry_SYSCALL_64_after_hwframe+0x49/0xb3
>>
>> Make io ring ctx's percpu_ref balanced.
>>
>> --- a/fs/io_uring.c
>> +++ b/fs/io_uring.c
>> @@ -5904,6 +5904,7 @@ static int io_submit_sqes(struct io_ring
>>  fail_req:
>>  			io_cqring_add_event(req, err);
>>  			io_double_put_req(req);
>> +			--submitted;
>>  			break;
>>  		}
> 
> 
> fs/io_uring.c
>   5880          for (i = 0; i < nr; i++) {
>   5881                  const struct io_uring_sqe *sqe;
>   5882                  struct io_kiocb *req;
>   5883                  int err;
>   5884  
>   5885                  sqe = io_get_sqe(ctx);
>   5886                  if (unlikely(!sqe)) {
>   5887                          io_consume_sqe(ctx);
>   5888                          break;
>   5889                  }
>   5890                  req = io_alloc_req(ctx, statep);
>   5891                  if (unlikely(!req)) {
>   5892                          if (!submitted)
>   5893                                  submitted = -EAGAIN;
>   5894                          break;
>   5895                  }
>   5896  
>   5897                  err = io_init_req(ctx, req, sqe, statep, async);
>                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> On the success path io_init_req() takes some references like:
> 
> 	get_cred(req->work.creds);

If a req have got into io_init_req(), than it'll be put at some point
with io_put_req(). io_req_work_drop_env() called from there will clean
up req->work.creds.

>  
> That one is probably buggy and should be put if the call to:
> 
> 	return io_req_set_file(state, req, fd, sqe_flags);
> 
> fails...  But io_req_set_file() takes some other references if it
> succeeds like percpu_ref_get(req->fixed_file_refs); and it's not clear
> that those are released if io_submit_sqe() fails.

The same should happen with req->fixed_file_refs, though I don't
remember in details.

> 
>   5898                  io_consume_sqe(ctx);
>   5899                  /* will complete beyond this point, count as submitted */
>   5900                  submitted++;

Regarding, "--submitted" patch -- we take 1 ctx->refs per request, which
is put in io_put_req(). So after a request passes the line above (5900),
it's ref will be eventually dropped in io_put_req() and friends.

And it's a bit more peculiar because io_submit_sqes() batch-takes N refs
first, and then puts unused back at the end.

>   5901  
>   5902                  if (unlikely(err)) {
>   5903  fail_req:
>   5904                          io_cqring_add_event(req, err);
>   5905                          io_double_put_req(req);
>   5906                          break;
>   5907                  }
>   5908  
>   5909                  trace_io_uring_submit_sqe(ctx, req->opcode, req->user_data,
>   5910                                                  true, async);
>   5911                  err = io_submit_sqe(req, sqe, statep, &link);
>                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> here
> 
>   5912                  if (err)
>   5913                          goto fail_req;
>   5914          }
> 
> regards,
> dan carpenter
> 

-- 
Pavel Begunkov

  reply	other threads:[~2020-04-20 12:57 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-04-18 18:59 INFO: rcu detected stall in io_uring_release syzbot
     [not found] ` <[email protected]>
2020-04-19 19:57   ` Jens Axboe
2020-04-20 11:47   ` Dan Carpenter
2020-04-20 12:57     ` Pavel Begunkov [this message]
2020-05-12  2:17 ` syzbot
2020-05-12 15:25   ` Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox