April 23, 2024

Tycho Andersen

The Compute crew at Netflix is charged with managing all AWS and containerized workloads at Netflix, together with autoscaling, deployment of containers, challenge remediation, and many others. As a part of this crew, I work on fixing unusual issues that customers report.

This specific challenge concerned a customized inside FUSE filesystem: ndrive. It had been festering for a while, however wanted somebody to sit down down and have a look at it in anger. This weblog put up describes how I poked at /procto get a way of what was occurring, earlier than posting the problem to the kernel mailing checklist and getting schooled on how the kernel’s wait code really works!

We had a caught docker API name:

goroutine 146 [select, 8817 minutes]:
web/http.(*persistConn).roundTrip(0xc000658fc0, 0xc0003fc080, 0x0, 0x0, 0x0)
/usr/native/go/src/web/http/transport.go:2610 +0x765
web/http.(*Transport).roundTrip(0xc000420140, 0xc000966200, 0x30, 0x1366f20, 0x162)
/usr/native/go/src/web/http/transport.go:592 +0xacb
web/http.(*Transport).RoundTrip(0xc000420140, 0xc000966200, 0xc000420140, 0x0, 0x0)
/usr/native/go/src/web/http/roundtrip.go:17 +0x35
web/http.ship(0xc000966200, 0x161eba0, 0xc000420140, 0x0, 0x0, 0x0, 0xc00000e050, 0x3, 0x1, 0x0)
/usr/native/go/src/web/http/consumer.go:251 +0x454
web/http.(*Shopper).ship(0xc000438480, 0xc000966200, 0x0, 0x0, 0x0, 0xc00000e050, 0x0, 0x1, 0x10000168e)
/usr/native/go/src/web/http/consumer.go:175 +0xff
web/http.(*Shopper).do(0xc000438480, 0xc000966200, 0x0, 0x0, 0x0)
/usr/native/go/src/web/http/consumer.go:717 +0x45f
golang.org/x/web/context/ctxhttp.Do(0x163bd48, 0xc000044090, 0xc000438480, 0xc000966100, 0x0, 0x0, 0x0)
/go/pkg/mod/golang.org/x/[email protected]/context/ctxhttp/ctxhttp.go:27 +0x10f
github.com/docker/docker/consumer.(*Shopper).doRequest(0xc0001a8200, 0x163bd48, 0xc000044090, 0xc000966100, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, ...)
/go/pkg/mod/github.com/moby/[email protected]/consumer/request.go:132 +0xbe
github.com/docker/docker/consumer.(*Shopper).sendRequest(0xc0001a8200, 0x163bd48, 0xc000044090, 0x13d8643, 0x3, 0xc00079a720, 0x51, 0x0, 0x0, 0x0, ...)
/go/pkg/mod/github.com/moby/[email protected]/consumer/request.go:122 +0x156
/go/pkg/mod/github.com/moby/[email protected]/consumer/request.go:37
github.com/docker/docker/consumer.(*Shopper).ContainerInspect(0xc0001a8200, 0x163bd48, 0xc000044090, 0xc0006a01c0, 0x40, 0x0, 0x0, 0x0, 0x0, 0x0, ...)
/go/pkg/mod/github.com/moby/[email protected]/consumer/container_inspect.go:18 +0x128
github.com/Netflix/titus-executor/executor/runtime/docker.(*DockerRuntime).Kill(0xc000215180, 0x163bdb8, 0xc000938600, 0x1, 0x0, 0x0)
/var/lib/buildkite-agent/builds/ip-192-168-1-90-1/netflix/titus-executor/executor/runtime/docker/docker.go:2835 +0x310
github.com/Netflix/titus-executor/executor/runner.(*Runner).doShutdown(0xc000432dc0, 0x163bd10, 0xc000938390, 0x1, 0xc000b821e0, 0x1d, 0xc0005e4710)
/var/lib/buildkite-agent/builds/ip-192-168-1-90-1/netflix/titus-executor/executor/runner/runner.go:326 +0x4f4
github.com/Netflix/titus-executor/executor/runner.(*Runner).startRunner(0xc000432dc0, 0x163bdb8, 0xc00071e0c0, 0xc0a502e28c08b488, 0x24572b8, 0x1df5980)
/var/lib/buildkite-agent/builds/ip-192-168-1-90-1/netflix/titus-executor/executor/runner/runner.go:122 +0x391
created by github.com/Netflix/titus-executor/executor/runner.StartTaskWithRuntime
/var/lib/buildkite-agent/builds/ip-192-168-1-90-1/netflix/titus-executor/executor/runner/runner.go:81 +0x411

Right here, our administration engine has made an HTTP name to the Docker API’s unix socket asking it to kill a container. Our containers are configured to be killed by way of SIGKILL. However that is unusual. kill(SIGKILL) ought to be comparatively deadly, so what’s the container doing?

$ docker exec -it 6643cd073492 bash
OCI runtime exec failed: exec failed: container_linux.go:380: beginning container course of prompted: process_linux.go:130: executing setns course of prompted: exit standing 1: unknown

Hmm. Looks like it’s alive, however setns(2) fails. Why would that be? If we have a look at the method tree by way of ps awwfux, we see:

_ containerd-shim -namespace moby -workdir /var/lib/containerd/io.containerd.runtime.v1.linux/moby/6643cd073492ba9166100ed30dbe389ff1caef0dc3d35
| _ [docker-init]
| _ [ndrive] <defunct>

Okay, so the container’s init course of continues to be alive, however it has one zombie baby. What may the container’s init course of probably be doing?

# cat /proc/1528591/stack
[<0>] do_wait+0x156/0x2f0
[<0>] kernel_wait4+0x8d/0x140
[<0>] zap_pid_ns_processes+0x104/0x180
[<0>] do_exit+0xa41/0xb80
[<0>] do_group_exit+0x3a/0xa0
[<0>] __x64_sys_exit_group+0x14/0x20
[<0>] do_syscall_64+0x37/0xb0
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xae

It’s within the strategy of exiting, however it appears caught. The one baby is the ndrive course of in Z (i.e. “zombie”) state, although. Zombies are processes which have efficiently exited, and are ready to be reaped by a corresponding wait() syscall from their mother and father. So how may the kernel be caught ready on a zombie?

# ls /proc/1544450/process
1544450 1544574

Ah ha, there are two threads within the thread group. One in every of them is a zombie, perhaps the opposite one isn’t:

# cat /proc/1544574/stack
[<0>] request_wait_answer+0x12f/0x210
[<0>] fuse_simple_request+0x109/0x2c0
[<0>] fuse_flush+0x16f/0x1b0
[<0>] filp_close+0x27/0x70
[<0>] put_files_struct+0x6b/0xc0
[<0>] do_exit+0x360/0xb80
[<0>] do_group_exit+0x3a/0xa0
[<0>] get_signal+0x140/0x870
[<0>] arch_do_signal_or_restart+0xae/0x7c0
[<0>] exit_to_user_mode_prepare+0x10f/0x1c0
[<0>] syscall_exit_to_user_mode+0x26/0x40
[<0>] do_syscall_64+0x46/0xb0
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xae

Certainly it isn’t a zombie. It’s attempting to turn out to be one as exhausting as it may possibly, however it’s blocking inside FUSE for some motive. To search out out why, let’s have a look at some kernel code. If we have a look at zap_pid_ns_processes(), it does:

* Reap the EXIT_ZOMBIE kids we had earlier than we ignored SIGCHLD.
* kernel_wait4() may even block till our youngsters traced from the
* guardian namespace are indifferent and turn out to be EXIT_DEAD.
rc = kernel_wait4(-1, NULL, __WALL, NULL);
whereas (rc != -ECHILD);

which is the place we’re caught, however earlier than that, it has carried out:

/* Do not enable any extra processes into the pid namespace */

which is why docker can’t setns() — the namespace is a zombie. Okay, so we will’t setns(2), however why are we caught in kernel_wait4()? To know why, let’s have a look at what the opposite thread was doing in FUSE’s request_wait_answer():

* Both request is already in userspace, or it was compelled.
* Wait it out.
wait_event(req->waitq, test_bit(FR_FINISHED, &req->flags));

Okay, so we’re ready for an occasion (on this case, that userspace has replied to the FUSE flush request). However zap_pid_ns_processes()despatched a SIGKILL! SIGKILL ought to be very deadly to a course of. If we have a look at the method, we will certainly see that there’s a pending SIGKILL:

# grep Pnd /proc/1544574/standing
SigPnd: 0000000000000000
ShdPnd: 0000000000000100

Viewing course of standing this manner, you’ll be able to see 0x100 (i.e. the ninth bit is about) underneath ShdPnd, which is the sign quantity akin to SIGKILL. Pending indicators are indicators which were generated by the kernel, however haven’t but been delivered to userspace. Indicators are solely delivered at sure instances, for instance when coming into or leaving a syscall, or when ready on occasions. If the kernel is presently doing one thing on behalf of the duty, the sign could also be pending. Indicators will also be blocked by a process, in order that they’re by no means delivered. Blocked indicators will present up of their respective pending units as nicely. Nonetheless, man 7 sign says: “The indicators SIGKILL and SIGSTOP can’t be caught, blocked, or ignored.” However right here the kernel is telling us that we’ve a pending SIGKILL, aka that it’s being ignored even whereas the duty is ready!

Properly that’s bizarre. The wait code (i.e. embody/linux/wait.h) is used in every single place within the kernel: semaphores, wait queues, completions, and many others. Certainly it is aware of to search for SIGKILLs. So what does wait_event() really do? Digging via the macro expansions and wrappers, the meat of it’s:

#outline ___wait_event(wq_head, situation, state, unique, ret, cmd)           
__label__ __out;
struct wait_queue_entry __wq_entry;
lengthy __ret = ret; /* express shadow */

init_wait_entry(&__wq_entry, unique ? WQ_FLAG_EXCLUSIVE : 0);
for (;;)
lengthy __int = prepare_to_wait_event(&wq_head, &__wq_entry, state);

if (situation)

if (___wait_is_interruptible(state) && __int)
__ret = __int;
goto __out;


finish_wait(&wq_head, &__wq_entry);
__out: __ret;

So it loops without end, doing prepare_to_wait_event(), checking the situation, then checking to see if we have to interrupt. Then it does cmd, which on this case is schedule(), i.e. “do one thing else for some time”. prepare_to_wait_event() seems to be like:

lengthy prepare_to_wait_event(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state)

unsigned lengthy flags;
lengthy ret = 0;

spin_lock_irqsave(&wq_head->lock, flags);
if (signal_pending_state(state, present))
* Unique waiter should not fail if it was chosen by wakeup,
* it ought to "devour" the situation we had been ready for.
* The caller will recheck the situation and return success if
* we had been already woken up, we cannot miss the occasion as a result of
* wakeup locks/unlocks the identical wq_head->lock.
* However we have to be certain that set-condition + wakeup after that
* cannot see us, it ought to get up one other unique waiter if
* we fail.
if (list_empty(&wq_entry->entry))
if (wq_entry->flags & WQ_FLAG_EXCLUSIVE)
__add_wait_queue_entry_tail(wq_head, wq_entry);
__add_wait_queue(wq_head, wq_entry);


spin_unlock_irqrestore(&wq_head->lock, flags);

return ret;


It seems to be like the one manner we will get away of this with a non-zero exit code is that if signal_pending_state() is true. Since our name web site was simply wait_event(), we all know that state right here is TASK_UNINTERRUPTIBLE; the definition of signal_pending_state() seems to be like:

static inline int signal_pending_state(unsigned int state, struct task_struct *p)
return 0;
if (!signal_pending(p))
return 0;

return (state & TASK_INTERRUPTIBLE)

Our process isn’t interruptible, so the primary if fails. Our process ought to have a sign pending, although, proper?

static inline int signal_pending(struct task_struct *p)

* TIF_NOTIFY_SIGNAL is not actually a sign, however it requires the identical
* conduct when it comes to making certain that we get away of wait loops
* in order that notify sign callbacks may be processed.
if (unlikely(test_tsk_thread_flag(p, TIF_NOTIFY_SIGNAL)))
return 1;
return task_sigpending(p);

Because the remark notes, TIF_NOTIFY_SIGNAL isn’t related right here, regardless of its identify, however let’s have a look at task_sigpending():

static inline int task_sigpending(struct task_struct *p)

return unlikely(test_tsk_thread_flag(p,TIF_SIGPENDING));

Hmm. Looks like we must always have that flag set, proper? To determine that out, let’s have a look at how sign supply works. After we’re shutting down the pid namespace in zap_pid_ns_processes(), it does:

group_send_sig_info(SIGKILL, SEND_SIG_PRIV, process, PIDTYPE_MAX);

which ultimately will get to __send_signal_locked(), which has:

pending = (sort != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
sigaddset(&pending->sign, sig);
complete_signal(sig, t, sort);

Utilizing PIDTYPE_MAX right here as the sort is a bit of bizarre, however it roughly signifies “that is very privileged kernel stuff sending this sign, it is best to undoubtedly ship it”. There’s a little bit of unintended consequence right here, although, in that __send_signal_locked() finally ends up sending the SIGKILL to the shared set, as an alternative of the person process’s set. If we have a look at the __fatal_signal_pending() code, we see:

static inline int __fatal_signal_pending(struct task_struct *p)

return unlikely(sigismember(&p->pending.sign, SIGKILL));

But it surely seems this can be a little bit of a purple herring (although it took a while for me to grasp that).

To know what’s actually occurring right here, we have to have a look at complete_signal(), because it unconditionally provides a SIGKILL to the duty’s pending set:

sigaddset(&t->pending.sign, SIGKILL);

however why doesn’t it work? On the high of the perform we’ve:

* Now discover a thread we will get up to take the sign off the queue.
* If the primary thread desires the sign, it will get first crack.
* Most likely the least shocking to the common bear.
if (wants_signal(sig, p))
t = p;
else if ((sort == PIDTYPE_PID) || thread_group_empty(p))
* There is only one thread and it doesn't should be woken.
* It is going to dequeue unblocked indicators earlier than it runs once more.

however as Eric Biederman described, mainly each thread can deal with a SIGKILL at any time. Right here’s wants_signal():

static inline bool wants_signal(int sig, struct task_struct *p)

if (sigismember(&p->blocked, sig))
return false;

if (p->flags & PF_EXITING)
return false;

if (sig == SIGKILL)
return true;

if (task_is_stopped_or_traced(p))
return false;

return task_curr(p)

So… if a thread is already exiting (i.e. it has PF_EXITING), it doesn’t desire a sign. Think about the next sequence of occasions:

1. a process opens a FUSE file, and doesn’t shut it, then exits. Throughout that exit, the kernel dutifully calls do_exit(), which does the next:

exit_signals(tsk); /* units PF_EXITING */

2. do_exit() continues on to exit_files(tsk);, which flushes all information which are nonetheless open, ensuing within the stack hint above.

3. the pid namespace exits, and enters zap_pid_ns_processes(), sends a SIGKILL to everybody (that it expects to be deadly), after which waits for everybody to exit.

4. this kills the FUSE daemon within the pid ns so it may possibly by no means reply.

5. complete_signal() for the FUSE process that was already exiting ignores the sign, because it has PF_EXITING.

6. Impasse. With out manually aborting the FUSE connection, issues will grasp without end.

It doesn’t actually make sense to attend for flushes on this case: the duty is dying, so there’s no person to inform the return code of flush() to. It additionally seems that this bug can occur with a number of filesystems (something that calls the kernel’s wait code in flush(), i.e. mainly something that talks to one thing outdoors the native kernel).

Particular person filesystems will should be patched within the meantime, for instance the repair for FUSE is here, which was launched on April 23 in Linux 6.3.

Whereas this weblog put up addresses FUSE deadlocks, there are undoubtedly points within the nfs code and elsewhere, which we’ve not hit in manufacturing but, however virtually actually will. You can too see it as a symptom of other filesystem bugs. One thing to look out for in case you have a pid namespace that received’t exit.

That is only a small style of the number of unusual points we encounter operating containers at scale at Netflix. Our crew is hiring, so please attain out in the event you additionally love purple herrings and kernel deadlocks!