Re: [PATCH 1/2] eventfd: Make wake counter work for single fd instead of all

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 4/9/20 3:37 AM, He Zhe wrote:
> 
> 
> On 4/8/20 4:06 AM, Jens Axboe wrote:
>> On 4/7/20 3:59 AM, zhe.he@xxxxxxxxxxxxx wrote:
>>> From: He Zhe <zhe.he@xxxxxxxxxxxxx>
>>>
>>> commit b5e683d5cab8 ("eventfd: track eventfd_signal() recursion depth")
>>> introduces a percpu counter that tracks the percpu recursion depth and
>>> warn if it greater than one, to avoid potential deadlock and stack
>>> overflow.
>>>
>>> However sometimes different eventfds may be used in parallel.
>>> Specifically, when high network load goes through kvm and vhost, working
>>> as below, it would trigger the following call trace.
>>>
>>> -  100.00%
>>>    - 66.51%
>>>         ret_from_fork
>>>         kthread
>>>       - vhost_worker
>>>          - 33.47% handle_tx_kick
>>>               handle_tx
>>>               handle_tx_copy
>>>               vhost_tx_batch.isra.0
>>>               vhost_add_used_and_signal_n
>>>               eventfd_signal
>>>          - 33.05% handle_rx_net
>>>               handle_rx
>>>               vhost_add_used_and_signal_n
>>>               eventfd_signal
>>>    - 33.49%
>>>         ioctl
>>>         entry_SYSCALL_64_after_hwframe
>>>         do_syscall_64
>>>         __x64_sys_ioctl
>>>         ksys_ioctl
>>>         do_vfs_ioctl
>>>         kvm_vcpu_ioctl
>>>         kvm_arch_vcpu_ioctl_run
>>>         vmx_handle_exit
>>>         handle_ept_misconfig
>>>         kvm_io_bus_write
>>>         __kvm_io_bus_write
>>>         eventfd_signal
>>>
>>> 001: WARNING: CPU: 1 PID: 1503 at fs/eventfd.c:73 eventfd_signal+0x85/0xa0
>>> ---- snip ----
>>> 001: Call Trace:
>>> 001:  vhost_signal+0x15e/0x1b0 [vhost]
>>> 001:  vhost_add_used_and_signal_n+0x2b/0x40 [vhost]
>>> 001:  handle_rx+0xb9/0x900 [vhost_net]
>>> 001:  handle_rx_net+0x15/0x20 [vhost_net]
>>> 001:  vhost_worker+0xbe/0x120 [vhost]
>>> 001:  kthread+0x106/0x140
>>> 001:  ? log_used.part.0+0x20/0x20 [vhost]
>>> 001:  ? kthread_park+0x90/0x90
>>> 001:  ret_from_fork+0x35/0x40
>>> 001: ---[ end trace 0000000000000003 ]---
>>>
>>> This patch moves the percpu counter into eventfd control structure and
>>> does the clean-ups, so that eventfd can still be protected from deadlock
>>> while allowing different ones to work in parallel.
>>>
>>> As to potential stack overflow, we might want to figure out a better
>>> solution in the future to warn when the stack is about to overflow so it
>>> can be better utilized, rather than break the working flow when just the
>>> second one comes.
>> This doesn't work for the infinite recursion case, the state has to be
>> global, or per thread.
> 
> Thanks, but I'm not very clear about why the counter has to be global
> or per thread.
> 
> If the recursion happens on the same eventfd, the attempt to re-grab
> the same ctx->wqh.lock would be blocked by the fd-specific counter in
> this patch.
> 
> If the recursion happens with a chain of different eventfds, that
> might lead to a stack overflow issue. The issue should be handled but
> it seems unnecessary to stop the just the second ring(when the counter
> is going to be 2) of the chain.
> 
> Specifically in the vhost case, it runs very likely with heavy network
> load which generates loads of eventfd_signal. Delaying the
> eventfd_signal to worker threads will still end up violating the
> global counter later and failing as above.
> 
> So we might want to take care of the potential overflow later,
> hopefully with a measurement that can tell us if it's about to
> overflow.

The worry is different eventfds, recursion on a single one could be
detected by keeping state in the ctx itself. And yeah, I agree that one
level isn't very deep, but wakeup chains can be deep and we can't allow
a whole lot more. I'm sure folks would be open to increasing it, if some
worst case kind of data was collected to prove it's fine to go deeper.

-- 
Jens Axboe




[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux