Re: [BUG] RCU hang with io_uring nvme polling

[Jens Axboe <axboe@kernel.dk>]

On 6/26/26 10:05 AM, Keith Busch wrote:
> On Fri, Jun 26, 2026 at 09:17:35AM -0600, Jens Axboe wrote:
>> On 6/26/26 9:09 AM, Ben Carey wrote:
>>> From a running QEMU image with the latest kernel:
>>> 1. Attach GDB to the running instance.
>>> 2. Enable io polling via sysfs (echo 1 > /sys/block/nvme0n1/queue/io_poll).
>>
>> That's not how that works at all. You need to setup poll queues on the
>> nvme driver side, using the nvme.poll_queues=XX kernel parameter, or if
>> using nvme as a module, load the module with poll_queues=XX where XX is
>> the number of poll queues. You're not doing any polled IO as-is, and the
>> above should also have dumped a dmesg message about how that does
>> absolutely nothing.
>>
>> That said, it should still work, just not doing polled IO. I'll take a
>> look sometime next week, OOO right now.
> 
> Yeah, the sysfs attribute does nothing, but Ben mentioned they had the
> correct kernel command line:
> 
>   BOOT_IMAGE=/vmlinuz-7.1.0-g3996771b8f75 root=/dev/mapper/ubuntu--vg-ubuntu--lv \
>     ro nvme.poll_queues=1 nokaslr
> 
> So they did enable polling, but the "echo" step is just confusing and
> unnecessary.
> 
> I tried out the test, and there does appear to be a problem here, so I'm
> looking into it.

Ah good catch, I missed that. Should've grepped! In general, IO should
either get polled, or if the device is misbehaving, then timeouts will
catch it. That said, haven't looked at the actual report yet, will do
so next week (unless you beat me to it...?)

-- 
Jens Axboe