On 2/7/24 9:37 AM, Jens Axboe wrote: > On 2/7/24 12:19 AM, Guangwu Zhang wrote: >> HI, >> Found the kernel issue with linux-block/for-next branch. >> kernel repo https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git >> >> Reproducer : >> 1. offline_online_cpu_in_bg >> 2. some_io_in_bg > > I don't know what these two things are. Would be nice with an actual > reproducer. I can trivially write something that offline and onlines CPU > in the background, and I did, but I cannot reproduce this issue. Ditto > on "some_io_in_bg", what does that mean?? I managed to reproduce this, it takes: - Using polled IO, but you didn't set hipri in the given command line you sent. What you sent cannot have been what was run, the trace as well shows that polled IO is being done. - NVMe device with _no_ poll queues defined. Anyway, the current branch should work fine. In the future, please do include the full setup description and exactly what was run. This one was easy to find, but may not be so lucky in the future. -- Jens Axboe
