Re: [PATCH 2/2] selftests: ublk: make the fault_inject delay a pure timeout

[Ming Lei]

On Thu, Oct 08, 2026 at 10:45:40PM +0800, Qiliang Yuan wrote:
> ublk_fault_inject_queue_io() queues the delay of each I/O as an
> IORING_OP_TIMEOUT with a completion count of 1. Such a timeout also
> completes as soon as any other CQE is posted on the ring.
> 
> With more than one I/O in flight, the completion of one I/O ends the
> delay of the others, and ublk_fault_inject_tgt_io_done() reports every
> early completion as "unexpected cqe res 0".
> 
> Pass a count of 0 so that only the expiry of the timer completes the
> timeout.
> 
> fio 4k random reads for 10 s at queue depth 4 on a fault_inject device
> with --delay_us 1000000:
> 
>                             before      after   expected
>   I/Os                     2263170         40         40
>   mean latency             16.3 us    1.000 s        1 s
>   "unexpected cqe res 0"   2263169          0          0
> 
> Fixes: 81586652bb1f ("selftests: ublk: add generic_06 for covering fault inject")
> Signed-off-by: Qiliang Yuan <odys.yuan@xxxxxxxxx>

Reviewed-by: Ming Lei <tom.leiming@xxxxxxxxx>

Thanks,
Ming