Jens,
On 2020/04/17 23:32, Jens Axboe wrote:
> On 4/16/20 5:30 AM, Shin'ichiro Kawasaki wrote:
>> Commit fb0259fb ("zbd: Ensure first I/O is write for random read/write to
>> sequential zones") introduced a step to change direction of io_u from
>> read to write when that is the first I/O of the random read/write
>> workload to zoned block devices. However, such direction adjustment
>> results in inconsistent I/O length when read block size and write block
>> size are different.
>>
>> To avoid the inconsistency between I/O direction and I/O length,
>> adjust the I/O direction before the I/O length is set. Move the step
>> from zbd_adjust_block() to set_rw_ddir(). To minimize changes in
>> set_rw_ddir(), introduce zbd_adjust_ddir() helper function.
>
> I'm not a huge fan of this zbd sprinkling everywhere, but that's not the
> fault of your fix. But I would really prefer if zbd got disentangled
> from the core/hot paths completely, maybe by adding a io engine fn for
> whatever it needs to do. At least that reduces the hot path overhead to
> check a pointer and calling that function instead of having various
> checks here and there.
Understood. We will work on that.
>
> Really not great to have a somewhat esoteric feature like zbd support
> add _any_ overhead to the hot path at all.
--
Damien Le Moal
Western Digital Research
