Hi Jeff,
On 07/23/2014 09:25 PM, Jeff Moyer wrote:
> Gu Zheng <guz.f...@cn.fujitsu.com> writes:
>
>> Previously, we only offer a single iovec to handle all the read/write cases,
>> so
>> the PREADV/PWRITEV request always need to alloc more iovec buffer when
>> copying
>> user vectors.
>> If we use a tmp iovec array rather than the single one, some small
>> PREADV/PWRITEV
>> workloads(vector size small than the tmp buffer) will not need to alloc more
>> iovec buffer when copying user vectors.
>
> Hi, Gu,
>
> This still doesn't explain why you decided to look into this.
The comment is clear, just want to avoid some needless memory allocation in
the io submit path.
> Did you
> notice a performance issue in this path? Do you have benchmarks that
> show some speedup due to this change?
Just some common tests based on fio, it gains a slight improvement(~3%) when the
iodepth in [5,6,7] than before. I did not paste these info here, because I think
other guys(especially the guys have production environment) who are interested
in this can give us more meaningful feedback.
Thanks,
Gu
>
> Thanks,
> Jeff
>
>>
>> Reviewed-by: Jeff Moyer <jmo...@redhat.com>
>> Signed-off-by: Gu Zheng <guz.f...@cn.fujitsu.com>
>> ---
>> fs/aio.c | 10 +++++-----
>> 1 files changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/fs/aio.c b/fs/aio.c
>> index f1fede2..df3491a 100644
>> --- a/fs/aio.c
>> +++ b/fs/aio.c
>> @@ -1267,12 +1267,12 @@ static ssize_t aio_setup_vectored_rw(struct kiocb
>> *kiocb,
>> if (compat)
>> ret = compat_rw_copy_check_uvector(rw,
>> (struct compat_iovec __user *)buf,
>> - *nr_segs, 1, *iovec, iovec);
>> + *nr_segs, UIO_FASTIOV, *iovec, iovec);
>> else
>> #endif
>> ret = rw_copy_check_uvector(rw,
>> (struct iovec __user *)buf,
>> - *nr_segs, 1, *iovec, iovec);
>> + *nr_segs, UIO_FASTIOV, *iovec, iovec);
>> if (ret < 0)
>> return ret;
>>
>> @@ -1309,7 +1309,7 @@ static ssize_t aio_run_iocb(struct kiocb *req,
>> unsigned opcode,
>> fmode_t mode;
>> aio_rw_op *rw_op;
>> rw_iter_op *iter_op;
>> - struct iovec inline_vec, *iovec = &inline_vec;
>> + struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
>> struct iov_iter iter;
>>
>> switch (opcode) {
>> @@ -1344,7 +1344,7 @@ rw_common:
>> if (!ret)
>> ret = rw_verify_area(rw, file, &req->ki_pos,
>> req->ki_nbytes);
>> if (ret < 0) {
>> - if (iovec != &inline_vec)
>> + if (iovec != inline_vecs)
>> kfree(iovec);
>> return ret;
>> }
>> @@ -1391,7 +1391,7 @@ rw_common:
>> return -EINVAL;
>> }
>>
>> - if (iovec != &inline_vec)
>> + if (iovec != inline_vecs)
>> kfree(iovec);
>>
>> if (ret != -EIOCBQUEUED) {
> .
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
