[v5,0/5] fsstress,fsx: add io_uring test and do some fix

Message

Zorro Lang Sept. 11, 2020, 7:15 a.m. UTC

This patchset tries to add new IO_URING test into fsstress [1/5] and fsx
[4/5 and 5/5]. And then do some changes and bug fix by the way [2/5 and 3/5].

fsstress and fsx are important tools in xfstests to do filesystem I/Os test,
lots of test cases use it. So add IO_URING operation into fsstress and fsx
will help to cover IO_URING test from fs side.

I'm not an IO_URING expert, so cc io-uring@ list, please feel free to
tell me if you find something wrong or have any suggestions to improve
the test.

V2 did below changes:
1) 1/4 change the definition of URING_ENTRIES to 1
2) 2/4 change the difinition of AIO_ENTRIES to 1, undo an unrelated changed line
3) 4/4 turn to use io_uring_prep_readv/io_uring_prep_writev, due to old
       liburing(0.2-2) doesn't support io_uring_prep_read/io_uring_prep_write.

V3 changed io_uring_submit(&ring) to io_uring_submit_and_wait(&ring, 1). I'm
not sure if this's the real mean of Jens Axboe's review point, please check.
  https://marc.info/?l=fstests&m=159811932808057&w=2

V4 did below changes:
1) 1/5 change the "goto" related code of do_uring_rw()
2) 3/5 similar change as above
3) 4/5 new patch, separated from original 4/4 patch
3) 5/5 change #elif to #else
4) 5/5 change __uring_rw to uring_rw.
5) 5/5 change the loop logic in uring_rw().

V5 did below changes:
1) 1/5 turn to use (fd != -1)
2) 3/5 turn to use (fd != -1)
3) 5/5 change res to ret, change res2 to res.

Thanks,
Zorro

Comments

Zorro Lang Sept. 21, 2020, 3:01 a.m. UTC | #1

On Fri, Sep 11, 2020 at 03:15:50PM +0800, Zorro Lang wrote:
> This patchset tries to add new IO_URING test into fsstress [1/5] and fsx
> [4/5 and 5/5]. And then do some changes and bug fix by the way [2/5 and 3/5].
> 
> fsstress and fsx are important tools in xfstests to do filesystem I/Os test,
> lots of test cases use it. So add IO_URING operation into fsstress and fsx
> will help to cover IO_URING test from fs side.
> 
> I'm not an IO_URING expert, so cc io-uring@ list, please feel free to
> tell me if you find something wrong or have any suggestions to improve
> the test.

Hi Eryu,

We're waiting for this patchset get merged. It's been reviewed-by Brian, do you
still need more ACK?

Thanks,
Zorro

> 
> V2 did below changes:
> 1) 1/4 change the definition of URING_ENTRIES to 1
> 2) 2/4 change the difinition of AIO_ENTRIES to 1, undo an unrelated changed line
> 3) 4/4 turn to use io_uring_prep_readv/io_uring_prep_writev, due to old
>        liburing(0.2-2) doesn't support io_uring_prep_read/io_uring_prep_write.
> 
> V3 changed io_uring_submit(&ring) to io_uring_submit_and_wait(&ring, 1). I'm
> not sure if this's the real mean of Jens Axboe's review point, please check.
>   https://marc.info/?l=fstests&m=159811932808057&w=2
> 
> V4 did below changes:
> 1) 1/5 change the "goto" related code of do_uring_rw()
> 2) 3/5 similar change as above
> 3) 4/5 new patch, separated from original 4/4 patch
> 3) 5/5 change #elif to #else
> 4) 5/5 change __uring_rw to uring_rw.
> 5) 5/5 change the loop logic in uring_rw().
> 
> V5 did below changes:
> 1) 1/5 turn to use (fd != -1)
> 2) 3/5 turn to use (fd != -1)
> 3) 5/5 change res to ret, change res2 to res.
> 
> Thanks,
> Zorro
> 
> 
>