Re: [PATCH v2 1/2] fs,block: Introduce RWF_ZONE_APPEND and handling in direct IO path

[Damien Le Moal]

On 2020/06/30 3:35, Kanchan Joshi wrote:
> On Fri, Jun 26, 2020 at 02:50:20AM +0000, Damien Le Moal wrote:
>> On 2020/06/26 2:18, Kanchan Joshi wrote:
>>> Introduce RWF_ZONE_APPEND flag to represent zone-append. User-space
>>> sends this with write. Add IOCB_ZONE_APPEND which is set in
>>> kiocb->ki_flags on receiving RWF_ZONE_APPEND.
>>> Make direct IO submission path use IOCB_ZONE_APPEND to send bio with
>>> append op. Direct IO completion returns zone-relative offset, in sector
>>> unit, to upper layer using kiocb->ki_complete interface.
>>> Report error if zone-append is requested on regular file or on sync
>>> kiocb (i.e. one without ki_complete).
>>>
>>> Signed-off-by: Kanchan Joshi <joshi.k@xxxxxxxxxxx>
>>> Signed-off-by: SelvaKumar S <selvakuma.s1@xxxxxxxxxxx>
>>> Signed-off-by: Arnav Dawn <a.dawn@xxxxxxxxxxx>
>>> Signed-off-by: Nitesh Shetty <nj.shetty@xxxxxxxxxxx>
>>> Signed-off-by: Javier Gonzalez <javier.gonz@xxxxxxxxxxx>
>>> ---
>>>  fs/block_dev.c          | 28 ++++++++++++++++++++++++----
>>>  include/linux/fs.h      |  9 +++++++++
>>>  include/uapi/linux/fs.h |  5 ++++-
>>>  3 files changed, 37 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/fs/block_dev.c b/fs/block_dev.c
>>> index 47860e5..5180268 100644
>>> --- a/fs/block_dev.c
>>> +++ b/fs/block_dev.c
>>> @@ -185,6 +185,10 @@ static unsigned int dio_bio_write_op(struct kiocb *iocb)
>>>  	/* avoid the need for a I/O completion work item */
>>>  	if (iocb->ki_flags & IOCB_DSYNC)
>>>  		op |= REQ_FUA;
>>> +
>>> +	if (iocb->ki_flags & IOCB_ZONE_APPEND)
>>> +		op |= REQ_OP_ZONE_APPEND;
>>
>> This is wrong. REQ_OP_WRITE is already set in the declaration of "op". How can
>> this work ?
> REQ_OP_ZONE_APPEND will override the REQ_WRITE op, while previously set op
> flags (REQ_FUA etc.) will be retained. But yes, this can be made to look
> cleaner.
> V3 will include the other changes you pointed out. Thanks for the review.
> 

REQ_OP_WRITE and REQ_OP_ZONE_APPEND are different bits, so there is no
"override". A well formed BIO bi_opf is one op+flags. Specifying multiple OP
codes does not make sense.

-- 
Damien Le Moal
Western Digital Research