Re: [PATCH v2 2/2] io_uring: add support for zone-append

[Pavel Begunkov]

On 25/06/2020 20:15, Kanchan Joshi wrote:
> From: Selvakumar S <selvakuma.s1@xxxxxxxxxxx>
> 
> For zone-append, block-layer will return zone-relative offset via ret2
> of ki_complete interface. Make changes to collect it, and send to
> user-space using ceq->flags.
> Detect and report early error if zone-append is requested with
> fixed-buffers.
> 
> Signed-off-by: Selvakumar S <selvakuma.s1@xxxxxxxxxxx>
> Signed-off-by: Kanchan Joshi <joshi.k@xxxxxxxxxxx>
> Signed-off-by: Nitesh Shetty <nj.shetty@xxxxxxxxxxx>
> Signed-off-by: Javier Gonzalez <javier.gonz@xxxxxxxxxxx>
> ---
>  fs/io_uring.c | 32 ++++++++++++++++++++++++++++++--
>  1 file changed, 30 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 155f3d8..31a9da58 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -402,6 +402,8 @@ struct io_rw {
>  	struct kiocb			kiocb;
>  	u64				addr;
>  	u64				len;
> +	/* zone-relative offset for append, in sectors */
> +	u32			append_offset;
>  };
>  
>  struct io_connect {
> @@ -541,6 +543,7 @@ enum {
>  	REQ_F_NO_FILE_TABLE_BIT,
>  	REQ_F_QUEUE_TIMEOUT_BIT,
>  	REQ_F_WORK_INITIALIZED_BIT,
> +	REQ_F_ZONE_APPEND_BIT,
>  
>  	/* not a real bit, just to check we're not overflowing the space */
>  	__REQ_F_LAST_BIT,
> @@ -598,6 +601,8 @@ enum {
>  	REQ_F_QUEUE_TIMEOUT	= BIT(REQ_F_QUEUE_TIMEOUT_BIT),
>  	/* io_wq_work is initialized */
>  	REQ_F_WORK_INITIALIZED	= BIT(REQ_F_WORK_INITIALIZED_BIT),
> +	/* to return zone relative offset for zone append*/
> +	REQ_F_ZONE_APPEND	= BIT(REQ_F_ZONE_APPEND_BIT),

Do we need a new flag? We can check for IOCB_ZONE_APPEND, flags are always
close by in req->rw.kiocb.ki_flags. May require to be careful about not
setting it for read, so not screwing buf select.

>  };
>  
>  struct async_poll {
> @@ -1745,6 +1750,8 @@ static void io_iopoll_complete(struct io_ring_ctx *ctx, unsigned int *nr_events,
>  
>  		if (req->flags & REQ_F_BUFFER_SELECTED)
>  			cflags = io_put_kbuf(req);
> +		if (req->flags & REQ_F_ZONE_APPEND)
> +			cflags = req->rw.append_offset;
>  
>  		__io_cqring_fill_event(req, req->result, cflags);
>  		(*nr_events)++;
> @@ -1943,7 +1950,7 @@ static inline void req_set_fail_links(struct io_kiocb *req)
>  		req->flags |= REQ_F_FAIL_LINK;
>  }
>  
> -static void io_complete_rw_common(struct kiocb *kiocb, long res)
> +static void io_complete_rw_common(struct kiocb *kiocb, long res, long res2)
>  {
>  	struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw.kiocb);
>  	int cflags = 0;
> @@ -1953,8 +1960,14 @@ static void io_complete_rw_common(struct kiocb *kiocb, long res)
>  
>  	if (res != req->result)
>  		req_set_fail_links(req);
> +
>  	if (req->flags & REQ_F_BUFFER_SELECTED)
>  		cflags = io_put_kbuf(req);
> +
> +	/* use cflags to return zone append completion result */
> +	if (req->flags & REQ_F_ZONE_APPEND)
> +		cflags = res2;
> +
>  	__io_cqring_add_event(req, res, cflags);
>  }
>  
> @@ -1962,7 +1975,7 @@ static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
>  {
>  	struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw.kiocb);
>  
> -	io_complete_rw_common(kiocb, res);
> +	io_complete_rw_common(kiocb, res, res2);
>  	io_put_req(req);
>  }
>  
> @@ -1975,6 +1988,9 @@ static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
>  
>  	if (res != req->result)
>  		req_set_fail_links(req);
> +	if (req->flags & REQ_F_ZONE_APPEND)
> +		req->rw.append_offset = res2;
> +
>  	req->result = res;
>  	if (res != -EAGAIN)
>  		WRITE_ONCE(req->iopoll_completed, 1);
> @@ -2127,6 +2143,9 @@ static int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe,
>  	if (kiocb->ki_flags & IOCB_NOWAIT)
>  		req->flags |= REQ_F_NOWAIT;
>  
> +	if (kiocb->ki_flags & IOCB_ZONE_APPEND)
> +		req->flags |= REQ_F_ZONE_APPEND;
> +
>  	if (force_nonblock)
>  		kiocb->ki_flags |= IOCB_NOWAIT;
>  
> @@ -2409,6 +2428,14 @@ static ssize_t io_import_iovec(int rw, struct io_kiocb *req,
>  
>  	opcode = req->opcode;
>  	if (opcode == IORING_OP_READ_FIXED || opcode == IORING_OP_WRITE_FIXED) {
> +		/*
> +		 * fixed-buffers not supported for zone-append.
> +		 * This check can be removed when block-layer starts
> +		 * supporting append with iov_iter of bvec type
> +		 */
> +		if (req->flags == REQ_F_ZONE_APPEND)

s/==/&/

> +			return -EINVAL;
> +
>  		*iovec = NULL;
>  		return io_import_fixed(req, rw, iter);
>  	}
> @@ -2704,6 +2731,7 @@ static int io_write(struct io_kiocb *req, bool force_nonblock)
>  		req->rw.kiocb.ki_flags &= ~IOCB_NOWAIT;
>  
>  	req->result = 0;
> +

Extra \n

>  	io_size = ret;
>  	if (req->flags & REQ_F_LINK_HEAD)
>  		req->result = io_size;
> 

-- 
Pavel Begunkov