Re: [PATCH 3/3] io_uring: add support for zone-append

[Damien Le Moal]

On 2020/06/18 2:27, Kanchan Joshi wrote:
> From: Selvakumar S <selvakuma.s1@xxxxxxxxxxx>
> 
> Introduce three new opcodes for zone-append -
> 
>    IORING_OP_ZONE_APPEND     : non-vectord, similiar to IORING_OP_WRITE
>    IORING_OP_ZONE_APPENDV    : vectored, similar to IORING_OP_WRITEV
>    IORING_OP_ZONE_APPEND_FIXED : append using fixed-buffers
> 
> Repurpose cqe->flags to return zone-relative offset.
> 
> 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                 | 72 +++++++++++++++++++++++++++++++++++++++++--
>  include/uapi/linux/io_uring.h |  8 ++++-
>  2 files changed, 77 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 155f3d8..c14c873 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -649,6 +649,10 @@ struct io_kiocb {
>  	unsigned long		fsize;
>  	u64			user_data;
>  	u32			result;
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	/* zone-relative offset for append, in bytes */
> +	u32			append_offset;

this can overflow. u64 is needed.

> +#endif
>  	u32			sequence;
>  
>  	struct list_head	link_list;
> @@ -875,6 +879,26 @@ static const struct io_op_def io_op_defs[] = {
>  		.hash_reg_file		= 1,
>  		.unbound_nonreg_file	= 1,
>  	},
> +	[IORING_OP_ZONE_APPEND] = {
> +		.needs_mm               = 1,
> +		.needs_file             = 1,
> +		.unbound_nonreg_file    = 1,
> +		.pollout		= 1,
> +	},
> +	[IORING_OP_ZONE_APPENDV] = {
> +	       .async_ctx              = 1,
> +	       .needs_mm               = 1,
> +	       .needs_file             = 1,
> +	       .hash_reg_file          = 1,
> +	       .unbound_nonreg_file    = 1,
> +	       .pollout			= 1,
> +	},
> +	[IORING_OP_ZONE_APPEND_FIXED] = {
> +	       .needs_file             = 1,
> +	       .hash_reg_file          = 1,
> +	       .unbound_nonreg_file    = 1,
> +	       .pollout			= 1,
> +	},
>  };
>  
>  static void io_wq_submit_work(struct io_wq_work **workptr);
> @@ -1285,7 +1309,16 @@ static void __io_cqring_fill_event(struct io_kiocb *req, long res, long cflags)
>  	if (likely(cqe)) {
>  		WRITE_ONCE(cqe->user_data, req->user_data);
>  		WRITE_ONCE(cqe->res, res);
> +#ifdef CONFIG_BLK_DEV_ZONED
> +		if (req->opcode == IORING_OP_ZONE_APPEND ||
> +				req->opcode == IORING_OP_ZONE_APPENDV ||
> +				req->opcode == IORING_OP_ZONE_APPEND_FIXED)
> +			WRITE_ONCE(cqe->res2, req->append_offset);
> +		else
> +			WRITE_ONCE(cqe->flags, cflags);
> +#else
>  		WRITE_ONCE(cqe->flags, cflags);
> +#endif
>  	} else if (ctx->cq_overflow_flushed) {
>  		WRITE_ONCE(ctx->rings->cq_overflow,
>  				atomic_inc_return(&ctx->cached_cq_overflow));
> @@ -1961,6 +1994,9 @@ static void io_complete_rw_common(struct kiocb *kiocb, long res)
>  static void io_complete_rw(struct kiocb *kiocb, long res, long res2)
>  {
>  	struct io_kiocb *req = container_of(kiocb, struct io_kiocb, rw.kiocb);
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	req->append_offset = (u32)res2;
> +#endif
>  
>  	io_complete_rw_common(kiocb, res);
>  	io_put_req(req);
> @@ -1976,6 +2012,9 @@ static void io_complete_rw_iopoll(struct kiocb *kiocb, long res, long res2)
>  	if (res != req->result)
>  		req_set_fail_links(req);
>  	req->result = res;
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	req->append_offset = (u32)res2;
> +#endif
>  	if (res != -EAGAIN)
>  		WRITE_ONCE(req->iopoll_completed, 1);
>  }
> @@ -2408,7 +2447,8 @@ static ssize_t io_import_iovec(int rw, struct io_kiocb *req,
>  	u8 opcode;
>  
>  	opcode = req->opcode;
> -	if (opcode == IORING_OP_READ_FIXED || opcode == IORING_OP_WRITE_FIXED) {
> +	if (opcode == IORING_OP_READ_FIXED || opcode == IORING_OP_WRITE_FIXED ||
> +			opcode == IORING_OP_ZONE_APPEND_FIXED) {
>  		*iovec = NULL;
>  		return io_import_fixed(req, rw, iter);
>  	}
> @@ -2417,7 +2457,8 @@ static ssize_t io_import_iovec(int rw, struct io_kiocb *req,
>  	if (req->buf_index && !(req->flags & REQ_F_BUFFER_SELECT))
>  		return -EINVAL;
>  
> -	if (opcode == IORING_OP_READ || opcode == IORING_OP_WRITE) {
> +	if (opcode == IORING_OP_READ || opcode == IORING_OP_WRITE ||
> +			opcode == IORING_OP_ZONE_APPEND) {
>  		if (req->flags & REQ_F_BUFFER_SELECT) {
>  			buf = io_rw_buffer_select(req, &sqe_len, needs_lock);
>  			if (IS_ERR(buf)) {
> @@ -2704,6 +2745,9 @@ static int io_write(struct io_kiocb *req, bool force_nonblock)
>  		req->rw.kiocb.ki_flags &= ~IOCB_NOWAIT;
>  
>  	req->result = 0;
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	req->append_offset = 0;
> +#endif
>  	io_size = ret;
>  	if (req->flags & REQ_F_LINK_HEAD)
>  		req->result = io_size;
> @@ -2738,6 +2782,13 @@ static int io_write(struct io_kiocb *req, bool force_nonblock)
>  			__sb_writers_release(file_inode(req->file)->i_sb,
>  						SB_FREEZE_WRITE);
>  		}
> +#ifdef CONFIG_BLK_DEV_ZONED
> +		if (req->opcode == IORING_OP_ZONE_APPEND ||
> +				req->opcode == IORING_OP_ZONE_APPENDV ||
> +				req->opcode == IORING_OP_ZONE_APPEND_FIXED)
> +			kiocb->ki_flags |= IOCB_ZONE_APPEND;
> +#endif
> +
>  		kiocb->ki_flags |= IOCB_WRITE;
>  
>  		if (!force_nonblock)
> @@ -4906,6 +4957,12 @@ static int io_req_defer_prep(struct io_kiocb *req,
>  	case IORING_OP_WRITEV:
>  	case IORING_OP_WRITE_FIXED:
>  	case IORING_OP_WRITE:
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	fallthrough;
> +	case IORING_OP_ZONE_APPEND:
> +	case IORING_OP_ZONE_APPENDV:
> +	case IORING_OP_ZONE_APPEND_FIXED:
> +#endif
>  		ret = io_write_prep(req, sqe, true);
>  		break;
>  	case IORING_OP_POLL_ADD:
> @@ -5038,6 +5095,12 @@ static void io_cleanup_req(struct io_kiocb *req)
>  	case IORING_OP_WRITEV:
>  	case IORING_OP_WRITE_FIXED:
>  	case IORING_OP_WRITE:
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	fallthrough;
> +	case IORING_OP_ZONE_APPEND:
> +	case IORING_OP_ZONE_APPENDV:
> +	case IORING_OP_ZONE_APPEND_FIXED:
> +#endif
>  		if (io->rw.iov != io->rw.fast_iov)
>  			kfree(io->rw.iov);
>  		break;
> @@ -5086,6 +5149,11 @@ static int io_issue_sqe(struct io_kiocb *req, const struct io_uring_sqe *sqe,
>  		}
>  		ret = io_read(req, force_nonblock);
>  		break;
> +#ifdef CONFIG_BLK_DEV_ZONED
> +	case IORING_OP_ZONE_APPEND:
> +	case IORING_OP_ZONE_APPENDV:
> +	case IORING_OP_ZONE_APPEND_FIXED:
> +#endif
>  	case IORING_OP_WRITEV:
>  	case IORING_OP_WRITE_FIXED:
>  	case IORING_OP_WRITE:
> diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
> index 92c2269..6c8e932 100644
> --- a/include/uapi/linux/io_uring.h
> +++ b/include/uapi/linux/io_uring.h
> @@ -130,6 +130,9 @@ enum {
>  	IORING_OP_PROVIDE_BUFFERS,
>  	IORING_OP_REMOVE_BUFFERS,
>  	IORING_OP_TEE,
> +	IORING_OP_ZONE_APPEND,
> +	IORING_OP_ZONE_APPENDV,
> +	IORING_OP_ZONE_APPEND_FIXED,
>  
>  	/* this goes last, obviously */
>  	IORING_OP_LAST,
> @@ -157,7 +160,10 @@ enum {
>  struct io_uring_cqe {
>  	__u64	user_data;	/* sqe->data submission passed back */
>  	__s32	res;		/* result code for this event */
> -	__u32	flags;
> +	union {
> +		__u32	res2; /* res2 like aio, currently used for zone-append */
> +		__u32	flags;
> +	};
>  };
>  
>  /*
> 


-- 
Damien Le Moal
Western Digital Research