Re: [Qemu-devel] [PATCH v3 41/44] nbd: Implement NBD_CMD_WRITE_ZEROES on server

[Pavel Borzenkov <pborzenkov@virtuozzo.com>]

On Fri, Apr 22, 2016 at 05:40:49PM -0600, Eric Blake wrote:
> Upstream NBD protocol recently added the ability to efficiently
> write zeroes without having to send the zeroes over the wire,
> along with a flag to control whether the client wants a hole.
> 
> Signed-off-by: Eric Blake <eblake@redhat.com>
> 
> ---
> v3: abandon NBD_CMD_CLOSE extension, rebase to use blk_pwrite_zeroes
> ---
>  include/block/nbd.h |  7 +++++--
>  nbd/server.c        | 42 ++++++++++++++++++++++++++++++++++++++++--
>  2 files changed, 45 insertions(+), 4 deletions(-)
> 
> diff --git a/include/block/nbd.h b/include/block/nbd.h
> index 05c0e48..1072d9e 100644
> --- a/include/block/nbd.h
> +++ b/include/block/nbd.h
> @@ -70,6 +70,7 @@ typedef struct nbd_reply nbd_reply;
>  #define NBD_FLAG_SEND_FUA       (1 << 3)        /* Send FUA (Force Unit Access) */
>  #define NBD_FLAG_ROTATIONAL     (1 << 4)        /* Use elevator algorithm - rotational media */
>  #define NBD_FLAG_SEND_TRIM      (1 << 5)        /* Send TRIM (discard) */
> +#define NBD_FLAG_SEND_WRITE_ZEROES (1 << 6)     /* Send WRITE_ZEROES */
> 
>  /* New-style handshake (global) flags, sent from server to client, and
>     control what will happen during handshake phase. */
> @@ -102,7 +103,8 @@ typedef struct nbd_reply nbd_reply;
>  #define NBD_INFO_DESCRIPTION    2
> 
>  /* Request flags, sent from client to server during transmission phase */
> -#define NBD_CMD_FLAG_FUA        (1 << 0)
> +#define NBD_CMD_FLAG_FUA        (1 << 0) /* 'force unit access' during write */
> +#define NBD_CMD_FLAG_NO_HOLE    (1 << 1) /* don't punch hole on zero run */
> 
>  /* Supported request types */
>  enum {
> @@ -110,7 +112,8 @@ enum {
>      NBD_CMD_WRITE = 1,
>      NBD_CMD_DISC = 2,
>      NBD_CMD_FLUSH = 3,
> -    NBD_CMD_TRIM = 4
> +    NBD_CMD_TRIM = 4,
> +    NBD_CMD_WRITE_ZEROES = 5,

It's defined to 6 by the spec.

>  };
> 
>  #define NBD_DEFAULT_PORT	10809
> diff --git a/nbd/server.c b/nbd/server.c
> index 1edb5f3..563afb2 100644
> --- a/nbd/server.c
> +++ b/nbd/server.c
> @@ -689,7 +689,8 @@ static coroutine_fn int nbd_negotiate(NBDClientNewData *data)
>      char buf[8 + 8 + 8 + 128];
>      int rc;
>      const uint16_t myflags = (NBD_FLAG_HAS_FLAGS | NBD_FLAG_SEND_TRIM |
> -                              NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA);
> +                              NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA |
> +                              NBD_FLAG_SEND_WRITE_ZEROES);
>      bool oldStyle;
>      size_t len;
> 
> @@ -1199,11 +1200,17 @@ static ssize_t nbd_co_receive_request(NBDRequest *req,
>          rc = -EINVAL;
>          goto out;
>      }
> -    if (request->flags & ~NBD_CMD_FLAG_FUA) {
> +    if (request->flags & ~(NBD_CMD_FLAG_FUA | NBD_CMD_FLAG_NO_HOLE)) {
>          LOG("unsupported flags (got 0x%x)", request->flags);
>          rc = -EINVAL;
>          goto out;
>      }
> +    if (request->type != NBD_CMD_WRITE_ZEROES &&
> +        (request->flags & NBD_CMD_FLAG_NO_HOLE)) {
> +        LOG("unexpected flags (got 0x%x)", request->flags);
> +        rc = -EINVAL;
> +        goto out;
> +    }
> 
>      rc = 0;
> 
> @@ -1308,6 +1315,37 @@ static void nbd_trip(void *opaque)
>          }
>          break;
> 
> +    case NBD_CMD_WRITE_ZEROES:
> +        TRACE("Request type is WRITE_ZEROES");
> +
> +        if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
> +            TRACE("Server is read-only, return error");
> +            reply.error = EROFS;
> +            goto error_reply;
> +        }
> +
> +        TRACE("Writing to device");
> +
> +        flags = 0;
> +        if (request.flags & NBD_CMD_FLAG_FUA) {
> +            flags |= BDRV_REQ_FUA;
> +        }
> +        if (!(request.flags & NBD_CMD_FLAG_NO_HOLE)) {
> +            flags |= BDRV_REQ_MAY_UNMAP;
> +        }
> +        ret = blk_pwrite_zeroes(exp->blk, request.from + exp->dev_offset,
> +                                request.len, flags);
> +        if (ret < 0) {
> +            LOG("writing to file failed");
> +            reply.error = -ret;
> +            goto error_reply;
> +        }
> +
> +        if (nbd_co_send_reply(req, &reply, 0) < 0) {
> +            goto out;
> +        }
> +        break;
> +
>      case NBD_CMD_DISC:
>          /* unreachable, thanks to special case in nbd_co_receive_request() */
>          abort();
> -- 
> 2.5.5
> 
>