On Mon, Aug 14, 2023 at 01:36:54PM -0400, Jeff Layton wrote:
> svc_tcp_sendmsg used to factor in the xdr->page_base when sending pages,
> but 5df5dd03a8f7 dropped that part of the handling. Fix it by setting
> the bv_offset of the first bvec.
>
> Fixes: 5df5dd03a8f7 ("sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage")
> Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
LGTM. However, nfsd-fixes does not have 5df5dd03a8f7 because the
previous nfsd-next was merged into v6.5 before David's
MSG_SPLICE_PAGES work was merged.
Unless someone has a better suggestion, I'll rebase nfsd-fixes to
v6.5-r6 and apply this fix to send to Linus.
> ---
> Changes in v2:
> - limit the change to just svc_tcp_sendmsg
> ---
> net/sunrpc/svcsock.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> index e43f26382411..2eb8df44f894 100644
> --- a/net/sunrpc/svcsock.c
> +++ b/net/sunrpc/svcsock.c
> @@ -1244,6 +1244,9 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
> if (ret != head->iov_len)
> goto out;
>
> + if (xdr_buf_pagecount(xdr))
> + xdr->bvec[0].bv_offset = offset_in_page(xdr->page_base);
> +
> msg.msg_flags = MSG_SPLICE_PAGES;
> iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
> xdr_buf_pagecount(xdr), xdr->page_len);
>
> ---
> base-commit: 2ccdd1b13c591d306f0401d98dedc4bdcd02b421
> change-id: 20230814-sendpage-b04874eed249
>
> Best regards,
> --
> Jeff Layton <jlayton@xxxxxxxxxx>
>
--
Chuck Lever
