From: Chuck Lever <chuck.lever@oracle.com>
To: linux-nfs@vger.kernel.org, linux-rdma@vger.kernel.org
Subject: [PATCH RFC 08/11] svcrdma: Add svc_rdma_skip_payloads()
Date: Thu, 19 Mar 2020 11:21:05 -0400 [thread overview]
Message-ID: <20200319152105.16298.14148.stgit@klimt.1015granger.net> (raw)
In-Reply-To: <20200319150136.16298.68813.stgit@klimt.1015granger.net>
We'll need a generic mechanism for processing only the parts of an
egress RPC message that are _not_ a READ payload. This will be used
in subsequent patches.
This is a separate patch to reduce the complexity of subsequent
patches, so that the logic of this new mechanism can be separately
reviewed.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
include/linux/sunrpc/svc_rdma.h | 4 ++
net/sunrpc/xprtrdma/svc_rdma_sendto.c | 72 +++++++++++++++++++++++++++++++++
2 files changed, 76 insertions(+)
diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h
index 37e4c597dc71..93642a889535 100644
--- a/include/linux/sunrpc/svc_rdma.h
+++ b/include/linux/sunrpc/svc_rdma.h
@@ -195,6 +195,10 @@ extern struct svc_rdma_send_ctxt *
svc_rdma_send_ctxt_get(struct svcxprt_rdma *rdma);
extern void svc_rdma_send_ctxt_put(struct svcxprt_rdma *rdma,
struct svc_rdma_send_ctxt *ctxt);
+extern int svc_rdma_skip_payloads(const struct xdr_buf *xdr,
+ const struct svc_rdma_recv_ctxt *rctxt,
+ int (*actor)(const struct xdr_buf *, void *),
+ void *data);
extern int svc_rdma_send(struct svcxprt_rdma *rdma, struct ib_send_wr *wr);
extern int svc_rdma_map_reply_msg(struct svcxprt_rdma *rdma,
struct svc_rdma_send_ctxt *sctxt,
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 9fe7b0d1e335..85c91d0debb4 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -515,6 +515,78 @@ svc_rdma_encode_reply_chunk(const struct svc_rdma_recv_ctxt *rctxt,
return svc_rdma_encode_write_chunk(sctxt, &payload);
}
+static inline int
+xdr_buf_process_region(const struct xdr_buf *xdr,
+ unsigned int offset, unsigned int length,
+ int (*actor)(const struct xdr_buf *, void *),
+ void *data)
+{
+ struct xdr_buf subbuf;
+
+ if (!length)
+ return 0;
+ if (xdr_buf_subsegment(xdr, &subbuf, offset, length))
+ return -EMSGSIZE;
+ return actor(&subbuf, data);
+}
+
+/**
+ * svc_rdma_skip_payloads - Call an actor for non-payload regions of @xdr
+ * @xdr: xdr_buf to process
+ * @rctxt: Write and Reply chunks provided by client
+ * @actor: function to invoke on that region
+ * @data: pointer to arguments for @actor
+ *
+ * This mechanism must ignore not only READ payloads that were already
+ * sent via RDMA Write, but also XDR padding for those payloads that
+ * the upper layer has added.
+ *
+ * Assumptions:
+ * The xdr->len and rp_ fields are aligned to 4-byte multiples.
+ *
+ * Returns:
+ * On success, zero,
+ * %-EMSGSIZE on XDR buffer overflow, or
+ * The return value of @actor
+ */
+int svc_rdma_skip_payloads(const struct xdr_buf *xdr,
+ const struct svc_rdma_recv_ctxt *rctxt,
+ int (*actor)(const struct xdr_buf *, void *),
+ void *data)
+{
+ const unsigned int num_payloads = rctxt ? rctxt->rc_cur_payload : 0;
+ unsigned int offset, length;
+ int i, ret;
+
+ if (likely(!num_payloads))
+ return actor(xdr, data);
+
+ /* Before the first READ payload */
+ offset = 0;
+ length = rctxt->rc_read_payloads[0].rp_offset;
+ ret = xdr_buf_process_region(xdr, offset, length, actor, data);
+ if (ret < 0)
+ return ret;
+
+ /* Any middle READ payloads */
+ for (i = 0; i + 1 < num_payloads; i++) {
+ offset = xdr_align_size(length + rctxt->rc_read_payloads[i].rp_length);
+ length = rctxt->rc_read_payloads[i + 1].rp_offset - offset;
+ ret = xdr_buf_process_region(xdr, offset, length, actor, data);
+ if (ret < 0)
+ return ret;
+ }
+
+ /* After the last READ payload */
+ offset = xdr_align_size(length + rctxt->rc_read_payloads[i].rp_length);
+ length = xdr->len - offset;
+ ret = xdr_buf_process_region(xdr, offset, length, actor, data);
+ if (ret < 0)
+ return ret;
+
+ return 0;
+}
+
static int svc_rdma_dma_map_page(struct svcxprt_rdma *rdma,
struct svc_rdma_send_ctxt *ctxt,
struct page *page,
next prev parent reply other threads:[~2020-03-19 15:21 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-19 15:20 [PATCH RFC 00/11] Linux NFS server support for multiple Write chunks Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 01/11] SUNRPC: Adjust synopsis of xdr_buf_subsegment() Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 02/11] svcrdma: Clean up RDMA Write path Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 03/11] NFSD: Invoke svc_encode_read_payload in "read" NFSD encoders Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 04/11] svcrdma: Post RDMA Writes while XDR encoding replies Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 05/11] svcrdma: Clean up svc_rdma_encode_reply_chunk() Chuck Lever
2020-03-19 15:20 ` [PATCH RFC 06/11] svcrdma: Cache number of Write chunks Chuck Lever
2020-03-19 15:21 ` [PATCH RFC 07/11] svcrdma: Add a data structure to track READ payloads Chuck Lever
2020-03-19 15:21 ` Chuck Lever [this message]
2020-03-19 15:21 ` [PATCH RFC 09/11] svcrdma: Support multiple READ payloads when pulling up Chuck Lever
2020-03-19 15:21 ` [PATCH RFC 10/11] svcrdma: Support multiple READ payloads in svc_rdma_map_reply_msg() Chuck Lever
2020-03-19 15:21 ` [PATCH RFC 11/11] svcrdma: Support multiple Write chunks in svc_rdma_send_reply_chunk Chuck Lever
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200319152105.16298.14148.stgit@klimt.1015granger.net \
--to=chuck.lever@oracle.com \
--cc=linux-nfs@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).