From 10b9d99a3dbbf5e9af838d1887a1047875dcafd9 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Sat, 18 Apr 2020 18:30:42 -0400
Subject: SUNRPC: Augment server-side rpcgss tracepoints

Add similar tracepoints to those that were recently added on the
client side to track failures in the integ and priv unwrap paths.

And, let's collect the seqno-specific tracepoints together with a
common naming convention.

Regarding the gss_check_seq_num() changes: everywhere else treats
the GSS sequence number as an unsigned 32-bit integer. As far back
as 2.6.12, I couldn't find a compelling reason to do things
differently here. As a defensive change it's better to eliminate
needless implicit sign conversions.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/auth_gss/svcauth_gss.c | 117 +++++++++++++++++++++++++-------------
 net/sunrpc/auth_gss/trace.c       |   3 +
 2 files changed, 82 insertions(+), 38 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index 46027d0c903f..7d83f54aaaa6 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -332,7 +332,7 @@ static struct rsi *rsi_update(struct cache_detail *cd, struct rsi *new, struct r
 
 struct gss_svc_seq_data {
 	/* highest seq number seen so far: */
-	int			sd_max;
+	u32			sd_max;
 	/* for i such that sd_max-GSS_SEQ_WIN < i <= sd_max, the i-th bit of
 	 * sd_win is nonzero iff sequence number i has been seen already: */
 	unsigned long		sd_win[GSS_SEQ_WIN/BITS_PER_LONG];
@@ -613,16 +613,29 @@ gss_svc_searchbyctx(struct cache_detail *cd, struct xdr_netobj *handle)
 	return found;
 }
 
-/* Implements sequence number algorithm as specified in RFC 2203. */
-static int
-gss_check_seq_num(struct rsc *rsci, int seq_num)
+/**
+ * gss_check_seq_num - GSS sequence number window check
+ * @rqstp: RPC Call to use when reporting errors
+ * @rsci: cached GSS context state (updated on return)
+ * @seq_num: sequence number to check
+ *
+ * Implements sequence number algorithm as specified in
+ * RFC 2203, Section 5.3.3.1. "Context Management".
+ *
+ * Return values:
+ *   %true: @rqstp's GSS sequence number is inside the window
+ *   %false: @rqstp's GSS sequence number is outside the window
+ */
+static bool gss_check_seq_num(const struct svc_rqst *rqstp, struct rsc *rsci,
+			      u32 seq_num)
 {
 	struct gss_svc_seq_data *sd = &rsci->seqdata;
+	bool result = false;
 
 	spin_lock(&sd->sd_lock);
 	if (seq_num > sd->sd_max) {
 		if (seq_num >= sd->sd_max + GSS_SEQ_WIN) {
-			memset(sd->sd_win,0,sizeof(sd->sd_win));
+			memset(sd->sd_win, 0, sizeof(sd->sd_win));
 			sd->sd_max = seq_num;
 		} else while (sd->sd_max < seq_num) {
 			sd->sd_max++;
@@ -631,17 +644,25 @@ gss_check_seq_num(struct rsc *rsci, int seq_num)
 		__set_bit(seq_num % GSS_SEQ_WIN, sd->sd_win);
 		goto ok;
 	} else if (seq_num <= sd->sd_max - GSS_SEQ_WIN) {
-		goto drop;
+		goto toolow;
 	}
-	/* sd_max - GSS_SEQ_WIN < seq_num <= sd_max */
 	if (__test_and_set_bit(seq_num % GSS_SEQ_WIN, sd->sd_win))
-		goto drop;
+		goto alreadyseen;
+
 ok:
+	result = true;
+out:
 	spin_unlock(&sd->sd_lock);
-	return 1;
-drop:
-	spin_unlock(&sd->sd_lock);
-	return 0;
+	return result;
+
+toolow:
+	trace_rpcgss_svc_seqno_low(rqstp, seq_num,
+				   sd->sd_max - GSS_SEQ_WIN,
+				   sd->sd_max);
+	goto out;
+alreadyseen:
+	trace_rpcgss_svc_seqno_seen(rqstp, seq_num);
+	goto out;
 }
 
 static inline u32 round_up_to_quad(u32 i)
@@ -721,14 +742,12 @@ gss_verify_header(struct svc_rqst *rqstp, struct rsc *rsci,
 	}
 
 	if (gc->gc_seq > MAXSEQ) {
-		trace_rpcgss_svc_large_seqno(rqstp->rq_xid, gc->gc_seq);
+		trace_rpcgss_svc_seqno_large(rqstp, gc->gc_seq);
 		*authp = rpcsec_gsserr_ctxproblem;
 		return SVC_DENIED;
 	}
-	if (!gss_check_seq_num(rsci, gc->gc_seq)) {
-		trace_rpcgss_svc_old_seqno(rqstp->rq_xid, gc->gc_seq);
+	if (!gss_check_seq_num(rqstp, rsci, gc->gc_seq))
 		return SVC_DROP;
-	}
 	return SVC_OK;
 }
 
@@ -866,11 +885,13 @@ read_u32_from_xdr_buf(struct xdr_buf *buf, int base, u32 *obj)
 static int
 unwrap_integ_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gss_ctx *ctx)
 {
+	u32 integ_len, rseqno, maj_stat;
 	int stat = -EINVAL;
-	u32 integ_len, maj_stat;
 	struct xdr_netobj mic;
 	struct xdr_buf integ_buf;
 
+	mic.data = NULL;
+
 	/* NFS READ normally uses splice to send data in-place. However
 	 * the data in cache can change after the reply's MIC is computed
 	 * but before the RPC reply is sent. To prevent the client from
@@ -885,34 +906,44 @@ unwrap_integ_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct g
 
 	integ_len = svc_getnl(&buf->head[0]);
 	if (integ_len & 3)
-		return stat;
+		goto unwrap_failed;
 	if (integ_len > buf->len)
-		return stat;
-	if (xdr_buf_subsegment(buf, &integ_buf, 0, integ_len)) {
-		WARN_ON_ONCE(1);
-		return stat;
-	}
+		goto unwrap_failed;
+	if (xdr_buf_subsegment(buf, &integ_buf, 0, integ_len))
+		goto unwrap_failed;
+
 	/* copy out mic... */
 	if (read_u32_from_xdr_buf(buf, integ_len, &mic.len))
-		return stat;
+		goto unwrap_failed;
 	if (mic.len > RPC_MAX_AUTH_SIZE)
-		return stat;
+		goto unwrap_failed;
 	mic.data = kmalloc(mic.len, GFP_KERNEL);
 	if (!mic.data)
-		return stat;
+		goto unwrap_failed;
 	if (read_bytes_from_xdr_buf(buf, integ_len + 4, mic.data, mic.len))
-		goto out;
+		goto unwrap_failed;
 	maj_stat = gss_verify_mic(ctx, &integ_buf, &mic);
 	if (maj_stat != GSS_S_COMPLETE)
-		goto out;
-	if (svc_getnl(&buf->head[0]) != seq)
-		goto out;
+		goto bad_mic;
+	rseqno = svc_getnl(&buf->head[0]);
+	if (rseqno != seq)
+		goto bad_seqno;
 	/* trim off the mic and padding at the end before returning */
 	xdr_buf_trim(buf, round_up_to_quad(mic.len) + 4);
 	stat = 0;
 out:
 	kfree(mic.data);
 	return stat;
+
+unwrap_failed:
+	trace_rpcgss_svc_unwrap_failed(rqstp);
+	goto out;
+bad_seqno:
+	trace_rpcgss_svc_seqno_bad(rqstp, seq, rseqno);
+	goto out;
+bad_mic:
+	trace_rpcgss_svc_mic(rqstp, maj_stat);
+	goto out;
 }
 
 static inline int
@@ -937,6 +968,7 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
 {
 	u32 priv_len, maj_stat;
 	int pad, remaining_len, offset;
+	u32 rseqno;
 
 	clear_bit(RQ_SPLICE_OK, &rqstp->rq_flags);
 
@@ -951,7 +983,7 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
 	 * not yet read from the head, so these two values are different: */
 	remaining_len = total_buf_len(buf);
 	if (priv_len > remaining_len)
-		return -EINVAL;
+		goto unwrap_failed;
 	pad = remaining_len - priv_len;
 	buf->len -= pad;
 	fix_priv_head(buf, pad);
@@ -972,11 +1004,22 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
 		fix_priv_head(buf, pad);
 	}
 	if (maj_stat != GSS_S_COMPLETE)
-		return -EINVAL;
+		goto bad_unwrap;
 out_seq:
-	if (svc_getnl(&buf->head[0]) != seq)
-		return -EINVAL;
+	rseqno = svc_getnl(&buf->head[0]);
+	if (rseqno != seq)
+		goto bad_seqno;
 	return 0;
+
+unwrap_failed:
+	trace_rpcgss_svc_unwrap_failed(rqstp);
+	return -EINVAL;
+bad_seqno:
+	trace_rpcgss_svc_seqno_bad(rqstp, seq, rseqno);
+	return -EINVAL;
+bad_unwrap:
+	trace_rpcgss_svc_unwrap(rqstp, maj_stat);
+	return -EINVAL;
 }
 
 struct gss_svc_data {
@@ -1314,8 +1357,7 @@ static int svcauth_gss_proxy_init(struct svc_rqst *rqstp,
 	if (status)
 		goto out;
 
-	trace_rpcgss_svc_accept_upcall(rqstp->rq_xid, ud.major_status,
-				       ud.minor_status);
+	trace_rpcgss_svc_accept_upcall(rqstp, ud.major_status, ud.minor_status);
 
 	switch (ud.major_status) {
 	case GSS_S_CONTINUE_NEEDED:
@@ -1490,8 +1532,6 @@ svcauth_gss_accept(struct svc_rqst *rqstp, __be32 *authp)
 	int		ret;
 	struct sunrpc_net *sn = net_generic(SVC_NET(rqstp), sunrpc_net_id);
 
-	trace_rpcgss_svc_accept(rqstp->rq_xid, argv->iov_len);
-
 	*authp = rpc_autherr_badcred;
 	if (!svcdata)
 		svcdata = kmalloc(sizeof(*svcdata), GFP_KERNEL);
@@ -1608,6 +1648,7 @@ svcauth_gss_accept(struct svc_rqst *rqstp, __be32 *authp)
 					GSS_C_QOP_DEFAULT,
 					gc->gc_svc);
 		ret = SVC_OK;
+		trace_rpcgss_svc_authenticate(rqstp, gc);
 		goto out;
 	}
 garbage_args:
diff --git a/net/sunrpc/auth_gss/trace.c b/net/sunrpc/auth_gss/trace.c
index 49fa583d7f91..d26036a57443 100644
--- a/net/sunrpc/auth_gss/trace.c
+++ b/net/sunrpc/auth_gss/trace.c
@@ -5,6 +5,9 @@
 
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/sched.h>
+#include <linux/sunrpc/svc.h>
+#include <linux/sunrpc/svc_xprt.h>
+#include <linux/sunrpc/auth_gss.h>
 #include <linux/sunrpc/gss_err.h>
 #include <linux/sunrpc/auth_gss.h>
 
-- 
cgit v1.2.3


From e814eecbe3bbeaa8b004d25a4b8974d232b765a9 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Thu, 11 Jun 2020 12:44:56 -0400
Subject: svcrdma: Fix page leak in svc_rdma_recv_read_chunk()

Commit 07d0ff3b0cd2 ("svcrdma: Clean up Read chunk path") moved the
page saver logic so that it gets executed event when an error occurs.
In that case, the I/O is never posted, and those pages are then
leaked. Errors in this path, however, are quite rare.

Fixes: 07d0ff3b0cd2 ("svcrdma: Clean up Read chunk path")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_rw.c | 28 +++++++++++++++++++++-------
 1 file changed, 21 insertions(+), 7 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index 5eb35309ecef..83806fa94def 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -684,7 +684,6 @@ static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
 				     struct svc_rdma_read_info *info,
 				     __be32 *p)
 {
-	unsigned int i;
 	int ret;
 
 	ret = -EINVAL;
@@ -707,12 +706,6 @@ static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
 		info->ri_chunklen += rs_length;
 	}
 
-	/* Pages under I/O have been copied to head->rc_pages.
-	 * Prevent their premature release by svc_xprt_release() .
-	 */
-	for (i = 0; i < info->ri_readctxt->rc_page_count; i++)
-		rqstp->rq_pages[i] = NULL;
-
 	return ret;
 }
 
@@ -807,6 +800,26 @@ out:
 	return ret;
 }
 
+/* Pages under I/O have been copied to head->rc_pages. Ensure they
+ * are not released by svc_xprt_release() until the I/O is complete.
+ *
+ * This has to be done after all Read WRs are constructed to properly
+ * handle a page that is part of I/O on behalf of two different RDMA
+ * segments.
+ *
+ * Do this only if I/O has been posted. Otherwise, we do indeed want
+ * svc_xprt_release() to clean things up properly.
+ */
+static void svc_rdma_save_io_pages(struct svc_rqst *rqstp,
+				   const unsigned int start,
+				   const unsigned int num_pages)
+{
+	unsigned int i;
+
+	for (i = start; i < num_pages + start; i++)
+		rqstp->rq_pages[i] = NULL;
+}
+
 /**
  * svc_rdma_recv_read_chunk - Pull a Read chunk from the client
  * @rdma: controlling RDMA transport
@@ -860,6 +873,7 @@ int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma, struct svc_rqst *rqstp,
 	ret = svc_rdma_post_chunk_ctxt(&info->ri_cc);
 	if (ret < 0)
 		goto out_err;
+	svc_rdma_save_io_pages(rqstp, 0, head->rc_page_count);
 	return 0;
 
 out_err:
-- 
cgit v1.2.3


From 6e9fab7073e5b6bdba4d5891cd486ffdd7de373f Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 25 Mar 2020 11:15:55 -0400
Subject: svcrdma: Remove save_io_pages() call from send_error_msg()

Commit 4757d90b15d8 ("svcrdma: Report Write/Reply chunk overruns")
made an effort to preserve I/O pages until RDMA Write completion.

In a subsequent patch, I intend to de-duplicate the two functions
that send ERR_CHUNK responses. Pull the save_io_pages() call out of
svc_rdma_send_error_msg() to make it more like
svc_rdma_send_error().

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 38e7c3c8c4a9..2f88d01e8d27 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -806,8 +806,7 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
 
 /* Given the client-provided Write and Reply chunks, the server was not
  * able to form a complete reply. Return an RDMA_ERROR message so the
- * client can retire this RPC transaction. As above, the Send completion
- * routine releases payload pages that were part of a previous RDMA Write.
+ * client can retire this RPC transaction.
  *
  * Remote Invalidation is skipped for simplicity.
  */
@@ -834,8 +833,6 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	*p   = err_chunk;
 	trace_svcrdma_err_chunk(*rdma_argp);
 
-	svc_rdma_save_io_pages(rqstp, ctxt);
-
 	ctxt->sc_send_wr.num_sge = 1;
 	ctxt->sc_send_wr.opcode = IB_WR_SEND;
 	ctxt->sc_sges[0].length = ctxt->sc_hdrbuf.len;
@@ -930,6 +927,10 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	if (ret != -E2BIG && ret != -EINVAL)
 		goto err1;
 
+	/* Send completion releases payload pages that were part
+	 * of previously posted RDMA Writes.
+	 */
+	svc_rdma_save_io_pages(rqstp, sctxt);
 	ret = svc_rdma_send_error_msg(rdma, sctxt, rqstp);
 	if (ret < 0)
 		goto err1;
-- 
cgit v1.2.3


From d1f6e2369c63f2cbc7a7441d3ee9b0eabfa9a327 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 25 Mar 2020 11:31:37 -0400
Subject: svcrdma: Add @rctxt parameter to svc_rdma_send_error() functions

Another step towards making svc_rdma_send_error_msg() and
svc_rdma_send_error() similar enough to eliminate one of them.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c |  9 +++++----
 net/sunrpc/xprtrdma/svc_rdma_sendto.c   | 23 +++++++++++------------
 2 files changed, 16 insertions(+), 16 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index e426fedb9524..60d855116ae7 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -715,10 +715,11 @@ static void rdma_read_complete(struct svc_rqst *rqstp,
 }
 
 static void svc_rdma_send_error(struct svcxprt_rdma *xprt,
-				__be32 *rdma_argp, int status)
+				struct svc_rdma_recv_ctxt *rctxt,
+				int status)
 {
+	__be32 *p, *rdma_argp = rctxt->rc_recv_buf;
 	struct svc_rdma_send_ctxt *ctxt;
-	__be32 *p;
 	int ret;
 
 	ctxt = svc_rdma_send_ctxt_get(xprt);
@@ -900,13 +901,13 @@ out_readchunk:
 	return 0;
 
 out_err:
-	svc_rdma_send_error(rdma_xprt, p, ret);
+	svc_rdma_send_error(rdma_xprt, ctxt, ret);
 	svc_rdma_recv_ctxt_put(rdma_xprt, ctxt);
 	return 0;
 
 out_postfail:
 	if (ret == -EINVAL)
-		svc_rdma_send_error(rdma_xprt, p, ret);
+		svc_rdma_send_error(rdma_xprt, ctxt, ret);
 	svc_rdma_recv_ctxt_put(rdma_xprt, ctxt);
 	return ret;
 
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 2f88d01e8d27..47ada61411c3 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -811,18 +811,17 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
  * Remote Invalidation is skipped for simplicity.
  */
 static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
-				   struct svc_rdma_send_ctxt *ctxt,
-				   struct svc_rqst *rqstp)
+				   struct svc_rdma_send_ctxt *sctxt,
+				   struct svc_rdma_recv_ctxt *rctxt)
 {
-	struct svc_rdma_recv_ctxt *rctxt = rqstp->rq_xprt_ctxt;
 	__be32 *rdma_argp = rctxt->rc_recv_buf;
 	__be32 *p;
 
-	rpcrdma_set_xdrlen(&ctxt->sc_hdrbuf, 0);
-	xdr_init_encode(&ctxt->sc_stream, &ctxt->sc_hdrbuf, ctxt->sc_xprt_buf,
-			NULL);
+	rpcrdma_set_xdrlen(&sctxt->sc_hdrbuf, 0);
+	xdr_init_encode(&sctxt->sc_stream, &sctxt->sc_hdrbuf,
+			sctxt->sc_xprt_buf, NULL);
 
-	p = xdr_reserve_space(&ctxt->sc_stream, RPCRDMA_HDRLEN_ERR);
+	p = xdr_reserve_space(&sctxt->sc_stream, RPCRDMA_HDRLEN_ERR);
 	if (!p)
 		return -ENOMSG;
 
@@ -833,10 +832,10 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	*p   = err_chunk;
 	trace_svcrdma_err_chunk(*rdma_argp);
 
-	ctxt->sc_send_wr.num_sge = 1;
-	ctxt->sc_send_wr.opcode = IB_WR_SEND;
-	ctxt->sc_sges[0].length = ctxt->sc_hdrbuf.len;
-	return svc_rdma_send(rdma, &ctxt->sc_send_wr);
+	sctxt->sc_send_wr.num_sge = 1;
+	sctxt->sc_send_wr.opcode = IB_WR_SEND;
+	sctxt->sc_sges[0].length = sctxt->sc_hdrbuf.len;
+	return svc_rdma_send(rdma, &sctxt->sc_send_wr);
 }
 
 /**
@@ -931,7 +930,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	 * of previously posted RDMA Writes.
 	 */
 	svc_rdma_save_io_pages(rqstp, sctxt);
-	ret = svc_rdma_send_error_msg(rdma, sctxt, rqstp);
+	ret = svc_rdma_send_error_msg(rdma, sctxt, rctxt);
 	if (ret < 0)
 		goto err1;
 	return 0;
-- 
cgit v1.2.3


From 4f200bd8affbccc3152d497b4ce5cfaca5a7c53f Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 25 Mar 2020 11:57:51 -0400
Subject: svcrdma: Add a @status parameter to svc_rdma_send_error_msg()

The common "send RDMA_ERR" function should be in svc_rdma_sendto.c,
since that is where the other Send-related functions are located.
So from here, I will beef up svc_rdma_send_error_msg() and deprecate
svc_rdma_send_error().

A generic svc_rdma_send_error_msg() will need to handle both
ERR_CHUNK and ERR_VERS. Copy that logic from svc_rdma_send_error()
to svc_rdma_send_error_msg().

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 32 ++++++++++++++++++++++++++------
 1 file changed, 26 insertions(+), 6 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 47ada61411c3..73fe7a213169 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -812,7 +812,8 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
  */
 static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 				   struct svc_rdma_send_ctxt *sctxt,
-				   struct svc_rdma_recv_ctxt *rctxt)
+				   struct svc_rdma_recv_ctxt *rctxt,
+				   int status)
 {
 	__be32 *rdma_argp = rctxt->rc_recv_buf;
 	__be32 *p;
@@ -821,16 +822,35 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	xdr_init_encode(&sctxt->sc_stream, &sctxt->sc_hdrbuf,
 			sctxt->sc_xprt_buf, NULL);
 
-	p = xdr_reserve_space(&sctxt->sc_stream, RPCRDMA_HDRLEN_ERR);
+	p = xdr_reserve_space(&sctxt->sc_stream,
+			      rpcrdma_fixed_maxsz * sizeof(*p));
 	if (!p)
 		return -ENOMSG;
 
 	*p++ = *rdma_argp;
 	*p++ = *(rdma_argp + 1);
 	*p++ = rdma->sc_fc_credits;
-	*p++ = rdma_error;
-	*p   = err_chunk;
-	trace_svcrdma_err_chunk(*rdma_argp);
+	*p = rdma_error;
+
+	switch (status) {
+	case -EPROTONOSUPPORT:
+		p = xdr_reserve_space(&sctxt->sc_stream, 3 * sizeof(*p));
+		if (!p)
+			return -ENOMSG;
+
+		*p++ = err_vers;
+		*p++ = rpcrdma_version;
+		*p = rpcrdma_version;
+		trace_svcrdma_err_vers(*rdma_argp);
+		break;
+	default:
+		p = xdr_reserve_space(&sctxt->sc_stream, sizeof(*p));
+		if (!p)
+			return -ENOMSG;
+
+		*p = err_chunk;
+		trace_svcrdma_err_chunk(*rdma_argp);
+	}
 
 	sctxt->sc_send_wr.num_sge = 1;
 	sctxt->sc_send_wr.opcode = IB_WR_SEND;
@@ -930,7 +950,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	 * of previously posted RDMA Writes.
 	 */
 	svc_rdma_save_io_pages(rqstp, sctxt);
-	ret = svc_rdma_send_error_msg(rdma, sctxt, rctxt);
+	ret = svc_rdma_send_error_msg(rdma, sctxt, rctxt, ret);
 	if (ret < 0)
 		goto err1;
 	return 0;
-- 
cgit v1.2.3


From 605c61bee5b13bc9d597192779865a1a707166ed Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 25 Mar 2020 14:38:07 -0400
Subject: svcrdma: Eliminate return value for svc_rdma_send_error_msg()

Like svc_rdma_send_error(), have svc_rdma_send_error_msg() handle
any error conditions internally, rather than duplicating that
recovery logic at every call site.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 25 ++++++++++++++-----------
 1 file changed, 14 insertions(+), 11 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 73fe7a213169..fb548b548c4b 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -810,10 +810,10 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
  *
  * Remote Invalidation is skipped for simplicity.
  */
-static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
-				   struct svc_rdma_send_ctxt *sctxt,
-				   struct svc_rdma_recv_ctxt *rctxt,
-				   int status)
+static void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
+				    struct svc_rdma_send_ctxt *sctxt,
+				    struct svc_rdma_recv_ctxt *rctxt,
+				    int status)
 {
 	__be32 *rdma_argp = rctxt->rc_recv_buf;
 	__be32 *p;
@@ -825,7 +825,7 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	p = xdr_reserve_space(&sctxt->sc_stream,
 			      rpcrdma_fixed_maxsz * sizeof(*p));
 	if (!p)
-		return -ENOMSG;
+		goto put_ctxt;
 
 	*p++ = *rdma_argp;
 	*p++ = *(rdma_argp + 1);
@@ -836,7 +836,7 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	case -EPROTONOSUPPORT:
 		p = xdr_reserve_space(&sctxt->sc_stream, 3 * sizeof(*p));
 		if (!p)
-			return -ENOMSG;
+			goto put_ctxt;
 
 		*p++ = err_vers;
 		*p++ = rpcrdma_version;
@@ -846,7 +846,7 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	default:
 		p = xdr_reserve_space(&sctxt->sc_stream, sizeof(*p));
 		if (!p)
-			return -ENOMSG;
+			goto put_ctxt;
 
 		*p = err_chunk;
 		trace_svcrdma_err_chunk(*rdma_argp);
@@ -855,7 +855,12 @@ static int svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	sctxt->sc_send_wr.num_sge = 1;
 	sctxt->sc_send_wr.opcode = IB_WR_SEND;
 	sctxt->sc_sges[0].length = sctxt->sc_hdrbuf.len;
-	return svc_rdma_send(rdma, &sctxt->sc_send_wr);
+	if (svc_rdma_send(rdma, &sctxt->sc_send_wr))
+		goto put_ctxt;
+	return;
+
+put_ctxt:
+	svc_rdma_send_ctxt_put(rdma, sctxt);
 }
 
 /**
@@ -950,9 +955,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	 * of previously posted RDMA Writes.
 	 */
 	svc_rdma_save_io_pages(rqstp, sctxt);
-	ret = svc_rdma_send_error_msg(rdma, sctxt, rctxt, ret);
-	if (ret < 0)
-		goto err1;
+	svc_rdma_send_error_msg(rdma, sctxt, rctxt, ret);
 	return 0;
 
  err1:
-- 
cgit v1.2.3


From c65b326b1eb983bca35ed43d0e453d1b15705f10 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 25 Mar 2020 14:41:46 -0400
Subject: svcrdma: Make svc_rdma_send_error_msg() a global function

Prepare for svc_rdma_send_error_msg() to be invoked from another
source file.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/svc_rdma.h       |  4 ++++
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 28 +++++++++++++++++++---------
 2 files changed, 23 insertions(+), 9 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h
index 7ed82625dc0b..1579f7a14ab4 100644
--- a/include/linux/sunrpc/svc_rdma.h
+++ b/include/linux/sunrpc/svc_rdma.h
@@ -195,6 +195,10 @@ extern int svc_rdma_map_reply_msg(struct svcxprt_rdma *rdma,
 				  struct svc_rdma_send_ctxt *sctxt,
 				  const struct svc_rdma_recv_ctxt *rctxt,
 				  struct xdr_buf *xdr);
+extern void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
+				    struct svc_rdma_send_ctxt *sctxt,
+				    struct svc_rdma_recv_ctxt *rctxt,
+				    int status);
 extern int svc_rdma_sendto(struct svc_rqst *);
 extern int svc_rdma_read_payload(struct svc_rqst *rqstp, unsigned int offset,
 				 unsigned int length);
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index fb548b548c4b..57041298fe4f 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -804,16 +804,25 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
 	return svc_rdma_send(rdma, &sctxt->sc_send_wr);
 }
 
-/* Given the client-provided Write and Reply chunks, the server was not
- * able to form a complete reply. Return an RDMA_ERROR message so the
- * client can retire this RPC transaction.
- *
- * Remote Invalidation is skipped for simplicity.
+/**
+ * svc_rdma_send_error_msg - Send an RPC/RDMA v1 error response
+ * @rdma: controlling transport context
+ * @sctxt: Send context for the response
+ * @rctxt: Receive context for incoming bad message
+ * @status: negative errno indicating error that occurred
+ *
+ * Given the client-provided Read, Write, and Reply chunks, the
+ * server was not able to parse the Call or form a complete Reply.
+ * Return an RDMA_ERROR message so the client can retire the RPC
+ * transaction.
+ *
+ * The caller does not have to release @sctxt. It is released by
+ * Send completion, or by this function on error.
  */
-static void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
-				    struct svc_rdma_send_ctxt *sctxt,
-				    struct svc_rdma_recv_ctxt *rctxt,
-				    int status)
+void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
+			     struct svc_rdma_send_ctxt *sctxt,
+			     struct svc_rdma_recv_ctxt *rctxt,
+			     int status)
 {
 	__be32 *rdma_argp = rctxt->rc_recv_buf;
 	__be32 *p;
@@ -852,6 +861,7 @@ static void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 		trace_svcrdma_err_chunk(*rdma_argp);
 	}
 
+	/* Remote Invalidation is skipped for simplicity. */
 	sctxt->sc_send_wr.num_sge = 1;
 	sctxt->sc_send_wr.opcode = IB_WR_SEND;
 	sctxt->sc_sges[0].length = sctxt->sc_hdrbuf.len;
-- 
cgit v1.2.3


From ba6cc97738a15751350d3f85a6b856e4d6d4c202 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 24 Mar 2020 17:57:18 -0400
Subject: svcrdma: Consolidate send_error helper functions

Final refactor: Replace internals of svc_rdma_send_error() with a
simple call to svc_rdma_send_error_msg().

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 52 ++++-----------------------------
 1 file changed, 5 insertions(+), 47 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 60d855116ae7..c072ce61b393 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -714,58 +714,16 @@ static void rdma_read_complete(struct svc_rqst *rqstp,
 	rqstp->rq_arg.buflen = head->rc_arg.buflen;
 }
 
-static void svc_rdma_send_error(struct svcxprt_rdma *xprt,
+static void svc_rdma_send_error(struct svcxprt_rdma *rdma,
 				struct svc_rdma_recv_ctxt *rctxt,
 				int status)
 {
-	__be32 *p, *rdma_argp = rctxt->rc_recv_buf;
-	struct svc_rdma_send_ctxt *ctxt;
-	int ret;
+	struct svc_rdma_send_ctxt *sctxt;
 
-	ctxt = svc_rdma_send_ctxt_get(xprt);
-	if (!ctxt)
+	sctxt = svc_rdma_send_ctxt_get(rdma);
+	if (!sctxt)
 		return;
-
-	p = xdr_reserve_space(&ctxt->sc_stream,
-			      rpcrdma_fixed_maxsz * sizeof(*p));
-	if (!p)
-		goto put_ctxt;
-
-	*p++ = *rdma_argp;
-	*p++ = *(rdma_argp + 1);
-	*p++ = xprt->sc_fc_credits;
-	*p = rdma_error;
-
-	switch (status) {
-	case -EPROTONOSUPPORT:
-		p = xdr_reserve_space(&ctxt->sc_stream, 3 * sizeof(*p));
-		if (!p)
-			goto put_ctxt;
-
-		*p++ = err_vers;
-		*p++ = rpcrdma_version;
-		*p = rpcrdma_version;
-		trace_svcrdma_err_vers(*rdma_argp);
-		break;
-	default:
-		p = xdr_reserve_space(&ctxt->sc_stream, sizeof(*p));
-		if (!p)
-			goto put_ctxt;
-
-		*p = err_chunk;
-		trace_svcrdma_err_chunk(*rdma_argp);
-	}
-
-	ctxt->sc_send_wr.num_sge = 1;
-	ctxt->sc_send_wr.opcode = IB_WR_SEND;
-	ctxt->sc_sges[0].length = ctxt->sc_hdrbuf.len;
-	ret = svc_rdma_send(xprt, &ctxt->sc_send_wr);
-	if (ret)
-		goto put_ctxt;
-	return;
-
-put_ctxt:
-	svc_rdma_send_ctxt_put(xprt, ctxt);
+	svc_rdma_send_error_msg(rdma, sctxt, rctxt, status);
 }
 
 /* By convention, backchannel calls arrive via rdma_msg type
-- 
cgit v1.2.3


From 3f8f25c696f9c4e352a4d705ba767af676421564 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Thu, 30 Apr 2020 14:17:40 -0400
Subject: svcrdma: Clean up trace_svcrdma_send_failed() tracepoint

- Use the _err naming convention instead
- Remove display of kernel memory address of the controlling xprt

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/trace/events/rpcrdma.h        | 7 ++-----
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 2 +-
 2 files changed, 3 insertions(+), 6 deletions(-)

(limited to 'net')

diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index 0f05a6e2b9cb..0eff80dee066 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1716,7 +1716,7 @@ TRACE_EVENT(svcrdma_send_pullup,
 	TP_printk("len=%u", __entry->len)
 );
 
-TRACE_EVENT(svcrdma_send_failed,
+TRACE_EVENT(svcrdma_send_err,
 	TP_PROTO(
 		const struct svc_rqst *rqst,
 		int status
@@ -1727,19 +1727,16 @@ TRACE_EVENT(svcrdma_send_failed,
 	TP_STRUCT__entry(
 		__field(int, status)
 		__field(u32, xid)
-		__field(const void *, xprt)
 		__string(addr, rqst->rq_xprt->xpt_remotebuf)
 	),
 
 	TP_fast_assign(
 		__entry->status = status;
 		__entry->xid = __be32_to_cpu(rqst->rq_xid);
-		__entry->xprt = rqst->rq_xprt;
 		__assign_str(addr, rqst->rq_xprt->xpt_remotebuf);
 	),
 
-	TP_printk("xprt=%p addr=%s xid=0x%08x status=%d",
-		__entry->xprt, __get_str(addr),
+	TP_printk("addr=%s xid=0x%08x status=%d", __get_str(addr),
 		__entry->xid, __entry->status
 	)
 );
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 57041298fe4f..f985f548346a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -971,7 +971,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
  err1:
 	svc_rdma_send_ctxt_put(rdma, sctxt);
  err0:
-	trace_svcrdma_send_failed(rqstp, ret);
+	trace_svcrdma_send_err(rqstp, ret);
 	set_bit(XPT_CLOSE, &xprt->xpt_flags);
 	return -ENOTCONN;
 }
-- 
cgit v1.2.3


From 07e9a6325a35fb9655f7b52e2b9dc632da6eef51 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Sat, 28 Mar 2020 13:43:22 -0400
Subject: SUNRPC: Add helpers for decoding list discriminators symbolically

Use these helpers in a few spots to demonstrate their use.

The remaining open-coded discriminator checks in rpcrdma will be
addressed in subsequent patches.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/xdr.h              | 26 ++++++++++++++++++++++++++
 net/sunrpc/xprtrdma/rpc_rdma.c          | 12 ++++++------
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 17 ++++++++---------
 3 files changed, 40 insertions(+), 15 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h
index 22c207b2425f..5a6a81b7cd9f 100644
--- a/include/linux/sunrpc/xdr.h
+++ b/include/linux/sunrpc/xdr.h
@@ -474,6 +474,32 @@ xdr_stream_encode_uint32_array(struct xdr_stream *xdr,
 	return ret;
 }
 
+/**
+ * xdr_item_is_absent - symbolically handle XDR discriminators
+ * @p: pointer to undecoded discriminator
+ *
+ * Return values:
+ *   %true if the following XDR item is absent
+ *   %false if the following XDR item is present
+ */
+static inline bool xdr_item_is_absent(const __be32 *p)
+{
+	return *p == xdr_zero;
+}
+
+/**
+ * xdr_item_is_present - symbolically handle XDR discriminators
+ * @p: pointer to undecoded discriminator
+ *
+ * Return values:
+ *   %true if the following XDR item is present
+ *   %false if the following XDR item is absent
+ */
+static inline bool xdr_item_is_present(const __be32 *p)
+{
+	return *p != xdr_zero;
+}
+
 /**
  * xdr_stream_decode_u32 - Decode a 32-bit integer
  * @xdr: pointer to xdr_stream
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 935bbef2f7be..feecd1f55f18 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -1133,11 +1133,11 @@ rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep)
 	p = xdr_inline_decode(xdr, 0);
 
 	/* Chunk lists */
-	if (*p++ != xdr_zero)
+	if (xdr_item_is_present(p++))
 		return false;
-	if (*p++ != xdr_zero)
+	if (xdr_item_is_present(p++))
 		return false;
-	if (*p++ != xdr_zero)
+	if (xdr_item_is_present(p++))
 		return false;
 
 	/* RPC header */
@@ -1215,7 +1215,7 @@ static int decode_read_list(struct xdr_stream *xdr)
 	p = xdr_inline_decode(xdr, sizeof(*p));
 	if (unlikely(!p))
 		return -EIO;
-	if (unlikely(*p != xdr_zero))
+	if (unlikely(xdr_item_is_present(p)))
 		return -EIO;
 	return 0;
 }
@@ -1234,7 +1234,7 @@ static int decode_write_list(struct xdr_stream *xdr, u32 *length)
 		p = xdr_inline_decode(xdr, sizeof(*p));
 		if (unlikely(!p))
 			return -EIO;
-		if (*p == xdr_zero)
+		if (xdr_item_is_absent(p))
 			break;
 		if (!first)
 			return -EIO;
@@ -1256,7 +1256,7 @@ static int decode_reply_chunk(struct xdr_stream *xdr, u32 *length)
 		return -EIO;
 
 	*length = 0;
-	if (*p != xdr_zero)
+	if (xdr_item_is_present(p))
 		if (decode_write_chunk(xdr, length))
 			return -EIO;
 	return 0;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index c072ce61b393..5e78067889f3 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -419,7 +419,7 @@ static bool xdr_check_read_list(struct svc_rdma_recv_ctxt *rctxt)
 
 	len = 0;
 	first = true;
-	while (*p != xdr_zero) {
+	while (xdr_item_is_present(p)) {
 		p = xdr_inline_decode(&rctxt->rc_stream,
 				      rpcrdma_readseg_maxsz * sizeof(*p));
 		if (!p)
@@ -500,7 +500,7 @@ static bool xdr_check_write_list(struct svc_rdma_recv_ctxt *rctxt)
 	if (!p)
 		return false;
 	rctxt->rc_write_list = p;
-	while (*p != xdr_zero) {
+	while (xdr_item_is_present(p)) {
 		if (!xdr_check_write_chunk(rctxt, MAX_BYTES_WRITE_CHUNK))
 			return false;
 		++chcount;
@@ -532,12 +532,11 @@ static bool xdr_check_reply_chunk(struct svc_rdma_recv_ctxt *rctxt)
 	p = xdr_inline_decode(&rctxt->rc_stream, sizeof(*p));
 	if (!p)
 		return false;
-	rctxt->rc_reply_chunk = p;
-	if (*p != xdr_zero) {
+	rctxt->rc_reply_chunk = NULL;
+	if (xdr_item_is_present(p)) {
 		if (!xdr_check_write_chunk(rctxt, MAX_BYTES_SPECIAL_CHUNK))
 			return false;
-	} else {
-		rctxt->rc_reply_chunk = NULL;
+		rctxt->rc_reply_chunk = p;
 	}
 	return true;
 }
@@ -568,7 +567,7 @@ static void svc_rdma_get_inv_rkey(struct svcxprt_rdma *rdma,
 	p += rpcrdma_fixed_maxsz;
 
 	/* Read list */
-	while (*p++ != xdr_zero) {
+	while (xdr_item_is_present(p++)) {
 		p++;	/* position */
 		if (inv_rkey == xdr_zero)
 			inv_rkey = *p;
@@ -578,7 +577,7 @@ static void svc_rdma_get_inv_rkey(struct svcxprt_rdma *rdma,
 	}
 
 	/* Write list */
-	while (*p++ != xdr_zero) {
+	while (xdr_item_is_present(p++)) {
 		segcount = be32_to_cpup(p++);
 		for (i = 0; i < segcount; i++) {
 			if (inv_rkey == xdr_zero)
@@ -590,7 +589,7 @@ static void svc_rdma_get_inv_rkey(struct svcxprt_rdma *rdma,
 	}
 
 	/* Reply chunk */
-	if (*p++ != xdr_zero) {
+	if (xdr_item_is_present(p++)) {
 		segcount = be32_to_cpup(p++);
 		for (i = 0; i < segcount; i++) {
 			if (inv_rkey == xdr_zero)
-- 
cgit v1.2.3


From f60a08697d28b138c73b14c3204947bc8e637197 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Sun, 29 Mar 2020 16:44:13 -0400
Subject: svcrdma: Add common XDR decoders for RDMA and Read segments

Clean up: De-duplicate some code.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/rpc_rdma.h          | 37 ++++++++++++++++++++++++++++++++
 net/sunrpc/xprtrdma/frwr_ops.c           |  1 -
 net/sunrpc/xprtrdma/rpc_rdma.c           |  5 +----
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c  |  4 +---
 net/sunrpc/xprtrdma/svc_rdma_rw.c        | 37 ++++++++++++++------------------
 net/sunrpc/xprtrdma/svc_rdma_sendto.c    |  5 +----
 net/sunrpc/xprtrdma/svc_rdma_transport.c |  1 -
 7 files changed, 56 insertions(+), 34 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/rpc_rdma.h b/include/linux/sunrpc/rpc_rdma.h
index 320c672d84de..db50380f64f4 100644
--- a/include/linux/sunrpc/rpc_rdma.h
+++ b/include/linux/sunrpc/rpc_rdma.h
@@ -124,4 +124,41 @@ rpcrdma_decode_buffer_size(u8 val)
 	return ((unsigned int)val + 1) << 10;
 }
 
+/**
+ * xdr_decode_rdma_segment - Decode contents of an RDMA segment
+ * @p: Pointer to the undecoded RDMA segment
+ * @handle: Upon return, the RDMA handle
+ * @length: Upon return, the RDMA length
+ * @offset: Upon return, the RDMA offset
+ *
+ * Return value:
+ *   Pointer to the XDR item that follows the RDMA segment
+ */
+static inline __be32 *xdr_decode_rdma_segment(__be32 *p, u32 *handle,
+					      u32 *length, u64 *offset)
+{
+	*handle = be32_to_cpup(p++);
+	*length = be32_to_cpup(p++);
+	return xdr_decode_hyper(p, offset);
+}
+
+/**
+ * xdr_decode_read_segment - Decode contents of a Read segment
+ * @p: Pointer to the undecoded Read segment
+ * @position: Upon return, the segment's position
+ * @handle: Upon return, the RDMA handle
+ * @length: Upon return, the RDMA length
+ * @offset: Upon return, the RDMA offset
+ *
+ * Return value:
+ *   Pointer to the XDR item that follows the Read segment
+ */
+static inline __be32 *xdr_decode_read_segment(__be32 *p, u32 *position,
+					      u32 *handle, u32 *length,
+					      u64 *offset)
+{
+	*position = be32_to_cpup(p++);
+	return xdr_decode_rdma_segment(p, handle, length, offset);
+}
+
 #endif				/* _LINUX_SUNRPC_RPC_RDMA_H */
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index b647562a26dd..7f94c9a19fd3 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -40,7 +40,6 @@
  * New MRs are created on demand.
  */
 
-#include <linux/sunrpc/rpc_rdma.h>
 #include <linux/sunrpc/svc_rdma.h>
 
 #include "xprt_rdma.h"
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index feecd1f55f18..5461f01eeca6 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -1176,10 +1176,7 @@ static int decode_rdma_segment(struct xdr_stream *xdr, u32 *length)
 	if (unlikely(!p))
 		return -EIO;
 
-	handle = be32_to_cpup(p++);
-	*length = be32_to_cpup(p++);
-	xdr_decode_hyper(p, &offset);
-
+	xdr_decode_rdma_segment(p, &handle, length, &offset);
 	trace_xprtrdma_decode_seg(handle, *length, offset);
 	return 0;
 }
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 5e78067889f3..c0587d3cd389 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -466,9 +466,7 @@ static bool xdr_check_write_chunk(struct svc_rdma_recv_ctxt *rctxt, u32 maxlen)
 		if (!p)
 			return false;
 
-		handle = be32_to_cpup(p++);
-		length = be32_to_cpup(p++);
-		xdr_decode_hyper(p, &offset);
+		xdr_decode_rdma_segment(p, &handle, &length, &offset);
 		trace_svcrdma_decode_wseg(handle, length, offset);
 
 		total += length;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index 83806fa94def..2038b1b286dd 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -7,6 +7,7 @@
 
 #include <rdma/rw.h>
 
+#include <linux/sunrpc/xdr.h>
 #include <linux/sunrpc/rpc_rdma.h>
 #include <linux/sunrpc/svc_rdma.h>
 
@@ -441,34 +442,32 @@ svc_rdma_build_writes(struct svc_rdma_write_info *info,
 	seg = info->wi_segs + info->wi_seg_no * rpcrdma_segment_maxsz;
 	do {
 		unsigned int write_len;
-		u32 seg_length, seg_handle;
-		u64 seg_offset;
+		u32 handle, length;
+		u64 offset;
 
 		if (info->wi_seg_no >= info->wi_nsegs)
 			goto out_overflow;
 
-		seg_handle = be32_to_cpup(seg);
-		seg_length = be32_to_cpup(seg + 1);
-		xdr_decode_hyper(seg + 2, &seg_offset);
-		seg_offset += info->wi_seg_off;
+		xdr_decode_rdma_segment(seg, &handle, &length, &offset);
+		offset += info->wi_seg_off;
 
-		write_len = min(remaining, seg_length - info->wi_seg_off);
+		write_len = min(remaining, length - info->wi_seg_off);
 		ctxt = svc_rdma_get_rw_ctxt(rdma,
 					    (write_len >> PAGE_SHIFT) + 2);
 		if (!ctxt)
 			return -ENOMEM;
 
 		constructor(info, write_len, ctxt);
-		ret = svc_rdma_rw_ctx_init(rdma, ctxt, seg_offset, seg_handle,
+		ret = svc_rdma_rw_ctx_init(rdma, ctxt, offset, handle,
 					   DMA_TO_DEVICE);
 		if (ret < 0)
 			return -EIO;
 
-		trace_svcrdma_send_wseg(seg_handle, write_len, seg_offset);
+		trace_svcrdma_send_wseg(handle, write_len, offset);
 
 		list_add(&ctxt->rw_list, &cc->cc_rwctxts);
 		cc->cc_sqecount += ret;
-		if (write_len == seg_length - info->wi_seg_off) {
+		if (write_len == length - info->wi_seg_off) {
 			seg += 4;
 			info->wi_seg_no++;
 			info->wi_seg_off = 0;
@@ -689,21 +688,17 @@ static int svc_rdma_build_read_chunk(struct svc_rqst *rqstp,
 	ret = -EINVAL;
 	info->ri_chunklen = 0;
 	while (*p++ != xdr_zero && be32_to_cpup(p++) == info->ri_position) {
-		u32 rs_handle, rs_length;
-		u64 rs_offset;
+		u32 handle, length;
+		u64 offset;
 
-		rs_handle = be32_to_cpup(p++);
-		rs_length = be32_to_cpup(p++);
-		p = xdr_decode_hyper(p, &rs_offset);
-
-		ret = svc_rdma_build_read_segment(info, rqstp,
-						  rs_handle, rs_length,
-						  rs_offset);
+		p = xdr_decode_rdma_segment(p, &handle, &length, &offset);
+		ret = svc_rdma_build_read_segment(info, rqstp, handle, length,
+						  offset);
 		if (ret < 0)
 			break;
 
-		trace_svcrdma_send_rseg(rs_handle, rs_length, rs_offset);
-		info->ri_chunklen += rs_length;
+		trace_svcrdma_send_rseg(handle, length, offset);
+		info->ri_chunklen += length;
 	}
 
 	return ret;
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index f985f548346a..a78f1d22e9bb 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -106,7 +106,6 @@
 #include <rdma/rdma_cm.h>
 
 #include <linux/sunrpc/debug.h>
-#include <linux/sunrpc/rpc_rdma.h>
 #include <linux/sunrpc/svc_rdma.h>
 
 #include "xprt_rdma.h"
@@ -375,9 +374,7 @@ static ssize_t svc_rdma_encode_write_segment(__be32 *src,
 	if (!p)
 		return -EMSGSIZE;
 
-	handle = be32_to_cpup(src++);
-	length = be32_to_cpup(src++);
-	xdr_decode_hyper(src, &offset);
+	xdr_decode_rdma_segment(src, &handle, &length, &offset);
 
 	*p++ = cpu_to_be32(handle);
 	if (*remaining < length) {
diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index d38be57b00ed..3da7901a49e6 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -55,7 +55,6 @@
 
 #include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/debug.h>
-#include <linux/sunrpc/rpc_rdma.h>
 #include <linux/sunrpc/svc_xprt.h>
 #include <linux/sunrpc/svc_rdma.h>
 
-- 
cgit v1.2.3


From 379c3bc6b4eb989ee37c4ce8ab403719e06fe35f Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 7 Apr 2020 15:32:14 -0400
Subject: svcrdma: Add common XDR encoders for RDMA and Read segments

Clean up: De-duplicate some code.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/rpc_rdma.h       | 37 +++++++++++++++++++++++++++++++++++
 net/sunrpc/xprtrdma/rpc_rdma.c        | 14 +++----------
 net/sunrpc/xprtrdma/svc_rdma_sendto.c |  4 +---
 3 files changed, 41 insertions(+), 14 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/rpc_rdma.h b/include/linux/sunrpc/rpc_rdma.h
index db50380f64f4..4af31bbc8802 100644
--- a/include/linux/sunrpc/rpc_rdma.h
+++ b/include/linux/sunrpc/rpc_rdma.h
@@ -124,6 +124,43 @@ rpcrdma_decode_buffer_size(u8 val)
 	return ((unsigned int)val + 1) << 10;
 }
 
+/**
+ * xdr_encode_rdma_segment - Encode contents of an RDMA segment
+ * @p: Pointer into a send buffer
+ * @handle: The RDMA handle to encode
+ * @length: The RDMA length to encode
+ * @offset: The RDMA offset to encode
+ *
+ * Return value:
+ *   Pointer to the XDR position that follows the encoded RDMA segment
+ */
+static inline __be32 *xdr_encode_rdma_segment(__be32 *p, u32 handle,
+					      u32 length, u64 offset)
+{
+	*p++ = cpu_to_be32(handle);
+	*p++ = cpu_to_be32(length);
+	return xdr_encode_hyper(p, offset);
+}
+
+/**
+ * xdr_encode_read_segment - Encode contents of a Read segment
+ * @p: Pointer into a send buffer
+ * @position: The position to encode
+ * @handle: The RDMA handle to encode
+ * @length: The RDMA length to encode
+ * @offset: The RDMA offset to encode
+ *
+ * Return value:
+ *   Pointer to the XDR position that follows the encoded Read segment
+ */
+static inline __be32 *xdr_encode_read_segment(__be32 *p, u32 position,
+					      u32 handle, u32 length,
+					      u64 offset)
+{
+	*p++ = cpu_to_be32(position);
+	return xdr_encode_rdma_segment(p, handle, length, offset);
+}
+
 /**
  * xdr_decode_rdma_segment - Decode contents of an RDMA segment
  * @p: Pointer to the undecoded RDMA segment
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 5461f01eeca6..73ed51893175 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -275,14 +275,6 @@ out:
 	return n;
 }
 
-static void
-xdr_encode_rdma_segment(__be32 *iptr, struct rpcrdma_mr *mr)
-{
-	*iptr++ = cpu_to_be32(mr->mr_handle);
-	*iptr++ = cpu_to_be32(mr->mr_length);
-	xdr_encode_hyper(iptr, mr->mr_offset);
-}
-
 static int
 encode_rdma_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr)
 {
@@ -292,7 +284,7 @@ encode_rdma_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr)
 	if (unlikely(!p))
 		return -EMSGSIZE;
 
-	xdr_encode_rdma_segment(p, mr);
+	xdr_encode_rdma_segment(p, mr->mr_handle, mr->mr_length, mr->mr_offset);
 	return 0;
 }
 
@@ -307,8 +299,8 @@ encode_read_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr,
 		return -EMSGSIZE;
 
 	*p++ = xdr_one;			/* Item present */
-	*p++ = cpu_to_be32(position);
-	xdr_encode_rdma_segment(p, mr);
+	xdr_encode_read_segment(p, position, mr->mr_handle, mr->mr_length,
+				mr->mr_offset);
 	return 0;
 }
 
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index a78f1d22e9bb..38d8f0ee35ec 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -376,7 +376,6 @@ static ssize_t svc_rdma_encode_write_segment(__be32 *src,
 
 	xdr_decode_rdma_segment(src, &handle, &length, &offset);
 
-	*p++ = cpu_to_be32(handle);
 	if (*remaining < length) {
 		/* segment only partly filled */
 		length = *remaining;
@@ -385,8 +384,7 @@ static ssize_t svc_rdma_encode_write_segment(__be32 *src,
 		/* entire segment was consumed */
 		*remaining -= length;
 	}
-	*p++ = cpu_to_be32(length);
-	xdr_encode_hyper(p, offset);
+	xdr_encode_rdma_segment(p, handle, length, offset);
 
 	trace_svcrdma_encode_wseg(handle, length, offset);
 	return len;
-- 
cgit v1.2.3


From 9b3bcf8c5c134038e30624db5b57992ae50b80a9 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 29 Apr 2020 16:22:26 -0400
Subject: svcrdma: Introduce Receive completion IDs

Set up a completion ID in each svc_rdma_recv_ctxt. The ID is used
to match an incoming Receive completion to a transport and to a
previous ib_post_recv().

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/svc_rdma.h         |  4 +++
 include/trace/events/rpcrdma.h          | 51 +++++++++++++--------------------
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 15 ++++++++--
 3 files changed, 36 insertions(+), 34 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h
index d28ca1b6f2eb..c3c1e46f510f 100644
--- a/include/linux/sunrpc/svc_rdma.h
+++ b/include/linux/sunrpc/svc_rdma.h
@@ -46,6 +46,7 @@
 #include <linux/sunrpc/xdr.h>
 #include <linux/sunrpc/svcsock.h>
 #include <linux/sunrpc/rpc_rdma.h>
+#include <linux/sunrpc/rpc_rdma_cid.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/rdma_cm.h>
 
@@ -109,6 +110,8 @@ struct svcxprt_rdma {
 	struct work_struct   sc_work;
 
 	struct llist_head    sc_recv_ctxts;
+
+	atomic_t	     sc_completion_ids;
 };
 /* sc_flags */
 #define RDMAXPRT_CONN_PENDING	3
@@ -129,6 +132,7 @@ struct svc_rdma_recv_ctxt {
 	struct list_head	rc_list;
 	struct ib_recv_wr	rc_recv_wr;
 	struct ib_cqe		rc_cqe;
+	struct rpc_rdma_cid	rc_cid;
 	struct ib_sge		rc_recv_sge;
 	void			*rc_recv_buf;
 	struct xdr_buf		rc_arg;
diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index 70ab989aa3b7..a0330a557e34 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -59,8 +59,6 @@ DECLARE_EVENT_CLASS(rpcrdma_completion_class,
 				),					\
 				TP_ARGS(wc, cid))
 
-DEFINE_COMPLETION_EVENT(dummy);
-
 DECLARE_EVENT_CLASS(xprtrdma_reply_event,
 	TP_PROTO(
 		const struct rpcrdma_rep *rep
@@ -1849,57 +1847,48 @@ DEFINE_SENDCOMP_EVENT(send);
 
 TRACE_EVENT(svcrdma_post_recv,
 	TP_PROTO(
-		const struct ib_recv_wr *wr,
-		int status
+		const struct svc_rdma_recv_ctxt *ctxt
 	),
 
-	TP_ARGS(wr, status),
+	TP_ARGS(ctxt),
 
 	TP_STRUCT__entry(
-		__field(const void *, cqe)
-		__field(int, status)
+		__field(u32, cq_id)
+		__field(int, completion_id)
 	),
 
 	TP_fast_assign(
-		__entry->cqe = wr->wr_cqe;
-		__entry->status = status;
+		__entry->cq_id = ctxt->rc_cid.ci_queue_id;
+		__entry->completion_id = ctxt->rc_cid.ci_completion_id;
 	),
 
-	TP_printk("cqe=%p status=%d",
-		__entry->cqe, __entry->status
+	TP_printk("cq.id=%d cid=%d",
+		__entry->cq_id, __entry->completion_id
 	)
 );
 
-TRACE_EVENT(svcrdma_wc_receive,
+DEFINE_COMPLETION_EVENT(svcrdma_wc_receive);
+
+TRACE_EVENT(svcrdma_rq_post_err,
 	TP_PROTO(
-		const struct ib_wc *wc
+		const struct svcxprt_rdma *rdma,
+		int status
 	),
 
-	TP_ARGS(wc),
+	TP_ARGS(rdma, status),
 
 	TP_STRUCT__entry(
-		__field(const void *, cqe)
-		__field(u32, byte_len)
-		__field(unsigned int, status)
-		__field(u32, vendor_err)
+		__field(int, status)
+		__string(addr, rdma->sc_xprt.xpt_remotebuf)
 	),
 
 	TP_fast_assign(
-		__entry->cqe = wc->wr_cqe;
-		__entry->status = wc->status;
-		if (wc->status) {
-			__entry->byte_len = 0;
-			__entry->vendor_err = wc->vendor_err;
-		} else {
-			__entry->byte_len = wc->byte_len;
-			__entry->vendor_err = 0;
-		}
+		__entry->status = status;
+		__assign_str(addr, rdma->sc_xprt.xpt_remotebuf);
 	),
 
-	TP_printk("cqe=%p byte_len=%u status=%s (%u/0x%x)",
-		__entry->cqe, __entry->byte_len,
-		rdma_show_wc_status(__entry->status),
-		__entry->status, __entry->vendor_err
+	TP_printk("addr=%s status=%d",
+		__get_str(addr), __entry->status
 	)
 );
 
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index c0587d3cd389..e6d7401232d2 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -117,6 +117,13 @@ svc_rdma_next_recv_ctxt(struct list_head *list)
 					rc_list);
 }
 
+static void svc_rdma_recv_cid_init(struct svcxprt_rdma *rdma,
+				   struct rpc_rdma_cid *cid)
+{
+	cid->ci_queue_id = rdma->sc_rq_cq->res.id;
+	cid->ci_completion_id = atomic_inc_return(&rdma->sc_completion_ids);
+}
+
 static struct svc_rdma_recv_ctxt *
 svc_rdma_recv_ctxt_alloc(struct svcxprt_rdma *rdma)
 {
@@ -135,6 +142,8 @@ svc_rdma_recv_ctxt_alloc(struct svcxprt_rdma *rdma)
 	if (ib_dma_mapping_error(rdma->sc_pd->device, addr))
 		goto fail2;
 
+	svc_rdma_recv_cid_init(rdma, &ctxt->rc_cid);
+
 	ctxt->rc_recv_wr.next = NULL;
 	ctxt->rc_recv_wr.wr_cqe = &ctxt->rc_cqe;
 	ctxt->rc_recv_wr.sg_list = &ctxt->rc_recv_sge;
@@ -249,13 +258,14 @@ static int __svc_rdma_post_recv(struct svcxprt_rdma *rdma,
 	int ret;
 
 	svc_xprt_get(&rdma->sc_xprt);
+	trace_svcrdma_post_recv(ctxt);
 	ret = ib_post_recv(rdma->sc_qp, &ctxt->rc_recv_wr, NULL);
-	trace_svcrdma_post_recv(&ctxt->rc_recv_wr, ret);
 	if (ret)
 		goto err_post;
 	return 0;
 
 err_post:
+	trace_svcrdma_rq_post_err(rdma, ret);
 	svc_rdma_recv_ctxt_put(rdma, ctxt);
 	svc_xprt_put(&rdma->sc_xprt);
 	return ret;
@@ -309,11 +319,10 @@ static void svc_rdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 	struct ib_cqe *cqe = wc->wr_cqe;
 	struct svc_rdma_recv_ctxt *ctxt;
 
-	trace_svcrdma_wc_receive(wc);
-
 	/* WARNING: Only wc->wr_cqe and wc->status are reliable */
 	ctxt = container_of(cqe, struct svc_rdma_recv_ctxt, rc_cqe);
 
+	trace_svcrdma_wc_receive(wc, &ctxt->rc_cid);
 	if (wc->status != IB_WC_SUCCESS)
 		goto flushed;
 
-- 
cgit v1.2.3


From 007140ee9b4fc4e59538677799c916890a2f13e2 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 29 Apr 2020 17:16:31 -0400
Subject: svcrdma: Record Receive completion ID in svc_rdma_decode_rqst

When recording a trace event in the Receive path, tie decoding
results and errors to an incoming Receive completion.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/trace/events/rpcrdma.h          | 34 ++++++++++++++++++++++++++-------
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 12 ++++++------
 2 files changed, 33 insertions(+), 13 deletions(-)

(limited to 'net')

diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index a0330a557e34..df49ae5d447b 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1369,13 +1369,16 @@ TRACE_DEFINE_ENUM(RDMA_ERROR);
 
 TRACE_EVENT(svcrdma_decode_rqst,
 	TP_PROTO(
+		const struct svc_rdma_recv_ctxt *ctxt,
 		__be32 *p,
 		unsigned int hdrlen
 	),
 
-	TP_ARGS(p, hdrlen),
+	TP_ARGS(ctxt, p, hdrlen),
 
 	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(int, completion_id)
 		__field(u32, xid)
 		__field(u32, vers)
 		__field(u32, proc)
@@ -1384,6 +1387,8 @@ TRACE_EVENT(svcrdma_decode_rqst,
 	),
 
 	TP_fast_assign(
+		__entry->cq_id = ctxt->rc_cid.ci_queue_id;
+		__entry->completion_id = ctxt->rc_cid.ci_completion_id;
 		__entry->xid = be32_to_cpup(p++);
 		__entry->vers = be32_to_cpup(p++);
 		__entry->credits = be32_to_cpup(p++);
@@ -1391,37 +1396,48 @@ TRACE_EVENT(svcrdma_decode_rqst,
 		__entry->hdrlen = hdrlen;
 	),
 
-	TP_printk("xid=0x%08x vers=%u credits=%u proc=%s hdrlen=%u",
+	TP_printk("cq.id=%u cid=%d xid=0x%08x vers=%u credits=%u proc=%s hdrlen=%u",
+		__entry->cq_id, __entry->completion_id,
 		__entry->xid, __entry->vers, __entry->credits,
 		show_rpcrdma_proc(__entry->proc), __entry->hdrlen)
 );
 
 TRACE_EVENT(svcrdma_decode_short_err,
 	TP_PROTO(
+		const struct svc_rdma_recv_ctxt *ctxt,
 		unsigned int hdrlen
 	),
 
-	TP_ARGS(hdrlen),
+	TP_ARGS(ctxt, hdrlen),
 
 	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(int, completion_id)
 		__field(unsigned int, hdrlen)
 	),
 
 	TP_fast_assign(
+		__entry->cq_id = ctxt->rc_cid.ci_queue_id;
+		__entry->completion_id = ctxt->rc_cid.ci_completion_id;
 		__entry->hdrlen = hdrlen;
 	),
 
-	TP_printk("hdrlen=%u", __entry->hdrlen)
+	TP_printk("cq.id=%u cid=%d hdrlen=%u",
+		__entry->cq_id, __entry->completion_id,
+		__entry->hdrlen)
 );
 
 DECLARE_EVENT_CLASS(svcrdma_badreq_event,
 	TP_PROTO(
+		const struct svc_rdma_recv_ctxt *ctxt,
 		__be32 *p
 	),
 
-	TP_ARGS(p),
+	TP_ARGS(ctxt, p),
 
 	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(int, completion_id)
 		__field(u32, xid)
 		__field(u32, vers)
 		__field(u32, proc)
@@ -1429,13 +1445,16 @@ DECLARE_EVENT_CLASS(svcrdma_badreq_event,
 	),
 
 	TP_fast_assign(
+		__entry->cq_id = ctxt->rc_cid.ci_queue_id;
+		__entry->completion_id = ctxt->rc_cid.ci_completion_id;
 		__entry->xid = be32_to_cpup(p++);
 		__entry->vers = be32_to_cpup(p++);
 		__entry->credits = be32_to_cpup(p++);
 		__entry->proc = be32_to_cpup(p);
 	),
 
-	TP_printk("xid=0x%08x vers=%u credits=%u proc=%u",
+	TP_printk("cq.id=%u cid=%d xid=0x%08x vers=%u credits=%u proc=%u",
+		__entry->cq_id, __entry->completion_id,
 		__entry->xid, __entry->vers, __entry->credits, __entry->proc)
 );
 
@@ -1443,9 +1462,10 @@ DECLARE_EVENT_CLASS(svcrdma_badreq_event,
 		DEFINE_EVENT(svcrdma_badreq_event,			\
 			     svcrdma_decode_##name##_err,		\
 				TP_PROTO(				\
+					const struct svc_rdma_recv_ctxt *ctxt,	\
 					__be32 *p			\
 				),					\
-				TP_ARGS(p))
+				TP_ARGS(ctxt, p))
 
 DEFINE_BADREQ_EVENT(badvers);
 DEFINE_BADREQ_EVENT(drop);
diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index e6d7401232d2..d5ec85cb652c 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -667,27 +667,27 @@ static int svc_rdma_xdr_decode_req(struct xdr_buf *rq_arg,
 	hdr_len = xdr_stream_pos(&rctxt->rc_stream);
 	rq_arg->head[0].iov_len -= hdr_len;
 	rq_arg->len -= hdr_len;
-	trace_svcrdma_decode_rqst(rdma_argp, hdr_len);
+	trace_svcrdma_decode_rqst(rctxt, rdma_argp, hdr_len);
 	return hdr_len;
 
 out_short:
-	trace_svcrdma_decode_short_err(rq_arg->len);
+	trace_svcrdma_decode_short_err(rctxt, rq_arg->len);
 	return -EINVAL;
 
 out_version:
-	trace_svcrdma_decode_badvers_err(rdma_argp);
+	trace_svcrdma_decode_badvers_err(rctxt, rdma_argp);
 	return -EPROTONOSUPPORT;
 
 out_drop:
-	trace_svcrdma_decode_drop_err(rdma_argp);
+	trace_svcrdma_decode_drop_err(rctxt, rdma_argp);
 	return 0;
 
 out_proc:
-	trace_svcrdma_decode_badproc_err(rdma_argp);
+	trace_svcrdma_decode_badproc_err(rctxt, rdma_argp);
 	return -EINVAL;
 
 out_inval:
-	trace_svcrdma_decode_parse_err(rdma_argp);
+	trace_svcrdma_decode_parse_err(rctxt, rdma_argp);
 	return -EINVAL;
 }
 
-- 
cgit v1.2.3


From 3ac56c2fb166fea25974d8c48bb4a72ee298361b Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Thu, 30 Apr 2020 13:47:07 -0400
Subject: svcrdma: Introduce Send completion IDs

Set up a completion ID in each svc_rdma_send_ctxt. The ID is used
to match an incoming Send completion to a transport and to a
previous ib_post_send().

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/svc_rdma.h       |  2 ++
 include/trace/events/rpcrdma.h        |  2 +-
 net/sunrpc/xprtrdma/svc_rdma_sendto.c | 15 ++++++++++++---
 3 files changed, 15 insertions(+), 4 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h
index c3c1e46f510f..c91e00bc937e 100644
--- a/include/linux/sunrpc/svc_rdma.h
+++ b/include/linux/sunrpc/svc_rdma.h
@@ -151,6 +151,8 @@ struct svc_rdma_recv_ctxt {
 
 struct svc_rdma_send_ctxt {
 	struct list_head	sc_list;
+	struct rpc_rdma_cid	sc_cid;
+
 	struct ib_send_wr	sc_send_wr;
 	struct ib_cqe		sc_cqe;
 	struct xdr_buf		sc_hdrbuf;
diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index df49ae5d447b..782a4d826a4b 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1863,7 +1863,7 @@ TRACE_EVENT(svcrdma_post_send,
 	)
 );
 
-DEFINE_SENDCOMP_EVENT(send);
+DEFINE_COMPLETION_EVENT(svcrdma_wc_send);
 
 TRACE_EVENT(svcrdma_post_recv,
 	TP_PROTO(
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 38d8f0ee35ec..c720dcf56231 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -122,6 +122,13 @@ svc_rdma_next_send_ctxt(struct list_head *list)
 					sc_list);
 }
 
+static void svc_rdma_send_cid_init(struct svcxprt_rdma *rdma,
+				   struct rpc_rdma_cid *cid)
+{
+	cid->ci_queue_id = rdma->sc_sq_cq->res.id;
+	cid->ci_completion_id = atomic_inc_return(&rdma->sc_completion_ids);
+}
+
 static struct svc_rdma_send_ctxt *
 svc_rdma_send_ctxt_alloc(struct svcxprt_rdma *rdma)
 {
@@ -144,6 +151,8 @@ svc_rdma_send_ctxt_alloc(struct svcxprt_rdma *rdma)
 	if (ib_dma_mapping_error(rdma->sc_pd->device, addr))
 		goto fail2;
 
+	svc_rdma_send_cid_init(rdma, &ctxt->sc_cid);
+
 	ctxt->sc_send_wr.next = NULL;
 	ctxt->sc_send_wr.wr_cqe = &ctxt->sc_cqe;
 	ctxt->sc_send_wr.sg_list = ctxt->sc_sges;
@@ -268,14 +277,14 @@ static void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 {
 	struct svcxprt_rdma *rdma = cq->cq_context;
 	struct ib_cqe *cqe = wc->wr_cqe;
-	struct svc_rdma_send_ctxt *ctxt;
+	struct svc_rdma_send_ctxt *ctxt =
+		container_of(cqe, struct svc_rdma_send_ctxt, sc_cqe);
 
-	trace_svcrdma_wc_send(wc);
+	trace_svcrdma_wc_send(wc, &ctxt->sc_cid);
 
 	atomic_inc(&rdma->sc_sq_avail);
 	wake_up(&rdma->sc_send_wait);
 
-	ctxt = container_of(cqe, struct svc_rdma_send_ctxt, sc_cqe);
 	svc_rdma_send_ctxt_put(rdma, ctxt);
 
 	if (unlikely(wc->status != IB_WC_SUCCESS)) {
-- 
cgit v1.2.3


From 17f70f8dd52be3723250d21093403bb3a9f2162f Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 29 Apr 2020 11:05:33 -0400
Subject: svcrdma: Record send_ctxt completion ID in trace_svcrdma_post_send()

First, refactor: Dereference the svc_rdma_send_ctxt inside
svc_rdma_send() instead of at every call site.

Then, it can be passed into trace_svcrdma_post_send() to get the
proper completion ID.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/linux/sunrpc/svc_rdma.h            |  3 ++-
 include/trace/events/rpcrdma.h             | 18 +++++++++++-------
 net/sunrpc/xprtrdma/svc_rdma_backchannel.c |  2 +-
 net/sunrpc/xprtrdma/svc_rdma_sendto.c      | 11 ++++++-----
 4 files changed, 20 insertions(+), 14 deletions(-)

(limited to 'net')

diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h
index c91e00bc937e..9dc3a3b88391 100644
--- a/include/linux/sunrpc/svc_rdma.h
+++ b/include/linux/sunrpc/svc_rdma.h
@@ -196,7 +196,8 @@ 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_send(struct svcxprt_rdma *rdma, struct ib_send_wr *wr);
+extern int svc_rdma_send(struct svcxprt_rdma *rdma,
+			 struct svc_rdma_send_ctxt *ctxt);
 extern int svc_rdma_map_reply_msg(struct svcxprt_rdma *rdma,
 				  struct svc_rdma_send_ctxt *sctxt,
 				  const struct svc_rdma_recv_ctxt *rctxt,
diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index 782a4d826a4b..aeeba9188ed5 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1839,27 +1839,31 @@ DECLARE_EVENT_CLASS(svcrdma_sendcomp_event,
 
 TRACE_EVENT(svcrdma_post_send,
 	TP_PROTO(
-		const struct ib_send_wr *wr
+		const struct svc_rdma_send_ctxt *ctxt
 	),
 
-	TP_ARGS(wr),
+	TP_ARGS(ctxt),
 
 	TP_STRUCT__entry(
-		__field(const void *, cqe)
+		__field(u32, cq_id)
+		__field(int, completion_id)
 		__field(unsigned int, num_sge)
 		__field(u32, inv_rkey)
 	),
 
 	TP_fast_assign(
-		__entry->cqe = wr->wr_cqe;
+		const struct ib_send_wr *wr = &ctxt->sc_send_wr;
+
+		__entry->cq_id = ctxt->sc_cid.ci_queue_id;
+		__entry->completion_id = ctxt->sc_cid.ci_completion_id;
 		__entry->num_sge = wr->num_sge;
 		__entry->inv_rkey = (wr->opcode == IB_WR_SEND_WITH_INV) ?
 					wr->ex.invalidate_rkey : 0;
 	),
 
-	TP_printk("cqe=%p num_sge=%u inv_rkey=0x%08x",
-		__entry->cqe, __entry->num_sge,
-		__entry->inv_rkey
+	TP_printk("cq_id=%u cid=%d num_sge=%u inv_rkey=0x%08x",
+		__entry->cq_id, __entry->completion_id,
+		__entry->num_sge, __entry->inv_rkey
 	)
 );
 
diff --git a/net/sunrpc/xprtrdma/svc_rdma_backchannel.c b/net/sunrpc/xprtrdma/svc_rdma_backchannel.c
index 1ee73f7cf931..5e7c4ba9e147 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_backchannel.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_backchannel.c
@@ -87,7 +87,7 @@ static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma,
 	 */
 	get_page(virt_to_page(rqst->rq_buffer));
 	ctxt->sc_send_wr.opcode = IB_WR_SEND;
-	return svc_rdma_send(rdma, &ctxt->sc_send_wr);
+	return svc_rdma_send(rdma, ctxt);
 }
 
 /* Server-side transport endpoint wants a whole page for its send
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index c720dcf56231..73d46e8cdc16 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -298,13 +298,14 @@ static void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 /**
  * svc_rdma_send - Post a single Send WR
  * @rdma: transport on which to post the WR
- * @wr: prepared Send WR to post
+ * @ctxt: send ctxt with a Send WR ready to post
  *
  * Returns zero the Send WR was posted successfully. Otherwise, a
  * negative errno is returned.
  */
-int svc_rdma_send(struct svcxprt_rdma *rdma, struct ib_send_wr *wr)
+int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 {
+	struct ib_send_wr *wr = &ctxt->sc_send_wr;
 	int ret;
 
 	might_sleep();
@@ -330,7 +331,7 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct ib_send_wr *wr)
 		}
 
 		svc_xprt_get(&rdma->sc_xprt);
-		trace_svcrdma_post_send(wr);
+		trace_svcrdma_post_send(ctxt);
 		ret = ib_post_send(rdma->sc_qp, wr, NULL);
 		if (ret)
 			break;
@@ -805,7 +806,7 @@ static int svc_rdma_send_reply_msg(struct svcxprt_rdma *rdma,
 	} else {
 		sctxt->sc_send_wr.opcode = IB_WR_SEND;
 	}
-	return svc_rdma_send(rdma, &sctxt->sc_send_wr);
+	return svc_rdma_send(rdma, sctxt);
 }
 
 /**
@@ -869,7 +870,7 @@ void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
 	sctxt->sc_send_wr.num_sge = 1;
 	sctxt->sc_send_wr.opcode = IB_WR_SEND;
 	sctxt->sc_sges[0].length = sctxt->sc_hdrbuf.len;
-	if (svc_rdma_send(rdma, &sctxt->sc_send_wr))
+	if (svc_rdma_send(rdma, sctxt))
 		goto put_ctxt;
 	return;
 
-- 
cgit v1.2.3


From 6787f0bea27a24e4c306616565b02234ee558cfb Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Wed, 29 Apr 2020 17:25:36 -0400
Subject: svcrdma: Display chunk completion ID when posting a rw_ctxt

Re-use the post_rw tracepoint (safely) to trace cc_info lifetime
events, including completion IDs.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/trace/events/rpcrdma.h    | 56 +++++++++------------------------------
 net/sunrpc/xprtrdma/svc_rdma_rw.c | 14 ++++++++--
 2 files changed, 24 insertions(+), 46 deletions(-)

(limited to 'net')

diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index aeeba9188ed5..abe942225637 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -1802,41 +1802,6 @@ TRACE_EVENT(svcrdma_send_err,
 	)
 );
 
-DECLARE_EVENT_CLASS(svcrdma_sendcomp_event,
-	TP_PROTO(
-		const struct ib_wc *wc
-	),
-
-	TP_ARGS(wc),
-
-	TP_STRUCT__entry(
-		__field(const void *, cqe)
-		__field(unsigned int, status)
-		__field(unsigned int, vendor_err)
-	),
-
-	TP_fast_assign(
-		__entry->cqe = wc->wr_cqe;
-		__entry->status = wc->status;
-		if (wc->status)
-			__entry->vendor_err = wc->vendor_err;
-		else
-			__entry->vendor_err = 0;
-	),
-
-	TP_printk("cqe=%p status=%s (%u/0x%x)",
-		__entry->cqe, rdma_show_wc_status(__entry->status),
-		__entry->status, __entry->vendor_err
-	)
-);
-
-#define DEFINE_SENDCOMP_EVENT(name)					\
-		DEFINE_EVENT(svcrdma_sendcomp_event, svcrdma_wc_##name,	\
-				TP_PROTO(				\
-					const struct ib_wc *wc		\
-				),					\
-				TP_ARGS(wc))
-
 TRACE_EVENT(svcrdma_post_send,
 	TP_PROTO(
 		const struct svc_rdma_send_ctxt *ctxt
@@ -1916,31 +1881,34 @@ TRACE_EVENT(svcrdma_rq_post_err,
 	)
 );
 
-TRACE_EVENT(svcrdma_post_rw,
+TRACE_EVENT(svcrdma_post_chunk,
 	TP_PROTO(
-		const void *cqe,
+		const struct rpc_rdma_cid *cid,
 		int sqecount
 	),
 
-	TP_ARGS(cqe, sqecount),
+	TP_ARGS(cid, sqecount),
 
 	TP_STRUCT__entry(
-		__field(const void *, cqe)
+		__field(u32, cq_id)
+		__field(int, completion_id)
 		__field(int, sqecount)
 	),
 
 	TP_fast_assign(
-		__entry->cqe = cqe;
+		__entry->cq_id = cid->ci_queue_id;
+		__entry->completion_id = cid->ci_completion_id;
 		__entry->sqecount = sqecount;
 	),
 
-	TP_printk("cqe=%p sqecount=%d",
-		__entry->cqe, __entry->sqecount
+	TP_printk("cq.id=%u cid=%d sqecount=%d",
+		__entry->cq_id, __entry->completion_id,
+		__entry->sqecount
 	)
 );
 
-DEFINE_SENDCOMP_EVENT(read);
-DEFINE_SENDCOMP_EVENT(write);
+DEFINE_COMPLETION_EVENT(svcrdma_wc_read);
+DEFINE_COMPLETION_EVENT(svcrdma_wc_write);
 
 TRACE_EVENT(svcrdma_qp_error,
 	TP_PROTO(
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index 2038b1b286dd..c16d10601d65 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -145,15 +145,24 @@ static int svc_rdma_rw_ctx_init(struct svcxprt_rdma *rdma,
  * demand, and not cached.
  */
 struct svc_rdma_chunk_ctxt {
+	struct rpc_rdma_cid	cc_cid;
 	struct ib_cqe		cc_cqe;
 	struct svcxprt_rdma	*cc_rdma;
 	struct list_head	cc_rwctxts;
 	int			cc_sqecount;
 };
 
+static void svc_rdma_cc_cid_init(struct svcxprt_rdma *rdma,
+				 struct rpc_rdma_cid *cid)
+{
+	cid->ci_queue_id = rdma->sc_sq_cq->res.id;
+	cid->ci_completion_id = atomic_inc_return(&rdma->sc_completion_ids);
+}
+
 static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
 			     struct svc_rdma_chunk_ctxt *cc)
 {
+	svc_rdma_cc_cid_init(rdma, &cc->cc_cid);
 	cc->cc_rdma = rdma;
 	svc_xprt_get(&rdma->sc_xprt);
 
@@ -237,7 +246,7 @@ static void svc_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc)
 	struct svc_rdma_write_info *info =
 			container_of(cc, struct svc_rdma_write_info, wi_cc);
 
-	trace_svcrdma_wc_write(wc);
+	trace_svcrdma_wc_write(wc, &cc->cc_cid);
 
 	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
 	wake_up(&rdma->sc_send_wait);
@@ -295,7 +304,7 @@ static void svc_rdma_wc_read_done(struct ib_cq *cq, struct ib_wc *wc)
 	struct svc_rdma_read_info *info =
 			container_of(cc, struct svc_rdma_read_info, ri_cc);
 
-	trace_svcrdma_wc_read(wc);
+	trace_svcrdma_wc_read(wc, &cc->cc_cid);
 
 	atomic_add(cc->cc_sqecount, &rdma->sc_sq_avail);
 	wake_up(&rdma->sc_send_wait);
@@ -351,6 +360,7 @@ static int svc_rdma_post_chunk_ctxt(struct svc_rdma_chunk_ctxt *cc)
 	do {
 		if (atomic_sub_return(cc->cc_sqecount,
 				      &rdma->sc_sq_avail) > 0) {
+			trace_svcrdma_post_chunk(&cc->cc_cid, cc->cc_sqecount);
 			ret = ib_post_send(rdma->sc_qp, first_wr, &bad_wr);
 			if (ret)
 				break;
-- 
cgit v1.2.3


From 986a4b63d3bc5f2c0eb4083b05aff2bf883b7b2f Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Fri, 24 Jul 2020 17:08:57 -0400
Subject: SUNRPC: Fix ("SUNRPC: Add "@len" parameter to gss_unwrap()")

Braino when converting "buf->len -=" to "buf->len = len -".

The result is under-estimation of the ralign and rslack values. On
krb5p mounts, this has caused READDIR to fail with EIO, and KASAN
splats when decoding READLINK replies.

As a result of fixing this oversight, the gss_unwrap method now
returns a buf->len that can be shorter than priv_len for small
RPC messages. The additional adjustment done in unwrap_priv_data()
can underflow buf->len. This causes the nfsd_request_too_large
check to fail during some NFSv3 operations.

Reported-by: Marian Rainer-Harbach
Reported-by: Pierre Sauter <pierre.sauter@stwm.de>
BugLink: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1886277
Fixes: 31c9590ae468 ("SUNRPC: Add "@len" parameter to gss_unwrap()")
Reviewed-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/auth_gss/gss_krb5_wrap.c | 2 +-
 net/sunrpc/auth_gss/svcauth_gss.c   | 1 -
 2 files changed, 1 insertion(+), 2 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/auth_gss/gss_krb5_wrap.c b/net/sunrpc/auth_gss/gss_krb5_wrap.c
index cf0fd170ac18..90b8329fef82 100644
--- a/net/sunrpc/auth_gss/gss_krb5_wrap.c
+++ b/net/sunrpc/auth_gss/gss_krb5_wrap.c
@@ -584,7 +584,7 @@ gss_unwrap_kerberos_v2(struct krb5_ctx *kctx, int offset, int len,
 							buf->head[0].iov_len);
 	memmove(ptr, ptr + GSS_KRB5_TOK_HDR_LEN + headskip, movelen);
 	buf->head[0].iov_len -= GSS_KRB5_TOK_HDR_LEN + headskip;
-	buf->len = len - GSS_KRB5_TOK_HDR_LEN + headskip;
+	buf->len = len - (GSS_KRB5_TOK_HDR_LEN + headskip);
 
 	/* Trim off the trailing "extra count" and checksum blob */
 	xdr_buf_trim(buf, ec + GSS_KRB5_TOK_HDR_LEN + tailskip);
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index 7d83f54aaaa6..258b04372f85 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -990,7 +990,6 @@ unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gs
 
 	maj_stat = gss_unwrap(ctx, 0, priv_len, buf);
 	pad = priv_len - buf->len;
-	buf->len -= pad;
 	/* The upper layers assume the buffer is aligned on 4-byte boundaries.
 	 * In the krb5p case, at least, the data ends up offset, so we need to
 	 * move it around. */
-- 
cgit v1.2.3


From 64d26422516b2e347b32e6d9b1d40b3c19a62aae Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 30 Jun 2020 15:55:45 -0400
Subject: svcrdma: Fix another Receive buffer leak

During a connection tear down, the Receive queue is flushed before
the device resources are freed. Typically, all the Receives flush
with IB_WR_FLUSH_ERR.

However, any pending successful Receives flush with IB_WR_SUCCESS,
and the server automatically posts a fresh Receive to replace the
completing one. This happens even after the connection has closed
and the RQ is drained. Receives that are posted after the RQ is
drained appear never to complete, causing a Receive resource leak.
The leaked Receive buffer is left DMA-mapped.

To prevent these late-posted recv_ctxt's from leaking, block new
Receive posting after XPT_CLOSE is set.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index d5ec85cb652c..5bb97b5f4606 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -275,6 +275,8 @@ static int svc_rdma_post_recv(struct svcxprt_rdma *rdma)
 {
 	struct svc_rdma_recv_ctxt *ctxt;
 
+	if (test_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags))
+		return 0;
 	ctxt = svc_rdma_recv_ctxt_get(rdma);
 	if (!ctxt)
 		return -ENOMEM;
-- 
cgit v1.2.3


From 365e9992b90fbbfda89f72866783fd52bbd58e64 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 30 Jun 2020 13:25:41 -0400
Subject: svcrdma: Remove transport reference counting

Jason tells me that a ULP cannot rely on getting an ESTABLISHED
and DISCONNECTED event pair for each connection, so transport
reference counting in the CM event handler will never be reliable.

Now that we have ib_drain_qp(), svcrdma should no longer need to
hold transport references while Sends and Receives are posted. So
remove the get/put call sites in the CM event handlers.

This eliminates a significant source of locked memory bus traffic.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c  |  6 +-----
 net/sunrpc/xprtrdma/svc_rdma_rw.c        |  2 --
 net/sunrpc/xprtrdma/svc_rdma_sendto.c    |  4 ----
 net/sunrpc/xprtrdma/svc_rdma_transport.c | 17 +----------------
 4 files changed, 2 insertions(+), 27 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 5bb97b5f4606..c6ea2903c21a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -257,7 +257,6 @@ static int __svc_rdma_post_recv(struct svcxprt_rdma *rdma,
 {
 	int ret;
 
-	svc_xprt_get(&rdma->sc_xprt);
 	trace_svcrdma_post_recv(ctxt);
 	ret = ib_post_recv(rdma->sc_qp, &ctxt->rc_recv_wr, NULL);
 	if (ret)
@@ -267,7 +266,6 @@ static int __svc_rdma_post_recv(struct svcxprt_rdma *rdma,
 err_post:
 	trace_svcrdma_rq_post_err(rdma, ret);
 	svc_rdma_recv_ctxt_put(rdma, ctxt);
-	svc_xprt_put(&rdma->sc_xprt);
 	return ret;
 }
 
@@ -344,15 +342,13 @@ static void svc_rdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 	spin_unlock(&rdma->sc_rq_dto_lock);
 	if (!test_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags))
 		svc_xprt_enqueue(&rdma->sc_xprt);
-	goto out;
+	return;
 
 flushed:
 post_err:
 	svc_rdma_recv_ctxt_put(rdma, ctxt);
 	set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
 	svc_xprt_enqueue(&rdma->sc_xprt);
-out:
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /**
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index c16d10601d65..fe54cbe97a46 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -164,7 +164,6 @@ static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
 {
 	svc_rdma_cc_cid_init(rdma, &cc->cc_cid);
 	cc->cc_rdma = rdma;
-	svc_xprt_get(&rdma->sc_xprt);
 
 	INIT_LIST_HEAD(&cc->cc_rwctxts);
 	cc->cc_sqecount = 0;
@@ -184,7 +183,6 @@ static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc,
 				    ctxt->rw_nents, dir);
 		svc_rdma_put_rw_ctxt(rdma, ctxt);
 	}
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /* State for sending a Write or Reply chunk.
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 73d46e8cdc16..7b94d971feb3 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -291,8 +291,6 @@ static void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
 		svc_xprt_enqueue(&rdma->sc_xprt);
 	}
-
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /**
@@ -330,7 +328,6 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 			continue;
 		}
 
-		svc_xprt_get(&rdma->sc_xprt);
 		trace_svcrdma_post_send(ctxt);
 		ret = ib_post_send(rdma->sc_qp, wr, NULL);
 		if (ret)
@@ -340,7 +337,6 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 
 	trace_svcrdma_sq_post_err(rdma, ret);
 	set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
-	svc_xprt_put(&rdma->sc_xprt);
 	wake_up(&rdma->sc_send_wait);
 	return ret;
 }
diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index 3da7901a49e6..aa60f75c8c1d 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -271,7 +271,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 	switch (event->event) {
 	case RDMA_CM_EVENT_ESTABLISHED:
 		/* Accept complete */
-		svc_xprt_get(xprt);
 		dprintk("svcrdma: Connection completed on DTO xprt=%p, "
 			"cm_id=%p\n", xprt, cma_id);
 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
@@ -282,7 +281,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 			xprt, cma_id);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
-		svc_xprt_put(xprt);
 		break;
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
 		dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, "
@@ -290,7 +288,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 			rdma_event_msg(event->event), event->event);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
-		svc_xprt_put(xprt);
 		break;
 	default:
 		dprintk("svcrdma: Unexpected event on DTO endpoint %p, "
@@ -539,24 +536,11 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	return NULL;
 }
 
-/*
- * When connected, an svc_xprt has at least two references:
- *
- * - A reference held by the cm_id between the ESTABLISHED and
- *   DISCONNECTED events. If the remote peer disconnected first, this
- *   reference could be gone.
- *
- * - A reference held by the svc_recv code that called this function
- *   as part of close processing.
- *
- * At a minimum one references should still be held.
- */
 static void svc_rdma_detach(struct svc_xprt *xprt)
 {
 	struct svcxprt_rdma *rdma =
 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
 
-	/* Disconnect and flush posted WQE */
 	rdma_disconnect(rdma->sc_cm_id);
 }
 
@@ -566,6 +550,7 @@ static void __svc_rdma_free(struct work_struct *work)
 		container_of(work, struct svcxprt_rdma, sc_work);
 	struct svc_xprt *xprt = &rdma->sc_xprt;
 
+	/* This blocks until the Completion Queues are empty */
 	if (rdma->sc_qp && !IS_ERR(rdma->sc_qp))
 		ib_drain_qp(rdma->sc_qp);
 
-- 
cgit v1.2.3


From b297fed699ad9e50315b27e78de42ac631c9990d Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 30 Jun 2020 17:16:35 -0400
Subject: svcrdma: CM event handler clean up

Now that there's a core tracepoint that reports these events, there's
no need to maintain dprintk() call sites in each arm of the switch
statements.

We also refresh the documenting comments.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_transport.c | 56 ++++++++++++++------------------
 1 file changed, 25 insertions(+), 31 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index aa60f75c8c1d..fb044792b571 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -237,62 +237,56 @@ static void handle_connect_req(struct rdma_cm_id *new_cma_id,
 	svc_xprt_enqueue(&listen_xprt->sc_xprt);
 }
 
-/*
- * Handles events generated on the listening endpoint. These events will be
- * either be incoming connect requests or adapter removal  events.
+/**
+ * svc_rdma_listen_handler - Handle CM events generated on a listening endpoint
+ * @cma_id: the server's listener rdma_cm_id
+ * @event: details of the event
+ *
+ * Return values:
+ *     %0: Do not destroy @cma_id
+ *     %1: Destroy @cma_id (never returned here)
+ *
+ * NB: There is never a DEVICE_REMOVAL event for INADDR_ANY listeners.
  */
-static int rdma_listen_handler(struct rdma_cm_id *cma_id,
-			       struct rdma_cm_event *event)
+static int svc_rdma_listen_handler(struct rdma_cm_id *cma_id,
+				   struct rdma_cm_event *event)
 {
 	switch (event->event) {
 	case RDMA_CM_EVENT_CONNECT_REQUEST:
-		dprintk("svcrdma: Connect request on cma_id=%p, xprt = %p, "
-			"event = %s (%d)\n", cma_id, cma_id->context,
-			rdma_event_msg(event->event), event->event);
 		handle_connect_req(cma_id, &event->param.conn);
 		break;
 	default:
-		/* NB: No device removal upcall for INADDR_ANY listeners */
-		dprintk("svcrdma: Unexpected event on listening endpoint %p, "
-			"event = %s (%d)\n", cma_id,
-			rdma_event_msg(event->event), event->event);
 		break;
 	}
-
 	return 0;
 }
 
-static int rdma_cma_handler(struct rdma_cm_id *cma_id,
-			    struct rdma_cm_event *event)
+/**
+ * svc_rdma_cma_handler - Handle CM events on client connections
+ * @cma_id: the server's listener rdma_cm_id
+ * @event: details of the event
+ *
+ * Return values:
+ *     %0: Do not destroy @cma_id
+ *     %1: Destroy @cma_id (never returned here)
+ */
+static int svc_rdma_cma_handler(struct rdma_cm_id *cma_id,
+				struct rdma_cm_event *event)
 {
 	struct svcxprt_rdma *rdma = cma_id->context;
 	struct svc_xprt *xprt = &rdma->sc_xprt;
 
 	switch (event->event) {
 	case RDMA_CM_EVENT_ESTABLISHED:
-		/* Accept complete */
-		dprintk("svcrdma: Connection completed on DTO xprt=%p, "
-			"cm_id=%p\n", xprt, cma_id);
 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
 		svc_xprt_enqueue(xprt);
 		break;
 	case RDMA_CM_EVENT_DISCONNECTED:
-		dprintk("svcrdma: Disconnect on DTO xprt=%p, cm_id=%p\n",
-			xprt, cma_id);
-		set_bit(XPT_CLOSE, &xprt->xpt_flags);
-		svc_xprt_enqueue(xprt);
-		break;
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
-		dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, "
-			"event = %s (%d)\n", cma_id, xprt,
-			rdma_event_msg(event->event), event->event);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
 		break;
 	default:
-		dprintk("svcrdma: Unexpected event on DTO endpoint %p, "
-			"event = %s (%d)\n", cma_id,
-			rdma_event_msg(event->event), event->event);
 		break;
 	}
 	return 0;
@@ -318,7 +312,7 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
 	set_bit(XPT_LISTENER, &cma_xprt->sc_xprt.xpt_flags);
 	strcpy(cma_xprt->sc_xprt.xpt_remotebuf, "listener");
 
-	listen_id = rdma_create_id(net, rdma_listen_handler, cma_xprt,
+	listen_id = rdma_create_id(net, svc_rdma_listen_handler, cma_xprt,
 				   RDMA_PS_TCP, IB_QPT_RC);
 	if (IS_ERR(listen_id)) {
 		ret = PTR_ERR(listen_id);
@@ -482,7 +476,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 		goto errout;
 
 	/* Swap out the handler */
-	newxprt->sc_cm_id->event_handler = rdma_cma_handler;
+	newxprt->sc_cm_id->event_handler = svc_rdma_cma_handler;
 
 	/* Construct RDMA-CM private message */
 	pmsg.cp_magic = rpcrdma_cmp_magic;
-- 
cgit v1.2.3