Merge tag 'nfsd-4.6' of git://linux-nfs.org/~bfields/linux

Pull nfsd updates from Bruce Fields:
 "Various bugfixes, a RDMA update from Chuck Lever, and support for a
  new pnfs layout type from Christoph Hellwig.  The new layout type is a
  variant of the block layout which uses SCSI features to offer improved
  fencing and device identification.

  (Also: note this pull request also includes the client side of SCSI
  layout, with Trond's permission.)"

* tag 'nfsd-4.6' of git://linux-nfs.org/~bfields/linux:
  sunrpc/cache: drop reference when sunrpc_cache_pipe_upcall() detects a race
  nfsd: recover: fix memory leak
  nfsd: fix deadlock secinfo+readdir compound
  nfsd4: resfh unused in nfsd4_secinfo
  svcrdma: Use new CQ API for RPC-over-RDMA server send CQs
  svcrdma: Use new CQ API for RPC-over-RDMA server receive CQs
  svcrdma: Remove close_out exit path
  svcrdma: Hook up the logic to return ERR_CHUNK
  svcrdma: Use correct XID in error replies
  svcrdma: Make RDMA_ERROR messages work
  rpcrdma: Add RPCRDMA_HDRLEN_ERR
  svcrdma: svc_rdma_post_recv() should close connection on error
  svcrdma: Close connection when a send error occurs
  nfsd: Lower NFSv4.1 callback message size limit
  svcrdma: Do not send Write chunk XDR pad with inline content
  svcrdma: Do not write xdr_buf::tail in a Write chunk
  svcrdma: Find client-provided write and reply chunks once per reply
  nfsd: Update NFS server comments related to RDMA support
  nfsd: Fix a memory leak when meeting unsupported state_protect_how4
  nfsd4: fix bad bounds checking

fs/nfsd/nfs4proc.c

@@ -864,12 +864,10 @@ static __be32
 nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	      struct nfsd4_secinfo *secinfo)
 {
-	struct svc_fh resfh;
 	struct svc_export *exp;
 	struct dentry *dentry;
 	__be32 err;
 
-	fh_init(&resfh, NFS4_FHSIZE);
 	err = fh_verify(rqstp, &cstate->current_fh, S_IFDIR, NFSD_MAY_EXEC);
 	if (err)
 		return err;
@@ -878,6 +876,7 @@ nfsd4_secinfo(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 				    &exp, &dentry);
 	if (err)
 		return err;
+	fh_unlock(&cstate->current_fh);
 	if (d_really_is_negative(dentry)) {
 		exp_put(exp);
 		err = nfserr_noent;

fs/nfsd/nfs4recover.c

@@ -1266,6 +1266,7 @@ nfsd4_umh_cltrack_init(struct net *net)
 	/* XXX: The usermode helper s not working in container yet. */
 	if (net != &init_net) {
 		pr_warn("NFSD: attempt to initialize umh client tracking in a container ignored.\n");
+		kfree(grace_start);
 		return -EINVAL;
 	}
 

fs/nfsd/nfs4state.c

@@ -2408,7 +2408,8 @@ nfsd4_exchange_id(struct svc_rqst *rqstp,
 	default:				/* checked by xdr code */
 		WARN_ON_ONCE(1);
 	case SP4_SSV:
-		return nfserr_encr_alg_unsupp;
+		status = nfserr_encr_alg_unsupp;
+		goto out_nolock;
 	}
 
 	/* Cases below refer to rfc 5661 section 18.35.4: */
@@ -2586,21 +2587,26 @@ static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfs
 	return nfs_ok;
 }
 
+/*
+ * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
+ * These are based on similar macros in linux/sunrpc/msg_prot.h .
+ */
+#define RPC_MAX_HEADER_WITH_AUTH_SYS \
+	(RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
+
+#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
+	(RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
+
 #define NFSD_CB_MAX_REQ_SZ	((NFS4_enc_cb_recall_sz + \
-				 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
+				 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
 #define NFSD_CB_MAX_RESP_SZ	((NFS4_dec_cb_recall_sz + \
-				 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
+				 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
+				 sizeof(__be32))
 
 static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
 {
 	ca->headerpadsz = 0;
 
-	/*
-	 * These RPC_MAX_HEADER macros are overkill, especially since we
-	 * don't even do gss on the backchannel yet.  But this is still
-	 * less than 1k.  Tighten up this estimate in the unlikely event
-	 * it turns out to be a problem for some client:
-	 */
 	if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
 		return nfserr_toosmall;
 	if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
@@ -2710,10 +2716,9 @@ nfsd4_create_session(struct svc_rqst *rqstp,
 		goto out_free_conn;
 	}
 	status = nfs_ok;
-	/*
-	 * We do not support RDMA or persistent sessions
-	 */
+	/* Persistent sessions are not supported */
 	cr_ses->flags &= ~SESSION4_PERSIST;
+	/* Upshifting from TCP to RDMA is not supported */
 	cr_ses->flags &= ~SESSION4_RDMA;
 
 	init_session(rqstp, new, conf, cr_ses);

fs/nfsd/nfs4xdr.c

@@ -1072,8 +1072,9 @@ nfsd4_decode_rename(struct nfsd4_compoundargs *argp, struct nfsd4_rename *rename
 
 	READ_BUF(4);
 	rename->rn_snamelen = be32_to_cpup(p++);
-	READ_BUF(rename->rn_snamelen + 4);
+	READ_BUF(rename->rn_snamelen);
 	SAVEMEM(rename->rn_sname, rename->rn_snamelen);
+	READ_BUF(4);
 	rename->rn_tnamelen = be32_to_cpup(p++);
 	READ_BUF(rename->rn_tnamelen);
 	SAVEMEM(rename->rn_tname, rename->rn_tnamelen);
@@ -1155,13 +1156,14 @@ nfsd4_decode_setclientid(struct nfsd4_compoundargs *argp, struct nfsd4_setclient
 	READ_BUF(8);
 	setclientid->se_callback_prog = be32_to_cpup(p++);
 	setclientid->se_callback_netid_len = be32_to_cpup(p++);
-
-	READ_BUF(setclientid->se_callback_netid_len + 4);
+	READ_BUF(setclientid->se_callback_netid_len);
 	SAVEMEM(setclientid->se_callback_netid_val, setclientid->se_callback_netid_len);
+	READ_BUF(4);
 	setclientid->se_callback_addr_len = be32_to_cpup(p++);
 
-	READ_BUF(setclientid->se_callback_addr_len + 4);
+	READ_BUF(setclientid->se_callback_addr_len);
 	SAVEMEM(setclientid->se_callback_addr_val, setclientid->se_callback_addr_len);
+	READ_BUF(4);
 	setclientid->se_callback_ident = be32_to_cpup(p++);
 
 	DECODE_TAIL;
@@ -1835,8 +1837,9 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
 
 	READ_BUF(4);
 	argp->taglen = be32_to_cpup(p++);
-	READ_BUF(argp->taglen + 8);
+	READ_BUF(argp->taglen);
 	SAVEMEM(argp->tag, argp->taglen);
+	READ_BUF(8);
 	argp->minorversion = be32_to_cpup(p++);
 	argp->opcnt = be32_to_cpup(p++);
 	max_reply += 4 + (XDR_QUADLEN(argp->taglen) << 2);
@@ -3060,7 +3063,7 @@ static __be32 nfsd4_encode_bind_conn_to_session(struct nfsd4_compoundres *resp,
 		p = xdr_encode_opaque_fixed(p, bcts->sessionid.data,
 						NFS4_MAX_SESSIONID_LEN);
 		*p++ = cpu_to_be32(bcts->dir);
-		/* Sorry, we do not yet support RDMA over 4.1: */
+		/* Upshifting from TCP to RDMA is not supported */
 		*p++ = cpu_to_be32(0);
 	}
 	return nfserr;

include/linux/sunrpc/auth.h

@@ -20,11 +20,18 @@
 #include <linux/uidgid.h>
 #include <linux/utsname.h>
 
+/*
+ * Maximum size of AUTH_NONE authentication information, in XDR words.
+ */
+#define NUL_CALLSLACK	(4)
+#define NUL_REPLYSLACK	(2)
+
 /*
  * Size of the nodename buffer. RFC1831 specifies a hard limit of 255 bytes,
  * but Linux hostnames are actually limited to __NEW_UTS_LEN bytes.
  */
 #define UNX_MAXNODENAME	__NEW_UTS_LEN
+#define UNX_CALLSLACK	(21 + XDR_QUADLEN(UNX_MAXNODENAME))
 
 struct rpcsec_gss_info;
 

include/linux/sunrpc/svc_rdma.h

@@ -75,8 +75,10 @@ struct svc_rdma_op_ctxt {
 	struct svc_rdma_fastreg_mr *frmr;
 	int hdr_count;
 	struct xdr_buf arg;
+	struct ib_cqe cqe;
+	struct ib_cqe reg_cqe;
+	struct ib_cqe inv_cqe;
 	struct list_head dto_q;
-	enum ib_wr_opcode wr_op;
 	enum ib_wc_status wc_status;
 	u32 byte_len;
 	u32 position;
@@ -174,8 +176,6 @@ struct svcxprt_rdma {
 	struct work_struct   sc_work;
 };
 /* sc_flags */
-#define RDMAXPRT_RQ_PENDING	1
-#define RDMAXPRT_SQ_PENDING	2
 #define RDMAXPRT_CONN_PENDING	3
 
 #define RPCRDMA_LISTEN_BACKLOG  10
@@ -199,7 +199,7 @@ extern int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt,
 				    struct xdr_buf *rcvbuf);
 
 /* svc_rdma_marshal.c */
-extern int svc_rdma_xdr_decode_req(struct rpcrdma_msg **, struct svc_rqst *);
+extern int svc_rdma_xdr_decode_req(struct rpcrdma_msg *, struct svc_rqst *);
 extern int svc_rdma_xdr_encode_error(struct svcxprt_rdma *,
 				     struct rpcrdma_msg *,
 				     enum rpcrdma_errcode, __be32 *);
@@ -224,16 +224,22 @@ extern int rdma_read_chunk_frmr(struct svcxprt_rdma *, struct svc_rqst *,
 
 /* svc_rdma_sendto.c */
 extern int svc_rdma_map_xdr(struct svcxprt_rdma *, struct xdr_buf *,
-			    struct svc_rdma_req_map *);
+			    struct svc_rdma_req_map *, bool);
 extern int svc_rdma_sendto(struct svc_rqst *);
 extern struct rpcrdma_read_chunk *
 	svc_rdma_get_read_chunk(struct rpcrdma_msg *);
+extern void svc_rdma_send_error(struct svcxprt_rdma *, struct rpcrdma_msg *,
+				int);
 
 /* svc_rdma_transport.c */
+extern void svc_rdma_wc_send(struct ib_cq *, struct ib_wc *);
+extern void svc_rdma_wc_write(struct ib_cq *, struct ib_wc *);
+extern void svc_rdma_wc_reg(struct ib_cq *, struct ib_wc *);
+extern void svc_rdma_wc_read(struct ib_cq *, struct ib_wc *);
+extern void svc_rdma_wc_inv(struct ib_cq *, struct ib_wc *);
 extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *);
-extern void svc_rdma_send_error(struct svcxprt_rdma *, struct rpcrdma_msg *,
-				enum rpcrdma_errcode);
 extern int svc_rdma_post_recv(struct svcxprt_rdma *, gfp_t);
+extern int svc_rdma_repost_recv(struct svcxprt_rdma *, gfp_t);
 extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *);
 extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *);
 extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int);

net/sunrpc/auth_null.c

@@ -113,8 +113,8 @@ const struct rpc_authops authnull_ops = {
 
 static
 struct rpc_auth null_auth = {
-	.au_cslack	= 4,
-	.au_rslack	= 2,
+	.au_cslack	= NUL_CALLSLACK,
+	.au_rslack	= NUL_REPLYSLACK,
 	.au_ops		= &authnull_ops,
 	.au_flavor	= RPC_AUTH_NULL,
 	.au_count	= ATOMIC_INIT(0),

net/sunrpc/auth_unix.c

@@ -23,8 +23,6 @@ struct unx_cred {
 };
 #define uc_uid			uc_base.cr_uid
 
-#define UNX_WRITESLACK		(21 + XDR_QUADLEN(UNX_MAXNODENAME))
-
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 # define RPCDBG_FACILITY	RPCDBG_AUTH
 #endif
@@ -228,8 +226,8 @@ const struct rpc_authops authunix_ops = {
 
 static
 struct rpc_auth		unix_auth = {
-	.au_cslack	= UNX_WRITESLACK,
-	.au_rslack	= 2,			/* assume AUTH_NULL verf */
+	.au_cslack	= UNX_CALLSLACK,
+	.au_rslack	= NUL_REPLYSLACK,
 	.au_ops		= &authunix_ops,
 	.au_flavor	= RPC_AUTH_UNIX,
 	.au_count	= ATOMIC_INIT(0),

net/sunrpc/cache.c

@@ -1182,14 +1182,14 @@ int sunrpc_cache_pipe_upcall(struct cache_detail *detail, struct cache_head *h)
 	}
 
 	crq->q.reader = 0;
-	crq->item = cache_get(h);
 	crq->buf = buf;
 	crq->len = 0;
 	crq->readers = 0;
 	spin_lock(&queue_lock);
-	if (test_bit(CACHE_PENDING, &h->flags))
+	if (test_bit(CACHE_PENDING, &h->flags)) {
+		crq->item = cache_get(h);
 		list_add_tail(&crq->q.list, &detail->queue);
-	else
+	} else
 		/* Lost a race, no longer PENDING, so don't enqueue */
 		ret = -EAGAIN;
 	spin_unlock(&queue_lock);

net/sunrpc/xprtrdma/svc_rdma_backchannel.c

@@ -107,26 +107,18 @@ static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma,
 	int ret;
 
 	vec = svc_rdma_get_req_map(rdma);
-	ret = svc_rdma_map_xdr(rdma, sndbuf, vec);
+	ret = svc_rdma_map_xdr(rdma, sndbuf, vec, false);
 	if (ret)
 		goto out_err;
 
-	/* Post a recv buffer to handle the reply for this request. */
-	ret = svc_rdma_post_recv(rdma, GFP_NOIO);
-	if (ret) {
-		pr_err("svcrdma: Failed to post bc receive buffer, err=%d.\n",
-		       ret);
-		pr_err("svcrdma: closing transport %p.\n", rdma);
-		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
-		ret = -ENOTCONN;
+	ret = svc_rdma_repost_recv(rdma, GFP_NOIO);
+	if (ret)
 		goto out_err;
-	}
 
 	ctxt = svc_rdma_get_context(rdma);
 	ctxt->pages[0] = virt_to_page(rqst->rq_buffer);
 	ctxt->count = 1;
 
-	ctxt->wr_op = IB_WR_SEND;
 	ctxt->direction = DMA_TO_DEVICE;
 	ctxt->sge[0].lkey = rdma->sc_pd->local_dma_lkey;
 	ctxt->sge[0].length = sndbuf->len;
@@ -140,7 +132,8 @@ static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma,
 	atomic_inc(&rdma->sc_dma_used);
 
 	memset(&send_wr, 0, sizeof(send_wr));
-	send_wr.wr_id = (unsigned long)ctxt;
+	ctxt->cqe.done = svc_rdma_wc_send;
+	send_wr.wr_cqe = &ctxt->cqe;
 	send_wr.sg_list = ctxt->sge;
 	send_wr.num_sge = 1;
 	send_wr.opcode = IB_WR_SEND;

net/sunrpc/xprtrdma/svc_rdma_marshal.c

@@ -145,29 +145,44 @@ static __be32 *decode_reply_array(__be32 *va, __be32 *vaend)
 	return (__be32 *)&ary->wc_array[nchunks];
 }
 
-int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req,
-			    struct svc_rqst *rqstp)
+int svc_rdma_xdr_decode_req(struct rpcrdma_msg *rmsgp, struct svc_rqst *rqstp)
 {
-	struct rpcrdma_msg *rmsgp = NULL;
 	__be32 *va, *vaend;
+	unsigned int len;
 	u32 hdr_len;
 
-	rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
-
 	/* Verify that there's enough bytes for header + something */
-	if (rqstp->rq_arg.len <= RPCRDMA_HDRLEN_MIN) {
+	if (rqstp->rq_arg.len <= RPCRDMA_HDRLEN_ERR) {
 		dprintk("svcrdma: header too short = %d\n",
 			rqstp->rq_arg.len);
 		return -EINVAL;
 	}
 
-	if (rmsgp->rm_vers != rpcrdma_version)
-		return -ENOSYS;
-
-	/* Pull in the extra for the padded case and bump our pointer */
-	if (rmsgp->rm_type == rdma_msgp) {
-		int hdrlen;
+	if (rmsgp->rm_vers != rpcrdma_version) {
+		dprintk("%s: bad version %u\n", __func__,
+			be32_to_cpu(rmsgp->rm_vers));
+		return -EPROTONOSUPPORT;
+	}
 
+	switch (be32_to_cpu(rmsgp->rm_type)) {
+	case RDMA_MSG:
+	case RDMA_NOMSG:
+		break;
+
+	case RDMA_DONE:
+		/* Just drop it */
+		dprintk("svcrdma: dropping RDMA_DONE message\n");
+		return 0;
+
+	case RDMA_ERROR:
+		/* Possible if this is a backchannel reply.
+		 * XXX: We should cancel this XID, though.
+		 */
+		dprintk("svcrdma: dropping RDMA_ERROR message\n");
+		return 0;
+
+	case RDMA_MSGP:
+		/* Pull in the extra for the padded case, bump our pointer */
 		rmsgp->rm_body.rm_padded.rm_align =
 			be32_to_cpu(rmsgp->rm_body.rm_padded.rm_align);
 		rmsgp->rm_body.rm_padded.rm_thresh =
@@ -175,11 +190,15 @@ int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req,
 
 		va = &rmsgp->rm_body.rm_padded.rm_pempty[4];
 		rqstp->rq_arg.head[0].iov_base = va;
-		hdrlen = (u32)((unsigned long)va - (unsigned long)rmsgp);
-		rqstp->rq_arg.head[0].iov_len -= hdrlen;
-		if (hdrlen > rqstp->rq_arg.len)
+		len = (u32)((unsigned long)va - (unsigned long)rmsgp);
+		rqstp->rq_arg.head[0].iov_len -= len;
+		if (len > rqstp->rq_arg.len)
 			return -EINVAL;
-		return hdrlen;
+		return len;
+	default:
+		dprintk("svcrdma: bad rdma procedure (%u)\n",
+			be32_to_cpu(rmsgp->rm_type));
+		return -EINVAL;
 	}
 
 	/* The chunk list may contain either a read chunk list or a write
@@ -188,20 +207,25 @@ int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req,
 	va = &rmsgp->rm_body.rm_chunks[0];
 	vaend = (__be32 *)((unsigned long)rmsgp + rqstp->rq_arg.len);
 	va = decode_read_list(va, vaend);
-	if (!va)
+	if (!va) {
+		dprintk("svcrdma: failed to decode read list\n");
 		return -EINVAL;
+	}
 	va = decode_write_list(va, vaend);
-	if (!va)
+	if (!va) {
+		dprintk("svcrdma: failed to decode write list\n");
 		return -EINVAL;
+	}
 	va = decode_reply_array(va, vaend);
-	if (!va)
+	if (!va) {
+		dprintk("svcrdma: failed to decode reply chunk\n");
 		return -EINVAL;
+	}
 
 	rqstp->rq_arg.head[0].iov_base = va;
 	hdr_len = (unsigned long)va - (unsigned long)rmsgp;
 	rqstp->rq_arg.head[0].iov_len -= hdr_len;
 
-	*rdma_req = rmsgp;
 	return hdr_len;
 }
 

net/sunrpc/xprtrdma/svc_rdma_recvfrom.c

@@ -180,9 +180,9 @@ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt,
 		clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags);
 
 	memset(&read_wr, 0, sizeof(read_wr));
-	read_wr.wr.wr_id = (unsigned long)ctxt;
+	ctxt->cqe.done = svc_rdma_wc_read;
+	read_wr.wr.wr_cqe = &ctxt->cqe;
 	read_wr.wr.opcode = IB_WR_RDMA_READ;
-	ctxt->wr_op = read_wr.wr.opcode;
 	read_wr.wr.send_flags = IB_SEND_SIGNALED;
 	read_wr.rkey = rs_handle;
 	read_wr.remote_addr = rs_offset;
@@ -299,8 +299,9 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
 	ctxt->read_hdr = head;
 
 	/* Prepare REG WR */
+	ctxt->reg_cqe.done = svc_rdma_wc_reg;
+	reg_wr.wr.wr_cqe = &ctxt->reg_cqe;
 	reg_wr.wr.opcode = IB_WR_REG_MR;
-	reg_wr.wr.wr_id = 0;
 	reg_wr.wr.send_flags = IB_SEND_SIGNALED;
 	reg_wr.wr.num_sge = 0;
 	reg_wr.mr = frmr->mr;
@@ -310,6 +311,8 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
 
 	/* Prepare RDMA_READ */
 	memset(&read_wr, 0, sizeof(read_wr));
+	ctxt->cqe.done = svc_rdma_wc_read;
+	read_wr.wr.wr_cqe = &ctxt->cqe;
 	read_wr.wr.send_flags = IB_SEND_SIGNALED;
 	read_wr.rkey = rs_handle;
 	read_wr.remote_addr = rs_offset;
@@ -317,19 +320,18 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
 	read_wr.wr.num_sge = 1;
 	if (xprt->sc_dev_caps & SVCRDMA_DEVCAP_READ_W_INV) {
 		read_wr.wr.opcode = IB_WR_RDMA_READ_WITH_INV;
-		read_wr.wr.wr_id = (unsigned long)ctxt;
 		read_wr.wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey;
 	} else {
 		read_wr.wr.opcode = IB_WR_RDMA_READ;
 		read_wr.wr.next = &inv_wr;
 		/* Prepare invalidate */
 		memset(&inv_wr, 0, sizeof(inv_wr));
-		inv_wr.wr_id = (unsigned long)ctxt;
+		ctxt->inv_cqe.done = svc_rdma_wc_inv;
+		inv_wr.wr_cqe = &ctxt->inv_cqe;
 		inv_wr.opcode = IB_WR_LOCAL_INV;
 		inv_wr.send_flags = IB_SEND_SIGNALED | IB_SEND_FENCE;
 		inv_wr.ex.invalidate_rkey = frmr->mr->lkey;
 	}
-	ctxt->wr_op = read_wr.wr.opcode;
 
 	/* Post the chain */
 	ret = svc_rdma_send(xprt, &reg_wr.wr);
@@ -612,7 +614,6 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 	struct svc_rdma_op_ctxt *ctxt = NULL;
 	struct rpcrdma_msg *rmsgp;
 	int ret = 0;
-	int len;
 
 	dprintk("svcrdma: rqstp=%p\n", rqstp);
 
@@ -642,8 +643,7 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 		 * transport list
 		 */
 		if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
-			goto close_out;
-
+			goto defer;
 		goto out;
 	}
 	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p, status=%d\n",
@@ -654,15 +654,13 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 	rdma_build_arg_xdr(rqstp, ctxt, ctxt->byte_len);
 
 	/* Decode the RDMA header. */
-	len = svc_rdma_xdr_decode_req(&rmsgp, rqstp);
-	rqstp->rq_xprt_hlen = len;
-
-	/* If the request is invalid, reply with an error */
-	if (len < 0) {
-		if (len == -ENOSYS)
-			svc_rdma_send_error(rdma_xprt, rmsgp, ERR_VERS);
-		goto close_out;
-	}
+	rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base;
+	ret = svc_rdma_xdr_decode_req(rmsgp, rqstp);
+	if (ret < 0)
+		goto out_err;
+	if (ret == 0)
+		goto out_drop;
+	rqstp->rq_xprt_hlen = ret;
 
 	if (svc_rdma_is_backchannel_reply(xprt, rmsgp)) {
 		ret = svc_rdma_handle_bc_reply(xprt->xpt_bc_xprt, rmsgp,
@@ -698,26 +696,16 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 	svc_xprt_copy_addrs(rqstp, xprt);
 	return ret;
 
- close_out:
-	if (ctxt)
-		svc_rdma_put_context(ctxt, 1);
-	dprintk("svcrdma: transport %p is closing\n", xprt);
-	/*
-	 * Set the close bit and enqueue it. svc_recv will see the
-	 * close bit and call svc_xprt_delete
-	 */
-	set_bit(XPT_CLOSE, &xprt->xpt_flags);
+out_err:
+	svc_rdma_send_error(rdma_xprt, rmsgp, ret);
+	svc_rdma_put_context(ctxt, 0);
+	return 0;
+
 defer:
 	return 0;
 
+out_drop:
+	svc_rdma_put_context(ctxt, 1);
 repost:
-	ret = svc_rdma_post_recv(rdma_xprt, GFP_KERNEL);
-	if (ret) {
-		pr_err("svcrdma: could not post a receive buffer, err=%d.\n",
-		       ret);
-		pr_err("svcrdma: closing transport %p.\n", rdma_xprt);
-		set_bit(XPT_CLOSE, &rdma_xprt->sc_xprt.xpt_flags);
-		ret = -ENOTCONN;
-	}
-	return ret;
+	return svc_rdma_repost_recv(rdma_xprt, GFP_KERNEL);
 }