The one new feature is support for a new NFSv4.2 mode_umask attribute

that makes ACL inheritance a little more useful in environments that
 default to restrictive umasks.  Requires client-side support, also on
 its way for 4.10.
 
 Other than that, miscellaneous smaller fixes and cleanup, especially to
 the server rdma code.
 -----BEGIN PGP SIGNATURE-----
 
 iQIcBAABAgAGBQJYVAEqAAoJECebzXlCjuG+VM0QAKaR+ibSM31Ahpnrgit5/wrb
 n630KDFztO7iqEeuHfPQ4/n05T2QR0JWsLpjLMFvx88Gy4gyXYk9cuDPIrNKX1IS
 3/nnhBo0+EVnjODjufommCrtbPZlqOSsS3N03vWkB7rTi8QYsWBOThh+XLRJYOXo
 LZzJE1WmXNeCXV1kXPBsauryywql1fmwTXBzmIf1HbzoGAVROMEA2qqh4Z3nb7BP
 sJuGchWx0STBOuAa278ighXQPUW2lUft9uzw2bssOtMwfNyOs/Pd6nx4F1Lg6WwD
 1UQXoiR8K3PqelZfoeFJ05v0css/sbNKep+huWRdOXZj3Kjpa20lKBX8xHfat7sN
 1OQ4FHx8ToigX3c+wwtlCqRMCcIxqUYkRjqzPHyeBiSSSp0rLrId44rI5x/K0yay
 3bkGw7hFDSzc0Nq2uZgmtlbyTC71hLNhkWe7ThofcVG/pS0JtAqBiKIVwXJPh/e0
 PLmVHYGU6Xowjag5edJlXY1tlIlxtWfqsWUarCXS5bfKUa3UjMVSjyuljsDqqJsn
 96fEWu7DiUo4HeGYmf8MJoeZYV2y0DKSQGeguVkUKWp2DoTzinQHTfdKvrZVwNuu
 hVE9/QeWzUvPY13HOUaKD2skozhbUChqv0NHESKUv8gxE3svTEpYZkXrE74WNqMk
 l/WXAhw+RdKZof4+qdjU
 =JANY
 -----END PGP SIGNATURE-----

Merge tag 'nfsd-4.10' of git://linux-nfs.org/~bfields/linux

Pull nfsd updates from Bruce Fields:
 "The one new feature is support for a new NFSv4.2 mode_umask attribute
  that makes ACL inheritance a little more useful in environments that
  default to restrictive umasks. Requires client-side support, also on
  its way for 4.10.

  Other than that, miscellaneous smaller fixes and cleanup, especially
  to the server rdma code"

[ The client side of the umask attribute was merged yesterday ]

* tag 'nfsd-4.10' of git://linux-nfs.org/~bfields/linux:
  nfsd: add support for the umask attribute
  sunrpc: use DEFINE_SPINLOCK()
  svcrdma: Further clean-up of svc_rdma_get_inv_rkey()
  svcrdma: Break up dprintk format in svc_rdma_accept()
  svcrdma: Remove unused variable in rdma_copy_tail()
  svcrdma: Remove unused variables in xprt_rdma_bc_allocate()
  svcrdma: Remove svc_rdma_op_ctxt::wc_status
  svcrdma: Remove DMA map accounting
  svcrdma: Remove BH-disabled spin locking in svc_rdma_send()
  svcrdma: Renovate sendto chunk list parsing
  svcauth_gss: Close connection when dropping an incoming message
  svcrdma: Clear xpt_bc_xps in xprt_setup_rdma_bc() error exit arm
  nfsd: constify reply_cache_stats_operations structure
  nfsd: update workqueue creation
  sunrpc: GFP_KERNEL should be GFP_NOFS in crypto code
  nfsd: catch errors in decode_fattr earlier
  nfsd: clean up supported attribute handling
  nfsd: fix error handling for clients that fail to return the layout
  nfsd: more robust allocation failure handling in nfsd_reply_cache_init

fs/nfsd/nfs4callback.c

@@ -1061,7 +1061,7 @@ static const struct rpc_call_ops nfsd4_cb_ops = {
 
 int nfsd4_create_callback_queue(void)
 {
-	callback_wq = create_singlethread_workqueue("nfsd4_callbacks");
+	callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
 	if (!callback_wq)
 		return -ENOMEM;
 	return 0;

fs/nfsd/nfs4layouts.c

@@ -686,10 +686,6 @@ nfsd4_cb_layout_done(struct nfsd4_callback *cb, struct rpc_task *task)
 			return 0;
 		}
 		/* Fallthrough */
-	case -NFS4ERR_NOMATCHING_LAYOUT:
-		trace_layout_recall_done(&ls->ls_stid.sc_stateid);
-		task->tk_status = 0;
-		return 1;
 	default:
 		/*
 		 * Unknown error or non-responding client, we'll need to fence.
@@ -702,6 +698,10 @@ nfsd4_cb_layout_done(struct nfsd4_callback *cb, struct rpc_task *task)
 		else
 			nfsd4_cb_layout_fail(ls);
 		return -1;
+	case -NFS4ERR_NOMATCHING_LAYOUT:
+		trace_layout_recall_done(&ls->ls_stid.sc_stateid);
+		task->tk_status = 0;
+		return 1;
 	}
 }
 

fs/nfsd/nfs4proc.c

@@ -96,33 +96,15 @@ check_attr_support(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 {
 	struct dentry *dentry = cstate->current_fh.fh_dentry;
 
-	/*
+	if (!nfsd_attrs_supported(cstate->minorversion, bmval))
-	 * Check about attributes are supported by the NFSv4 server or not.
-	 * According to spec, unsupported attributes return ERR_ATTRNOTSUPP.
-	 */
-	if ((bmval[0] & ~nfsd_suppattrs0(cstate->minorversion)) ||
-	    (bmval[1] & ~nfsd_suppattrs1(cstate->minorversion)) ||
-	    (bmval[2] & ~nfsd_suppattrs2(cstate->minorversion)))
 		return nfserr_attrnotsupp;
-
+	if ((bmval[0] & FATTR4_WORD0_ACL) && !IS_POSIXACL(d_inode(dentry)))
-	/*
+		return nfserr_attrnotsupp;
-	 * Check FATTR4_WORD0_ACL can be supported
+	if (writable && !bmval_is_subset(bmval, writable))
-	 * in current environment or not.
+		return nfserr_inval;
-	 */
+	if (writable && (bmval[2] & FATTR4_WORD2_MODE_UMASK) &&
-	if (bmval[0] & FATTR4_WORD0_ACL) {
+			(bmval[1] & FATTR4_WORD1_MODE))
-		if (!IS_POSIXACL(d_inode(dentry)))
+		return nfserr_inval;
-			return nfserr_attrnotsupp;
-	}
-
-	/*
-	 * According to spec, read-only attributes return ERR_INVAL.
-	 */
-	if (writable) {
-		if ((bmval[0] & ~writable[0]) || (bmval[1] & ~writable[1]) ||
-		    (bmval[2] & ~writable[2]))
-			return nfserr_inval;
-	}
-
 	return nfs_ok;
 }
 
@@ -695,9 +677,9 @@ nfsd4_getattr(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	if (getattr->ga_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 		return nfserr_inval;
 
-	getattr->ga_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
+	getattr->ga_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
-	getattr->ga_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
+	getattr->ga_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
-	getattr->ga_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
+	getattr->ga_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 
 	getattr->ga_fhp = &cstate->current_fh;
 	return nfs_ok;
@@ -799,9 +781,9 @@ nfsd4_readdir(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	if (readdir->rd_bmval[1] & NFSD_WRITEONLY_ATTRS_WORD1)
 		return nfserr_inval;
 
-	readdir->rd_bmval[0] &= nfsd_suppattrs0(cstate->minorversion);
+	readdir->rd_bmval[0] &= nfsd_suppattrs[cstate->minorversion][0];
-	readdir->rd_bmval[1] &= nfsd_suppattrs1(cstate->minorversion);
+	readdir->rd_bmval[1] &= nfsd_suppattrs[cstate->minorversion][1];
-	readdir->rd_bmval[2] &= nfsd_suppattrs2(cstate->minorversion);
+	readdir->rd_bmval[2] &= nfsd_suppattrs[cstate->minorversion][2];
 
 	if ((cookie == 1) || (cookie == 2) ||
 	    (cookie == 0 && memcmp(readdir->rd_verf.data, zeroverf.data, NFS4_VERIFIER_SIZE)))

fs/nfsd/nfs4xdr.c

@@ -33,6 +33,7 @@
  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
+#include <linux/fs_struct.h>
 #include <linux/file.h>
 #include <linux/slab.h>
 #include <linux/namei.h>
@@ -57,6 +58,20 @@
 
 #define NFSDDBG_FACILITY		NFSDDBG_XDR
 
+u32 nfsd_suppattrs[3][3] = {
+	{NFSD4_SUPPORTED_ATTRS_WORD0,
+	 NFSD4_SUPPORTED_ATTRS_WORD1,
+	 NFSD4_SUPPORTED_ATTRS_WORD2},
+
+	{NFSD4_1_SUPPORTED_ATTRS_WORD0,
+	 NFSD4_1_SUPPORTED_ATTRS_WORD1,
+	 NFSD4_1_SUPPORTED_ATTRS_WORD2},
+
+	{NFSD4_1_SUPPORTED_ATTRS_WORD0,
+	 NFSD4_1_SUPPORTED_ATTRS_WORD1,
+	 NFSD4_2_SUPPORTED_ATTRS_WORD2},
+};
+
 /*
  * As per referral draft, the fsid for a referral MUST be different from the fsid of the containing
  * directory in order to indicate to the client that a filesystem boundary is present
@@ -285,7 +300,7 @@ nfsd4_decode_bitmap(struct nfsd4_compoundargs *argp, u32 *bmval)
 static __be32
 nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval,
 		   struct iattr *iattr, struct nfs4_acl **acl,
-		   struct xdr_netobj *label)
+		   struct xdr_netobj *label, int *umask)
 {
 	int expected_len, len = 0;
 	u32 dummy32;
@@ -296,6 +311,14 @@ nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval,
 	if ((status = nfsd4_decode_bitmap(argp, bmval)))
 		return status;
 
+	if (bmval[0] & ~NFSD_WRITEABLE_ATTRS_WORD0
+	    || bmval[1] & ~NFSD_WRITEABLE_ATTRS_WORD1
+	    || bmval[2] & ~NFSD_WRITEABLE_ATTRS_WORD2) {
+		if (nfsd_attrs_supported(argp->minorversion, bmval))
+			return nfserr_inval;
+		return nfserr_attrnotsupp;
+	}
+
 	READ_BUF(4);
 	expected_len = be32_to_cpup(p++);
 
@@ -435,12 +458,18 @@ nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval,
 			return nfserr_jukebox;
 	}
 #endif
-
+	if (bmval[2] & FATTR4_WORD2_MODE_UMASK) {
-	if (bmval[0] & ~NFSD_WRITEABLE_ATTRS_WORD0
+		if (!umask)
-	    || bmval[1] & ~NFSD_WRITEABLE_ATTRS_WORD1
+			goto xdr_error;
-	    || bmval[2] & ~NFSD_WRITEABLE_ATTRS_WORD2)
+		READ_BUF(8);
-		READ_BUF(expected_len - len);
+		len += 8;
-	else if (len != expected_len)
+		dummy32 = be32_to_cpup(p++);
+		iattr->ia_mode = dummy32 & (S_IFMT | S_IALLUGO);
+		dummy32 = be32_to_cpup(p++);
+		*umask = dummy32 & S_IRWXUGO;
+		iattr->ia_valid |= ATTR_MODE;
+	}
+	if (len != expected_len)
 		goto xdr_error;
 
 	DECODE_TAIL;
@@ -634,7 +663,8 @@ nfsd4_decode_create(struct nfsd4_compoundargs *argp, struct nfsd4_create *create
 		return status;
 
 	status = nfsd4_decode_fattr(argp, create->cr_bmval, &create->cr_iattr,
-				    &create->cr_acl, &create->cr_label);
+				    &create->cr_acl, &create->cr_label,
+				    &current->fs->umask);
 	if (status)
 		goto out;
 
@@ -879,13 +909,15 @@ nfsd4_decode_open(struct nfsd4_compoundargs *argp, struct nfsd4_open *open)
 	case NFS4_OPEN_NOCREATE:
 		break;
 	case NFS4_OPEN_CREATE:
+		current->fs->umask = 0;
 		READ_BUF(4);
 		open->op_createmode = be32_to_cpup(p++);
 		switch (open->op_createmode) {
 		case NFS4_CREATE_UNCHECKED:
 		case NFS4_CREATE_GUARDED:
 			status = nfsd4_decode_fattr(argp, open->op_bmval,
-				&open->op_iattr, &open->op_acl, &open->op_label);
+				&open->op_iattr, &open->op_acl, &open->op_label,
+				&current->fs->umask);
 			if (status)
 				goto out;
 			break;
@@ -899,7 +931,8 @@ nfsd4_decode_open(struct nfsd4_compoundargs *argp, struct nfsd4_open *open)
 			READ_BUF(NFS4_VERIFIER_SIZE);
 			COPYMEM(open->op_verf.data, NFS4_VERIFIER_SIZE);
 			status = nfsd4_decode_fattr(argp, open->op_bmval,
-				&open->op_iattr, &open->op_acl, &open->op_label);
+				&open->op_iattr, &open->op_acl, &open->op_label,
+				&current->fs->umask);
 			if (status)
 				goto out;
 			break;
@@ -1136,7 +1169,7 @@ nfsd4_decode_setattr(struct nfsd4_compoundargs *argp, struct nfsd4_setattr *seta
 	if (status)
 		return status;
 	return nfsd4_decode_fattr(argp, setattr->sa_bmval, &setattr->sa_iattr,
-				  &setattr->sa_acl, &setattr->sa_label);
+				  &setattr->sa_acl, &setattr->sa_label, NULL);
 }
 
 static __be32
@@ -2340,9 +2373,7 @@ nfsd4_encode_fattr(struct xdr_stream *xdr, struct svc_fh *fhp,
 	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
 
 	BUG_ON(bmval1 & NFSD_WRITEONLY_ATTRS_WORD1);
-	BUG_ON(bmval0 & ~nfsd_suppattrs0(minorversion));
+	BUG_ON(!nfsd_attrs_supported(minorversion, bmval));
-	BUG_ON(bmval1 & ~nfsd_suppattrs1(minorversion));
-	BUG_ON(bmval2 & ~nfsd_suppattrs2(minorversion));
 
 	if (exp->ex_fslocs.migrated) {
 		status = fattr_handle_absent_fs(&bmval0, &bmval1, &bmval2, &rdattr_err);
@@ -2409,29 +2440,27 @@ nfsd4_encode_fattr(struct xdr_stream *xdr, struct svc_fh *fhp,
 	p++;                /* to be backfilled later */
 
 	if (bmval0 & FATTR4_WORD0_SUPPORTED_ATTRS) {
-		u32 word0 = nfsd_suppattrs0(minorversion);
+		u32 *supp = nfsd_suppattrs[minorversion];
-		u32 word1 = nfsd_suppattrs1(minorversion);
-		u32 word2 = nfsd_suppattrs2(minorversion);
 
 		if (!IS_POSIXACL(dentry->d_inode))
-			word0 &= ~FATTR4_WORD0_ACL;
+			supp[0] &= ~FATTR4_WORD0_ACL;
 		if (!contextsupport)
-			word2 &= ~FATTR4_WORD2_SECURITY_LABEL;
+			supp[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
-		if (!word2) {
+		if (!supp[2]) {
 			p = xdr_reserve_space(xdr, 12);
 			if (!p)
 				goto out_resource;
 			*p++ = cpu_to_be32(2);
-			*p++ = cpu_to_be32(word0);
+			*p++ = cpu_to_be32(supp[0]);
-			*p++ = cpu_to_be32(word1);
+			*p++ = cpu_to_be32(supp[1]);
 		} else {
 			p = xdr_reserve_space(xdr, 16);
 			if (!p)
 				goto out_resource;
 			*p++ = cpu_to_be32(3);
-			*p++ = cpu_to_be32(word0);
+			*p++ = cpu_to_be32(supp[0]);
-			*p++ = cpu_to_be32(word1);
+			*p++ = cpu_to_be32(supp[1]);
-			*p++ = cpu_to_be32(word2);
+			*p++ = cpu_to_be32(supp[2]);
 		}
 	}
 	if (bmval0 & FATTR4_WORD0_TYPE) {

fs/nfsd/nfscache.c

@@ -9,6 +9,7 @@
  */
 
 #include <linux/slab.h>
+#include <linux/vmalloc.h>
 #include <linux/sunrpc/addr.h>
 #include <linux/highmem.h>
 #include <linux/log2.h>
@@ -174,8 +175,12 @@ int nfsd_reply_cache_init(void)
 		goto out_nomem;
 
 	drc_hashtbl = kcalloc(hashsize, sizeof(*drc_hashtbl), GFP_KERNEL);
-	if (!drc_hashtbl)
+	if (!drc_hashtbl) {
-		goto out_nomem;
+		drc_hashtbl = vzalloc(hashsize * sizeof(*drc_hashtbl));
+		if (!drc_hashtbl)
+			goto out_nomem;
+	}
+
 	for (i = 0; i < hashsize; i++) {
 		INIT_LIST_HEAD(&drc_hashtbl[i].lru_head);
 		spin_lock_init(&drc_hashtbl[i].cache_lock);
@@ -204,7 +209,7 @@ void nfsd_reply_cache_shutdown(void)
 		}
 	}
 
-	kfree (drc_hashtbl);
+	kvfree(drc_hashtbl);
 	drc_hashtbl = NULL;
 	drc_hashsize = 0;
 

fs/nfsd/nfsctl.c

@@ -217,7 +217,7 @@ static const struct file_operations pool_stats_operations = {
 	.release	= nfsd_pool_stats_release,
 };
 
-static struct file_operations reply_cache_stats_operations = {
+static const struct file_operations reply_cache_stats_operations = {
 	.open		= nfsd_reply_cache_stats_open,
 	.read		= seq_read,
 	.llseek		= seq_lseek,

fs/nfsd/nfsd.h

@@ -359,44 +359,46 @@ void		nfsd_lockd_shutdown(void);
 
 #define NFSD4_2_SUPPORTED_ATTRS_WORD2 \
 	(NFSD4_1_SUPPORTED_ATTRS_WORD2 | \
+	FATTR4_WORD2_MODE_UMASK | \
 	NFSD4_2_SECURITY_ATTRS)
 
-static inline u32 nfsd_suppattrs0(u32 minorversion)
+extern u32 nfsd_suppattrs[3][3];
+
+static inline bool bmval_is_subset(u32 *bm1, u32 *bm2)
 {
-	return minorversion ? NFSD4_1_SUPPORTED_ATTRS_WORD0
+	return !((bm1[0] & ~bm2[0]) ||
-			    : NFSD4_SUPPORTED_ATTRS_WORD0;
+	         (bm1[1] & ~bm2[1]) ||
+		 (bm1[2] & ~bm2[2]));
 }
 
-static inline u32 nfsd_suppattrs1(u32 minorversion)
+static inline bool nfsd_attrs_supported(u32 minorversion, u32 *bmval)
 {
-	return minorversion ? NFSD4_1_SUPPORTED_ATTRS_WORD1
+	return bmval_is_subset(bmval, nfsd_suppattrs[minorversion]);
-			    : NFSD4_SUPPORTED_ATTRS_WORD1;
-}
-
-static inline u32 nfsd_suppattrs2(u32 minorversion)
-{
-	switch (minorversion) {
-	default: return NFSD4_2_SUPPORTED_ATTRS_WORD2;
-	case 1:  return NFSD4_1_SUPPORTED_ATTRS_WORD2;
-	case 0:  return NFSD4_SUPPORTED_ATTRS_WORD2;
-	}
 }
 
 /* These will return ERR_INVAL if specified in GETATTR or READDIR. */
 #define NFSD_WRITEONLY_ATTRS_WORD1 \
 	(FATTR4_WORD1_TIME_ACCESS_SET   | FATTR4_WORD1_TIME_MODIFY_SET)
 
-/* These are the only attrs allowed in CREATE/OPEN/SETATTR. */
+/*
+ * These are the only attrs allowed in CREATE/OPEN/SETATTR. Don't add
+ * a writeable attribute here without also adding code to parse it to
+ * nfsd4_decode_fattr().
+ */
 #define NFSD_WRITEABLE_ATTRS_WORD0 \
 	(FATTR4_WORD0_SIZE | FATTR4_WORD0_ACL)
 #define NFSD_WRITEABLE_ATTRS_WORD1 \
 	(FATTR4_WORD1_MODE | FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP \
 	| FATTR4_WORD1_TIME_ACCESS_SET | FATTR4_WORD1_TIME_MODIFY_SET)
 #ifdef CONFIG_NFSD_V4_SECURITY_LABEL
-#define NFSD_WRITEABLE_ATTRS_WORD2 FATTR4_WORD2_SECURITY_LABEL
+#define MAYBE_FATTR4_WORD2_SECURITY_LABEL \
+	FATTR4_WORD2_SECURITY_LABEL
 #else
-#define NFSD_WRITEABLE_ATTRS_WORD2 0
+#define MAYBE_FATTR4_WORD2_SECURITY_LABEL 0
 #endif
+#define NFSD_WRITEABLE_ATTRS_WORD2 \
+	(FATTR4_WORD2_MODE_UMASK \
+	| MAYBE_FATTR4_WORD2_SECURITY_LABEL)
 
 #define NFSD_SUPPATTR_EXCLCREAT_WORD0 \
 	NFSD_WRITEABLE_ATTRS_WORD0

fs/nfsd/nfssvc.c

@@ -661,8 +661,8 @@ nfsd(void *vrqstp)
 	mutex_lock(&nfsd_mutex);
 
 	/* At this point, the thread shares current->fs
-	 * with the init process. We need to create files with a
+	 * with the init process. We need to create files with the
-	 * umask of 0 instead of init's umask. */
+	 * umask as defined by the client instead of init's umask. */
 	if (unshare_fs_struct() < 0) {
 		printk("Unable to start nfsd thread: out of memory\n");
 		goto out;

include/linux/sunrpc/svc_rdma.h

@@ -79,7 +79,6 @@ struct svc_rdma_op_ctxt {
 	struct ib_cqe reg_cqe;
 	struct ib_cqe inv_cqe;
 	struct list_head dto_q;
-	enum ib_wc_status wc_status;
 	u32 byte_len;
 	u32 position;
 	struct svcxprt_rdma *xprt;
@@ -139,7 +138,7 @@ struct svcxprt_rdma {
 	int                  sc_max_sge_rd;	/* max sge for read target */
 	bool		     sc_snd_w_inv;	/* OK to use Send With Invalidate */
 
-	atomic_t             sc_sq_count;	/* Number of SQ WR on queue */
+	atomic_t             sc_sq_avail;	/* SQEs ready to be consumed */
 	unsigned int	     sc_sq_depth;	/* Depth of SQ */
 	unsigned int	     sc_rq_depth;	/* Depth of RQ */
 	u32		     sc_max_requests;	/* Forward credits */
@@ -148,7 +147,6 @@ struct svcxprt_rdma {
 
 	struct ib_pd         *sc_pd;
 
-	atomic_t	     sc_dma_used;
 	spinlock_t	     sc_ctxt_lock;
 	struct list_head     sc_ctxts;
 	int		     sc_ctxt_used;
@@ -200,7 +198,6 @@ static inline void svc_rdma_count_mappings(struct svcxprt_rdma *rdma,
 					   struct svc_rdma_op_ctxt *ctxt)
 {
 	ctxt->mapped_sges++;
-	atomic_inc(&rdma->sc_dma_used);
 }
 
 /* svc_rdma_backchannel.c */
@@ -236,8 +233,6 @@ extern int rdma_read_chunk_frmr(struct svcxprt_rdma *, struct svc_rqst *,
 extern int svc_rdma_map_xdr(struct svcxprt_rdma *, struct xdr_buf *,
 			    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);
 

net/sunrpc/auth_gss/gss_krb5_crypto.c

@@ -200,7 +200,7 @@ make_checksum_hmac_md5(struct krb5_ctx *kctx, char *header, int hdrlen,
 	if (IS_ERR(hmac_md5))
 		goto out_free_md5;
 
-	req = ahash_request_alloc(md5, GFP_KERNEL);
+	req = ahash_request_alloc(md5, GFP_NOFS);
 	if (!req)
 		goto out_free_hmac_md5;
 
@@ -230,7 +230,7 @@ make_checksum_hmac_md5(struct krb5_ctx *kctx, char *header, int hdrlen,
 		goto out;
 
 	ahash_request_free(req);
-	req = ahash_request_alloc(hmac_md5, GFP_KERNEL);
+	req = ahash_request_alloc(hmac_md5, GFP_NOFS);
 	if (!req)
 		goto out_free_hmac_md5;
 
@@ -299,7 +299,7 @@ make_checksum(struct krb5_ctx *kctx, char *header, int hdrlen,
 	if (IS_ERR(tfm))
 		goto out_free_cksum;
 
-	req = ahash_request_alloc(tfm, GFP_KERNEL);
+	req = ahash_request_alloc(tfm, GFP_NOFS);
 	if (!req)
 		goto out_free_ahash;
 
@@ -397,7 +397,7 @@ make_checksum_v2(struct krb5_ctx *kctx, char *header, int hdrlen,
 		goto out_free_cksum;
 	checksumlen = crypto_ahash_digestsize(tfm);
 
-	req = ahash_request_alloc(tfm, GFP_KERNEL);
+	req = ahash_request_alloc(tfm, GFP_NOFS);
 	if (!req)
 		goto out_free_ahash;
 
@@ -963,7 +963,7 @@ krb5_rc4_setup_seq_key(struct krb5_ctx *kctx, struct crypto_skcipher *cipher,
 	}
 
 	desc = kmalloc(sizeof(*desc) + crypto_shash_descsize(hmac),
-		       GFP_KERNEL);
+		       GFP_NOFS);
 	if (!desc) {
 		dprintk("%s: failed to allocate shash descriptor for '%s'\n",
 			__func__, kctx->gk5e->cksum_name);
@@ -1030,7 +1030,7 @@ krb5_rc4_setup_enc_key(struct krb5_ctx *kctx, struct crypto_skcipher *cipher,
 	}
 
 	desc = kmalloc(sizeof(*desc) + crypto_shash_descsize(hmac),
-		       GFP_KERNEL);
+		       GFP_NOFS);
 	if (!desc) {
 		dprintk("%s: failed to allocate shash descriptor for '%s'\n",
 			__func__, kctx->gk5e->cksum_name);

net/sunrpc/auth_gss/gss_krb5_mech.c

@@ -451,8 +451,7 @@ context_derive_keys_rc4(struct krb5_ctx *ctx)
 		goto out_err_free_hmac;
 
 
-	desc = kmalloc(sizeof(*desc) + crypto_shash_descsize(hmac),
+	desc = kmalloc(sizeof(*desc) + crypto_shash_descsize(hmac), GFP_NOFS);
-		       GFP_KERNEL);
 	if (!desc) {
 		dprintk("%s: failed to allocate hash descriptor for '%s'\n",
 			__func__, ctx->gk5e->cksum_name);

net/sunrpc/auth_gss/svcauth_gss.c

@@ -1548,7 +1548,7 @@ complete:
 	ret = SVC_COMPLETE;
 	goto out;
 drop:
-	ret = SVC_DROP;
+	ret = SVC_CLOSE;
 out:
 	if (rsci)
 		cache_put(&rsci->h, sn->rsc_cache);

net/sunrpc/svc.c

@@ -1155,8 +1155,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 	case SVC_DENIED:
 		goto err_bad_auth;
 	case SVC_CLOSE:
-		if (test_bit(XPT_TEMP, &rqstp->rq_xprt->xpt_flags))
+		goto close;
-			svc_close_xprt(rqstp->rq_xprt);
 	case SVC_DROP:
 		goto dropit;
 	case SVC_COMPLETE:
@@ -1246,7 +1245,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 
  sendit:
 	if (svc_authorise(rqstp))
-		goto dropit;
+		goto close;
 	return 1;		/* Caller can now send it */
 
  dropit:
@@ -1254,11 +1253,16 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 	dprintk("svc: svc_process dropit\n");
 	return 0;
 
+ close:
+	if (test_bit(XPT_TEMP, &rqstp->rq_xprt->xpt_flags))
+		svc_close_xprt(rqstp->rq_xprt);
+	dprintk("svc: svc_process close\n");
+	return 0;
+
 err_short_len:
 	svc_printk(rqstp, "short len %Zd, dropping request\n",
 			argv->iov_len);
-
+	goto close;
-	goto dropit;			/* drop request */
 
 err_bad_rpc:
 	serv->sv_stats->rpcbadfmt++;

net/sunrpc/svcauth.c

@@ -124,8 +124,7 @@ EXPORT_SYMBOL_GPL(svc_auth_unregister);
 #define	DN_HASHMAX	(1<<DN_HASHBITS)
 
 static struct hlist_head	auth_domain_table[DN_HASHMAX];
-static spinlock_t	auth_domain_lock =
+static DEFINE_SPINLOCK(auth_domain_lock);
-	__SPIN_LOCK_UNLOCKED(auth_domain_lock);
 
 void auth_domain_put(struct auth_domain *dom)
 {

net/sunrpc/xprtrdma/svc_rdma_backchannel.c

@@ -164,13 +164,9 @@ static int
 xprt_rdma_bc_allocate(struct rpc_task *task)
 {
 	struct rpc_rqst *rqst = task->tk_rqstp;
-	struct svc_xprt *sxprt = rqst->rq_xprt->bc_xprt;
 	size_t size = rqst->rq_callsize;
-	struct svcxprt_rdma *rdma;
 	struct page *page;
 
-	rdma = container_of(sxprt, struct svcxprt_rdma, sc_xprt);
-
 	if (size > PAGE_SIZE) {
 		WARN_ONCE(1, "svcrdma: large bc buffer request (size %zu)\n",
 			  size);
@@ -359,6 +355,7 @@ xprt_setup_rdma_bc(struct xprt_create *args)
 out_fail:
 	xprt_rdma_free_addresses(xprt);
 	args->bc_xprt->xpt_bc_xprt = NULL;
+	args->bc_xprt->xpt_bc_xps = NULL;
 	xprt_put(xprt);
 	xprt_free(xprt);
 	return ERR_PTR(-EINVAL);

net/sunrpc/xprtrdma/svc_rdma_recvfrom.c

@@ -279,7 +279,6 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt,
 		       frmr->sg);
 		return -ENOMEM;
 	}
-	atomic_inc(&xprt->sc_dma_used);
 
 	n = ib_map_mr_sg(frmr->mr, frmr->sg, frmr->sg_nents, NULL, PAGE_SIZE);
 	if (unlikely(n != frmr->sg_nents)) {
@@ -374,9 +373,7 @@ rdma_copy_tail(struct svc_rqst *rqstp, struct svc_rdma_op_ctxt *head,
 	       u32 position, u32 byte_count, u32 page_offset, int page_no)
 {
 	char *srcp, *destp;
-	int ret;
 
-	ret = 0;
 	srcp = head->arg.head[0].iov_base + position;
 	byte_count = head->arg.head[0].iov_len - position;
 	if (byte_count > PAGE_SIZE) {
@@ -415,6 +412,20 @@ done:
 	return 1;
 }
 
+/* Returns the address of the first read chunk or <nul> if no read chunk
+ * is present
+ */
+static struct rpcrdma_read_chunk *
+svc_rdma_get_read_chunk(struct rpcrdma_msg *rmsgp)
+{
+	struct rpcrdma_read_chunk *ch =
+		(struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
+
+	if (ch->rc_discrim == xdr_zero)
+		return NULL;
+	return ch;
+}
+
 static int rdma_read_chunks(struct svcxprt_rdma *xprt,
 			    struct rpcrdma_msg *rmsgp,
 			    struct svc_rqst *rqstp,
@@ -627,8 +638,8 @@ int svc_rdma_recvfrom(struct svc_rqst *rqstp)
 			goto defer;
 		goto out;
 	}
-	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p, status=%d\n",
+	dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p\n",
-		ctxt, rdma_xprt, rqstp, ctxt->wc_status);
+		ctxt, rdma_xprt, rqstp);
 	atomic_inc(&rdma_stat_recv);
 
 	/* Build up the XDR from the receive buffers. */

net/sunrpc/xprtrdma/svc_rdma_sendto.c

@@ -153,76 +153,35 @@ static dma_addr_t dma_map_xdr(struct svcxprt_rdma *xprt,
 	return dma_addr;
 }
 
-/* Returns the address of the first read chunk or <nul> if no read chunk
+/* Parse the RPC Call's transport header.
- * is present
  */
-struct rpcrdma_read_chunk *
+static void svc_rdma_get_write_arrays(struct rpcrdma_msg *rmsgp,
-svc_rdma_get_read_chunk(struct rpcrdma_msg *rmsgp)
+				      struct rpcrdma_write_array **write,
+				      struct rpcrdma_write_array **reply)
 {
-	struct rpcrdma_read_chunk *ch =
+	__be32 *p;
-		(struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0];
 
-	if (ch->rc_discrim == xdr_zero)
+	p = (__be32 *)&rmsgp->rm_body.rm_chunks[0];
-		return NULL;
-	return ch;
-}
 
-/* Returns the address of the first read write array element or <nul>
+	/* Read list */
- * if no write array list is present
+	while (*p++ != xdr_zero)
- */
+		p += 5;
-static struct rpcrdma_write_array *
-svc_rdma_get_write_array(struct rpcrdma_msg *rmsgp)
-{
-	if (rmsgp->rm_body.rm_chunks[0] != xdr_zero ||
-	    rmsgp->rm_body.rm_chunks[1] == xdr_zero)
-		return NULL;
-	return (struct rpcrdma_write_array *)&rmsgp->rm_body.rm_chunks[1];
-}
 
-/* Returns the address of the first reply array element or <nul> if no
+	/* Write list */
- * reply array is present
+	if (*p != xdr_zero) {
- */
+		*write = (struct rpcrdma_write_array *)p;
-static struct rpcrdma_write_array *
+		while (*p++ != xdr_zero)
-svc_rdma_get_reply_array(struct rpcrdma_msg *rmsgp,
+			p += 1 + be32_to_cpu(*p) * 4;
-			 struct rpcrdma_write_array *wr_ary)
+	} else {
-{
+		*write = NULL;
-	struct rpcrdma_read_chunk *rch;
+		p++;
-	struct rpcrdma_write_array *rp_ary;
-
-	/* XXX: Need to fix when reply chunk may occur with read list
-	 *	and/or write list.
-	 */
-	if (rmsgp->rm_body.rm_chunks[0] != xdr_zero ||
-	    rmsgp->rm_body.rm_chunks[1] != xdr_zero)
-		return NULL;
-
-	rch = svc_rdma_get_read_chunk(rmsgp);
-	if (rch) {
-		while (rch->rc_discrim != xdr_zero)
-			rch++;
-
-		/* The reply chunk follows an empty write array located
-		 * at 'rc_position' here. The reply array is at rc_target.
-		 */
-		rp_ary = (struct rpcrdma_write_array *)&rch->rc_target;
-		goto found_it;
 	}
 
-	if (wr_ary) {
+	/* Reply chunk */
-		int chunk = be32_to_cpu(wr_ary->wc_nchunks);
+	if (*p != xdr_zero)
-
+		*reply = (struct rpcrdma_write_array *)p;
-		rp_ary = (struct rpcrdma_write_array *)
+	else
-			 &wr_ary->wc_array[chunk].wc_target.rs_length;
+		*reply = NULL;
-		goto found_it;
-	}
-
-	/* No read list, no write list */
-	rp_ary = (struct rpcrdma_write_array *)&rmsgp->rm_body.rm_chunks[2];
-
- found_it:
-	if (rp_ary->wc_discrim == xdr_zero)
-		return NULL;
-	return rp_ary;
 }
 
 /* RPC-over-RDMA Version One private extension: Remote Invalidation.
@@ -240,31 +199,22 @@ static u32 svc_rdma_get_inv_rkey(struct rpcrdma_msg *rdma_argp,
 {
 	struct rpcrdma_read_chunk *rd_ary;
 	struct rpcrdma_segment *arg_ch;
-	u32 inv_rkey;
 
-	inv_rkey = 0;
+	rd_ary = (struct rpcrdma_read_chunk *)&rdma_argp->rm_body.rm_chunks[0];
-
+	if (rd_ary->rc_discrim != xdr_zero)
-	rd_ary = svc_rdma_get_read_chunk(rdma_argp);
+		return be32_to_cpu(rd_ary->rc_target.rs_handle);
-	if (rd_ary) {
-		inv_rkey = be32_to_cpu(rd_ary->rc_target.rs_handle);
-		goto out;
-	}
 
 	if (wr_ary && be32_to_cpu(wr_ary->wc_nchunks)) {
 		arg_ch = &wr_ary->wc_array[0].wc_target;
-		inv_rkey = be32_to_cpu(arg_ch->rs_handle);
+		return be32_to_cpu(arg_ch->rs_handle);
-		goto out;
 	}
 
 	if (rp_ary && be32_to_cpu(rp_ary->wc_nchunks)) {
 		arg_ch = &rp_ary->wc_array[0].wc_target;
-		inv_rkey = be32_to_cpu(arg_ch->rs_handle);
+		return be32_to_cpu(arg_ch->rs_handle);
-		goto out;
 	}
 
-out:
+	return 0;
-	dprintk("svcrdma: Send With Invalidate rkey=%08x\n", inv_rkey);
-	return inv_rkey;
 }
 
 /* Assumptions:
@@ -622,8 +572,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 	 * places this at the start of page 0.
 	 */
 	rdma_argp = page_address(rqstp->rq_pages[0]);
-	wr_ary = svc_rdma_get_write_array(rdma_argp);
+	svc_rdma_get_write_arrays(rdma_argp, &wr_ary, &rp_ary);
-	rp_ary = svc_rdma_get_reply_array(rdma_argp, wr_ary);
 
 	inv_rkey = 0;
 	if (rdma->sc_snd_w_inv)
@@ -636,7 +585,12 @@ int svc_rdma_sendto(struct svc_rqst *rqstp)
 		goto err0;
 	inline_bytes = rqstp->rq_res.len;
 
-	/* Create the RDMA response header */
+	/* Create the RDMA response header. xprt->xpt_mutex,
+	 * acquired in svc_send(), serializes RPC replies. The
+	 * code path below that inserts the credit grant value
+	 * into each transport header runs only inside this
+	 * critical section.
+	 */
 	ret = -ENOMEM;
 	res_page = alloc_page(GFP_KERNEL);
 	if (!res_page)

net/sunrpc/xprtrdma/svc_rdma_transport.c

@@ -41,6 +41,7 @@
  */
 
 #include <linux/sunrpc/svc_xprt.h>
+#include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/debug.h>
 #include <linux/sunrpc/rpc_rdma.h>
 #include <linux/interrupt.h>
@@ -226,25 +227,22 @@ void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt)
 	struct svcxprt_rdma *xprt = ctxt->xprt;
 	struct ib_device *device = xprt->sc_cm_id->device;
 	u32 lkey = xprt->sc_pd->local_dma_lkey;
-	unsigned int i, count;
+	unsigned int i;
 
-	for (count = 0, i = 0; i < ctxt->mapped_sges; i++) {
+	for (i = 0; i < ctxt->mapped_sges; i++) {
 		/*
 		 * Unmap the DMA addr in the SGE if the lkey matches
 		 * the local_dma_lkey, otherwise, ignore it since it is
 		 * an FRMR lkey and will be unmapped later when the
 		 * last WR that uses it completes.
 		 */
-		if (ctxt->sge[i].lkey == lkey) {
+		if (ctxt->sge[i].lkey == lkey)
-			count++;
 			ib_dma_unmap_page(device,
 					    ctxt->sge[i].addr,
 					    ctxt->sge[i].length,
 					    ctxt->direction);
-		}
 	}
 	ctxt->mapped_sges = 0;
-	atomic_sub(count, &xprt->sc_dma_used);
 }
 
 void svc_rdma_put_context(struct svc_rdma_op_ctxt *ctxt, int free_pages)
@@ -398,7 +396,6 @@ static void svc_rdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 
 	/* WARNING: Only wc->wr_cqe and wc->status are reliable */
 	ctxt = container_of(cqe, struct svc_rdma_op_ctxt, cqe);
-	ctxt->wc_status = wc->status;
 	svc_rdma_unmap_dma(ctxt);
 
 	if (wc->status != IB_WC_SUCCESS)
@@ -436,7 +433,7 @@ static void svc_rdma_send_wc_common(struct svcxprt_rdma *xprt,
 		goto err;
 
 out:
-	atomic_dec(&xprt->sc_sq_count);
+	atomic_inc(&xprt->sc_sq_avail);
 	wake_up(&xprt->sc_send_wait);
 	return;
 
@@ -946,7 +943,6 @@ void svc_rdma_put_frmr(struct svcxprt_rdma *rdma,
 	if (frmr) {
 		ib_dma_unmap_sg(rdma->sc_cm_id->device,
 				frmr->sg, frmr->sg_nents, frmr->direction);
-		atomic_dec(&rdma->sc_dma_used);
 		spin_lock_bh(&rdma->sc_frmr_q_lock);
 		WARN_ON_ONCE(!list_empty(&frmr->frmr_list));
 		list_add(&frmr->frmr_list, &rdma->sc_frmr_q);
@@ -973,6 +969,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	struct rpcrdma_connect_private pmsg;
 	struct ib_qp_init_attr qp_attr;
 	struct ib_device *dev;
+	struct sockaddr *sap;
 	unsigned int i;
 	int ret = 0;
 
@@ -1010,6 +1007,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	newxprt->sc_rq_depth = newxprt->sc_max_requests +
 			       newxprt->sc_max_bc_requests;
 	newxprt->sc_sq_depth = RPCRDMA_SQ_DEPTH_MULT * newxprt->sc_rq_depth;
+	atomic_set(&newxprt->sc_sq_avail, newxprt->sc_sq_depth);
 
 	if (!svc_rdma_prealloc_ctxts(newxprt))
 		goto errout;
@@ -1052,18 +1050,12 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	qp_attr.qp_type = IB_QPT_RC;
 	qp_attr.send_cq = newxprt->sc_sq_cq;
 	qp_attr.recv_cq = newxprt->sc_rq_cq;
-	dprintk("svcrdma: newxprt->sc_cm_id=%p, newxprt->sc_pd=%p\n"
+	dprintk("svcrdma: newxprt->sc_cm_id=%p, newxprt->sc_pd=%p\n",
-		"    cm_id->device=%p, sc_pd->device=%p\n"
+		newxprt->sc_cm_id, newxprt->sc_pd);
-		"    cap.max_send_wr = %d\n"
+	dprintk("    cap.max_send_wr = %d, cap.max_recv_wr = %d\n",
-		"    cap.max_recv_wr = %d\n"
+		qp_attr.cap.max_send_wr, qp_attr.cap.max_recv_wr);
-		"    cap.max_send_sge = %d\n"
+	dprintk("    cap.max_send_sge = %d, cap.max_recv_sge = %d\n",
-		"    cap.max_recv_sge = %d\n",
+		qp_attr.cap.max_send_sge, qp_attr.cap.max_recv_sge);
-		newxprt->sc_cm_id, newxprt->sc_pd,
-		dev, newxprt->sc_pd->device,
-		qp_attr.cap.max_send_wr,
-		qp_attr.cap.max_recv_wr,
-		qp_attr.cap.max_send_sge,
-		qp_attr.cap.max_recv_sge);
 
 	ret = rdma_create_qp(newxprt->sc_cm_id, newxprt->sc_pd, &qp_attr);
 	if (ret) {
@@ -1146,31 +1138,16 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 		goto errout;
 	}
 
-	dprintk("svcrdma: new connection %p accepted with the following "
+	dprintk("svcrdma: new connection %p accepted:\n", newxprt);
-		"attributes:\n"
+	sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.src_addr;
-		"    local_ip        : %pI4\n"
+	dprintk("    local address   : %pIS:%u\n", sap, rpc_get_port(sap));
-		"    local_port	     : %d\n"
+	sap = (struct sockaddr *)&newxprt->sc_cm_id->route.addr.dst_addr;
-		"    remote_ip       : %pI4\n"
+	dprintk("    remote address  : %pIS:%u\n", sap, rpc_get_port(sap));
-		"    remote_port     : %d\n"
+	dprintk("    max_sge         : %d\n", newxprt->sc_max_sge);
-		"    max_sge         : %d\n"
+	dprintk("    max_sge_rd      : %d\n", newxprt->sc_max_sge_rd);
-		"    max_sge_rd      : %d\n"
+	dprintk("    sq_depth        : %d\n", newxprt->sc_sq_depth);
-		"    sq_depth        : %d\n"
+	dprintk("    max_requests    : %d\n", newxprt->sc_max_requests);
-		"    max_requests    : %d\n"
+	dprintk("    ord             : %d\n", newxprt->sc_ord);
-		"    ord             : %d\n",
-		newxprt,
-		&((struct sockaddr_in *)&newxprt->sc_cm_id->
-			 route.addr.src_addr)->sin_addr.s_addr,
-		ntohs(((struct sockaddr_in *)&newxprt->sc_cm_id->
-		       route.addr.src_addr)->sin_port),
-		&((struct sockaddr_in *)&newxprt->sc_cm_id->
-			 route.addr.dst_addr)->sin_addr.s_addr,
-		ntohs(((struct sockaddr_in *)&newxprt->sc_cm_id->
-		       route.addr.dst_addr)->sin_port),
-		newxprt->sc_max_sge,
-		newxprt->sc_max_sge_rd,
-		newxprt->sc_sq_depth,
-		newxprt->sc_max_requests,
-		newxprt->sc_ord);
 
 	return &newxprt->sc_xprt;
 
@@ -1257,9 +1234,6 @@ static void __svc_rdma_free(struct work_struct *work)
 	if (rdma->sc_ctxt_used != 0)
 		pr_err("svcrdma: ctxt still in use? (%d)\n",
 		       rdma->sc_ctxt_used);
-	if (atomic_read(&rdma->sc_dma_used) != 0)
-		pr_err("svcrdma: dma still in use? (%d)\n",
-		       atomic_read(&rdma->sc_dma_used));
 
 	/* Final put of backchannel client transport */
 	if (xprt->xpt_bc_xprt) {
@@ -1339,15 +1313,13 @@ int svc_rdma_send(struct svcxprt_rdma *xprt, struct ib_send_wr *wr)
 
 	/* If the SQ is full, wait until an SQ entry is available */
 	while (1) {
-		spin_lock_bh(&xprt->sc_lock);
+		if ((atomic_sub_return(wr_count, &xprt->sc_sq_avail) < 0)) {
-		if (xprt->sc_sq_depth < atomic_read(&xprt->sc_sq_count) + wr_count) {
-			spin_unlock_bh(&xprt->sc_lock);
 			atomic_inc(&rdma_stat_sq_starve);
 
 			/* Wait until SQ WR available if SQ still full */
+			atomic_add(wr_count, &xprt->sc_sq_avail);
 			wait_event(xprt->sc_send_wait,
-				   atomic_read(&xprt->sc_sq_count) <
+				   atomic_read(&xprt->sc_sq_avail) > wr_count);
-				   xprt->sc_sq_depth);
 			if (test_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags))
 				return -ENOTCONN;
 			continue;
@@ -1357,21 +1329,17 @@ int svc_rdma_send(struct svcxprt_rdma *xprt, struct ib_send_wr *wr)
 			svc_xprt_get(&xprt->sc_xprt);
 
 		/* Bump used SQ WR count and post */
-		atomic_add(wr_count, &xprt->sc_sq_count);
 		ret = ib_post_send(xprt->sc_qp, wr, &bad_wr);
 		if (ret) {
 			set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags);
-			atomic_sub(wr_count, &xprt->sc_sq_count);
 			for (i = 0; i < wr_count; i ++)
 				svc_xprt_put(&xprt->sc_xprt);
-			dprintk("svcrdma: failed to post SQ WR rc=%d, "
+			dprintk("svcrdma: failed to post SQ WR rc=%d\n", ret);
-			       "sc_sq_count=%d, sc_sq_depth=%d\n",
+			dprintk("    sc_sq_avail=%d, sc_sq_depth=%d\n",
-			       ret, atomic_read(&xprt->sc_sq_count),
+				atomic_read(&xprt->sc_sq_avail),
-			       xprt->sc_sq_depth);
+				xprt->sc_sq_depth);
-		}
-		spin_unlock_bh(&xprt->sc_lock);
-		if (ret)
 			wake_up(&xprt->sc_send_wait);
+		}
 		break;
 	}
 	return ret;