Merge tag 'nfs-for-4.16-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client updates from Trond Myklebust:
 "Highlights include:

  Stable bugfixes:

   - Fix breakages in the nfsstat utility due to the inclusion of the
     NFSv4 LOOKUPP operation

   - Fix a NULL pointer dereference in nfs_idmap_prepare_pipe_upcall()
     due to nfs_idmap_legacy_upcall() being called without an 'aux'
     parameter

   - Fix a refcount leak in the standard O_DIRECT error path

   - Fix a refcount leak in the pNFS O_DIRECT fallback to MDS path

   - Fix CPU latency issues with nfs_commit_release_pages()

   - Fix the LAYOUTUNAVAILABLE error case in the file layout type

   - NFS: Fix a race between mmap() and O_DIRECT

  Features:

   - Support the statx() mask and query flags to enable optimisations
     when the user is requesting only attributes that are already up to
     date in the inode cache, or is specifying the AT_STATX_DONT_SYNC
     flag

   - Add a module alias for the SCSI pNFS layout type

  Bugfixes:

   - Automounting when resolving a NFSv4 referral should preserve the
     RDMA transport protocol settings

   - Various other RDMA bugfixes from Chuck

   - pNFS block layout fixes

   - Always set NFS_LOCK_LOST when a lock is lost"

* tag 'nfs-for-4.16-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (69 commits)
  NFS: Fix a race between mmap() and O_DIRECT
  NFS: Remove a redundant call to unmap_mapping_range()
  pnfs/blocklayout: Ensure disk address in block device map
  pnfs/blocklayout: pnfs_block_dev_map uses bytes, not sectors
  lockd: Fix server refcounting
  SUNRPC: Fix null rpc_clnt dereference in rpc_task_queued tracepoint
  SUNRPC: Micro-optimize __rpc_execute
  SUNRPC: task_run_action should display tk_callback
  sunrpc: Format RPC events consistently for display
  SUNRPC: Trace xprt_timer events
  xprtrdma: Correct some documenting comments
  xprtrdma: Fix "bytes registered" accounting
  xprtrdma: Instrument allocation/release of rpcrdma_req/rep objects
  xprtrdma: Add trace points to instrument QP and CQ access upcalls
  xprtrdma: Add trace points in the client-side backchannel code paths
  xprtrdma: Add trace points for connect events
  xprtrdma: Add trace points to instrument MR allocation and recovery
  xprtrdma: Add trace points to instrument memory invalidation
  xprtrdma: Add trace points in reply decoder path
  xprtrdma: Add trace points to instrument memory registration
  ..

fs/lockd/clntproc.c

@@ -48,13 +48,13 @@ void nlmclnt_next_cookie(struct nlm_cookie *c)
 
 static struct nlm_lockowner *nlm_get_lockowner(struct nlm_lockowner *lockowner)
 {
-	atomic_inc(&lockowner->count);
+	refcount_inc(&lockowner->count);
 	return lockowner;
 }
 
 static void nlm_put_lockowner(struct nlm_lockowner *lockowner)
 {
-	if (!atomic_dec_and_lock(&lockowner->count, &lockowner->host->h_lock))
+	if (!refcount_dec_and_lock(&lockowner->count, &lockowner->host->h_lock))
 		return;
 	list_del(&lockowner->list);
 	spin_unlock(&lockowner->host->h_lock);
@@ -105,7 +105,7 @@ static struct nlm_lockowner *nlm_find_lockowner(struct nlm_host *host, fl_owner_
 		res = __nlm_find_lockowner(host, owner);
 		if (res == NULL && new != NULL) {
 			res = new;
-			atomic_set(&new->count, 1);
+			refcount_set(&new->count, 1);
 			new->owner = owner;
 			new->pid = __nlm_alloc_pid(host);
 			new->host = nlm_get_host(host);
@@ -204,7 +204,7 @@ struct nlm_rqst *nlm_alloc_call(struct nlm_host *host)
 	for(;;) {
 		call = kzalloc(sizeof(*call), GFP_KERNEL);
 		if (call != NULL) {
-			atomic_set(&call->a_count, 1);
+			refcount_set(&call->a_count, 1);
 			locks_init_lock(&call->a_args.lock.fl);
 			locks_init_lock(&call->a_res.lock.fl);
 			call->a_host = nlm_get_host(host);
@@ -222,7 +222,7 @@ void nlmclnt_release_call(struct nlm_rqst *call)
 {
 	const struct nlmclnt_operations *nlmclnt_ops = call->a_host->h_nlmclnt_ops;
 
-	if (!atomic_dec_and_test(&call->a_count))
+	if (!refcount_dec_and_test(&call->a_count))
 		return;
 	if (nlmclnt_ops && nlmclnt_ops->nlmclnt_release_call)
 		nlmclnt_ops->nlmclnt_release_call(call->a_callback_data);
@@ -678,7 +678,7 @@ nlmclnt_unlock(struct nlm_rqst *req, struct file_lock *fl)
 		goto out;
 	}
 
-	atomic_inc(&req->a_count);
+	refcount_inc(&req->a_count);
 	status = nlmclnt_async_call(nfs_file_cred(fl->fl_file), req,
 			NLMPROC_UNLOCK, &nlmclnt_unlock_ops);
 	if (status < 0)
@@ -769,7 +769,7 @@ static int nlmclnt_cancel(struct nlm_host *host, int block, struct file_lock *fl
 	nlmclnt_setlockargs(req, fl);
 	req->a_args.block = block;
 
-	atomic_inc(&req->a_count);
+	refcount_inc(&req->a_count);
 	status = nlmclnt_async_call(nfs_file_cred(fl->fl_file), req,
 			NLMPROC_CANCEL, &nlmclnt_cancel_ops);
 	if (status == 0 && req->a_res.status == nlm_lck_denied)

fs/lockd/host.c

@@ -114,7 +114,7 @@ static struct nlm_host *nlm_alloc_host(struct nlm_lookup_host_info *ni,
 	unsigned long now = jiffies;
 
 	if (nsm != NULL)
-		atomic_inc(&nsm->sm_count);
+		refcount_inc(&nsm->sm_count);
 	else {
 		host = NULL;
 		nsm = nsm_get_handle(ni->net, ni->sap, ni->salen,
@@ -151,7 +151,7 @@ static struct nlm_host *nlm_alloc_host(struct nlm_lookup_host_info *ni,
 	host->h_state      = 0;
 	host->h_nsmstate   = 0;
 	host->h_pidcount   = 0;
-	atomic_set(&host->h_count, 1);
+	refcount_set(&host->h_count, 1);
 	mutex_init(&host->h_mutex);
 	host->h_nextrebind = now + NLM_HOST_REBIND;
 	host->h_expires    = now + NLM_HOST_EXPIRE;
@@ -290,7 +290,7 @@ void nlmclnt_release_host(struct nlm_host *host)
 
 	WARN_ON_ONCE(host->h_server);
 
-	if (atomic_dec_and_test(&host->h_count)) {
+	if (refcount_dec_and_test(&host->h_count)) {
 		WARN_ON_ONCE(!list_empty(&host->h_lockowners));
 		WARN_ON_ONCE(!list_empty(&host->h_granted));
 		WARN_ON_ONCE(!list_empty(&host->h_reclaim));
@@ -388,6 +388,8 @@ struct nlm_host *nlmsvc_lookup_host(const struct svc_rqst *rqstp,
 	ln->nrhosts++;
 	nrhosts++;
 
+	refcount_inc(&host->h_count);
+
 	dprintk("lockd: %s created host %s (%s)\n",
 		__func__, host->h_name, host->h_addrbuf);
 
@@ -410,7 +412,7 @@ void nlmsvc_release_host(struct nlm_host *host)
 	dprintk("lockd: release server host %s\n", host->h_name);
 
 	WARN_ON_ONCE(!host->h_server);
-	atomic_dec(&host->h_count);
+	refcount_dec(&host->h_count);
 }
 
 /*
@@ -504,7 +506,7 @@ struct nlm_host * nlm_get_host(struct nlm_host *host)
 {
 	if (host) {
 		dprintk("lockd: get host %s\n", host->h_name);
-		atomic_inc(&host->h_count);
+		refcount_inc(&host->h_count);
 		host->h_expires = jiffies + NLM_HOST_EXPIRE;
 	}
 	return host;
@@ -593,7 +595,7 @@ static void nlm_complain_hosts(struct net *net)
 		if (net && host->net != net)
 			continue;
 		dprintk("       %s (cnt %d use %d exp %ld net %x)\n",
-			host->h_name, atomic_read(&host->h_count),
+			host->h_name, refcount_read(&host->h_count),
 			host->h_inuse, host->h_expires, host->net->ns.inum);
 	}
 }
@@ -662,16 +664,16 @@ nlm_gc_hosts(struct net *net)
 	for_each_host_safe(host, next, chain, nlm_server_hosts) {
 		if (net && host->net != net)
 			continue;
-		if (atomic_read(&host->h_count) || host->h_inuse
-		 || time_before(jiffies, host->h_expires)) {
+		if (host->h_inuse || time_before(jiffies, host->h_expires)) {
 			dprintk("nlm_gc_hosts skipping %s "
 				"(cnt %d use %d exp %ld net %x)\n",
-				host->h_name, atomic_read(&host->h_count),
+				host->h_name, refcount_read(&host->h_count),
 				host->h_inuse, host->h_expires,
 				host->net->ns.inum);
 			continue;
 		}
-		nlm_destroy_host_locked(host);
+		if (refcount_dec_if_one(&host->h_count))
+			nlm_destroy_host_locked(host);
 	}
 
 	if (net) {

fs/lockd/mon.c

@@ -191,7 +191,7 @@ void nsm_unmonitor(const struct nlm_host *host)
 	struct nsm_res	res;
 	int status;
 
-	if (atomic_read(&nsm->sm_count) == 1
+	if (refcount_read(&nsm->sm_count) == 1
 	 && nsm->sm_monitored && !nsm->sm_sticky) {
 		dprintk("lockd: nsm_unmonitor(%s)\n", nsm->sm_name);
 
@@ -279,7 +279,7 @@ static struct nsm_handle *nsm_create_handle(const struct sockaddr *sap,
 	if (unlikely(new == NULL))
 		return NULL;
 
-	atomic_set(&new->sm_count, 1);
+	refcount_set(&new->sm_count, 1);
 	new->sm_name = (char *)(new + 1);
 	memcpy(nsm_addr(new), sap, salen);
 	new->sm_addrlen = salen;
@@ -337,13 +337,13 @@ struct nsm_handle *nsm_get_handle(const struct net *net,
 		cached = nsm_lookup_addr(&ln->nsm_handles, sap);
 
 	if (cached != NULL) {
-		atomic_inc(&cached->sm_count);
+		refcount_inc(&cached->sm_count);
 		spin_unlock(&nsm_lock);
 		kfree(new);
 		dprintk("lockd: found nsm_handle for %s (%s), "
 				"cnt %d\n", cached->sm_name,
 				cached->sm_addrbuf,
-				atomic_read(&cached->sm_count));
+				refcount_read(&cached->sm_count));
 		return cached;
 	}
 
@@ -388,12 +388,12 @@ struct nsm_handle *nsm_reboot_lookup(const struct net *net,
 		return cached;
 	}
 
-	atomic_inc(&cached->sm_count);
+	refcount_inc(&cached->sm_count);
 	spin_unlock(&nsm_lock);
 
 	dprintk("lockd: host %s (%s) rebooted, cnt %d\n",
 			cached->sm_name, cached->sm_addrbuf,
-			atomic_read(&cached->sm_count));
+			refcount_read(&cached->sm_count));
 	return cached;
 }
 
@@ -404,7 +404,7 @@ struct nsm_handle *nsm_reboot_lookup(const struct net *net,
  */
 void nsm_release(struct nsm_handle *nsm)
 {
-	if (atomic_dec_and_lock(&nsm->sm_count, &nsm_lock)) {
+	if (refcount_dec_and_lock(&nsm->sm_count, &nsm_lock)) {
 		list_del(&nsm->sm_link);
 		spin_unlock(&nsm_lock);
 		dprintk("lockd: destroyed nsm_handle for %s (%s)\n",

fs/lockd/svcproc.c

@@ -295,7 +295,7 @@ static void nlmsvc_callback_exit(struct rpc_task *task, void *data)
 
 void nlmsvc_release_call(struct nlm_rqst *call)
 {
-	if (!atomic_dec_and_test(&call->a_count))
+	if (!refcount_dec_and_test(&call->a_count))
 		return;
 	nlmsvc_release_host(call->a_host);
 	kfree(call);

fs/nfs/blocklayout/blocklayout.c

@@ -137,6 +137,11 @@ bl_alloc_init_bio(int npg, struct block_device *bdev, sector_t disk_sector,
 	return bio;
 }
 
+static bool offset_in_map(u64 offset, struct pnfs_block_dev_map *map)
+{
+	return offset >= map->start && offset < map->start + map->len;
+}
+
 static struct bio *
 do_add_page_to_bio(struct bio *bio, int npg, int rw, sector_t isect,
 		struct page *page, struct pnfs_block_dev_map *map,
@@ -156,8 +161,8 @@ do_add_page_to_bio(struct bio *bio, int npg, int rw, sector_t isect,
 
 	/* translate to physical disk offset */
 	disk_addr = (u64)isect << SECTOR_SHIFT;
-	if (disk_addr < map->start || disk_addr >= map->start + map->len) {
-		if (!dev->map(dev, disk_addr, map))
+	if (!offset_in_map(disk_addr, map)) {
+		if (!dev->map(dev, disk_addr, map) || !offset_in_map(disk_addr, map))
 			return ERR_PTR(-EIO);
 		bio = bl_submit_bio(bio);
 	}
@@ -184,6 +189,29 @@ do_add_page_to_bio(struct bio *bio, int npg, int rw, sector_t isect,
 	return bio;
 }
 
+static void bl_mark_devices_unavailable(struct nfs_pgio_header *header, bool rw)
+{
+	struct pnfs_block_layout *bl = BLK_LSEG2EXT(header->lseg);
+	size_t bytes_left = header->args.count;
+	sector_t isect, extent_length = 0;
+	struct pnfs_block_extent be;
+
+	isect = header->args.offset >> SECTOR_SHIFT;
+	bytes_left += header->args.offset - (isect << SECTOR_SHIFT);
+
+	while (bytes_left > 0) {
+		if (!ext_tree_lookup(bl, isect, &be, rw))
+				return;
+		extent_length = be.be_length - (isect - be.be_f_offset);
+		nfs4_mark_deviceid_unavailable(be.be_device);
+		isect += extent_length;
+		if (bytes_left > extent_length << SECTOR_SHIFT)
+			bytes_left -= extent_length << SECTOR_SHIFT;
+		else
+			bytes_left = 0;
+	}
+}
+
 static void bl_end_io_read(struct bio *bio)
 {
 	struct parallel_io *par = bio->bi_private;
@@ -194,6 +222,7 @@ static void bl_end_io_read(struct bio *bio)
 		if (!header->pnfs_error)
 			header->pnfs_error = -EIO;
 		pnfs_set_lo_fail(header->lseg);
+		bl_mark_devices_unavailable(header, false);
 	}
 
 	bio_put(bio);
@@ -323,6 +352,7 @@ static void bl_end_io_write(struct bio *bio)
 		if (!header->pnfs_error)
 			header->pnfs_error = -EIO;
 		pnfs_set_lo_fail(header->lseg);
+		bl_mark_devices_unavailable(header, true);
 	}
 	bio_put(bio);
 	put_parallel(par);
@@ -552,6 +582,31 @@ static int decode_sector_number(__be32 **rp, sector_t *sp)
 	return 0;
 }
 
+static struct nfs4_deviceid_node *
+bl_find_get_deviceid(struct nfs_server *server,
+		const struct nfs4_deviceid *id, struct rpc_cred *cred,
+		gfp_t gfp_mask)
+{
+	struct nfs4_deviceid_node *node;
+	unsigned long start, end;
+
+retry:
+	node = nfs4_find_get_deviceid(server, id, cred, gfp_mask);
+	if (!node)
+		return ERR_PTR(-ENODEV);
+
+	if (test_bit(NFS_DEVICEID_UNAVAILABLE, &node->flags) == 0)
+		return node;
+
+	end = jiffies;
+	start = end - PNFS_DEVICE_RETRY_TIMEOUT;
+	if (!time_in_range(node->timestamp_unavailable, start, end)) {
+		nfs4_delete_deviceid(node->ld, node->nfs_client, id);
+		goto retry;
+	}
+	return ERR_PTR(-ENODEV);
+}
+
 static int
 bl_alloc_extent(struct xdr_stream *xdr, struct pnfs_layout_hdr *lo,
 		struct layout_verification *lv, struct list_head *extents,
@@ -573,16 +628,18 @@ bl_alloc_extent(struct xdr_stream *xdr, struct pnfs_layout_hdr *lo,
 	memcpy(&id, p, NFS4_DEVICEID4_SIZE);
 	p += XDR_QUADLEN(NFS4_DEVICEID4_SIZE);
 
-	error = -EIO;
-	be->be_device = nfs4_find_get_deviceid(NFS_SERVER(lo->plh_inode), &id,
+	be->be_device = bl_find_get_deviceid(NFS_SERVER(lo->plh_inode), &id,
 						lo->plh_lc_cred, gfp_mask);
-	if (!be->be_device)
+	if (IS_ERR(be->be_device)) {
+		error = PTR_ERR(be->be_device);
 		goto out_free_be;
+	}
 
 	/*
 	 * The next three values are read in as bytes, but stored in the
 	 * extent structure in 512-byte granularity.
 	 */
+	error = -EIO;
 	if (decode_sector_number(&p, &be->be_f_offset) < 0)
 		goto out_put_deviceid;
 	if (decode_sector_number(&p, &be->be_length) < 0)
@@ -692,11 +749,16 @@ bl_alloc_lseg(struct pnfs_layout_hdr *lo, struct nfs4_layoutget_res *lgr,
 	__free_page(scratch);
 out:
 	dprintk("%s returns %d\n", __func__, status);
-	if (status) {
+	switch (status) {
+	case -ENODEV:
+		/* Our extent block devices are unavailable */
+		set_bit(NFS_LSEG_UNAVAILABLE, &lseg->pls_flags);
+	case 0:
+		return lseg;
+	default:
 		kfree(lseg);
 		return ERR_PTR(status);
 	}
-	return lseg;
 }
 
 static void
@@ -798,6 +860,13 @@ bl_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *req)
 	}
 
 	pnfs_generic_pg_init_read(pgio, req);
+
+	if (pgio->pg_lseg &&
+		test_bit(NFS_LSEG_UNAVAILABLE, &pgio->pg_lseg->pls_flags)) {
+		pnfs_error_mark_layout_for_return(pgio->pg_inode, pgio->pg_lseg);
+		pnfs_set_lo_fail(pgio->pg_lseg);
+		nfs_pageio_reset_read_mds(pgio);
+	}
 }
 
 /*
@@ -853,6 +922,14 @@ bl_pg_init_write(struct nfs_pageio_descriptor *pgio, struct nfs_page *req)
 		wb_size = nfs_dreq_bytes_left(pgio->pg_dreq);
 
 	pnfs_generic_pg_init_write(pgio, req, wb_size);
+
+	if (pgio->pg_lseg &&
+		test_bit(NFS_LSEG_UNAVAILABLE, &pgio->pg_lseg->pls_flags)) {
+
+		pnfs_error_mark_layout_for_return(pgio->pg_inode, pgio->pg_lseg);
+		pnfs_set_lo_fail(pgio->pg_lseg);
+		nfs_pageio_reset_write_mds(pgio);
+	}
 }
 
 /*
@@ -887,6 +964,7 @@ static struct pnfs_layoutdriver_type blocklayout_type = {
 	.name				= "LAYOUT_BLOCK_VOLUME",
 	.owner				= THIS_MODULE,
 	.flags				= PNFS_LAYOUTRET_ON_SETATTR |
+					  PNFS_LAYOUTRET_ON_ERROR |
 					  PNFS_READ_WHOLE_PAGE,
 	.read_pagelist			= bl_read_pagelist,
 	.write_pagelist			= bl_write_pagelist,
@@ -910,6 +988,7 @@ static struct pnfs_layoutdriver_type scsilayout_type = {
 	.name				= "LAYOUT_SCSI",
 	.owner				= THIS_MODULE,
 	.flags				= PNFS_LAYOUTRET_ON_SETATTR |
+					  PNFS_LAYOUTRET_ON_ERROR |
 					  PNFS_READ_WHOLE_PAGE,
 	.read_pagelist			= bl_read_pagelist,
 	.write_pagelist			= bl_write_pagelist,
@@ -967,6 +1046,7 @@ static void __exit nfs4blocklayout_exit(void)
 }
 
 MODULE_ALIAS("nfs-layouttype4-3");
+MODULE_ALIAS("nfs-layouttype4-5");
 
 module_init(nfs4blocklayout_init);
 module_exit(nfs4blocklayout_exit);

fs/nfs/blocklayout/blocklayout.h

@@ -92,10 +92,9 @@ struct pnfs_block_volume {
 };
 
 struct pnfs_block_dev_map {
-	sector_t			start;
-	sector_t			len;
-
-	sector_t			disk_offset;
+	u64			start;
+	u64			len;
+	u64			disk_offset;
 	struct block_device		*bdev;
 };
 

fs/nfs/blocklayout/dev.c

@@ -533,14 +533,11 @@ bl_alloc_deviceid_node(struct nfs_server *server, struct pnfs_device *pdev,
 		goto out_free_volumes;
 
 	ret = bl_parse_deviceid(server, top, volumes, nr_volumes - 1, gfp_mask);
-	if (ret) {
-		bl_free_device(top);
-		kfree(top);
-		goto out_free_volumes;
-	}
 
 	node = &top->node;
 	nfs4_init_deviceid_node(node, server, &pdev->dev_id);
+	if (ret)
+		nfs4_mark_deviceid_unavailable(node);
 
 out_free_volumes:
 	kfree(volumes);

fs/nfs/direct.c

@@ -775,10 +775,8 @@ static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
 
 	spin_lock(&dreq->lock);
 
-	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
-		dreq->flags = 0;
+	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags))
 		dreq->error = hdr->error;
-	}
 	if (dreq->error == 0) {
 		nfs_direct_good_bytes(dreq, hdr);
 		if (nfs_write_need_commit(hdr)) {

fs/nfs/export.c

@@ -48,10 +48,6 @@ nfs_encode_fh(struct inode *inode, __u32 *p, int *max_len, struct inode *parent)
 		*max_len = len;
 		return FILEID_INVALID;
 	}
-	if (IS_AUTOMOUNT(inode)) {
-		*max_len = FILEID_INVALID;
-		goto out;
-	}
 
 	p[FILEID_HIGH_OFF] = NFS_FILEID(inode) >> 32;
 	p[FILEID_LOW_OFF] = NFS_FILEID(inode);
@@ -59,7 +55,6 @@ nfs_encode_fh(struct inode *inode, __u32 *p, int *max_len, struct inode *parent)
 	p[len - 1] = 0; /* Padding */
 	nfs_copy_fh(clnt_fh, server_fh);
 	*max_len = len;
-out:
 	dprintk("%s: result fh fileid %llu mode %u size %d\n",
 		__func__, NFS_FILEID(inode), inode->i_mode, *max_len);
 	return *max_len;

fs/nfs/filelayout/filelayout.c

@@ -895,9 +895,7 @@ fl_pnfs_update_layout(struct inode *ino,
 
 	lseg = pnfs_update_layout(ino, ctx, pos, count, iomode, strict_iomode,
 				  gfp_flags);
-	if (!lseg)
-		lseg = ERR_PTR(-ENOMEM);
-	if (IS_ERR(lseg))
+	if (IS_ERR_OR_NULL(lseg))
 		goto out;
 
 	lo = NFS_I(ino)->layout;

fs/nfs/inode.c

@@ -735,12 +735,20 @@ int nfs_getattr(const struct path *path, struct kstat *stat,
 		u32 request_mask, unsigned int query_flags)
 {
 	struct inode *inode = d_inode(path->dentry);
-	int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
+	struct nfs_server *server = NFS_SERVER(inode);
+	unsigned long cache_validity;
 	int err = 0;
+	bool force_sync = query_flags & AT_STATX_FORCE_SYNC;
+	bool do_update = false;
 
 	trace_nfs_getattr_enter(inode);
+
+	if ((query_flags & AT_STATX_DONT_SYNC) && !force_sync)
+		goto out_no_update;
+
 	/* Flush out writes to the server in order to update c/mtime.  */
-	if (S_ISREG(inode->i_mode)) {
+	if ((request_mask & (STATX_CTIME|STATX_MTIME)) &&
+			S_ISREG(inode->i_mode)) {
 		err = filemap_write_and_wait(inode->i_mapping);
 		if (err)
 			goto out;
@@ -757,24 +765,42 @@ int nfs_getattr(const struct path *path, struct kstat *stat,
 	 */
 	if ((path->mnt->mnt_flags & MNT_NOATIME) ||
 	    ((path->mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
-		need_atime = 0;
-
-	if (need_atime || nfs_need_revalidate_inode(inode)) {
-		struct nfs_server *server = NFS_SERVER(inode);
-
+		request_mask &= ~STATX_ATIME;
+
+	/* Is the user requesting attributes that might need revalidation? */
+	if (!(request_mask & (STATX_MODE|STATX_NLINK|STATX_ATIME|STATX_CTIME|
+					STATX_MTIME|STATX_UID|STATX_GID|
+					STATX_SIZE|STATX_BLOCKS)))
+		goto out_no_revalidate;
+
+	/* Check whether the cached attributes are stale */
+	do_update |= force_sync || nfs_attribute_cache_expired(inode);
+	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
+	do_update |= cache_validity &
+		(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_LABEL);
+	if (request_mask & STATX_ATIME)
+		do_update |= cache_validity & NFS_INO_INVALID_ATIME;
+	if (request_mask & (STATX_CTIME|STATX_MTIME))
+		do_update |= cache_validity & NFS_INO_REVAL_PAGECACHE;
+	if (do_update) {
+		/* Update the attribute cache */
 		if (!(server->flags & NFS_MOUNT_NOAC))
 			nfs_readdirplus_parent_cache_miss(path->dentry);
 		else
 			nfs_readdirplus_parent_cache_hit(path->dentry);
 		err = __nfs_revalidate_inode(server, inode);
+		if (err)
+			goto out;
 	} else
 		nfs_readdirplus_parent_cache_hit(path->dentry);
-	if (!err) {
-		generic_fillattr(inode, stat);
-		stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode));
-		if (S_ISDIR(inode->i_mode))
-			stat->blksize = NFS_SERVER(inode)->dtsize;
-	}
+out_no_revalidate:
+	/* Only return attributes that were revalidated. */
+	stat->result_mask &= request_mask;
+out_no_update:
+	generic_fillattr(inode, stat);
+	stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode));
+	if (S_ISDIR(inode->i_mode))
+		stat->blksize = NFS_SERVER(inode)->dtsize;
 out:
 	trace_nfs_getattr_exit(inode, err);
 	return err;
@@ -1144,7 +1170,6 @@ static int nfs_invalidate_mapping(struct inode *inode, struct address_space *map
 
 	if (mapping->nrpages != 0) {
 		if (S_ISREG(inode->i_mode)) {
-			unmap_mapping_range(mapping, 0, 0, 0);
 			ret = nfs_sync_mapping(mapping);
 			if (ret < 0)
 				return ret;

fs/nfs/io.c

@@ -99,7 +99,7 @@ static void nfs_block_buffered(struct nfs_inode *nfsi, struct inode *inode)
 {
 	if (!test_bit(NFS_INO_ODIRECT, &nfsi->flags)) {
 		set_bit(NFS_INO_ODIRECT, &nfsi->flags);
-		nfs_wb_all(inode);
+		nfs_sync_mapping(inode->i_mapping);
 	}
 }
 

fs/nfs/nfs4client.c

@@ -861,6 +861,7 @@ static int nfs4_set_client(struct nfs_server *server,
 		set_bit(NFS_CS_MIGRATION, &cl_init.init_flags);
 	if (test_bit(NFS_MIG_TSM_POSSIBLE, &server->mig_status))
 		set_bit(NFS_CS_TSM_POSSIBLE, &cl_init.init_flags);
+	server->port = rpc_get_port(addr);
 
 	/* Allocate or find a client reference we can use */
 	clp = nfs_get_client(&cl_init);
@@ -1123,19 +1124,36 @@ struct nfs_server *nfs4_create_referral_server(struct nfs_clone_mount *data,
 	/* Initialise the client representation from the parent server */
 	nfs_server_copy_userdata(server, parent_server);
 
-	/* Get a client representation.
-	 * Note: NFSv4 always uses TCP, */
+	/* Get a client representation */
+#ifdef CONFIG_SUNRPC_XPRT_RDMA
+	rpc_set_port(data->addr, NFS_RDMA_PORT);
 	error = nfs4_set_client(server, data->hostname,
 				data->addr,
 				data->addrlen,
 				parent_client->cl_ipaddr,
-				rpc_protocol(parent_server->client),
+				XPRT_TRANSPORT_RDMA,
+				parent_server->client->cl_timeout,
+				parent_client->cl_mvops->minor_version,
+				parent_client->cl_net);
+	if (!error)
+		goto init_server;
+#endif	/* CONFIG_SUNRPC_XPRT_RDMA */
+
+	rpc_set_port(data->addr, NFS_PORT);
+	error = nfs4_set_client(server, data->hostname,
+				data->addr,
+				data->addrlen,
+				parent_client->cl_ipaddr,
+				XPRT_TRANSPORT_TCP,
 				parent_server->client->cl_timeout,
 				parent_client->cl_mvops->minor_version,
 				parent_client->cl_net);
 	if (error < 0)
 		goto error;
 
+#ifdef CONFIG_SUNRPC_XPRT_RDMA
+init_server:
+#endif
 	error = nfs_init_server_rpcclient(server, parent_server->client->cl_timeout, data->authflavor);
 	if (error < 0)
 		goto error;

fs/nfs/nfs4idmap.c

@@ -568,9 +568,13 @@ static int nfs_idmap_legacy_upcall(struct key_construction *cons,
 	struct idmap_msg *im;
 	struct idmap *idmap = (struct idmap *)aux;
 	struct key *key = cons->key;
-	int ret = -ENOMEM;
+	int ret = -ENOKEY;
+
+	if (!aux)
+		goto out1;
 
 	/* msg and im are freed in idmap_pipe_destroy_msg */
+	ret = -ENOMEM;
 	data = kzalloc(sizeof(*data), GFP_KERNEL);
 	if (!data)
 		goto out1;

fs/nfs/nfs4namespace.c

@@ -270,8 +270,6 @@ static struct vfsmount *try_location(struct nfs_clone_mount *mountdata,
 		if (mountdata->addrlen == 0)
 			continue;
 
-		rpc_set_port(mountdata->addr, NFS_PORT);
-
 		memcpy(page2, buf->data, buf->len);
 		page2[buf->len] = '\0';
 		mountdata->hostname = page2;

fs/nfs/nfs4proc.c

@@ -2020,7 +2020,7 @@ static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *sta
 	return ret;
 }
 
-static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, int err)
+static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
 {
 	switch (err) {
 		default:
@@ -2067,7 +2067,11 @@ static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct
 			return -EAGAIN;
 		case -ENOMEM:
 		case -NFS4ERR_DENIED:
-			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
+			if (fl) {
+				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
+				if (lsp)
+					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
+			}
 			return 0;
 	}
 	return err;
@@ -2103,7 +2107,7 @@ int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
 		err = nfs4_open_recover_helper(opendata, FMODE_READ);
 	}
 	nfs4_opendata_put(opendata);
-	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
+	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
 }
 
 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
@@ -3150,6 +3154,11 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 	struct nfs4_state *state = calldata->state;
 	struct nfs_server *server = NFS_SERVER(calldata->inode);
 	nfs4_stateid *res_stateid = NULL;
+	struct nfs4_exception exception = {
+		.state = state,
+		.inode = calldata->inode,
+		.stateid = &calldata->arg.stateid,
+	};
 
 	dprintk("%s: begin!\n", __func__);
 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
@@ -3215,7 +3224,9 @@ static void nfs4_close_done(struct rpc_task *task, void *data)
 		case -NFS4ERR_BAD_STATEID:
 			break;
 		default:
-			if (nfs4_async_handle_error(task, server, state, NULL) == -EAGAIN)
+			task->tk_status = nfs4_async_handle_exception(task,
+					server, task->tk_status, &exception);
+			if (exception.retry)
 				goto out_restart;
 	}
 	nfs_clear_open_stateid(state, &calldata->arg.stateid,
@@ -5759,6 +5770,10 @@ struct nfs4_delegreturndata {
 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 {
 	struct nfs4_delegreturndata *data = calldata;
+	struct nfs4_exception exception = {
+		.inode = data->inode,
+		.stateid = &data->stateid,
+	};
 
 	if (!nfs4_sequence_done(task, &data->res.seq_res))
 		return;
@@ -5820,10 +5835,11 @@ static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
 		}
 		/* Fallthrough */
 	default:
-		if (nfs4_async_handle_error(task, data->res.server,
-					    NULL, NULL) == -EAGAIN) {
+		task->tk_status = nfs4_async_handle_exception(task,
+				data->res.server, task->tk_status,
+				&exception);
+		if (exception.retry)
 			goto out_restart;
-		}
 	}
 	data->rpc_status = task->tk_status;
 	return;
@@ -6061,6 +6077,10 @@ static void nfs4_locku_release_calldata(void *data)
 static void nfs4_locku_done(struct rpc_task *task, void *data)
 {
 	struct nfs4_unlockdata *calldata = data;
+	struct nfs4_exception exception = {
+		.inode = calldata->lsp->ls_state->inode,
+		.stateid = &calldata->arg.stateid,
+	};
 
 	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
 		return;
@@ -6084,8 +6104,10 @@ static void nfs4_locku_done(struct rpc_task *task, void *data)
 				rpc_restart_call_prepare(task);
 			break;
 		default:
-			if (nfs4_async_handle_error(task, calldata->server,
-						    NULL, NULL) == -EAGAIN)
+			task->tk_status = nfs4_async_handle_exception(task,
+					calldata->server, task->tk_status,
+					&exception);
+			if (exception.retry)
 				rpc_restart_call_prepare(task);
 	}
 	nfs_release_seqid(calldata->arg.seqid);
@@ -6741,7 +6763,7 @@ int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state,
 	if (err != 0)
 		return err;
 	err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
-	return nfs4_handle_delegation_recall_error(server, state, stateid, err);
+	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
 }
 
 struct nfs_release_lockowner_data {

fs/nfs/nfs4state.c

@@ -1482,6 +1482,7 @@ static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_
 	struct inode *inode = state->inode;
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct file_lock *fl;
+	struct nfs4_lock_state *lsp;
 	int status = 0;
 	struct file_lock_context *flctx = inode->i_flctx;
 	struct list_head *list;
@@ -1522,7 +1523,9 @@ static int nfs4_reclaim_locks(struct nfs4_state *state, const struct nfs4_state_
 		case -NFS4ERR_DENIED:
 		case -NFS4ERR_RECLAIM_BAD:
 		case -NFS4ERR_RECLAIM_CONFLICT:
-			/* kill_proc(fl->fl_pid, SIGLOST, 1); */
+			lsp = fl->fl_u.nfs4_fl.owner;
+			if (lsp)
+				set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
 			status = 0;
 		}
 		spin_lock(&flctx->flc_lock);

fs/nfs/nfs4sysctl.c

@@ -32,7 +32,7 @@ static struct ctl_table nfs4_cb_sysctls[] = {
 		.data = &nfs_idmap_cache_timeout,
 		.maxlen = sizeof(int),
 		.mode = 0644,
-		.proc_handler = proc_dointvec_jiffies,
+		.proc_handler = proc_dointvec,
 	},
 	{ }
 };

fs/nfs/nfs4xdr.c

@@ -7678,6 +7678,22 @@ nfs4_stat_to_errno(int stat)
 	.p_name = #proc,	\
 }
 
+#if defined(CONFIG_NFS_V4_1)
+#define PROC41(proc, argtype, restype)				\
+	PROC(proc, argtype, restype)
+#else
+#define PROC41(proc, argtype, restype)				\
+	STUB(proc)
+#endif
+
+#if defined(CONFIG_NFS_V4_2)
+#define PROC42(proc, argtype, restype)				\
+	PROC(proc, argtype, restype)
+#else
+#define PROC42(proc, argtype, restype)				\
+	STUB(proc)
+#endif
+
 const struct rpc_procinfo nfs4_procedures[] = {
 	PROC(READ,		enc_read,		dec_read),
 	PROC(WRITE,		enc_write,		dec_write),
@@ -7698,7 +7714,6 @@ const struct rpc_procinfo nfs4_procedures[] = {
 	PROC(ACCESS,		enc_access,		dec_access),
 	PROC(GETATTR,		enc_getattr,		dec_getattr),
 	PROC(LOOKUP,		enc_lookup,		dec_lookup),
-	PROC(LOOKUPP,		enc_lookupp,		dec_lookupp),
 	PROC(LOOKUP_ROOT,	enc_lookup_root,	dec_lookup_root),
 	PROC(REMOVE,		enc_remove,		dec_remove),
 	PROC(RENAME,		enc_rename,		dec_rename),
@@ -7717,33 +7732,30 @@ const struct rpc_procinfo nfs4_procedures[] = {
 	PROC(RELEASE_LOCKOWNER,	enc_release_lockowner,	dec_release_lockowner),
 	PROC(SECINFO,		enc_secinfo,		dec_secinfo),
 	PROC(FSID_PRESENT,	enc_fsid_present,	dec_fsid_present),
-#if defined(CONFIG_NFS_V4_1)
-	PROC(EXCHANGE_ID,	enc_exchange_id,	dec_exchange_id),
-	PROC(CREATE_SESSION,	enc_create_session,	dec_create_session),
-	PROC(DESTROY_SESSION,	enc_destroy_session,	dec_destroy_session),
-	PROC(SEQUENCE,		enc_sequence,		dec_sequence),
-	PROC(GET_LEASE_TIME,	enc_get_lease_time,	dec_get_lease_time),
-	PROC(RECLAIM_COMPLETE,	enc_reclaim_complete,	dec_reclaim_complete),
-	PROC(GETDEVICEINFO,	enc_getdeviceinfo,	dec_getdeviceinfo),
-	PROC(LAYOUTGET,		enc_layoutget,		dec_layoutget),
-	PROC(LAYOUTCOMMIT,	enc_layoutcommit,	dec_layoutcommit),
-	PROC(LAYOUTRETURN,	enc_layoutreturn,	dec_layoutreturn),
-	PROC(SECINFO_NO_NAME,	enc_secinfo_no_name,	dec_secinfo_no_name),
-	PROC(TEST_STATEID,	enc_test_stateid,	dec_test_stateid),
-	PROC(FREE_STATEID,	enc_free_stateid,	dec_free_stateid),
+	PROC41(EXCHANGE_ID,	enc_exchange_id,	dec_exchange_id),
+	PROC41(CREATE_SESSION,	enc_create_session,	dec_create_session),
+	PROC41(DESTROY_SESSION,	enc_destroy_session,	dec_destroy_session),
+	PROC41(SEQUENCE,	enc_sequence,		dec_sequence),
+	PROC41(GET_LEASE_TIME,	enc_get_lease_time,	dec_get_lease_time),
+	PROC41(RECLAIM_COMPLETE,enc_reclaim_complete,	dec_reclaim_complete),
+	PROC41(GETDEVICEINFO,	enc_getdeviceinfo,	dec_getdeviceinfo),
+	PROC41(LAYOUTGET,	enc_layoutget,		dec_layoutget),
+	PROC41(LAYOUTCOMMIT,	enc_layoutcommit,	dec_layoutcommit),
+	PROC41(LAYOUTRETURN,	enc_layoutreturn,	dec_layoutreturn),
+	PROC41(SECINFO_NO_NAME,	enc_secinfo_no_name,	dec_secinfo_no_name),
+	PROC41(TEST_STATEID,	enc_test_stateid,	dec_test_stateid),
+	PROC41(FREE_STATEID,	enc_free_stateid,	dec_free_stateid),
 	STUB(GETDEVICELIST),
-	PROC(BIND_CONN_TO_SESSION,
+	PROC41(BIND_CONN_TO_SESSION,
 			enc_bind_conn_to_session, dec_bind_conn_to_session),
-	PROC(DESTROY_CLIENTID,	enc_destroy_clientid,	dec_destroy_clientid),
-#endif /* CONFIG_NFS_V4_1 */
-#ifdef CONFIG_NFS_V4_2
-	PROC(SEEK,		enc_seek,		dec_seek),
-	PROC(ALLOCATE,		enc_allocate,		dec_allocate),
-	PROC(DEALLOCATE,	enc_deallocate,		dec_deallocate),
-	PROC(LAYOUTSTATS,	enc_layoutstats,	dec_layoutstats),
-	PROC(CLONE,		enc_clone,		dec_clone),
-	PROC(COPY,		enc_copy,		dec_copy),
-#endif /* CONFIG_NFS_V4_2 */
+	PROC41(DESTROY_CLIENTID,enc_destroy_clientid,	dec_destroy_clientid),
+	PROC42(SEEK,		enc_seek,		dec_seek),
+	PROC42(ALLOCATE,	enc_allocate,		dec_allocate),
+	PROC42(DEALLOCATE,	enc_deallocate,		dec_deallocate),
+	PROC42(LAYOUTSTATS,	enc_layoutstats,	dec_layoutstats),
+	PROC42(CLONE,		enc_clone,		dec_clone),
+	PROC42(COPY,		enc_copy,		dec_copy),
+	PROC(LOOKUPP,		enc_lookupp,		dec_lookupp),
 };
 
 static unsigned int nfs_version4_counts[ARRAY_SIZE(nfs4_procedures)];

fs/nfs/nfstrace.h

@@ -797,15 +797,15 @@ TRACE_EVENT(nfs_readpage_done,
 		)
 );
 
-/*
- * XXX: I tried using NFS_UNSTABLE and friends in this table, but they
- * all evaluate to 0 for some reason, even if I include linux/nfs.h.
- */
+TRACE_DEFINE_ENUM(NFS_UNSTABLE);
+TRACE_DEFINE_ENUM(NFS_DATA_SYNC);
+TRACE_DEFINE_ENUM(NFS_FILE_SYNC);
+
 #define nfs_show_stable(stable) \
 	__print_symbolic(stable, \
-			{ 0, " (UNSTABLE)" }, \
-			{ 1, " (DATA_SYNC)" }, \
-			{ 2, " (FILE_SYNC)" })
+			{ NFS_UNSTABLE, "UNSTABLE" }, \
+			{ NFS_DATA_SYNC, "DATA_SYNC" }, \
+			{ NFS_FILE_SYNC, "FILE_SYNC" })
 
 TRACE_EVENT(nfs_initiate_write,
 		TP_PROTO(
@@ -838,12 +838,12 @@ TRACE_EVENT(nfs_initiate_write,
 
 		TP_printk(
 			"fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"offset=%lld count=%lu stable=%d%s",
+			"offset=%lld count=%lu stable=%s",
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
 			__entry->fhandle,
 			__entry->offset, __entry->count,
-			__entry->stable, nfs_show_stable(__entry->stable)
+			nfs_show_stable(__entry->stable)
 		)
 );
 
@@ -882,13 +882,13 @@ TRACE_EVENT(nfs_writeback_done,
 
 		TP_printk(
 			"fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"offset=%lld status=%d stable=%d%s "
+			"offset=%lld status=%d stable=%s "
 			"verifier 0x%016llx",
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
 			__entry->fhandle,
 			__entry->offset, __entry->status,
-			__entry->stable, nfs_show_stable(__entry->stable),
+			nfs_show_stable(__entry->stable),
 			__entry->verifier
 		)
 );

fs/nfs/pagelist.c

@@ -537,7 +537,7 @@ EXPORT_SYMBOL_GPL(nfs_pgio_header_free);
  * @cinfo: Commit information for the call (writes only)
  */
 static void nfs_pgio_rpcsetup(struct nfs_pgio_header *hdr,
-			      unsigned int count, unsigned int offset,
+			      unsigned int count,
 			      int how, struct nfs_commit_info *cinfo)
 {
 	struct nfs_page *req = hdr->req;
@@ -546,10 +546,10 @@ static void nfs_pgio_rpcsetup(struct nfs_pgio_header *hdr,
 	 * NB: take care not to mess about with hdr->commit et al. */
 
 	hdr->args.fh     = NFS_FH(hdr->inode);
-	hdr->args.offset = req_offset(req) + offset;
+	hdr->args.offset = req_offset(req);
 	/* pnfs_set_layoutcommit needs this */
 	hdr->mds_offset = hdr->args.offset;
-	hdr->args.pgbase = req->wb_pgbase + offset;
+	hdr->args.pgbase = req->wb_pgbase;
 	hdr->args.pages  = hdr->page_array.pagevec;
 	hdr->args.count  = count;
 	hdr->args.context = get_nfs_open_context(req->wb_context);
@@ -789,7 +789,7 @@ int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
 		desc->pg_ioflags &= ~FLUSH_COND_STABLE;
 
 	/* Set up the argument struct */
-	nfs_pgio_rpcsetup(hdr, mirror->pg_count, 0, desc->pg_ioflags, &cinfo);
+	nfs_pgio_rpcsetup(hdr, mirror->pg_count, desc->pg_ioflags, &cinfo);
 	desc->pg_rpc_callops = &nfs_pgio_common_ops;
 	return 0;
 }

fs/nfs/pnfs.c

@@ -655,7 +655,7 @@ pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 		return 0;
 	list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list)
 		if (pnfs_match_lseg_recall(lseg, recall_range, seq)) {
-			dprintk("%s: freeing lseg %p iomode %d seq %u"
+			dprintk("%s: freeing lseg %p iomode %d seq %u "
 				"offset %llu length %llu\n", __func__,
 				lseg, lseg->pls_range.iomode, lseg->pls_seq,
 				lseg->pls_range.offset, lseg->pls_range.length);
@@ -2255,7 +2255,7 @@ pnfs_write_through_mds(struct nfs_pageio_descriptor *desc,
 		nfs_pageio_reset_write_mds(desc);
 		mirror->pg_recoalesce = 1;
 	}
-	hdr->release(hdr);
+	hdr->completion_ops->completion(hdr);
 }
 
 static enum pnfs_try_status
@@ -2378,7 +2378,7 @@ pnfs_read_through_mds(struct nfs_pageio_descriptor *desc,
 		nfs_pageio_reset_read_mds(desc);
 		mirror->pg_recoalesce = 1;
 	}
-	hdr->release(hdr);
+	hdr->completion_ops->completion(hdr);
 }
 
 /*

fs/nfs/pnfs.h

@@ -40,6 +40,7 @@ enum {
 	NFS_LSEG_ROC,		/* roc bit received from server */
 	NFS_LSEG_LAYOUTCOMMIT,	/* layoutcommit bit set for layoutcommit */
 	NFS_LSEG_LAYOUTRETURN,	/* layoutreturn bit set for layoutreturn */
+	NFS_LSEG_UNAVAILABLE,	/* unavailable bit set for temporary problem */
 };
 
 /* Individual ip address */
@@ -86,6 +87,7 @@ enum pnfs_try_status {
  */
 #define NFS4_DEF_DS_TIMEO   600 /* in tenths of a second */
 #define NFS4_DEF_DS_RETRANS 5
+#define PNFS_DEVICE_RETRY_TIMEOUT (120*HZ)
 
 /* error codes for internal use */
 #define NFS4ERR_RESET_TO_MDS   12001
@@ -524,8 +526,10 @@ static inline int pnfs_return_layout(struct inode *ino)
 	struct nfs_inode *nfsi = NFS_I(ino);
 	struct nfs_server *nfss = NFS_SERVER(ino);
 
-	if (pnfs_enabled_sb(nfss) && nfsi->layout)
+	if (pnfs_enabled_sb(nfss) && nfsi->layout) {
+		set_bit(NFS_LAYOUT_RETURN_REQUESTED, &nfsi->layout->plh_flags);
 		return _pnfs_return_layout(ino);
+	}
 
 	return 0;
 }

fs/nfs/pnfs_dev.c

@@ -43,7 +43,6 @@
 #define NFS4_DEVICE_ID_HASH_SIZE	(1 << NFS4_DEVICE_ID_HASH_BITS)
 #define NFS4_DEVICE_ID_HASH_MASK	(NFS4_DEVICE_ID_HASH_SIZE - 1)
 
-#define PNFS_DEVICE_RETRY_TIMEOUT (120*HZ)
 
 static struct hlist_head nfs4_deviceid_cache[NFS4_DEVICE_ID_HASH_SIZE];
 static DEFINE_SPINLOCK(nfs4_deviceid_lock);

fs/nfs/write.c

@@ -1835,6 +1835,8 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data)
 		set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
 	next:
 		nfs_unlock_and_release_request(req);
+		/* Latency breaker */
+		cond_resched();
 	}
 	nfss = NFS_SERVER(data->inode);
 	if (atomic_long_read(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)

include/linux/lockd/lockd.h

@@ -17,6 +17,7 @@
 #include <net/ipv6.h>
 #include <linux/fs.h>
 #include <linux/kref.h>
+#include <linux/refcount.h>
 #include <linux/utsname.h>
 #include <linux/lockd/bind.h>
 #include <linux/lockd/xdr.h>
@@ -58,7 +59,7 @@ struct nlm_host {
 	u32			h_state;	/* pseudo-state counter */
 	u32			h_nsmstate;	/* true remote NSM state */
 	u32			h_pidcount;	/* Pseudopids */
-	atomic_t		h_count;	/* reference count */
+	refcount_t		h_count;	/* reference count */
 	struct mutex		h_mutex;	/* mutex for pmap binding */
 	unsigned long		h_nextrebind;	/* next portmap call */
 	unsigned long		h_expires;	/* eligible for GC */
@@ -83,7 +84,7 @@ struct nlm_host {
 
 struct nsm_handle {
 	struct list_head	sm_link;
-	atomic_t		sm_count;
+	refcount_t		sm_count;
 	char			*sm_mon_name;
 	char			*sm_name;
 	struct sockaddr_storage	sm_addr;
@@ -122,7 +123,7 @@ static inline struct sockaddr *nlm_srcaddr(const struct nlm_host *host)
  */
 struct nlm_lockowner {
 	struct list_head list;
-	atomic_t count;
+	refcount_t count;
 
 	struct nlm_host *host;
 	fl_owner_t owner;
@@ -136,7 +137,7 @@ struct nlm_wait;
  */
 #define NLMCLNT_OHSIZE		((__NEW_UTS_LEN) + 10u)
 struct nlm_rqst {
-	atomic_t		a_count;
+	refcount_t		a_count;
 	unsigned int		a_flags;	/* initial RPC task flags */
 	struct nlm_host *	a_host;		/* host handle */
 	struct nlm_args		a_args;		/* arguments */

include/linux/nfs4.h

@@ -457,7 +457,12 @@ enum lock_type4 {
 
 #define NFS4_DEBUG 1
 
-/* Index of predefined Linux client operations */
+/*
+ * Index of predefined Linux client operations
+ *
+ * To ensure that /proc/net/rpc/nfs remains correctly ordered, please
+ * append only to this enum when adding new client operations.
+ */
 
 enum {
 	NFSPROC4_CLNT_NULL = 0,		/* Unused */
@@ -480,7 +485,6 @@ enum {
 	NFSPROC4_CLNT_ACCESS,
 	NFSPROC4_CLNT_GETATTR,
 	NFSPROC4_CLNT_LOOKUP,
-	NFSPROC4_CLNT_LOOKUPP,
 	NFSPROC4_CLNT_LOOKUP_ROOT,
 	NFSPROC4_CLNT_REMOVE,
 	NFSPROC4_CLNT_RENAME,
@@ -500,7 +504,6 @@ enum {
 	NFSPROC4_CLNT_SECINFO,
 	NFSPROC4_CLNT_FSID_PRESENT,
 
-	/* nfs41 */
 	NFSPROC4_CLNT_EXCHANGE_ID,
 	NFSPROC4_CLNT_CREATE_SESSION,
 	NFSPROC4_CLNT_DESTROY_SESSION,
@@ -518,13 +521,14 @@ enum {
 	NFSPROC4_CLNT_BIND_CONN_TO_SESSION,
 	NFSPROC4_CLNT_DESTROY_CLIENTID,
 
-	/* nfs42 */
 	NFSPROC4_CLNT_SEEK,
 	NFSPROC4_CLNT_ALLOCATE,
 	NFSPROC4_CLNT_DEALLOCATE,
 	NFSPROC4_CLNT_LAYOUTSTATS,
 	NFSPROC4_CLNT_CLONE,
 	NFSPROC4_CLNT_COPY,
+
+	NFSPROC4_CLNT_LOOKUPP,
 };
 
 /* nfs41 types */

include/linux/sunrpc/clnt.h

@@ -179,7 +179,6 @@ struct rpc_task *rpc_call_null(struct rpc_clnt *clnt, struct rpc_cred *cred,
 int		rpc_restart_call_prepare(struct rpc_task *);
 int		rpc_restart_call(struct rpc_task *);
 void		rpc_setbufsize(struct rpc_clnt *, unsigned int, unsigned int);
-int		rpc_protocol(struct rpc_clnt *);
 struct net *	rpc_net_ns(struct rpc_clnt *);
 size_t		rpc_max_payload(struct rpc_clnt *);
 size_t		rpc_max_bc_payload(struct rpc_clnt *);

include/linux/sunrpc/xprtrdma.h

@@ -64,7 +64,7 @@ enum rpcrdma_memreg {
 	RPCRDMA_MEMWINDOWS,
 	RPCRDMA_MEMWINDOWS_ASYNC,
 	RPCRDMA_MTHCAFMR,
-	RPCRDMA_FRMR,
+	RPCRDMA_FRWR,
 	RPCRDMA_ALLPHYSICAL,
 	RPCRDMA_LAST
 };

include/trace/events/rdma.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2017 Oracle.  All rights reserved.
+ */
+
+/*
+ * enum ib_event_type, from include/rdma/ib_verbs.h
+ */
+
+#define IB_EVENT_LIST				\
+	ib_event(CQ_ERR)			\
+	ib_event(QP_FATAL)			\
+	ib_event(QP_REQ_ERR)			\
+	ib_event(QP_ACCESS_ERR)			\
+	ib_event(COMM_EST)			\
+	ib_event(SQ_DRAINED)			\
+	ib_event(PATH_MIG)			\
+	ib_event(PATH_MIG_ERR)			\
+	ib_event(DEVICE_FATAL)			\
+	ib_event(PORT_ACTIVE)			\
+	ib_event(PORT_ERR)			\
+	ib_event(LID_CHANGE)			\
+	ib_event(PKEY_CHANGE)			\
+	ib_event(SM_CHANGE)			\
+	ib_event(SRQ_ERR)			\
+	ib_event(SRQ_LIMIT_REACHED)		\
+	ib_event(QP_LAST_WQE_REACHED)		\
+	ib_event(CLIENT_REREGISTER)		\
+	ib_event(GID_CHANGE)			\
+	ib_event_end(WQ_FATAL)
+
+#undef ib_event
+#undef ib_event_end
+
+#define ib_event(x)		TRACE_DEFINE_ENUM(IB_EVENT_##x);
+#define ib_event_end(x)		TRACE_DEFINE_ENUM(IB_EVENT_##x);
+
+IB_EVENT_LIST
+
+#undef ib_event
+#undef ib_event_end
+
+#define ib_event(x)		{ IB_EVENT_##x, #x },
+#define ib_event_end(x)		{ IB_EVENT_##x, #x }
+
+#define rdma_show_ib_event(x) \
+		__print_symbolic(x, IB_EVENT_LIST)
+
+/*
+ * enum ib_wc_status type, from include/rdma/ib_verbs.h
+ */
+#define IB_WC_STATUS_LIST			\
+	ib_wc_status(SUCCESS)			\
+	ib_wc_status(LOC_LEN_ERR)		\
+	ib_wc_status(LOC_QP_OP_ERR)		\
+	ib_wc_status(LOC_EEC_OP_ERR)		\
+	ib_wc_status(LOC_PROT_ERR)		\
+	ib_wc_status(WR_FLUSH_ERR)		\
+	ib_wc_status(MW_BIND_ERR)		\
+	ib_wc_status(BAD_RESP_ERR)		\
+	ib_wc_status(LOC_ACCESS_ERR)		\
+	ib_wc_status(REM_INV_REQ_ERR)		\
+	ib_wc_status(REM_ACCESS_ERR)		\
+	ib_wc_status(REM_OP_ERR)		\
+	ib_wc_status(RETRY_EXC_ERR)		\
+	ib_wc_status(RNR_RETRY_EXC_ERR)		\
+	ib_wc_status(LOC_RDD_VIOL_ERR)		\
+	ib_wc_status(REM_INV_RD_REQ_ERR)	\
+	ib_wc_status(REM_ABORT_ERR)		\
+	ib_wc_status(INV_EECN_ERR)		\
+	ib_wc_status(INV_EEC_STATE_ERR)		\
+	ib_wc_status(FATAL_ERR)			\
+	ib_wc_status(RESP_TIMEOUT_ERR)		\
+	ib_wc_status_end(GENERAL_ERR)
+
+#undef ib_wc_status
+#undef ib_wc_status_end
+
+#define ib_wc_status(x)		TRACE_DEFINE_ENUM(IB_WC_##x);
+#define ib_wc_status_end(x)	TRACE_DEFINE_ENUM(IB_WC_##x);
+
+IB_WC_STATUS_LIST
+
+#undef ib_wc_status
+#undef ib_wc_status_end
+
+#define ib_wc_status(x)		{ IB_WC_##x, #x },
+#define ib_wc_status_end(x)	{ IB_WC_##x, #x }
+
+#define rdma_show_wc_status(x) \
+		__print_symbolic(x, IB_WC_STATUS_LIST)
+
+/*
+ * enum rdma_cm_event_type, from include/rdma/rdma_cm.h
+ */
+#define RDMA_CM_EVENT_LIST			\
+	rdma_cm_event(ADDR_RESOLVED)		\
+	rdma_cm_event(ADDR_ERROR)		\
+	rdma_cm_event(ROUTE_RESOLVED)		\
+	rdma_cm_event(ROUTE_ERROR)		\
+	rdma_cm_event(CONNECT_REQUEST)		\
+	rdma_cm_event(CONNECT_RESPONSE)		\
+	rdma_cm_event(CONNECT_ERROR)		\
+	rdma_cm_event(UNREACHABLE)		\
+	rdma_cm_event(REJECTED)			\
+	rdma_cm_event(ESTABLISHED)		\
+	rdma_cm_event(DISCONNECTED)		\
+	rdma_cm_event(DEVICE_REMOVAL)		\
+	rdma_cm_event(MULTICAST_JOIN)		\
+	rdma_cm_event(MULTICAST_ERROR)		\
+	rdma_cm_event(ADDR_CHANGE)		\
+	rdma_cm_event_end(TIMEWAIT_EXIT)
+
+#undef rdma_cm_event
+#undef rdma_cm_event_end
+
+#define rdma_cm_event(x)	TRACE_DEFINE_ENUM(RDMA_CM_EVENT_##x);
+#define rdma_cm_event_end(x)	TRACE_DEFINE_ENUM(RDMA_CM_EVENT_##x);
+
+RDMA_CM_EVENT_LIST
+
+#undef rdma_cm_event
+#undef rdma_cm_event_end
+
+#define rdma_cm_event(x)	{ RDMA_CM_EVENT_##x, #x },
+#define rdma_cm_event_end(x)	{ RDMA_CM_EVENT_##x, #x }
+
+#define rdma_show_cm_event(x) \
+		__print_symbolic(x, RDMA_CM_EVENT_LIST)

include/trace/events/sunrpc.h

@@ -32,7 +32,7 @@ DECLARE_EVENT_CLASS(rpc_task_status,
 		__entry->status = task->tk_status;
 	),
 
-	TP_printk("task:%u@%u, status %d",
+	TP_printk("task:%u@%u status=%d",
 		__entry->task_id, __entry->client_id,
 		__entry->status)
 );
@@ -66,7 +66,7 @@ TRACE_EVENT(rpc_connect_status,
 		__entry->status = status;
 	),
 
-	TP_printk("task:%u@%u, status %d",
+	TP_printk("task:%u@%u status=%d",
 		__entry->task_id, __entry->client_id,
 		__entry->status)
 );
@@ -175,7 +175,7 @@ DECLARE_EVENT_CLASS(rpc_task_queued,
 		),
 
 	TP_fast_assign(
-		__entry->client_id = clnt->cl_clid;
+		__entry->client_id = clnt ? clnt->cl_clid : -1;
 		__entry->task_id = task->tk_pid;
 		__entry->timeout = task->tk_timeout;
 		__entry->runstate = task->tk_runstate;
@@ -184,7 +184,7 @@ DECLARE_EVENT_CLASS(rpc_task_queued,
 		__assign_str(q_name, rpc_qname(q));
 		),
 
-	TP_printk("task:%u@%u flags=%4.4x state=%4.4lx status=%d timeout=%lu queue=%s",
+	TP_printk("task:%u@%d flags=%4.4x state=%4.4lx status=%d timeout=%lu queue=%s",
 		__entry->task_id, __entry->client_id,
 		__entry->flags,
 		__entry->runstate,
@@ -390,6 +390,10 @@ DECLARE_EVENT_CLASS(rpc_xprt_event,
 			__entry->status)
 );
 
+DEFINE_EVENT(rpc_xprt_event, xprt_timer,
+	TP_PROTO(struct rpc_xprt *xprt, __be32 xid, int status),
+	TP_ARGS(xprt, xid, status));
+
 DEFINE_EVENT(rpc_xprt_event, xprt_lookup_rqst,
 	TP_PROTO(struct rpc_xprt *xprt, __be32 xid, int status),
 	TP_ARGS(xprt, xid, status));

include/uapi/linux/nfs.h

@@ -12,6 +12,7 @@
 
 #define NFS_PROGRAM	100003
 #define NFS_PORT	2049
+#define NFS_RDMA_PORT	20049
 #define NFS_MAXDATA	8192
 #define NFS_MAXPATHLEN	1024
 #define NFS_MAXNAMLEN	255

net/sunrpc/clnt.c

@@ -1375,22 +1375,6 @@ rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize
 }
 EXPORT_SYMBOL_GPL(rpc_setbufsize);
 
-/**
- * rpc_protocol - Get transport protocol number for an RPC client
- * @clnt: RPC client to query
- *
- */
-int rpc_protocol(struct rpc_clnt *clnt)
-{
-	int protocol;
-
-	rcu_read_lock();
-	protocol = rcu_dereference(clnt->cl_xprt)->prot;
-	rcu_read_unlock();
-	return protocol;
-}
-EXPORT_SYMBOL_GPL(rpc_protocol);
-
 /**
  * rpc_net_ns - Get the network namespace for this RPC client
  * @clnt: RPC client to query

net/sunrpc/sched.c

@@ -755,22 +755,20 @@ static void __rpc_execute(struct rpc_task *task)
 		void (*do_action)(struct rpc_task *);
 
 		/*
-		 * Execute any pending callback first.
+		 * Perform the next FSM step or a pending callback.
+		 *
+		 * tk_action may be NULL if the task has been killed.
+		 * In particular, note that rpc_killall_tasks may
+		 * do this at any time, so beware when dereferencing.
 		 */
-		do_action = task->tk_callback;
-		task->tk_callback = NULL;
-		if (do_action == NULL) {
-			/*
-			 * Perform the next FSM step.
-			 * tk_action may be NULL if the task has been killed.
-			 * In particular, note that rpc_killall_tasks may
-			 * do this at any time, so beware when dereferencing.
-			 */
-			do_action = task->tk_action;
-			if (do_action == NULL)
-				break;
+		do_action = task->tk_action;
+		if (task->tk_callback) {
+			do_action = task->tk_callback;
+			task->tk_callback = NULL;
 		}
-		trace_rpc_task_run_action(task->tk_client, task, task->tk_action);
+		if (!do_action)
+			break;
+		trace_rpc_task_run_action(task->tk_client, task, do_action);
 		do_action(task);
 
 		/*

net/sunrpc/xprt.c

@@ -940,8 +940,8 @@ static void xprt_timer(struct rpc_task *task)
 
 	if (task->tk_status != -ETIMEDOUT)
 		return;
-	dprintk("RPC: %5u xprt_timer\n", task->tk_pid);
 
+	trace_xprt_timer(xprt, req->rq_xid, task->tk_status);
 	if (!req->rq_reply_bytes_recvd) {
 		if (xprt->ops->timer)
 			xprt->ops->timer(xprt, task);

net/sunrpc/xprtrdma/backchannel.c

@@ -43,7 +43,6 @@ static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
 	req = rpcrdma_create_req(r_xprt);
 	if (IS_ERR(req))
 		return PTR_ERR(req);
-	__set_bit(RPCRDMA_REQ_F_BACKCHANNEL, &req->rl_flags);
 
 	rb = rpcrdma_alloc_regbuf(RPCRDMA_HDRBUF_SIZE,
 				  DMA_TO_DEVICE, GFP_KERNEL);
@@ -74,21 +73,13 @@ static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
 static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt,
 				 unsigned int count)
 {
-	struct rpcrdma_rep *rep;
 	int rc = 0;
 
 	while (count--) {
-		rep = rpcrdma_create_rep(r_xprt);
-		if (IS_ERR(rep)) {
-			pr_err("RPC:       %s: reply buffer alloc failed\n",
-			       __func__);
-			rc = PTR_ERR(rep);
+		rc = rpcrdma_create_rep(r_xprt);
+		if (rc)
 			break;
-		}
-
-		rpcrdma_recv_buffer_put(rep);
 	}
-
 	return rc;
 }
 
@@ -129,6 +120,7 @@ int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs)
 		rqst->rq_xprt = &r_xprt->rx_xprt;
 		INIT_LIST_HEAD(&rqst->rq_list);
 		INIT_LIST_HEAD(&rqst->rq_bc_list);
+		__set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
 
 		if (rpcrdma_bc_setup_rqst(r_xprt, rqst))
 			goto out_free;
@@ -148,7 +140,7 @@ int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs)
 
 	buffer->rb_bc_srv_max_requests = reqs;
 	request_module("svcrdma");
-
+	trace_xprtrdma_cb_setup(r_xprt, reqs);
 	return 0;
 
 out_free:
@@ -196,13 +188,7 @@ size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *xprt)
 	return maxmsg - RPCRDMA_HDRLEN_MIN;
 }
 
-/**
- * rpcrdma_bc_marshal_reply - Send backwards direction reply
- * @rqst: buffer containing RPC reply data
- *
- * Returns zero on success.
- */
-int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
+static int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
@@ -226,7 +212,46 @@ int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
 	if (rpcrdma_prepare_send_sges(r_xprt, req, RPCRDMA_HDRLEN_MIN,
 				      &rqst->rq_snd_buf, rpcrdma_noch))
 		return -EIO;
+
+	trace_xprtrdma_cb_reply(rqst);
+	return 0;
+}
+
+/**
+ * xprt_rdma_bc_send_reply - marshal and send a backchannel reply
+ * @rqst: RPC rqst with a backchannel RPC reply in rq_snd_buf
+ *
+ * Caller holds the transport's write lock.
+ *
+ * Returns:
+ *	%0 if the RPC message has been sent
+ *	%-ENOTCONN if the caller should reconnect and call again
+ *	%-EIO if a permanent error occurred and the request was not
+ *		sent. Do not try to send this message again.
+ */
+int xprt_rdma_bc_send_reply(struct rpc_rqst *rqst)
+{
+	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
+	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
+	int rc;
+
+	if (!xprt_connected(rqst->rq_xprt))
+		goto drop_connection;
+
+	rc = rpcrdma_bc_marshal_reply(rqst);
+	if (rc < 0)
+		goto failed_marshal;
+
+	if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req))
+		goto drop_connection;
 	return 0;
+
+failed_marshal:
+	if (rc != -ENOTCONN)
+		return rc;
+drop_connection:
+	xprt_disconnect_done(rqst->rq_xprt);
+	return -ENOTCONN;
 }
 
 /**
@@ -262,11 +287,6 @@ void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
 	dprintk("RPC:       %s: freeing rqst %p (req %p)\n",
 		__func__, rqst, rpcr_to_rdmar(rqst));
 
-	smp_mb__before_atomic();
-	WARN_ON_ONCE(!test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state));
-	clear_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
-	smp_mb__after_atomic();
-
 	spin_lock_bh(&xprt->bc_pa_lock);
 	list_add_tail(&rqst->rq_bc_pa_list, &xprt->bc_pa_list);
 	spin_unlock_bh(&xprt->bc_pa_lock);
@@ -274,7 +294,7 @@ void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
 
 /**
  * rpcrdma_bc_receive_call - Handle a backward direction call
- * @xprt: transport receiving the call
+ * @r_xprt: transport receiving the call
  * @rep: receive buffer containing the call
  *
  * Operational assumptions:
@@ -313,7 +333,6 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 				struct rpc_rqst, rq_bc_pa_list);
 	list_del(&rqst->rq_bc_pa_list);
 	spin_unlock(&xprt->bc_pa_lock);
-	dprintk("RPC:       %s: using rqst %p\n", __func__, rqst);
 
 	/* Prepare rqst */
 	rqst->rq_reply_bytes_recvd = 0;
@@ -321,7 +340,6 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	rqst->rq_xid = *p;
 
 	rqst->rq_private_buf.len = size;
-	set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state);
 
 	buf = &rqst->rq_rcv_buf;
 	memset(buf, 0, sizeof(*buf));
@@ -335,12 +353,8 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	 * the Upper Layer is done decoding it.
 	 */
 	req = rpcr_to_rdmar(rqst);
-	dprintk("RPC:       %s: attaching rep %p to req %p\n",
-		__func__, rep, req);
 	req->rl_reply = rep;
-
-	/* Defeat the retransmit detection logic in send_request */
-	req->rl_connect_cookie = 0;
+	trace_xprtrdma_cb_call(rqst);
 
 	/* Queue rqst for ULP's callback service */
 	bc_serv = xprt->bc_serv;

net/sunrpc/xprtrdma/fmr_ops.c

 // SPDX-License-Identifier: GPL-2.0
 /*
- * Copyright (c) 2015 Oracle.  All rights reserved.
+ * Copyright (c) 2015, 2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  */
 
@@ -47,7 +47,7 @@ fmr_is_supported(struct rpcrdma_ia *ia)
 }
 
 static int
-fmr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mw *mw)
+fmr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mr *mr)
 {
 	static struct ib_fmr_attr fmr_attr = {
 		.max_pages	= RPCRDMA_MAX_FMR_SGES,
@@ -55,106 +55,108 @@ fmr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mw *mw)
 		.page_shift	= PAGE_SHIFT
 	};
 
-	mw->fmr.fm_physaddrs = kcalloc(RPCRDMA_MAX_FMR_SGES,
+	mr->fmr.fm_physaddrs = kcalloc(RPCRDMA_MAX_FMR_SGES,
 				       sizeof(u64), GFP_KERNEL);
-	if (!mw->fmr.fm_physaddrs)
+	if (!mr->fmr.fm_physaddrs)
 		goto out_free;
 
-	mw->mw_sg = kcalloc(RPCRDMA_MAX_FMR_SGES,
-			    sizeof(*mw->mw_sg), GFP_KERNEL);
-	if (!mw->mw_sg)
+	mr->mr_sg = kcalloc(RPCRDMA_MAX_FMR_SGES,
+			    sizeof(*mr->mr_sg), GFP_KERNEL);
+	if (!mr->mr_sg)
 		goto out_free;
 
-	sg_init_table(mw->mw_sg, RPCRDMA_MAX_FMR_SGES);
+	sg_init_table(mr->mr_sg, RPCRDMA_MAX_FMR_SGES);
 
-	mw->fmr.fm_mr = ib_alloc_fmr(ia->ri_pd, RPCRDMA_FMR_ACCESS_FLAGS,
+	mr->fmr.fm_mr = ib_alloc_fmr(ia->ri_pd, RPCRDMA_FMR_ACCESS_FLAGS,
 				     &fmr_attr);
-	if (IS_ERR(mw->fmr.fm_mr))
+	if (IS_ERR(mr->fmr.fm_mr))
 		goto out_fmr_err;
 
 	return 0;
 
 out_fmr_err:
 	dprintk("RPC:       %s: ib_alloc_fmr returned %ld\n", __func__,
-		PTR_ERR(mw->fmr.fm_mr));
+		PTR_ERR(mr->fmr.fm_mr));
 
 out_free:
-	kfree(mw->mw_sg);
-	kfree(mw->fmr.fm_physaddrs);
+	kfree(mr->mr_sg);
+	kfree(mr->fmr.fm_physaddrs);
 	return -ENOMEM;
 }
 
 static int
-__fmr_unmap(struct rpcrdma_mw *mw)
+__fmr_unmap(struct rpcrdma_mr *mr)
 {
 	LIST_HEAD(l);
 	int rc;
 
-	list_add(&mw->fmr.fm_mr->list, &l);
+	list_add(&mr->fmr.fm_mr->list, &l);
 	rc = ib_unmap_fmr(&l);
-	list_del(&mw->fmr.fm_mr->list);
+	list_del(&mr->fmr.fm_mr->list);
 	return rc;
 }
 
 static void
-fmr_op_release_mr(struct rpcrdma_mw *r)
+fmr_op_release_mr(struct rpcrdma_mr *mr)
 {
 	LIST_HEAD(unmap_list);
 	int rc;
 
 	/* Ensure MW is not on any rl_registered list */
-	if (!list_empty(&r->mw_list))
-		list_del(&r->mw_list);
+	if (!list_empty(&mr->mr_list))
+		list_del(&mr->mr_list);
 
-	kfree(r->fmr.fm_physaddrs);
-	kfree(r->mw_sg);
+	kfree(mr->fmr.fm_physaddrs);
+	kfree(mr->mr_sg);
 
 	/* In case this one was left mapped, try to unmap it
 	 * to prevent dealloc_fmr from failing with EBUSY
 	 */
-	rc = __fmr_unmap(r);
+	rc = __fmr_unmap(mr);
 	if (rc)
 		pr_err("rpcrdma: final ib_unmap_fmr for %p failed %i\n",
-		       r, rc);
+		       mr, rc);
 
-	rc = ib_dealloc_fmr(r->fmr.fm_mr);
+	rc = ib_dealloc_fmr(mr->fmr.fm_mr);
 	if (rc)
 		pr_err("rpcrdma: final ib_dealloc_fmr for %p returned %i\n",
-		       r, rc);
+		       mr, rc);
 
-	kfree(r);
+	kfree(mr);
 }
 
 /* Reset of a single FMR.
  */
 static void
-fmr_op_recover_mr(struct rpcrdma_mw *mw)
+fmr_op_recover_mr(struct rpcrdma_mr *mr)
 {
-	struct rpcrdma_xprt *r_xprt = mw->mw_xprt;
+	struct rpcrdma_xprt *r_xprt = mr->mr_xprt;
 	int rc;
 
 	/* ORDER: invalidate first */
-	rc = __fmr_unmap(mw);
-
-	/* ORDER: then DMA unmap */
-	ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
-			mw->mw_sg, mw->mw_nents, mw->mw_dir);
+	rc = __fmr_unmap(mr);
 	if (rc)
 		goto out_release;
 
-	rpcrdma_put_mw(r_xprt, mw);
+	/* ORDER: then DMA unmap */
+	rpcrdma_mr_unmap_and_put(mr);
+
 	r_xprt->rx_stats.mrs_recovered++;
 	return;
 
 out_release:
-	pr_err("rpcrdma: FMR reset failed (%d), %p released\n", rc, mw);
+	pr_err("rpcrdma: FMR reset failed (%d), %p released\n", rc, mr);
 	r_xprt->rx_stats.mrs_orphaned++;
 
-	spin_lock(&r_xprt->rx_buf.rb_mwlock);
-	list_del(&mw->mw_all);
-	spin_unlock(&r_xprt->rx_buf.rb_mwlock);
+	trace_xprtrdma_dma_unmap(mr);
+	ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
+			mr->mr_sg, mr->mr_nents, mr->mr_dir);
+
+	spin_lock(&r_xprt->rx_buf.rb_mrlock);
+	list_del(&mr->mr_all);
+	spin_unlock(&r_xprt->rx_buf.rb_mrlock);
 
-	fmr_op_release_mr(mw);
+	fmr_op_release_mr(mr);
 }
 
 static int
@@ -180,15 +182,15 @@ fmr_op_maxpages(struct rpcrdma_xprt *r_xprt)
  */
 static struct rpcrdma_mr_seg *
 fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
-	   int nsegs, bool writing, struct rpcrdma_mw **out)
+	   int nsegs, bool writing, struct rpcrdma_mr **out)
 {
 	struct rpcrdma_mr_seg *seg1 = seg;
 	int len, pageoff, i, rc;
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 	u64 *dma_pages;
 
-	mw = rpcrdma_get_mw(r_xprt);
-	if (!mw)
+	mr = rpcrdma_mr_get(r_xprt);
+	if (!mr)
 		return ERR_PTR(-ENOBUFS);
 
 	pageoff = offset_in_page(seg1->mr_offset);
@@ -199,12 +201,12 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		nsegs = RPCRDMA_MAX_FMR_SGES;
 	for (i = 0; i < nsegs;) {
 		if (seg->mr_page)
-			sg_set_page(&mw->mw_sg[i],
+			sg_set_page(&mr->mr_sg[i],
 				    seg->mr_page,
 				    seg->mr_len,
 				    offset_in_page(seg->mr_offset));
 		else
-			sg_set_buf(&mw->mw_sg[i], seg->mr_offset,
+			sg_set_buf(&mr->mr_sg[i], seg->mr_offset,
 				   seg->mr_len);
 		len += seg->mr_len;
 		++seg;
@@ -214,40 +216,38 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
 			break;
 	}
-	mw->mw_dir = rpcrdma_data_dir(writing);
+	mr->mr_dir = rpcrdma_data_dir(writing);
 
-	mw->mw_nents = ib_dma_map_sg(r_xprt->rx_ia.ri_device,
-				     mw->mw_sg, i, mw->mw_dir);
-	if (!mw->mw_nents)
+	mr->mr_nents = ib_dma_map_sg(r_xprt->rx_ia.ri_device,
+				     mr->mr_sg, i, mr->mr_dir);
+	if (!mr->mr_nents)
 		goto out_dmamap_err;
 
-	for (i = 0, dma_pages = mw->fmr.fm_physaddrs; i < mw->mw_nents; i++)
-		dma_pages[i] = sg_dma_address(&mw->mw_sg[i]);
-	rc = ib_map_phys_fmr(mw->fmr.fm_mr, dma_pages, mw->mw_nents,
+	for (i = 0, dma_pages = mr->fmr.fm_physaddrs; i < mr->mr_nents; i++)
+		dma_pages[i] = sg_dma_address(&mr->mr_sg[i]);
+	rc = ib_map_phys_fmr(mr->fmr.fm_mr, dma_pages, mr->mr_nents,
 			     dma_pages[0]);
 	if (rc)
 		goto out_maperr;
 
-	mw->mw_handle = mw->fmr.fm_mr->rkey;
-	mw->mw_length = len;
-	mw->mw_offset = dma_pages[0] + pageoff;
+	mr->mr_handle = mr->fmr.fm_mr->rkey;
+	mr->mr_length = len;
+	mr->mr_offset = dma_pages[0] + pageoff;
 
-	*out = mw;
+	*out = mr;
 	return seg;
 
 out_dmamap_err:
 	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
-	       mw->mw_sg, i);
-	rpcrdma_put_mw(r_xprt, mw);
+	       mr->mr_sg, i);
+	rpcrdma_mr_put(mr);
 	return ERR_PTR(-EIO);
 
 out_maperr:
 	pr_err("rpcrdma: ib_map_phys_fmr %u@0x%llx+%i (%d) status %i\n",
 	       len, (unsigned long long)dma_pages[0],
-	       pageoff, mw->mw_nents, rc);
-	ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
-			mw->mw_sg, mw->mw_nents, mw->mw_dir);
-	rpcrdma_put_mw(r_xprt, mw);
+	       pageoff, mr->mr_nents, rc);
+	rpcrdma_mr_unmap_and_put(mr);
 	return ERR_PTR(-EIO);
 }
 
@@ -256,13 +256,13 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
  * Sleeps until it is safe for the host CPU to access the
  * previously mapped memory regions.
  *
- * Caller ensures that @mws is not empty before the call. This
+ * Caller ensures that @mrs is not empty before the call. This
  * function empties the list.
  */
 static void
-fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
+fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mrs)
 {
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 	LIST_HEAD(unmap_list);
 	int rc;
 
@@ -271,10 +271,11 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	 * ib_unmap_fmr() is slow, so use a single call instead
 	 * of one call per mapped FMR.
 	 */
-	list_for_each_entry(mw, mws, mw_list) {
+	list_for_each_entry(mr, mrs, mr_list) {
 		dprintk("RPC:       %s: unmapping fmr %p\n",
-			__func__, &mw->fmr);
-		list_add_tail(&mw->fmr.fm_mr->list, &unmap_list);
+			__func__, &mr->fmr);
+		trace_xprtrdma_localinv(mr);
+		list_add_tail(&mr->fmr.fm_mr->list, &unmap_list);
 	}
 	r_xprt->rx_stats.local_inv_needed++;
 	rc = ib_unmap_fmr(&unmap_list);
@@ -284,14 +285,10 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	/* ORDER: Now DMA unmap all of the req's MRs, and return
 	 * them to the free MW list.
 	 */
-	while (!list_empty(mws)) {
-		mw = rpcrdma_pop_mw(mws);
-		dprintk("RPC:       %s: DMA unmapping fmr %p\n",
-			__func__, &mw->fmr);
-		list_del(&mw->fmr.fm_mr->list);
-		ib_dma_unmap_sg(r_xprt->rx_ia.ri_device,
-				mw->mw_sg, mw->mw_nents, mw->mw_dir);
-		rpcrdma_put_mw(r_xprt, mw);
+	while (!list_empty(mrs)) {
+		mr = rpcrdma_mr_pop(mrs);
+		list_del(&mr->fmr.fm_mr->list);
+		rpcrdma_mr_unmap_and_put(mr);
 	}
 
 	return;
@@ -299,10 +296,10 @@ fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 out_reset:
 	pr_err("rpcrdma: ib_unmap_fmr failed (%i)\n", rc);
 
-	while (!list_empty(mws)) {
-		mw = rpcrdma_pop_mw(mws);
-		list_del(&mw->fmr.fm_mr->list);
-		fmr_op_recover_mr(mw);
+	while (!list_empty(mrs)) {
+		mr = rpcrdma_mr_pop(mrs);
+		list_del(&mr->fmr.fm_mr->list);
+		fmr_op_recover_mr(mr);
 	}
 }
 

net/sunrpc/xprtrdma/frwr_ops.c

 // SPDX-License-Identifier: GPL-2.0
 /*
- * Copyright (c) 2015 Oracle.  All rights reserved.
+ * Copyright (c) 2015, 2017 Oracle.  All rights reserved.
  * Copyright (c) 2003-2007 Network Appliance, Inc. All rights reserved.
  */
 
 /* Lightweight memory registration using Fast Registration Work
- * Requests (FRWR). Also referred to sometimes as FRMR mode.
+ * Requests (FRWR).
  *
  * FRWR features ordered asynchronous registration and deregistration
  * of arbitrarily sized memory regions. This is the fastest and safest
@@ -15,9 +15,9 @@
 /* Normal operation
  *
  * A Memory Region is prepared for RDMA READ or WRITE using a FAST_REG
- * Work Request (frmr_op_map). When the RDMA operation is finished, this
+ * Work Request (frwr_op_map). When the RDMA operation is finished, this
  * Memory Region is invalidated using a LOCAL_INV Work Request
- * (frmr_op_unmap).
+ * (frwr_op_unmap_sync).
  *
  * Typically these Work Requests are not signaled, and neither are RDMA
  * SEND Work Requests (with the exception of signaling occasionally to
@@ -26,7 +26,7 @@
  *
  * As an optimization, frwr_op_unmap marks MRs INVALID before the
  * LOCAL_INV WR is posted. If posting succeeds, the MR is placed on
- * rb_mws immediately so that no work (like managing a linked list
+ * rb_mrs immediately so that no work (like managing a linked list
  * under a spinlock) is needed in the completion upcall.
  *
  * But this means that frwr_op_map() can occasionally encounter an MR
@@ -60,7 +60,7 @@
  * When frwr_op_map encounters FLUSHED and VALID MRs, they are recovered
  * with ib_dereg_mr and then are re-initialized. Because MR recovery
  * allocates fresh resources, it is deferred to a workqueue, and the
- * recovered MRs are placed back on the rb_mws list when recovery is
+ * recovered MRs are placed back on the rb_mrs list when recovery is
  * complete. frwr_op_map allocates another MR for the current RPC while
  * the broken MR is reset.
  *
@@ -96,26 +96,26 @@ frwr_is_supported(struct rpcrdma_ia *ia)
 }
 
 static int
-frwr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mw *r)
+frwr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mr *mr)
 {
-	unsigned int depth = ia->ri_max_frmr_depth;
-	struct rpcrdma_frmr *f = &r->frmr;
+	unsigned int depth = ia->ri_max_frwr_depth;
+	struct rpcrdma_frwr *frwr = &mr->frwr;
 	int rc;
 
-	f->fr_mr = ib_alloc_mr(ia->ri_pd, ia->ri_mrtype, depth);
-	if (IS_ERR(f->fr_mr))
+	frwr->fr_mr = ib_alloc_mr(ia->ri_pd, ia->ri_mrtype, depth);
+	if (IS_ERR(frwr->fr_mr))
 		goto out_mr_err;
 
-	r->mw_sg = kcalloc(depth, sizeof(*r->mw_sg), GFP_KERNEL);
-	if (!r->mw_sg)
+	mr->mr_sg = kcalloc(depth, sizeof(*mr->mr_sg), GFP_KERNEL);
+	if (!mr->mr_sg)
 		goto out_list_err;
 
-	sg_init_table(r->mw_sg, depth);
-	init_completion(&f->fr_linv_done);
+	sg_init_table(mr->mr_sg, depth);
+	init_completion(&frwr->fr_linv_done);
 	return 0;
 
 out_mr_err:
-	rc = PTR_ERR(f->fr_mr);
+	rc = PTR_ERR(frwr->fr_mr);
 	dprintk("RPC:       %s: ib_alloc_mr status %i\n",
 		__func__, rc);
 	return rc;
@@ -124,83 +124,85 @@ frwr_op_init_mr(struct rpcrdma_ia *ia, struct rpcrdma_mw *r)
 	rc = -ENOMEM;
 	dprintk("RPC:       %s: sg allocation failure\n",
 		__func__);
-	ib_dereg_mr(f->fr_mr);
+	ib_dereg_mr(frwr->fr_mr);
 	return rc;
 }
 
 static void
-frwr_op_release_mr(struct rpcrdma_mw *r)
+frwr_op_release_mr(struct rpcrdma_mr *mr)
 {
 	int rc;
 
-	/* Ensure MW is not on any rl_registered list */
-	if (!list_empty(&r->mw_list))
-		list_del(&r->mw_list);
+	/* Ensure MR is not on any rl_registered list */
+	if (!list_empty(&mr->mr_list))
+		list_del(&mr->mr_list);
 
-	rc = ib_dereg_mr(r->frmr.fr_mr);
+	rc = ib_dereg_mr(mr->frwr.fr_mr);
 	if (rc)
 		pr_err("rpcrdma: final ib_dereg_mr for %p returned %i\n",
-		       r, rc);
-	kfree(r->mw_sg);
-	kfree(r);
+		       mr, rc);
+	kfree(mr->mr_sg);
+	kfree(mr);
 }
 
 static int
-__frwr_reset_mr(struct rpcrdma_ia *ia, struct rpcrdma_mw *r)
+__frwr_mr_reset(struct rpcrdma_ia *ia, struct rpcrdma_mr *mr)
 {
-	struct rpcrdma_frmr *f = &r->frmr;
+	struct rpcrdma_frwr *frwr = &mr->frwr;
 	int rc;
 
-	rc = ib_dereg_mr(f->fr_mr);
+	rc = ib_dereg_mr(frwr->fr_mr);
 	if (rc) {
 		pr_warn("rpcrdma: ib_dereg_mr status %d, frwr %p orphaned\n",
-			rc, r);
+			rc, mr);
 		return rc;
 	}
 
-	f->fr_mr = ib_alloc_mr(ia->ri_pd, ia->ri_mrtype,
-			       ia->ri_max_frmr_depth);
-	if (IS_ERR(f->fr_mr)) {
+	frwr->fr_mr = ib_alloc_mr(ia->ri_pd, ia->ri_mrtype,
+				  ia->ri_max_frwr_depth);
+	if (IS_ERR(frwr->fr_mr)) {
 		pr_warn("rpcrdma: ib_alloc_mr status %ld, frwr %p orphaned\n",
-			PTR_ERR(f->fr_mr), r);
-		return PTR_ERR(f->fr_mr);
+			PTR_ERR(frwr->fr_mr), mr);
+		return PTR_ERR(frwr->fr_mr);
 	}
 
-	dprintk("RPC:       %s: recovered FRMR %p\n", __func__, f);
-	f->fr_state = FRMR_IS_INVALID;
+	dprintk("RPC:       %s: recovered FRWR %p\n", __func__, frwr);
+	frwr->fr_state = FRWR_IS_INVALID;
 	return 0;
 }
 
-/* Reset of a single FRMR. Generate a fresh rkey by replacing the MR.
+/* Reset of a single FRWR. Generate a fresh rkey by replacing the MR.
  */
 static void
-frwr_op_recover_mr(struct rpcrdma_mw *mw)
+frwr_op_recover_mr(struct rpcrdma_mr *mr)
 {
-	enum rpcrdma_frmr_state state = mw->frmr.fr_state;
-	struct rpcrdma_xprt *r_xprt = mw->mw_xprt;
+	enum rpcrdma_frwr_state state = mr->frwr.fr_state;
+	struct rpcrdma_xprt *r_xprt = mr->mr_xprt;
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	int rc;
 
-	rc = __frwr_reset_mr(ia, mw);
-	if (state != FRMR_FLUSHED_LI)
+	rc = __frwr_mr_reset(ia, mr);
+	if (state != FRWR_FLUSHED_LI) {
+		trace_xprtrdma_dma_unmap(mr);
 		ib_dma_unmap_sg(ia->ri_device,
-				mw->mw_sg, mw->mw_nents, mw->mw_dir);
+				mr->mr_sg, mr->mr_nents, mr->mr_dir);
+	}
 	if (rc)
 		goto out_release;
 
-	rpcrdma_put_mw(r_xprt, mw);
+	rpcrdma_mr_put(mr);
 	r_xprt->rx_stats.mrs_recovered++;
 	return;
 
 out_release:
-	pr_err("rpcrdma: FRMR reset failed %d, %p release\n", rc, mw);
+	pr_err("rpcrdma: FRWR reset failed %d, %p release\n", rc, mr);
 	r_xprt->rx_stats.mrs_orphaned++;
 
-	spin_lock(&r_xprt->rx_buf.rb_mwlock);
-	list_del(&mw->mw_all);
-	spin_unlock(&r_xprt->rx_buf.rb_mwlock);
+	spin_lock(&r_xprt->rx_buf.rb_mrlock);
+	list_del(&mr->mr_all);
+	spin_unlock(&r_xprt->rx_buf.rb_mrlock);
 
-	frwr_op_release_mr(mw);
+	frwr_op_release_mr(mr);
 }
 
 static int
@@ -214,31 +216,31 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 	if (attrs->device_cap_flags & IB_DEVICE_SG_GAPS_REG)
 		ia->ri_mrtype = IB_MR_TYPE_SG_GAPS;
 
-	ia->ri_max_frmr_depth =
+	ia->ri_max_frwr_depth =
 			min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
 			      attrs->max_fast_reg_page_list_len);
 	dprintk("RPC:       %s: device's max FR page list len = %u\n",
-		__func__, ia->ri_max_frmr_depth);
-
-	/* Add room for frmr register and invalidate WRs.
-	 * 1. FRMR reg WR for head
-	 * 2. FRMR invalidate WR for head
-	 * 3. N FRMR reg WRs for pagelist
-	 * 4. N FRMR invalidate WRs for pagelist
-	 * 5. FRMR reg WR for tail
-	 * 6. FRMR invalidate WR for tail
+		__func__, ia->ri_max_frwr_depth);
+
+	/* Add room for frwr register and invalidate WRs.
+	 * 1. FRWR reg WR for head
+	 * 2. FRWR invalidate WR for head
+	 * 3. N FRWR reg WRs for pagelist
+	 * 4. N FRWR invalidate WRs for pagelist
+	 * 5. FRWR reg WR for tail
+	 * 6. FRWR invalidate WR for tail
 	 * 7. The RDMA_SEND WR
 	 */
 	depth = 7;
 
-	/* Calculate N if the device max FRMR depth is smaller than
+	/* Calculate N if the device max FRWR depth is smaller than
 	 * RPCRDMA_MAX_DATA_SEGS.
 	 */
-	if (ia->ri_max_frmr_depth < RPCRDMA_MAX_DATA_SEGS) {
-		delta = RPCRDMA_MAX_DATA_SEGS - ia->ri_max_frmr_depth;
+	if (ia->ri_max_frwr_depth < RPCRDMA_MAX_DATA_SEGS) {
+		delta = RPCRDMA_MAX_DATA_SEGS - ia->ri_max_frwr_depth;
 		do {
-			depth += 2; /* FRMR reg + invalidate */
-			delta -= ia->ri_max_frmr_depth;
+			depth += 2; /* FRWR reg + invalidate */
+			delta -= ia->ri_max_frwr_depth;
 		} while (delta > 0);
 	}
 
@@ -252,7 +254,7 @@ frwr_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep,
 	}
 
 	ia->ri_max_segs = max_t(unsigned int, 1, RPCRDMA_MAX_DATA_SEGS /
-				ia->ri_max_frmr_depth);
+				ia->ri_max_frwr_depth);
 	return 0;
 }
 
@@ -265,7 +267,7 @@ frwr_op_maxpages(struct rpcrdma_xprt *r_xprt)
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 
 	return min_t(unsigned int, RPCRDMA_MAX_DATA_SEGS,
-		     RPCRDMA_MAX_HDR_SEGS * ia->ri_max_frmr_depth);
+		     RPCRDMA_MAX_HDR_SEGS * ia->ri_max_frwr_depth);
 }
 
 static void
@@ -286,16 +288,16 @@ __frwr_sendcompletion_flush(struct ib_wc *wc, const char *wr)
 static void
 frwr_wc_fastreg(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct rpcrdma_frmr *frmr;
-	struct ib_cqe *cqe;
+	struct ib_cqe *cqe = wc->wr_cqe;
+	struct rpcrdma_frwr *frwr =
+			container_of(cqe, struct rpcrdma_frwr, fr_cqe);
 
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	if (wc->status != IB_WC_SUCCESS) {
-		cqe = wc->wr_cqe;
-		frmr = container_of(cqe, struct rpcrdma_frmr, fr_cqe);
-		frmr->fr_state = FRMR_FLUSHED_FR;
+		frwr->fr_state = FRWR_FLUSHED_FR;
 		__frwr_sendcompletion_flush(wc, "fastreg");
 	}
+	trace_xprtrdma_wc_fastreg(wc, frwr);
 }
 
 /**
@@ -307,16 +309,16 @@ frwr_wc_fastreg(struct ib_cq *cq, struct ib_wc *wc)
 static void
 frwr_wc_localinv(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct rpcrdma_frmr *frmr;
-	struct ib_cqe *cqe;
+	struct ib_cqe *cqe = wc->wr_cqe;
+	struct rpcrdma_frwr *frwr = container_of(cqe, struct rpcrdma_frwr,
+						 fr_cqe);
 
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	if (wc->status != IB_WC_SUCCESS) {
-		cqe = wc->wr_cqe;
-		frmr = container_of(cqe, struct rpcrdma_frmr, fr_cqe);
-		frmr->fr_state = FRMR_FLUSHED_LI;
+		frwr->fr_state = FRWR_FLUSHED_LI;
 		__frwr_sendcompletion_flush(wc, "localinv");
 	}
+	trace_xprtrdma_wc_li(wc, frwr);
 }
 
 /**
@@ -329,17 +331,17 @@ frwr_wc_localinv(struct ib_cq *cq, struct ib_wc *wc)
 static void
 frwr_wc_localinv_wake(struct ib_cq *cq, struct ib_wc *wc)
 {
-	struct rpcrdma_frmr *frmr;
-	struct ib_cqe *cqe;
+	struct ib_cqe *cqe = wc->wr_cqe;
+	struct rpcrdma_frwr *frwr = container_of(cqe, struct rpcrdma_frwr,
+						 fr_cqe);
 
 	/* WARNING: Only wr_cqe and status are reliable at this point */
-	cqe = wc->wr_cqe;
-	frmr = container_of(cqe, struct rpcrdma_frmr, fr_cqe);
 	if (wc->status != IB_WC_SUCCESS) {
-		frmr->fr_state = FRMR_FLUSHED_LI;
+		frwr->fr_state = FRWR_FLUSHED_LI;
 		__frwr_sendcompletion_flush(wc, "localinv");
 	}
-	complete(&frmr->fr_linv_done);
+	complete(&frwr->fr_linv_done);
+	trace_xprtrdma_wc_li_wake(wc, frwr);
 }
 
 /* Post a REG_MR Work Request to register a memory region
@@ -347,41 +349,39 @@ frwr_wc_localinv_wake(struct ib_cq *cq, struct ib_wc *wc)
  */
 static struct rpcrdma_mr_seg *
 frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
-	    int nsegs, bool writing, struct rpcrdma_mw **out)
+	    int nsegs, bool writing, struct rpcrdma_mr **out)
 {
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	bool holes_ok = ia->ri_mrtype == IB_MR_TYPE_SG_GAPS;
-	struct rpcrdma_mw *mw;
-	struct rpcrdma_frmr *frmr;
-	struct ib_mr *mr;
+	struct rpcrdma_frwr *frwr;
+	struct rpcrdma_mr *mr;
+	struct ib_mr *ibmr;
 	struct ib_reg_wr *reg_wr;
 	struct ib_send_wr *bad_wr;
 	int rc, i, n;
 	u8 key;
 
-	mw = NULL;
+	mr = NULL;
 	do {
-		if (mw)
-			rpcrdma_defer_mr_recovery(mw);
-		mw = rpcrdma_get_mw(r_xprt);
-		if (!mw)
+		if (mr)
+			rpcrdma_mr_defer_recovery(mr);
+		mr = rpcrdma_mr_get(r_xprt);
+		if (!mr)
 			return ERR_PTR(-ENOBUFS);
-	} while (mw->frmr.fr_state != FRMR_IS_INVALID);
-	frmr = &mw->frmr;
-	frmr->fr_state = FRMR_IS_VALID;
-	mr = frmr->fr_mr;
-	reg_wr = &frmr->fr_regwr;
-
-	if (nsegs > ia->ri_max_frmr_depth)
-		nsegs = ia->ri_max_frmr_depth;
+	} while (mr->frwr.fr_state != FRWR_IS_INVALID);
+	frwr = &mr->frwr;
+	frwr->fr_state = FRWR_IS_VALID;
+
+	if (nsegs > ia->ri_max_frwr_depth)
+		nsegs = ia->ri_max_frwr_depth;
 	for (i = 0; i < nsegs;) {
 		if (seg->mr_page)
-			sg_set_page(&mw->mw_sg[i],
+			sg_set_page(&mr->mr_sg[i],
 				    seg->mr_page,
 				    seg->mr_len,
 				    offset_in_page(seg->mr_offset));
 		else
-			sg_set_buf(&mw->mw_sg[i], seg->mr_offset,
+			sg_set_buf(&mr->mr_sg[i], seg->mr_offset,
 				   seg->mr_len);
 
 		++seg;
@@ -392,30 +392,29 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		    offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len))
 			break;
 	}
-	mw->mw_dir = rpcrdma_data_dir(writing);
+	mr->mr_dir = rpcrdma_data_dir(writing);
 
-	mw->mw_nents = ib_dma_map_sg(ia->ri_device, mw->mw_sg, i, mw->mw_dir);
-	if (!mw->mw_nents)
+	mr->mr_nents = ib_dma_map_sg(ia->ri_device, mr->mr_sg, i, mr->mr_dir);
+	if (!mr->mr_nents)
 		goto out_dmamap_err;
 
-	n = ib_map_mr_sg(mr, mw->mw_sg, mw->mw_nents, NULL, PAGE_SIZE);
-	if (unlikely(n != mw->mw_nents))
+	ibmr = frwr->fr_mr;
+	n = ib_map_mr_sg(ibmr, mr->mr_sg, mr->mr_nents, NULL, PAGE_SIZE);
+	if (unlikely(n != mr->mr_nents))
 		goto out_mapmr_err;
 
-	dprintk("RPC:       %s: Using frmr %p to map %u segments (%llu bytes)\n",
-		__func__, frmr, mw->mw_nents, mr->length);
-
-	key = (u8)(mr->rkey & 0x000000FF);
-	ib_update_fast_reg_key(mr, ++key);
+	key = (u8)(ibmr->rkey & 0x000000FF);
+	ib_update_fast_reg_key(ibmr, ++key);
 
+	reg_wr = &frwr->fr_regwr;
 	reg_wr->wr.next = NULL;
 	reg_wr->wr.opcode = IB_WR_REG_MR;
-	frmr->fr_cqe.done = frwr_wc_fastreg;
-	reg_wr->wr.wr_cqe = &frmr->fr_cqe;
+	frwr->fr_cqe.done = frwr_wc_fastreg;
+	reg_wr->wr.wr_cqe = &frwr->fr_cqe;
 	reg_wr->wr.num_sge = 0;
 	reg_wr->wr.send_flags = 0;
-	reg_wr->mr = mr;
-	reg_wr->key = mr->rkey;
+	reg_wr->mr = ibmr;
+	reg_wr->key = ibmr->rkey;
 	reg_wr->access = writing ?
 			 IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
 			 IB_ACCESS_REMOTE_READ;
@@ -424,47 +423,64 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	if (rc)
 		goto out_senderr;
 
-	mw->mw_handle = mr->rkey;
-	mw->mw_length = mr->length;
-	mw->mw_offset = mr->iova;
+	mr->mr_handle = ibmr->rkey;
+	mr->mr_length = ibmr->length;
+	mr->mr_offset = ibmr->iova;
 
-	*out = mw;
+	*out = mr;
 	return seg;
 
 out_dmamap_err:
 	pr_err("rpcrdma: failed to DMA map sg %p sg_nents %d\n",
-	       mw->mw_sg, i);
-	frmr->fr_state = FRMR_IS_INVALID;
-	rpcrdma_put_mw(r_xprt, mw);
+	       mr->mr_sg, i);
+	frwr->fr_state = FRWR_IS_INVALID;
+	rpcrdma_mr_put(mr);
 	return ERR_PTR(-EIO);
 
 out_mapmr_err:
 	pr_err("rpcrdma: failed to map mr %p (%d/%d)\n",
-	       frmr->fr_mr, n, mw->mw_nents);
-	rpcrdma_defer_mr_recovery(mw);
+	       frwr->fr_mr, n, mr->mr_nents);
+	rpcrdma_mr_defer_recovery(mr);
 	return ERR_PTR(-EIO);
 
 out_senderr:
-	pr_err("rpcrdma: FRMR registration ib_post_send returned %i\n", rc);
-	rpcrdma_defer_mr_recovery(mw);
+	pr_err("rpcrdma: FRWR registration ib_post_send returned %i\n", rc);
+	rpcrdma_mr_defer_recovery(mr);
 	return ERR_PTR(-ENOTCONN);
 }
 
+/* Handle a remotely invalidated mr on the @mrs list
+ */
+static void
+frwr_op_reminv(struct rpcrdma_rep *rep, struct list_head *mrs)
+{
+	struct rpcrdma_mr *mr;
+
+	list_for_each_entry(mr, mrs, mr_list)
+		if (mr->mr_handle == rep->rr_inv_rkey) {
+			list_del(&mr->mr_list);
+			trace_xprtrdma_remoteinv(mr);
+			mr->frwr.fr_state = FRWR_IS_INVALID;
+			rpcrdma_mr_unmap_and_put(mr);
+			break;	/* only one invalidated MR per RPC */
+		}
+}
+
 /* Invalidate all memory regions that were registered for "req".
  *
  * Sleeps until it is safe for the host CPU to access the
  * previously mapped memory regions.
  *
- * Caller ensures that @mws is not empty before the call. This
+ * Caller ensures that @mrs is not empty before the call. This
  * function empties the list.
  */
 static void
-frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
+frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mrs)
 {
 	struct ib_send_wr *first, **prev, *last, *bad_wr;
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
-	struct rpcrdma_frmr *f;
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_frwr *frwr;
+	struct rpcrdma_mr *mr;
 	int count, rc;
 
 	/* ORDER: Invalidate all of the MRs first
@@ -472,31 +488,27 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	 * Chain the LOCAL_INV Work Requests and post them with
 	 * a single ib_post_send() call.
 	 */
-	f = NULL;
+	frwr = NULL;
 	count = 0;
 	prev = &first;
-	list_for_each_entry(mw, mws, mw_list) {
-		mw->frmr.fr_state = FRMR_IS_INVALID;
+	list_for_each_entry(mr, mrs, mr_list) {
+		mr->frwr.fr_state = FRWR_IS_INVALID;
 
-		if (mw->mw_flags & RPCRDMA_MW_F_RI)
-			continue;
+		frwr = &mr->frwr;
+		trace_xprtrdma_localinv(mr);
 
-		f = &mw->frmr;
-		dprintk("RPC:       %s: invalidating frmr %p\n",
-			__func__, f);
-
-		f->fr_cqe.done = frwr_wc_localinv;
-		last = &f->fr_invwr;
+		frwr->fr_cqe.done = frwr_wc_localinv;
+		last = &frwr->fr_invwr;
 		memset(last, 0, sizeof(*last));
-		last->wr_cqe = &f->fr_cqe;
+		last->wr_cqe = &frwr->fr_cqe;
 		last->opcode = IB_WR_LOCAL_INV;
-		last->ex.invalidate_rkey = mw->mw_handle;
+		last->ex.invalidate_rkey = mr->mr_handle;
 		count++;
 
 		*prev = last;
 		prev = &last->next;
 	}
-	if (!f)
+	if (!frwr)
 		goto unmap;
 
 	/* Strong send queue ordering guarantees that when the
@@ -504,8 +516,8 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	 * are complete.
 	 */
 	last->send_flags = IB_SEND_SIGNALED;
-	f->fr_cqe.done = frwr_wc_localinv_wake;
-	reinit_completion(&f->fr_linv_done);
+	frwr->fr_cqe.done = frwr_wc_localinv_wake;
+	reinit_completion(&frwr->fr_linv_done);
 
 	/* Transport disconnect drains the receive CQ before it
 	 * replaces the QP. The RPC reply handler won't call us
@@ -515,36 +527,32 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 	bad_wr = NULL;
 	rc = ib_post_send(ia->ri_id->qp, first, &bad_wr);
 	if (bad_wr != first)
-		wait_for_completion(&f->fr_linv_done);
+		wait_for_completion(&frwr->fr_linv_done);
 	if (rc)
 		goto reset_mrs;
 
 	/* ORDER: Now DMA unmap all of the MRs, and return
-	 * them to the free MW list.
+	 * them to the free MR list.
 	 */
 unmap:
-	while (!list_empty(mws)) {
-		mw = rpcrdma_pop_mw(mws);
-		dprintk("RPC:       %s: DMA unmapping frmr %p\n",
-			__func__, &mw->frmr);
-		ib_dma_unmap_sg(ia->ri_device,
-				mw->mw_sg, mw->mw_nents, mw->mw_dir);
-		rpcrdma_put_mw(r_xprt, mw);
+	while (!list_empty(mrs)) {
+		mr = rpcrdma_mr_pop(mrs);
+		rpcrdma_mr_unmap_and_put(mr);
 	}
 	return;
 
 reset_mrs:
-	pr_err("rpcrdma: FRMR invalidate ib_post_send returned %i\n", rc);
+	pr_err("rpcrdma: FRWR invalidate ib_post_send returned %i\n", rc);
 
 	/* Find and reset the MRs in the LOCAL_INV WRs that did not
 	 * get posted.
 	 */
 	while (bad_wr) {
-		f = container_of(bad_wr, struct rpcrdma_frmr,
-				 fr_invwr);
-		mw = container_of(f, struct rpcrdma_mw, frmr);
+		frwr = container_of(bad_wr, struct rpcrdma_frwr,
+				    fr_invwr);
+		mr = container_of(frwr, struct rpcrdma_mr, frwr);
 
-		__frwr_reset_mr(ia, mw);
+		__frwr_mr_reset(ia, mr);
 
 		bad_wr = bad_wr->next;
 	}
@@ -553,6 +561,7 @@ frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct list_head *mws)
 
 const struct rpcrdma_memreg_ops rpcrdma_frwr_memreg_ops = {
 	.ro_map				= frwr_op_map,
+	.ro_reminv			= frwr_op_reminv,
 	.ro_unmap_sync			= frwr_op_unmap_sync,
 	.ro_recover_mr			= frwr_op_recover_mr,
 	.ro_open			= frwr_op_open,

net/sunrpc/xprtrdma/module.c

 /*
- * Copyright (c) 2015 Oracle.  All rights reserved.
+ * Copyright (c) 2015, 2017 Oracle.  All rights reserved.
  */
 
 /* rpcrdma.ko module initialization
  */
 
+#include <linux/types.h>
+#include <linux/compiler.h>
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/sunrpc/svc_rdma.h>
-#include "xprt_rdma.h"
 
-#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
-# define RPCDBG_FACILITY	RPCDBG_TRANS
-#endif
+#include <asm/swab.h>
+
+#define CREATE_TRACE_POINTS
+#include "xprt_rdma.h"
 
 MODULE_AUTHOR("Open Grid Computing and Network Appliance, Inc.");
 MODULE_DESCRIPTION("RPC/RDMA Transport");

net/sunrpc/xprtrdma/rpc_rdma.c

@@ -292,15 +292,15 @@ encode_item_not_present(struct xdr_stream *xdr)
 }
 
 static void
-xdr_encode_rdma_segment(__be32 *iptr, struct rpcrdma_mw *mw)
+xdr_encode_rdma_segment(__be32 *iptr, struct rpcrdma_mr *mr)
 {
-	*iptr++ = cpu_to_be32(mw->mw_handle);
-	*iptr++ = cpu_to_be32(mw->mw_length);
-	xdr_encode_hyper(iptr, mw->mw_offset);
+	*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_mw *mw)
+encode_rdma_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr)
 {
 	__be32 *p;
 
@@ -308,12 +308,12 @@ encode_rdma_segment(struct xdr_stream *xdr, struct rpcrdma_mw *mw)
 	if (unlikely(!p))
 		return -EMSGSIZE;
 
-	xdr_encode_rdma_segment(p, mw);
+	xdr_encode_rdma_segment(p, mr);
 	return 0;
 }
 
 static int
-encode_read_segment(struct xdr_stream *xdr, struct rpcrdma_mw *mw,
+encode_read_segment(struct xdr_stream *xdr, struct rpcrdma_mr *mr,
 		    u32 position)
 {
 	__be32 *p;
@@ -324,7 +324,7 @@ encode_read_segment(struct xdr_stream *xdr, struct rpcrdma_mw *mw,
 
 	*p++ = xdr_one;			/* Item present */
 	*p++ = cpu_to_be32(position);
-	xdr_encode_rdma_segment(p, mw);
+	xdr_encode_rdma_segment(p, mr);
 	return 0;
 }
 
@@ -348,7 +348,7 @@ rpcrdma_encode_read_list(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 {
 	struct xdr_stream *xdr = &req->rl_stream;
 	struct rpcrdma_mr_seg *seg;
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 	unsigned int pos;
 	int nsegs;
 
@@ -363,21 +363,17 @@ rpcrdma_encode_read_list(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 
 	do {
 		seg = r_xprt->rx_ia.ri_ops->ro_map(r_xprt, seg, nsegs,
-						   false, &mw);
+						   false, &mr);
 		if (IS_ERR(seg))
 			return PTR_ERR(seg);
-		rpcrdma_push_mw(mw, &req->rl_registered);
+		rpcrdma_mr_push(mr, &req->rl_registered);
 
-		if (encode_read_segment(xdr, mw, pos) < 0)
+		if (encode_read_segment(xdr, mr, pos) < 0)
 			return -EMSGSIZE;
 
-		dprintk("RPC: %5u %s: pos %u %u@0x%016llx:0x%08x (%s)\n",
-			rqst->rq_task->tk_pid, __func__, pos,
-			mw->mw_length, (unsigned long long)mw->mw_offset,
-			mw->mw_handle, mw->mw_nents < nsegs ? "more" : "last");
-
+		trace_xprtrdma_read_chunk(rqst->rq_task, pos, mr, nsegs);
 		r_xprt->rx_stats.read_chunk_count++;
-		nsegs -= mw->mw_nents;
+		nsegs -= mr->mr_nents;
 	} while (nsegs);
 
 	return 0;
@@ -404,7 +400,7 @@ rpcrdma_encode_write_list(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 {
 	struct xdr_stream *xdr = &req->rl_stream;
 	struct rpcrdma_mr_seg *seg;
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 	int nsegs, nchunks;
 	__be32 *segcount;
 
@@ -425,23 +421,19 @@ rpcrdma_encode_write_list(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 	nchunks = 0;
 	do {
 		seg = r_xprt->rx_ia.ri_ops->ro_map(r_xprt, seg, nsegs,
-						   true, &mw);
+						   true, &mr);
 		if (IS_ERR(seg))
 			return PTR_ERR(seg);
-		rpcrdma_push_mw(mw, &req->rl_registered);
+		rpcrdma_mr_push(mr, &req->rl_registered);
 
-		if (encode_rdma_segment(xdr, mw) < 0)
+		if (encode_rdma_segment(xdr, mr) < 0)
 			return -EMSGSIZE;
 
-		dprintk("RPC: %5u %s: %u@0x016%llx:0x%08x (%s)\n",
-			rqst->rq_task->tk_pid, __func__,
-			mw->mw_length, (unsigned long long)mw->mw_offset,
-			mw->mw_handle, mw->mw_nents < nsegs ? "more" : "last");
-
+		trace_xprtrdma_write_chunk(rqst->rq_task, mr, nsegs);
 		r_xprt->rx_stats.write_chunk_count++;
-		r_xprt->rx_stats.total_rdma_request += seg->mr_len;
+		r_xprt->rx_stats.total_rdma_request += mr->mr_length;
 		nchunks++;
-		nsegs -= mw->mw_nents;
+		nsegs -= mr->mr_nents;
 	} while (nsegs);
 
 	/* Update count of segments in this Write chunk */
@@ -468,7 +460,7 @@ rpcrdma_encode_reply_chunk(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 {
 	struct xdr_stream *xdr = &req->rl_stream;
 	struct rpcrdma_mr_seg *seg;
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 	int nsegs, nchunks;
 	__be32 *segcount;
 
@@ -487,23 +479,19 @@ rpcrdma_encode_reply_chunk(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req,
 	nchunks = 0;
 	do {
 		seg = r_xprt->rx_ia.ri_ops->ro_map(r_xprt, seg, nsegs,
-						   true, &mw);
+						   true, &mr);
 		if (IS_ERR(seg))
 			return PTR_ERR(seg);
-		rpcrdma_push_mw(mw, &req->rl_registered);
+		rpcrdma_mr_push(mr, &req->rl_registered);
 
-		if (encode_rdma_segment(xdr, mw) < 0)
+		if (encode_rdma_segment(xdr, mr) < 0)
 			return -EMSGSIZE;
 
-		dprintk("RPC: %5u %s: %u@0x%016llx:0x%08x (%s)\n",
-			rqst->rq_task->tk_pid, __func__,
-			mw->mw_length, (unsigned long long)mw->mw_offset,
-			mw->mw_handle, mw->mw_nents < nsegs ? "more" : "last");
-
+		trace_xprtrdma_reply_chunk(rqst->rq_task, mr, nsegs);
 		r_xprt->rx_stats.reply_chunk_count++;
-		r_xprt->rx_stats.total_rdma_request += seg->mr_len;
+		r_xprt->rx_stats.total_rdma_request += mr->mr_length;
 		nchunks++;
-		nsegs -= mw->mw_nents;
+		nsegs -= mr->mr_nents;
 	} while (nsegs);
 
 	/* Update count of segments in the Reply chunk */
@@ -524,9 +512,6 @@ rpcrdma_unmap_sendctx(struct rpcrdma_sendctx *sc)
 	struct ib_sge *sge;
 	unsigned int count;
 
-	dprintk("RPC:       %s: unmapping %u sges for sc=%p\n",
-		__func__, sc->sc_unmap_count, sc);
-
 	/* The first two SGEs contain the transport header and
 	 * the inline buffer. These are always left mapped so
 	 * they can be cheaply re-used.
@@ -754,11 +739,6 @@ rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst)
 	__be32 *p;
 	int ret;
 
-#if defined(CONFIG_SUNRPC_BACKCHANNEL)
-	if (test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state))
-		return rpcrdma_bc_marshal_reply(rqst);
-#endif
-
 	rpcrdma_set_xdrlen(&req->rl_hdrbuf, 0);
 	xdr_init_encode(xdr, &req->rl_hdrbuf,
 			req->rl_rdmabuf->rg_base);
@@ -821,6 +801,17 @@ rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst)
 		rtype = rpcrdma_areadch;
 	}
 
+	/* If this is a retransmit, discard previously registered
+	 * chunks. Very likely the connection has been replaced,
+	 * so these registrations are invalid and unusable.
+	 */
+	while (unlikely(!list_empty(&req->rl_registered))) {
+		struct rpcrdma_mr *mr;
+
+		mr = rpcrdma_mr_pop(&req->rl_registered);
+		rpcrdma_mr_defer_recovery(mr);
+	}
+
 	/* This implementation supports the following combinations
 	 * of chunk lists in one RPC-over-RDMA Call message:
 	 *
@@ -868,10 +859,7 @@ rpcrdma_marshal_req(struct rpcrdma_xprt *r_xprt, struct rpc_rqst *rqst)
 	if (ret)
 		goto out_err;
 
-	dprintk("RPC: %5u %s: %s/%s: hdrlen %u rpclen\n",
-		rqst->rq_task->tk_pid, __func__,
-		transfertypes[rtype], transfertypes[wtype],
-		xdr_stream_pos(xdr));
+	trace_xprtrdma_marshal(rqst, xdr_stream_pos(xdr), rtype, wtype);
 
 	ret = rpcrdma_prepare_send_sges(r_xprt, req, xdr_stream_pos(xdr),
 					&rqst->rq_snd_buf, rtype);
@@ -926,8 +914,7 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
 	curlen = rqst->rq_rcv_buf.head[0].iov_len;
 	if (curlen > copy_len)
 		curlen = copy_len;
-	dprintk("RPC:       %s: srcp 0x%p len %d hdrlen %d\n",
-		__func__, srcp, copy_len, curlen);
+	trace_xprtrdma_fixup(rqst, copy_len, curlen);
 	srcp += curlen;
 	copy_len -= curlen;
 
@@ -947,9 +934,8 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
 			if (curlen > pagelist_len)
 				curlen = pagelist_len;
 
-			dprintk("RPC:       %s: page %d"
-				" srcp 0x%p len %d curlen %d\n",
-				__func__, i, srcp, copy_len, curlen);
+			trace_xprtrdma_fixup_pg(rqst, i, srcp,
+						copy_len, curlen);
 			destp = kmap_atomic(ppages[i]);
 			memcpy(destp + page_base, srcp, curlen);
 			flush_dcache_page(ppages[i]);
@@ -984,24 +970,6 @@ rpcrdma_inline_fixup(struct rpc_rqst *rqst, char *srcp, int copy_len, int pad)
 	return fixup_copy_count;
 }
 
-/* Caller must guarantee @rep remains stable during this call.
- */
-static void
-rpcrdma_mark_remote_invalidation(struct list_head *mws,
-				 struct rpcrdma_rep *rep)
-{
-	struct rpcrdma_mw *mw;
-
-	if (!(rep->rr_wc_flags & IB_WC_WITH_INVALIDATE))
-		return;
-
-	list_for_each_entry(mw, mws, mw_list)
-		if (mw->mw_handle == rep->rr_inv_rkey) {
-			mw->mw_flags = RPCRDMA_MW_F_RI;
-			break; /* only one invalidated MR per RPC */
-		}
-}
-
 /* By convention, backchannel calls arrive via rdma_msg type
  * messages, and never populate the chunk lists. This makes
  * the RPC/RDMA header small and fixed in size, so it is
@@ -1058,26 +1026,19 @@ rpcrdma_is_bcall(struct rpcrdma_xprt *r_xprt, struct rpcrdma_rep *rep)
 
 static int decode_rdma_segment(struct xdr_stream *xdr, u32 *length)
 {
+	u32 handle;
+	u64 offset;
 	__be32 *p;
 
 	p = xdr_inline_decode(xdr, 4 * sizeof(*p));
 	if (unlikely(!p))
 		return -EIO;
 
-	ifdebug(FACILITY) {
-		u64 offset;
-		u32 handle;
-
-		handle = be32_to_cpup(p++);
-		*length = be32_to_cpup(p++);
-		xdr_decode_hyper(p, &offset);
-		dprintk("RPC:       %s:   segment %u@0x%016llx:0x%08x\n",
-			__func__, *length, (unsigned long long)offset,
-			handle);
-	} else {
-		*length = be32_to_cpup(p + 1);
-	}
+	handle = be32_to_cpup(p++);
+	*length = be32_to_cpup(p++);
+	xdr_decode_hyper(p, &offset);
 
+	trace_xprtrdma_decode_seg(handle, *length, offset);
 	return 0;
 }
 
@@ -1098,8 +1059,6 @@ static int decode_write_chunk(struct xdr_stream *xdr, u32 *length)
 		*length += seglength;
 	}
 
-	dprintk("RPC:       %s: segcount=%u, %u bytes\n",
-		__func__, be32_to_cpup(p), *length);
 	return 0;
 }
 
@@ -1296,8 +1255,7 @@ void rpcrdma_complete_rqst(struct rpcrdma_rep *rep)
  * being marshaled.
  */
 out_badheader:
-	dprintk("RPC: %5u %s: invalid rpcrdma reply (type %u)\n",
-		rqst->rq_task->tk_pid, __func__, be32_to_cpu(rep->rr_proc));
+	trace_xprtrdma_reply_hdr(rep);
 	r_xprt->rx_stats.bad_reply_count++;
 	status = -EIO;
 	goto out;
@@ -1339,9 +1297,12 @@ void rpcrdma_deferred_completion(struct work_struct *work)
 	struct rpcrdma_rep *rep =
 			container_of(work, struct rpcrdma_rep, rr_work);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rep->rr_rqst);
+	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
 
-	rpcrdma_mark_remote_invalidation(&req->rl_registered, rep);
-	rpcrdma_release_rqst(rep->rr_rxprt, req);
+	trace_xprtrdma_defer_cmp(rep);
+	if (rep->rr_wc_flags & IB_WC_WITH_INVALIDATE)
+		r_xprt->rx_ia.ri_ops->ro_reminv(rep, &req->rl_registered);
+	rpcrdma_release_rqst(r_xprt, req);
 	rpcrdma_complete_rqst(rep);
 }
 
@@ -1360,8 +1321,6 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 	u32 credits;
 	__be32 *p;
 
-	dprintk("RPC:       %s: incoming rep %p\n", __func__, rep);
-
 	if (rep->rr_hdrbuf.head[0].iov_len == 0)
 		goto out_badstatus;
 
@@ -1405,8 +1364,7 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 	rep->rr_rqst = rqst;
 	clear_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags);
 
-	dprintk("RPC:       %s: reply %p completes request %p (xid 0x%08x)\n",
-		__func__, rep, req, be32_to_cpu(rep->rr_xid));
+	trace_xprtrdma_reply(rqst->rq_task, rep, req, credits);
 
 	queue_work_on(req->rl_cpu, rpcrdma_receive_wq, &rep->rr_work);
 	return;
@@ -1420,8 +1378,7 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 	return;
 
 out_badversion:
-	dprintk("RPC:       %s: invalid version %d\n",
-		__func__, be32_to_cpu(rep->rr_vers));
+	trace_xprtrdma_reply_vers(rep);
 	goto repost;
 
 /* The RPC transaction has already been terminated, or the header
@@ -1429,12 +1386,11 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
  */
 out_norqst:
 	spin_unlock(&xprt->recv_lock);
-	dprintk("RPC:       %s: no match for incoming xid 0x%08x\n",
-		__func__, be32_to_cpu(rep->rr_xid));
+	trace_xprtrdma_reply_rqst(rep);
 	goto repost;
 
 out_shortreply:
-	dprintk("RPC:       %s: short/invalid reply\n", __func__);
+	trace_xprtrdma_reply_short(rep);
 
 /* If no pending RPC transaction was matched, post a replacement
  * receive buffer before returning.

net/sunrpc/xprtrdma/transport.c

@@ -67,8 +67,7 @@
 static unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
 unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE;
 static unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE;
-static unsigned int xprt_rdma_inline_write_padding;
-unsigned int xprt_rdma_memreg_strategy		= RPCRDMA_FRMR;
+unsigned int xprt_rdma_memreg_strategy		= RPCRDMA_FRWR;
 int xprt_rdma_pad_optimize;
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
@@ -81,6 +80,7 @@ static unsigned int zero;
 static unsigned int max_padding = PAGE_SIZE;
 static unsigned int min_memreg = RPCRDMA_BOUNCEBUFFERS;
 static unsigned int max_memreg = RPCRDMA_LAST - 1;
+static unsigned int dummy;
 
 static struct ctl_table_header *sunrpc_table_header;
 
@@ -114,7 +114,7 @@ static struct ctl_table xr_tunables_table[] = {
 	},
 	{
 		.procname	= "rdma_inline_write_padding",
-		.data		= &xprt_rdma_inline_write_padding,
+		.data		= &dummy,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec_minmax,
@@ -259,13 +259,10 @@ xprt_rdma_connect_worker(struct work_struct *work)
 
 	xprt_clear_connected(xprt);
 
-	dprintk("RPC:       %s: %sconnect\n", __func__,
-			r_xprt->rx_ep.rep_connected != 0 ? "re" : "");
 	rc = rpcrdma_ep_connect(&r_xprt->rx_ep, &r_xprt->rx_ia);
 	if (rc)
 		xprt_wake_pending_tasks(xprt, rc);
 
-	dprintk("RPC:       %s: exit\n", __func__);
 	xprt_clear_connecting(xprt);
 }
 
@@ -275,7 +272,7 @@ xprt_rdma_inject_disconnect(struct rpc_xprt *xprt)
 	struct rpcrdma_xprt *r_xprt = container_of(xprt, struct rpcrdma_xprt,
 						   rx_xprt);
 
-	pr_info("rpcrdma: injecting transport disconnect on xprt=%p\n", xprt);
+	trace_xprtrdma_inject_dsc(r_xprt);
 	rdma_disconnect(r_xprt->rx_ia.ri_id);
 }
 
@@ -295,7 +292,7 @@ xprt_rdma_destroy(struct rpc_xprt *xprt)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 
-	dprintk("RPC:       %s: called\n", __func__);
+	trace_xprtrdma_destroy(r_xprt);
 
 	cancel_delayed_work_sync(&r_xprt->rx_connect_worker);
 
@@ -306,11 +303,8 @@ xprt_rdma_destroy(struct rpc_xprt *xprt)
 	rpcrdma_ia_close(&r_xprt->rx_ia);
 
 	xprt_rdma_free_addresses(xprt);
-
 	xprt_free(xprt);
 
-	dprintk("RPC:       %s: returning\n", __func__);
-
 	module_put(THIS_MODULE);
 }
 
@@ -361,9 +355,7 @@ xprt_setup_rdma(struct xprt_create *args)
 	/*
 	 * Set up RDMA-specific connect data.
 	 */
-
-	sap = (struct sockaddr *)&cdata.addr;
-	memcpy(sap, args->dstaddr, args->addrlen);
+	sap = args->dstaddr;
 
 	/* Ensure xprt->addr holds valid server TCP (not RDMA)
 	 * address, for any side protocols which peek at it */
@@ -373,6 +365,7 @@ xprt_setup_rdma(struct xprt_create *args)
 
 	if (rpc_get_port(sap))
 		xprt_set_bound(xprt);
+	xprt_rdma_format_addresses(xprt, sap);
 
 	cdata.max_requests = xprt->max_reqs;
 
@@ -387,8 +380,6 @@ xprt_setup_rdma(struct xprt_create *args)
 	if (cdata.inline_rsize > cdata.rsize)
 		cdata.inline_rsize = cdata.rsize;
 
-	cdata.padding = xprt_rdma_inline_write_padding;
-
 	/*
 	 * Create new transport instance, which includes initialized
 	 *  o ia
@@ -398,7 +389,7 @@ xprt_setup_rdma(struct xprt_create *args)
 
 	new_xprt = rpcx_to_rdmax(xprt);
 
-	rc = rpcrdma_ia_open(new_xprt, sap);
+	rc = rpcrdma_ia_open(new_xprt);
 	if (rc)
 		goto out1;
 
@@ -407,31 +398,19 @@ xprt_setup_rdma(struct xprt_create *args)
 	 */
 	new_xprt->rx_data = cdata;
 	new_ep = &new_xprt->rx_ep;
-	new_ep->rep_remote_addr = cdata.addr;
 
 	rc = rpcrdma_ep_create(&new_xprt->rx_ep,
 				&new_xprt->rx_ia, &new_xprt->rx_data);
 	if (rc)
 		goto out2;
 
-	/*
-	 * Allocate pre-registered send and receive buffers for headers and
-	 * any inline data. Also specify any padding which will be provided
-	 * from a preregistered zero buffer.
-	 */
 	rc = rpcrdma_buffer_create(new_xprt);
 	if (rc)
 		goto out3;
 
-	/*
-	 * Register a callback for connection events. This is necessary because
-	 * connection loss notification is async. We also catch connection loss
-	 * when reaping receives.
-	 */
 	INIT_DELAYED_WORK(&new_xprt->rx_connect_worker,
 			  xprt_rdma_connect_worker);
 
-	xprt_rdma_format_addresses(xprt, sap);
 	xprt->max_payload = new_xprt->rx_ia.ri_ops->ro_maxpages(new_xprt);
 	if (xprt->max_payload == 0)
 		goto out4;
@@ -445,16 +424,19 @@ xprt_setup_rdma(struct xprt_create *args)
 	dprintk("RPC:       %s: %s:%s\n", __func__,
 		xprt->address_strings[RPC_DISPLAY_ADDR],
 		xprt->address_strings[RPC_DISPLAY_PORT]);
+	trace_xprtrdma_create(new_xprt);
 	return xprt;
 
 out4:
-	xprt_rdma_free_addresses(xprt);
-	rc = -EINVAL;
+	rpcrdma_buffer_destroy(&new_xprt->rx_buf);
+	rc = -ENODEV;
 out3:
 	rpcrdma_ep_destroy(new_ep, &new_xprt->rx_ia);
 out2:
 	rpcrdma_ia_close(&new_xprt->rx_ia);
 out1:
+	trace_xprtrdma_destroy(new_xprt);
+	xprt_rdma_free_addresses(xprt);
 	xprt_free(xprt);
 	return ERR_PTR(rc);
 }
@@ -488,16 +470,34 @@ xprt_rdma_close(struct rpc_xprt *xprt)
 	rpcrdma_ep_disconnect(ep, ia);
 }
 
+/**
+ * xprt_rdma_set_port - update server port with rpcbind result
+ * @xprt: controlling RPC transport
+ * @port: new port value
+ *
+ * Transport connect status is unchanged.
+ */
 static void
 xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 {
-	struct sockaddr_in *sap;
+	struct sockaddr *sap = (struct sockaddr *)&xprt->addr;
+	char buf[8];
 
-	sap = (struct sockaddr_in *)&xprt->addr;
-	sap->sin_port = htons(port);
-	sap = (struct sockaddr_in *)&rpcx_to_rdmad(xprt).addr;
-	sap->sin_port = htons(port);
-	dprintk("RPC:       %s: %u\n", __func__, port);
+	dprintk("RPC:       %s: setting port for xprt %p (%s:%s) to %u\n",
+		__func__, xprt,
+		xprt->address_strings[RPC_DISPLAY_ADDR],
+		xprt->address_strings[RPC_DISPLAY_PORT],
+		port);
+
+	rpc_set_port(sap, port);
+
+	kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
+	snprintf(buf, sizeof(buf), "%u", port);
+	xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
+
+	kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
+	snprintf(buf, sizeof(buf), "%4hx", port);
+	xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
 }
 
 /**
@@ -516,8 +516,6 @@ xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 static void
 xprt_rdma_timer(struct rpc_xprt *xprt, struct rpc_task *task)
 {
-	dprintk("RPC: %5u %s: xprt = %p\n", task->tk_pid, __func__, xprt);
-
 	xprt_force_disconnect(xprt);
 }
 
@@ -640,7 +638,7 @@ xprt_rdma_allocate(struct rpc_task *task)
 
 	req = rpcrdma_buffer_get(&r_xprt->rx_buf);
 	if (req == NULL)
-		return -ENOMEM;
+		goto out_get;
 
 	flags = RPCRDMA_DEF_GFP;
 	if (RPC_IS_SWAPPER(task))
@@ -653,19 +651,18 @@ xprt_rdma_allocate(struct rpc_task *task)
 	if (!rpcrdma_get_recvbuf(r_xprt, req, rqst->rq_rcvsize, flags))
 		goto out_fail;
 
-	dprintk("RPC: %5u %s: send size = %zd, recv size = %zd, req = %p\n",
-		task->tk_pid, __func__, rqst->rq_callsize,
-		rqst->rq_rcvsize, req);
-
 	req->rl_cpu = smp_processor_id();
 	req->rl_connect_cookie = 0;	/* our reserved value */
 	rpcrdma_set_xprtdata(rqst, req);
 	rqst->rq_buffer = req->rl_sendbuf->rg_base;
 	rqst->rq_rbuffer = req->rl_recvbuf->rg_base;
+	trace_xprtrdma_allocate(task, req);
 	return 0;
 
 out_fail:
 	rpcrdma_buffer_put(req);
+out_get:
+	trace_xprtrdma_allocate(task, NULL);
 	return -ENOMEM;
 }
 
@@ -682,13 +679,9 @@ xprt_rdma_free(struct rpc_task *task)
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
 
-	if (test_bit(RPCRDMA_REQ_F_BACKCHANNEL, &req->rl_flags))
-		return;
-
-	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);
-
 	if (test_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags))
 		rpcrdma_release_rqst(r_xprt, req);
+	trace_xprtrdma_rpc_done(task, req);
 	rpcrdma_buffer_put(req);
 }
 
@@ -698,22 +691,12 @@ xprt_rdma_free(struct rpc_task *task)
  *
  * Caller holds the transport's write lock.
  *
- * Return values:
- *        0:	The request has been sent
- * ENOTCONN:	Caller needs to invoke connect logic then call again
- *  ENOBUFS:	Call again later to send the request
- *      EIO:	A permanent error occurred. The request was not sent,
- *		and don't try it again
- *
- * send_request invokes the meat of RPC RDMA. It must do the following:
- *
- *  1.  Marshal the RPC request into an RPC RDMA request, which means
- *	putting a header in front of data, and creating IOVs for RDMA
- *	from those in the request.
- *  2.  In marshaling, detect opportunities for RDMA, and use them.
- *  3.  Post a recv message to set up asynch completion, then send
- *	the request (rpcrdma_ep_post).
- *  4.  No partial sends are possible in the RPC-RDMA protocol (as in UDP).
+ * Returns:
+ *	%0 if the RPC message has been sent
+ *	%-ENOTCONN if the caller should reconnect and call again
+ *	%-ENOBUFS if the caller should call again later
+ *	%-EIO if a permanent error occurred and the request was not
+ *		sent. Do not try to send this message again.
  */
 static int
 xprt_rdma_send_request(struct rpc_task *task)
@@ -724,14 +707,14 @@ xprt_rdma_send_request(struct rpc_task *task)
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 	int rc = 0;
 
+#if defined(CONFIG_SUNRPC_BACKCHANNEL)
+	if (unlikely(!rqst->rq_buffer))
+		return xprt_rdma_bc_send_reply(rqst);
+#endif	/* CONFIG_SUNRPC_BACKCHANNEL */
+
 	if (!xprt_connected(xprt))
 		goto drop_connection;
 
-	/* On retransmit, remove any previously registered chunks */
-	if (unlikely(!list_empty(&req->rl_registered)))
-		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt,
-						    &req->rl_registered);
-
 	rc = rpcrdma_marshal_req(r_xprt, rqst);
 	if (rc < 0)
 		goto failed_marshal;
@@ -744,7 +727,7 @@ xprt_rdma_send_request(struct rpc_task *task)
 		goto drop_connection;
 	req->rl_connect_cookie = xprt->connect_cookie;
 
-	set_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags);
+	__set_bit(RPCRDMA_REQ_F_PENDING, &req->rl_flags);
 	if (rpcrdma_ep_post(&r_xprt->rx_ia, &r_xprt->rx_ep, req))
 		goto drop_connection;
 
@@ -904,8 +887,7 @@ int xprt_rdma_init(void)
 		"\tMaxInlineRead %d\n\tMaxInlineWrite %d\n",
 		xprt_rdma_slot_table_entries,
 		xprt_rdma_max_inline_read, xprt_rdma_max_inline_write);
-	dprintk("\tPadding %d\n\tMemreg %d\n",
-		xprt_rdma_inline_write_padding, xprt_rdma_memreg_strategy);
+	dprintk("\tPadding 0\n\tMemreg %d\n", xprt_rdma_memreg_strategy);
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 	if (!sunrpc_table_header)

net/sunrpc/xprtrdma/xprt_rdma.h

@@ -73,11 +73,10 @@ struct rpcrdma_ia {
 	struct completion	ri_remove_done;
 	int			ri_async_rc;
 	unsigned int		ri_max_segs;
-	unsigned int		ri_max_frmr_depth;
+	unsigned int		ri_max_frwr_depth;
 	unsigned int		ri_max_inline_write;
 	unsigned int		ri_max_inline_read;
 	unsigned int		ri_max_send_sges;
-	bool			ri_reminv_expected;
 	bool			ri_implicit_roundup;
 	enum ib_mr_type		ri_mrtype;
 	unsigned long		ri_flags;
@@ -101,7 +100,6 @@ struct rpcrdma_ep {
 	wait_queue_head_t 	rep_connect_wait;
 	struct rpcrdma_connect_private	rep_cm_private;
 	struct rdma_conn_param	rep_remote_cma;
-	struct sockaddr_storage	rep_remote_addr;
 	struct delayed_work	rep_connect_worker;
 };
 
@@ -232,29 +230,29 @@ enum {
 };
 
 /*
- * struct rpcrdma_mw - external memory region metadata
+ * struct rpcrdma_mr - external memory region metadata
  *
  * An external memory region is any buffer or page that is registered
  * on the fly (ie, not pre-registered).
  *
- * Each rpcrdma_buffer has a list of free MWs anchored in rb_mws. During
+ * Each rpcrdma_buffer has a list of free MWs anchored in rb_mrs. During
  * call_allocate, rpcrdma_buffer_get() assigns one to each segment in
  * an rpcrdma_req. Then rpcrdma_register_external() grabs these to keep
  * track of registration metadata while each RPC is pending.
  * rpcrdma_deregister_external() uses this metadata to unmap and
  * release these resources when an RPC is complete.
  */
-enum rpcrdma_frmr_state {
-	FRMR_IS_INVALID,	/* ready to be used */
-	FRMR_IS_VALID,		/* in use */
-	FRMR_FLUSHED_FR,	/* flushed FASTREG WR */
-	FRMR_FLUSHED_LI,	/* flushed LOCALINV WR */
+enum rpcrdma_frwr_state {
+	FRWR_IS_INVALID,	/* ready to be used */
+	FRWR_IS_VALID,		/* in use */
+	FRWR_FLUSHED_FR,	/* flushed FASTREG WR */
+	FRWR_FLUSHED_LI,	/* flushed LOCALINV WR */
 };
 
-struct rpcrdma_frmr {
+struct rpcrdma_frwr {
 	struct ib_mr			*fr_mr;
 	struct ib_cqe			fr_cqe;
-	enum rpcrdma_frmr_state		fr_state;
+	enum rpcrdma_frwr_state		fr_state;
 	struct completion		fr_linv_done;
 	union {
 		struct ib_reg_wr	fr_regwr;
@@ -267,26 +265,20 @@ struct rpcrdma_fmr {
 	u64			*fm_physaddrs;
 };
 
-struct rpcrdma_mw {
-	struct list_head	mw_list;
-	struct scatterlist	*mw_sg;
-	int			mw_nents;
-	enum dma_data_direction	mw_dir;
-	unsigned long		mw_flags;
+struct rpcrdma_mr {
+	struct list_head	mr_list;
+	struct scatterlist	*mr_sg;
+	int			mr_nents;
+	enum dma_data_direction	mr_dir;
 	union {
 		struct rpcrdma_fmr	fmr;
-		struct rpcrdma_frmr	frmr;
+		struct rpcrdma_frwr	frwr;
 	};
-	struct rpcrdma_xprt	*mw_xprt;
-	u32			mw_handle;
-	u32			mw_length;
-	u64			mw_offset;
-	struct list_head	mw_all;
-};
-
-/* mw_flags */
-enum {
-	RPCRDMA_MW_F_RI		= 1,
+	struct rpcrdma_xprt	*mr_xprt;
+	u32			mr_handle;
+	u32			mr_length;
+	u64			mr_offset;
+	struct list_head	mr_all;
 };
 
 /*
@@ -362,8 +354,7 @@ struct rpcrdma_req {
 
 /* rl_flags */
 enum {
-	RPCRDMA_REQ_F_BACKCHANNEL = 0,
-	RPCRDMA_REQ_F_PENDING,
+	RPCRDMA_REQ_F_PENDING = 0,
 	RPCRDMA_REQ_F_TX_RESOURCES,
 };
 
@@ -374,25 +365,25 @@ rpcrdma_set_xprtdata(struct rpc_rqst *rqst, struct rpcrdma_req *req)
 }
 
 static inline struct rpcrdma_req *
-rpcr_to_rdmar(struct rpc_rqst *rqst)
+rpcr_to_rdmar(const struct rpc_rqst *rqst)
 {
 	return rqst->rq_xprtdata;
 }
 
 static inline void
-rpcrdma_push_mw(struct rpcrdma_mw *mw, struct list_head *list)
+rpcrdma_mr_push(struct rpcrdma_mr *mr, struct list_head *list)
 {
-	list_add_tail(&mw->mw_list, list);
+	list_add_tail(&mr->mr_list, list);
 }
 
-static inline struct rpcrdma_mw *
-rpcrdma_pop_mw(struct list_head *list)
+static inline struct rpcrdma_mr *
+rpcrdma_mr_pop(struct list_head *list)
 {
-	struct rpcrdma_mw *mw;
+	struct rpcrdma_mr *mr;
 
-	mw = list_first_entry(list, struct rpcrdma_mw, mw_list);
-	list_del(&mw->mw_list);
-	return mw;
+	mr = list_first_entry(list, struct rpcrdma_mr, mr_list);
+	list_del(&mr->mr_list);
+	return mr;
 }
 
 /*
@@ -402,8 +393,8 @@ rpcrdma_pop_mw(struct list_head *list)
  * One of these is associated with a transport instance
  */
 struct rpcrdma_buffer {
-	spinlock_t		rb_mwlock;	/* protect rb_mws list */
-	struct list_head	rb_mws;
+	spinlock_t		rb_mrlock;	/* protect rb_mrs list */
+	struct list_head	rb_mrs;
 	struct list_head	rb_all;
 
 	unsigned long		rb_sc_head;
@@ -438,13 +429,11 @@ struct rpcrdma_buffer {
  * This data should be set with mount options
  */
 struct rpcrdma_create_data_internal {
-	struct sockaddr_storage	addr;	/* RDMA server address */
 	unsigned int	max_requests;	/* max requests (slots) in flight */
 	unsigned int	rsize;		/* mount rsize - max read hdr+data */
 	unsigned int	wsize;		/* mount wsize - max write hdr+data */
 	unsigned int	inline_rsize;	/* max non-rdma read data payload */
 	unsigned int	inline_wsize;	/* max non-rdma write data payload */
-	unsigned int	padding;	/* non-rdma write header padding */
 };
 
 /*
@@ -484,17 +473,19 @@ struct rpcrdma_memreg_ops {
 	struct rpcrdma_mr_seg *
 			(*ro_map)(struct rpcrdma_xprt *,
 				  struct rpcrdma_mr_seg *, int, bool,
-				  struct rpcrdma_mw **);
+				  struct rpcrdma_mr **);
+	void		(*ro_reminv)(struct rpcrdma_rep *rep,
+				     struct list_head *mrs);
 	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
 					 struct list_head *);
-	void		(*ro_recover_mr)(struct rpcrdma_mw *);
+	void		(*ro_recover_mr)(struct rpcrdma_mr *mr);
 	int		(*ro_open)(struct rpcrdma_ia *,
 				   struct rpcrdma_ep *,
 				   struct rpcrdma_create_data_internal *);
 	size_t		(*ro_maxpages)(struct rpcrdma_xprt *);
 	int		(*ro_init_mr)(struct rpcrdma_ia *,
-				      struct rpcrdma_mw *);
-	void		(*ro_release_mr)(struct rpcrdma_mw *);
+				      struct rpcrdma_mr *);
+	void		(*ro_release_mr)(struct rpcrdma_mr *mr);
 	const char	*ro_displayname;
 	const int	ro_send_w_inv_ok;
 };
@@ -525,6 +516,18 @@ struct rpcrdma_xprt {
 #define rpcx_to_rdmax(x) container_of(x, struct rpcrdma_xprt, rx_xprt)
 #define rpcx_to_rdmad(x) (rpcx_to_rdmax(x)->rx_data)
 
+static inline const char *
+rpcrdma_addrstr(const struct rpcrdma_xprt *r_xprt)
+{
+	return r_xprt->rx_xprt.address_strings[RPC_DISPLAY_ADDR];
+}
+
+static inline const char *
+rpcrdma_portstr(const struct rpcrdma_xprt *r_xprt)
+{
+	return r_xprt->rx_xprt.address_strings[RPC_DISPLAY_PORT];
+}
+
 /* Setting this to 0 ensures interoperability with early servers.
  * Setting this to 1 enhances certain unaligned read/write performance.
  * Default is 0, see sysctl entry and rpc_rdma.c rpcrdma_convert_iovs() */
@@ -538,7 +541,7 @@ extern unsigned int xprt_rdma_memreg_strategy;
 /*
  * Interface Adapter calls - xprtrdma/verbs.c
  */
-int rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr);
+int rpcrdma_ia_open(struct rpcrdma_xprt *xprt);
 void rpcrdma_ia_remove(struct rpcrdma_ia *ia);
 void rpcrdma_ia_close(struct rpcrdma_ia *);
 bool frwr_is_supported(struct rpcrdma_ia *);
@@ -564,22 +567,23 @@ int rpcrdma_ep_post_recv(struct rpcrdma_ia *, struct rpcrdma_rep *);
  * Buffer calls - xprtrdma/verbs.c
  */
 struct rpcrdma_req *rpcrdma_create_req(struct rpcrdma_xprt *);
-struct rpcrdma_rep *rpcrdma_create_rep(struct rpcrdma_xprt *);
 void rpcrdma_destroy_req(struct rpcrdma_req *);
+int rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt);
 int rpcrdma_buffer_create(struct rpcrdma_xprt *);
 void rpcrdma_buffer_destroy(struct rpcrdma_buffer *);
 struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_buffer *buf);
 void rpcrdma_sendctx_put_locked(struct rpcrdma_sendctx *sc);
 
-struct rpcrdma_mw *rpcrdma_get_mw(struct rpcrdma_xprt *);
-void rpcrdma_put_mw(struct rpcrdma_xprt *, struct rpcrdma_mw *);
+struct rpcrdma_mr *rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt);
+void rpcrdma_mr_put(struct rpcrdma_mr *mr);
+void rpcrdma_mr_unmap_and_put(struct rpcrdma_mr *mr);
+void rpcrdma_mr_defer_recovery(struct rpcrdma_mr *mr);
+
 struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *);
 void rpcrdma_buffer_put(struct rpcrdma_req *);
 void rpcrdma_recv_buffer_get(struct rpcrdma_req *);
 void rpcrdma_recv_buffer_put(struct rpcrdma_rep *);
 
-void rpcrdma_defer_mr_recovery(struct rpcrdma_mw *);
-
 struct rpcrdma_regbuf *rpcrdma_alloc_regbuf(size_t, enum dma_data_direction,
 					    gfp_t);
 bool __rpcrdma_dma_map_regbuf(struct rpcrdma_ia *, struct rpcrdma_regbuf *);
@@ -663,7 +667,7 @@ int xprt_rdma_bc_up(struct svc_serv *, struct net *);
 size_t xprt_rdma_bc_maxpayload(struct rpc_xprt *);
 int rpcrdma_bc_post_recv(struct rpcrdma_xprt *, unsigned int);
 void rpcrdma_bc_receive_call(struct rpcrdma_xprt *, struct rpcrdma_rep *);
-int rpcrdma_bc_marshal_reply(struct rpc_rqst *);
+int xprt_rdma_bc_send_reply(struct rpc_rqst *rqst);
 void xprt_rdma_bc_free_rqst(struct rpc_rqst *);
 void xprt_rdma_bc_destroy(struct rpc_xprt *, unsigned int);
 #endif	/* CONFIG_SUNRPC_BACKCHANNEL */
@@ -671,3 +675,5 @@ void xprt_rdma_bc_destroy(struct rpc_xprt *, unsigned int);
 extern struct xprt_class xprt_rdma_bc;
 
 #endif				/* _LINUX_SUNRPC_XPRT_RDMA_H */
+
+#include <trace/events/rpcrdma.h>

net/sunrpc/xprtsock.c

@@ -52,6 +52,8 @@
 
 #include "sunrpc.h"
 
+#define RPC_TCP_READ_CHUNK_SZ	(3*512*1024)
+
 static void xs_close(struct rpc_xprt *xprt);
 static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 		struct socket *sock);
@@ -1003,6 +1005,7 @@ static void xs_local_data_receive(struct sock_xprt *transport)
 	struct sock *sk;
 	int err;
 
+restart:
 	mutex_lock(&transport->recv_mutex);
 	sk = transport->inet;
 	if (sk == NULL)
@@ -1016,6 +1019,11 @@ static void xs_local_data_receive(struct sock_xprt *transport)
 		}
 		if (!test_and_clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
 			break;
+		if (need_resched()) {
+			mutex_unlock(&transport->recv_mutex);
+			cond_resched();
+			goto restart;
+		}
 	}
 out:
 	mutex_unlock(&transport->recv_mutex);
@@ -1094,6 +1102,7 @@ static void xs_udp_data_receive(struct sock_xprt *transport)
 	struct sock *sk;
 	int err;
 
+restart:
 	mutex_lock(&transport->recv_mutex);
 	sk = transport->inet;
 	if (sk == NULL)
@@ -1107,6 +1116,11 @@ static void xs_udp_data_receive(struct sock_xprt *transport)
 		}
 		if (!test_and_clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
 			break;
+		if (need_resched()) {
+			mutex_unlock(&transport->recv_mutex);
+			cond_resched();
+			goto restart;
+		}
 	}
 out:
 	mutex_unlock(&transport->recv_mutex);
@@ -1479,6 +1493,7 @@ static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, uns
 		.offset	= offset,
 		.count	= len,
 	};
+	size_t ret;
 
 	dprintk("RPC:       xs_tcp_data_recv started\n");
 	do {
@@ -1507,9 +1522,14 @@ static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, uns
 		/* Skip over any trailing bytes on short reads */
 		xs_tcp_read_discard(transport, &desc);
 	} while (desc.count);
+	ret = len - desc.count;
+	if (ret < rd_desc->count)
+		rd_desc->count -= ret;
+	else
+		rd_desc->count = 0;
 	trace_xs_tcp_data_recv(transport);
 	dprintk("RPC:       xs_tcp_data_recv done\n");
-	return len - desc.count;
+	return ret;
 }
 
 static void xs_tcp_data_receive(struct sock_xprt *transport)
@@ -1517,30 +1537,34 @@ static void xs_tcp_data_receive(struct sock_xprt *transport)
 	struct rpc_xprt *xprt = &transport->xprt;
 	struct sock *sk;
 	read_descriptor_t rd_desc = {
-		.count = 2*1024*1024,
 		.arg.data = xprt,
 	};
 	unsigned long total = 0;
-	int loop;
 	int read = 0;
 
+restart:
 	mutex_lock(&transport->recv_mutex);
 	sk = transport->inet;
 	if (sk == NULL)
 		goto out;
 
 	/* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
-	for (loop = 0; loop < 64; loop++) {
+	for (;;) {
+		rd_desc.count = RPC_TCP_READ_CHUNK_SZ;
 		lock_sock(sk);
 		read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
-		if (read <= 0) {
+		if (rd_desc.count != 0 || read < 0) {
 			clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state);
 			release_sock(sk);
 			break;
 		}
 		release_sock(sk);
 		total += read;
-		rd_desc.count = 65536;
+		if (need_resched()) {
+			mutex_unlock(&transport->recv_mutex);
+			cond_resched();
+			goto restart;
+		}
 	}
 	if (test_bit(XPRT_SOCK_DATA_READY, &transport->sock_state))
 		queue_work(xprtiod_workqueue, &transport->recv_worker);