Merge branch 'pnfs_generic'

* pnfs_generic:
  NFSv4.1/pNFS: Cleanup constify struct pnfs_layout_range arguments
  NFSv4.1/pnfs: Cleanup copying of pnfs_layout_range structures
  NFSv4.1/pNFS: Cleanup pnfs_mark_matching_lsegs_invalid()
  NFSv4.1/pNFS: Fix a race in initiate_file_draining()
  NFSv4.1/pNFS: pnfs_error_mark_layout_for_return() must always return layout
  NFSv4.1/pNFS: pnfs_mark_matching_lsegs_return() should set the iomode
  NFSv4.1/pNFS: Use nfs4_stateid_copy for copying stateids
  NFSv4.1/pNFS: Don't pass stateids by value to pnfs_send_layoutreturn()
  NFS: Relax requirements in nfs_flush_incompatible
  NFSv4.1/pNFS: Don't queue up a new commit if the layout segment is invalid
  NFS: Allow multiple commit requests in flight per file
  NFS/pNFS: Fix up pNFS write reschedule layering violations and bugs
  NFSv4: List stateid information in the callback tracepoints
  NFSv4.1/pNFS: Don't return NFS4ERR_DELAY unnecessarily in CB_LAYOUTRECALL
  NFSv4.1/pNFS: Ensure we enforce RFC5661 Section 12.5.5.2.1
  pNFS: If we have to delay the layout callback, mark the layout for return
  NFSv4.1/pNFS: Add a helper to mark the layout as returned
  pNFS: Ensure nfs4_layoutget_prepare returns the correct error

15 files changed, 296 insertions, 117 deletions

diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c
index 807eb6ef4f91..f0939d097406 100644
--- a/fs/nfs/callback_proc.c
+++ b/fs/nfs/callback_proc.c
@@ -83,8 +83,11 @@ __be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy,
 
 	res = htonl(NFS4ERR_BADHANDLE);
 	inode = nfs_delegation_find_inode(cps->clp, &args->fh);
-	if (inode == NULL)
+	if (inode == NULL) {
+		trace_nfs4_cb_recall(cps->clp, &args->fh, NULL,
+				&args->stateid, -ntohl(res));
 		goto out;
+	}
 	/* Set up a helper thread to actually return the delegation */
 	switch (nfs_async_inode_return_delegation(inode, &args->stateid)) {
 	case 0:
@@ -96,7 +99,8 @@ __be32 nfs4_callback_recall(struct cb_recallargs *args, void *dummy,
 	default:
 		res = htonl(NFS4ERR_RESOURCE);
 	}
-	trace_nfs4_recall_delegation(inode, -ntohl(res));
+	trace_nfs4_cb_recall(cps->clp, &args->fh, inode,
+			&args->stateid, -ntohl(res));
 	iput(inode);
 out:
 	dprintk("%s: exit with status = %d\n", __func__, ntohl(res));
@@ -160,6 +164,22 @@ static struct pnfs_layout_hdr * get_layout_by_fh(struct nfs_client *clp,
 	return lo;
 }
 
+/*
+ * Enforce RFC5661 section 12.5.5.2.1. (Layout Recall and Return Sequencing)
+ */
+static bool pnfs_check_stateid_sequence(struct pnfs_layout_hdr *lo,
+					const nfs4_stateid *new)
+{
+	u32 oldseq, newseq;
+
+	oldseq = be32_to_cpu(lo->plh_stateid.seqid);
+	newseq = be32_to_cpu(new->seqid);
+
+	if (newseq > oldseq + 1)
+		return false;
+	return true;
+}
+
 static u32 initiate_file_draining(struct nfs_client *clp,
 				  struct cb_layoutrecallargs *args)
 {
@@ -169,34 +189,52 @@ static u32 initiate_file_draining(struct nfs_client *clp,
 	LIST_HEAD(free_me_list);
 
 	lo = get_layout_by_fh(clp, &args->cbl_fh, &args->cbl_stateid);
-	if (!lo)
+	if (!lo) {
+		trace_nfs4_cb_layoutrecall_file(clp, &args->cbl_fh, NULL,
+				&args->cbl_stateid, -rv);
 		goto out;
+	}
 
 	ino = lo->plh_inode;
 
 	spin_lock(&ino->i_lock);
+	if (!pnfs_check_stateid_sequence(lo, &args->cbl_stateid)) {
+		rv = NFS4ERR_DELAY;
+		goto unlock;
+	}
 	pnfs_set_layout_stateid(lo, &args->cbl_stateid, true);
 	spin_unlock(&ino->i_lock);
 
 	pnfs_layoutcommit_inode(ino, false);
 
 	spin_lock(&ino->i_lock);
-	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags) ||
-	    pnfs_mark_matching_lsegs_invalid(lo, &free_me_list,
-					&args->cbl_range)) {
+	/*
+	 * Enforce RFC5661 Section 12.5.5.2.1.5 (Bulk Recall and Return)
+	 */
+	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) {
 		rv = NFS4ERR_DELAY;
 		goto unlock;
 	}
 
+	if (pnfs_mark_matching_lsegs_return(lo, &free_me_list,
+					&args->cbl_range)) {
+		rv = NFS4_OK;
+		goto unlock;
+	}
+
 	if (NFS_SERVER(ino)->pnfs_curr_ld->return_range) {
 		NFS_SERVER(ino)->pnfs_curr_ld->return_range(lo,
 			&args->cbl_range);
 	}
+	pnfs_mark_layout_returned_if_empty(lo);
 unlock:
 	spin_unlock(&ino->i_lock);
 	pnfs_free_lseg_list(&free_me_list);
+	/* Free all lsegs that are attached to commit buckets */
+	nfs_commit_inode(ino, 0);
 	pnfs_put_layout_hdr(lo);
-	trace_nfs4_cb_layoutrecall_inode(clp, &args->cbl_fh, ino, -rv);
+	trace_nfs4_cb_layoutrecall_file(clp, &args->cbl_fh, ino,
+			&args->cbl_stateid, -rv);
 	iput(ino);
 out:
 	return rv;
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 2e7142bcb4c8..7ab7ec9f4eed 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -117,12 +117,6 @@ static inline int put_dreq(struct nfs_direct_req *dreq)
 	return atomic_dec_and_test(&dreq->io_count);
 }
 
-void nfs_direct_set_resched_writes(struct nfs_direct_req *dreq)
-{
-	dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
-}
-EXPORT_SYMBOL_GPL(nfs_direct_set_resched_writes);
-
 static void
 nfs_direct_good_bytes(struct nfs_direct_req *dreq, struct nfs_pgio_header *hdr)
 {
@@ -735,14 +729,20 @@ static void nfs_direct_commit_complete(struct nfs_commit_data *data)
 		nfs_direct_write_complete(dreq, data->inode);
 }
 
-static void nfs_direct_error_cleanup(struct nfs_inode *nfsi)
+static void nfs_direct_resched_write(struct nfs_commit_info *cinfo,
+		struct nfs_page *req)
 {
-	/* There is no lock to clear */
+	struct nfs_direct_req *dreq = cinfo->dreq;
+
+	spin_lock(&dreq->lock);
+	dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+	spin_unlock(&dreq->lock);
+	nfs_mark_request_commit(req, NULL, cinfo, 0);
 }
 
 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
 	.completion = nfs_direct_commit_complete,
-	.error_cleanup = nfs_direct_error_cleanup,
+	.resched_write = nfs_direct_resched_write,
 };
 
 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
@@ -847,10 +847,25 @@ static void nfs_write_sync_pgio_error(struct list_head *head)
 	}
 }
 
+static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr)
+{
+	struct nfs_direct_req *dreq = hdr->dreq;
+
+	spin_lock(&dreq->lock);
+	if (dreq->error == 0) {
+		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+		/* fake unstable write to let common nfs resend pages */
+		hdr->verf.committed = NFS_UNSTABLE;
+		hdr->good_bytes = hdr->args.count;
+	}
+	spin_unlock(&dreq->lock);
+}
+
 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
 	.error_cleanup = nfs_write_sync_pgio_error,
 	.init_hdr = nfs_direct_pgio_init,
 	.completion = nfs_direct_write_completion,
+	.reschedule_io = nfs_direct_write_reschedule_io,
 };
 
 
diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index f188dd071dfc..178ec8da028f 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -514,7 +514,7 @@ static void nfs_check_dirty_writeback(struct page *page,
 	 * so it will not block due to pages that will shortly be freeable.
 	 */
 	nfsi = NFS_I(mapping->host);
-	if (test_bit(NFS_INO_COMMIT, &nfsi->flags)) {
+	if (atomic_read(&nfsi->commit_info.rpcs_out)) {
 		*writeback = true;
 		return;
 	}
diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c
index 2981cd190bfd..18c329b84ffb 100644
--- a/fs/nfs/flexfilelayout/flexfilelayout.c
+++ b/fs/nfs/flexfilelayout/flexfilelayout.c
@@ -941,18 +941,7 @@ static void ff_layout_reset_write(struct nfs_pgio_header *hdr, bool retry_pnfs)
 			hdr->args.count,
 			(unsigned long long)hdr->args.offset);
 
-		if (!hdr->dreq) {
-			struct nfs_open_context *ctx;
-
-			ctx = nfs_list_entry(hdr->pages.next)->wb_context;
-			set_bit(NFS_CONTEXT_RESEND_WRITES, &ctx->flags);
-			hdr->completion_ops->error_cleanup(&hdr->pages);
-		} else {
-			nfs_direct_set_resched_writes(hdr->dreq);
-			/* fake unstable write to let common nfs resend pages */
-			hdr->verf.committed = NFS_UNSTABLE;
-			hdr->good_bytes = hdr->args.count;
-		}
+		hdr->completion_ops->reschedule_io(hdr);
 		return;
 	}
 
diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index 68f773dc226e..ee81792d2886 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -264,6 +264,12 @@ static inline bool nfs_pgio_has_mirroring(struct nfs_pageio_descriptor *desc)
 	return desc->pg_mirror_count > 1;
 }
 
+static inline bool nfs_match_open_context(const struct nfs_open_context *ctx1,
+		const struct nfs_open_context *ctx2)
+{
+	return ctx1->cred == ctx2->cred && ctx1->state == ctx2->state;
+}
+
 /* nfs2xdr.c */
 extern struct rpc_procinfo nfs_procedures[];
 extern int nfs2_decode_dirent(struct xdr_stream *,
@@ -519,7 +525,6 @@ static inline void nfs_inode_dio_wait(struct inode *inode)
 	inode_dio_wait(inode);
 }
 extern ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq);
-extern void nfs_direct_set_resched_writes(struct nfs_direct_req *dreq);
 
 /* nfs4proc.c */
 extern void __nfs4_read_done_cb(struct nfs_pgio_header *);
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index adae525edec4..5e5062c9b92b 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -7776,6 +7776,7 @@ nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
 	struct nfs4_layoutget *lgp = calldata;
 	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
 	struct nfs4_session *session = nfs4_get_session(server);
+	int ret;
 
 	dprintk("--> %s\n", __func__);
 	/* Note the is a race here, where a CB_LAYOUTRECALL can come in
@@ -7786,12 +7787,12 @@ nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
 	if (nfs41_setup_sequence(session, &lgp->args.seq_args,
 				&lgp->res.seq_res, task))
 		return;
-	if (pnfs_choose_layoutget_stateid(&lgp->args.stateid,
+	ret = pnfs_choose_layoutget_stateid(&lgp->args.stateid,
 					  NFS_I(lgp->args.inode)->layout,
 					  &lgp->args.range,
-					  lgp->args.ctx->state)) {
-		rpc_exit(task, NFS4_OK);
-	}
+					  lgp->args.ctx->state);
+	if (ret < 0)
+		rpc_exit(task, ret);
 }
 
 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
@@ -8073,9 +8074,10 @@ static void nfs4_layoutreturn_release(void *calldata)
 
 	dprintk("--> %s\n", __func__);
 	spin_lock(&lo->plh_inode->i_lock);
+	pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
+	pnfs_mark_layout_returned_if_empty(lo);
 	if (lrp->res.lrs_present)
 		pnfs_set_layout_stateid(lo, &lrp->res.stateid, true);
-	pnfs_mark_matching_lsegs_invalid(lo, &freeme, &lrp->args.range);
 	pnfs_clear_layoutreturn_waitbit(lo);
 	lo->plh_block_lgets--;
 	spin_unlock(&lo->plh_inode->i_lock);
diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h
index d08d0c84b778..88b6b14ce71b 100644
--- a/fs/nfs/nfs4trace.h
+++ b/fs/nfs/nfs4trace.h
@@ -982,7 +982,6 @@ DEFINE_NFS4_INODE_EVENT(nfs4_set_acl);
 DEFINE_NFS4_INODE_EVENT(nfs4_get_security_label);
 DEFINE_NFS4_INODE_EVENT(nfs4_set_security_label);
 #endif /* CONFIG_NFS_V4_SECURITY_LABEL */
-DEFINE_NFS4_INODE_EVENT(nfs4_recall_delegation);
 
 DECLARE_EVENT_CLASS(nfs4_inode_stateid_event,
 		TP_PROTO(
@@ -1145,8 +1144,74 @@ DECLARE_EVENT_CLASS(nfs4_inode_callback_event,
 			), \
 			TP_ARGS(clp, fhandle, inode, error))
 DEFINE_NFS4_INODE_CALLBACK_EVENT(nfs4_cb_getattr);
-DEFINE_NFS4_INODE_CALLBACK_EVENT(nfs4_cb_layoutrecall_inode);
 
+DECLARE_EVENT_CLASS(nfs4_inode_stateid_callback_event,
+		TP_PROTO(
+			const struct nfs_client *clp,
+			const struct nfs_fh *fhandle,
+			const struct inode *inode,
+			const nfs4_stateid *stateid,
+			int error
+		),
+
+		TP_ARGS(clp, fhandle, inode, stateid, error),
+
+		TP_STRUCT__entry(
+			__field(int, error)
+			__field(dev_t, dev)
+			__field(u32, fhandle)
+			__field(u64, fileid)
+			__string(dstaddr, clp ?
+				rpc_peeraddr2str(clp->cl_rpcclient,
+					RPC_DISPLAY_ADDR) : "unknown")
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+		),
+
+		TP_fast_assign(
+			__entry->error = error;
+			__entry->fhandle = nfs_fhandle_hash(fhandle);
+			if (inode != NULL) {
+				__entry->fileid = NFS_FILEID(inode);
+				__entry->dev = inode->i_sb->s_dev;
+			} else {
+				__entry->fileid = 0;
+				__entry->dev = 0;
+			}
+			__assign_str(dstaddr, clp ?
+				rpc_peeraddr2str(clp->cl_rpcclient,
+					RPC_DISPLAY_ADDR) : "unknown")
+			__entry->stateid_seq =
+				be32_to_cpu(stateid->seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(stateid);
+		),
+
+		TP_printk(
+			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x "
+			"stateid=%d:0x%08x dstaddr=%s",
+			__entry->error,
+			show_nfsv4_errors(__entry->error),
+			MAJOR(__entry->dev), MINOR(__entry->dev),
+			(unsigned long long)__entry->fileid,
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash,
+			__get_str(dstaddr)
+		)
+);
+
+#define DEFINE_NFS4_INODE_STATEID_CALLBACK_EVENT(name) \
+	DEFINE_EVENT(nfs4_inode_stateid_callback_event, name, \
+			TP_PROTO( \
+				const struct nfs_client *clp, \
+				const struct nfs_fh *fhandle, \
+				const struct inode *inode, \
+				const nfs4_stateid *stateid, \
+				int error \
+			), \
+			TP_ARGS(clp, fhandle, inode, stateid, error))
+DEFINE_NFS4_INODE_STATEID_CALLBACK_EVENT(nfs4_cb_recall);
+DEFINE_NFS4_INODE_STATEID_CALLBACK_EVENT(nfs4_cb_layoutrecall_file);
 
 DECLARE_EVENT_CLASS(nfs4_idmap_event,
 		TP_PROTO(
diff --git a/fs/nfs/nfstrace.h b/fs/nfs/nfstrace.h
index 59f838cdc009..9f80a086b612 100644
--- a/fs/nfs/nfstrace.h
+++ b/fs/nfs/nfstrace.h
@@ -39,7 +39,6 @@
 			{ 1 << NFS_INO_INVALIDATING, "INVALIDATING" }, \
 			{ 1 << NFS_INO_FLUSHING, "FLUSHING" }, \
 			{ 1 << NFS_INO_FSCACHE, "FSCACHE" }, \
-			{ 1 << NFS_INO_COMMIT, "COMMIT" }, \
 			{ 1 << NFS_INO_LAYOUTCOMMIT, "NEED_LAYOUTCOMMIT" }, \
 			{ 1 << NFS_INO_LAYOUTCOMMITTING, "LAYOUTCOMMIT" })
 
diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index 7c71b71016b5..eeddbf0bf4c4 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -899,12 +899,6 @@ static void nfs_pageio_cleanup_mirroring(struct nfs_pageio_descriptor *pgio)
 	pgio->pg_mirrors_dynamic = NULL;
 }
 
-static bool nfs_match_open_context(const struct nfs_open_context *ctx1,
-		const struct nfs_open_context *ctx2)
-{
-	return ctx1->cred == ctx2->cred && ctx1->state == ctx2->state;
-}
-
 static bool nfs_match_lock_context(const struct nfs_lock_context *l1,
 		const struct nfs_lock_context *l2)
 {
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 113c3b327e24..a3592cc34a20 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -53,7 +53,7 @@ static DEFINE_SPINLOCK(pnfs_spinlock);
 static LIST_HEAD(pnfs_modules_tbl);
 
 static int
-pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, nfs4_stateid stateid,
+pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, const nfs4_stateid *stateid,
 		       enum pnfs_iomode iomode, bool sync);
 
 /* Return the registered pnfs layout driver module matching given id */
@@ -385,13 +385,13 @@ static void pnfs_layoutreturn_before_put_lseg(struct pnfs_layout_segment *lseg,
 		enum pnfs_iomode iomode;
 		bool send;
 
-		stateid = lo->plh_stateid;
+		nfs4_stateid_copy(&stateid, &lo->plh_stateid);
 		iomode = lo->plh_return_iomode;
 		send = pnfs_prepare_layoutreturn(lo);
 		spin_unlock(&inode->i_lock);
 		if (send) {
 			/* Send an async layoutreturn so we dont deadlock */
-			pnfs_send_layoutreturn(lo, stateid, iomode, false);
+			pnfs_send_layoutreturn(lo, &stateid, iomode, false);
 		}
 	} else
 		spin_unlock(&inode->i_lock);
@@ -566,10 +566,10 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg,
 int
 pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 			    struct list_head *tmp_list,
-			    struct pnfs_layout_range *recall_range)
+			    const struct pnfs_layout_range *recall_range)
 {
 	struct pnfs_layout_segment *lseg, *next;
-	int invalid = 0, removed = 0;
+	int remaining = 0;
 
 	dprintk("%s:Begin lo %p\n", __func__, lo);
 
@@ -582,11 +582,11 @@ pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 				"offset %llu length %llu\n", __func__,
 				lseg, lseg->pls_range.iomode, lseg->pls_range.offset,
 				lseg->pls_range.length);
-			invalid++;
-			removed += mark_lseg_invalid(lseg, tmp_list);
+			if (!mark_lseg_invalid(lseg, tmp_list))
+				remaining++;
 		}
-	dprintk("%s:Return %i\n", __func__, invalid - removed);
-	return invalid - removed;
+	dprintk("%s:Return %i\n", __func__, remaining);
+	return remaining;
 }
 
 /* note free_me must contain lsegs from a single layout_hdr */
@@ -702,6 +702,8 @@ pnfs_layout_free_bulk_destroy_list(struct list_head *layout_list,
 			ret = -EAGAIN;
 		spin_unlock(&inode->i_lock);
 		pnfs_free_lseg_list(&lseg_list);
+		/* Free all lsegs that are attached to commit buckets */
+		nfs_commit_inode(inode, 0);
 		pnfs_put_layout_hdr(lo);
 		iput(inode);
 	}
@@ -825,7 +827,7 @@ pnfs_layoutgets_blocked(const struct pnfs_layout_hdr *lo)
 
 int
 pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo,
-			      struct pnfs_layout_range *range,
+			      const struct pnfs_layout_range *range,
 			      struct nfs4_state *open_state)
 {
 	int status = 0;
@@ -860,7 +862,7 @@ pnfs_choose_layoutget_stateid(nfs4_stateid *dst, struct pnfs_layout_hdr *lo,
 static struct pnfs_layout_segment *
 send_layoutget(struct pnfs_layout_hdr *lo,
 	   struct nfs_open_context *ctx,
-	   struct pnfs_layout_range *range,
+	   const struct pnfs_layout_range *range,
 	   gfp_t gfp_flags)
 {
 	struct inode *ino = lo->plh_inode;
@@ -893,7 +895,7 @@ send_layoutget(struct pnfs_layout_hdr *lo,
 				lgp->args.minlength = i_size - range->offset;
 		}
 		lgp->args.maxcount = PNFS_LAYOUT_MAXSIZE;
-		lgp->args.range = *range;
+		pnfs_copy_range(&lgp->args.range, range);
 		lgp->args.type = server->pnfs_curr_ld->id;
 		lgp->args.inode = ino;
 		lgp->args.ctx = get_nfs_open_context(ctx);
@@ -936,7 +938,7 @@ void pnfs_clear_layoutreturn_waitbit(struct pnfs_layout_hdr *lo)
 }
 
 static int
-pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, nfs4_stateid stateid,
+pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, const nfs4_stateid *stateid,
 		       enum pnfs_iomode iomode, bool sync)
 {
 	struct inode *ino = lo->plh_inode;
@@ -953,7 +955,7 @@ pnfs_send_layoutreturn(struct pnfs_layout_hdr *lo, nfs4_stateid stateid,
 		goto out;
 	}
 
-	lrp->args.stateid = stateid;
+	nfs4_stateid_copy(&lrp->args.stateid, stateid);
 	lrp->args.layout_type = NFS_SERVER(ino)->pnfs_curr_ld->id;
 	lrp->args.inode = ino;
 	lrp->args.range.iomode = iomode;
@@ -996,7 +998,7 @@ _pnfs_return_layout(struct inode *ino)
 		dprintk("NFS: %s no layout to return\n", __func__);
 		goto out;
 	}
-	stateid = nfsi->layout->plh_stateid;
+	nfs4_stateid_copy(&stateid, &nfsi->layout->plh_stateid);
 	/* Reference matched in nfs4_layoutreturn_release */
 	pnfs_get_layout_hdr(lo);
 	empty = list_empty(&lo->plh_segs);
@@ -1024,7 +1026,7 @@ _pnfs_return_layout(struct inode *ino)
 	spin_unlock(&ino->i_lock);
 	pnfs_free_lseg_list(&tmp_list);
 	if (send)
-		status = pnfs_send_layoutreturn(lo, stateid, IOMODE_ANY, true);
+		status = pnfs_send_layoutreturn(lo, &stateid, IOMODE_ANY, true);
 out_put_layout_hdr:
 	pnfs_put_layout_hdr(lo);
 out:
@@ -1087,7 +1089,7 @@ bool pnfs_roc(struct inode *ino)
 			goto out_noroc;
 	}
 
-	stateid = lo->plh_stateid;
+	nfs4_stateid_copy(&stateid, &lo->plh_stateid);
 	/* always send layoutreturn if being marked so */
 	if (test_and_clear_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE,
 				   &lo->plh_flags))
@@ -1114,7 +1116,7 @@ out_noroc:
 	pnfs_free_lseg_list(&tmp_list);
 	pnfs_layoutcommit_inode(ino, true);
 	if (layoutreturn)
-		pnfs_send_layoutreturn(lo, stateid, IOMODE_ANY, true);
+		pnfs_send_layoutreturn(lo, &stateid, IOMODE_ANY, true);
 	return roc;
 }
 
@@ -1139,6 +1141,7 @@ void pnfs_roc_set_barrier(struct inode *ino, u32 barrier)
 
 	spin_lock(&ino->i_lock);
 	lo = NFS_I(ino)->layout;
+	pnfs_mark_layout_returned_if_empty(lo);
 	if (pnfs_seqid_is_newer(barrier, lo->plh_barrier))
 		lo->plh_barrier = barrier;
 	spin_unlock(&ino->i_lock);
@@ -1734,16 +1737,29 @@ out_forget_reply:
 }
 
 static void
+pnfs_set_plh_return_iomode(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode)
+{
+	if (lo->plh_return_iomode == iomode)
+		return;
+	if (lo->plh_return_iomode != 0)
+		iomode = IOMODE_ANY;
+	lo->plh_return_iomode = iomode;
+}
+
+int
 pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 				struct list_head *tmp_list,
-				struct pnfs_layout_range *return_range)
+				const struct pnfs_layout_range *return_range)
 {
 	struct pnfs_layout_segment *lseg, *next;
+	int remaining = 0;
 
 	dprintk("%s:Begin lo %p\n", __func__, lo);
 
 	if (list_empty(&lo->plh_segs))
-		return;
+		return 0;
+
+	assert_spin_locked(&lo->plh_inode->i_lock);
 
 	list_for_each_entry_safe(lseg, next, &lo->plh_segs, pls_list)
 		if (should_free_lseg(&lseg->pls_range, return_range)) {
@@ -1753,10 +1769,13 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 				lseg->pls_range.offset,
 				lseg->pls_range.length);
 			set_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags);
-			mark_lseg_invalid(lseg, tmp_list);
+			pnfs_set_plh_return_iomode(lo, return_range->iomode);
+			if (!mark_lseg_invalid(lseg, tmp_list))
+				remaining++;
 			set_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE,
 					&lo->plh_flags);
 		}
+	return remaining;
 }
 
 void pnfs_error_mark_layout_for_return(struct inode *inode,
@@ -1769,19 +1788,28 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 		.length = NFS4_MAX_UINT64,
 	};
 	LIST_HEAD(free_me);
+	bool return_now = false;
 
 	spin_lock(&inode->i_lock);
-	if (lo->plh_return_iomode == 0)
-		lo->plh_return_iomode = range.iomode;
-	else if (lo->plh_return_iomode != range.iomode)
-		lo->plh_return_iomode = IOMODE_ANY;
+	pnfs_set_plh_return_iomode(lo, range.iomode);
 	/*
 	 * mark all matching lsegs so that we are sure to have no live
 	 * segments at hand when sending layoutreturn. See pnfs_put_lseg()
 	 * for how it works.
 	 */
-	pnfs_mark_matching_lsegs_return(lo, &free_me, &range);
-	spin_unlock(&inode->i_lock);
+	if (!pnfs_mark_matching_lsegs_return(lo, &free_me, &range)) {
+		nfs4_stateid stateid;
+		enum pnfs_iomode iomode = lo->plh_return_iomode;
+
+		nfs4_stateid_copy(&stateid, &lo->plh_stateid);
+		return_now = pnfs_prepare_layoutreturn(lo);
+		spin_unlock(&inode->i_lock);
+		if (return_now)
+			pnfs_send_layoutreturn(lo, &stateid, iomode, false);
+	} else {
+		spin_unlock(&inode->i_lock);
+		nfs_commit_inode(inode, 0);
+	}
 	pnfs_free_lseg_list(&free_me);
 }
 EXPORT_SYMBOL_GPL(pnfs_error_mark_layout_for_return);
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 6916ff4e86f9..9f4e2a47f4aa 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -260,11 +260,14 @@ void pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo,
 			     bool update_barrier);
 int pnfs_choose_layoutget_stateid(nfs4_stateid *dst,
 				  struct pnfs_layout_hdr *lo,
-				  struct pnfs_layout_range *range,
+				  const struct pnfs_layout_range *range,
 				  struct nfs4_state *open_state);
 int pnfs_mark_matching_lsegs_invalid(struct pnfs_layout_hdr *lo,
 				struct list_head *tmp_list,
-				struct pnfs_layout_range *recall_range);
+				const struct pnfs_layout_range *recall_range);
+int pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
+				struct list_head *tmp_list,
+				const struct pnfs_layout_range *recall_range);
 bool pnfs_roc(struct inode *ino);
 void pnfs_roc_release(struct inode *ino);
 void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
@@ -388,6 +391,12 @@ pnfs_get_lseg(struct pnfs_layout_segment *lseg)
 	return lseg;
 }
 
+static inline bool
+pnfs_is_valid_lseg(struct pnfs_layout_segment *lseg)
+{
+	return test_bit(NFS_LSEG_VALID, &lseg->pls_flags) != 0;
+}
+
 /* Return true if a layout driver is being used for this mountpoint */
 static inline int pnfs_enabled_sb(struct nfs_server *nfss)
 {
@@ -535,6 +544,26 @@ pnfs_calc_offset_length(u64 offset, u64 end)
 	return 1 + end - offset;
 }
 
+/**
+ * pnfs_mark_layout_returned_if_empty - marks the layout as returned
+ * @lo: layout header
+ *
+ * Note: Caller must hold inode->i_lock
+ */
+static inline void
+pnfs_mark_layout_returned_if_empty(struct pnfs_layout_hdr *lo)
+{
+	if (list_empty(&lo->plh_segs))
+		set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
+}
+
+static inline void
+pnfs_copy_range(struct pnfs_layout_range *dst,
+		const struct pnfs_layout_range *src)
+{
+	memcpy(dst, src, sizeof(*dst));
+}
+
 extern unsigned int layoutstats_timer;
 
 #ifdef NFS_DEBUG
diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c
index 24655b807d44..81ac6480f9e7 100644
--- a/fs/nfs/pnfs_nfs.c
+++ b/fs/nfs/pnfs_nfs.c
@@ -266,17 +266,14 @@ pnfs_generic_commit_pagelist(struct inode *inode, struct list_head *mds_pages,
 		} else {
 			nfs_retry_commit(mds_pages, NULL, cinfo, 0);
 			pnfs_generic_retry_commit(cinfo, 0);
-			cinfo->completion_ops->error_cleanup(NFS_I(inode));
 			return -ENOMEM;
 		}
 	}
 
 	nreq += pnfs_generic_alloc_ds_commits(cinfo, &list);
 
-	if (nreq == 0) {
-		cinfo->completion_ops->error_cleanup(NFS_I(inode));
+	if (nreq == 0)
 		goto out;
-	}
 
 	atomic_add(nreq, &cinfo->mds->rpcs_out);
 
@@ -871,6 +868,11 @@ pnfs_layout_mark_request_commit(struct nfs_page *req,
 	buckets = cinfo->ds->buckets;
 	list = &buckets[ds_commit_idx].written;
 	if (list_empty(list)) {
+		if (!pnfs_is_valid_lseg(lseg)) {
+			spin_unlock(cinfo->lock);
+			cinfo->completion_ops->resched_write(cinfo, req);
+			return;
+		}
 		/* Non-empty buckets hold a reference on the lseg.  That ref
 		 * is normally transferred to the COMMIT call and released
 		 * there.  It could also be released if the last req is pulled
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 4d254232d728..94828b3f8c95 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -21,6 +21,8 @@
 #include <linux/nfs_page.h>
 #include <linux/backing-dev.h>
 #include <linux/export.h>
+#include <linux/freezer.h>
+#include <linux/wait.h>
 
 #include <asm/uaccess.h>
 
@@ -1155,7 +1157,8 @@ int nfs_flush_incompatible(struct file *file, struct page *page)
 		if (req == NULL)
 			return 0;
 		l_ctx = req->wb_lock_context;
-		do_flush = req->wb_page != page || req->wb_context != ctx;
+		do_flush = req->wb_page != page ||
+			!nfs_match_open_context(req->wb_context, ctx);
 		/* for now, flush if more than 1 request in page_group */
 		do_flush |= req->wb_this_page != req;
 		if (l_ctx && flctx &&
@@ -1353,9 +1356,15 @@ static void nfs_async_write_error(struct list_head *head)
 	}
 }
 
+static void nfs_async_write_reschedule_io(struct nfs_pgio_header *hdr)
+{
+	nfs_async_write_error(&hdr->pages);
+}
+
 static const struct nfs_pgio_completion_ops nfs_async_write_completion_ops = {
 	.error_cleanup = nfs_async_write_error,
 	.completion = nfs_write_completion,
+	.reschedule_io = nfs_async_write_reschedule_io,
 };
 
 void nfs_pageio_init_write(struct nfs_pageio_descriptor *pgio,
@@ -1556,27 +1565,29 @@ static void nfs_writeback_result(struct rpc_task *task,
 	}
 }
 
+static int nfs_wait_atomic_killable(atomic_t *key)
+{
+	if (fatal_signal_pending(current))
+		return -ERESTARTSYS;
+	freezable_schedule_unsafe();
+	return 0;
+}
 
-static int nfs_commit_set_lock(struct nfs_inode *nfsi, int may_wait)
+static int wait_on_commit(struct nfs_mds_commit_info *cinfo)
 {
-	int ret;
+	return wait_on_atomic_t(&cinfo->rpcs_out,
+			nfs_wait_atomic_killable, TASK_KILLABLE);
+}
 
-	if (!test_and_set_bit(NFS_INO_COMMIT, &nfsi->flags))
-		return 1;
-	if (!may_wait)
-		return 0;
-	ret = out_of_line_wait_on_bit_lock(&nfsi->flags,
-				NFS_INO_COMMIT,
-				nfs_wait_bit_killable,
-				TASK_KILLABLE);
-	return (ret < 0) ? ret : 1;
+static void nfs_commit_begin(struct nfs_mds_commit_info *cinfo)
+{
+	atomic_inc(&cinfo->rpcs_out);
 }
 
-static void nfs_commit_clear_lock(struct nfs_inode *nfsi)
+static void nfs_commit_end(struct nfs_mds_commit_info *cinfo)
 {
-	clear_bit(NFS_INO_COMMIT, &nfsi->flags);
-	smp_mb__after_atomic();
-	wake_up_bit(&nfsi->flags, NFS_INO_COMMIT);
+	if (atomic_dec_and_test(&cinfo->rpcs_out))
+		wake_up_atomic_t(&cinfo->rpcs_out);
 }
 
 void nfs_commitdata_release(struct nfs_commit_data *data)
@@ -1693,6 +1704,13 @@ void nfs_retry_commit(struct list_head *page_list,
 }
 EXPORT_SYMBOL_GPL(nfs_retry_commit);
 
+static void
+nfs_commit_resched_write(struct nfs_commit_info *cinfo,
+		struct nfs_page *req)
+{
+	__set_page_dirty_nobuffers(req->wb_page);
+}
+
 /*
  * Commit dirty pages
  */
@@ -1714,7 +1732,6 @@ nfs_commit_list(struct inode *inode, struct list_head *head, int how,
 				   data->mds_ops, how, 0);
  out_bad:
 	nfs_retry_commit(head, NULL, cinfo, 0);
-	cinfo->completion_ops->error_cleanup(NFS_I(inode));
 	return -ENOMEM;
 }
 
@@ -1776,8 +1793,7 @@ static void nfs_commit_release_pages(struct nfs_commit_data *data)
 		clear_bdi_congested(&nfss->backing_dev_info, BLK_RW_ASYNC);
 
 	nfs_init_cinfo(&cinfo, data->inode, data->dreq);
-	if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
-		nfs_commit_clear_lock(NFS_I(data->inode));
+	nfs_commit_end(cinfo.mds);
 }
 
 static void nfs_commit_release(void *calldata)
@@ -1796,7 +1812,7 @@ static const struct rpc_call_ops nfs_commit_ops = {
 
 static const struct nfs_commit_completion_ops nfs_commit_completion_ops = {
 	.completion = nfs_commit_release_pages,
-	.error_cleanup = nfs_commit_clear_lock,
+	.resched_write = nfs_commit_resched_write,
 };
 
 int nfs_generic_commit_list(struct inode *inode, struct list_head *head,
@@ -1815,30 +1831,25 @@ int nfs_commit_inode(struct inode *inode, int how)
 	LIST_HEAD(head);
 	struct nfs_commit_info cinfo;
 	int may_wait = how & FLUSH_SYNC;
+	int error = 0;
 	int res;
 
-	res = nfs_commit_set_lock(NFS_I(inode), may_wait);
-	if (res <= 0)
-		goto out_mark_dirty;
 	nfs_init_cinfo_from_inode(&cinfo, inode);
+	nfs_commit_begin(cinfo.mds);
 	res = nfs_scan_commit(inode, &head, &cinfo);
-	if (res) {
-		int error;
-
+	if (res)
 		error = nfs_generic_commit_list(inode, &head, how, &cinfo);
-		if (error < 0)
-			return error;
-		if (!may_wait)
-			goto out_mark_dirty;
-		error = wait_on_bit_action(&NFS_I(inode)->flags,
-				NFS_INO_COMMIT,
-				nfs_wait_bit_killable,
-				TASK_KILLABLE);
-		if (error < 0)
-			return error;
-	} else
-		nfs_commit_clear_lock(NFS_I(inode));
+	nfs_commit_end(cinfo.mds);
+	if (error < 0)
+		goto out_error;
+	if (!may_wait)
+		goto out_mark_dirty;
+	error = wait_on_commit(cinfo.mds);
+	if (error < 0)
+		return error;
 	return res;
+out_error:
+	res = error;
 	/* Note: If we exit without ensuring that the commit is complete,
 	 * we must mark the inode as dirty. Otherwise, future calls to
 	 * sync_inode() with the WB_SYNC_ALL flag set will fail to ensure
@@ -1848,6 +1859,7 @@ out_mark_dirty:
 	__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
 	return res;
 }
+EXPORT_SYMBOL_GPL(nfs_commit_inode);
 
 int nfs_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index b88fc46cfbb8..9eee972863a7 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -216,7 +216,6 @@ struct nfs_inode {
 #define NFS_INO_FLUSHING	(4)		/* inode is flushing out data */
 #define NFS_INO_FSCACHE		(5)		/* inode can be cached by FS-Cache */
 #define NFS_INO_FSCACHE_LOCK	(6)		/* FS-Cache cookie management lock */
-#define NFS_INO_COMMIT		(7)		/* inode is committing unstable writes */
 #define NFS_INO_LAYOUTCOMMIT	(9)		/* layoutcommit required */
 #define NFS_INO_LAYOUTCOMMITTING (10)		/* layoutcommit inflight */
 #define NFS_INO_LAYOUTSTATS	(11)		/* layoutstats inflight */
diff --git a/include/linux/nfs_xdr.h b/include/linux/nfs_xdr.h
index 7b30ac0c7def..791098a08a87 100644
--- a/include/linux/nfs_xdr.h
+++ b/include/linux/nfs_xdr.h
@@ -1421,11 +1421,12 @@ struct nfs_mds_commit_info {
 	struct list_head	list;
 };
 
+struct nfs_commit_info;
 struct nfs_commit_data;
 struct nfs_inode;
 struct nfs_commit_completion_ops {
-	void (*error_cleanup) (struct nfs_inode *nfsi);
 	void (*completion) (struct nfs_commit_data *data);
+	void (*resched_write) (struct nfs_commit_info *, struct nfs_page *);
 };
 
 struct nfs_commit_info {
@@ -1462,6 +1463,7 @@ struct nfs_pgio_completion_ops {
 	void	(*error_cleanup)(struct list_head *head);
 	void	(*init_hdr)(struct nfs_pgio_header *hdr);
 	void	(*completion)(struct nfs_pgio_header *hdr);
+	void	(*reschedule_io)(struct nfs_pgio_header *hdr);
 };
 
 struct nfs_unlinkdata {