Merge tag 'nfs-for-4.5-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

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

  Stable fixes:
   - Fix a regression in the SunRPC socket polling code
   - Fix the attribute cache revalidation code
   - Fix race in __update_open_stateid()
   - Fix an lo->plh_block_lgets imbalance in layoutreturn
   - Fix an Oopsable typo in ff_mirror_match_fh()

  Features:
   - pNFS layout recall performance improvements.
   - pNFS/flexfiles: Support server-supplied layoutstats sampling period

  Bugfixes + cleanups:
   - NFSv4: Don't perform cached access checks before we've OPENed the
     file
   - Fix starvation issues with background flushes
   - Reclaim writes should be flushed as unstable writes if there are
     already entries in the commit lists
   - Various bugfixes from Chuck to fix NFS/RDMA send queue ordering
     problems
   - Ensure that we propagate fatal layoutget errors back to the
     application
   - Fixes for sundry flexfiles layoutstats bugs
   - Fix files/flexfiles to not cache invalidated layouts in the DS
     commit buckets"

* tag 'nfs-for-4.5-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (68 commits)
  NFS: Fix a compile warning about unused variable in nfs_generic_pg_pgios()
  NFSv4: Fix a compile warning about no prototype for nfs4_ioctl()
  NFS: Use wait_on_atomic_t() for unlock after readahead
  SUNRPC: Fixup socket wait for memory
  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
  SUNRPC: Fix a missing break in rpc_anyaddr()
  pNFS/flexfiles: Fix an Oopsable typo in ff_mirror_match_fh()
  NFS: Fix attribute cache revalidation
  NFS: Ensure we revalidate attributes before using execute_ok()
  ...

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;

fs/nfs/dir.c

@@ -2431,6 +2431,20 @@ int nfs_may_open(struct inode *inode, struct rpc_cred *cred, int openflags)
 }
 EXPORT_SYMBOL_GPL(nfs_may_open);
 
+static int nfs_execute_ok(struct inode *inode, int mask)
+{
+	struct nfs_server *server = NFS_SERVER(inode);
+	int ret;
+
+	if (mask & MAY_NOT_BLOCK)
+		ret = nfs_revalidate_inode_rcu(server, inode);
+	else
+		ret = nfs_revalidate_inode(server, inode);
+	if (ret == 0 && !execute_ok(inode))
+		ret = -EACCES;
+	return ret;
+}
+
 int nfs_permission(struct inode *inode, int mask)
 {
 	struct rpc_cred *cred;
@@ -2448,6 +2462,9 @@ int nfs_permission(struct inode *inode, int mask)
 		case S_IFLNK:
 			goto out;
 		case S_IFREG:
+			if ((mask & MAY_OPEN) &&
+			   nfs_server_capable(inode, NFS_CAP_ATOMIC_OPEN))
+				return 0;
 			break;
 		case S_IFDIR:
 			/*
@@ -2480,8 +2497,8 @@ force_lookup:
 			res = PTR_ERR(cred);
 	}
 out:
-	if (!res && (mask & MAY_EXEC) && !execute_ok(inode))
-		res = -EACCES;
+	if (!res && (mask & MAY_EXEC))
+		res = nfs_execute_ok(inode, mask);
 
 	dfprintk(VFS, "NFS: permission(%s/%lu), mask=0x%x, res=%d\n",
 		inode->i_sb->s_id, inode->i_ino, mask, res);

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)
 {
@@ -670,6 +664,10 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 
 	req = nfs_list_entry(reqs.next);
 	nfs_direct_setup_mirroring(dreq, &desc, req);
+	if (desc.pg_error < 0) {
+		list_splice_init(&reqs, &failed);
+		goto out_failed;
+	}
 
 	list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
 		if (!nfs_pageio_add_request(&desc, req)) {
@@ -677,13 +675,17 @@ static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 			nfs_list_add_request(req, &failed);
 			spin_lock(cinfo.lock);
 			dreq->flags = 0;
-			dreq->error = -EIO;
+			if (desc.pg_error < 0)
+				dreq->error = desc.pg_error;
+			else
+				dreq->error = -EIO;
 			spin_unlock(cinfo.lock);
 		}
 		nfs_release_request(req);
 	}
 	nfs_pageio_complete(&desc);
 
+out_failed:
 	while (!list_empty(&failed)) {
 		req = nfs_list_entry(failed.next);
 		nfs_list_remove_request(req);
@@ -727,14 +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)
@@ -839,10 +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,
 };
 
 
@@ -900,6 +923,11 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
 			}
 
 			nfs_direct_setup_mirroring(dreq, &desc, req);
+			if (desc.pg_error < 0) {
+				nfs_free_request(req);
+				result = desc.pg_error;
+				break;
+			}
 
 			nfs_lock_request(req);
 			req->wb_index = pos >> PAGE_SHIFT;

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;
 	}
@@ -545,7 +545,7 @@ static int nfs_launder_page(struct page *page)
 		inode->i_ino, (long long)page_offset(page));
 
 	nfs_fscache_wait_on_page_write(nfsi, page);
-	return nfs_wb_page(inode, page);
+	return nfs_wb_launder_page(inode, page);
 }
 
 static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file,
@@ -756,7 +756,7 @@ do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
 
 	l_ctx = nfs_get_lock_context(nfs_file_open_context(filp));
 	if (!IS_ERR(l_ctx)) {
-		status = nfs_iocounter_wait(&l_ctx->io_count);
+		status = nfs_iocounter_wait(l_ctx);
 		nfs_put_lock_context(l_ctx);
 		if (status < 0)
 			return status;

fs/nfs/filelayout/filelayout.c

@@ -202,6 +202,7 @@ static int filelayout_async_handle_error(struct rpc_task *task,
 			task->tk_status);
 		nfs4_mark_deviceid_unavailable(devid);
 		pnfs_error_mark_layout_for_return(inode, lseg);
+		pnfs_set_lo_fail(lseg);
 		rpc_wake_up(&tbl->slot_tbl_waitq);
 		/* fall through */
 	default:
@@ -883,13 +884,19 @@ static void
 filelayout_pg_init_read(struct nfs_pageio_descriptor *pgio,
 			struct nfs_page *req)
 {
-	if (!pgio->pg_lseg)
+	if (!pgio->pg_lseg) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 					   req->wb_context,
 					   0,
 					   NFS4_MAX_UINT64,
 					   IOMODE_READ,
 					   GFP_KERNEL);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
+	}
 	/* If no lseg, fall back to read through mds */
 	if (pgio->pg_lseg == NULL)
 		nfs_pageio_reset_read_mds(pgio);
@@ -902,13 +909,20 @@ filelayout_pg_init_write(struct nfs_pageio_descriptor *pgio,
 	struct nfs_commit_info cinfo;
 	int status;
 
-	if (!pgio->pg_lseg)
+	if (!pgio->pg_lseg) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 					   req->wb_context,
 					   0,
 					   NFS4_MAX_UINT64,
 					   IOMODE_RW,
 					   GFP_NOFS);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
+	}
+
 	/* If no lseg, fall back to write through mds */
 	if (pgio->pg_lseg == NULL)
 		goto out_mds;

fs/nfs/flexfilelayout/flexfilelayout.c

@@ -145,7 +145,7 @@ static bool ff_mirror_match_fh(const struct nfs4_ff_layout_mirror *m1,
 		return false;
 	for (i = 0; i < m1->fh_versions_cnt; i++) {
 		bool found_fh = false;
-		for (j = 0; j < m2->fh_versions_cnt; i++) {
+		for (j = 0; j < m2->fh_versions_cnt; j++) {
 			if (nfs_compare_fh(&m1->fh_versions[i],
 					&m2->fh_versions[j]) == 0) {
 				found_fh = true;
@@ -505,9 +505,17 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
 	}
 
 	p = xdr_inline_decode(&stream, 4);
-	if (p)
-		fls->flags = be32_to_cpup(p);
+	if (!p)
+		goto out_sort_mirrors;
+	fls->flags = be32_to_cpup(p);
+
+	p = xdr_inline_decode(&stream, 4);
+	if (!p)
+		goto out_sort_mirrors;
+	for (i=0; i < fls->mirror_array_cnt; i++)
+		fls->mirror_array[i]->report_interval = be32_to_cpup(p);
 
+out_sort_mirrors:
 	ff_layout_sort_mirrors(fls);
 	rc = ff_layout_check_layout(lgr);
 	if (rc)
@@ -603,7 +611,9 @@ nfs4_ff_layoutstat_start_io(struct nfs4_ff_layout_mirror *mirror,
 		mirror->start_time = now;
 	if (ktime_equal(mirror->last_report_time, notime))
 		mirror->last_report_time = now;
-	if (layoutstats_timer != 0)
+	if (mirror->report_interval != 0)
+		report_interval = (s64)mirror->report_interval * 1000LL;
+	else if (layoutstats_timer != 0)
 		report_interval = (s64)layoutstats_timer * 1000LL;
 	if (ktime_to_ms(ktime_sub(now, mirror->last_report_time)) >=
 			report_interval) {
@@ -785,13 +795,19 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio,
 	int ds_idx;
 
 	/* Use full layout for now */
-	if (!pgio->pg_lseg)
+	if (!pgio->pg_lseg) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 						   req->wb_context,
 						   0,
 						   NFS4_MAX_UINT64,
 						   IOMODE_READ,
 						   GFP_KERNEL);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
+	}
 	/* If no lseg, fall back to read through mds */
 	if (pgio->pg_lseg == NULL)
 		goto out_mds;
@@ -825,13 +841,19 @@ ff_layout_pg_init_write(struct nfs_pageio_descriptor *pgio,
 	int i;
 	int status;
 
-	if (!pgio->pg_lseg)
+	if (!pgio->pg_lseg) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 						   req->wb_context,
 						   0,
 						   NFS4_MAX_UINT64,
 						   IOMODE_RW,
 						   GFP_NOFS);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
+	}
 	/* If no lseg, fall back to write through mds */
 	if (pgio->pg_lseg == NULL)
 		goto out_mds;
@@ -867,18 +889,25 @@ static unsigned int
 ff_layout_pg_get_mirror_count_write(struct nfs_pageio_descriptor *pgio,
 				    struct nfs_page *req)
 {
-	if (!pgio->pg_lseg)
+	if (!pgio->pg_lseg) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 						   req->wb_context,
 						   0,
 						   NFS4_MAX_UINT64,
 						   IOMODE_RW,
 						   GFP_NOFS);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			goto out;
+		}
+	}
 	if (pgio->pg_lseg)
 		return FF_LAYOUT_MIRROR_COUNT(pgio->pg_lseg);
 
 	/* no lseg means that pnfs is not in use, so no mirroring here */
 	nfs_pageio_reset_write_mds(pgio);
+out:
 	return 1;
 }
 
@@ -912,18 +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;
 	}
 
@@ -1101,7 +1119,7 @@ static int ff_layout_async_handle_error_v3(struct rpc_task *task,
 	return -NFS4ERR_RESET_TO_PNFS;
 out_retry:
 	task->tk_status = 0;
-	rpc_restart_call(task);
+	rpc_restart_call_prepare(task);
 	rpc_delay(task, NFS_JUKEBOX_RETRY_TIME);
 	return -EAGAIN;
 }
@@ -1159,6 +1177,14 @@ static void ff_layout_io_track_ds_error(struct pnfs_layout_segment *lseg,
 		}
 	}
 
+	switch (status) {
+	case NFS4ERR_DELAY:
+	case NFS4ERR_GRACE:
+		return;
+	default:
+		break;
+	}
+
 	mirror = FF_LAYOUT_COMP(lseg, idx);
 	err = ff_layout_track_ds_error(FF_LAYOUT_FROM_HDR(lseg->pls_layout),
 				       mirror, offset, length, status, opnum,
@@ -1242,14 +1268,31 @@ ff_layout_reset_to_mds(struct pnfs_layout_segment *lseg, int idx)
 	return ff_layout_test_devid_unavailable(node);
 }
 
-static int ff_layout_read_prepare_common(struct rpc_task *task,
-					 struct nfs_pgio_header *hdr)
+static void ff_layout_read_record_layoutstats_start(struct rpc_task *task,
+		struct nfs_pgio_header *hdr)
 {
+	if (test_and_set_bit(NFS_IOHDR_STAT, &hdr->flags))
+		return;
 	nfs4_ff_layout_stat_io_start_read(hdr->inode,
 			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
 			hdr->args.count,
 			task->tk_start);
+}
+
+static void ff_layout_read_record_layoutstats_done(struct rpc_task *task,
+		struct nfs_pgio_header *hdr)
+{
+	if (!test_and_clear_bit(NFS_IOHDR_STAT, &hdr->flags))
+		return;
+	nfs4_ff_layout_stat_io_end_read(task,
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count,
+			hdr->res.count);
+}
 
+static int ff_layout_read_prepare_common(struct rpc_task *task,
+					 struct nfs_pgio_header *hdr)
+{
 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) {
 		rpc_exit(task, -EIO);
 		return -EIO;
@@ -1265,6 +1308,7 @@ static int ff_layout_read_prepare_common(struct rpc_task *task,
 	}
 	hdr->pgio_done_cb = ff_layout_read_done_cb;
 
+	ff_layout_read_record_layoutstats_start(task, hdr);
 	return 0;
 }
 
@@ -1323,10 +1367,6 @@ static void ff_layout_read_call_done(struct rpc_task *task, void *data)
 
 	dprintk("--> %s task->tk_status %d\n", __func__, task->tk_status);
 
-	nfs4_ff_layout_stat_io_end_read(task,
-			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
-			hdr->args.count, hdr->res.count);
-
 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags) &&
 	    task->tk_status == 0) {
 		nfs4_sequence_done(task, &hdr->res.seq_res);
@@ -1341,10 +1381,20 @@ static void ff_layout_read_count_stats(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
+	ff_layout_read_record_layoutstats_done(task, hdr);
 	rpc_count_iostats_metrics(task,
 	    &NFS_CLIENT(hdr->inode)->cl_metrics[NFSPROC4_CLNT_READ]);
 }
 
+static void ff_layout_read_release(void *data)
+{
+	struct nfs_pgio_header *hdr = data;
+
+	ff_layout_read_record_layoutstats_done(&hdr->task, hdr);
+	pnfs_generic_rw_release(data);
+}
+
+
 static int ff_layout_write_done_cb(struct rpc_task *task,
 				struct nfs_pgio_header *hdr)
 {
@@ -1362,15 +1412,12 @@ static int ff_layout_write_done_cb(struct rpc_task *task,
 
 	switch (err) {
 	case -NFS4ERR_RESET_TO_PNFS:
-		pnfs_set_retry_layoutget(hdr->lseg->pls_layout);
 		ff_layout_reset_write(hdr, true);
 		return task->tk_status;
 	case -NFS4ERR_RESET_TO_MDS:
-		pnfs_clear_retry_layoutget(hdr->lseg->pls_layout);
 		ff_layout_reset_write(hdr, false);
 		return task->tk_status;
 	case -EAGAIN:
-		rpc_restart_call_prepare(task);
 		return -EAGAIN;
 	}
 
@@ -1402,11 +1449,9 @@ static int ff_layout_commit_done_cb(struct rpc_task *task,
 
 	switch (err) {
 	case -NFS4ERR_RESET_TO_PNFS:
-		pnfs_set_retry_layoutget(data->lseg->pls_layout);
 		pnfs_generic_prepare_to_resend_writes(data);
 		return -EAGAIN;
 	case -NFS4ERR_RESET_TO_MDS:
-		pnfs_clear_retry_layoutget(data->lseg->pls_layout);
 		pnfs_generic_prepare_to_resend_writes(data);
 		return -EAGAIN;
 	case -EAGAIN:
@@ -1421,14 +1466,31 @@ static int ff_layout_commit_done_cb(struct rpc_task *task,
 	return 0;
 }
 
-static int ff_layout_write_prepare_common(struct rpc_task *task,
-					  struct nfs_pgio_header *hdr)
+static void ff_layout_write_record_layoutstats_start(struct rpc_task *task,
+		struct nfs_pgio_header *hdr)
 {
+	if (test_and_set_bit(NFS_IOHDR_STAT, &hdr->flags))
+		return;
 	nfs4_ff_layout_stat_io_start_write(hdr->inode,
 			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
 			hdr->args.count,
 			task->tk_start);
+}
+
+static void ff_layout_write_record_layoutstats_done(struct rpc_task *task,
+		struct nfs_pgio_header *hdr)
+{
+	if (!test_and_clear_bit(NFS_IOHDR_STAT, &hdr->flags))
+		return;
+	nfs4_ff_layout_stat_io_end_write(task,
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count, hdr->res.count,
+			hdr->res.verf->committed);
+}
 
+static int ff_layout_write_prepare_common(struct rpc_task *task,
+					  struct nfs_pgio_header *hdr)
+{
 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) {
 		rpc_exit(task, -EIO);
 		return -EIO;
@@ -1445,6 +1507,7 @@ static int ff_layout_write_prepare_common(struct rpc_task *task,
 		return -EAGAIN;
 	}
 
+	ff_layout_write_record_layoutstats_start(task, hdr);
 	return 0;
 }
 
@@ -1480,11 +1543,6 @@ static void ff_layout_write_call_done(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
-	nfs4_ff_layout_stat_io_end_write(task,
-			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
-			hdr->args.count, hdr->res.count,
-			hdr->res.verf->committed);
-
 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags) &&
 	    task->tk_status == 0) {
 		nfs4_sequence_done(task, &hdr->res.seq_res);
@@ -1499,18 +1557,53 @@ static void ff_layout_write_count_stats(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
+	ff_layout_write_record_layoutstats_done(task, hdr);
 	rpc_count_iostats_metrics(task,
 	    &NFS_CLIENT(hdr->inode)->cl_metrics[NFSPROC4_CLNT_WRITE]);
 }
 
-static void ff_layout_commit_prepare_common(struct rpc_task *task,
+static void ff_layout_write_release(void *data)
+{
+	struct nfs_pgio_header *hdr = data;
+
+	ff_layout_write_record_layoutstats_done(&hdr->task, hdr);
+	pnfs_generic_rw_release(data);
+}
+
+static void ff_layout_commit_record_layoutstats_start(struct rpc_task *task,
 		struct nfs_commit_data *cdata)
 {
+	if (test_and_set_bit(NFS_IOHDR_STAT, &cdata->flags))
+		return;
 	nfs4_ff_layout_stat_io_start_write(cdata->inode,
 			FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index),
 			0, task->tk_start);
 }
 
+static void ff_layout_commit_record_layoutstats_done(struct rpc_task *task,
+		struct nfs_commit_data *cdata)
+{
+	struct nfs_page *req;
+	__u64 count = 0;
+
+	if (!test_and_clear_bit(NFS_IOHDR_STAT, &cdata->flags))
+		return;
+
+	if (task->tk_status == 0) {
+		list_for_each_entry(req, &cdata->pages, wb_list)
+			count += req->wb_bytes;
+	}
+	nfs4_ff_layout_stat_io_end_write(task,
+			FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index),
+			count, count, NFS_FILE_SYNC);
+}
+
+static void ff_layout_commit_prepare_common(struct rpc_task *task,
+		struct nfs_commit_data *cdata)
+{
+	ff_layout_commit_record_layoutstats_start(task, cdata);
+}
+
 static void ff_layout_commit_prepare_v3(struct rpc_task *task, void *data)
 {
 	ff_layout_commit_prepare_common(task, data);
@@ -1531,19 +1624,6 @@ static void ff_layout_commit_prepare_v4(struct rpc_task *task, void *data)
 
 static void ff_layout_commit_done(struct rpc_task *task, void *data)
 {
-	struct nfs_commit_data *cdata = data;
-	struct nfs_page *req;
-	__u64 count = 0;
-
-	if (task->tk_status == 0) {
-		list_for_each_entry(req, &cdata->pages, wb_list)
-			count += req->wb_bytes;
-	}
-
-	nfs4_ff_layout_stat_io_end_write(task,
-			FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index),
-			count, count, NFS_FILE_SYNC);
-
 	pnfs_generic_write_commit_done(task, data);
 }
 
@@ -1551,50 +1631,59 @@ static void ff_layout_commit_count_stats(struct rpc_task *task, void *data)
 {
 	struct nfs_commit_data *cdata = data;
 
+	ff_layout_commit_record_layoutstats_done(task, cdata);
 	rpc_count_iostats_metrics(task,
 	    &NFS_CLIENT(cdata->inode)->cl_metrics[NFSPROC4_CLNT_COMMIT]);
 }
 
+static void ff_layout_commit_release(void *data)
+{
+	struct nfs_commit_data *cdata = data;
+
+	ff_layout_commit_record_layoutstats_done(&cdata->task, cdata);
+	pnfs_generic_commit_release(data);
+}
+
 static const struct rpc_call_ops ff_layout_read_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_read_prepare_v3,
 	.rpc_call_done = ff_layout_read_call_done,
 	.rpc_count_stats = ff_layout_read_count_stats,
-	.rpc_release = pnfs_generic_rw_release,
+	.rpc_release = ff_layout_read_release,
 };
 
 static const struct rpc_call_ops ff_layout_read_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_read_prepare_v4,
 	.rpc_call_done = ff_layout_read_call_done,
 	.rpc_count_stats = ff_layout_read_count_stats,
-	.rpc_release = pnfs_generic_rw_release,
+	.rpc_release = ff_layout_read_release,
 };
 
 static const struct rpc_call_ops ff_layout_write_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_write_prepare_v3,
 	.rpc_call_done = ff_layout_write_call_done,
 	.rpc_count_stats = ff_layout_write_count_stats,
-	.rpc_release = pnfs_generic_rw_release,
+	.rpc_release = ff_layout_write_release,
 };
 
 static const struct rpc_call_ops ff_layout_write_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_write_prepare_v4,
 	.rpc_call_done = ff_layout_write_call_done,
 	.rpc_count_stats = ff_layout_write_count_stats,
-	.rpc_release = pnfs_generic_rw_release,
+	.rpc_release = ff_layout_write_release,
 };
 
 static const struct rpc_call_ops ff_layout_commit_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_commit_prepare_v3,
 	.rpc_call_done = ff_layout_commit_done,
 	.rpc_count_stats = ff_layout_commit_count_stats,
-	.rpc_release = pnfs_generic_commit_release,
+	.rpc_release = ff_layout_commit_release,
 };
 
 static const struct rpc_call_ops ff_layout_commit_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_commit_prepare_v4,
 	.rpc_call_done = ff_layout_commit_done,
 	.rpc_count_stats = ff_layout_commit_count_stats,
-	.rpc_release = pnfs_generic_commit_release,
+	.rpc_release = ff_layout_commit_release,
 };
 
 static enum pnfs_try_status

fs/nfs/flexfilelayout/flexfilelayout.h

@@ -85,6 +85,7 @@ struct nfs4_ff_layout_mirror {
 	struct nfs4_ff_layoutstat	write_stat;
 	ktime_t				start_time;
 	ktime_t				last_report_time;
+	u32				report_interval;
 };
 
 struct nfs4_ff_layout_segment {

fs/nfs/flexfilelayout/flexfilelayoutdev.c

@@ -429,22 +429,14 @@ nfs4_ff_layout_prepare_ds(struct pnfs_layout_segment *lseg, u32 ds_idx,
 					 mirror, lseg->pls_range.offset,
 					 lseg->pls_range.length, NFS4ERR_NXIO,
 					 OP_ILLEGAL, GFP_NOIO);
-		if (fail_return) {
-			pnfs_error_mark_layout_for_return(ino, lseg);
-			if (ff_layout_has_available_ds(lseg))
-				pnfs_set_retry_layoutget(lseg->pls_layout);
-			else
-				pnfs_clear_retry_layoutget(lseg->pls_layout);
-
-		} else {
+		if (!fail_return) {
 			if (ff_layout_has_available_ds(lseg))
 				set_bit(NFS_LAYOUT_RETURN_BEFORE_CLOSE,
 					&lseg->pls_layout->plh_flags);
-			else {
+			else
 				pnfs_error_mark_layout_for_return(ino, lseg);
-				pnfs_clear_retry_layoutget(lseg->pls_layout);
-			}
-		}
+		} else
+			pnfs_error_mark_layout_for_return(ino, lseg);
 	}
 out_update_creds:
 	if (ff_layout_update_mirror_cred(mirror, ds))

fs/nfs/inode.c

@@ -71,19 +71,25 @@ nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
 	return nfs_fileid_to_ino_t(fattr->fileid);
 }
 
-/**
- * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
- * @word: long word containing the bit lock
- */
-int nfs_wait_bit_killable(struct wait_bit_key *key, int mode)
+static int nfs_wait_killable(int mode)
 {
 	freezable_schedule_unsafe();
 	if (signal_pending_state(mode, current))
 		return -ERESTARTSYS;
 	return 0;
 }
+
+int nfs_wait_bit_killable(struct wait_bit_key *key, int mode)
+{
+	return nfs_wait_killable(mode);
+}
 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable);
 
+int nfs_wait_atomic_killable(atomic_t *p)
+{
+	return nfs_wait_killable(TASK_KILLABLE);
+}
+
 /**
  * nfs_compat_user_ino64 - returns the user-visible inode number
  * @fileid: 64-bit fileid
@@ -700,7 +706,7 @@ static void nfs_init_lock_context(struct nfs_lock_context *l_ctx)
 	l_ctx->lockowner.l_owner = current->files;
 	l_ctx->lockowner.l_pid = current->tgid;
 	INIT_LIST_HEAD(&l_ctx->list);
-	nfs_iocounter_init(&l_ctx->io_count);
+	atomic_set(&l_ctx->io_count, 0);
 }
 
 static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context *ctx)
@@ -913,6 +919,12 @@ void nfs_file_clear_open_context(struct file *filp)
 	if (ctx) {
 		struct inode *inode = d_inode(ctx->dentry);
 
+		/*
+		 * We fatal error on write before. Try to writeback
+		 * every page again.
+		 */
+		if (ctx->error < 0)
+			invalidate_inode_pages2(inode->i_mapping);
 		filp->private_data = NULL;
 		spin_lock(&inode->i_lock);
 		list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
@@ -1663,6 +1675,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 	unsigned long invalid = 0;
 	unsigned long now = jiffies;
 	unsigned long save_cache_validity;
+	bool cache_revalidated = true;
 
 	dfprintk(VFS, "NFS: %s(%s/%lu fh_crc=0x%08x ct=%d info=0x%x)\n",
 			__func__, inode->i_sb->s_id, inode->i_ino,
@@ -1724,22 +1737,28 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 				nfs_force_lookup_revalidate(inode);
 			inode->i_version = fattr->change_attr;
 		}
-	} else
+	} else {
 		nfsi->cache_validity |= save_cache_validity;
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_MTIME) {
 		memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
-	} else if (server->caps & NFS_CAP_MTIME)
+	} else if (server->caps & NFS_CAP_MTIME) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_CTIME) {
 		memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
-	} else if (server->caps & NFS_CAP_CTIME)
+	} else if (server->caps & NFS_CAP_CTIME) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	/* Check if our cached file size is stale */
 	if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
@@ -1759,19 +1778,23 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 					(long long)cur_isize,
 					(long long)new_isize);
 		}
-	} else
+	} else {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_REVAL_PAGECACHE
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 
 	if (fattr->valid & NFS_ATTR_FATTR_ATIME)
 		memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
-	else if (server->caps & NFS_CAP_ATIME)
+	else if (server->caps & NFS_CAP_ATIME) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATIME
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_MODE) {
 		if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) {
@@ -1780,36 +1803,42 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 			inode->i_mode = newmode;
 			invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
 		}
-	} else if (server->caps & NFS_CAP_MODE)
+	} else if (server->caps & NFS_CAP_MODE) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_INVALID_ACCESS
 				| NFS_INO_INVALID_ACL
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_OWNER) {
 		if (!uid_eq(inode->i_uid, fattr->uid)) {
 			invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
 			inode->i_uid = fattr->uid;
 		}
-	} else if (server->caps & NFS_CAP_OWNER)
+	} else if (server->caps & NFS_CAP_OWNER) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_INVALID_ACCESS
 				| NFS_INO_INVALID_ACL
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_GROUP) {
 		if (!gid_eq(inode->i_gid, fattr->gid)) {
 			invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
 			inode->i_gid = fattr->gid;
 		}
-	} else if (server->caps & NFS_CAP_OWNER_GROUP)
+	} else if (server->caps & NFS_CAP_OWNER_GROUP) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_INVALID_ACCESS
 				| NFS_INO_INVALID_ACL
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_NLINK) {
 		if (inode->i_nlink != fattr->nlink) {
@@ -1818,19 +1847,22 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 				invalid |= NFS_INO_INVALID_DATA;
 			set_nlink(inode, fattr->nlink);
 		}
-	} else if (server->caps & NFS_CAP_NLINK)
+	} else if (server->caps & NFS_CAP_NLINK) {
 		nfsi->cache_validity |= save_cache_validity &
 				(NFS_INO_INVALID_ATTR
 				| NFS_INO_REVAL_FORCED);
+		cache_revalidated = false;
+	}
 
 	if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
 		/*
 		 * report the blocks in 512byte units
 		 */
 		inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
- 	}
-	if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
+	} else if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
 		inode->i_blocks = fattr->du.nfs2.blocks;
+	else
+		cache_revalidated = false;
 
 	/* Update attrtimeo value if we're out of the unstable period */
 	if (invalid & NFS_INO_INVALID_ATTR) {
@@ -1840,9 +1872,13 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 		/* Set barrier to be more recent than all outstanding updates */
 		nfsi->attr_gencount = nfs_inc_attr_generation_counter();
 	} else {
-		if (!time_in_range_open(now, nfsi->attrtimeo_timestamp, nfsi->attrtimeo_timestamp + nfsi->attrtimeo)) {
-			if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
-				nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
+		if (cache_revalidated) {
+			if (!time_in_range_open(now, nfsi->attrtimeo_timestamp,
+				nfsi->attrtimeo_timestamp + nfsi->attrtimeo)) {
+				nfsi->attrtimeo <<= 1;
+				if (nfsi->attrtimeo > NFS_MAXATTRTIMEO(inode))
+					nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
+			}
 			nfsi->attrtimeo_timestamp = now;
 		}
 		/* Set the barrier to be more recent than this fattr */
@@ -1851,7 +1887,7 @@ static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
 	}
 
 	/* Don't declare attrcache up to date if there were no attrs! */
-	if (fattr->valid != 0)
+	if (cache_revalidated)
 		invalid &= ~NFS_INO_INVALID_ATTR;
 
 	/* Don't invalidate the data if we were to blame */

fs/nfs/internal.h

@@ -238,7 +238,7 @@ extern void nfs_pgheader_init(struct nfs_pageio_descriptor *desc,
 			      struct nfs_pgio_header *hdr,
 			      void (*release)(struct nfs_pgio_header *hdr));
 void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos);
-int nfs_iocounter_wait(struct nfs_io_counter *c);
+int nfs_iocounter_wait(struct nfs_lock_context *l_ctx);
 
 extern const struct nfs_pageio_ops nfs_pgio_rw_ops;
 struct nfs_pgio_header *nfs_pgio_header_alloc(const struct nfs_rw_ops *);
@@ -252,18 +252,18 @@ void nfs_free_request(struct nfs_page *req);
 struct nfs_pgio_mirror *
 nfs_pgio_current_mirror(struct nfs_pageio_descriptor *desc);
 
-static inline void nfs_iocounter_init(struct nfs_io_counter *c)
-{
-	c->flags = 0;
-	atomic_set(&c->io_count, 0);
-}
-
 static inline bool nfs_pgio_has_mirroring(struct nfs_pageio_descriptor *desc)
 {
 	WARN_ON_ONCE(desc->pg_mirror_count < 1);
 	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 *,
@@ -380,6 +380,7 @@ extern void nfs_clear_inode(struct inode *);
 extern void nfs_evict_inode(struct inode *);
 void nfs_zap_acl_cache(struct inode *inode);
 extern int nfs_wait_bit_killable(struct wait_bit_key *key, int mode);
+extern int nfs_wait_atomic_killable(atomic_t *p);
 
 /* super.c */
 extern const struct super_operations nfs_sops;
@@ -519,7 +520,6 @@ 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 *);
@@ -696,9 +696,32 @@ static inline u32 nfs_fhandle_hash(const struct nfs_fh *fh)
 {
 	return ~crc32_le(0xFFFFFFFF, &fh->data[0], fh->size);
 }
+static inline u32 nfs_stateid_hash(const nfs4_stateid *stateid)
+{
+	return ~crc32_le(0xFFFFFFFF, &stateid->other[0],
+				NFS4_STATEID_OTHER_SIZE);
+}
 #else
 static inline u32 nfs_fhandle_hash(const struct nfs_fh *fh)
 {
 	return 0;
 }
+static inline u32 nfs_stateid_hash(nfs4_stateid *stateid)
+{
+	return 0;
+}
 #endif
+
+static inline bool nfs_error_is_fatal(int err)
+{
+	switch (err) {
+	case -ERESTARTSYS:
+	case -EIO:
+	case -ENOSPC:
+	case -EROFS:
+	case -E2BIG:
+		return true;
+	default:
+		return false;
+	}
+}

fs/nfs/nfs42proc.c

@@ -204,6 +204,8 @@ static void
 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
 {
 	struct nfs42_layoutstat_data *data = calldata;
+	struct inode *inode = data->inode;
+	struct pnfs_layout_hdr *lo;
 
 	if (!nfs4_sequence_done(task, &data->res.seq_res))
 		return;
@@ -211,12 +213,35 @@ nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
 	switch (task->tk_status) {
 	case 0:
 		break;
+	case -NFS4ERR_EXPIRED:
+	case -NFS4ERR_STALE_STATEID:
+	case -NFS4ERR_OLD_STATEID:
+	case -NFS4ERR_BAD_STATEID:
+		spin_lock(&inode->i_lock);
+		lo = NFS_I(inode)->layout;
+		if (lo && nfs4_stateid_match(&data->args.stateid,
+					     &lo->plh_stateid)) {
+			LIST_HEAD(head);
+
+			/*
+			 * Mark the bad layout state as invalid, then retry
+			 * with the current stateid.
+			 */
+			set_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags);
+			pnfs_mark_matching_lsegs_invalid(lo, &head, NULL);
+			spin_unlock(&inode->i_lock);
+			pnfs_free_lseg_list(&head);
+		} else
+			spin_unlock(&inode->i_lock);
+		break;
 	case -ENOTSUPP:
 	case -EOPNOTSUPP:
-		NFS_SERVER(data->inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
+		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
 	default:
-		dprintk("%s server returns %d\n", __func__, task->tk_status);
+		break;
 	}
+
+	dprintk("%s server returns %d\n", __func__, task->tk_status);
 }
 
 static void

fs/nfs/nfs4proc.c

@@ -208,6 +208,9 @@ static const u32 nfs4_pnfs_open_bitmap[3] = {
 	| FATTR4_WORD1_TIME_METADATA
 	| FATTR4_WORD1_TIME_MODIFY,
 	FATTR4_WORD2_MDSTHRESHOLD
+#ifdef CONFIG_NFS_V4_SECURITY_LABEL
+	| FATTR4_WORD2_SECURITY_LABEL
+#endif
 };
 
 static const u32 nfs4_open_noattr_bitmap[3] = {
@@ -1385,6 +1388,7 @@ static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_s
 	 * Protect the call to nfs4_state_set_mode_locked and
 	 * serialise the stateid update
 	 */
+	spin_lock(&state->owner->so_lock);
 	write_seqlock(&state->seqlock);
 	if (deleg_stateid != NULL) {
 		nfs4_stateid_copy(&state->stateid, deleg_stateid);
@@ -1393,7 +1397,6 @@ static void __update_open_stateid(struct nfs4_state *state, nfs4_stateid *open_s
 	if (open_stateid != NULL)
 		nfs_set_open_stateid_locked(state, open_stateid, fmode);
 	write_sequnlock(&state->seqlock);
-	spin_lock(&state->owner->so_lock);
 	update_open_stateflags(state, fmode);
 	spin_unlock(&state->owner->so_lock);
 }
@@ -1598,6 +1601,7 @@ _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
 
 	if (!data->rpc_done) {
 		state = nfs4_try_open_cached(data);
+		trace_nfs4_cached_open(data->state);
 		goto out;
 	}
 
@@ -2015,6 +2019,7 @@ static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
 	}
 	return;
 unlock_no_action:
+	trace_nfs4_cached_open(data->state);
 	rcu_read_unlock();
 out_no_action:
 	task->tk_action = NULL;
@@ -2703,6 +2708,7 @@ static int _nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
 	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
 	if (status == 0 && state != NULL)
 		renew_lease(server, timestamp);
+	trace_nfs4_setattr(inode, &arg.stateid, status);
 	return status;
 }
 
@@ -2719,7 +2725,6 @@ static int nfs4_do_setattr(struct inode *inode, struct rpc_cred *cred,
 	int err;
 	do {
 		err = _nfs4_do_setattr(inode, cred, fattr, sattr, state, ilabel, olabel);
-		trace_nfs4_setattr(inode, err);
 		switch (err) {
 		case -NFS4ERR_OPENMODE:
 			if (!(sattr->ia_valid & ATTR_SIZE)) {
@@ -5048,7 +5053,6 @@ static void nfs4_init_boot_verifier(const struct nfs_client *clp,
 static int
 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
 {
-	int result;
 	size_t len;
 	char *str;
 
@@ -5076,7 +5080,7 @@ nfs4_init_nonuniform_client_string(struct nfs_client *clp)
 		return -ENOMEM;
 
 	rcu_read_lock();
-	result = scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
+	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
 			clp->cl_ipaddr,
 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
 			rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));
@@ -5089,7 +5093,6 @@ nfs4_init_nonuniform_client_string(struct nfs_client *clp)
 static int
 nfs4_init_uniquifier_client_string(struct nfs_client *clp)
 {
-	int result;
 	size_t len;
 	char *str;
 
@@ -5109,7 +5112,7 @@ nfs4_init_uniquifier_client_string(struct nfs_client *clp)
 	if (!str)
 		return -ENOMEM;
 
-	result = scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
+	scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
 			clp->rpc_ops->version, clp->cl_minorversion,
 			nfs4_client_id_uniquifier,
 			clp->cl_rpcclient->cl_nodename);
@@ -5120,7 +5123,6 @@ nfs4_init_uniquifier_client_string(struct nfs_client *clp)
 static int
 nfs4_init_uniform_client_string(struct nfs_client *clp)
 {
-	int result;
 	size_t len;
 	char *str;
 
@@ -5145,7 +5147,7 @@ nfs4_init_uniform_client_string(struct nfs_client *clp)
 	if (!str)
 		return -ENOMEM;
 
-	result = scnprintf(str, len, "Linux NFSv%u.%u %s",
+	scnprintf(str, len, "Linux NFSv%u.%u %s",
 			clp->rpc_ops->version, clp->cl_minorversion,
 			clp->cl_rpcclient->cl_nodename);
 	clp->cl_owner_id = str;
@@ -5384,6 +5386,11 @@ static int _nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, co
 	if (data == NULL)
 		return -ENOMEM;
 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1);
+
+	nfs4_state_protect(server->nfs_client,
+			NFS_SP4_MACH_CRED_CLEANUP,
+			&task_setup_data.rpc_client, &msg);
+
 	data->args.fhandle = &data->fh;
 	data->args.stateid = &data->stateid;
 	data->args.bitmask = server->cache_consistency_bitmask;
@@ -5426,7 +5433,7 @@ int nfs4_proc_delegreturn(struct inode *inode, struct rpc_cred *cred, const nfs4
 	int err;
 	do {
 		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
-		trace_nfs4_delegreturn(inode, err);
+		trace_nfs4_delegreturn(inode, stateid, err);
 		switch (err) {
 			case -NFS4ERR_STALE_STATEID:
 			case -NFS4ERR_EXPIRED:
@@ -5936,6 +5943,7 @@ static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *f
 		data->cancelled = 1;
 	rpc_put_task(task);
 	dprintk("%s: done, ret = %d!\n", __func__, ret);
+	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
 	return ret;
 }
 
@@ -5952,7 +5960,6 @@ static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request
 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
 			return 0;
 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
-		trace_nfs4_lock_reclaim(request, state, F_SETLK, err);
 		if (err != -NFS4ERR_DELAY)
 			break;
 		nfs4_handle_exception(server, err, &exception);
@@ -5979,7 +5986,6 @@ static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request
 		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
 			return 0;
 		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
-		trace_nfs4_lock_expired(request, state, F_SETLK, err);
 		switch (err) {
 		default:
 			goto out;
@@ -6087,7 +6093,6 @@ static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *
 
 	do {
 		err = _nfs4_proc_setlk(state, cmd, request);
-		trace_nfs4_set_lock(request, state, cmd, err);
 		if (err == -NFS4ERR_DENIED)
 			err = -EAGAIN;
 		err = nfs4_handle_exception(NFS_SERVER(state->inode),
@@ -6847,10 +6852,13 @@ static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
 	},
 	.allow.u.words = {
 		[0] = 1 << (OP_CLOSE) |
+		      1 << (OP_OPEN_DOWNGRADE) |
 		      1 << (OP_LOCKU) |
+		      1 << (OP_DELEGRETURN) |
 		      1 << (OP_COMMIT),
 		[1] = 1 << (OP_SECINFO - 32) |
 		      1 << (OP_SECINFO_NO_NAME - 32) |
+		      1 << (OP_LAYOUTRETURN - 32) |
 		      1 << (OP_TEST_STATEID - 32) |
 		      1 << (OP_FREE_STATEID - 32) |
 		      1 << (OP_WRITE - 32)
@@ -6915,11 +6923,19 @@ static int nfs4_sp4_select_mode(struct nfs_client *clp,
 		}
 
 		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
+		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
+		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
 		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
 			dfprintk(MOUNT, "  cleanup mode enabled\n");
 			set_bit(NFS_SP4_MACH_CRED_CLEANUP, &clp->cl_sp4_flags);
 		}
 
+		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
+			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
+			set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP,
+				&clp->cl_sp4_flags);
+		}
+
 		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
 		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
 			dfprintk(MOUNT, "  secinfo mode enabled\n");
@@ -7748,6 +7764,7 @@ 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
@@ -7758,12 +7775,12 @@ 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)
@@ -7783,6 +7800,15 @@ static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
 	switch (task->tk_status) {
 	case 0:
 		goto out;
+
+	/*
+	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
+	 * on the file. set tk_status to -ENODATA to tell upper layer to
+	 * retry go inband.
+	 */
+	case -NFS4ERR_LAYOUTUNAVAILABLE:
+		task->tk_status = -ENODATA;
+		goto out;
 	/*
 	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
 	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
@@ -7979,6 +8005,7 @@ nfs4_proc_layoutget(struct nfs4_layoutget *lgp, gfp_t gfp_flags)
 	trace_nfs4_layoutget(lgp->args.ctx,
 			&lgp->args.range,
 			&lgp->res.range,
+			&lgp->res.stateid,
 			status);
 	/* if layoutp->len is 0, nfs4_layoutget_prepare called rpc_exit */
 	if (status == 0 && lgp->res.layoutp->len)
@@ -8035,11 +8062,11 @@ 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);
 	pnfs_free_lseg_list(&freeme);
 	pnfs_put_layout_hdr(lrp->args.layout);
@@ -8071,6 +8098,10 @@ int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
 	};
 	int status = 0;
 
+	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
+			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
+			&task_setup_data.rpc_client, &msg);
+
 	dprintk("--> %s\n", __func__);
 	if (!sync) {
 		lrp->inode = nfs_igrab_and_active(lrp->args.inode);
@@ -8086,7 +8117,7 @@ int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
 		return PTR_ERR(task);
 	if (sync)
 		status = task->tk_status;
-	trace_nfs4_layoutreturn(lrp->args.inode, status);
+	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
 	dprintk("<-- %s status=%d\n", __func__, status);
 	rpc_put_task(task);
 	return status;
@@ -8234,7 +8265,7 @@ nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
 		return PTR_ERR(task);
 	if (sync)
 		status = task->tk_status;
-	trace_nfs4_layoutcommit(data->args.inode, status);
+	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
 	dprintk("%s: status %d\n", __func__, status);
 	rpc_put_task(task);
 	return status;

fs/nfs/nfs4sysctl.c

@@ -12,7 +12,7 @@
 #include "nfs4idmap.h"
 #include "callback.h"
 
-static const int nfs_set_port_min = 0;
+static const int nfs_set_port_min;
 static const int nfs_set_port_max = 65535;
 static struct ctl_table_header *nfs4_callback_sysctl_table;
 

fs/nfs/nfs4trace.c

@@ -6,6 +6,7 @@
 #include "internal.h"
 #include "nfs4session.h"
 #include "callback.h"
+#include "pnfs.h"
 
 #define CREATE_TRACE_POINTS
 #include "nfs4trace.h"

fs/nfs/nfs4trace.h

@@ -321,6 +321,7 @@ TRACE_EVENT(nfs4_sequence_done,
 			__entry->highest_slotid = res->sr_highest_slotid;
 			__entry->target_highest_slotid =
 					res->sr_target_highest_slotid;
+			__entry->status_flags = res->sr_status_flags;
 			__entry->error = res->sr_status;
 		),
 		TP_printk(
@@ -399,6 +400,10 @@ DECLARE_EVENT_CLASS(nfs4_open_event,
 			__field(u64, fileid)
 			__field(u64, dir)
 			__string(name, ctx->dentry->d_name.name)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+			__field(int, openstateid_seq)
+			__field(u32, openstateid_hash)
 		),
 
 		TP_fast_assign(
@@ -409,8 +414,22 @@ DECLARE_EVENT_CLASS(nfs4_open_event,
 			__entry->flags = flags;
 			__entry->fmode = (__force unsigned int)ctx->mode;
 			__entry->dev = ctx->dentry->d_sb->s_dev;
-			if (!IS_ERR_OR_NULL(state))
+			if (!IS_ERR_OR_NULL(state)) {
 				inode = state->inode;
+				__entry->stateid_seq =
+					be32_to_cpu(state->stateid.seqid);
+				__entry->stateid_hash =
+					nfs_stateid_hash(&state->stateid);
+				__entry->openstateid_seq =
+					be32_to_cpu(state->open_stateid.seqid);
+				__entry->openstateid_hash =
+					nfs_stateid_hash(&state->open_stateid);
+			} else {
+				__entry->stateid_seq = 0;
+				__entry->stateid_hash = 0;
+				__entry->openstateid_seq = 0;
+				__entry->openstateid_hash = 0;
+			}
 			if (inode != NULL) {
 				__entry->fileid = NFS_FILEID(inode);
 				__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
@@ -425,7 +444,8 @@ DECLARE_EVENT_CLASS(nfs4_open_event,
 		TP_printk(
 			"error=%d (%s) flags=%d (%s) fmode=%s "
 			"fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"name=%02x:%02x:%llu/%s",
+			"name=%02x:%02x:%llu/%s stateid=%d:0x%08x "
+			"openstateid=%d:0x%08x",
 			 __entry->error,
 			 show_nfsv4_errors(__entry->error),
 			 __entry->flags,
@@ -436,7 +456,9 @@ DECLARE_EVENT_CLASS(nfs4_open_event,
 			 __entry->fhandle,
 			 MAJOR(__entry->dev), MINOR(__entry->dev),
 			 (unsigned long long)__entry->dir,
-			 __get_str(name)
+			 __get_str(name),
+			 __entry->stateid_seq, __entry->stateid_hash,
+			 __entry->openstateid_seq, __entry->openstateid_hash
 		)
 );
 
@@ -452,6 +474,45 @@ DEFINE_NFS4_OPEN_EVENT(nfs4_open_reclaim);
 DEFINE_NFS4_OPEN_EVENT(nfs4_open_expired);
 DEFINE_NFS4_OPEN_EVENT(nfs4_open_file);
 
+TRACE_EVENT(nfs4_cached_open,
+		TP_PROTO(
+			const struct nfs4_state *state
+		),
+		TP_ARGS(state),
+		TP_STRUCT__entry(
+			__field(dev_t, dev)
+			__field(u32, fhandle)
+			__field(u64, fileid)
+			__field(unsigned int, fmode)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+		),
+
+		TP_fast_assign(
+			const struct inode *inode = state->inode;
+
+			__entry->dev = inode->i_sb->s_dev;
+			__entry->fileid = NFS_FILEID(inode);
+			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->fmode = (__force unsigned int)state->state;
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
+		),
+
+		TP_printk(
+			"fmode=%s fileid=%02x:%02x:%llu "
+			"fhandle=0x%08x stateid=%d:0x%08x",
+			__entry->fmode ?  show_fmode_flags(__entry->fmode) :
+					  "closed",
+			MAJOR(__entry->dev), MINOR(__entry->dev),
+			(unsigned long long)__entry->fileid,
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash
+		)
+);
+
 TRACE_EVENT(nfs4_close,
 		TP_PROTO(
 			const struct nfs4_state *state,
@@ -468,6 +529,8 @@ TRACE_EVENT(nfs4_close,
 			__field(u64, fileid)
 			__field(unsigned int, fmode)
 			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
@@ -478,18 +541,23 @@ TRACE_EVENT(nfs4_close,
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
 			__entry->fmode = (__force unsigned int)state->state;
 			__entry->error = error;
+			__entry->stateid_seq =
+				be32_to_cpu(args->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&args->stateid);
 		),
 
 		TP_printk(
 			"error=%d (%s) fmode=%s fileid=%02x:%02x:%llu "
-			"fhandle=0x%08x",
+			"fhandle=0x%08x openstateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			__entry->fmode ?  show_fmode_flags(__entry->fmode) :
 					  "closed",
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
-			__entry->fhandle
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 
@@ -523,6 +591,8 @@ DECLARE_EVENT_CLASS(nfs4_lock_event,
 			__field(dev_t, dev)
 			__field(u32, fhandle)
 			__field(u64, fileid)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
@@ -536,11 +606,16 @@ DECLARE_EVENT_CLASS(nfs4_lock_event,
 			__entry->dev = inode->i_sb->s_dev;
 			__entry->fileid = NFS_FILEID(inode);
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
 		),
 
 		TP_printk(
 			"error=%d (%s) cmd=%s:%s range=%lld:%lld "
-			"fileid=%02x:%02x:%llu fhandle=0x%08x",
+			"fileid=%02x:%02x:%llu fhandle=0x%08x "
+			"stateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			show_lock_cmd(__entry->cmd),
@@ -549,7 +624,8 @@ DECLARE_EVENT_CLASS(nfs4_lock_event,
 			(long long)__entry->end,
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
-			__entry->fhandle
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 
@@ -563,11 +639,73 @@ DECLARE_EVENT_CLASS(nfs4_lock_event,
 			), \
 			TP_ARGS(request, state, cmd, error))
 DEFINE_NFS4_LOCK_EVENT(nfs4_get_lock);
-DEFINE_NFS4_LOCK_EVENT(nfs4_set_lock);
-DEFINE_NFS4_LOCK_EVENT(nfs4_lock_reclaim);
-DEFINE_NFS4_LOCK_EVENT(nfs4_lock_expired);
 DEFINE_NFS4_LOCK_EVENT(nfs4_unlock);
 
+TRACE_EVENT(nfs4_set_lock,
+		TP_PROTO(
+			const struct file_lock *request,
+			const struct nfs4_state *state,
+			const nfs4_stateid *lockstateid,
+			int cmd,
+			int error
+		),
+
+		TP_ARGS(request, state, lockstateid, cmd, error),
+
+		TP_STRUCT__entry(
+			__field(int, error)
+			__field(int, cmd)
+			__field(char, type)
+			__field(loff_t, start)
+			__field(loff_t, end)
+			__field(dev_t, dev)
+			__field(u32, fhandle)
+			__field(u64, fileid)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+			__field(int, lockstateid_seq)
+			__field(u32, lockstateid_hash)
+		),
+
+		TP_fast_assign(
+			const struct inode *inode = state->inode;
+
+			__entry->error = error;
+			__entry->cmd = cmd;
+			__entry->type = request->fl_type;
+			__entry->start = request->fl_start;
+			__entry->end = request->fl_end;
+			__entry->dev = inode->i_sb->s_dev;
+			__entry->fileid = NFS_FILEID(inode);
+			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
+			__entry->lockstateid_seq =
+				be32_to_cpu(lockstateid->seqid);
+			__entry->lockstateid_hash =
+				nfs_stateid_hash(lockstateid);
+		),
+
+		TP_printk(
+			"error=%d (%s) cmd=%s:%s range=%lld:%lld "
+			"fileid=%02x:%02x:%llu fhandle=0x%08x "
+			"stateid=%d:0x%08x lockstateid=%d:0x%08x",
+			__entry->error,
+			show_nfsv4_errors(__entry->error),
+			show_lock_cmd(__entry->cmd),
+			show_lock_type(__entry->type),
+			(long long)__entry->start,
+			(long long)__entry->end,
+			MAJOR(__entry->dev), MINOR(__entry->dev),
+			(unsigned long long)__entry->fileid,
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash,
+			__entry->lockstateid_seq, __entry->lockstateid_hash
+		)
+);
+
 DECLARE_EVENT_CLASS(nfs4_set_delegation_event,
 		TP_PROTO(
 			const struct inode *inode,
@@ -621,20 +759,28 @@ TRACE_EVENT(nfs4_delegreturn_exit,
 			__field(dev_t, dev)
 			__field(u32, fhandle)
 			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
 			__entry->dev = res->server->s_dev;
 			__entry->fhandle = nfs_fhandle_hash(args->fhandle);
 			__entry->error = error;
+			__entry->stateid_seq =
+				be32_to_cpu(args->stateid->seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(args->stateid);
 		),
 
 		TP_printk(
-			"error=%d (%s) dev=%02x:%02x fhandle=0x%08x",
+			"error=%d (%s) dev=%02x:%02x fhandle=0x%08x "
+			"stateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			MAJOR(__entry->dev), MINOR(__entry->dev),
-			__entry->fhandle
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 
@@ -653,6 +799,8 @@ DECLARE_EVENT_CLASS(nfs4_test_stateid_event,
 			__field(dev_t, dev)
 			__field(u32, fhandle)
 			__field(u64, fileid)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
@@ -662,15 +810,21 @@ DECLARE_EVENT_CLASS(nfs4_test_stateid_event,
 			__entry->dev = inode->i_sb->s_dev;
 			__entry->fileid = NFS_FILEID(inode);
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
 		),
 
 		TP_printk(
-			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x",
+			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x "
+			"stateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
-			__entry->fhandle
+			__entry->fhandle,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 
@@ -820,7 +974,6 @@ DECLARE_EVENT_CLASS(nfs4_inode_event,
 			), \
 			TP_ARGS(inode, error))
 
-DEFINE_NFS4_INODE_EVENT(nfs4_setattr);
 DEFINE_NFS4_INODE_EVENT(nfs4_access);
 DEFINE_NFS4_INODE_EVENT(nfs4_readlink);
 DEFINE_NFS4_INODE_EVENT(nfs4_readdir);
@@ -830,8 +983,59 @@ 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);
-DEFINE_NFS4_INODE_EVENT(nfs4_delegreturn);
+
+DECLARE_EVENT_CLASS(nfs4_inode_stateid_event,
+		TP_PROTO(
+			const struct inode *inode,
+			const nfs4_stateid *stateid,
+			int error
+		),
+
+		TP_ARGS(inode, stateid, error),
+
+		TP_STRUCT__entry(
+			__field(dev_t, dev)
+			__field(u32, fhandle)
+			__field(u64, fileid)
+			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+		),
+
+		TP_fast_assign(
+			__entry->dev = inode->i_sb->s_dev;
+			__entry->fileid = NFS_FILEID(inode);
+			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->error = error;
+			__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",
+			__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
+		)
+);
+
+#define DEFINE_NFS4_INODE_STATEID_EVENT(name) \
+	DEFINE_EVENT(nfs4_inode_stateid_event, name, \
+			TP_PROTO( \
+				const struct inode *inode, \
+				const nfs4_stateid *stateid, \
+				int error \
+			), \
+			TP_ARGS(inode, stateid, error))
+
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_setattr);
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_delegreturn);
 
 DECLARE_EVENT_CLASS(nfs4_getattr_event,
 		TP_PROTO(
@@ -941,8 +1145,74 @@ 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(
@@ -1005,28 +1275,37 @@ DECLARE_EVENT_CLASS(nfs4_read_event,
 			__field(loff_t, offset)
 			__field(size_t, count)
 			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
 			const struct inode *inode = hdr->inode;
+			const struct nfs4_state *state =
+				hdr->args.context->state;
 			__entry->dev = inode->i_sb->s_dev;
 			__entry->fileid = NFS_FILEID(inode);
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
 			__entry->offset = hdr->args.offset;
 			__entry->count = hdr->args.count;
 			__entry->error = error;
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
 		),
 
 		TP_printk(
 			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"offset=%lld count=%zu",
+			"offset=%lld count=%zu stateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
 			__entry->fhandle,
 			(long long)__entry->offset,
-			__entry->count
+			__entry->count,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 #define DEFINE_NFS4_READ_EVENT(name) \
@@ -1056,28 +1335,37 @@ DECLARE_EVENT_CLASS(nfs4_write_event,
 			__field(loff_t, offset)
 			__field(size_t, count)
 			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
 		),
 
 		TP_fast_assign(
 			const struct inode *inode = hdr->inode;
+			const struct nfs4_state *state =
+				hdr->args.context->state;
 			__entry->dev = inode->i_sb->s_dev;
 			__entry->fileid = NFS_FILEID(inode);
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
 			__entry->offset = hdr->args.offset;
 			__entry->count = hdr->args.count;
 			__entry->error = error;
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
 		),
 
 		TP_printk(
 			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"offset=%lld count=%zu",
+			"offset=%lld count=%zu stateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			MAJOR(__entry->dev), MINOR(__entry->dev),
 			(unsigned long long)__entry->fileid,
 			__entry->fhandle,
 			(long long)__entry->offset,
-			__entry->count
+			__entry->count,
+			__entry->stateid_seq, __entry->stateid_hash
 		)
 );
 
@@ -1154,10 +1442,11 @@ TRACE_EVENT(nfs4_layoutget,
 			const struct nfs_open_context *ctx,
 			const struct pnfs_layout_range *args,
 			const struct pnfs_layout_range *res,
+			const nfs4_stateid *layout_stateid,
 			int error
 		),
 
-		TP_ARGS(ctx, args, res, error),
+		TP_ARGS(ctx, args, res, layout_stateid, error),
 
 		TP_STRUCT__entry(
 			__field(dev_t, dev)
@@ -1167,10 +1456,15 @@ TRACE_EVENT(nfs4_layoutget,
 			__field(u64, offset)
 			__field(u64, count)
 			__field(int, error)
+			__field(int, stateid_seq)
+			__field(u32, stateid_hash)
+			__field(int, layoutstateid_seq)
+			__field(u32, layoutstateid_hash)
 		),
 
 		TP_fast_assign(
 			const struct inode *inode = d_inode(ctx->dentry);
+			const struct nfs4_state *state = ctx->state;
 			__entry->dev = inode->i_sb->s_dev;
 			__entry->fileid = NFS_FILEID(inode);
 			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
@@ -1178,11 +1472,25 @@ TRACE_EVENT(nfs4_layoutget,
 			__entry->offset = args->offset;
 			__entry->count = args->length;
 			__entry->error = error;
+			__entry->stateid_seq =
+				be32_to_cpu(state->stateid.seqid);
+			__entry->stateid_hash =
+				nfs_stateid_hash(&state->stateid);
+			if (!error) {
+				__entry->layoutstateid_seq =
+				be32_to_cpu(layout_stateid->seqid);
+				__entry->layoutstateid_hash =
+				nfs_stateid_hash(layout_stateid);
+			} else {
+				__entry->layoutstateid_seq = 0;
+				__entry->layoutstateid_hash = 0;
+			}
 		),
 
 		TP_printk(
 			"error=%d (%s) fileid=%02x:%02x:%llu fhandle=0x%08x "
-			"iomode=%s offset=%llu count=%llu",
+			"iomode=%s offset=%llu count=%llu stateid=%d:0x%08x "
+			"layoutstateid=%d:0x%08x",
 			__entry->error,
 			show_nfsv4_errors(__entry->error),
 			MAJOR(__entry->dev), MINOR(__entry->dev),
@@ -1190,14 +1498,83 @@ TRACE_EVENT(nfs4_layoutget,
 			__entry->fhandle,
 			show_pnfs_iomode(__entry->iomode),
 			(unsigned long long)__entry->offset,
-			(unsigned long long)__entry->count
+			(unsigned long long)__entry->count,
+			__entry->stateid_seq, __entry->stateid_hash,
+			__entry->layoutstateid_seq, __entry->layoutstateid_hash
 		)
 );
 
-DEFINE_NFS4_INODE_EVENT(nfs4_layoutcommit);
-DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn);
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_layoutcommit);
+DEFINE_NFS4_INODE_STATEID_EVENT(nfs4_layoutreturn);
 DEFINE_NFS4_INODE_EVENT(nfs4_layoutreturn_on_close);
 
+#define show_pnfs_update_layout_reason(reason)				\
+	__print_symbolic(reason,					\
+		{ PNFS_UPDATE_LAYOUT_UNKNOWN, "unknown" },		\
+		{ PNFS_UPDATE_LAYOUT_NO_PNFS, "no pnfs" },		\
+		{ PNFS_UPDATE_LAYOUT_RD_ZEROLEN, "read+zerolen" },	\
+		{ PNFS_UPDATE_LAYOUT_MDSTHRESH, "mdsthresh" },		\
+		{ PNFS_UPDATE_LAYOUT_NOMEM, "nomem" },			\
+		{ PNFS_UPDATE_LAYOUT_BULK_RECALL, "bulk recall" },	\
+		{ PNFS_UPDATE_LAYOUT_IO_TEST_FAIL, "io test fail" },	\
+		{ PNFS_UPDATE_LAYOUT_FOUND_CACHED, "found cached" },	\
+		{ PNFS_UPDATE_LAYOUT_RETURN, "layoutreturn" },		\
+		{ PNFS_UPDATE_LAYOUT_BLOCKED, "layouts blocked" },	\
+		{ PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET, "sent layoutget" })
+
+TRACE_EVENT(pnfs_update_layout,
+		TP_PROTO(struct inode *inode,
+			loff_t pos,
+			u64 count,
+			enum pnfs_iomode iomode,
+			struct pnfs_layout_hdr *lo,
+			enum pnfs_update_layout_reason reason
+		),
+		TP_ARGS(inode, pos, count, iomode, lo, reason),
+		TP_STRUCT__entry(
+			__field(dev_t, dev)
+			__field(u64, fileid)
+			__field(u32, fhandle)
+			__field(loff_t, pos)
+			__field(u64, count)
+			__field(enum pnfs_iomode, iomode)
+			__field(int, layoutstateid_seq)
+			__field(u32, layoutstateid_hash)
+			__field(enum pnfs_update_layout_reason, reason)
+		),
+		TP_fast_assign(
+			__entry->dev = inode->i_sb->s_dev;
+			__entry->fileid = NFS_FILEID(inode);
+			__entry->fhandle = nfs_fhandle_hash(NFS_FH(inode));
+			__entry->pos = pos;
+			__entry->count = count;
+			__entry->iomode = iomode;
+			__entry->reason = reason;
+			if (lo != NULL) {
+				__entry->layoutstateid_seq =
+				be32_to_cpu(lo->plh_stateid.seqid);
+				__entry->layoutstateid_hash =
+				nfs_stateid_hash(&lo->plh_stateid);
+			} else {
+				__entry->layoutstateid_seq = 0;
+				__entry->layoutstateid_hash = 0;
+			}
+		),
+		TP_printk(
+			"fileid=%02x:%02x:%llu fhandle=0x%08x "
+			"iomode=%s pos=%llu count=%llu "
+			"layoutstateid=%d:0x%08x (%s)",
+			MAJOR(__entry->dev), MINOR(__entry->dev),
+			(unsigned long long)__entry->fileid,
+			__entry->fhandle,
+			show_pnfs_iomode(__entry->iomode),
+			(unsigned long long)__entry->pos,
+			(unsigned long long)__entry->count,
+			__entry->layoutstateid_seq, __entry->layoutstateid_hash,
+			show_pnfs_update_layout_reason(__entry->reason)
+		)
+);
+
 #endif /* CONFIG_NFS_V4_1 */
 
 #endif /* _TRACE_NFS4_H */

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" })
 

fs/nfs/pagelist.c

@@ -101,53 +101,18 @@ nfs_page_free(struct nfs_page *p)
 	kmem_cache_free(nfs_page_cachep, p);
 }
 
-static void
-nfs_iocounter_inc(struct nfs_io_counter *c)
-{
-	atomic_inc(&c->io_count);
-}
-
-static void
-nfs_iocounter_dec(struct nfs_io_counter *c)
-{
-	if (atomic_dec_and_test(&c->io_count)) {
-		clear_bit(NFS_IO_INPROGRESS, &c->flags);
-		smp_mb__after_atomic();
-		wake_up_bit(&c->flags, NFS_IO_INPROGRESS);
-	}
-}
-
-static int
-__nfs_iocounter_wait(struct nfs_io_counter *c)
-{
-	wait_queue_head_t *wq = bit_waitqueue(&c->flags, NFS_IO_INPROGRESS);
-	DEFINE_WAIT_BIT(q, &c->flags, NFS_IO_INPROGRESS);
-	int ret = 0;
-
-	do {
-		prepare_to_wait(wq, &q.wait, TASK_KILLABLE);
-		set_bit(NFS_IO_INPROGRESS, &c->flags);
-		if (atomic_read(&c->io_count) == 0)
-			break;
-		ret = nfs_wait_bit_killable(&q.key, TASK_KILLABLE);
-	} while (atomic_read(&c->io_count) != 0 && !ret);
-	finish_wait(wq, &q.wait);
-	return ret;
-}
-
 /**
  * nfs_iocounter_wait - wait for i/o to complete
- * @c: nfs_io_counter to use
+ * @l_ctx: nfs_lock_context with io_counter to use
  *
  * returns -ERESTARTSYS if interrupted by a fatal signal.
  * Otherwise returns 0 once the io_count hits 0.
  */
 int
-nfs_iocounter_wait(struct nfs_io_counter *c)
+nfs_iocounter_wait(struct nfs_lock_context *l_ctx)
 {
-	if (atomic_read(&c->io_count) == 0)
-		return 0;
-	return __nfs_iocounter_wait(c);
+	return wait_on_atomic_t(&l_ctx->io_count, nfs_wait_atomic_killable,
+			TASK_KILLABLE);
 }
 
 /*
@@ -370,7 +335,7 @@ nfs_create_request(struct nfs_open_context *ctx, struct page *page,
 		return ERR_CAST(l_ctx);
 	}
 	req->wb_lock_context = l_ctx;
-	nfs_iocounter_inc(&l_ctx->io_count);
+	atomic_inc(&l_ctx->io_count);
 
 	/* Initialize the request struct. Initially, we assume a
 	 * long write-back delay. This will be adjusted in
@@ -431,7 +396,8 @@ static void nfs_clear_request(struct nfs_page *req)
 		req->wb_page = NULL;
 	}
 	if (l_ctx != NULL) {
-		nfs_iocounter_dec(&l_ctx->io_count);
+		if (atomic_dec_and_test(&l_ctx->io_count))
+			wake_up_atomic_t(&l_ctx->io_count);
 		nfs_put_lock_context(l_ctx);
 		req->wb_lock_context = NULL;
 	}
@@ -664,22 +630,11 @@ EXPORT_SYMBOL_GPL(nfs_initiate_pgio);
  * @desc: IO descriptor
  * @hdr: pageio header
  */
-static int nfs_pgio_error(struct nfs_pageio_descriptor *desc,
-			  struct nfs_pgio_header *hdr)
+static void nfs_pgio_error(struct nfs_pgio_header *hdr)
 {
-	struct nfs_pgio_mirror *mirror;
-	u32 midx;
-
 	set_bit(NFS_IOHDR_REDO, &hdr->flags);
 	nfs_pgio_data_destroy(hdr);
 	hdr->completion_ops->completion(hdr);
-	/* TODO: Make sure it's right to clean up all mirrors here
-	 *       and not just hdr->pgio_mirror_idx */
-	for (midx = 0; midx < desc->pg_mirror_count; midx++) {
-		mirror = &desc->pg_mirrors[midx];
-		desc->pg_completion_ops->error_cleanup(&mirror->pg_list);
-	}
-	return -ENOMEM;
 }
 
 /**
@@ -800,8 +755,11 @@ int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
 	unsigned int pagecount, pageused;
 
 	pagecount = nfs_page_array_len(mirror->pg_base, mirror->pg_count);
-	if (!nfs_pgarray_set(&hdr->page_array, pagecount))
-		return nfs_pgio_error(desc, hdr);
+	if (!nfs_pgarray_set(&hdr->page_array, pagecount)) {
+		nfs_pgio_error(hdr);
+		desc->pg_error = -ENOMEM;
+		return desc->pg_error;
+	}
 
 	nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
 	pages = hdr->page_array.pagevec;
@@ -819,8 +777,11 @@ int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
 			*pages++ = last_page = req->wb_page;
 		}
 	}
-	if (WARN_ON_ONCE(pageused != pagecount))
-		return nfs_pgio_error(desc, hdr);
+	if (WARN_ON_ONCE(pageused != pagecount)) {
+		nfs_pgio_error(hdr);
+		desc->pg_error = -EINVAL;
+		return desc->pg_error;
+	}
 
 	if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
 	    (desc->pg_moreio || nfs_reqs_to_commit(&cinfo)))
@@ -835,18 +796,13 @@ EXPORT_SYMBOL_GPL(nfs_generic_pgio);
 
 static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
 {
-	struct nfs_pgio_mirror *mirror;
 	struct nfs_pgio_header *hdr;
 	int ret;
 
-	mirror = nfs_pgio_current_mirror(desc);
-
 	hdr = nfs_pgio_header_alloc(desc->pg_rw_ops);
 	if (!hdr) {
-		/* TODO: make sure this is right with mirroring - or
-		 *       should it back out all mirrors? */
-		desc->pg_completion_ops->error_cleanup(&mirror->pg_list);
-		return -ENOMEM;
+		desc->pg_error = -ENOMEM;
+		return desc->pg_error;
 	}
 	nfs_pgheader_init(desc, hdr, nfs_pgio_header_free);
 	ret = nfs_generic_pgio(desc, hdr);
@@ -874,6 +830,9 @@ static int nfs_pageio_setup_mirroring(struct nfs_pageio_descriptor *pgio,
 
 	mirror_count = pgio->pg_ops->pg_get_mirror_count(pgio, req);
 
+	if (pgio->pg_error < 0)
+		return pgio->pg_error;
+
 	if (!mirror_count || mirror_count > NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX)
 		return -EINVAL;
 
@@ -903,12 +862,6 @@ 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)
 {
@@ -982,6 +935,8 @@ static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
 	} else {
 		if (desc->pg_ops->pg_init)
 			desc->pg_ops->pg_init(desc, req);
+		if (desc->pg_error < 0)
+			return 0;
 		mirror->pg_base = req->wb_pgbase;
 	}
 	if (!nfs_can_coalesce_requests(prev, req, desc))
@@ -1147,6 +1102,8 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 	bytes = req->wb_bytes;
 
 	nfs_pageio_setup_mirroring(desc, req);
+	if (desc->pg_error < 0)
+		goto out_failed;
 
 	for (midx = 0; midx < desc->pg_mirror_count; midx++) {
 		if (midx) {
@@ -1163,7 +1120,8 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 
 			if (IS_ERR(dupreq)) {
 				nfs_page_group_unlock(req);
-				return 0;
+				desc->pg_error = PTR_ERR(dupreq);
+				goto out_failed;
 			}
 
 			nfs_lock_request(dupreq);
@@ -1176,10 +1134,32 @@ int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
 		if (nfs_pgio_has_mirroring(desc))
 			desc->pg_mirror_idx = midx;
 		if (!nfs_pageio_add_request_mirror(desc, dupreq))
-			return 0;
+			goto out_failed;
 	}
 
 	return 1;
+
+out_failed:
+	/*
+	 * We might have failed before sending any reqs over wire.
+	 * Clean up rest of the reqs in mirror pg_list.
+	 */
+	if (desc->pg_error) {
+		struct nfs_pgio_mirror *mirror;
+		void (*func)(struct list_head *);
+
+		/* remember fatal errors */
+		if (nfs_error_is_fatal(desc->pg_error))
+			mapping_set_error(desc->pg_inode->i_mapping,
+					  desc->pg_error);
+
+		func = desc->pg_completion_ops->error_cleanup;
+		for (midx = 0; midx < desc->pg_mirror_count; midx++) {
+			mirror = &desc->pg_mirrors[midx];
+			func(&mirror->pg_list);
+		}
+	}
+	return 0;
 }
 
 /*
@@ -1232,7 +1212,7 @@ int nfs_pageio_resend(struct nfs_pageio_descriptor *desc,
 	nfs_pageio_complete(desc);
 	if (!list_empty(&failed)) {
 		list_move(&failed, &hdr->pages);
-		return -EIO;
+		return desc->pg_error < 0 ? desc->pg_error : -EIO;
 	}
 	return 0;
 }

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 */
@@ -618,7 +618,6 @@ pnfs_destroy_layout(struct nfs_inode *nfsi)
 		pnfs_get_layout_hdr(lo);
 		pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RO_FAILED);
 		pnfs_layout_clear_fail_bit(lo, NFS_LAYOUT_RW_FAILED);
-		pnfs_clear_retry_layoutget(lo);
 		spin_unlock(&nfsi->vfs_inode.i_lock);
 		pnfs_free_lseg_list(&tmp_list);
 		pnfs_put_layout_hdr(lo);
@@ -703,6 +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);
 	}
@@ -826,7 +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;
@@ -861,7 +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;
@@ -894,7 +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);
@@ -904,17 +905,9 @@ send_layoutget(struct pnfs_layout_hdr *lo,
 		lseg = nfs4_proc_layoutget(lgp, gfp_flags);
 	} while (lseg == ERR_PTR(-EAGAIN));
 
-	if (IS_ERR(lseg)) {
-		switch (PTR_ERR(lseg)) {
-		case -ENOMEM:
-		case -ERESTARTSYS:
-			break;
-		default:
-			/* remember that LAYOUTGET failed and suspend trying */
-			pnfs_layout_io_set_failed(lo, range->iomode);
-		}
-		return NULL;
-	} else
+	if (IS_ERR(lseg) && !nfs_error_is_fatal(PTR_ERR(lseg)))
+		lseg = NULL;
+	else
 		pnfs_layout_clear_fail_bit(lo,
 				pnfs_iomode_to_fail_bit(range->iomode));
 
@@ -945,7 +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;
@@ -962,7 +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;
@@ -1005,7 +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);
@@ -1033,7 +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:
@@ -1096,13 +1089,12 @@ 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))
 		layoutreturn = pnfs_prepare_layoutreturn(lo);
 
-	pnfs_clear_retry_layoutget(lo);
 	list_for_each_entry_safe(lseg, tmp, &lo->plh_segs, pls_list)
 		/* If we are sending layoutreturn, invalidate all valid lsegs */
 		if (layoutreturn || test_bit(NFS_LSEG_ROC, &lseg->pls_flags)) {
@@ -1124,7 +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;
 }
 
@@ -1149,6 +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);
@@ -1465,25 +1458,15 @@ static bool pnfs_within_mdsthreshold(struct nfs_open_context *ctx,
 	return ret;
 }
 
-/* stop waiting if someone clears NFS_LAYOUT_RETRY_LAYOUTGET bit. */
-static int pnfs_layoutget_retry_bit_wait(struct wait_bit_key *key, int mode)
-{
-	if (!test_bit(NFS_LAYOUT_RETRY_LAYOUTGET, key->flags))
-		return 1;
-	return nfs_wait_bit_killable(key, mode);
-}
-
 static bool pnfs_prepare_to_retry_layoutget(struct pnfs_layout_hdr *lo)
 {
-	if (!pnfs_should_retry_layoutget(lo))
-		return false;
 	/*
 	 * send layoutcommit as it can hold up layoutreturn due to lseg
 	 * reference
 	 */
 	pnfs_layoutcommit_inode(lo->plh_inode, false);
 	return !wait_on_bit_action(&lo->plh_flags, NFS_LAYOUT_RETURN,
-				   pnfs_layoutget_retry_bit_wait,
+				   nfs_wait_bit_killable,
 				   TASK_UNINTERRUPTIBLE);
 }
 
@@ -1520,14 +1503,23 @@ pnfs_update_layout(struct inode *ino,
 	struct pnfs_layout_segment *lseg = NULL;
 	bool first;
 
-	if (!pnfs_enabled_sb(NFS_SERVER(ino)))
+	if (!pnfs_enabled_sb(NFS_SERVER(ino))) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, NULL,
+				 PNFS_UPDATE_LAYOUT_NO_PNFS);
 		goto out;
+	}
 
-	if (iomode == IOMODE_READ && i_size_read(ino) == 0)
+	if (iomode == IOMODE_READ && i_size_read(ino) == 0) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, NULL,
+				 PNFS_UPDATE_LAYOUT_RD_ZEROLEN);
 		goto out;
+	}
 
-	if (pnfs_within_mdsthreshold(ctx, ino, iomode))
+	if (pnfs_within_mdsthreshold(ctx, ino, iomode)) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, NULL,
+				 PNFS_UPDATE_LAYOUT_MDSTHRESH);
 		goto out;
+	}
 
 lookup_again:
 	first = false;
@@ -1535,19 +1527,25 @@ lookup_again:
 	lo = pnfs_find_alloc_layout(ino, ctx, gfp_flags);
 	if (lo == NULL) {
 		spin_unlock(&ino->i_lock);
+		trace_pnfs_update_layout(ino, pos, count, iomode, NULL,
+				 PNFS_UPDATE_LAYOUT_NOMEM);
 		goto out;
 	}
 
 	/* Do we even need to bother with this? */
 	if (test_bit(NFS_LAYOUT_BULK_RECALL, &lo->plh_flags)) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+				 PNFS_UPDATE_LAYOUT_BULK_RECALL);
 		dprintk("%s matches recall, use MDS\n", __func__);
 		goto out_unlock;
 	}
 
 	/* if LAYOUTGET already failed once we don't try again */
-	if (pnfs_layout_io_test_failed(lo, iomode) &&
-	    !pnfs_should_retry_layoutget(lo))
+	if (pnfs_layout_io_test_failed(lo, iomode)) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+				 PNFS_UPDATE_LAYOUT_IO_TEST_FAIL);
 		goto out_unlock;
+	}
 
 	first = list_empty(&lo->plh_segs);
 	if (first) {
@@ -1567,8 +1565,11 @@ lookup_again:
 		 * already exists
 		 */
 		lseg = pnfs_find_lseg(lo, &arg);
-		if (lseg)
+		if (lseg) {
+			trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+					PNFS_UPDATE_LAYOUT_FOUND_CACHED);
 			goto out_unlock;
+		}
 	}
 
 	/*
@@ -1585,11 +1586,16 @@ lookup_again:
 			dprintk("%s retrying\n", __func__);
 			goto lookup_again;
 		}
+		trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+				PNFS_UPDATE_LAYOUT_RETURN);
 		goto out_put_layout_hdr;
 	}
 
-	if (pnfs_layoutgets_blocked(lo))
+	if (pnfs_layoutgets_blocked(lo)) {
+		trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+				PNFS_UPDATE_LAYOUT_BLOCKED);
 		goto out_unlock;
+	}
 	atomic_inc(&lo->plh_outstanding);
 	spin_unlock(&ino->i_lock);
 
@@ -1612,8 +1618,9 @@ lookup_again:
 		arg.length = PAGE_CACHE_ALIGN(arg.length);
 
 	lseg = send_layoutget(lo, ctx, &arg, gfp_flags);
-	pnfs_clear_retry_layoutget(lo);
 	atomic_dec(&lo->plh_outstanding);
+	trace_pnfs_update_layout(ino, pos, count, iomode, lo,
+				 PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET);
 out_put_layout_hdr:
 	if (first)
 		pnfs_clear_first_layoutget(lo);
@@ -1623,7 +1630,7 @@ out:
 			"(%s, offset: %llu, length: %llu)\n",
 			__func__, ino->i_sb->s_id,
 			(unsigned long long)NFS_FILEID(ino),
-			lseg == NULL ? "not found" : "found",
+			IS_ERR_OR_NULL(lseg) ? "not found" : "found",
 			iomode==IOMODE_RW ?  "read/write" : "read-only",
 			(unsigned long long)pos,
 			(unsigned long long)count);
@@ -1730,16 +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)) {
@@ -1749,38 +1769,47 @@ 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,
 				       struct pnfs_layout_segment *lseg)
 {
 	struct pnfs_layout_hdr *lo = NFS_I(inode)->layout;
-	int iomode = pnfs_iomode_to_fail_bit(lseg->pls_range.iomode);
 	struct pnfs_layout_range range = {
 		.iomode = lseg->pls_range.iomode,
 		.offset = 0,
 		.length = NFS4_MAX_UINT64,
 	};
 	LIST_HEAD(free_me);
+	bool return_now = false;
 
 	spin_lock(&inode->i_lock);
-	/* set failure bit so that pnfs path will be retried later */
-	pnfs_layout_set_fail_bit(lo, iomode);
-	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);
@@ -1802,6 +1831,11 @@ pnfs_generic_pg_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *r
 						   rd_size,
 						   IOMODE_READ,
 						   GFP_KERNEL);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
 	}
 	/* If no lseg, fall back to read through mds */
 	if (pgio->pg_lseg == NULL)
@@ -1814,13 +1848,19 @@ void
 pnfs_generic_pg_init_write(struct nfs_pageio_descriptor *pgio,
 			   struct nfs_page *req, u64 wb_size)
 {
-	if (pgio->pg_lseg == NULL)
+	if (pgio->pg_lseg == NULL) {
 		pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode,
 						   req->wb_context,
 						   req_offset(req),
 						   wb_size,
 						   IOMODE_RW,
 						   GFP_NOFS);
+		if (IS_ERR(pgio->pg_lseg)) {
+			pgio->pg_error = PTR_ERR(pgio->pg_lseg);
+			pgio->pg_lseg = NULL;
+			return;
+		}
+	}
 	/* If no lseg, fall back to write through mds */
 	if (pgio->pg_lseg == NULL)
 		nfs_pageio_reset_write_mds(pgio);
@@ -1988,15 +2028,13 @@ static void pnfs_writehdr_free(struct nfs_pgio_header *hdr)
 int
 pnfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc)
 {
-	struct nfs_pgio_mirror *mirror = nfs_pgio_current_mirror(desc);
-
 	struct nfs_pgio_header *hdr;
 	int ret;
 
 	hdr = nfs_pgio_header_alloc(desc->pg_rw_ops);
 	if (!hdr) {
-		desc->pg_completion_ops->error_cleanup(&mirror->pg_list);
-		return -ENOMEM;
+		desc->pg_error = -ENOMEM;
+		return desc->pg_error;
 	}
 	nfs_pgheader_init(desc, hdr, pnfs_writehdr_free);
 
@@ -2119,15 +2157,13 @@ static void pnfs_readhdr_free(struct nfs_pgio_header *hdr)
 int
 pnfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc)
 {
-	struct nfs_pgio_mirror *mirror = nfs_pgio_current_mirror(desc);
-
 	struct nfs_pgio_header *hdr;
 	int ret;
 
 	hdr = nfs_pgio_header_alloc(desc->pg_rw_ops);
 	if (!hdr) {
-		desc->pg_completion_ops->error_cleanup(&mirror->pg_list);
-		return -ENOMEM;
+		desc->pg_error = -ENOMEM;
+		return desc->pg_error;
 	}
 	nfs_pgheader_init(desc, hdr, pnfs_readhdr_free);
 	hdr->lseg = pnfs_get_lseg(desc->pg_lseg);

fs/nfs/pnfs.h

@@ -98,7 +98,6 @@ enum {
 	NFS_LAYOUT_RETURN_BEFORE_CLOSE,	/* Return this layout before close */
 	NFS_LAYOUT_INVALID_STID,	/* layout stateid id is invalid */
 	NFS_LAYOUT_FIRST_LAYOUTGET,	/* Serialize first layoutget */
-	NFS_LAYOUT_RETRY_LAYOUTGET,	/* Retry layoutget */
 };
 
 enum layoutdriver_policy_flags {
@@ -261,11 +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);
@@ -379,26 +381,6 @@ nfs4_get_deviceid(struct nfs4_deviceid_node *d)
 	return d;
 }
 
-static inline void pnfs_set_retry_layoutget(struct pnfs_layout_hdr *lo)
-{
-	if (!test_and_set_bit(NFS_LAYOUT_RETRY_LAYOUTGET, &lo->plh_flags))
-		atomic_inc(&lo->plh_refcount);
-}
-
-static inline void pnfs_clear_retry_layoutget(struct pnfs_layout_hdr *lo)
-{
-	if (test_and_clear_bit(NFS_LAYOUT_RETRY_LAYOUTGET, &lo->plh_flags)) {
-		atomic_dec(&lo->plh_refcount);
-		/* wake up waiters for LAYOUTRETURN as that is not needed */
-		wake_up_bit(&lo->plh_flags, NFS_LAYOUT_RETURN);
-	}
-}
-
-static inline bool pnfs_should_retry_layoutget(struct pnfs_layout_hdr *lo)
-{
-	return test_bit(NFS_LAYOUT_RETRY_LAYOUTGET, &lo->plh_flags);
-}
-
 static inline struct pnfs_layout_segment *
 pnfs_get_lseg(struct pnfs_layout_segment *lseg)
 {
@@ -409,6 +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)
 {
@@ -556,6 +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

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

fs/nfs/read.c

@@ -85,6 +85,23 @@ void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio)
 }
 EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds);
 
+static void nfs_readpage_release(struct nfs_page *req)
+{
+	struct inode *inode = d_inode(req->wb_context->dentry);
+
+	dprintk("NFS: read done (%s/%llu %d@%lld)\n", inode->i_sb->s_id,
+		(unsigned long long)NFS_FILEID(inode), req->wb_bytes,
+		(long long)req_offset(req));
+
+	if (nfs_page_group_sync_on_bit(req, PG_UNLOCKPAGE)) {
+		if (PageUptodate(req->wb_page))
+			nfs_readpage_to_fscache(inode, req->wb_page, 0);
+
+		unlock_page(req->wb_page);
+	}
+	nfs_release_request(req);
+}
+
 int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
 		       struct page *page)
 {
@@ -106,7 +123,10 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
 
 	nfs_pageio_init_read(&pgio, inode, false,
 			     &nfs_async_read_completion_ops);
-	nfs_pageio_add_request(&pgio, new);
+	if (!nfs_pageio_add_request(&pgio, new)) {
+		nfs_list_remove_request(new);
+		nfs_readpage_release(new);
+	}
 	nfs_pageio_complete(&pgio);
 
 	/* It doesn't make sense to do mirrored reads! */
@@ -115,24 +135,7 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
 	pgm = &pgio.pg_mirrors[0];
 	NFS_I(inode)->read_io += pgm->pg_bytes_written;
 
-	return 0;
-}
-
-static void nfs_readpage_release(struct nfs_page *req)
-{
-	struct inode *inode = d_inode(req->wb_context->dentry);
-
-	dprintk("NFS: read done (%s/%llu %d@%lld)\n", inode->i_sb->s_id,
-		(unsigned long long)NFS_FILEID(inode), req->wb_bytes,
-		(long long)req_offset(req));
-
-	if (nfs_page_group_sync_on_bit(req, PG_UNLOCKPAGE)) {
-		if (PageUptodate(req->wb_page))
-			nfs_readpage_to_fscache(inode, req->wb_page, 0);
-
-		unlock_page(req->wb_page);
-	}
-	nfs_release_request(req);
+	return pgio.pg_error < 0 ? pgio.pg_error : 0;
 }
 
 static void nfs_page_group_set_uptodate(struct nfs_page *req)
@@ -361,6 +364,8 @@ readpage_async_filler(void *data, struct page *page)
 	if (len < PAGE_CACHE_SIZE)
 		zero_user_segment(page, len, PAGE_CACHE_SIZE);
 	if (!nfs_pageio_add_request(desc->pgio, new)) {
+		nfs_list_remove_request(new);
+		nfs_readpage_release(new);
 		error = desc->pgio->pg_error;
 		goto out_unlock;
 	}

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>
 
@@ -244,11 +246,9 @@ static int wb_priority(struct writeback_control *wbc)
 {
 	int ret = 0;
 	if (wbc->for_reclaim)
-		return FLUSH_HIGHPRI | FLUSH_STABLE;
+		return FLUSH_HIGHPRI | FLUSH_COND_STABLE;
 	if (wbc->sync_mode == WB_SYNC_ALL)
 		ret = FLUSH_COND_STABLE;
-	if (wbc->for_kupdate || wbc->for_background)
-		ret |= FLUSH_LOWPRI;
 	return ret;
 }
 
@@ -545,12 +545,22 @@ try_again:
 	return head;
 }
 
+static void nfs_write_error_remove_page(struct nfs_page *req)
+{
+	nfs_unlock_request(req);
+	nfs_end_page_writeback(req);
+	nfs_release_request(req);
+	generic_error_remove_page(page_file_mapping(req->wb_page),
+				  req->wb_page);
+}
+
 /*
  * Find an associated nfs write request, and prepare to flush it out
  * May return an error if the user signalled nfs_wait_on_request().
  */
 static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
-				struct page *page, bool nonblock)
+				struct page *page, bool nonblock,
+				bool launder)
 {
 	struct nfs_page *req;
 	int ret = 0;
@@ -567,8 +577,21 @@ static int nfs_page_async_flush(struct nfs_pageio_descriptor *pgio,
 
 	ret = 0;
 	if (!nfs_pageio_add_request(pgio, req)) {
-		nfs_redirty_request(req);
 		ret = pgio->pg_error;
+		/*
+		 * Remove the problematic req upon fatal errors
+		 * in launder case, while other dirty pages can
+		 * still be around until they get flushed.
+		 */
+		if (nfs_error_is_fatal(ret)) {
+			nfs_context_set_write_error(req->wb_context, ret);
+			if (launder) {
+				nfs_write_error_remove_page(req);
+				goto out;
+			}
+		}
+		nfs_redirty_request(req);
+		ret = -EAGAIN;
 	} else
 		nfs_add_stats(page_file_mapping(page)->host,
 				NFSIOS_WRITEPAGES, 1);
@@ -576,12 +599,14 @@ out:
 	return ret;
 }
 
-static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, struct nfs_pageio_descriptor *pgio)
+static int nfs_do_writepage(struct page *page, struct writeback_control *wbc,
+			    struct nfs_pageio_descriptor *pgio, bool launder)
 {
 	int ret;
 
 	nfs_pageio_cond_complete(pgio, page_file_index(page));
-	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE);
+	ret = nfs_page_async_flush(pgio, page, wbc->sync_mode == WB_SYNC_NONE,
+				   launder);
 	if (ret == -EAGAIN) {
 		redirty_page_for_writepage(wbc, page);
 		ret = 0;
@@ -592,7 +617,9 @@ static int nfs_do_writepage(struct page *page, struct writeback_control *wbc, st
 /*
  * Write an mmapped page to the server.
  */
-static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc)
+static int nfs_writepage_locked(struct page *page,
+				struct writeback_control *wbc,
+				bool launder)
 {
 	struct nfs_pageio_descriptor pgio;
 	struct inode *inode = page_file_mapping(page)->host;
@@ -601,7 +628,7 @@ static int nfs_writepage_locked(struct page *page, struct writeback_control *wbc
 	nfs_inc_stats(inode, NFSIOS_VFSWRITEPAGE);
 	nfs_pageio_init_write(&pgio, inode, wb_priority(wbc),
 				false, &nfs_async_write_completion_ops);
-	err = nfs_do_writepage(page, wbc, &pgio);
+	err = nfs_do_writepage(page, wbc, &pgio, launder);
 	nfs_pageio_complete(&pgio);
 	if (err < 0)
 		return err;
@@ -614,7 +641,7 @@ int nfs_writepage(struct page *page, struct writeback_control *wbc)
 {
 	int ret;
 
-	ret = nfs_writepage_locked(page, wbc);
+	ret = nfs_writepage_locked(page, wbc, false);
 	unlock_page(page);
 	return ret;
 }
@@ -623,7 +650,7 @@ static int nfs_writepages_callback(struct page *page, struct writeback_control *
 {
 	int ret;
 
-	ret = nfs_do_writepage(page, wbc, data);
+	ret = nfs_do_writepage(page, wbc, data, false);
 	unlock_page(page);
 	return ret;
 }
@@ -1128,7 +1155,8 @@ 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 &&
@@ -1326,9 +1354,15 @@ 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,
@@ -1529,27 +1563,21 @@ static void nfs_writeback_result(struct rpc_task *task,
 	}
 }
 
-
-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)
@@ -1666,6 +1694,13 @@ 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
  */
@@ -1687,7 +1722,6 @@ 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;
 }
 
@@ -1749,8 +1783,7 @@ 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)
@@ -1769,7 +1802,7 @@ 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,
@@ -1788,30 +1821,25 @@ 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
@@ -1821,6 +1849,7 @@ 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)
 {
@@ -1911,7 +1940,7 @@ int nfs_wb_page_cancel(struct inode *inode, struct page *page)
 /*
  * Write back all requests on one page - we do this before reading it.
  */
-int nfs_wb_page(struct inode *inode, struct page *page)
+int nfs_wb_single_page(struct inode *inode, struct page *page, bool launder)
 {
 	loff_t range_start = page_file_offset(page);
 	loff_t range_end = range_start + (loff_t)(PAGE_CACHE_SIZE - 1);
@@ -1928,7 +1957,7 @@ int nfs_wb_page(struct inode *inode, struct page *page)
 	for (;;) {
 		wait_on_page_writeback(page);
 		if (clear_page_dirty_for_io(page)) {
-			ret = nfs_writepage_locked(page, &wbc);
+			ret = nfs_writepage_locked(page, &wbc, launder);
 			if (ret < 0)
 				goto out_error;
 			continue;

include/linux/nfs4.h

@@ -592,4 +592,18 @@ enum data_content4 {
 	NFS4_CONTENT_HOLE		= 1,
 };
 
+enum pnfs_update_layout_reason {
+	PNFS_UPDATE_LAYOUT_UNKNOWN = 0,
+	PNFS_UPDATE_LAYOUT_NO_PNFS,
+	PNFS_UPDATE_LAYOUT_RD_ZEROLEN,
+	PNFS_UPDATE_LAYOUT_MDSTHRESH,
+	PNFS_UPDATE_LAYOUT_NOMEM,
+	PNFS_UPDATE_LAYOUT_BULK_RECALL,
+	PNFS_UPDATE_LAYOUT_IO_TEST_FAIL,
+	PNFS_UPDATE_LAYOUT_FOUND_CACHED,
+	PNFS_UPDATE_LAYOUT_RETURN,
+	PNFS_UPDATE_LAYOUT_BLOCKED,
+	PNFS_UPDATE_LAYOUT_SEND_LAYOUTGET,
+};
+
 #endif

include/linux/nfs_fs.h

@@ -60,18 +60,12 @@ struct nfs_lockowner {
 	pid_t l_pid;
 };
 
-#define NFS_IO_INPROGRESS 0
-struct nfs_io_counter {
-	unsigned long flags;
-	atomic_t io_count;
-};
-
 struct nfs_lock_context {
 	atomic_t count;
 	struct list_head list;
 	struct nfs_open_context *open_context;
 	struct nfs_lockowner lockowner;
-	struct nfs_io_counter io_count;
+	atomic_t io_count;
 };
 
 struct nfs4_state;
@@ -216,7 +210,6 @@ 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 */
@@ -518,12 +511,24 @@ extern int  nfs_updatepage(struct file *, struct page *, unsigned int, unsigned
  */
 extern int nfs_sync_inode(struct inode *inode);
 extern int nfs_wb_all(struct inode *inode);
-extern int nfs_wb_page(struct inode *inode, struct page* page);
+extern int nfs_wb_single_page(struct inode *inode, struct page *page, bool launder);
 extern int nfs_wb_page_cancel(struct inode *inode, struct page* page);
 extern int  nfs_commit_inode(struct inode *, int);
 extern struct nfs_commit_data *nfs_commitdata_alloc(void);
 extern void nfs_commit_free(struct nfs_commit_data *data);
 
+static inline int
+nfs_wb_launder_page(struct inode *inode, struct page *page)
+{
+	return nfs_wb_single_page(inode, page, true);
+}
+
+static inline int
+nfs_wb_page(struct inode *inode, struct page *page)
+{
+	return nfs_wb_single_page(inode, page, false);
+}
+
 static inline int
 nfs_have_writebacks(struct inode *inode)
 {

include/linux/nfs_fs_sb.h

@@ -102,6 +102,7 @@ struct nfs_client {
 #define NFS_SP4_MACH_CRED_STATEID  4	/* TEST_STATEID and FREE_STATEID */
 #define NFS_SP4_MACH_CRED_WRITE    5	/* WRITE */
 #define NFS_SP4_MACH_CRED_COMMIT   6	/* COMMIT */
+#define NFS_SP4_MACH_CRED_PNFS_CLEANUP  7 /* LAYOUTRETURN */
 #endif /* CONFIG_NFS_V4 */
 
 	/* Our own IP address, as a null-terminated string.

include/linux/nfs_xdr.h

@@ -1375,6 +1375,7 @@ enum {
 	NFS_IOHDR_ERROR = 0,
 	NFS_IOHDR_EOF,
 	NFS_IOHDR_REDO,
+	NFS_IOHDR_STAT,
 };
 
 struct nfs_pgio_header {
@@ -1420,11 +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 {
@@ -1454,12 +1456,14 @@ struct nfs_commit_data {
 	const struct rpc_call_ops *mds_ops;
 	const struct nfs_commit_completion_ops *completion_ops;
 	int (*commit_done_cb) (struct rpc_task *task, struct nfs_commit_data *data);
+	unsigned long		flags;
 };
 
 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 {

net/sunrpc/clnt.c

@@ -1217,6 +1217,7 @@ static int rpc_anyaddr(int family, struct sockaddr *buf, size_t buflen)
 			return -EINVAL;
 		memcpy(buf, &rpc_in6addr_loopback,
 				sizeof(rpc_in6addr_loopback));
+		break;
 	default:
 		dprintk("RPC:       %s: address family not supported\n",
 			__func__);

net/sunrpc/xprtrdma/backchannel.c

@@ -15,7 +15,7 @@
 # define RPCDBG_FACILITY	RPCDBG_TRANS
 #endif
 
-#define RPCRDMA_BACKCHANNEL_DEBUG
+#undef RPCRDMA_BACKCHANNEL_DEBUG
 
 static void rpcrdma_bc_free_rqst(struct rpcrdma_xprt *r_xprt,
 				 struct rpc_rqst *rqst)
@@ -42,8 +42,8 @@ static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt,
 	size_t size;
 
 	req = rpcrdma_create_req(r_xprt);
-	if (!req)
-		return -ENOMEM;
+	if (IS_ERR(req))
+		return PTR_ERR(req);
 	req->rl_backchannel = true;
 
 	size = RPCRDMA_INLINE_WRITE_THRESHOLD(rqst);
@@ -84,9 +84,7 @@ out_fail:
 static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt,
 				 unsigned int count)
 {
-	struct rpcrdma_buffer *buffers = &r_xprt->rx_buf;
 	struct rpcrdma_rep *rep;
-	unsigned long flags;
 	int rc = 0;
 
 	while (count--) {
@@ -98,9 +96,7 @@ static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt,
 			break;
 		}
 
-		spin_lock_irqsave(&buffers->rb_lock, flags);
-		list_add(&rep->rr_list, &buffers->rb_recv_bufs);
-		spin_unlock_irqrestore(&buffers->rb_lock, flags);
+		rpcrdma_recv_buffer_put(rep);
 	}
 
 	return rc;
@@ -140,6 +136,7 @@ int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs)
 			       __func__);
 			goto out_free;
 		}
+		dprintk("RPC:       %s: new rqst %p\n", __func__, rqst);
 
 		rqst->rq_xprt = &r_xprt->rx_xprt;
 		INIT_LIST_HEAD(&rqst->rq_list);
@@ -220,12 +217,14 @@ int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst)
 
 	rpclen = rqst->rq_svec[0].iov_len;
 
+#ifdef RPCRDMA_BACKCHANNEL_DEBUG
 	pr_info("RPC:       %s: rpclen %zd headerp 0x%p lkey 0x%x\n",
 		__func__, rpclen, headerp, rdmab_lkey(req->rl_rdmabuf));
 	pr_info("RPC:       %s: RPC/RDMA: %*ph\n",
 		__func__, (int)RPCRDMA_HDRLEN_MIN, headerp);
 	pr_info("RPC:       %s:      RPC: %*ph\n",
 		__func__, (int)rpclen, rqst->rq_svec[0].iov_base);
+#endif
 
 	req->rl_send_iov[0].addr = rdmab_addr(req->rl_rdmabuf);
 	req->rl_send_iov[0].length = RPCRDMA_HDRLEN_MIN;
@@ -269,6 +268,9 @@ void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst)
 {
 	struct rpc_xprt *xprt = rqst->rq_xprt;
 
+	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);
@@ -333,9 +335,7 @@ 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);
-#ifdef RPCRDMA_BACKCHANNEL_DEBUG
-	pr_info("RPC:       %s: using rqst %p\n", __func__, rqst);
-#endif
+	dprintk("RPC:       %s: using rqst %p\n", __func__, rqst);
 
 	/* Prepare rqst */
 	rqst->rq_reply_bytes_recvd = 0;
@@ -355,10 +355,8 @@ void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt,
 	 * direction reply.
 	 */
 	req = rpcr_to_rdmar(rqst);
-#ifdef RPCRDMA_BACKCHANNEL_DEBUG
-	pr_info("RPC:       %s: attaching rep %p to req %p\n",
+	dprintk("RPC:       %s: attaching rep %p to req %p\n",
 		__func__, rep, req);
-#endif
 	req->rl_reply = rep;
 
 	/* Defeat the retransmit detection logic in send_request */

net/sunrpc/xprtrdma/fmr_ops.c

@@ -179,6 +179,69 @@ out_maperr:
 	return rc;
 }
 
+static void
+__fmr_dma_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
+{
+	struct ib_device *device = r_xprt->rx_ia.ri_device;
+	struct rpcrdma_mw *mw = seg->rl_mw;
+	int nsegs = seg->mr_nsegs;
+
+	seg->rl_mw = NULL;
+
+	while (nsegs--)
+		rpcrdma_unmap_one(device, seg++);
+
+	rpcrdma_put_mw(r_xprt, mw);
+}
+
+/* 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.
+ */
+static void
+fmr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+{
+	struct rpcrdma_mr_seg *seg;
+	unsigned int i, nchunks;
+	struct rpcrdma_mw *mw;
+	LIST_HEAD(unmap_list);
+	int rc;
+
+	dprintk("RPC:       %s: req %p\n", __func__, req);
+
+	/* ORDER: Invalidate all of the req's MRs first
+	 *
+	 * ib_unmap_fmr() is slow, so use a single call instead
+	 * of one call per mapped MR.
+	 */
+	for (i = 0, nchunks = req->rl_nchunks; nchunks; nchunks--) {
+		seg = &req->rl_segments[i];
+		mw = seg->rl_mw;
+
+		list_add(&mw->r.fmr.fmr->list, &unmap_list);
+
+		i += seg->mr_nsegs;
+	}
+	rc = ib_unmap_fmr(&unmap_list);
+	if (rc)
+		pr_warn("%s: ib_unmap_fmr failed (%i)\n", __func__, rc);
+
+	/* ORDER: Now DMA unmap all of the req's MRs, and return
+	 * them to the free MW list.
+	 */
+	for (i = 0, nchunks = req->rl_nchunks; nchunks; nchunks--) {
+		seg = &req->rl_segments[i];
+
+		__fmr_dma_unmap(r_xprt, seg);
+
+		i += seg->mr_nsegs;
+		seg->mr_nsegs = 0;
+	}
+
+	req->rl_nchunks = 0;
+}
+
 /* Use the ib_unmap_fmr() verb to prevent further remote
  * access via RDMA READ or RDMA WRITE.
  */
@@ -231,6 +294,7 @@ fmr_op_destroy(struct rpcrdma_buffer *buf)
 
 const struct rpcrdma_memreg_ops rpcrdma_fmr_memreg_ops = {
 	.ro_map				= fmr_op_map,
+	.ro_unmap_sync			= fmr_op_unmap_sync,
 	.ro_unmap			= fmr_op_unmap,
 	.ro_open			= fmr_op_open,
 	.ro_maxpages			= fmr_op_maxpages,

net/sunrpc/xprtrdma/frwr_ops.c

@@ -245,12 +245,14 @@ frwr_op_maxpages(struct rpcrdma_xprt *r_xprt)
 		     rpcrdma_max_segments(r_xprt) * ia->ri_max_frmr_depth);
 }
 
-/* If FAST_REG or LOCAL_INV failed, indicate the frmr needs to be reset. */
+/* If FAST_REG or LOCAL_INV failed, indicate the frmr needs
+ * to be reset.
+ *
+ * WARNING: Only wr_id and status are reliable at this point
+ */
 static void
-frwr_sendcompletion(struct ib_wc *wc)
+__frwr_sendcompletion_flush(struct ib_wc *wc, struct rpcrdma_mw *r)
 {
-	struct rpcrdma_mw *r;
-
 	if (likely(wc->status == IB_WC_SUCCESS))
 		return;
 
@@ -261,9 +263,23 @@ frwr_sendcompletion(struct ib_wc *wc)
 	else
 		pr_warn("RPC:       %s: frmr %p error, status %s (%d)\n",
 			__func__, r, ib_wc_status_msg(wc->status), wc->status);
+
 	r->r.frmr.fr_state = FRMR_IS_STALE;
 }
 
+static void
+frwr_sendcompletion(struct ib_wc *wc)
+{
+	struct rpcrdma_mw *r = (struct rpcrdma_mw *)(unsigned long)wc->wr_id;
+	struct rpcrdma_frmr *f = &r->r.frmr;
+
+	if (unlikely(wc->status != IB_WC_SUCCESS))
+		__frwr_sendcompletion_flush(wc, r);
+
+	if (f->fr_waiter)
+		complete(&f->fr_linv_done);
+}
+
 static int
 frwr_op_init(struct rpcrdma_xprt *r_xprt)
 {
@@ -319,7 +335,7 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	struct rpcrdma_mw *mw;
 	struct rpcrdma_frmr *frmr;
 	struct ib_mr *mr;
-	struct ib_reg_wr reg_wr;
+	struct ib_reg_wr *reg_wr;
 	struct ib_send_wr *bad_wr;
 	int rc, i, n, dma_nents;
 	u8 key;
@@ -335,7 +351,9 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	} while (mw->r.frmr.fr_state != FRMR_IS_INVALID);
 	frmr = &mw->r.frmr;
 	frmr->fr_state = FRMR_IS_VALID;
+	frmr->fr_waiter = false;
 	mr = frmr->fr_mr;
+	reg_wr = &frmr->fr_regwr;
 
 	if (nsegs > ia->ri_max_frmr_depth)
 		nsegs = ia->ri_max_frmr_depth;
@@ -381,19 +399,19 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	key = (u8)(mr->rkey & 0x000000FF);
 	ib_update_fast_reg_key(mr, ++key);
 
-	reg_wr.wr.next = NULL;
-	reg_wr.wr.opcode = IB_WR_REG_MR;
-	reg_wr.wr.wr_id = (uintptr_t)mw;
-	reg_wr.wr.num_sge = 0;
-	reg_wr.wr.send_flags = 0;
-	reg_wr.mr = mr;
-	reg_wr.key = mr->rkey;
-	reg_wr.access = writing ?
-			IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
-			IB_ACCESS_REMOTE_READ;
+	reg_wr->wr.next = NULL;
+	reg_wr->wr.opcode = IB_WR_REG_MR;
+	reg_wr->wr.wr_id = (uintptr_t)mw;
+	reg_wr->wr.num_sge = 0;
+	reg_wr->wr.send_flags = 0;
+	reg_wr->mr = mr;
+	reg_wr->key = mr->rkey;
+	reg_wr->access = writing ?
+			 IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE :
+			 IB_ACCESS_REMOTE_READ;
 
 	DECR_CQCOUNT(&r_xprt->rx_ep);
-	rc = ib_post_send(ia->ri_id->qp, &reg_wr.wr, &bad_wr);
+	rc = ib_post_send(ia->ri_id->qp, &reg_wr->wr, &bad_wr);
 	if (rc)
 		goto out_senderr;
 
@@ -413,6 +431,116 @@ out_senderr:
 	return rc;
 }
 
+static struct ib_send_wr *
+__frwr_prepare_linv_wr(struct rpcrdma_mr_seg *seg)
+{
+	struct rpcrdma_mw *mw = seg->rl_mw;
+	struct rpcrdma_frmr *f = &mw->r.frmr;
+	struct ib_send_wr *invalidate_wr;
+
+	f->fr_waiter = false;
+	f->fr_state = FRMR_IS_INVALID;
+	invalidate_wr = &f->fr_invwr;
+
+	memset(invalidate_wr, 0, sizeof(*invalidate_wr));
+	invalidate_wr->wr_id = (unsigned long)(void *)mw;
+	invalidate_wr->opcode = IB_WR_LOCAL_INV;
+	invalidate_wr->ex.invalidate_rkey = f->fr_mr->rkey;
+
+	return invalidate_wr;
+}
+
+static void
+__frwr_dma_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
+		 int rc)
+{
+	struct ib_device *device = r_xprt->rx_ia.ri_device;
+	struct rpcrdma_mw *mw = seg->rl_mw;
+	struct rpcrdma_frmr *f = &mw->r.frmr;
+
+	seg->rl_mw = NULL;
+
+	ib_dma_unmap_sg(device, f->sg, f->sg_nents, seg->mr_dir);
+
+	if (!rc)
+		rpcrdma_put_mw(r_xprt, mw);
+	else
+		__frwr_queue_recovery(mw);
+}
+
+/* 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.
+ */
+static void
+frwr_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+{
+	struct ib_send_wr *invalidate_wrs, *pos, *prev, *bad_wr;
+	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
+	struct rpcrdma_mr_seg *seg;
+	unsigned int i, nchunks;
+	struct rpcrdma_frmr *f;
+	int rc;
+
+	dprintk("RPC:       %s: req %p\n", __func__, req);
+
+	/* ORDER: Invalidate all of the req's MRs first
+	 *
+	 * Chain the LOCAL_INV Work Requests and post them with
+	 * a single ib_post_send() call.
+	 */
+	invalidate_wrs = pos = prev = NULL;
+	seg = NULL;
+	for (i = 0, nchunks = req->rl_nchunks; nchunks; nchunks--) {
+		seg = &req->rl_segments[i];
+
+		pos = __frwr_prepare_linv_wr(seg);
+
+		if (!invalidate_wrs)
+			invalidate_wrs = pos;
+		else
+			prev->next = pos;
+		prev = pos;
+
+		i += seg->mr_nsegs;
+	}
+	f = &seg->rl_mw->r.frmr;
+
+	/* Strong send queue ordering guarantees that when the
+	 * last WR in the chain completes, all WRs in the chain
+	 * are complete.
+	 */
+	f->fr_invwr.send_flags = IB_SEND_SIGNALED;
+	f->fr_waiter = true;
+	init_completion(&f->fr_linv_done);
+	INIT_CQCOUNT(&r_xprt->rx_ep);
+
+	/* Transport disconnect drains the receive CQ before it
+	 * replaces the QP. The RPC reply handler won't call us
+	 * unless ri_id->qp is a valid pointer.
+	 */
+	rc = ib_post_send(ia->ri_id->qp, invalidate_wrs, &bad_wr);
+	if (rc)
+		pr_warn("%s: ib_post_send failed %i\n", __func__, rc);
+
+	wait_for_completion(&f->fr_linv_done);
+
+	/* ORDER: Now DMA unmap all of the req's MRs, and return
+	 * them to the free MW list.
+	 */
+	for (i = 0, nchunks = req->rl_nchunks; nchunks; nchunks--) {
+		seg = &req->rl_segments[i];
+
+		__frwr_dma_unmap(r_xprt, seg, rc);
+
+		i += seg->mr_nsegs;
+		seg->mr_nsegs = 0;
+	}
+
+	req->rl_nchunks = 0;
+}
+
 /* Post a LOCAL_INV Work Request to prevent further remote access
  * via RDMA READ or RDMA WRITE.
  */
@@ -423,23 +551,24 @@ frwr_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rpcrdma_mw *mw = seg1->rl_mw;
 	struct rpcrdma_frmr *frmr = &mw->r.frmr;
-	struct ib_send_wr invalidate_wr, *bad_wr;
+	struct ib_send_wr *invalidate_wr, *bad_wr;
 	int rc, nsegs = seg->mr_nsegs;
 
 	dprintk("RPC:       %s: FRMR %p\n", __func__, mw);
 
 	seg1->rl_mw = NULL;
 	frmr->fr_state = FRMR_IS_INVALID;
+	invalidate_wr = &mw->r.frmr.fr_invwr;
 
-	memset(&invalidate_wr, 0, sizeof(invalidate_wr));
-	invalidate_wr.wr_id = (unsigned long)(void *)mw;
-	invalidate_wr.opcode = IB_WR_LOCAL_INV;
-	invalidate_wr.ex.invalidate_rkey = frmr->fr_mr->rkey;
+	memset(invalidate_wr, 0, sizeof(*invalidate_wr));
+	invalidate_wr->wr_id = (uintptr_t)mw;
+	invalidate_wr->opcode = IB_WR_LOCAL_INV;
+	invalidate_wr->ex.invalidate_rkey = frmr->fr_mr->rkey;
 	DECR_CQCOUNT(&r_xprt->rx_ep);
 
 	ib_dma_unmap_sg(ia->ri_device, frmr->sg, frmr->sg_nents, seg1->mr_dir);
 	read_lock(&ia->ri_qplock);
-	rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr);
+	rc = ib_post_send(ia->ri_id->qp, invalidate_wr, &bad_wr);
 	read_unlock(&ia->ri_qplock);
 	if (rc)
 		goto out_err;
@@ -471,6 +600,7 @@ frwr_op_destroy(struct rpcrdma_buffer *buf)
 
 const struct rpcrdma_memreg_ops rpcrdma_frwr_memreg_ops = {
 	.ro_map				= frwr_op_map,
+	.ro_unmap_sync			= frwr_op_unmap_sync,
 	.ro_unmap			= frwr_op_unmap,
 	.ro_open			= frwr_op_open,
 	.ro_maxpages			= frwr_op_maxpages,

net/sunrpc/xprtrdma/physical_ops.c

@@ -83,6 +83,18 @@ physical_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
 	return 1;
 }
 
+/* DMA unmap all memory regions that were mapped for "req".
+ */
+static void
+physical_op_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
+{
+	struct ib_device *device = r_xprt->rx_ia.ri_device;
+	unsigned int i;
+
+	for (i = 0; req->rl_nchunks; --req->rl_nchunks)
+		rpcrdma_unmap_one(device, &req->rl_segments[i++]);
+}
+
 static void
 physical_op_destroy(struct rpcrdma_buffer *buf)
 {
@@ -90,6 +102,7 @@ physical_op_destroy(struct rpcrdma_buffer *buf)
 
 const struct rpcrdma_memreg_ops rpcrdma_physical_memreg_ops = {
 	.ro_map				= physical_op_map,
+	.ro_unmap_sync			= physical_op_unmap_sync,
 	.ro_unmap			= physical_op_unmap,
 	.ro_open			= physical_op_open,
 	.ro_maxpages			= physical_op_maxpages,

net/sunrpc/xprtrdma/rpc_rdma.c

@@ -804,6 +804,11 @@ rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 	if (req->rl_reply)
 		goto out_duplicate;
 
+	/* Sanity checking has passed. We are now committed
+	 * to complete this transaction.
+	 */
+	list_del_init(&rqst->rq_list);
+	spin_unlock_bh(&xprt->transport_lock);
 	dprintk("RPC:       %s: reply 0x%p completes request 0x%p\n"
 		"                   RPC request 0x%p xid 0x%08x\n",
 			__func__, rep, req, rqst,
@@ -888,12 +893,23 @@ badheader:
 		break;
 	}
 
+	/* Invalidate and flush the data payloads before waking the
+	 * waiting application. This guarantees the memory region is
+	 * properly fenced from the server before the application
+	 * accesses the data. It also ensures proper send flow
+	 * control: waking the next RPC waits until this RPC has
+	 * relinquished all its Send Queue entries.
+	 */
+	if (req->rl_nchunks)
+		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt, req);
+
 	credits = be32_to_cpu(headerp->rm_credit);
 	if (credits == 0)
 		credits = 1;	/* don't deadlock */
 	else if (credits > r_xprt->rx_buf.rb_max_requests)
 		credits = r_xprt->rx_buf.rb_max_requests;
 
+	spin_lock_bh(&xprt->transport_lock);
 	cwnd = xprt->cwnd;
 	xprt->cwnd = credits << RPC_CWNDSHIFT;
 	if (xprt->cwnd > cwnd)

net/sunrpc/xprtrdma/transport.c

@@ -576,6 +576,9 @@ xprt_rdma_free(void *buffer)
 
 	rb = container_of(buffer, struct rpcrdma_regbuf, rg_base[0]);
 	req = rb->rg_owner;
+	if (req->rl_backchannel)
+		return;
+
 	r_xprt = container_of(req->rl_buffer, struct rpcrdma_xprt, rx_buf);
 
 	dprintk("RPC:       %s: called on 0x%p\n", __func__, req->rl_reply);

net/sunrpc/xprtrdma/verbs.c

@@ -616,10 +616,8 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia,
 
 	/* set trigger for requesting send completion */
 	ep->rep_cqinit = ep->rep_attr.cap.max_send_wr/2 - 1;
-	if (ep->rep_cqinit > RPCRDMA_MAX_UNSIGNALED_SENDS)
-		ep->rep_cqinit = RPCRDMA_MAX_UNSIGNALED_SENDS;
-	else if (ep->rep_cqinit <= 2)
-		ep->rep_cqinit = 0;
+	if (ep->rep_cqinit <= 2)
+		ep->rep_cqinit = 0;	/* always signal? */
 	INIT_CQCOUNT(ep);
 	init_waitqueue_head(&ep->rep_connect_wait);
 	INIT_DELAYED_WORK(&ep->rep_connect_worker, rpcrdma_connect_worker);
@@ -852,10 +850,11 @@ retry:
 
 		if (extras) {
 			rc = rpcrdma_ep_post_extra_recv(r_xprt, extras);
-			if (rc)
+			if (rc) {
 				pr_warn("%s: rpcrdma_ep_post_extra_recv: %i\n",
 					__func__, rc);
 				rc = 0;
+			}
 		}
 	}
 
@@ -1337,15 +1336,14 @@ rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *r_xprt, unsigned int count)
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rpcrdma_ep *ep = &r_xprt->rx_ep;
 	struct rpcrdma_rep *rep;
-	unsigned long flags;
 	int rc;
 
 	while (count--) {
-		spin_lock_irqsave(&buffers->rb_lock, flags);
+		spin_lock(&buffers->rb_lock);
 		if (list_empty(&buffers->rb_recv_bufs))
 			goto out_reqbuf;
 		rep = rpcrdma_buffer_get_rep_locked(buffers);
-		spin_unlock_irqrestore(&buffers->rb_lock, flags);
+		spin_unlock(&buffers->rb_lock);
 
 		rc = rpcrdma_ep_post_recv(ia, ep, rep);
 		if (rc)
@@ -1355,7 +1353,7 @@ rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *r_xprt, unsigned int count)
 	return 0;
 
 out_reqbuf:
-	spin_unlock_irqrestore(&buffers->rb_lock, flags);
+	spin_unlock(&buffers->rb_lock);
 	pr_warn("%s: no extra receive buffers\n", __func__);
 	return -ENOMEM;
 

net/sunrpc/xprtrdma/xprt_rdma.h

@@ -88,12 +88,6 @@ struct rpcrdma_ep {
 	struct delayed_work	rep_connect_worker;
 };
 
-/*
- * Force a signaled SEND Work Request every so often,
- * in case the provider needs to do some housekeeping.
- */
-#define RPCRDMA_MAX_UNSIGNALED_SENDS	(32)
-
 #define INIT_CQCOUNT(ep) atomic_set(&(ep)->rep_cqcount, (ep)->rep_cqinit)
 #define DECR_CQCOUNT(ep) atomic_sub_return(1, &(ep)->rep_cqcount)
 
@@ -207,6 +201,12 @@ struct rpcrdma_frmr {
 	enum rpcrdma_frmr_state		fr_state;
 	struct work_struct		fr_work;
 	struct rpcrdma_xprt		*fr_xprt;
+	bool				fr_waiter;
+	struct completion		fr_linv_done;;
+	union {
+		struct ib_reg_wr	fr_regwr;
+		struct ib_send_wr	fr_invwr;
+	};
 };
 
 struct rpcrdma_fmr {
@@ -364,6 +364,8 @@ struct rpcrdma_xprt;
 struct rpcrdma_memreg_ops {
 	int		(*ro_map)(struct rpcrdma_xprt *,
 				  struct rpcrdma_mr_seg *, int, bool);
+	void		(*ro_unmap_sync)(struct rpcrdma_xprt *,
+					 struct rpcrdma_req *);
 	int		(*ro_unmap)(struct rpcrdma_xprt *,
 				    struct rpcrdma_mr_seg *);
 	int		(*ro_open)(struct rpcrdma_ia *,

net/sunrpc/xprtsock.c

@@ -398,7 +398,6 @@ static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen,
 	if (unlikely(!sock))
 		return -ENOTSOCK;
 
-	clear_bit(SOCKWQ_ASYNC_NOSPACE, &sock->flags);
 	if (base != 0) {
 		addr = NULL;
 		addrlen = 0;
@@ -442,7 +441,6 @@ static void xs_nospace_callback(struct rpc_task *task)
 	struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
 
 	transport->inet->sk_write_pending--;
-	clear_bit(SOCKWQ_ASYNC_NOSPACE, &transport->sock->flags);
 }
 
 /**
@@ -467,20 +465,11 @@ static int xs_nospace(struct rpc_task *task)
 
 	/* Don't race with disconnect */
 	if (xprt_connected(xprt)) {
-		if (test_bit(SOCKWQ_ASYNC_NOSPACE, &transport->sock->flags)) {
-			/*
-			 * Notify TCP that we're limited by the application
-			 * window size
-			 */
-			set_bit(SOCK_NOSPACE, &transport->sock->flags);
-			sk->sk_write_pending++;
-			/* ...and wait for more buffer space */
-			xprt_wait_for_buffer_space(task, xs_nospace_callback);
-		}
-	} else {
-		clear_bit(SOCKWQ_ASYNC_NOSPACE, &transport->sock->flags);
+		/* wait for more buffer space */
+		sk->sk_write_pending++;
+		xprt_wait_for_buffer_space(task, xs_nospace_callback);
+	} else
 		ret = -ENOTCONN;
-	}
 
 	spin_unlock_bh(&xprt->transport_lock);
 
@@ -616,9 +605,6 @@ process_status:
 	case -EAGAIN:
 		status = xs_nospace(task);
 		break;
-	default:
-		dprintk("RPC:       sendmsg returned unrecognized error %d\n",
-			-status);
 	case -ENETUNREACH:
 	case -ENOBUFS:
 	case -EPIPE:
@@ -626,7 +612,10 @@ process_status:
 	case -EPERM:
 		/* When the server has died, an ICMP port unreachable message
 		 * prompts ECONNREFUSED. */
-		clear_bit(SOCKWQ_ASYNC_NOSPACE, &transport->sock->flags);
+		break;
+	default:
+		dprintk("RPC:       sendmsg returned unrecognized error %d\n",
+			-status);
 	}
 
 	return status;
@@ -706,16 +695,16 @@ static int xs_tcp_send_request(struct rpc_task *task)
 	case -EAGAIN:
 		status = xs_nospace(task);
 		break;
-	default:
-		dprintk("RPC:       sendmsg returned unrecognized error %d\n",
-			-status);
 	case -ECONNRESET:
 	case -ECONNREFUSED:
 	case -ENOTCONN:
 	case -EADDRINUSE:
 	case -ENOBUFS:
 	case -EPIPE:
-		clear_bit(SOCKWQ_ASYNC_NOSPACE, &transport->sock->flags);
+		break;
+	default:
+		dprintk("RPC:       sendmsg returned unrecognized error %d\n",
+			-status);
 	}
 
 	return status;
@@ -1609,19 +1598,23 @@ static void xs_tcp_state_change(struct sock *sk)
 
 static void xs_write_space(struct sock *sk)
 {
-	struct socket *sock;
+	struct socket_wq *wq;
 	struct rpc_xprt *xprt;
 
-	if (unlikely(!(sock = sk->sk_socket)))
+	if (!sk->sk_socket)
 		return;
-	clear_bit(SOCK_NOSPACE, &sock->flags);
+	clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
 
 	if (unlikely(!(xprt = xprt_from_sock(sk))))
 		return;
-	if (test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sock->flags) == 0)
-		return;
+	rcu_read_lock();
+	wq = rcu_dereference(sk->sk_wq);
+	if (!wq || test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags) == 0)
+		goto out;
 
 	xprt_write_space(xprt);
+out:
+	rcu_read_unlock();
 }
 
 /**
@@ -1907,18 +1900,6 @@ static inline void xs_reclassify_socket(int family, struct socket *sock)
 	}
 }
 #else
-static inline void xs_reclassify_socketu(struct socket *sock)
-{
-}
-
-static inline void xs_reclassify_socket4(struct socket *sock)
-{
-}
-
-static inline void xs_reclassify_socket6(struct socket *sock)
-{
-}
-
 static inline void xs_reclassify_socket(int family, struct socket *sock)
 {
 }
@@ -2008,7 +1989,7 @@ static int xs_local_setup_socket(struct sock_xprt *transport)
 			"transport socket (%d).\n", -status);
 		goto out;
 	}
-	xs_reclassify_socketu(sock);
+	xs_reclassify_socket(AF_LOCAL, sock);
 
 	dprintk("RPC:       worker connecting xprt %p via AF_LOCAL to %s\n",
 			xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);