[PATCH 1/7] NFSv4: Ensure layout headers are RCU safe

[PATCH 1/7] NFSv4: Ensure layout headers are RCU safe

[Trond Myklebust]

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/blocklayout/blocklayout.c       |  2 +-
 fs/nfs/filelayout/filelayout.c         |  2 +-
 fs/nfs/flexfilelayout/flexfilelayout.c |  6 +++---
 fs/nfs/pnfs.c                          | 12 ++++++------
 fs/nfs/pnfs.h                          |  2 ++
 5 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c
index 690221747b47..d1a0e2c8b1b4 100644
--- a/fs/nfs/blocklayout/blocklayout.c
+++ b/fs/nfs/blocklayout/blocklayout.c
@@ -476,7 +476,7 @@ static void bl_free_layout_hdr(struct pnfs_layout_hdr *lo)
 	err = ext_tree_remove(bl, true, 0, LLONG_MAX);
 	WARN_ON(err);
 
-	kfree(bl);
+	kfree_rcu(bl, bl_layout.plh_rcu);
 }
 
 static struct pnfs_layout_hdr *__bl_alloc_layout_hdr(struct inode *inode,
diff --git a/fs/nfs/filelayout/filelayout.c b/fs/nfs/filelayout/filelayout.c
index c9b605f6c9cb..bd234394a87c 100644
--- a/fs/nfs/filelayout/filelayout.c
+++ b/fs/nfs/filelayout/filelayout.c
@@ -1146,7 +1146,7 @@ filelayout_alloc_layout_hdr(struct inode *inode, gfp_t gfp_flags)
 static void
 filelayout_free_layout_hdr(struct pnfs_layout_hdr *lo)
 {
-	kfree(FILELAYOUT_FROM_HDR(lo));
+	kfree_rcu(FILELAYOUT_FROM_HDR(lo), generic_hdr.plh_rcu);
 }
 
 static struct pnfs_ds_commit_info *
diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c b/fs/nfs/flexfilelayout/flexfilelayout.c
index 8b8171b48893..e7d8ae4d0cc5 100644
--- a/fs/nfs/flexfilelayout/flexfilelayout.c
+++ b/fs/nfs/flexfilelayout/flexfilelayout.c
@@ -59,14 +59,14 @@ ff_layout_alloc_layout_hdr(struct inode *inode, gfp_t gfp_flags)
 static void
 ff_layout_free_layout_hdr(struct pnfs_layout_hdr *lo)
 {
+	struct nfs4_flexfile_layout *ffl = FF_LAYOUT_FROM_HDR(lo);
 	struct nfs4_ff_layout_ds_err *err, *n;
 
-	list_for_each_entry_safe(err, n, &FF_LAYOUT_FROM_HDR(lo)->error_list,
-				 list) {
+	list_for_each_entry_safe(err, n, &ffl->error_list, list) {
 		list_del(&err->list);
 		kfree(err);
 	}
-	kfree(FF_LAYOUT_FROM_HDR(lo));
+	kfree_rcu(ffl, generic_hdr.plh_rcu);
 }
 
 static int decode_pnfs_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index cb99ac954688..268e7b9ff54e 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -268,11 +268,11 @@ pnfs_free_layout_hdr(struct pnfs_layout_hdr *lo)
 	struct nfs_server *server = NFS_SERVER(lo->plh_inode);
 	struct pnfs_layoutdriver_type *ld = server->pnfs_curr_ld;
 
-	if (!list_empty(&lo->plh_layouts)) {
+	if (test_and_clear_bit(NFS_LAYOUT_HASHED, &lo->plh_flags)) {
 		struct nfs_client *clp = server->nfs_client;
 
 		spin_lock(&clp->cl_lock);
-		list_del_init(&lo->plh_layouts);
+		list_del_rcu(&lo->plh_layouts);
 		spin_unlock(&clp->cl_lock);
 	}
 	put_cred(lo->plh_lc_cred);
@@ -784,7 +784,8 @@ pnfs_layout_bulk_destroy_byserver_locked(struct nfs_client *clp,
 			break;
 		inode = igrab(lo->plh_inode);
 		if (inode != NULL) {
-			list_del_init(&lo->plh_layouts);
+			if (test_and_clear_bit(NFS_LAYOUT_HASHED, &lo->plh_flags))
+				list_del_rcu(&lo->plh_layouts);
 			if (pnfs_layout_add_bulk_destroy_list(inode,
 						layout_list))
 				continue;
@@ -1870,15 +1871,14 @@ static void pnfs_clear_first_layoutget(struct pnfs_layout_hdr *lo)
 static void _add_to_server_list(struct pnfs_layout_hdr *lo,
 				struct nfs_server *server)
 {
-	if (list_empty(&lo->plh_layouts)) {
+	if (!test_and_set_bit(NFS_LAYOUT_HASHED, &lo->plh_flags)) {
 		struct nfs_client *clp = server->nfs_client;
 
 		/* The lo must be on the clp list if there is any
 		 * chance of a CB_LAYOUTRECALL(FILE) coming in.
 		 */
 		spin_lock(&clp->cl_lock);
-		if (list_empty(&lo->plh_layouts))
-			list_add_tail(&lo->plh_layouts, &server->layouts);
+		list_add_tail_rcu(&lo->plh_layouts, &server->layouts);
 		spin_unlock(&clp->cl_lock);
 	}
 }
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index cfb89d47c79d..8df9aa02d336 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -105,6 +105,7 @@ enum {
 	NFS_LAYOUT_INVALID_STID,	/* layout stateid id is invalid */
 	NFS_LAYOUT_FIRST_LAYOUTGET,	/* Serialize first layoutget */
 	NFS_LAYOUT_INODE_FREEING,	/* The inode is being freed */
+	NFS_LAYOUT_HASHED,		/* The layout visible */
 };
 
 enum layoutdriver_policy_flags {
@@ -203,6 +204,7 @@ struct pnfs_layout_hdr {
 	loff_t			plh_lwb; /* last write byte for layoutcommit */
 	const struct cred	*plh_lc_cred; /* layoutcommit cred */
 	struct inode		*plh_inode;
+	struct rcu_head		plh_rcu;
 };
 
 struct pnfs_device {
-- 
2.24.1

[PATCH 2/7] NFSv4/pnfs: Clean up nfs_layout_find_inode()

[Trond Myklebust]

Now that we can rely on just the rcu_read_lock(), remove the
clp->cl_lock and clean up.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/callback_proc.c | 52 +++++++++++++++++-------------------------
 1 file changed, 21 insertions(+), 31 deletions(-)

diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c
index eb9d035451a2..97084804a953 100644
--- a/fs/nfs/callback_proc.c
+++ b/fs/nfs/callback_proc.c
@@ -121,33 +121,31 @@ __be32 nfs4_callback_recall(void *argp, void *resp,
  */
 static struct inode *nfs_layout_find_inode_by_stateid(struct nfs_client *clp,
 		const nfs4_stateid *stateid)
+	__must_hold(RCU)
 {
 	struct nfs_server *server;
 	struct inode *inode;
 	struct pnfs_layout_hdr *lo;
 
+	rcu_read_lock();
 	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
-		list_for_each_entry(lo, &server->layouts, plh_layouts) {
+		list_for_each_entry_rcu(lo, &server->layouts, plh_layouts) {
 			if (!pnfs_layout_is_valid(lo))
 				continue;
 			if (stateid != NULL &&
 			    !nfs4_stateid_match_other(stateid, &lo->plh_stateid))
 				continue;
+			if (!nfs_sb_active(server->super))
+				continue;
 			inode = igrab(lo->plh_inode);
-			if (!inode)
-				return ERR_PTR(-EAGAIN);
-			if (!nfs_sb_active(inode->i_sb)) {
-				rcu_read_unlock();
-				spin_unlock(&clp->cl_lock);
-				iput(inode);
-				spin_lock(&clp->cl_lock);
-				rcu_read_lock();
-				return ERR_PTR(-EAGAIN);
-			}
-			return inode;
+			rcu_read_unlock();
+			if (inode)
+				return inode;
+			nfs_sb_deactive(server->super);
+			return ERR_PTR(-EAGAIN);
 		}
 	}
-
+	rcu_read_unlock();
 	return ERR_PTR(-ENOENT);
 }
 
@@ -165,28 +163,25 @@ static struct inode *nfs_layout_find_inode_by_fh(struct nfs_client *clp,
 	struct inode *inode;
 	struct pnfs_layout_hdr *lo;
 
+	rcu_read_lock();
 	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
-		list_for_each_entry(lo, &server->layouts, plh_layouts) {
+		list_for_each_entry_rcu(lo, &server->layouts, plh_layouts) {
 			nfsi = NFS_I(lo->plh_inode);
 			if (nfs_compare_fh(fh, &nfsi->fh))
 				continue;
 			if (nfsi->layout != lo)
 				continue;
+			if (!nfs_sb_active(server->super))
+				continue;
 			inode = igrab(lo->plh_inode);
-			if (!inode)
-				return ERR_PTR(-EAGAIN);
-			if (!nfs_sb_active(inode->i_sb)) {
-				rcu_read_unlock();
-				spin_unlock(&clp->cl_lock);
-				iput(inode);
-				spin_lock(&clp->cl_lock);
-				rcu_read_lock();
-				return ERR_PTR(-EAGAIN);
-			}
-			return inode;
+			rcu_read_unlock();
+			if (inode)
+				return inode;
+			nfs_sb_deactive(server->super);
+			return ERR_PTR(-EAGAIN);
 		}
 	}
-
+	rcu_read_unlock();
 	return ERR_PTR(-ENOENT);
 }
 
@@ -196,14 +191,9 @@ static struct inode *nfs_layout_find_inode(struct nfs_client *clp,
 {
 	struct inode *inode;
 
-	spin_lock(&clp->cl_lock);
-	rcu_read_lock();
 	inode = nfs_layout_find_inode_by_stateid(clp, stateid);
 	if (inode == ERR_PTR(-ENOENT))
 		inode = nfs_layout_find_inode_by_fh(clp, fh);
-	rcu_read_unlock();
-	spin_unlock(&clp->cl_lock);
-
 	return inode;
 }
 
-- 
2.24.1

[PATCH 3/7] NFS: Add a helper nfs_client_for_each_server()

[Trond Myklebust]

Add a helper nfs_client_for_each_server() to iterate through all the
filesystems that are attached to a struct nfs_client, and apply
a function to all the active ones.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/internal.h |  4 +++-
 fs/nfs/super.c    | 35 +++++++++++++++++++++++++++++++++++
 2 files changed, 38 insertions(+), 1 deletion(-)

diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index f80c47d5ff27..3b6fa9edc9b5 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -417,7 +417,9 @@ extern int __init register_nfs_fs(void);
 extern void __exit unregister_nfs_fs(void);
 extern bool nfs_sb_active(struct super_block *sb);
 extern void nfs_sb_deactive(struct super_block *sb);
-
+extern int nfs_client_for_each_server(struct nfs_client *clp,
+				      int (*fn)(struct nfs_server *, void *),
+				      void *data);
 /* io.c */
 extern void nfs_start_io_read(struct inode *inode);
 extern void nfs_end_io_read(struct inode *inode);
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index dada09b391c6..eb3a85492396 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -176,6 +176,41 @@ void nfs_sb_deactive(struct super_block *sb)
 }
 EXPORT_SYMBOL_GPL(nfs_sb_deactive);
 
+static int __nfs_list_for_each_server(struct list_head *head,
+		int (*fn)(struct nfs_server *, void *),
+		void *data)
+{
+	struct nfs_server *server, *last = NULL;
+	int ret = 0;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(server, head, client_link) {
+		if (!nfs_sb_active(server->super))
+			continue;
+		rcu_read_unlock();
+		if (last)
+			nfs_sb_deactive(last->super);
+		last = server;
+		ret = fn(server, data);
+		if (ret)
+			goto out;
+		rcu_read_lock();
+	}
+	rcu_read_unlock();
+out:
+	if (last)
+		nfs_sb_deactive(last->super);
+	return ret;
+}
+
+int nfs_client_for_each_server(struct nfs_client *clp,
+		int (*fn)(struct nfs_server *, void *),
+		void *data)
+{
+	return __nfs_list_for_each_server(&clp->cl_superblocks, fn, data);
+}
+EXPORT_SYMBOL_GPL(nfs_client_for_each_server);
+
 /*
  * Deliver file system statistics to userspace
  */
-- 
2.24.1

[PATCH 4/7] NFSv4: Clean up nfs_client_return_marked_delegations()

[Trond Myklebust]

Convert it to use the nfs_client_for_each_server() helper, and
make it more efficient by skipping delegations for inodes we
know are in the process of being freed. Also improve the efficiency
of the cursor by skipping delegations that are being freed.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/delegation.c | 129 +++++++++++++++++++++-----------------------
 1 file changed, 60 insertions(+), 69 deletions(-)

diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index 509b7235b132..19f66d3e58e8 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -563,21 +563,11 @@ static bool nfs_delegation_need_return(struct nfs_delegation *delegation)
 	return ret;
 }
 
-/**
- * nfs_client_return_marked_delegations - return previously marked delegations
- * @clp: nfs_client to process
- *
- * Note that this function is designed to be called by the state
- * manager thread. For this reason, it cannot flush the dirty data,
- * since that could deadlock in case of a state recovery error.
- *
- * Returns zero on success, or a negative errno value.
- */
-int nfs_client_return_marked_delegations(struct nfs_client *clp)
+static int nfs_server_return_marked_delegations(struct nfs_server *server,
+		void __always_unused *data)
 {
 	struct nfs_delegation *delegation;
 	struct nfs_delegation *prev;
-	struct nfs_server *server;
 	struct inode *inode;
 	struct inode *place_holder = NULL;
 	struct nfs_delegation *place_holder_deleg = NULL;
@@ -587,78 +577,79 @@ int nfs_client_return_marked_delegations(struct nfs_client *clp)
 	/*
 	 * To avoid quadratic looping we hold a reference
 	 * to an inode place_holder.  Each time we restart, we
-	 * list nfs_servers from the server of that inode, and
-	 * delegation in the server from the delegations of that
-	 * inode.
+	 * list delegation in the server from the delegations
+	 * of that inode.
 	 * prev is an RCU-protected pointer to a delegation which
 	 * wasn't marked for return and might be a good choice for
 	 * the next place_holder.
 	 */
-	rcu_read_lock();
 	prev = NULL;
+	delegation = NULL;
+	rcu_read_lock();
 	if (place_holder)
-		server = NFS_SERVER(place_holder);
-	else
-		server = list_entry_rcu(clp->cl_superblocks.next,
-					struct nfs_server, client_link);
-	list_for_each_entry_from_rcu(server, &clp->cl_superblocks, client_link) {
-		delegation = NULL;
-		if (place_holder && server == NFS_SERVER(place_holder))
-			delegation = rcu_dereference(NFS_I(place_holder)->delegation);
-		if (!delegation || delegation != place_holder_deleg)
-			delegation = list_entry_rcu(server->delegations.next,
-						    struct nfs_delegation, super_list);
-		list_for_each_entry_from_rcu(delegation, &server->delegations, super_list) {
-			struct inode *to_put = NULL;
-
-			if (!nfs_delegation_need_return(delegation)) {
+		delegation = rcu_dereference(NFS_I(place_holder)->delegation);
+	if (!delegation || delegation != place_holder_deleg)
+		delegation = list_entry_rcu(server->delegations.next,
+					    struct nfs_delegation, super_list);
+	list_for_each_entry_from_rcu(delegation, &server->delegations, super_list) {
+		struct inode *to_put = NULL;
+
+		if (test_bit(NFS_DELEGATION_INODE_FREEING, &delegation->flags))
+			continue;
+		if (!nfs_delegation_need_return(delegation)) {
+			if (nfs4_is_valid_delegation(delegation, 0))
 				prev = delegation;
-				continue;
-			}
-			if (!nfs_sb_active(server->super))
-				break; /* continue in outer loop */
-
-			if (prev) {
-				struct inode *tmp;
+			continue;
+		}
 
-				tmp = nfs_delegation_grab_inode(prev);
-				if (tmp) {
-					to_put = place_holder;
-					place_holder = tmp;
-					place_holder_deleg = prev;
-				}
+		if (prev) {
+			struct inode *tmp = nfs_delegation_grab_inode(prev);
+			if (tmp) {
+				to_put = place_holder;
+				place_holder = tmp;
+				place_holder_deleg = prev;
 			}
+		}
 
-			inode = nfs_delegation_grab_inode(delegation);
-			if (inode == NULL) {
-				rcu_read_unlock();
-				if (to_put)
-					iput(to_put);
-				nfs_sb_deactive(server->super);
-				goto restart;
-			}
-			delegation = nfs_start_delegation_return_locked(NFS_I(inode));
+		inode = nfs_delegation_grab_inode(delegation);
+		if (inode == NULL) {
 			rcu_read_unlock();
+			iput(to_put);
+			goto restart;
+		}
+		delegation = nfs_start_delegation_return_locked(NFS_I(inode));
+		rcu_read_unlock();
 
-			if (to_put)
-				iput(to_put);
+		iput(to_put);
 
-			err = nfs_end_delegation_return(inode, delegation, 0);
-			iput(inode);
-			nfs_sb_deactive(server->super);
-			cond_resched();
-			if (!err)
-				goto restart;
-			set_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state);
-			if (place_holder)
-				iput(place_holder);
-			return err;
-		}
+		err = nfs_end_delegation_return(inode, delegation, 0);
+		iput(inode);
+		cond_resched();
+		if (!err)
+			goto restart;
+		set_bit(NFS4CLNT_DELEGRETURN, &server->nfs_client->cl_state);
+		goto out;
 	}
 	rcu_read_unlock();
-	if (place_holder)
-		iput(place_holder);
-	return 0;
+out:
+	iput(place_holder);
+	return err;
+}
+
+/**
+ * nfs_client_return_marked_delegations - return previously marked delegations
+ * @clp: nfs_client to process
+ *
+ * Note that this function is designed to be called by the state
+ * manager thread. For this reason, it cannot flush the dirty data,
+ * since that could deadlock in case of a state recovery error.
+ *
+ * Returns zero on success, or a negative errno value.
+ */
+int nfs_client_return_marked_delegations(struct nfs_client *clp)
+{
+	return nfs_client_for_each_server(clp,
+			nfs_server_return_marked_delegations, NULL);
 }
 
 /**
-- 
2.24.1

[PATCH 5/7] NFSv4: Clean up nfs_delegation_reap_unclaimed()

[Trond Myklebust]

Convert nfs_delegation_reap_unclaimed() to use nfs_client_for_each_server()
for efficiency.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/delegation.c | 76 ++++++++++++++++++++++-----------------------
 1 file changed, 37 insertions(+), 39 deletions(-)

diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index 19f66d3e58e8..cb03ba99ae51 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -1092,53 +1092,51 @@ void nfs_delegation_mark_reclaim(struct nfs_client *clp)
 	rcu_read_unlock();
 }
 
-/**
- * nfs_delegation_reap_unclaimed - reap unclaimed delegations after reboot recovery is done
- * @clp: nfs_client to process
- *
- */
-void nfs_delegation_reap_unclaimed(struct nfs_client *clp)
+static int nfs_server_reap_unclaimed_delegations(struct nfs_server *server,
+		void __always_unused *data)
 {
 	struct nfs_delegation *delegation;
-	struct nfs_server *server;
 	struct inode *inode;
-
 restart:
 	rcu_read_lock();
-	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
-		list_for_each_entry_rcu(delegation, &server->delegations,
-								super_list) {
-			if (test_bit(NFS_DELEGATION_INODE_FREEING,
-						&delegation->flags) ||
-			    test_bit(NFS_DELEGATION_RETURNING,
-						&delegation->flags) ||
-			    test_bit(NFS_DELEGATION_NEED_RECLAIM,
-						&delegation->flags) == 0)
-				continue;
-			if (!nfs_sb_active(server->super))
-				break; /* continue in outer loop */
-			inode = nfs_delegation_grab_inode(delegation);
-			if (inode == NULL) {
-				rcu_read_unlock();
-				nfs_sb_deactive(server->super);
-				goto restart;
-			}
-			delegation = nfs_start_delegation_return_locked(NFS_I(inode));
-			rcu_read_unlock();
-			if (delegation != NULL) {
-				if (nfs_detach_delegation(NFS_I(inode), delegation,
-							server) != NULL)
-					nfs_free_delegation(delegation);
-				/* Match nfs_start_delegation_return_locked */
-				nfs_put_delegation(delegation);
-			}
-			iput(inode);
-			nfs_sb_deactive(server->super);
-			cond_resched();
-			goto restart;
+restart_locked:
+	list_for_each_entry_rcu(delegation, &server->delegations, super_list) {
+		if (test_bit(NFS_DELEGATION_INODE_FREEING,
+					&delegation->flags) ||
+		    test_bit(NFS_DELEGATION_RETURNING,
+					&delegation->flags) ||
+		    test_bit(NFS_DELEGATION_NEED_RECLAIM,
+					&delegation->flags) == 0)
+			continue;
+		inode = nfs_delegation_grab_inode(delegation);
+		if (inode == NULL)
+			goto restart_locked;
+		delegation = nfs_start_delegation_return_locked(NFS_I(inode));
+		rcu_read_unlock();
+		if (delegation != NULL) {
+			if (nfs_detach_delegation(NFS_I(inode), delegation,
+						server) != NULL)
+				nfs_free_delegation(delegation);
+			/* Match nfs_start_delegation_return_locked */
+			nfs_put_delegation(delegation);
 		}
+		iput(inode);
+		cond_resched();
+		goto restart;
 	}
 	rcu_read_unlock();
+	return 0;
+}
+
+/**
+ * nfs_delegation_reap_unclaimed - reap unclaimed delegations after reboot recovery is done
+ * @clp: nfs_client to process
+ *
+ */
+void nfs_delegation_reap_unclaimed(struct nfs_client *clp)
+{
+	nfs_client_for_each_server(clp, nfs_server_reap_unclaimed_delegations,
+			NULL);
 }
 
 static inline bool nfs4_server_rebooted(const struct nfs_client *clp)
-- 
2.24.1

[PATCH 6/7] NFSv4: Clean up nfs_delegation_reap_expired()

[Trond Myklebust]

Convert to use nfs_client_for_each_server() for efficiency.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/delegation.c | 83 ++++++++++++++++++++++-----------------------
 1 file changed, 40 insertions(+), 43 deletions(-)

diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index cb03ba99ae51..01974f17afc9 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -1222,62 +1222,59 @@ nfs_delegation_test_free_expired(struct inode *inode,
 		nfs_remove_bad_delegation(inode, stateid);
 }
 
-/**
- * nfs_reap_expired_delegations - reap expired delegations
- * @clp: nfs_client to process
- *
- * Iterates through all the delegations associated with this server and
- * checks if they have may have been revoked. This function is usually
- * expected to be called in cases where the server may have lost its
- * lease.
- */
-void nfs_reap_expired_delegations(struct nfs_client *clp)
+static int nfs_server_reap_expired_delegations(struct nfs_server *server,
+		void __always_unused *data)
 {
 	struct nfs_delegation *delegation;
-	struct nfs_server *server;
 	struct inode *inode;
 	const struct cred *cred;
 	nfs4_stateid stateid;
-
 restart:
 	rcu_read_lock();
-	list_for_each_entry_rcu(server, &clp->cl_superblocks, client_link) {
-		list_for_each_entry_rcu(delegation, &server->delegations,
-								super_list) {
-			if (test_bit(NFS_DELEGATION_INODE_FREEING,
-						&delegation->flags) ||
-			    test_bit(NFS_DELEGATION_RETURNING,
-						&delegation->flags) ||
-			    test_bit(NFS_DELEGATION_TEST_EXPIRED,
-						&delegation->flags) == 0)
-				continue;
-			if (!nfs_sb_active(server->super))
-				break; /* continue in outer loop */
-			inode = nfs_delegation_grab_inode(delegation);
-			if (inode == NULL) {
-				rcu_read_unlock();
-				nfs_sb_deactive(server->super);
-				goto restart;
-			}
-			cred = get_cred_rcu(delegation->cred);
-			nfs4_stateid_copy(&stateid, &delegation->stateid);
-			clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags);
-			rcu_read_unlock();
-			nfs_delegation_test_free_expired(inode, &stateid, cred);
-			put_cred(cred);
-			if (nfs4_server_rebooted(clp)) {
-				nfs_inode_mark_test_expired_delegation(server,inode);
-				iput(inode);
-				nfs_sb_deactive(server->super);
-				return;
-			}
+restart_locked:
+	list_for_each_entry_rcu(delegation, &server->delegations, super_list) {
+		if (test_bit(NFS_DELEGATION_INODE_FREEING,
+					&delegation->flags) ||
+		    test_bit(NFS_DELEGATION_RETURNING,
+					&delegation->flags) ||
+		    test_bit(NFS_DELEGATION_TEST_EXPIRED,
+					&delegation->flags) == 0)
+			continue;
+		inode = nfs_delegation_grab_inode(delegation);
+		if (inode == NULL)
+			goto restart_locked;
+		cred = get_cred_rcu(delegation->cred);
+		nfs4_stateid_copy(&stateid, &delegation->stateid);
+		clear_bit(NFS_DELEGATION_TEST_EXPIRED, &delegation->flags);
+		rcu_read_unlock();
+		nfs_delegation_test_free_expired(inode, &stateid, cred);
+		put_cred(cred);
+		if (!nfs4_server_rebooted(server->nfs_client)) {
 			iput(inode);
-			nfs_sb_deactive(server->super);
 			cond_resched();
 			goto restart;
 		}
+		nfs_inode_mark_test_expired_delegation(server,inode);
+		iput(inode);
+		return -EAGAIN;
 	}
 	rcu_read_unlock();
+	return 0;
+}
+
+/**
+ * nfs_reap_expired_delegations - reap expired delegations
+ * @clp: nfs_client to process
+ *
+ * Iterates through all the delegations associated with this server and
+ * checks if they have may have been revoked. This function is usually
+ * expected to be called in cases where the server may have lost its
+ * lease.
+ */
+void nfs_reap_expired_delegations(struct nfs_client *clp)
+{
+	nfs_client_for_each_server(clp, nfs_server_reap_expired_delegations,
+			NULL);
 }
 
 void nfs_inode_find_delegation_state_and_recover(struct inode *inode,
-- 
2.24.1

[PATCH 7/7] NFSv4: Add support for CB_RECALL_ANY for flexfiles layouts

[Trond Myklebust]

When we receive a CB_RECALL_ANY that asks us to return flexfiles
layouts, we iterate through all the layouts and look at whether or
not there are active open file descriptors that might need them
for I/O. If there are no such descriptors, we return the layouts.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/callback.h      |   4 +-
 fs/nfs/callback_proc.c |  13 ++++
 fs/nfs/nfs4_fs.h       |   4 +-
 fs/nfs/nfs4state.c     |  24 ++++++-
 fs/nfs/nfs4trace.h     |   8 ++-
 fs/nfs/pnfs.c          | 148 +++++++++++++++++++++++++++++++++++++----
 fs/nfs/pnfs.h          |   3 +
 7 files changed, 186 insertions(+), 18 deletions(-)

diff --git a/fs/nfs/callback.h b/fs/nfs/callback.h
index 549350259840..6a2033131c06 100644
--- a/fs/nfs/callback.h
+++ b/fs/nfs/callback.h
@@ -127,7 +127,9 @@ extern __be32 nfs4_callback_sequence(void *argp, void *resp,
 #define RCA4_TYPE_MASK_OBJ_LAYOUT_MAX  9
 #define RCA4_TYPE_MASK_OTHER_LAYOUT_MIN 12
 #define RCA4_TYPE_MASK_OTHER_LAYOUT_MAX 15
-#define RCA4_TYPE_MASK_ALL 0xf31f
+#define PNFS_FF_RCA4_TYPE_MASK_READ 16
+#define PNFS_FF_RCA4_TYPE_MASK_RW 17
+#define RCA4_TYPE_MASK_ALL 0x3f31f
 
 struct cb_recallanyargs {
 	uint32_t	craa_objs_to_keep;
diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c
index 97084804a953..e61dbc9b86ae 100644
--- a/fs/nfs/callback_proc.c
+++ b/fs/nfs/callback_proc.c
@@ -597,6 +597,7 @@ __be32 nfs4_callback_recallany(void *argp, void *resp,
 	struct cb_recallanyargs *args = argp;
 	__be32 status;
 	fmode_t flags = 0;
+	bool schedule_manager = false;
 
 	status = cpu_to_be32(NFS4ERR_OP_NOT_IN_SESSION);
 	if (!cps->clp) /* set in cb_sequence */
@@ -619,6 +620,18 @@ __be32 nfs4_callback_recallany(void *argp, void *resp,
 
 	if (args->craa_type_mask & BIT(RCA4_TYPE_MASK_FILE_LAYOUT))
 		pnfs_recall_all_layouts(cps->clp);
+
+	if (args->craa_type_mask & BIT(PNFS_FF_RCA4_TYPE_MASK_READ)) {
+		set_bit(NFS4CLNT_RECALL_ANY_LAYOUT_READ, &cps->clp->cl_state);
+		schedule_manager = true;
+	}
+	if (args->craa_type_mask & BIT(PNFS_FF_RCA4_TYPE_MASK_RW)) {
+		set_bit(NFS4CLNT_RECALL_ANY_LAYOUT_RW, &cps->clp->cl_state);
+		schedule_manager = true;
+	}
+	if (schedule_manager)
+		nfs4_schedule_state_manager(cps->clp);
+
 out:
 	dprintk("%s: exit with status = %d\n", __func__, ntohl(status));
 	return status;
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index 8be1ba7c62bb..2b7f6dcd2eb8 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -42,7 +42,9 @@ enum nfs4_client_state {
 	NFS4CLNT_LEASE_MOVED,
 	NFS4CLNT_DELEGATION_EXPIRED,
 	NFS4CLNT_RUN_MANAGER,
-	NFS4CLNT_DELEGRETURN_RUNNING,
+	NFS4CLNT_RECALL_RUNNING,
+	NFS4CLNT_RECALL_ANY_LAYOUT_READ,
+	NFS4CLNT_RECALL_ANY_LAYOUT_RW,
 };
 
 #define NFS4_RENEW_TIMEOUT		0x01
diff --git a/fs/nfs/nfs4state.c b/fs/nfs/nfs4state.c
index f7723d221945..ac93715c05a4 100644
--- a/fs/nfs/nfs4state.c
+++ b/fs/nfs/nfs4state.c
@@ -2524,6 +2524,21 @@ static int nfs4_bind_conn_to_session(struct nfs_client *clp)
 	}
 	return 0;
 }
+
+static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
+{
+	int iomode = 0;
+
+	if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_READ, &clp->cl_state))
+		iomode += IOMODE_READ;
+	if (test_and_clear_bit(NFS4CLNT_RECALL_ANY_LAYOUT_RW, &clp->cl_state))
+		iomode += IOMODE_RW;
+	/* Note: IOMODE_READ + IOMODE_RW == IOMODE_ANY */
+	if (iomode) {
+		pnfs_layout_return_unused_byclid(clp, iomode);
+		set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
+	}
+}
 #else /* CONFIG_NFS_V4_1 */
 static int nfs4_reset_session(struct nfs_client *clp) { return 0; }
 
@@ -2531,6 +2546,10 @@ static int nfs4_bind_conn_to_session(struct nfs_client *clp)
 {
 	return 0;
 }
+
+static void nfs4_layoutreturn_any_run(struct nfs_client *clp)
+{
+}
 #endif /* CONFIG_NFS_V4_1 */
 
 static void nfs4_state_manager(struct nfs_client *clp)
@@ -2635,12 +2654,13 @@ static void nfs4_state_manager(struct nfs_client *clp)
 		nfs4_end_drain_session(clp);
 		nfs4_clear_state_manager_bit(clp);
 
-		if (!test_and_set_bit(NFS4CLNT_DELEGRETURN_RUNNING, &clp->cl_state)) {
+		if (!test_and_set_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state)) {
 			if (test_and_clear_bit(NFS4CLNT_DELEGRETURN, &clp->cl_state)) {
 				nfs_client_return_marked_delegations(clp);
 				set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
 			}
-			clear_bit(NFS4CLNT_DELEGRETURN_RUNNING, &clp->cl_state);
+			nfs4_layoutreturn_any_run(clp);
+			clear_bit(NFS4CLNT_RECALL_RUNNING, &clp->cl_state);
 		}
 
 		/* Did we race with an attempt to give us more work? */
diff --git a/fs/nfs/nfs4trace.h b/fs/nfs/nfs4trace.h
index 1e97e5e04cb4..543541173a3d 100644
--- a/fs/nfs/nfs4trace.h
+++ b/fs/nfs/nfs4trace.h
@@ -584,7 +584,9 @@ TRACE_DEFINE_ENUM(NFS4CLNT_MOVED);
 TRACE_DEFINE_ENUM(NFS4CLNT_LEASE_MOVED);
 TRACE_DEFINE_ENUM(NFS4CLNT_DELEGATION_EXPIRED);
 TRACE_DEFINE_ENUM(NFS4CLNT_RUN_MANAGER);
-TRACE_DEFINE_ENUM(NFS4CLNT_DELEGRETURN_RUNNING);
+TRACE_DEFINE_ENUM(NFS4CLNT_RECALL_RUNNING);
+TRACE_DEFINE_ENUM(NFS4CLNT_RECALL_ANY_LAYOUT_READ);
+TRACE_DEFINE_ENUM(NFS4CLNT_RECALL_ANY_LAYOUT_RW);
 
 #define show_nfs4_clp_state(state) \
 	__print_flags(state, "|", \
@@ -605,7 +607,9 @@ TRACE_DEFINE_ENUM(NFS4CLNT_DELEGRETURN_RUNNING);
 		{ NFS4CLNT_LEASE_MOVED,		"LEASE_MOVED" }, \
 		{ NFS4CLNT_DELEGATION_EXPIRED,	"DELEGATION_EXPIRED" }, \
 		{ NFS4CLNT_RUN_MANAGER,		"RUN_MANAGER" }, \
-		{ NFS4CLNT_DELEGRETURN_RUNNING,	"DELEGRETURN_RUNNING" })
+		{ NFS4CLNT_RECALL_RUNNING,	"RECALL_RUNNING" }, \
+		{ NFS4CLNT_RECALL_ANY_LAYOUT_READ, "RECALL_ANY_LAYOUT_READ" }, \
+		{ NFS4CLNT_RECALL_ANY_LAYOUT_RW, "RECALL_ANY_LAYOUT_RW" })
 
 TRACE_EVENT(nfs4_state_mgr,
 		TP_PROTO(
diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 268e7b9ff54e..6b25117fca5f 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -309,6 +309,16 @@ pnfs_put_layout_hdr(struct pnfs_layout_hdr *lo)
 	}
 }
 
+static struct inode *
+pnfs_grab_inode_layout_hdr(struct pnfs_layout_hdr *lo)
+{
+	struct inode *inode = igrab(lo->plh_inode);
+	if (inode)
+		return inode;
+	set_bit(NFS_LAYOUT_INODE_FREEING, &lo->plh_flags);
+	return NULL;
+}
+
 static void
 pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 			 u32 seq)
@@ -782,7 +792,7 @@ pnfs_layout_bulk_destroy_byserver_locked(struct nfs_client *clp,
 		/* If the sb is being destroyed, just bail */
 		if (!nfs_sb_active(server->super))
 			break;
-		inode = igrab(lo->plh_inode);
+		inode = pnfs_grab_inode_layout_hdr(lo);
 		if (inode != NULL) {
 			if (test_and_clear_bit(NFS_LAYOUT_HASHED, &lo->plh_flags))
 				list_del_rcu(&lo->plh_layouts);
@@ -795,7 +805,6 @@ pnfs_layout_bulk_destroy_byserver_locked(struct nfs_client *clp,
 		} else {
 			rcu_read_unlock();
 			spin_unlock(&clp->cl_lock);
-			set_bit(NFS_LAYOUT_INODE_FREEING, &lo->plh_flags);
 		}
 		nfs_sb_deactive(server->super);
 		spin_lock(&clp->cl_lock);
@@ -2434,29 +2443,26 @@ pnfs_mark_matching_lsegs_return(struct pnfs_layout_hdr *lo,
 	return -ENOENT;
 }
 
-void pnfs_error_mark_layout_for_return(struct inode *inode,
-				       struct pnfs_layout_segment *lseg)
+static void
+pnfs_mark_layout_for_return(struct inode *inode,
+			    const struct pnfs_layout_range *range)
 {
-	struct pnfs_layout_hdr *lo = NFS_I(inode)->layout;
-	struct pnfs_layout_range range = {
-		.iomode = lseg->pls_range.iomode,
-		.offset = 0,
-		.length = NFS4_MAX_UINT64,
-	};
+	struct pnfs_layout_hdr *lo;
 	bool return_now = false;
 
 	spin_lock(&inode->i_lock);
+	lo = NFS_I(inode)->layout;
 	if (!pnfs_layout_is_valid(lo)) {
 		spin_unlock(&inode->i_lock);
 		return;
 	}
-	pnfs_set_plh_return_info(lo, range.iomode, 0);
+	pnfs_set_plh_return_info(lo, range->iomode, 0);
 	/*
 	 * 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.
 	 */
-	if (pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs, &range, 0) != -EBUSY) {
+	if (pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs, range, 0) != -EBUSY) {
 		nfs4_stateid stateid;
 		enum pnfs_iomode iomode;
 
@@ -2469,8 +2475,126 @@ void pnfs_error_mark_layout_for_return(struct inode *inode,
 		nfs_commit_inode(inode, 0);
 	}
 }
+
+void pnfs_error_mark_layout_for_return(struct inode *inode,
+				       struct pnfs_layout_segment *lseg)
+{
+	struct pnfs_layout_range range = {
+		.iomode = lseg->pls_range.iomode,
+		.offset = 0,
+		.length = NFS4_MAX_UINT64,
+	};
+
+	pnfs_mark_layout_for_return(inode, &range);
+}
 EXPORT_SYMBOL_GPL(pnfs_error_mark_layout_for_return);
 
+static bool
+pnfs_layout_can_be_returned(struct pnfs_layout_hdr *lo)
+{
+	return pnfs_layout_is_valid(lo) &&
+		!test_bit(NFS_LAYOUT_INODE_FREEING, &lo->plh_flags) &&
+		!test_bit(NFS_LAYOUT_RETURN, &lo->plh_flags);
+}
+
+static struct pnfs_layout_segment *
+pnfs_find_first_lseg(struct pnfs_layout_hdr *lo,
+		     const struct pnfs_layout_range *range,
+		     enum pnfs_iomode iomode)
+{
+	struct pnfs_layout_segment *lseg;
+
+	list_for_each_entry(lseg, &lo->plh_segs, pls_list) {
+		if (!test_bit(NFS_LSEG_VALID, &lseg->pls_flags))
+			continue;
+		if (test_bit(NFS_LSEG_LAYOUTRETURN, &lseg->pls_flags))
+			continue;
+		if (lseg->pls_range.iomode != iomode && iomode != IOMODE_ANY)
+			continue;
+		if (pnfs_lseg_range_intersecting(&lseg->pls_range, range))
+			return lseg;
+	}
+	return NULL;
+}
+
+/* Find open file states whose mode matches that of the range */
+static bool
+pnfs_should_return_unused_layout(struct pnfs_layout_hdr *lo,
+				 const struct pnfs_layout_range *range)
+{
+	struct list_head *head;
+	struct nfs_open_context *ctx;
+	fmode_t mode = 0;
+
+	if (!pnfs_layout_can_be_returned(lo) ||
+	    !pnfs_find_first_lseg(lo, range, range->iomode))
+		return false;
+
+	head = &NFS_I(lo->plh_inode)->open_files;
+	list_for_each_entry_rcu(ctx, head, list) {
+		if (ctx->state)
+			mode |= ctx->state->state & (FMODE_READ|FMODE_WRITE);
+	}
+
+	switch (range->iomode) {
+	default:
+		break;
+	case IOMODE_READ:
+		mode &= ~FMODE_WRITE;
+		break;
+	case IOMODE_RW:
+		if (pnfs_find_first_lseg(lo, range, IOMODE_READ))
+			mode &= ~FMODE_READ;
+	}
+	return mode == 0;
+}
+
+static int
+pnfs_layout_return_unused_byserver(struct nfs_server *server, void *data)
+{
+	const struct pnfs_layout_range *range = data;
+	struct pnfs_layout_hdr *lo;
+	struct inode *inode;
+restart:
+	rcu_read_lock();
+	list_for_each_entry_rcu(lo, &server->layouts, plh_layouts) {
+		if (!pnfs_layout_can_be_returned(lo) ||
+		    test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags))
+			continue;
+		inode = lo->plh_inode;
+		spin_lock(&inode->i_lock);
+		if (!pnfs_should_return_unused_layout(lo, range)) {
+			spin_unlock(&inode->i_lock);
+			continue;
+		}
+		spin_unlock(&inode->i_lock);
+		inode = pnfs_grab_inode_layout_hdr(lo);
+		if (!inode)
+			continue;
+		rcu_read_unlock();
+		pnfs_mark_layout_for_return(inode, range);
+		iput(inode);
+		cond_resched();
+		goto restart;
+	}
+	rcu_read_unlock();
+	return 0;
+}
+
+void
+pnfs_layout_return_unused_byclid(struct nfs_client *clp,
+				 enum pnfs_iomode iomode)
+{
+	struct pnfs_layout_range range = {
+		.iomode = iomode,
+		.offset = 0,
+		.length = NFS4_MAX_UINT64,
+	};
+
+	nfs_client_for_each_server(clp, pnfs_layout_return_unused_byserver,
+			&range);
+}
+
 void
 pnfs_generic_pg_check_layout(struct nfs_pageio_descriptor *pgio)
 {
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index 8df9aa02d336..7bfb6970134a 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -329,6 +329,9 @@ int pnfs_write_done_resend_to_mds(struct nfs_pgio_header *);
 struct nfs4_threshold *pnfs_mdsthreshold_alloc(void);
 void pnfs_error_mark_layout_for_return(struct inode *inode,
 				       struct pnfs_layout_segment *lseg);
+void pnfs_layout_return_unused_byclid(struct nfs_client *clp,
+				      enum pnfs_iomode iomode);
+
 /* nfs4_deviceid_flags */
 enum {
 	NFS_DEVICEID_INVALID = 0,       /* set when MDS clientid recalled */
-- 
2.24.1