ocfs2: fix rare stale inode errors when exporting via nfs

For nfs exporting, ocfs2_get_dentry() returns the dentry for fh.
ocfs2_get_dentry() may read from disk when the inode is not in memory,
without any cross cluster lock. this leads to the file system loading a
stale inode.

This patch fixes above problem.

Solution is that in case of inode is not in memory, we get the cluster
lock(PR) of alloc inode where the inode in question is allocated from (this
causes node on which deletion is done sync the alloc inode) before reading
out the inode itsself. then we check the bitmap in the group (the inode in
question allcated from) to see if the bit is clear. if it's clear then it's
stale. if the bit is set, we then check generation as the existing code
does.

We have to read out the inode in question from disk first to know its alloc
slot and allot bit. And if its not stale we read it out using ocfs2_iget().
The second read should then be from cache.

And also we have to add a per superblock nfs_sync_lock to cover the lock for
alloc inode and that for inode in question. this is because ocfs2_get_dentry()
and ocfs2_delete_inode() lock on them in reverse order. nfs_sync_lock is locked
in EX mode in ocfs2_get_dentry() and in PR mode in ocfs2_delete_inode(). so
that mutliple ocfs2_delete_inode() can run concurrently in normal case.

[mfasheh@suse.com: build warning fixes and comment cleanups]
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Acked-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>

9 files changed, 319 insertions, 8 deletions

diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 7219a86..e15fc7d 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -244,6 +244,10 @@ static struct ocfs2_lock_res_ops ocfs2_rename_lops = {
 	.flags		= 0,
 };
 
+static struct ocfs2_lock_res_ops ocfs2_nfs_sync_lops = {
+	.flags		= 0,
+};
+
 static struct ocfs2_lock_res_ops ocfs2_dentry_lops = {
 	.get_osb	= ocfs2_get_dentry_osb,
 	.post_unlock	= ocfs2_dentry_post_unlock,
@@ -622,6 +626,17 @@ static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
 				   &ocfs2_rename_lops, osb);
 }
 
+static void ocfs2_nfs_sync_lock_res_init(struct ocfs2_lock_res *res,
+					 struct ocfs2_super *osb)
+{
+	/* nfs_sync lockres doesn't come from a slab so we call init
+	 * once on it manually.  */
+	ocfs2_lock_res_init_once(res);
+	ocfs2_build_lock_name(OCFS2_LOCK_TYPE_NFS_SYNC, 0, 0, res->l_name);
+	ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_NFS_SYNC,
+				   &ocfs2_nfs_sync_lops, osb);
+}
+
 void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres,
 			      struct ocfs2_file_private *fp)
 {
@@ -2417,6 +2432,34 @@ void ocfs2_rename_unlock(struct ocfs2_super *osb)
 		ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX);
 }
 
+int ocfs2_nfs_sync_lock(struct ocfs2_super *osb, int ex)
+{
+	int status;
+	struct ocfs2_lock_res *lockres = &osb->osb_nfs_sync_lockres;
+
+	if (ocfs2_is_hard_readonly(osb))
+		return -EROFS;
+
+	if (ocfs2_mount_local(osb))
+		return 0;
+
+	status = ocfs2_cluster_lock(osb, lockres, ex ? LKM_EXMODE : LKM_PRMODE,
+				    0, 0);
+	if (status < 0)
+		mlog(ML_ERROR, "lock on nfs sync lock failed %d\n", status);
+
+	return status;
+}
+
+void ocfs2_nfs_sync_unlock(struct ocfs2_super *osb, int ex)
+{
+	struct ocfs2_lock_res *lockres = &osb->osb_nfs_sync_lockres;
+
+	if (!ocfs2_mount_local(osb))
+		ocfs2_cluster_unlock(osb, lockres,
+				     ex ? LKM_EXMODE : LKM_PRMODE);
+}
+
 int ocfs2_dentry_lock(struct dentry *dentry, int ex)
 {
 	int ret;
@@ -2798,6 +2841,7 @@ int ocfs2_dlm_init(struct ocfs2_super *osb)
 local:
 	ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
 	ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
+	ocfs2_nfs_sync_lock_res_init(&osb->osb_nfs_sync_lockres, osb);
 
 	osb->cconn = conn;
 
@@ -2833,6 +2877,7 @@ void ocfs2_dlm_shutdown(struct ocfs2_super *osb,
 
 	ocfs2_lock_res_free(&osb->osb_super_lockres);
 	ocfs2_lock_res_free(&osb->osb_rename_lockres);
+	ocfs2_lock_res_free(&osb->osb_nfs_sync_lockres);
 
 	ocfs2_cluster_disconnect(osb->cconn, hangup_pending);
 	osb->cconn = NULL;
@@ -3015,6 +3060,7 @@ static void ocfs2_drop_osb_locks(struct ocfs2_super *osb)
 {
 	ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres);
 	ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres);
+	ocfs2_simple_drop_lockres(osb, &osb->osb_nfs_sync_lockres);
 }
 
 int ocfs2_drop_inode_locks(struct inode *inode)
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
index 3f8d998..e1fd572 100644
--- a/fs/ocfs2/dlmglue.h
+++ b/fs/ocfs2/dlmglue.h
@@ -115,6 +115,8 @@ void ocfs2_super_unlock(struct ocfs2_super *osb,
 			int ex);
 int ocfs2_rename_lock(struct ocfs2_super *osb);
 void ocfs2_rename_unlock(struct ocfs2_super *osb);
+int ocfs2_nfs_sync_lock(struct ocfs2_super *osb, int ex);
+void ocfs2_nfs_sync_unlock(struct ocfs2_super *osb, int ex);
 int ocfs2_dentry_lock(struct dentry *dentry, int ex);
 void ocfs2_dentry_unlock(struct dentry *dentry, int ex);
 int ocfs2_file_lock(struct file *file, int ex, int trylock);
diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c
index 2f27b33..de3da8e 100644
--- a/fs/ocfs2/export.c
+++ b/fs/ocfs2/export.c
@@ -31,6 +31,7 @@
 
 #include "ocfs2.h"
 
+#include "alloc.h"
 #include "dir.h"
 #include "dlmglue.h"
 #include "dcache.h"
@@ -38,6 +39,7 @@
 #include "inode.h"
 
 #include "buffer_head_io.h"
+#include "suballoc.h"
 
 struct ocfs2_inode_handle
 {
@@ -49,29 +51,97 @@ static struct dentry *ocfs2_get_dentry(struct super_block *sb,
 		struct ocfs2_inode_handle *handle)
 {
 	struct inode *inode;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	u64 blkno = handle->ih_blkno;
+	int status, set;
 	struct dentry *result;
 
 	mlog_entry("(0x%p, 0x%p)\n", sb, handle);
 
-	if (handle->ih_blkno == 0) {
-		mlog_errno(-ESTALE);
-		return ERR_PTR(-ESTALE);
+	if (blkno == 0) {
+		mlog(0, "nfs wants inode with blkno: 0\n");
+		result = ERR_PTR(-ESTALE);
+		goto bail;
+	}
+
+	inode = ocfs2_ilookup(sb, blkno);
+	/*
+	 * If the inode exists in memory, we only need to check it's
+	 * generation number
+	 */
+	if (inode)
+		goto check_gen;
+
+	/*
+	 * This will synchronize us against ocfs2_delete_inode() on
+	 * all nodes
+	 */
+	status = ocfs2_nfs_sync_lock(osb, 1);
+	if (status < 0) {
+		mlog(ML_ERROR, "getting nfs sync lock(EX) failed %d\n", status);
+		goto check_err;
+	}
+
+	status = ocfs2_test_inode_bit(osb, blkno, &set);
+	if (status < 0) {
+		if (status == -EINVAL) {
+			/*
+			 * The blkno NFS gave us doesn't even show up
+			 * as an inode, we return -ESTALE to be
+			 * nice
+			 */
+			mlog(0, "test inode bit failed %d\n", status);
+			status = -ESTALE;
+		} else {
+			mlog(ML_ERROR, "test inode bit failed %d\n", status);
+		}
+		goto unlock_nfs_sync;
+	}
+
+	/* If the inode allocator bit is clear, this inode must be stale */
+	if (!set) {
+		mlog(0, "inode %llu suballoc bit is clear\n", blkno);
+		status = -ESTALE;
+		goto unlock_nfs_sync;
 	}
 
-	inode = ocfs2_iget(OCFS2_SB(sb), handle->ih_blkno, 0, 0);
+	inode = ocfs2_iget(osb, blkno, 0, 0);
 
-	if (IS_ERR(inode))
-		return (void *)inode;
+unlock_nfs_sync:
+	ocfs2_nfs_sync_unlock(osb, 1);
 
+check_err:
+	if (status < 0) {
+		if (status == -ESTALE) {
+			mlog(0, "stale inode ino: %llu generation: %u\n",
+			     blkno, handle->ih_generation);
+		}
+		result = ERR_PTR(status);
+		goto bail;
+	}
+
+	if (IS_ERR(inode)) {
+		mlog_errno(PTR_ERR(inode));
+		result = (void *)inode;
+		goto bail;
+	}
+
+check_gen:
 	if (handle->ih_generation != inode->i_generation) {
 		iput(inode);
-		return ERR_PTR(-ESTALE);
+		mlog(0, "stale inode ino: %llu generation: %u\n", blkno,
+		     handle->ih_generation);
+		result = ERR_PTR(-ESTALE);
+		goto bail;
 	}
 
 	result = d_obtain_alias(inode);
 	if (!IS_ERR(result))
 		result->d_op = &ocfs2_dentry_ops;
+	else
+		mlog_errno(PTR_ERR(result));
 
+bail:
 	mlog_exit_ptr(result);
 	return result;
 }
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
index 4a88bce..10e1fa87 100644
--- a/fs/ocfs2/inode.c
+++ b/fs/ocfs2/inode.c
@@ -113,6 +113,17 @@ void ocfs2_get_inode_flags(struct ocfs2_inode_info *oi)
 		oi->ip_attr |= OCFS2_DIRSYNC_FL;
 }
 
+struct inode *ocfs2_ilookup(struct super_block *sb, u64 blkno)
+{
+	struct ocfs2_find_inode_args args;
+
+	args.fi_blkno = blkno;
+	args.fi_flags = 0;
+	args.fi_ino = ino_from_blkno(sb, blkno);
+	args.fi_sysfile_type = 0;
+
+	return ilookup5(sb, blkno, ocfs2_find_actor, &args);
+}
 struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags,
 			 int sysfile_type)
 {
@@ -961,6 +972,17 @@ void ocfs2_delete_inode(struct inode *inode)
 		goto bail;
 	}
 
+	/*
+	 * Synchronize us against ocfs2_get_dentry. We take this in
+	 * shared mode so that all nodes can still concurrently
+	 * process deletes.
+	 */
+	status = ocfs2_nfs_sync_lock(OCFS2_SB(inode->i_sb), 0);
+	if (status < 0) {
+		mlog(ML_ERROR, "getting nfs sync lock(PR) failed %d\n", status);
+		ocfs2_cleanup_delete_inode(inode, 0);
+		goto bail_unblock;
+	}
 	/* Lock down the inode. This gives us an up to date view of
 	 * it's metadata (for verification), and allows us to
 	 * serialize delete_inode on multiple nodes.
@@ -974,7 +996,7 @@ void ocfs2_delete_inode(struct inode *inode)
 		if (status != -ENOENT)
 			mlog_errno(status);
 		ocfs2_cleanup_delete_inode(inode, 0);
-		goto bail_unblock;
+		goto bail_unlock_nfs_sync;
 	}
 
 	/* Query the cluster. This will be the final decision made
@@ -1017,6 +1039,10 @@ void ocfs2_delete_inode(struct inode *inode)
 bail_unlock_inode:
 	ocfs2_inode_unlock(inode, 1);
 	brelse(di_bh);
+
+bail_unlock_nfs_sync:
+	ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0);
+
 bail_unblock:
 	status = sigprocmask(SIG_SETMASK, &oldset, NULL);
 	if (status < 0)
diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
index e1978ac..ea71525 100644
--- a/fs/ocfs2/inode.h
+++ b/fs/ocfs2/inode.h
@@ -128,6 +128,7 @@ void ocfs2_drop_inode(struct inode *inode);
 /* Flags for ocfs2_iget() */
 #define OCFS2_FI_FLAG_SYSFILE		0x1
 #define OCFS2_FI_FLAG_ORPHAN_RECOVERY	0x2
+struct inode *ocfs2_ilookup(struct super_block *sb, u64 feoff);
 struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 feoff, unsigned flags,
 			 int sysfile_type);
 int ocfs2_inode_init_private(struct inode *inode);
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index b65d19c..558bd27 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -300,6 +300,7 @@ struct ocfs2_super
 	struct ocfs2_cluster_connection *cconn;
 	struct ocfs2_lock_res osb_super_lockres;
 	struct ocfs2_lock_res osb_rename_lockres;
+	struct ocfs2_lock_res osb_nfs_sync_lockres;
 	struct ocfs2_dlm_debug *osb_dlm_debug;
 
 	struct dentry *osb_debug_root;
diff --git a/fs/ocfs2/ocfs2_lockid.h b/fs/ocfs2/ocfs2_lockid.h
index eb6f50c..a53ce87 100644
--- a/fs/ocfs2/ocfs2_lockid.h
+++ b/fs/ocfs2/ocfs2_lockid.h
@@ -47,6 +47,7 @@ enum ocfs2_lock_type {
 	OCFS2_LOCK_TYPE_OPEN,
 	OCFS2_LOCK_TYPE_FLOCK,
 	OCFS2_LOCK_TYPE_QINFO,
+	OCFS2_LOCK_TYPE_NFS_SYNC,
 	OCFS2_NUM_LOCK_TYPES
 };
 
@@ -81,6 +82,9 @@ static inline char ocfs2_lock_type_char(enum ocfs2_lock_type type)
 		case OCFS2_LOCK_TYPE_QINFO:
 			c = 'Q';
 			break;
+		case OCFS2_LOCK_TYPE_NFS_SYNC:
+			c = 'Y';
+			break;
 		default:
 			c = '\0';
 	}
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index 4c1399c..b4ca591 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -2185,3 +2185,162 @@ out:
 
 	return ret;
 }
+
+/*
+ * Read the inode specified by blkno to get suballoc_slot and
+ * suballoc_bit.
+ */
+static int ocfs2_get_suballoc_slot_bit(struct ocfs2_super *osb, u64 blkno,
+				       u16 *suballoc_slot, u16 *suballoc_bit)
+{
+	int status;
+	struct buffer_head *inode_bh = NULL;
+	struct ocfs2_dinode *inode_fe;
+
+	mlog_entry("blkno: %llu\n", blkno);
+
+	/* dirty read disk */
+	status = ocfs2_read_blocks_sync(osb, blkno, 1, &inode_bh);
+	if (status < 0) {
+		mlog(ML_ERROR, "read block %llu failed %d\n", blkno, status);
+		goto bail;
+	}
+
+	inode_fe = (struct ocfs2_dinode *) inode_bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(inode_fe)) {
+		mlog(ML_ERROR, "invalid inode %llu requested\n", blkno);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	if (le16_to_cpu(inode_fe->i_suballoc_slot) != OCFS2_INVALID_SLOT &&
+	    (u32)le16_to_cpu(inode_fe->i_suballoc_slot) > osb->max_slots - 1) {
+		mlog(ML_ERROR, "inode %llu has invalid suballoc slot %u\n",
+		     blkno, (u32)le16_to_cpu(inode_fe->i_suballoc_slot));
+		status = -EINVAL;
+		goto bail;
+	}
+
+	if (suballoc_slot)
+		*suballoc_slot = le16_to_cpu(inode_fe->i_suballoc_slot);
+	if (suballoc_bit)
+		*suballoc_bit = le16_to_cpu(inode_fe->i_suballoc_bit);
+
+bail:
+	brelse(inode_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * test whether bit is SET in allocator bitmap or not.  on success, 0
+ * is returned and *res is 1 for SET; 0 otherwise.  when fails, errno
+ * is returned and *res is meaningless.  Call this after you have
+ * cluster locked against suballoc, or you may get a result based on
+ * non-up2date contents
+ */
+static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
+				   struct inode *suballoc,
+				   struct buffer_head *alloc_bh, u64 blkno,
+				   u16 bit, int *res)
+{
+	struct ocfs2_dinode *alloc_fe;
+	struct ocfs2_group_desc *group;
+	struct buffer_head *group_bh = NULL;
+	u64 bg_blkno;
+	int status;
+
+	mlog_entry("blkno: %llu bit: %u\n", blkno, (unsigned int)bit);
+
+	alloc_fe = (struct ocfs2_dinode *)alloc_bh->b_data;
+	if ((bit + 1) > ocfs2_bits_per_group(&alloc_fe->id2.i_chain)) {
+		mlog(ML_ERROR, "suballoc bit %u out of range of %u\n",
+		     (unsigned int)bit,
+		     ocfs2_bits_per_group(&alloc_fe->id2.i_chain));
+		status = -EINVAL;
+		goto bail;
+	}
+
+	bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
+	status = ocfs2_read_group_descriptor(suballoc, alloc_fe, bg_blkno,
+					     &group_bh);
+	if (status < 0) {
+		mlog(ML_ERROR, "read group %llu failed %d\n", bg_blkno, status);
+		goto bail;
+	}
+
+	group = (struct ocfs2_group_desc *) group_bh->b_data;
+	*res = ocfs2_test_bit(bit, (unsigned long *)group->bg_bitmap);
+
+bail:
+	brelse(group_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Test if the bit representing this inode (blkno) is set in the
+ * suballocator.
+ *
+ * On success, 0 is returned and *res is 1 for SET; 0 otherwise.
+ *
+ * In the event of failure, a negative value is returned and *res is
+ * meaningless.
+ *
+ * Callers must make sure to hold nfs_sync_lock to prevent
+ * ocfs2_delete_inode() on another node from accessing the same
+ * suballocator concurrently.
+ */
+int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res)
+{
+	int status;
+	u16 suballoc_bit = 0, suballoc_slot = 0;
+	struct inode *inode_alloc_inode;
+	struct buffer_head *alloc_bh = NULL;
+
+	mlog_entry("blkno: %llu", blkno);
+
+	status = ocfs2_get_suballoc_slot_bit(osb, blkno, &suballoc_slot,
+					     &suballoc_bit);
+	if (status < 0) {
+		mlog(ML_ERROR, "get alloc slot and bit failed %d\n", status);
+		goto bail;
+	}
+
+	inode_alloc_inode =
+		ocfs2_get_system_file_inode(osb, INODE_ALLOC_SYSTEM_INODE,
+					    suballoc_slot);
+	if (!inode_alloc_inode) {
+		/* the error code could be inaccurate, but we are not able to
+		 * get the correct one. */
+		status = -EINVAL;
+		mlog(ML_ERROR, "unable to get alloc inode in slot %u\n",
+		     (u32)suballoc_slot);
+		goto bail;
+	}
+
+	mutex_lock(&inode_alloc_inode->i_mutex);
+	status = ocfs2_inode_lock(inode_alloc_inode, &alloc_bh, 0);
+	if (status < 0) {
+		mutex_unlock(&inode_alloc_inode->i_mutex);
+		mlog(ML_ERROR, "lock on alloc inode on slot %u failed %d\n",
+		     (u32)suballoc_slot, status);
+		goto bail;
+	}
+
+	status = ocfs2_test_suballoc_bit(osb, inode_alloc_inode, alloc_bh,
+					 blkno, suballoc_bit, res);
+	if (status < 0)
+		mlog(ML_ERROR, "test suballoc bit failed %d\n", status);
+
+	ocfs2_inode_unlock(inode_alloc_inode, 0);
+	mutex_unlock(&inode_alloc_inode->i_mutex);
+
+	iput(inode_alloc_inode);
+	brelse(alloc_bh);
+bail:
+	mlog_exit(status);
+	return status;
+}
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index ea85a4c..8c9a78a 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -188,4 +188,6 @@ int ocfs2_lock_allocators(struct inode *inode, struct ocfs2_extent_tree *et,
 			  u32 clusters_to_add, u32 extents_to_split,
 			  struct ocfs2_alloc_context **data_ac,
 			  struct ocfs2_alloc_context **meta_ac);
+
+int ocfs2_test_inode_bit(struct ocfs2_super *osb, u64 blkno, int *res);
 #endif /* _CHAINALLOC_H_ */