From 9cfa1098dcfb34f71c5f3b7bcdbbb435a0cecab2 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:26:18 -0700
Subject: ceph: use flag bit for at_end readdir flag

This saves us a word of memory per file.

Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index ef8f08c343e..53b441fe78f 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -133,7 +133,7 @@ more:
 		     d_unhashed(dentry) ? "!hashed" : "hashed",
 		     parent->d_subdirs.prev, parent->d_subdirs.next);
 		if (p == &parent->d_subdirs) {
-			fi->at_end = 1;
+			fi->flags |= CEPH_F_ATEND;
 			goto out_unlock;
 		}
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
@@ -234,7 +234,7 @@ static int ceph_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	const int max_bytes = fsc->mount_options->max_readdir_bytes;
 
 	dout("readdir %p filp %p frag %u off %u\n", inode, filp, frag, off);
-	if (fi->at_end)
+	if (fi->flags & CEPH_F_ATEND)
 		return 0;
 
 	/* always start with . and .. */
@@ -403,7 +403,7 @@ more:
 		dout("readdir next frag is %x\n", frag);
 		goto more;
 	}
-	fi->at_end = 1;
+	fi->flags |= CEPH_F_ATEND;
 
 	/*
 	 * if dir_release_count still matches the dir, no dentries
@@ -435,7 +435,7 @@ static void reset_readdir(struct ceph_file_info *fi)
 		dput(fi->dentry);
 		fi->dentry = NULL;
 	}
-	fi->at_end = 0;
+	fi->flags &= ~CEPH_F_ATEND;
 }
 
 static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int origin)
@@ -458,7 +458,7 @@ static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int origin)
 		if (offset != file->f_pos) {
 			file->f_pos = offset;
 			file->f_version = 0;
-			fi->at_end = 0;
+			fi->flags &= ~CEPH_F_ATEND;
 		}
 		retval = offset;
 
-- 
cgit v1.2.3-70-g09d2


From 468640e32c7f6bfdaaa011095cc388786755d159 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:28:11 -0700
Subject: ceph: fix ceph_lookup_open intent usage

We weren't properly calling lookup_instantiate_filp when setting up the
lookup intent, which could lead to file leakage on errors.  So:

 - use separate helper for the hidden snapdir translation, immediately
   following the mds request
 - use ceph_finish_lookup for the final dentry/return value dance in the
   exit path
 - lookup_instantiate_filp on success

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c   | 31 ++++++++++++++++++++-----------
 fs/ceph/file.c  | 23 +++++++++++++++--------
 fs/ceph/super.h |  2 ++
 3 files changed, 37 insertions(+), 19 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 53b441fe78f..f39a409db0e 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -482,18 +482,10 @@ static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int origin)
 }
 
 /*
- * Process result of a lookup/open request.
- *
- * Mainly, make sure we return the final req->r_dentry (if it already
- * existed) in place of the original VFS-provided dentry when they
- * differ.
- *
- * Gracefully handle the case where the MDS replies with -ENOENT and
- * no trace (which it may do, at its discretion, e.g., if it doesn't
- * care to issue a lease on the negative dentry).
+ * Handle lookups for the hidden .snap directory.
  */
-struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
-				  struct dentry *dentry, int err)
+int ceph_handle_snapdir(struct ceph_mds_request *req,
+			struct dentry *dentry, int err)
 {
 	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
 	struct inode *parent = dentry->d_parent->d_inode;
@@ -510,7 +502,23 @@ struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 		d_add(dentry, inode);
 		err = 0;
 	}
+	return err;
+}
 
+/*
+ * Figure out final result of a lookup/open request.
+ *
+ * Mainly, make sure we return the final req->r_dentry (if it already
+ * existed) in place of the original VFS-provided dentry when they
+ * differ.
+ *
+ * Gracefully handle the case where the MDS replies with -ENOENT and
+ * no trace (which it may do, at its discretion, e.g., if it doesn't
+ * care to issue a lease on the negative dentry).
+ */
+struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
+				  struct dentry *dentry, int err)
+{
 	if (err == -ENOENT) {
 		/* no trace? */
 		err = 0;
@@ -605,6 +613,7 @@ static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
 	req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
 	req->r_locked_dir = dir;
 	err = ceph_mdsc_do_request(mdsc, NULL, req);
+	err = ceph_handle_snapdir(req, dentry, err);
 	dentry = ceph_finish_lookup(req, dentry, err);
 	ceph_mdsc_put_request(req);  /* will dput(dentry) */
 	dout("lookup result=%p\n", dentry);
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index e0115eb4e9b..f34d47d66e7 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -223,8 +223,9 @@ struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 {
 	struct ceph_fs_client *fsc = ceph_sb_to_client(dir->i_sb);
 	struct ceph_mds_client *mdsc = fsc->mdsc;
-	struct file *file = nd->intent.open.file;
+	struct file *file;
 	struct ceph_mds_request *req;
+	struct dentry *ret;
 	int err;
 	int flags = nd->intent.open.flags - 1;  /* silly vfs! */
 
@@ -245,15 +246,21 @@ struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 	err = ceph_mdsc_do_request(mdsc,
 				   (flags & (O_CREAT|O_TRUNC)) ? dir : NULL,
 				   req);
-	dentry = ceph_finish_lookup(req, dentry, err);
-	if (!err && (flags & O_CREAT) && !req->r_reply_info.head->is_dentry)
+	err = ceph_handle_snapdir(req, dentry, err);
+	if (err)
+		goto out;
+	if ((flags & O_CREAT) && !req->r_reply_info.head->is_dentry)
 		err = ceph_handle_notrace_create(dir, dentry);
-	if (!err)
-		err = ceph_init_file(req->r_dentry->d_inode, file,
-				     req->r_fmode);
+	if (err)
+		goto out;
+	file = lookup_instantiate_filp(nd, req->r_dentry, ceph_open);
+	if (IS_ERR(file))
+		err = PTR_ERR(file);
+out:
+	ret = ceph_finish_lookup(req, dentry, err);
 	ceph_mdsc_put_request(req);
-	dout("ceph_lookup_open result=%p\n", dentry);
-	return dentry;
+	dout("ceph_lookup_open result=%p\n", ret);
+	return ret;
 }
 
 int ceph_release(struct inode *inode, struct file *file)
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index a8a27332024..c24891a5bec 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -791,6 +791,8 @@ extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 	ceph_snapdir_dentry_ops;
 
 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
+extern int ceph_handle_snapdir(struct ceph_mds_request *req,
+			       struct dentry *dentry, int err);
 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 					 struct dentry *dentry, int err);
 
-- 
cgit v1.2.3-70-g09d2


From 48d0cbd1242aac969560ef8b90f26ee3b09a6a5c Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:30:15 -0700
Subject: ceph: handle racing calls to ceph_init_dentry

The ->lookup() and prepopulate_readdir() callers are working with unhashed
dentries, so we don't have to worry.  The export.c callers, though, need
to initialize something they got back from d_obtain_alias() and are
potentially racing with other callers.  Make sure we don't return unless
the dentry is properly initialized (by us or someone else).

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index f39a409db0e..883c9546111 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -40,14 +40,6 @@ int ceph_init_dentry(struct dentry *dentry)
 	if (dentry->d_fsdata)
 		return 0;
 
-	if (dentry->d_parent == NULL ||   /* nfs fh_to_dentry */
-	    ceph_snap(dentry->d_parent->d_inode) == CEPH_NOSNAP)
-		d_set_d_op(dentry, &ceph_dentry_ops);
-	else if (ceph_snap(dentry->d_parent->d_inode) == CEPH_SNAPDIR)
-		d_set_d_op(dentry, &ceph_snapdir_dentry_ops);
-	else
-		d_set_d_op(dentry, &ceph_snap_dentry_ops);
-
 	di = kmem_cache_alloc(ceph_dentry_cachep, GFP_NOFS | __GFP_ZERO);
 	if (!di)
 		return -ENOMEM;          /* oh well */
@@ -58,10 +50,21 @@ int ceph_init_dentry(struct dentry *dentry)
 		kmem_cache_free(ceph_dentry_cachep, di);
 		goto out_unlock;
 	}
+
+	if (dentry->d_parent == NULL ||   /* nfs fh_to_dentry */
+	    ceph_snap(dentry->d_parent->d_inode) == CEPH_NOSNAP)
+		d_set_d_op(dentry, &ceph_dentry_ops);
+	else if (ceph_snap(dentry->d_parent->d_inode) == CEPH_SNAPDIR)
+		d_set_d_op(dentry, &ceph_snapdir_dentry_ops);
+	else
+		d_set_d_op(dentry, &ceph_snap_dentry_ops);
+
 	di->dentry = dentry;
 	di->lease_session = NULL;
-	dentry->d_fsdata = di;
 	dentry->d_time = jiffies;
+	/* avoid reordering d_fsdata setup so that the check above is safe */
+	smp_mb();
+	dentry->d_fsdata = di;
 	ceph_dentry_lru_add(dentry);
 out_unlock:
 	spin_unlock(&dentry->d_lock);
-- 
cgit v1.2.3-70-g09d2


From 5f21c96dd5c615341963036ae8f5e4f5227a818d Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:30:29 -0700
Subject: ceph: protect access to d_parent

d_parent is protected by d_lock: use it when looking up a dentry's parent
directory inode.  Also take a reference and drop it in the caller to avoid
a use-after-free.

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c   | 15 +++++++++++++++
 fs/ceph/file.c  |  8 +++++---
 fs/ceph/inode.c |  4 +++-
 fs/ceph/ioctl.c |  4 +++-
 fs/ceph/super.h |  9 +--------
 fs/ceph/xattr.c |  8 ++++++--
 6 files changed, 33 insertions(+), 15 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 883c9546111..ed296ec121d 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -71,6 +71,21 @@ out_unlock:
 	return 0;
 }
 
+struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry)
+{
+	struct inode *inode = NULL;
+
+	if (!dentry)
+		return NULL;
+
+	spin_lock(&dentry->d_lock);
+	if (dentry->d_parent) {
+		inode = dentry->d_parent->d_inode;
+		ihold(inode);
+	}
+	spin_unlock(&dentry->d_lock);
+	return inode;
+}
 
 
 /*
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index f34d47d66e7..45fbd69daab 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -122,7 +122,7 @@ int ceph_open(struct inode *inode, struct file *file)
 	struct ceph_mds_client *mdsc = fsc->mdsc;
 	struct ceph_mds_request *req;
 	struct ceph_file_info *cf = file->private_data;
-	struct inode *parent_inode = file->f_dentry->d_parent->d_inode;
+	struct inode *parent_inode = NULL;
 	int err;
 	int flags, fmode, wanted;
 
@@ -194,8 +194,10 @@ int ceph_open(struct inode *inode, struct file *file)
 	req->r_inode = inode;
 	ihold(inode);
 	req->r_num_caps = 1;
-	err = ceph_mdsc_do_request(mdsc, (flags & (O_CREAT|O_TRUNC)) ?
-				   parent_inode : NULL, req);
+	if (flags & (O_CREAT|O_TRUNC))
+		parent_inode = ceph_get_dentry_parent_inode(file->f_dentry);
+	err = ceph_mdsc_do_request(mdsc, parent_inode, req);
+	iput(parent_inode);
 	if (!err)
 		err = ceph_init_file(inode, file, req->r_fmode);
 	ceph_mdsc_put_request(req);
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 2717dc4e443..a7db56f1523 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1562,7 +1562,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
 {
 	struct inode *inode = dentry->d_inode;
 	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct inode *parent_inode = dentry->d_parent->d_inode;
+	struct inode *parent_inode;
 	const unsigned int ia_valid = attr->ia_valid;
 	struct ceph_mds_request *req;
 	struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc;
@@ -1745,7 +1745,9 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
 		req->r_inode_drop = release;
 		req->r_args.setattr.mask = cpu_to_le32(mask);
 		req->r_num_caps = 1;
+		parent_inode = ceph_get_dentry_parent_inode(dentry);
 		err = ceph_mdsc_do_request(mdsc, parent_inode, req);
+		iput(parent_inode);
 	}
 	dout("setattr %p result=%d (%s locally, %d remote)\n", inode, err,
 	     ceph_cap_string(dirtied), mask);
diff --git a/fs/ceph/ioctl.c b/fs/ceph/ioctl.c
index a757a568057..3b256b50f7d 100644
--- a/fs/ceph/ioctl.c
+++ b/fs/ceph/ioctl.c
@@ -38,7 +38,7 @@ static long ceph_ioctl_get_layout(struct file *file, void __user *arg)
 static long ceph_ioctl_set_layout(struct file *file, void __user *arg)
 {
 	struct inode *inode = file->f_dentry->d_inode;
-	struct inode *parent_inode = file->f_dentry->d_parent->d_inode;
+	struct inode *parent_inode;
 	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
 	struct ceph_mds_request *req;
 	struct ceph_ioctl_layout l;
@@ -87,7 +87,9 @@ static long ceph_ioctl_set_layout(struct file *file, void __user *arg)
 	req->r_args.setlayout.layout.fl_pg_preferred =
 		cpu_to_le32(l.preferred_osd);
 
+	parent_inode = ceph_get_dentry_parent_inode(file->f_dentry);
 	err = ceph_mdsc_do_request(mdsc, parent_inode, req);
+	iput(parent_inode);
 	ceph_mdsc_put_request(req);
 	return err;
 }
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index c24891a5bec..c1eb9a014b7 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -801,6 +801,7 @@ extern void ceph_dentry_lru_touch(struct dentry *dn);
 extern void ceph_dentry_lru_del(struct dentry *dn);
 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 extern unsigned ceph_dentry_hash(struct dentry *dn);
+extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
 
 /*
  * our d_ops vary depending on whether the inode is live,
@@ -823,14 +824,6 @@ extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 			     int p_locks, int f_locks);
 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 
-static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
-{
-	if (dentry && dentry->d_parent)
-		return dentry->d_parent->d_inode;
-
-	return NULL;
-}
-
 /* debugfs.c */
 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c
index f42d730f1b6..96c6739a028 100644
--- a/fs/ceph/xattr.c
+++ b/fs/ceph/xattr.c
@@ -629,7 +629,7 @@ static int ceph_sync_setxattr(struct dentry *dentry, const char *name,
 	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
 	struct inode *inode = dentry->d_inode;
 	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct inode *parent_inode = dentry->d_parent->d_inode;
+	struct inode *parent_inode;
 	struct ceph_mds_request *req;
 	struct ceph_mds_client *mdsc = fsc->mdsc;
 	int err;
@@ -677,7 +677,9 @@ static int ceph_sync_setxattr(struct dentry *dentry, const char *name,
 	req->r_data_len = size;
 
 	dout("xattr.ver (before): %lld\n", ci->i_xattrs.version);
+	parent_inode = ceph_get_dentry_parent_inode(dentry);
 	err = ceph_mdsc_do_request(mdsc, parent_inode, req);
+	iput(parent_inode);
 	ceph_mdsc_put_request(req);
 	dout("xattr.ver (after): %lld\n", ci->i_xattrs.version);
 
@@ -788,7 +790,7 @@ static int ceph_send_removexattr(struct dentry *dentry, const char *name)
 	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
 	struct ceph_mds_client *mdsc = fsc->mdsc;
 	struct inode *inode = dentry->d_inode;
-	struct inode *parent_inode = dentry->d_parent->d_inode;
+	struct inode *parent_inode;
 	struct ceph_mds_request *req;
 	int err;
 
@@ -802,7 +804,9 @@ static int ceph_send_removexattr(struct dentry *dentry, const char *name)
 	req->r_num_caps = 1;
 	req->r_path2 = kstrdup(name, GFP_NOFS);
 
+	parent_inode = ceph_get_dentry_parent_inode(dentry);
 	err = ceph_mdsc_do_request(mdsc, parent_inode, req);
+	iput(parent_inode);
 	ceph_mdsc_put_request(req);
 	return err;
 }
-- 
cgit v1.2.3-70-g09d2


From bf1c6aca96c9d2f117dc7e590c2bc2304e7febe1 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:30:43 -0700
Subject: ceph: protect d_parent access in ceph_d_revalidate

Protect d_parent with d_lock.  Carry a reference.  Simplify the flow so
that there is a single exit point and cleanup.

Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c | 32 +++++++++++++++++---------------
 1 file changed, 17 insertions(+), 15 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index ed296ec121d..31d27f8f826 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -1024,36 +1024,38 @@ static int dir_lease_is_valid(struct inode *dir, struct dentry *dentry)
  */
 static int ceph_d_revalidate(struct dentry *dentry, struct nameidata *nd)
 {
+	int valid = 0;
 	struct inode *dir;
 
 	if (nd && nd->flags & LOOKUP_RCU)
 		return -ECHILD;
 
-	dir = dentry->d_parent->d_inode;
-
 	dout("d_revalidate %p '%.*s' inode %p offset %lld\n", dentry,
 	     dentry->d_name.len, dentry->d_name.name, dentry->d_inode,
 	     ceph_dentry(dentry)->offset);
 
+	dir = ceph_get_dentry_parent_inode(dentry);
+
 	/* always trust cached snapped dentries, snapdir dentry */
 	if (ceph_snap(dir) != CEPH_NOSNAP) {
 		dout("d_revalidate %p '%.*s' inode %p is SNAPPED\n", dentry,
 		     dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
-		goto out_touch;
+		valid = 1;
+	} else if (dentry->d_inode &&
+		   ceph_snap(dentry->d_inode) == CEPH_SNAPDIR) {
+		valid = 1;
+	} else if (dentry_lease_is_valid(dentry) ||
+		   dir_lease_is_valid(dir, dentry)) {
+		valid = 1;
 	}
-	if (dentry->d_inode && ceph_snap(dentry->d_inode) == CEPH_SNAPDIR)
-		goto out_touch;
 
-	if (dentry_lease_is_valid(dentry) ||
-	    dir_lease_is_valid(dir, dentry))
-		goto out_touch;
-
-	dout("d_revalidate %p invalid\n", dentry);
-	d_drop(dentry);
-	return 0;
-out_touch:
-	ceph_dentry_lru_touch(dentry);
-	return 1;
+	dout("d_revalidate %p %s\n", dentry, valid ? "valid" : "invalid");
+	if (valid)
+		ceph_dentry_lru_touch(dentry);
+	else
+		d_drop(dentry);
+	iput(dir);
+	return valid;
 }
 
 /*
-- 
cgit v1.2.3-70-g09d2


From e5f86dc377e7ff2b4195831153a85a3e76fefff2 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:30:55 -0700
Subject: ceph: avoid d_parent in ceph_dentry_hash; fix ceph_encode_fh()
 hashing bug

Have caller pass in a safely-obtained reference to the parent directory
for calculating a dentry's hash valud.

While we're here, simpify the flow through ceph_encode_fh() so that there
is a single exit point and cleanup.

Also fix a bug with the dentry hash calculation: calculate the hash for the
dentry we were given, not its parent.

Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c        |  3 +--
 fs/ceph/export.c     | 24 +++++++++++++++---------
 fs/ceph/mds_client.c |  2 +-
 fs/ceph/super.h      |  2 +-
 4 files changed, 18 insertions(+), 13 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 31d27f8f826..33a19df7228 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -1244,9 +1244,8 @@ void ceph_dentry_lru_del(struct dentry *dn)
  * Return name hash for a given dentry.  This is dependent on
  * the parent directory's hash function.
  */
-unsigned ceph_dentry_hash(struct dentry *dn)
+unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn)
 {
-	struct inode *dir = dn->d_parent->d_inode;
 	struct ceph_inode_info *dci = ceph_inode(dir);
 
 	switch (dci->i_dir_layout.dl_dir_hash) {
diff --git a/fs/ceph/export.c b/fs/ceph/export.c
index f67b687550d..9fbcdecaacc 100644
--- a/fs/ceph/export.c
+++ b/fs/ceph/export.c
@@ -46,7 +46,7 @@ static int ceph_encode_fh(struct dentry *dentry, u32 *rawfh, int *max_len,
 	int type;
 	struct ceph_nfs_fh *fh = (void *)rawfh;
 	struct ceph_nfs_confh *cfh = (void *)rawfh;
-	struct dentry *parent = dentry->d_parent;
+	struct dentry *parent;
 	struct inode *inode = dentry->d_inode;
 	int connected_handle_length = sizeof(*cfh)/4;
 	int handle_length = sizeof(*fh)/4;
@@ -55,26 +55,33 @@ static int ceph_encode_fh(struct dentry *dentry, u32 *rawfh, int *max_len,
 	if (ceph_snap(inode) != CEPH_NOSNAP)
 		return -EINVAL;
 
+	spin_lock(&dentry->d_lock);
+	parent = dget(dentry->d_parent);
+	spin_unlock(&dentry->d_lock);
+
 	if (*max_len >= connected_handle_length) {
 		dout("encode_fh %p connectable\n", dentry);
 		cfh->ino = ceph_ino(dentry->d_inode);
 		cfh->parent_ino = ceph_ino(parent->d_inode);
-		cfh->parent_name_hash = ceph_dentry_hash(parent);
+		cfh->parent_name_hash = ceph_dentry_hash(parent->d_inode,
+							 dentry);
 		*max_len = connected_handle_length;
 		type = 2;
 	} else if (*max_len >= handle_length) {
 		if (connectable) {
 			*max_len = connected_handle_length;
-			return 255;
+			type = 255;
+		} else {
+			dout("encode_fh %p\n", dentry);
+			fh->ino = ceph_ino(dentry->d_inode);
+			*max_len = handle_length;
+			type = 1;
 		}
-		dout("encode_fh %p\n", dentry);
-		fh->ino = ceph_ino(dentry->d_inode);
-		*max_len = handle_length;
-		type = 1;
 	} else {
 		*max_len = handle_length;
-		return 255;
+		type = 255;
 	}
+	dput(parent);
 	return type;
 }
 
@@ -123,7 +130,6 @@ static struct dentry *__fh_to_dentry(struct super_block *sb,
 		return dentry;
 	}
 	err = ceph_init_dentry(dentry);
-
 	if (err < 0) {
 		iput(inode);
 		return ERR_PTR(err);
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 3b1e743b8c8..8a09cd5a659 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -670,7 +670,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 		} else {
 			/* dir + name */
 			inode = dir;
-			hash = ceph_dentry_hash(req->r_dentry);
+			hash = ceph_dentry_hash(dir, req->r_dentry);
 			is_hash = true;
 		}
 	}
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index c1eb9a014b7..35dc9656e49 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -800,7 +800,7 @@ extern void ceph_dentry_lru_add(struct dentry *dn);
 extern void ceph_dentry_lru_touch(struct dentry *dn);
 extern void ceph_dentry_lru_del(struct dentry *dn);
 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
-extern unsigned ceph_dentry_hash(struct dentry *dn);
+extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
 
 /*
-- 
cgit v1.2.3-70-g09d2


From 41b02e1f9bb87b07d792b64aaeb7af3d00d69cd2 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:31:14 -0700
Subject: ceph: explicitly reference rename old_dentry parent dir in request

We carry a pin on the parent directory for the rename source and dest
dentries.  For the source it's r_locked_dir; we need to explicitly
reference the old_dentry parent as well, since the dentry's d_parent may
change between when the request was created and pinned and when it is
freed.

Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/debugfs.c    |  2 +-
 fs/ceph/dir.c        |  2 ++
 fs/ceph/mds_client.c | 23 +++++++++++++----------
 fs/ceph/mds_client.h |  1 +
 4 files changed, 17 insertions(+), 11 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
index 0dba6915712..fb962efdace 100644
--- a/fs/ceph/debugfs.c
+++ b/fs/ceph/debugfs.c
@@ -102,7 +102,7 @@ static int mdsc_show(struct seq_file *s, void *p)
 				path = NULL;
 			spin_lock(&req->r_old_dentry->d_lock);
 			seq_printf(s, " #%llx/%.*s (%s)",
-			   ceph_ino(req->r_old_dentry->d_parent->d_inode),
+			   ceph_ino(req->r_old_dentry_dir),
 				   req->r_old_dentry->d_name.len,
 				   req->r_old_dentry->d_name.name,
 				   path ? path : "");
diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 33a19df7228..7263f825d42 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -811,6 +811,7 @@ static int ceph_link(struct dentry *old_dentry, struct inode *dir,
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
 	req->r_old_dentry = dget(old_dentry); /* or inode? hrm. */
+	req->r_old_dentry_dir = ceph_get_dentry_parent_inode(old_dentry);
 	req->r_locked_dir = dir;
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
@@ -909,6 +910,7 @@ static int ceph_rename(struct inode *old_dir, struct dentry *old_dentry,
 	req->r_dentry = dget(new_dentry);
 	req->r_num_caps = 2;
 	req->r_old_dentry = dget(old_dentry);
+	req->r_old_dentry_dir = ceph_get_dentry_parent_inode(old_dentry);
 	req->r_locked_dir = new_dir;
 	req->r_old_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_old_dentry_unless = CEPH_CAP_FILE_EXCL;
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 8a09cd5a659..66a8939cc51 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -483,22 +483,26 @@ void ceph_mdsc_release_request(struct kref *kref)
 		destroy_reply_info(&req->r_reply_info);
 	}
 	if (req->r_inode) {
-		ceph_put_cap_refs(ceph_inode(req->r_inode),
-				  CEPH_CAP_PIN);
+		ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
 		iput(req->r_inode);
 	}
 	if (req->r_locked_dir)
-		ceph_put_cap_refs(ceph_inode(req->r_locked_dir),
-				  CEPH_CAP_PIN);
+		ceph_put_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
 	if (req->r_target_inode)
 		iput(req->r_target_inode);
 	if (req->r_dentry)
 		dput(req->r_dentry);
 	if (req->r_old_dentry) {
-		ceph_put_cap_refs(
-			ceph_inode(req->r_old_dentry->d_parent->d_inode),
-			CEPH_CAP_PIN);
+		/*
+		 * track (and drop pins for) r_old_dentry_dir
+		 * separately, since r_old_dentry's d_parent may have
+		 * changed between the dir mutex being dropped and
+		 * this request being freed.
+		 */
+		ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
+				  CEPH_CAP_PIN);
 		dput(req->r_old_dentry);
+		iput(req->r_old_dentry_dir);
 	}
 	kfree(req->r_path1);
 	kfree(req->r_path2);
@@ -1931,9 +1935,8 @@ int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
 	if (req->r_locked_dir)
 		ceph_get_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
 	if (req->r_old_dentry)
-		ceph_get_cap_refs(
-			ceph_inode(req->r_old_dentry->d_parent->d_inode),
-			CEPH_CAP_PIN);
+		ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
+				  CEPH_CAP_PIN);
 
 	/* issue */
 	mutex_lock(&mdsc->mutex);
diff --git a/fs/ceph/mds_client.h b/fs/ceph/mds_client.h
index 8a40f0636ba..4bb239921db 100644
--- a/fs/ceph/mds_client.h
+++ b/fs/ceph/mds_client.h
@@ -171,6 +171,7 @@ struct ceph_mds_request {
 	struct inode *r_inode;              /* arg1 */
 	struct dentry *r_dentry;            /* arg1 */
 	struct dentry *r_old_dentry;        /* arg2: rename from or link from */
+	struct inode *r_old_dentry_dir;     /* arg2: old dentry's parent dir */
 	char *r_path1, *r_path2;
 	struct ceph_vino r_ino1, r_ino2;
 
-- 
cgit v1.2.3-70-g09d2


From d79698da32b317e96216236f265a9b72b78ae568 Mon Sep 17 00:00:00 2001
From: Sage Weil <sage@newdream.net>
Date: Tue, 26 Jul 2011 11:31:26 -0700
Subject: ceph: document unlocked d_parent accesses

For the most part we don't care about racing with rename when directing
MDS requests; either the old or new parent is fine.  Document that, and
do some minor cleanup.

Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
---
 fs/ceph/dir.c        |  2 +-
 fs/ceph/mds_client.c | 13 ++++++++++---
 2 files changed, 11 insertions(+), 4 deletions(-)

(limited to 'fs/ceph/dir.c')

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 7263f825d42..852ff8600ac 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -506,7 +506,7 @@ int ceph_handle_snapdir(struct ceph_mds_request *req,
 			struct dentry *dentry, int err)
 {
 	struct ceph_fs_client *fsc = ceph_sb_to_client(dentry->d_sb);
-	struct inode *parent = dentry->d_parent->d_inode;
+	struct inode *parent = dentry->d_parent->d_inode; /* we hold i_mutex */
 
 	/* .snap dir? */
 	if (err == -ENOENT &&
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 66a8939cc51..fee028b5332 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -621,6 +621,12 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
  */
 struct dentry *get_nonsnap_parent(struct dentry *dentry)
 {
+	/*
+	 * we don't need to worry about protecting the d_parent access
+	 * here because we never renaming inside the snapped namespace
+	 * except to resplice to another snapdir, and either the old or new
+	 * result is a valid result.
+	 */
 	while (!IS_ROOT(dentry) && ceph_snap(dentry->d_inode) != CEPH_NOSNAP)
 		dentry = dentry->d_parent;
 	return dentry;
@@ -656,7 +662,9 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 	if (req->r_inode) {
 		inode = req->r_inode;
 	} else if (req->r_dentry) {
-		struct inode *dir = req->r_dentry->d_parent->d_inode;
+		/* ignore race with rename; old or new d_parent is okay */
+		struct dentry *parent = req->r_dentry->d_parent;
+		struct inode *dir = parent->d_inode;
 
 		if (dir->i_sb != mdsc->fsc->sb) {
 			/* not this fs! */
@@ -664,8 +672,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 		} else if (ceph_snap(dir) != CEPH_NOSNAP) {
 			/* direct snapped/virtual snapdir requests
 			 * based on parent dir inode */
-			struct dentry *dn =
-				get_nonsnap_parent(req->r_dentry->d_parent);
+			struct dentry *dn = get_nonsnap_parent(parent);
 			inode = dn->d_inode;
 			dout("__choose_mds using nonsnap parent %p\n", inode);
 		} else if (req->r_dentry->d_inode) {
-- 
cgit v1.2.3-70-g09d2