sysfs: fix race condition around sd->s_dentry, take#2

Allowing attribute and symlink dentries to be reclaimed means
sd->s_dentry can change dynamically.  However, updates to the field
are unsynchronized leading to race conditions.  This patch adds
sysfs_lock and use it to synchronize updates to sd->s_dentry.

Due to the locking around ->d_iput, the check in sysfs_drop_dentry()
is complex.  sysfs_lock only protect sd->s_dentry pointer itself.  The
validity of the dentry is protected by dcache_lock, so whether dentry
is alive or not can only be tested while holding both locks.

This is minimal backport of sysfs_drop_dentry() rewrite in devel
branch.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

3 files changed, 38 insertions, 3 deletions

diff --git a/fs/sysfs/dir.c b/fs/sysfs/dir.c
index 17a819151b91..c4342a019972 100644
--- a/fs/sysfs/dir.c
+++ b/fs/sysfs/dir.c
@@ -13,14 +13,26 @@
 #include "sysfs.h"
 
 DECLARE_RWSEM(sysfs_rename_sem);
+spinlock_t sysfs_lock = SPIN_LOCK_UNLOCKED;
 
 static void sysfs_d_iput(struct dentry * dentry, struct inode * inode)
 {
         struct sysfs_dirent * sd = dentry->d_fsdata;
 
         if (sd) {
-                BUG_ON(sd->s_dentry != dentry);
-                sd->s_dentry = NULL;
+                /* sd->s_dentry is protected with sysfs_lock.  This
+                 * allows sysfs_drop_dentry() to dereference it.
+                 */
+                spin_lock(&sysfs_lock);
+
+                /* The dentry might have been deleted or another
+                 * lookup could have happened updating sd->s_dentry to
+                 * point the new dentry.  Ignore if it isn't pointing
+                 * to this dentry.
+                 */
+                if (sd->s_dentry == dentry)
+                        sd->s_dentry = NULL;
+                spin_unlock(&sysfs_lock);
                 sysfs_put(sd);
         }
         iput(inode);
@@ -247,7 +259,10 @@ static int sysfs_attach_attr(struct sysfs_dirent * sd, struct dentry * dentry)
         }
 
         dentry->d_fsdata = sysfs_get(sd);
+        /* protect sd->s_dentry against sysfs_d_iput */
+        spin_lock(&sysfs_lock);
         sd->s_dentry = dentry;
+        spin_unlock(&sysfs_lock);
         error = sysfs_create(dentry, (attr->mode & S_IALLUGO) | S_IFREG, init);
         if (error) {
                 sysfs_put(sd);
@@ -269,7 +284,10 @@ static int sysfs_attach_link(struct sysfs_dirent * sd, struct dentry * dentry)
         int err = 0;
 
         dentry->d_fsdata = sysfs_get(sd);
+        /* protect sd->s_dentry against sysfs_d_iput */
+        spin_lock(&sysfs_lock);
         sd->s_dentry = dentry;
+        spin_unlock(&sysfs_lock);
         err = sysfs_create(dentry, S_IFLNK|S_IRWXUGO, init_symlink);
         if (!err) {
                 dentry->d_op = &sysfs_dentry_ops;
diff --git a/fs/sysfs/inode.c b/fs/sysfs/inode.c
index 38bbe071cc15..5266eec15f6e 100644
--- a/fs/sysfs/inode.c
+++ b/fs/sysfs/inode.c
@@ -246,9 +246,23 @@ static inline void orphan_all_buffers(struct inode *node)
  */
 void sysfs_drop_dentry(struct sysfs_dirent * sd, struct dentry * parent)
 {
-        struct dentry * dentry = sd->s_dentry;
+        struct dentry *dentry = NULL;
         struct inode *inode;
 
+        /* We're not holding a reference to ->s_dentry dentry but the
+         * field will stay valid as long as sysfs_lock is held.
+         */
+        spin_lock(&sysfs_lock);
+        spin_lock(&dcache_lock);
+
+        /* dget dentry if it's still alive */
+        if (sd->s_dentry && sd->s_dentry->d_inode)
+                dentry = dget_locked(sd->s_dentry);
+
+        spin_unlock(&dcache_lock);
+        spin_unlock(&sysfs_lock);
+
+        /* drop dentry */
         if (dentry) {
                 spin_lock(&dcache_lock);
                 spin_lock(&dentry->d_lock);
@@ -268,6 +282,8 @@ void sysfs_drop_dentry(struct sysfs_dirent * sd, struct dentry * parent)
                         spin_unlock(&dentry->d_lock);
                         spin_unlock(&dcache_lock);
                 }
+
+                dput(dentry);
         }
 }
 
diff --git a/fs/sysfs/sysfs.h b/fs/sysfs/sysfs.h
index 1966e1a0a015..502c949c402d 100644
--- a/fs/sysfs/sysfs.h
+++ b/fs/sysfs/sysfs.h
@@ -33,6 +33,7 @@ extern const unsigned char * sysfs_get_name(struct sysfs_dirent *sd);
 extern void sysfs_drop_dentry(struct sysfs_dirent *sd, struct dentry *parent);
 extern int sysfs_setattr(struct dentry *dentry, struct iattr *iattr);
 
+extern spinlock_t sysfs_lock;
 extern struct rw_semaphore sysfs_rename_sem;
 extern struct super_block * sysfs_sb;
 extern const struct file_operations sysfs_dir_operations;