cifs: fix potential races in cifs_revalidate_mapping

The handling of the CIFS_INO_INVALID_MAPPING flag is racy. It's possible
for two tasks to attempt to revalidate the mapping at the same time. The
first sees that CIFS_INO_INVALID_MAPPING is set. It clears the flag and
then calls invalidate_inode_pages2 to start shooting down the pagecache.

While that's going on, another task checks the flag and sees that it's
clear. It then ends up trusting the pagecache to satisfy a read when it
shouldn't.

Fix this by adding a bitlock to ensure that the clearing of the flag is
atomic with respect to the actual cache invalidation. Also, move the
other existing users of cifs_invalidate_mapping to use a new
cifs_zap_mapping() function that just sets the INVALID_MAPPING bit and
then uses the standard codepath to handle the invalidation.
Signed-off-by: NJeff Layton <jlayton@poochiereds.net>
Signed-off-by: NSteve French <smfrench@gmail.com>

fs/cifs/cifsfs.h

@@ -76,6 +76,7 @@ extern int cifs_revalidate_file(struct file *filp);
 extern int cifs_revalidate_dentry(struct dentry *);
 extern int cifs_invalidate_mapping(struct inode *inode);
 extern int cifs_revalidate_mapping(struct inode *inode);
+extern int cifs_zap_mapping(struct inode *inode);
 extern int cifs_getattr(struct vfsmount *, struct dentry *, struct kstat *);
 extern int cifs_setattr(struct dentry *, struct iattr *);
 

fs/cifs/cifsglob.h

@@ -1118,6 +1118,7 @@ struct cifsInodeInfo {
 #define CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2 (2) /* Downgrade oplock to L2 */
 #define CIFS_INO_DELETE_PENDING		  (3) /* delete pending on server */
 #define CIFS_INO_INVALID_MAPPING	  (4) /* pagecache is invalid */
+#define CIFS_INO_LOCK			  (5) /* lock bit for synchronization */
 	unsigned long flags;
 	spinlock_t writers_lock;
 	unsigned int writers;		/* Number of writers on this inode */

fs/cifs/file.c

@@ -335,7 +335,7 @@ cifs_new_fileinfo(struct cifs_fid *fid, struct file *file,
 	spin_unlock(&cifs_file_list_lock);
 
 	if (fid->purge_cache)
-		cifs_invalidate_mapping(inode);
+		cifs_zap_mapping(inode);
 
 	file->private_data = cfile;
 	return cfile;
@@ -1529,7 +1529,7 @@ cifs_setlk(struct file *file, struct file_lock *flock, __u32 type,
 		 */
 		if (!CIFS_CACHE_WRITE(CIFS_I(inode)) &&
 					CIFS_CACHE_READ(CIFS_I(inode))) {
-			cifs_invalidate_mapping(inode);
+			cifs_zap_mapping(inode);
 			cifs_dbg(FYI, "Set no oplock for inode=%p due to mand locks\n",
 				 inode);
 			CIFS_I(inode)->oplock = 0;
@@ -2218,7 +2218,7 @@ int cifs_strict_fsync(struct file *file, loff_t start, loff_t end,
 		 file->f_path.dentry->d_name.name, datasync);
 
 	if (!CIFS_CACHE_READ(CIFS_I(inode))) {
-		rc = cifs_invalidate_mapping(inode);
+		rc = cifs_zap_mapping(inode);
 		if (rc) {
 			cifs_dbg(FYI, "rc: %d during invalidate phase\n", rc);
 			rc = 0; /* don't care about it in fsync */
@@ -2649,7 +2649,7 @@ cifs_strict_writev(struct kiocb *iocb, const struct iovec *iov,
 		 * request comes - break it on the client to prevent reading
 		 * an old data.
 		 */
-		cifs_invalidate_mapping(inode);
+		cifs_zap_mapping(inode);
 		cifs_dbg(FYI, "Set no oplock for inode=%p after a write operation\n",
 			 inode);
 		cinode->oplock = 0;
@@ -3112,7 +3112,7 @@ int cifs_file_strict_mmap(struct file *file, struct vm_area_struct *vma)
 	xid = get_xid();
 
 	if (!CIFS_CACHE_READ(CIFS_I(inode))) {
-		rc = cifs_invalidate_mapping(inode);
+		rc = cifs_zap_mapping(inode);
 		if (rc)
 			return rc;
 	}
@@ -3670,7 +3670,7 @@ void cifs_oplock_break(struct work_struct *work)
 		if (!CIFS_CACHE_READ(cinode)) {
 			rc = filemap_fdatawait(inode->i_mapping);
 			mapping_set_error(inode->i_mapping, rc);
-			cifs_invalidate_mapping(inode);
+			cifs_zap_mapping(inode);
 		}
 		cifs_dbg(FYI, "Oplock flush inode %p rc %d\n", inode, rc);
 	}

fs/cifs/inode.c

@@ -22,6 +22,7 @@
 #include <linux/stat.h>
 #include <linux/slab.h>
 #include <linux/pagemap.h>
+#include <linux/freezer.h>
 #include <asm/div64.h>
 #include "cifsfs.h"
 #include "cifspdu.h"
@@ -1762,29 +1763,60 @@ int
 cifs_invalidate_mapping(struct inode *inode)
 {
 	int rc = 0;
-	struct cifsInodeInfo *cifs_i = CIFS_I(inode);
-
-	clear_bit(CIFS_INO_INVALID_MAPPING, &cifs_i->flags);
 
 	if (inode->i_mapping && inode->i_mapping->nrpages != 0) {
 		rc = invalidate_inode_pages2(inode->i_mapping);
-		if (rc) {
+		if (rc)
 			cifs_dbg(VFS, "%s: could not invalidate inode %p\n",
 				 __func__, inode);
-			set_bit(CIFS_INO_INVALID_MAPPING, &cifs_i->flags);
-		}
 	}
 
 	cifs_fscache_reset_inode_cookie(inode);
 	return rc;
 }
 
+/**
+ * cifs_wait_bit_killable - helper for functions that are sleeping on bit locks
+ * @word: long word containing the bit lock
+ */
+static int
+cifs_wait_bit_killable(void *word)
+{
+	if (fatal_signal_pending(current))
+		return -ERESTARTSYS;
+	freezable_schedule_unsafe();
+	return 0;
+}
+
 int
 cifs_revalidate_mapping(struct inode *inode)
 {
-	if (test_bit(CIFS_INO_INVALID_MAPPING, &CIFS_I(inode)->flags))
-		return cifs_invalidate_mapping(inode);
-	return 0;
+	int rc;
+	unsigned long *flags = &CIFS_I(inode)->flags;
+
+	rc = wait_on_bit_lock(flags, CIFS_INO_LOCK, cifs_wait_bit_killable,
+				TASK_KILLABLE);
+	if (rc)
+		return rc;
+
+	if (test_and_clear_bit(CIFS_INO_INVALID_MAPPING, flags)) {
+		rc = cifs_invalidate_mapping(inode);
+		if (rc)
+			set_bit(CIFS_INO_INVALID_MAPPING, flags);
+	}
+
+	clear_bit_unlock(CIFS_INO_LOCK, flags);
+	smp_mb__after_clear_bit();
+	wake_up_bit(flags, CIFS_INO_LOCK);
+
+	return rc;
+}
+
+int
+cifs_zap_mapping(struct inode *inode)
+{
+	set_bit(CIFS_INO_INVALID_MAPPING, &CIFS_I(inode)->flags);
+	return cifs_revalidate_mapping(inode);
 }
 
 int cifs_revalidate_file_attr(struct file *filp)