OSDN Git Service

cifs: Fix cifsInodeInfo lock_sem deadlock when reconnect occurs
authorDave Wysochanski <dwysocha@redhat.com>
Wed, 23 Oct 2019 09:02:33 +0000 (05:02 -0400)
committerSteve French <stfrench@microsoft.com>
Fri, 25 Oct 2019 02:35:04 +0000 (21:35 -0500)
There's a deadlock that is possible and can easily be seen with
a test where multiple readers open/read/close of the same file
and a disruption occurs causing reconnect.  The deadlock is due
a reader thread inside cifs_strict_readv calling down_read and
obtaining lock_sem, and then after reconnect inside
cifs_reopen_file calling down_read a second time.  If in
between the two down_read calls, a down_write comes from
another process, deadlock occurs.

        CPU0                    CPU1
        ----                    ----
cifs_strict_readv()
 down_read(&cifsi->lock_sem);
                               _cifsFileInfo_put
                                  OR
                               cifs_new_fileinfo
                                down_write(&cifsi->lock_sem);
cifs_reopen_file()
 down_read(&cifsi->lock_sem);

Fix the above by changing all down_write(lock_sem) calls to
down_write_trylock(lock_sem)/msleep() loop, which in turn
makes the second down_read call benign since it will never
block behind the writer while holding lock_sem.

Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
Suggested-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed--by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
fs/cifs/cifsglob.h
fs/cifs/cifsproto.h
fs/cifs/file.c
fs/cifs/smb2file.c

index 50dfd90..d78bfcc 100644 (file)
@@ -1391,6 +1391,11 @@ void cifsFileInfo_put(struct cifsFileInfo *cifs_file);
 struct cifsInodeInfo {
        bool can_cache_brlcks;
        struct list_head llist; /* locks helb by this inode */
+       /*
+        * NOTE: Some code paths call down_read(lock_sem) twice, so
+        * we must always use use cifs_down_write() instead of down_write()
+        * for this semaphore to avoid deadlocks.
+        */
        struct rw_semaphore lock_sem;   /* protect the fields above */
        /* BB add in lists for dirty pages i.e. write caching info for oplock */
        struct list_head openFileList;
index e53e9f6..fe597d3 100644 (file)
@@ -170,6 +170,7 @@ extern int cifs_unlock_range(struct cifsFileInfo *cfile,
                             struct file_lock *flock, const unsigned int xid);
 extern int cifs_push_mandatory_locks(struct cifsFileInfo *cfile);
 
+extern void cifs_down_write(struct rw_semaphore *sem);
 extern struct cifsFileInfo *cifs_new_fileinfo(struct cifs_fid *fid,
                                              struct file *file,
                                              struct tcon_link *tlink,
index 6482793..fa7b0fa 100644 (file)
@@ -281,6 +281,13 @@ cifs_has_mand_locks(struct cifsInodeInfo *cinode)
        return has_locks;
 }
 
+void
+cifs_down_write(struct rw_semaphore *sem)
+{
+       while (!down_write_trylock(sem))
+               msleep(10);
+}
+
 struct cifsFileInfo *
 cifs_new_fileinfo(struct cifs_fid *fid, struct file *file,
                  struct tcon_link *tlink, __u32 oplock)
@@ -306,7 +313,7 @@ cifs_new_fileinfo(struct cifs_fid *fid, struct file *file,
        INIT_LIST_HEAD(&fdlocks->locks);
        fdlocks->cfile = cfile;
        cfile->llist = fdlocks;
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        list_add(&fdlocks->llist, &cinode->llist);
        up_write(&cinode->lock_sem);
 
@@ -464,7 +471,7 @@ void _cifsFileInfo_put(struct cifsFileInfo *cifs_file, bool wait_oplock_handler)
         * Delete any outstanding lock records. We'll lose them when the file
         * is closed anyway.
         */
-       down_write(&cifsi->lock_sem);
+       cifs_down_write(&cifsi->lock_sem);
        list_for_each_entry_safe(li, tmp, &cifs_file->llist->locks, llist) {
                list_del(&li->llist);
                cifs_del_lock_waiters(li);
@@ -1027,7 +1034,7 @@ static void
 cifs_lock_add(struct cifsFileInfo *cfile, struct cifsLockInfo *lock)
 {
        struct cifsInodeInfo *cinode = CIFS_I(d_inode(cfile->dentry));
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        list_add_tail(&lock->llist, &cfile->llist->locks);
        up_write(&cinode->lock_sem);
 }
@@ -1049,7 +1056,7 @@ cifs_lock_add_if(struct cifsFileInfo *cfile, struct cifsLockInfo *lock,
 
 try_again:
        exist = false;
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
 
        exist = cifs_find_lock_conflict(cfile, lock->offset, lock->length,
                                        lock->type, lock->flags, &conf_lock,
@@ -1072,7 +1079,7 @@ try_again:
                                        (lock->blist.next == &lock->blist));
                if (!rc)
                        goto try_again;
-               down_write(&cinode->lock_sem);
+               cifs_down_write(&cinode->lock_sem);
                list_del_init(&lock->blist);
        }
 
@@ -1125,7 +1132,7 @@ cifs_posix_lock_set(struct file *file, struct file_lock *flock)
                return rc;
 
 try_again:
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        if (!cinode->can_cache_brlcks) {
                up_write(&cinode->lock_sem);
                return rc;
@@ -1331,7 +1338,7 @@ cifs_push_locks(struct cifsFileInfo *cfile)
        int rc = 0;
 
        /* we are going to update can_cache_brlcks here - need a write access */
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        if (!cinode->can_cache_brlcks) {
                up_write(&cinode->lock_sem);
                return rc;
@@ -1522,7 +1529,7 @@ cifs_unlock_range(struct cifsFileInfo *cfile, struct file_lock *flock,
        if (!buf)
                return -ENOMEM;
 
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        for (i = 0; i < 2; i++) {
                cur = buf;
                num = 0;
index e6a1fc7..8b0b512 100644 (file)
@@ -145,7 +145,7 @@ smb2_unlock_range(struct cifsFileInfo *cfile, struct file_lock *flock,
 
        cur = buf;
 
-       down_write(&cinode->lock_sem);
+       cifs_down_write(&cinode->lock_sem);
        list_for_each_entry_safe(li, tmp, &cfile->llist->locks, llist) {
                if (flock->fl_start > li->offset ||
                    (flock->fl_start + length) <