]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
dax: Don't access a freed inode
authorMatthew Wilcox <willy@infradead.org>
Tue, 27 Nov 2018 21:16:34 +0000 (13:16 -0800)
committerDan Williams <dan.j.williams@intel.com>
Wed, 28 Nov 2018 19:08:42 +0000 (11:08 -0800)
After we drop the i_pages lock, the inode can be freed at any time.
The get_unlocked_entry() code has no choice but to reacquire the lock,
so it can't be used here.  Create a new wait_entry_unlocked() which takes
care not to acquire the lock or dereference the address_space in any way.

Fixes: c2a7d2a11552 ("filesystem-dax: Introduce dax_lock_mapping_entry()")
Cc: <stable@vger.kernel.org>
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
fs/dax.c

index e69fc231833b7204f47c9abb8ef43d860e67e312..3f592dc18d67c6fbd804d529a599c55a6ac6ed4f 100644 (file)
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -232,6 +232,34 @@ static void *get_unlocked_entry(struct xa_state *xas)
        }
 }
 
+/*
+ * The only thing keeping the address space around is the i_pages lock
+ * (it's cycled in clear_inode() after removing the entries from i_pages)
+ * After we call xas_unlock_irq(), we cannot touch xas->xa.
+ */
+static void wait_entry_unlocked(struct xa_state *xas, void *entry)
+{
+       struct wait_exceptional_entry_queue ewait;
+       wait_queue_head_t *wq;
+
+       init_wait(&ewait.wait);
+       ewait.wait.func = wake_exceptional_entry_func;
+
+       wq = dax_entry_waitqueue(xas, entry, &ewait.key);
+       prepare_to_wait_exclusive(wq, &ewait.wait, TASK_UNINTERRUPTIBLE);
+       xas_unlock_irq(xas);
+       schedule();
+       finish_wait(wq, &ewait.wait);
+
+       /*
+        * Entry lock waits are exclusive. Wake up the next waiter since
+        * we aren't sure we will acquire the entry lock and thus wake
+        * the next waiter up on unlock.
+        */
+       if (waitqueue_active(wq))
+               __wake_up(wq, TASK_NORMAL, 1, &ewait.key);
+}
+
 static void put_unlocked_entry(struct xa_state *xas, void *entry)
 {
        /* If we were the only waiter woken, wake the next one */
@@ -389,9 +417,7 @@ bool dax_lock_mapping_entry(struct page *page)
                entry = xas_load(&xas);
                if (dax_is_locked(entry)) {
                        rcu_read_unlock();
-                       entry = get_unlocked_entry(&xas);
-                       xas_unlock_irq(&xas);
-                       put_unlocked_entry(&xas, entry);
+                       wait_entry_unlocked(&xas, entry);
                        rcu_read_lock();
                        continue;
                }