]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
NTFS: Fix a stupid bug causing writes to non-initialized pages to segfault.
authorAnton Altaparmakov <aia21@cantab.net>
Tue, 1 Nov 2005 15:49:31 +0000 (15:49 +0000)
committerAnton Altaparmakov <aia21@cantab.net>
Tue, 1 Nov 2005 15:49:31 +0000 (15:49 +0000)
Signed-off-by: Anton Altaparmakov <aia21@cantab.net>
fs/ntfs/file.c

index cf3e6ced2d01d12256c9bc66c9d5a108ee2765ae..7275338918132fe31eed51bddf2d48485d52c1b9 100644 (file)
@@ -668,10 +668,10 @@ map_buffer_cached:
                                 * to, we need to read it in before the write,
                                 * i.e. now.
                                 */
-                               if (!buffer_uptodate(bh) && ((bh_pos < pos &&
-                                               bh_end > pos) ||
-                                               (bh_end > end &&
-                                               bh_end > end))) {
+                               if (!buffer_uptodate(bh) && bh_pos < end &&
+                                               bh_end > pos &&
+                                               (bh_pos < pos ||
+                                               bh_end > end)) {
                                        /*
                                         * If the buffer is fully or partially
                                         * within the initialized size, do an
@@ -784,10 +784,11 @@ retry_remap:
                                                blocksize_bits);
                                cdelta = 0;
                                /*
-                                * If the number of remaining clusters in the
-                                * @pages is smaller or equal to the number of
-                                * cached clusters, unlock the runlist as the
-                                * map cache will be used from now on.
+                                * If the number of remaining clusters touched
+                                * by the write is smaller or equal to the
+                                * number of cached clusters, unlock the
+                                * runlist as the map cache will be used from
+                                * now on.
                                 */
                                if (likely(vcn + vcn_len >= cend)) {
                                        if (rl_write_locked) {