]> www.pilppa.org Git - linux-2.6-omap-h63xx.git/commitdiff
ext4: Fix fallocate to update the file size in each transaction
authorAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Tue, 29 Apr 2008 12:11:12 +0000 (08:11 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Tue, 29 Apr 2008 12:11:12 +0000 (08:11 -0400)
ext4_fallocate needs to update file size in each transaction.  Otherwise
if we crash the file size won't be seen.  We were also not marking
the inode dirty after updating file size before.  Also when we try to
retry allocation due to ENOSPC, make sure we reset the variable ret so
that we actually do a retry.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/extents.c

index 668d82e494dd081ea36c5cbfd381b83474ca707c..c2b004e29ad8f239c2cc66bb80f331d0b1e7ca79 100644 (file)
@@ -2785,6 +2785,28 @@ int ext4_ext_writepage_trans_blocks(struct inode *inode, int num)
        return needed;
 }
 
+static void ext4_falloc_update_inode(struct inode *inode,
+                               int mode, loff_t new_size, int update_ctime)
+{
+       struct timespec now;
+
+       if (update_ctime) {
+               now = current_fs_time(inode->i_sb);
+               if (!timespec_equal(&inode->i_ctime, &now))
+                       inode->i_ctime = now;
+       }
+       /*
+        * Update only when preallocation was requested beyond
+        * the file size.
+        */
+       if (!(mode & FALLOC_FL_KEEP_SIZE) &&
+                               new_size > i_size_read(inode)) {
+               i_size_write(inode, new_size);
+               EXT4_I(inode)->i_disksize = new_size;
+       }
+
+}
+
 /*
  * preallocate space for a file. This implements ext4's fallocate inode
  * operation, which gets called from sys_fallocate system call.
@@ -2796,8 +2818,8 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
 {
        handle_t *handle;
        ext4_lblk_t block;
+       loff_t new_size;
        unsigned long max_blocks;
-       ext4_fsblk_t nblocks = 0;
        int ret = 0;
        int ret2 = 0;
        int retries = 0;
@@ -2816,9 +2838,12 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
                return -ENODEV;
 
        block = offset >> blkbits;
+       /*
+        * We can't just convert len to max_blocks because
+        * If blocksize = 4096 offset = 3072 and len = 2048
+        */
        max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
-                       - block;
-
+                                                       - block;
        /*
         * credits to insert 1 extent into extent tree + buffers to be able to
         * modify 1 super block, 1 block bitmap and 1 group descriptor.
@@ -2834,7 +2859,6 @@ retry:
                        ret = PTR_ERR(handle);
                        break;
                }
-
                ret = ext4_get_blocks_wrap(handle, inode, block,
                                          max_blocks, &map_bh,
                                          EXT4_CREATE_UNINITIALIZED_EXT, 0);
@@ -2850,61 +2874,24 @@ retry:
                        ret2 = ext4_journal_stop(handle);
                        break;
                }
-               if (ret > 0) {
-                       /* check wrap through sign-bit/zero here */
-                       if ((block + ret) < 0 || (block + ret) < block) {
-                               ret = -EIO;
-                               ext4_mark_inode_dirty(handle, inode);
-                               ret2 = ext4_journal_stop(handle);
-                               break;
-                       }
-                       if (buffer_new(&map_bh) && ((block + ret) >
-                           (EXT4_BLOCK_ALIGN(i_size_read(inode), blkbits)
-                           >> blkbits)))
-                                       nblocks = nblocks + ret;
-               }
-
-               /* Update ctime if new blocks get allocated */
-               if (nblocks) {
-                       struct timespec now;
-
-                       now = current_fs_time(inode->i_sb);
-                       if (!timespec_equal(&inode->i_ctime, &now))
-                               inode->i_ctime = now;
-               }
+               if ((block + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
+                                               blkbits) >> blkbits))
+                       new_size = offset + len;
+               else
+                       new_size = (block + ret) << blkbits;
 
+               ext4_falloc_update_inode(inode, mode, new_size,
+                                               buffer_new(&map_bh));
                ext4_mark_inode_dirty(handle, inode);
                ret2 = ext4_journal_stop(handle);
                if (ret2)
                        break;
        }
-
-       if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
+       if (ret == -ENOSPC &&
+                       ext4_should_retry_alloc(inode->i_sb, &retries)) {
+               ret = 0;
                goto retry;
-
-       /*
-        * Time to update the file size.
-        * Update only when preallocation was requested beyond the file size.
-        */
-       if (!(mode & FALLOC_FL_KEEP_SIZE) &&
-           (offset + len) > i_size_read(inode)) {
-               if (ret > 0) {
-                       /*
-                        * if no error, we assume preallocation succeeded
-                        * completely
-                        */
-                       i_size_write(inode, offset + len);
-                       EXT4_I(inode)->i_disksize = i_size_read(inode);
-               } else if (ret < 0 && nblocks) {
-                       /* Handle partial allocation scenario */
-                       loff_t newsize;
-
-                       newsize  = (nblocks << blkbits) + i_size_read(inode);
-                       i_size_write(inode, EXT4_BLOCK_ALIGN(newsize, blkbits));
-                       EXT4_I(inode)->i_disksize = i_size_read(inode);
-               }
        }
-
        mutex_unlock(&inode->i_mutex);
        return ret > 0 ? ret2 : ret;
 }