Since it was ignored on LKML, I guess this should be the right list. -------- Original Message -------- Subject: [PATCH] UFS i_blocks handling Date: Wed, 02 May 2012 18:34:18 +0200 From: Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@xxxxxxxxx> To: linux-kernel@xxxxxxxxxxxxxxx Current UFS code doesn't maintain i_blocks. It creates 2 glitches: 1) state always says that no blocks are allocated 2) long symlinks are corrupted. For (2) in GRUB there is a workaround to check whether link is a fast link based on its length (link is considered a fast one if its size is small to fit in inline space and i_blocks == 0) however neither Linux nor BSD-using OS have such workaround and this should be fixed anyway. Please consider attached patch. It may be of bad quality, I haven't learned the code in question in more details. Please CC me when answering. -- Regards Vladimir 'φ-coder/phcoder' Serbinenko
diff -ur linux-source-3.2/fs/ufs/balloc.c /home/phcoder/projects/ufs/balloc.c
--- linux-source-3.2/fs/ufs/balloc.c 2012-04-23 00:31:32.000000000 +0200
+++ /home/phcoder/projects/ufs/balloc.c 2012-05-02 18:24:25.444822068 +0200
@@ -147,6 +147,8 @@
UFSD("ENTER, fragment %llu, count %u\n",
(unsigned long long)fragment, count);
+
+ inode->i_blocks -= count;
if ((fragment & uspi->s_fpbmask) || (count & uspi->s_fpbmask)) {
ufs_error (sb, "ufs_free_blocks", "internal error, "
@@ -423,6 +425,7 @@
if (result) {
ufs_cpu_to_data_ptr(sb, p, result);
*err = 0;
+ inode->i_blocks += count;
UFS_I(inode)->i_lastfrag =
max(UFS_I(inode)->i_lastfrag, fragment + count);
ufs_clear_frags(inode, result + oldcount,
Attachment:
signature.asc
Description: OpenPGP digital signature
