Hi everyone,
This patch changes reiserfs to use get_bh and put_bh everywhere,
instead of mucking around with b_count directly. In general,
reiserfs uses brelse when doing the final release on a buffer,
and it already has the buffer pinned during i/o because it wants
to wait on the buffer later. In other words, we should not have
the same races as in fs/buffer.c, but I made this patch to keep
track with the changes there.
One hunk depends on the direct->indirect cleanup patch
I posted last week (linus, I called this flushlist-3.diff in the
batch I just sent you). I can supply one against pure pre7 if you
would like.
Reiserfs already had a marco called get_bh, which was used
to get a pointer to the last buffer head in a path in the tree. I've
renamed that to get_last_bh.
Linus, please apply:
diff -Nru a/fs/reiserfs/do_balan.c b/fs/reiserfs/do_balan.c
--- a/fs/reiserfs/do_balan.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/do_balan.c Thu Jul 12 10:57:27 2001
@@ -1626,7 +1626,7 @@
for (i = 0; i < sizeof (tb->thrown)/sizeof (tb->thrown[0]); i ++)
if (!tb->thrown[i]) {
tb->thrown[i] = bh;
- atomic_inc(&bh->b_count) ; /* decremented in free_thrown */
+ get_bh(bh) ; /* free_thrown puts this */
return;
}
reiserfs_warning ("store_thrown: too many thrown buffers\n");
diff -Nru a/fs/reiserfs/fix_node.c b/fs/reiserfs/fix_node.c
--- a/fs/reiserfs/fix_node.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/fix_node.c Thu Jul 12 10:57:27 2001
@@ -1030,7 +1030,7 @@
left, left);
}
#endif
- atomic_dec (&(left->b_count));
+ put_bh(left) ;
return 1;
}
@@ -1123,7 +1123,7 @@
n_first_last_position = B_NR_ITEMS (p_s_parent);
if ( n_position != n_first_last_position ) {
*pp_s_com_father = p_s_parent;
- atomic_inc (&((*pp_s_com_father)->b_count));
+ get_bh(*pp_s_com_father) ;
/*(*pp_s_com_father = p_s_parent)->b_count++;*/
break;
}
@@ -1228,8 +1228,8 @@
/* Current node is not the first child of its parent. */
/*(p_s_curf = p_s_curcf = PATH_OFFSET_PBUFFER(p_s_path, n_path_offset - 1))->b_count += 2;*/
p_s_curf = p_s_curcf = PATH_OFFSET_PBUFFER(p_s_path, n_path_offset - 1);
- atomic_inc (&(p_s_curf->b_count));
- atomic_inc (&(p_s_curf->b_count));
+ get_bh(p_s_curf) ;
+ get_bh(p_s_curf) ;
p_s_tb->lkey[n_h] = n_position - 1;
}
else {
@@ -1267,8 +1267,8 @@
/* Current node is not the last child of its parent F[n_h]. */
/*(p_s_curf = p_s_curcf = PATH_OFFSET_PBUFFER(p_s_path, n_path_offset - 1))->b_count += 2;*/
p_s_curf = p_s_curcf = PATH_OFFSET_PBUFFER(p_s_path, n_path_offset - 1);
- atomic_inc (&(p_s_curf->b_count));
- atomic_inc (&(p_s_curf->b_count));
+ get_bh(p_s_curf) ;
+ get_bh(p_s_curf) ;
p_s_tb->rkey[n_h] = n_position;
}
diff -Nru a/fs/reiserfs/inode.c b/fs/reiserfs/inode.c
--- a/fs/reiserfs/inode.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/inode.c Thu Jul 12 10:57:27 2001
@@ -274,7 +274,7 @@
}
//
- bh = get_bh (&path);
+ bh = get_last_bh (&path);
ih = get_ih (&path);
if (is_indirect_le_ih (ih)) {
__u32 * ind_item = (__u32 *)B_I_PITEM (bh, ih);
@@ -369,7 +369,7 @@
if (search_for_position_by_key (inode->i_sb, &key, &path) != POSITION_FOUND)
// we read something from tail, even if now we got IO_ERROR
break;
- bh = get_bh (&path);
+ bh = get_last_bh (&path);
ih = get_ih (&path);
} while (1);
@@ -577,7 +577,7 @@
goto failure;
}
- bh = get_bh (&path);
+ bh = get_last_bh (&path);
ih = get_ih (&path);
item = get_item (&path);
pos_in_item = path.pos_in_item;
@@ -823,7 +823,7 @@
pathrelse(&path) ;
goto failure;
}
- bh = get_bh (&path);
+ bh = get_last_bh (&path);
ih = get_ih (&path);
item = get_item (&path);
pos_in_item = path.pos_in_item;
@@ -1050,7 +1050,7 @@
** FS might change. We have to detect that, and loop back to the
** search if the stat data item has moved
*/
- bh = get_bh(&path) ;
+ bh = get_last_bh(&path) ;
ih = get_ih(&path) ;
copy_item_head (&tmp_ih, ih);
fs_gen = get_generation (inode->i_sb);
@@ -1679,7 +1679,7 @@
goto out ;
}
- bh = get_bh(&path) ;
+ bh = get_last_bh(&path) ;
ih = get_ih(&path) ;
item = get_item(&path) ;
pos_in_item = path.pos_in_item ;
@@ -1761,7 +1761,7 @@
for(i = 0 ; i < nr ; i++) {
bh = bhp[i] ;
lock_buffer(bh) ;
- atomic_inc(&bh->b_count) ; /* async end_io handler decs this */
+ get_bh(bh) ; /* async end_io handler puts this */
set_buffer_async_io(bh) ;
/* submit_bh doesn't care if the buffer is dirty, but nobody
** later on in the call chain will be cleaning it. So, we
diff -Nru a/fs/reiserfs/journal.c b/fs/reiserfs/journal.c
--- a/fs/reiserfs/journal.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/journal.c Thu Jul 12 10:57:27 2001
@@ -707,7 +707,7 @@
}
ll_rw_block(WRITE, 1, &tbh) ;
count++ ;
- atomic_dec(&(tbh->b_count)) ; /* once for our get_hash */
+ put_bh(tbh) ; /* once for our get_hash */
}
}
@@ -722,7 +722,7 @@
if (!buffer_uptodate(tbh)) {
reiserfs_panic(s, "journal-601, buffer write failed\n") ;
}
- atomic_dec(&(tbh->b_count)) ; /* once for our get_hash */
+ put_bh(tbh) ; /* once for our get_hash */
bforget(tbh) ; /* once due to original getblk in do_journal_end */
atomic_dec(&(jl->j_commit_left)) ;
}
@@ -869,9 +869,11 @@
}
mark_buffer_uptodate(bh, uptodate) ;
unlock_buffer(bh) ;
+ put_bh(bh) ;
}
static void submit_logged_buffer(struct buffer_head *bh) {
lock_buffer(bh) ;
+ get_bh(bh) ;
bh->b_end_io = reiserfs_end_buffer_io_sync ;
mark_buffer_notjournal_new(bh) ;
clear_bit(BH_Dirty, &bh->b_state) ;
@@ -967,13 +969,13 @@
/* we do this to make sure nobody releases the buffer while
** we are working with it
*/
- atomic_inc(&(saved_bh->b_count)) ;
+ get_bh(saved_bh) ;
if (buffer_journal_dirty(saved_bh)) {
was_jwait = 1 ;
mark_buffer_notjournal_dirty(saved_bh) ;
- /* brelse the inc from journal_mark_dirty */
- atomic_dec(&(saved_bh->b_count)) ;
+ /* undo the inc from journal_mark_dirty */
+ put_bh(saved_bh) ;
}
if (can_dirty(cn)) {
was_dirty = 1 ;
@@ -1016,7 +1018,7 @@
}
if (was_dirty) {
/* we inc again because saved_bh gets decremented at free_cnode */
- atomic_inc(&(saved_bh->b_count)) ;
+ get_bh(saved_bh) ;
set_bit(BLOCK_NEEDS_FLUSH, &cn->state) ;
submit_logged_buffer(saved_bh) ;
count++ ;
@@ -1029,7 +1031,7 @@
cn = cn->next ;
if (saved_bh) {
/* we incremented this to keep others from taking the buffer head away */
- atomic_dec(&(saved_bh->b_count));
+ put_bh(saved_bh) ;
if (atomic_read(&(saved_bh->b_count)) < 0) {
printk("journal-945: saved_bh->b_count < 0") ;
}
@@ -2176,7 +2178,7 @@
cn->jlist = NULL ;
insert_journal_hash(SB_JOURNAL(p_s_sb)->j_hash_table, cn) ;
if (!count_already_incd) {
- atomic_inc(&(bh->b_count)) ;
+ get_bh(bh) ;
}
}
cn->next = NULL ;
@@ -2248,7 +2250,7 @@
if (!already_cleaned) {
mark_buffer_notjournal_dirty(bh) ;
- atomic_dec(&(bh->b_count)) ;
+ put_bh(bh) ;
if (atomic_read(&(bh->b_count)) < 0) {
printk("journal-1752: remove from trans, b_count < 0\n") ;
}
@@ -2580,7 +2582,7 @@
*/
mark_buffer_notjournal_dirty(cn->bh) ;
cleaned = 1 ;
- atomic_dec(&(cn->bh->b_count)) ;
+ put_bh(cn->bh) ;
if (atomic_read(&(cn->bh->b_count)) < 0) {
printk("journal-2138: cn->bh->b_count < 0\n") ;
}
@@ -2597,7 +2599,7 @@
if (bh) {
reiserfs_clean_and_file_buffer(bh) ;
- atomic_dec(&(bh->b_count)) ; /* get_hash incs this */
+ put_bh(bh) ; /* get_hash grabs the buffer */
if (atomic_read(&(bh->b_count)) < 0) {
printk("journal-2165: bh->b_count < 0\n") ;
}
diff -Nru a/fs/reiserfs/namei.c b/fs/reiserfs/namei.c
--- a/fs/reiserfs/namei.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/namei.c Thu Jul 12 10:57:27 2001
@@ -72,7 +72,7 @@
// comment? maybe something like set de to point to what the path points to?
static inline void set_de_item_location (struct reiserfs_dir_entry * de, struct path * path)
{
- de->de_bh = get_bh (path);
+ de->de_bh = get_last_bh (path);
de->de_ih = get_ih (path);
de->de_deh = B_I_DEH (de->de_bh, de->de_ih);
de->de_item_num = PATH_LAST_POSITION (path);
diff -Nru a/fs/reiserfs/stree.c b/fs/reiserfs/stree.c
--- a/fs/reiserfs/stree.c Thu Jul 12 10:57:27 2001
+++ b/fs/reiserfs/stree.c Thu Jul 12 10:57:27 2001
@@ -436,7 +436,7 @@
) {
if ( p_s_bh ) {
if ( atomic_read (&(p_s_bh->b_count)) ) {
- atomic_dec (&(p_s_bh->b_count));
+ put_bh(p_s_bh) ;
return;
}
reiserfs_panic(NULL, "PAP-5070: decrement_bcount: trying to free free buffer %b", p_s_bh);
@@ -1032,7 +1032,7 @@
}
/* Cut one record from the directory item. */
- *cut_size = -(DEH_SIZE + entry_length (get_bh (path), le_ih, pos_in_item (path)));
+ *cut_size = -(DEH_SIZE + entry_length (get_last_bh (path), le_ih, pos_in_item (path)));
return M_CUT;
}
@@ -1945,15 +1945,15 @@
struct item_head * found_ih = get_ih (path);
if (is_direct_le_ih (found_ih)) {
- if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_bh (path)->b_size) !=
+ if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_last_bh (path)->b_size) !=
cpu_key_k_offset (p_s_key) ||
- op_bytes_number (found_ih, get_bh (path)->b_size) != pos_in_item (path))
+ op_bytes_number (found_ih, get_last_bh (path)->b_size) != pos_in_item (path))
reiserfs_panic (0, "PAP-5720: check_research_for_paste: "
"found direct item %h or position (%d) does not match to key %K",
found_ih, pos_in_item (path), p_s_key);
}
if (is_indirect_le_ih (found_ih)) {
- if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_bh (path)->b_size) != cpu_key_k_offset (p_s_key) ||
+ if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_last_bh (path)->b_size) != cpu_key_k_offset (p_s_key) ||
I_UNFM_NUM (found_ih) != pos_in_item (path) ||
get_ih_free_space (found_ih) != 0)
reiserfs_panic (0, "PAP-5730: check_research_for_paste: "
diff -Nru a/fs/buffer.c b/fs/buffer.c
--- a/fs/buffer.c Thu Jul 12 11:01:21 2001
+++ b/fs/buffer.c Thu Jul 12 11:01:21 2001
@@ -71,17 +71,6 @@
#define BH_ENTRY(list) list_entry((list), struct buffer_head, b_inode_buffers)
-static inline void get_bh(struct buffer_head * bh)
-{
- atomic_inc(&(bh)->b_count);
-}
-
-static inline void put_bh(struct buffer_head *bh)
-{
- smp_mb__before_atomic_dec();
- atomic_dec(&bh->b_count);
-}
-
/*
* Hash table gook..
*/
diff -Nru a/include/linux/fs.h b/include/linux/fs.h
--- a/include/linux/fs.h Thu Jul 12 11:01:21 2001
+++ b/include/linux/fs.h Thu Jul 12 11:01:21 2001
@@ -1069,6 +1069,17 @@
#define BUF_PROTECTED 3 /* Ramdisk persistent storage */
#define NR_LIST 4
+static inline void get_bh(struct buffer_head * bh)
+{
+ atomic_inc(&(bh)->b_count);
+}
+
+static inline void put_bh(struct buffer_head *bh)
+{
+ smp_mb__before_atomic_dec();
+ atomic_dec(&bh->b_count);
+}
+
/*
* This is called by bh->b_end_io() handlers when I/O has completed.
*/
diff -Nru a/include/linux/reiserfs_fs.h b/include/linux/reiserfs_fs.h
--- a/include/linux/reiserfs_fs.h Thu Jul 12 11:01:21 2001
+++ b/include/linux/reiserfs_fs.h Thu Jul 12 11:01:21 2001
@@ -1102,7 +1106,7 @@
#define PATH_H_PATH_OFFSET(p_s_path, n_h) ((p_s_path)->path_length - (n_h))
-#define get_bh(path) PATH_PLAST_BUFFER(path)
+#define get_last_bh(path) PATH_PLAST_BUFFER(path)
#define get_ih(path) PATH_PITEM_HEAD(path)
#define get_item_pos(path) PATH_LAST_POSITION(path)
#define get_item(path) ((void *)B_N_PITEM(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION (path)))
On Wednesday, July 18, 2001 05:53:42 PM +0200 Andrea Arcangeli
<[email protected]> wrote:
> On Wed, Jul 18, 2001 at 10:25:20AM -0400, Chris Mason wrote:
>> @@ -2597,7 +2599,7 @@
>>
>> if (bh) {
>> reiserfs_clean_and_file_buffer(bh) ;
>> - atomic_dec(&(bh->b_count)) ; /* get_hash incs this */
>> + put_bh(bh) ; /* get_hash grabs the buffer */
>> if (atomic_read(&(bh->b_count)) < 0) {
>> printk("journal-2165: bh->b_count < 0\n") ;
>> }
>
> in mainline you aren't calling reiserfs_clean_and_file_buffer above, so
> it rejects.
Correct, it goes cleanly on top of the direct->indirect cleanup patch that
was sent to l-k last week, and sent to linus for inclusion. ac already has
it, so it should apply there.
-chris
On Wed, Jul 18, 2001 at 10:25:20AM -0400, Chris Mason wrote:
> @@ -2597,7 +2599,7 @@
>
> if (bh) {
> reiserfs_clean_and_file_buffer(bh) ;
> - atomic_dec(&(bh->b_count)) ; /* get_hash incs this */
> + put_bh(bh) ; /* get_hash grabs the buffer */
> if (atomic_read(&(bh->b_count)) < 0) {
> printk("journal-2165: bh->b_count < 0\n") ;
> }
in mainline you aren't calling reiserfs_clean_and_file_buffer above, so
it rejects.
Andrea