From: Mingming Cao <cmm@us.ibm.com>
This patch is trying to do lazy discard: keep the old reservation window
temporarilly until we find the new reservation window, only do remove/add if
the new reservation window locate different than the old one. (The current
reservation code will discard the old one first,then search the new one). Two
reasons:
a. If the ext3_find_goal() does a good job, the reservation windows on the
list should not very close to each other. So an inode's new reservation
window is likely located just next to it's old one, it's position in the
whole list is unchanged, no need to do remove and then add the new one to
the list in the same location. Just update the start block and end
block.
b. If we failed to find a new reservation in the goal group and move on
the search to the next group, having the old reservation around temporally
could allow us to search the list directly after the old window. Otherwise
we lost where we were and has to start from the beginning of the list.
Eventually the old window will be discard when we found a new one.
DESC
ext3 discard reservation in last iput fix patch
EDESC
From: Mingming Cao <cmm@us.ibm.com>
ext3_discard_reservation() should not be called on every iput(). Now it is
moved to ext3_delete_inode(), so it is only called on the last iput().
DESC
Fix lazy reservation discard
EDESC
->delete_inode() is only called when the inode is being deleted. So we're
leaving inodes on the reservation list when they are reclaimed by the VM.
So take them off the reservation list in ->clear_inode instead.
DESC
ext3 reservations: bad_inode fix
EDESC
From: Mingming Cao <cmm@us.ibm.com>
Thanks Badari!! The easiest fix is checking whether it is a bad inode in
ext3_clear_inode() (like what ext2_put_inode() does):
DESC
ext3 reservation discard race fix
EDESC
From: Mingming Cao <cmm@us.ibm.com>
This should fix a reservation window race between multiple processes when
one process closed the file while another one is in the middle of block
allocation using the inode's reservation window. reservation window should
be discard on the last writer on the inode, not the last writer on the
filp.
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/fs/ext3/balloc.c | 22 +++++++++++++++-------
25-akpm/fs/ext3/file.c | 4 +++-
25-akpm/fs/ext3/inode.c | 13 -------------
25-akpm/fs/ext3/super.c | 11 ++++-------
4 files changed, 22 insertions(+), 28 deletions(-)
diff -puN fs/ext3/balloc.c~ext3-lazy-discard-reservation-window-patch fs/ext3/balloc.c
--- 25/fs/ext3/balloc.c~ext3-lazy-discard-reservation-window-patch 2004-08-30 00:50:24.706008520 -0700
+++ 25-akpm/fs/ext3/balloc.c 2004-08-30 00:50:24.717006848 -0700
@@ -726,7 +726,7 @@ static int alloc_new_reservation(struct
start_block = goal + group_first_block;
size = atomic_read(&my_rsv->rsv_goal_size);
- /* if we have a old reservation, discard it first */
+ /* if we have a old reservation, start the search from the old rsv */
if (!rsv_is_empty(my_rsv)) {
/*
* if the old reservation is cross group boundary
@@ -748,8 +748,7 @@ static int alloc_new_reservation(struct
/* remember where we are before we discard the old one */
if (my_rsv->rsv_end + 1 > start_block)
start_block = my_rsv->rsv_end + 1;
- search_head = list_entry(my_rsv->rsv_list.prev,
- struct reserve_window, rsv_list);
+ search_head = my_rsv;
if ((my_rsv->rsv_alloc_hit > (my_rsv->rsv_end - my_rsv->rsv_start + 1) / 2)) {
/*
* if we previously allocation hit ration is greater than half
@@ -761,7 +760,6 @@ static int alloc_new_reservation(struct
size = EXT3_MAX_RESERVE_BLOCKS;
atomic_set(&my_rsv->rsv_goal_size, size);
}
- rsv_window_remove(my_rsv);
}
else {
/*
@@ -828,9 +826,16 @@ retry:
found_rsv_window:
/*
* great! the reservable space contains some free blocks.
- * Insert it to the list.
- */
- rsv_window_add(my_rsv, prev_rsv);
+ * if the search returns that we should add the new
+ * window just next to where the old window, we don't
+ * need to remove the old window first then add it to the
+ * same place, just update the new start and new end.
+ */
+ if (my_rsv != prev_rsv) {
+ if (!rsv_is_empty(my_rsv))
+ rsv_window_remove(my_rsv);
+ rsv_window_add(my_rsv, prev_rsv);
+ }
my_rsv->rsv_start = reservable_space_start;
my_rsv->rsv_end = my_rsv->rsv_start + size - 1;
return 0; /* succeed */
@@ -932,6 +937,9 @@ ext3_try_to_allocate_with_rsv(struct sup
if (!goal_in_my_reservation(my_rsv, goal, group, sb))
goal = -1;
}
+ if ((my_rsv->rsv_start >= group_first_block + EXT3_BLOCKS_PER_GROUP(sb))
+ || (my_rsv->rsv_end < group_first_block))
+ BUG();
ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh, goal,
my_rsv);
if (ret >= 0)
diff -puN fs/ext3/file.c~ext3-lazy-discard-reservation-window-patch fs/ext3/file.c
--- 25/fs/ext3/file.c~ext3-lazy-discard-reservation-window-patch 2004-08-30 00:50:24.707008368 -0700
+++ 25-akpm/fs/ext3/file.c 2004-08-30 00:50:24.717006848 -0700
@@ -33,7 +33,9 @@
*/
static int ext3_release_file (struct inode * inode, struct file * filp)
{
- if (filp->f_mode & FMODE_WRITE)
+ /* if we are the last writer on the inode, drop the block reservation */
+ if ((filp->f_mode & FMODE_WRITE) &&
+ (atomic_read(&inode->i_writecount) == 1))
ext3_discard_reservation(inode);
if (is_dx(inode) && filp->private_data)
ext3_htree_free_dir_info(filp->private_data);
diff -puN fs/ext3/inode.c~ext3-lazy-discard-reservation-window-patch fs/ext3/inode.c
--- 25/fs/ext3/inode.c~ext3-lazy-discard-reservation-window-patch 2004-08-30 00:50:24.709008064 -0700
+++ 25-akpm/fs/ext3/inode.c 2004-08-30 00:50:24.720006392 -0700
@@ -179,19 +179,6 @@ static int ext3_journal_test_restart(han
}
/*
- * Called at each iput()
- *
- * The inode may be "bad" if ext3_read_inode() saw an error from
- * ext3_get_inode(), so we need to check that to avoid freeing random disk
- * blocks.
- */
-void ext3_put_inode(struct inode *inode)
-{
- if (!is_bad_inode(inode))
- ext3_discard_reservation(inode);
-}
-
-/*
* Called at the last iput() if i_nlink is zero.
*/
void ext3_delete_inode (struct inode * inode)
diff -puN fs/ext3/super.c~ext3-lazy-discard-reservation-window-patch fs/ext3/super.c
--- 25/fs/ext3/super.c~ext3-lazy-discard-reservation-window-patch 2004-08-30 00:50:24.712007608 -0700
+++ 25-akpm/fs/ext3/super.c 2004-08-30 00:50:24.721006240 -0700
@@ -490,10 +490,9 @@ static void destroy_inodecache(void)
printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
}
-#ifdef CONFIG_EXT3_FS_POSIX_ACL
-
static void ext3_clear_inode(struct inode *inode)
{
+#ifdef CONFIG_EXT3_FS_POSIX_ACL
if (EXT3_I(inode)->i_acl &&
EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
posix_acl_release(EXT3_I(inode)->i_acl);
@@ -504,11 +503,10 @@ static void ext3_clear_inode(struct inod
posix_acl_release(EXT3_I(inode)->i_default_acl);
EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
}
-}
-
-#else
-# define ext3_clear_inode NULL
#endif
+ if (!is_bad_inode(inode))
+ ext3_discard_reservation(inode);
+}
#ifdef CONFIG_QUOTA
@@ -558,7 +556,6 @@ static struct super_operations ext3_sops
.read_inode = ext3_read_inode,
.write_inode = ext3_write_inode,
.dirty_inode = ext3_dirty_inode,
- .put_inode = ext3_put_inode,
.delete_inode = ext3_delete_inode,
.put_super = ext3_put_super,
.write_super = ext3_write_super,
_