* Re: [PATCH] Btrfs: fix use-after-free when cloning a trailing file hole
2014-06-29 19:43 [PATCH] Btrfs: fix use-after-free when cloning a trailing file hole Filipe David Borba Manana
@ 2014-06-29 19:23 ` Chris Mason
2014-06-29 19:28 ` Filipe David Manana
2014-06-29 20:45 ` [PATCH v2] " Filipe David Borba Manana
1 sibling, 1 reply; 4+ messages in thread
From: Chris Mason @ 2014-06-29 19:23 UTC (permalink / raw)
To: Filipe David Borba Manana, linux-btrfs
On 06/29/2014 03:43 PM, Filipe David Borba Manana wrote:
> The transaction handle was being used after being freed.
>
> Cc: Chris Mason <clm@fb.com>
> Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
> ---
> fs/btrfs/ioctl.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
> index 02dc64b..2562dc7 100644
> --- a/fs/btrfs/ioctl.c
> +++ b/fs/btrfs/ioctl.c
> @@ -3565,12 +3565,10 @@ process_slot:
> btrfs_end_transaction(trans, root);
> goto out;
> }
> - ret = clone_finish_inode_update(trans, inode, destoff + len,
> - destoff, olen);
> - if (ret)
> - goto out;
> clone_update_extent_map(inode, trans, path, NULL, last_dest_end,
> destoff + len - last_dest_end);
> + ret = clone_finish_inode_update(trans, inode, destoff + len,
> + destoff, olen);
> }
>
> out:
>
What about the path? It has been released, so it should either be NULL
or we have other problems ;)
-chris
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH v2] Btrfs: fix use-after-free when cloning a trailing file hole
2014-06-29 19:43 [PATCH] Btrfs: fix use-after-free when cloning a trailing file hole Filipe David Borba Manana
2014-06-29 19:23 ` Chris Mason
@ 2014-06-29 20:45 ` Filipe David Borba Manana
1 sibling, 0 replies; 4+ messages in thread
From: Filipe David Borba Manana @ 2014-06-29 20:45 UTC (permalink / raw)
To: linux-btrfs; +Cc: Filipe David Borba Manana, Chris Mason
The transaction handle was being used after being freed.
Cc: Chris Mason <clm@fb.com>
Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
---
V2: Removed file extent item argument to clone_update_extent_map() for
more clarity.
fs/btrfs/ioctl.c | 20 +++++++++-----------
1 file changed, 9 insertions(+), 11 deletions(-)
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 02dc64b..2a99f49 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -3142,7 +3142,6 @@ out:
static void clone_update_extent_map(struct inode *inode,
const struct btrfs_trans_handle *trans,
const struct btrfs_path *path,
- struct btrfs_file_extent_item *fi,
const u64 hole_offset,
const u64 hole_len)
{
@@ -3157,7 +3156,11 @@ static void clone_update_extent_map(struct inode *inode,
return;
}
- if (fi) {
+ if (path) {
+ struct btrfs_file_extent_item *fi;
+
+ fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
+ struct btrfs_file_extent_item);
btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
em->generation = -1;
if (btrfs_file_extent_type(path->nodes[0], fi) ==
@@ -3511,18 +3514,15 @@ process_slot:
btrfs_item_ptr_offset(leaf, slot),
size);
inode_add_bytes(inode, datal);
- extent = btrfs_item_ptr(leaf, slot,
- struct btrfs_file_extent_item);
}
/* If we have an implicit hole (NO_HOLES feature). */
if (drop_start < new_key.offset)
clone_update_extent_map(inode, trans,
- path, NULL, drop_start,
+ NULL, drop_start,
new_key.offset - drop_start);
- clone_update_extent_map(inode, trans, path,
- extent, 0, 0);
+ clone_update_extent_map(inode, trans, path, 0, 0);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
@@ -3565,12 +3565,10 @@ process_slot:
btrfs_end_transaction(trans, root);
goto out;
}
+ clone_update_extent_map(inode, trans, NULL, last_dest_end,
+ destoff + len - last_dest_end);
ret = clone_finish_inode_update(trans, inode, destoff + len,
destoff, olen);
- if (ret)
- goto out;
- clone_update_extent_map(inode, trans, path, NULL, last_dest_end,
- destoff + len - last_dest_end);
}
out:
--
1.9.1
^ permalink raw reply related [flat|nested] 4+ messages in thread