2017-10-30 13:32:19

by Artem Blagodarenko

[permalink] [raw]
Subject: [PATCH] e2fsck: Fix access after free for dx_db structure

dx_db structure is freed after fixing of PR_2_HTREE_BAD_ROOT
problem. Next code block use this structure to unerstand if leaf
is beeng processed.

If dx_db is freed, then root block is being processed and if_leaf
need to be set to 0.

Signed-off-by: Artem Blagodarenko <[email protected]>
---
e2fsck/pass2.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/e2fsck/pass2.c b/e2fsck/pass2.c
index 09c16179..7bda37a7 100644
--- a/e2fsck/pass2.c
+++ b/e2fsck/pass2.c
@@ -1109,7 +1109,7 @@ inline_read_fail:
((fs->blocksize - (8 + dx_csum_size)) /
sizeof(struct ext2_dx_entry))))
dx_db->type = DX_DIRBLOCK_NODE;
- is_leaf = (dx_db->type == DX_DIRBLOCK_LEAF);
+ is_leaf = dx_db ? (dx_db->type == DX_DIRBLOCK_LEAF) : 0;
}
out_htree:

--
2.13.5 (Apple Git-94)


2017-11-03 20:00:16

by Andreas Dilger

[permalink] [raw]
Subject: Re: [PATCH] e2fsck: Fix access after free for dx_db structure


> On Oct 30, 2017, at 7:32 AM, Artem Blagodarenko <[email protected]> wrote:
>
> dx_db structure is freed after fixing of PR_2_HTREE_BAD_ROOT
> problem. Next code block use this structure to unerstand if leaf
> is beeng processed.
>
> If dx_db is freed, then root block is being processed and if_leaf
> need to be set to 0.
>
> Signed-off-by: Artem Blagodarenko <[email protected]>
> ---
> e2fsck/pass2.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/e2fsck/pass2.c b/e2fsck/pass2.c
> index 09c16179..7bda37a7 100644
> --- a/e2fsck/pass2.c
> +++ b/e2fsck/pass2.c
> @@ -1109,7 +1109,7 @@ inline_read_fail:
> ((fs->blocksize - (8 + dx_csum_size)) /
> sizeof(struct ext2_dx_entry))))
> dx_db->type = DX_DIRBLOCK_NODE;
> - is_leaf = (dx_db->type == DX_DIRBLOCK_LEAF);
> + is_leaf = dx_db ? (dx_db->type == DX_DIRBLOCK_LEAF) : 0;

Ah, this code is a bit confusing. From looking at this patch, and even
looking at the code directly, I thought "the previous line is also
accessing dx_db, so what is the problem?" However, there really IS a
problem, because the previous line is part of a different if/else block.

It would be good to add braces around the "else" part of the condition
to make it clear that this is a separate block of code, which is why we
typically ask for matching braces for both legs of an if/else block. It
would also be good to set "dx_db = NULL" instead of setting it to "0".

Cheers, Andreas






Attachments:
signature.asc (195.00 B)
Message signed with OpenPGP