[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20241025043848.1981317-1-lizhi.xu@windriver.com>
Date: Fri, 25 Oct 2024 12:38:48 +0800
From: Lizhi Xu <lizhi.xu@...driver.com>
To: <quwenruo.btrfs@....com>
CC: <clm@...com>, <dsterba@...e.com>, <josef@...icpanda.com>,
<linux-btrfs@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
<lizhi.xu@...driver.com>,
<syzbot+3030e17bd57a73d39bd7@...kaller.appspotmail.com>,
<syzkaller-bugs@...glegroups.com>
Subject: Re: [syzbot] [bcachefs?] KASAN: slab-use-after-free Read in bch2_reconstruct_alloc
On Fri, 25 Oct 2024 14:49:48 +1030, Qu Wenruo wrote:
> > use the input logical can't find the extent root, so add sanity check for
> > extent root before search slot.
> >
> > #syz test
> >
> > diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
> > index f8e1d5b2c512..87eaf5dd2d5d 100644
> > --- a/fs/btrfs/backref.c
> > +++ b/fs/btrfs/backref.c
> > @@ -2213,6 +2213,9 @@ int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
> > key.objectid = logical;
> > key.offset = (u64)-1;
> >
> > + if (!extent_root)
> > + return -ENOENT;
>
> Considering we have a lot of such btrfs_search_slot() without checking
> if the csum/extent root is NULL, can we move the check into
> btrfs_search_slot()?
Yes, judging in btrfs_search_slot can fix the current issue while also
avoiding similar problems.
#syz test
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 0cc919d15b14..9c05cab473f5 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -2010,7 +2010,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
const struct btrfs_key *key, struct btrfs_path *p,
int ins_len, int cow)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
+ struct btrfs_fs_info *fs_info;
struct extent_buffer *b;
int slot;
int ret;
@@ -2023,6 +2023,10 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
int min_write_lock_level;
int prev_cmp;
+ if (!root)
+ return -EINVAL;
+
+ fs_info = root->fs_info;
might_sleep();
lowest_level = p->lowest_level;
Powered by blists - more mailing lists