[Linux-kernel-mentees] [PATCH net] Bluetooth: Fix NULL pointer dereference in amp_read_loc_assoc_final_data()

[Peilin Ye]

#syz test: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master

[syzbot]

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Reported-and-tested-by: syzbot+f4fb0e...@syzkaller.appspotmail.com

Tested on:

commit: 049eb096 Merge tag 'pci-v5.9-changes' of git://git.kernel...
git tree: upstream
kernel config: https://syzkaller.appspot.com/x/.config?x=8dfc9a51cda5095a
dashboard link: https://syzkaller.appspot.com/bug?extid=f4fb0eaafdb51c32a153
compiler: gcc (GCC) 10.1.0-syz 20200507
patch: https://syzkaller.appspot.com/x/patch.diff?x=109369e6900000

Note: testing is done by a robot and is best-effort only.

[Peilin Ye]

Prevent amp_read_loc_assoc_final_data() from dereferencing `mgr` as NULL.

Reported-and-tested-by: syzbot+f4fb0e...@syzkaller.appspotmail.com
Fixes: 9495b2ee757f ("Bluetooth: AMP: Process Chan Selected event")
Signed-off-by: Peilin Ye <yepei...@gmail.com>
---
net/bluetooth/amp.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/net/bluetooth/amp.c b/net/bluetooth/amp.c
index 9c711f0dfae3..be2d469d6369 100644
--- a/net/bluetooth/amp.c
+++ b/net/bluetooth/amp.c
@@ -297,6 +297,9 @@ void amp_read_loc_assoc_final_data(struct hci_dev *hdev,
struct hci_request req;
int err;

+ if (!mgr)
+ return;
+
cp.phy_handle = hcon->handle;
cp.len_so_far = cpu_to_le16(0);
cp.max_len = cpu_to_le16(hdev->amp_assoc_size);
--
2.25.1

[Dmitry Vyukov]

Not sure what happened here, but the merged patch somehow has a
different author and no Reported-by tag:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e8bd76ede155fd54d8c41d045dda43cd3174d506
so let's tell syzbot what fixed it manually:
#syz fix:
Bluetooth: Fix null pointer dereference in amp_read_loc_assoc_final_data

[Gopal Tiwari]

Hi,

I tried to search the patch for one of the bugzilla reported (Internal) https://bugzilla.redhat.com/show_bug.cgi?id=1916057 with the traces

[ 405.938525] Workqueue: hci0 hci_rx_work [bluetooth]
[ 405.941360] RIP: 0010:amp_read_loc_assoc_final_data+0xfc/0x1c0 [bluetooth]
[ 405.944740] Code: 89 44 24 29 48 b8 00 00 00 00 00 fc ff df 0f b6 04 02 84 c0 74 08 3c 01 0f 8e 9d 00 00 00 0f b7 85 c0 03 00 00 66 89 44 24 2b <f0> 41 80 4c 24 30 04 4c 8d 64 24 68 48 89 ee 4c 89 e7 e8 3d 48 fe
[ 405.952396] RSP: 0018:ffff88802ea0f838 EFLAGS: 00010246
[ 405.955368] RAX: 0000000000000000 RBX: 1ffff11005d41f08 RCX: dffffc0000000000
[ 405.958669] RDX: 1ffff110254cc878 RSI: ffff88802eeee000 RDI: ffff88812a6643c0
[ 405.961980] RBP: ffff88812a664000 R08: 0000000000000000 R09: 0000000000000000
[ 405.965319] R10: ffff88802ea0fd00 R11: 0000000000000000 R12: 0000000000000000
[ 405.968624] R13: 0000000000000041 R14: ffff88802b836800 R15: ffff8881250570c0
[ 405.971989] FS: 0000000000000000(0000) GS:ffff888055a00000(0000) knlGS:0000000000000000
[ 405.975645] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 405.978755] CR2: 0000000000000030 CR3: 000000002d200000 CR4: 0000000000340ee0
[ 405.982150] Call Trace:
[ 405.984768] ? amp_read_loc_assoc+0x170/0x170 [bluetooth]
[ 405.987875] ? rcu_read_unlock+0x50/0x50
[ 405.990663] ? deref_stack_reg+0xf0/0xf0
[ 405.993403] ? __module_address+0x3f/0x370
[ 405.996184] ? hci_cmd_work+0x180/0x330 [bluetooth]
[ 405.999170] ? hci_conn_hash_lookup_handle+0x1a1/0x270 [bluetooth]
[ 406.002354] hci_event_packet+0x1476/0x7e00 [bluetooth]
[ 406.005407] ? arch_stack_walk+0x8f/0xf0
[ 406.008206] ? ret_from_fork+0x27/0x50
[ 406.010887] ? hci_cmd_complete_evt+0xbf70/0xbf70 [bluetooth]
[ 406.013933] ? stack_trace_save+0x8a/0xb0
[ 406.016618] ? do_profile_hits.isra.4.cold.9+0x2d/0x2d
[ 406.019483] ? lock_acquire+0x1a3/0x970
[ 406.022092] ? __wake_up_common_lock+0xaf/0x130


I didn't found any solution upstream. After the vmcore analysis I found what is wrong. And took reference from the following patch, which seems to be on the similar line

commit 6dfccd13db2ff2b709ef60a50163925d477549aa
Author: Anmol Karn <anmol.k...@gmail.com>
Date: Wed Sep 30 19:48:13 2020 +0530

Bluetooth: Fix null pointer dereference in hci_event_packet()

AMP_MGR is getting derefernced in hci_phy_link_complete_evt(), when called
from hci_event_packet() and there is a possibility, that hcon->amp_mgr may
not be found when accessing after initialization of hcon.

- net/bluetooth/hci_event.c:4945

How we can avoid this scenario. So I made the chages and tested. It worked or avoided the kernel panic. But I really don't know that some one has already posted the patch. I would have love to backport the patch, I was more of looking for the fix. That's where I didn't applied the reported-by tag as I thought it reported internal only.

Thanks & regards,
Gopal Tiwari

[Dmitry Vyukov]

Hi Gopal,

I think it's somewhat inherent to the current kernel unstructured
processes with bugs being reported on mailing lists, bugzilla,
distro-specific trackers.
One useful thing, though, is searching Lore, e.g. searching for just
the crashing function:
https://lore.kernel.org/lkml/?q=amp_read_loc_assoc_final_data
gives the report and the patch (if we filter out all entries produced
by your patch, which obviously wasn't yet there before you wrote it
:)):

12. [Linux-kernel-mentees] [PATCH net] Bluetooth: Fix NULL pointer
dereference in amp_read_loc_assoc_final_data()
- by Peilin Ye @ 2020-08-08 4:04 UTC [21%]

13. KASAN: null-ptr-deref Write in amp_read_loc_assoc_final_data
- by syzbot @ 2020-07-31 17:04 UTC [13%]