Message-ID: <1a975830-96e8-92cb-560a-2c1ab41b8796@eknoes.de>
Date:   Tue, 4 Jan 2022 22:31:10 +0100
From:   Sönke Huster <soenke.huster@...oes.de>
To:     Luiz Augusto von Dentz <luiz.dentz@...il.com>
Cc:     "linux-bluetooth@...r.kernel.org" <linux-bluetooth@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Luiz Augusto Von Dentz <luiz.von.dentz@...el.com>,
        Marcel Holtmann <marcel@...tmann.org>
Subject: Re: [BUG] Page Fault in hci_inquiry_result_with_rssi_evt

Hi Luiz,

On 04.01.22 22:04, Luiz Augusto von Dentz wrote:
> Hi Sönke,
> 
> On Tue, Jan 4, 2022 at 11:31 AM Luiz Augusto von Dentz
> <luiz.dentz@...il.com> wrote:
>>
>> Hi Sönke,
>>
>> On Tue, Jan 4, 2022 at 3:50 AM Sönke Huster <soenke.huster@...oes.de> wrote:
>>>
>>> Hi Luiz,
>>>
>>> On 04.01.22 01:38, Luiz Augusto von Dentz wrote:
>>>> Hi Sönke,
>>>>
>>>> On Mon, Jan 3, 2022 at 3:41 PM Sönke Huster <soenke.huster@...oes.de> wrote:
>>>>>
>>>>> Hello,
>>>>>
>>>>> While fuzzing bluetooth-next I found the following bug:
>>>>>
>>>>> [   27.333034] BUG: unable to handle page fault for address: fffff61a1a1a1a1a
>>>>> [   27.333241] #PF: supervisor read access in kernel mode
>>>>> [   27.333241] #PF: error_code(0x0000) - not-present page
>>>>> [   27.333241] PGD 6dfd2067 P4D 6dfd2067 PUD 0
>>>>> [   27.333241] Oops: 0000 [#1] PREEMPT SMP KASAN NOPTI
>>>>> [   27.333241] CPU: 0 PID: 45 Comm: kworker/u3:2 Not tainted 5.16.0-rc1+ #81
>>>>> [   27.333241] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
>>>>> [   27.333241] Workqueue: hci0 hci_rx_work
>>>>> [   27.333241] RIP: 0010:hci_inquiry_result_with_rssi_evt+0xbc/0x950
>>>>> [   27.333241] Code: 8b 04 24 48 c1 e8 03 42 80 3c 28 00 0f 85 20 07 00 00 48 8b 04 24 4c 8b 28 48 b8 00 00 00 00 00 fc ff df 4c 89 0
>>>>> [   27.333241] RSP: 0018:ffffc900004ff9c8 EFLAGS: 00010212
>>>>> [   27.333241] RAX: dffffc0000000000 RBX: 0000000000000022 RCX: ffffffff834663d1
>>>>> [   27.333241] RDX: 1ffffa1a1a1a1a1a RSI: 0000000000000012 RDI: ffff88800affb074
>>>>> [   27.333241] RBP: ffff88800aae0000 R08: ffffffff844ef360 R09: ffffffff83487b35
>>>>> [   27.333241] R10: 000000000000002c R11: 0000000000000022 R12: ffff88800affb000
>>>>> [   27.333241] R13: ffffd0d0d0d0d0d0 R14: 0000000000000000 R15: ffff88800aae0000
>>>>> [   27.333241] FS:  0000000000000000(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
>>>>> [   27.333241] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>>>> [   27.333241] CR2: fffff61a1a1a1a1a CR3: 0000000004a26000 CR4: 00000000000006f0
>>>>> [   27.333241] Call Trace:
>>>>> [   27.333241]  <TASK>
>>>>> [   27.333241]  ? wait_for_completion_io+0x270/0x270
>>>>> [   27.333241]  ? hci_inquiry_result_evt+0x4b0/0x4b0
>>>>> [   27.333241]  hci_event_packet+0x3b11/0x7b10
>>>>> [   27.333241]  ? lock_chain_count+0x20/0x20
>>>>> [   27.333241]  ? hci_cmd_status_evt.constprop.0+0x4ea0/0x4ea0
>>>>> [   27.333241]  ? sysvec_reboot+0x50/0xc0
>>>>> [   27.333241]  ? find_held_lock+0x2c/0x110
>>>>> [   27.333241]  ? lock_release+0x3b2/0x6f0
>>>>> [   27.333241]  ? skb_dequeue+0x110/0x1a0
>>>>> [   27.333241]  ? mark_held_locks+0x9e/0xe0
>>>>> [   27.333241]  ? lockdep_hardirqs_on_prepare+0x17b/0x400
>>>>> [   27.333241]  hci_rx_work+0x4d3/0xb90
>>>>> [   27.333241]  process_one_work+0x904/0x1590
>>>>> [   27.333241]  ? lock_release+0x6f0/0x6f0
>>>>> [   27.333241]  ? pwq_dec_nr_in_flight+0x230/0x230
>>>>> [   27.333241]  ? rwlock_bug.part.0+0x90/0x90
>>>>> [   27.333241]  ? _raw_spin_lock_irq+0x41/0x50
>>>>> [   27.333241]  worker_thread+0x578/0x1310
>>>>> [   27.333241]  ? process_one_work+0x1590/0x1590
>>>>> [   27.333241]  kthread+0x3b2/0x490
>>>>> [   27.333241]  ? _raw_spin_unlock_irq+0x1f/0x40
>>>>> [   27.333241]  ? set_kthread_struct+0x100/0x100
>>>>> [   27.333241]  ret_from_fork+0x22/0x30
>>>>> [   27.333241]  </TASK>
>>>>> [   27.333241] Modules linked in:
>>>>> [   27.333241] CR2: fffff61a1a1a1a1a
>>>>> [   27.333241] ---[ end trace 6a6825484c8fefa6 ]---
>>>>> [   27.333241] RIP: 0010:hci_inquiry_result_with_rssi_evt+0xbc/0x950
>>>>> [   27.333241] Code: 8b 04 24 48 c1 e8 03 42 80 3c 28 00 0f 85 20 07 00 00 48 8b 04 24 4c 8b 28 48 b8 00 00 00 00 00 fc ff df 4c 89 0
>>>>> [   27.333241] RSP: 0018:ffffc900004ff9c8 EFLAGS: 00010212
>>>>> [   27.333241] RAX: dffffc0000000000 RBX: 0000000000000022 RCX: ffffffff834663d1
>>>>> [   27.333241] RDX: 1ffffa1a1a1a1a1a RSI: 0000000000000012 RDI: ffff88800affb074
>>>>> [   27.333241] RBP: ffff88800aae0000 R08: ffffffff844ef360 R09: ffffffff83487b35
>>>>> [   27.333241] R10: 000000000000002c R11: 0000000000000022 R12: ffff88800affb000
>>>>> [   27.333241] R13: ffffd0d0d0d0d0d0 R14: 0000000000000000 R15: ffff88800aae0000
>>>>> [   27.333241] FS:  0000000000000000(0000) GS:ffff88806ce00000(0000) knlGS:0000000000000000
>>>>> [   27.333241] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>>>> [   27.333241] CR2: fffff61a1a1a1a1a CR3: 0000000004a26000 CR4: 00000000000006f0
>>>>> [   27.379996] kworker/u3:2 (45) used greatest stack depth: 27736 bytes left
>>>>>
>>>>> It occurs when sending the following frame to the kernel:
>>>>>
>>>>> $ xxd crashes/hci_inquiry_result_with_rssi_evt
>>>>> 00000000: 0422 24d0 d0d0 d0d0 d0ff ff              ."$........
>>>>>
>>>>> The bug was introduced with the commit "Bluetooth: HCI: Use skb_pull_data to parse Inquiry Result with RSSI event" (https://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next.git/commit/?id=8d08d324fdcb7).
>>>>
>>>> That is pretty weird, the data seems to be the following:
>>>>
>>>> 04 -> HCI_EVENT_PKT
>>>> 22 -> HCI_EV_INQUIRY_RESULT_WITH_RSSI
>>>> 24 -> hci_ev_inquiry_result_rssi.num
>>>> d0 d0d0 d0d0 d0ff ff
>>>>
>>>> But this should never evaluate to true for:
>>>>
>>>> if (skb->len == flex_array_size(ev, res2->info, ev->res2->num)) {
>>>> ...
>>>> } else if (skb->len == flex_array_size(ev, res1->info, ev->res1->num)) {
>>>>
>>>
>>> I'm sorry, I forgot to mention the affected source code, the check seems to be too late. GDB says it is already happening in net/bluetooth/hci_event.c:4519:
>>>
>>> (gdb) list *hci_inquiry_result_with_rssi_evt+0x9b
>>> 0xffffffff83470d8b is in hci_inquiry_result_with_rssi_evt (net/bluetooth/hci_event.c:4519).
>>> 4514            struct inquiry_data data;
>>> 4515            int i;
>>> 4516
>>> 4517            bt_dev_dbg(hdev, "num_rsp %d", ev->res1->num);
>>> 4518
>>> 4519            if (!ev->res1->num) # <- page fault here
>>> 4520                    return;
>>>
>>> I just reproduced it on the HEAD of bluetooth-next (for-net-next-2021-12-29).
>>
>> That would mean there is less data than the hci_ev.min_len, but that
>> is checked in hci_event_func prior to calling the hci_ev.func:
>>
>> if (skb->len < ev->min_len) {
>> bt_dev_err(hdev, "unexpected event 0x%2.2x length: %u < %u",
>>    event, skb->len, ev->min_len);
>> return;
>> }
>>
>>>
>>>> These requires the data to be multiple of sizeof(struct
>>>> inquiry_info_rssi_pscan) = 15 bytes or sizeof(struct
>>>> inquiry_info_rssi) = 14 bytes respectively where the data left is just
>>>> 8 bytes long, besides with the number of entries being 0x24 this shall
>>>> be well beyond skb->len which shall have cause the else clause:
>>>>
>>>>   } else {
>>>>       bt_dev_err(hdev, "Malformed HCI Event: 0x%2.2x",
>>>>                          HCI_EV_INQUIRY_RESULT_WITH_RSSI);
>>>>   }
>>>>
>>>
>>> I think prior to the commit that introduced that, the check was made before casting it to the struct, so from the "raw" skb->data:
>>>
>>> -       int num_rsp = *((__u8 *) skb->data);
>>> -       if ((skb->len - 1) / num_rsp != sizeof(struct inquiry_info_with_rssi)) {
>>
>> That is still being checked, anyway that would cause a page fault when
>> accessing num_rsp so I don't understand why it would cause a page
>> fault now when previously it didn't even check if there were any bytes
>> to be read, perhaps there is some memory alignment at play.
>>
>>>> Anyway the bluetooth-next has been updated since last week so I first
>>>> attempt to reproduce with a fresh clone of it since we did some fixups
>>>> since then.
>>>>
>>>
>>> Btw, what is the best way to provide an easily reproducible bug report here, I did not figure that out yet.
>>>
>>> Also, when searching for bugs (to ideally provide patches), I am currently searching on the HEAD of bluetooth-next. As far as I understood it, the tags "for-net-..." should be more or less stable, as they are merged to net-next which makes its way to mainline, right?
>>
>> Yep, those are the tags we sent to net-next but in general
>> bluetooth-next shall also be stable since we run CI on every patch
>> nowadays.
> 
> Ive just sent a patch trying to eliminate the use of the union since
> that perhaps cause the compiler to use a different alignment, Im also

The patch fixes it, I can't reproduce it anymore and the kernel log shows the expected  "Bluetooth: hci0: Malformed HCI Event: 0x22".

> thinking on introducing some mechanism to our emulator to generate
> arbitrary events in a form a command that generates an event based on
> its parameter.
> 
> https://patchwork.kernel.org/project/bluetooth/patch/20220104210051.2982871-1-luiz.dentz@gmail.com/
> 

Powered by blists - more mailing lists