Re: usb/net/qmi_wwan: divide error in qmi_wwan_probe/usbnet_probe

[Andrey Konovalov]

On Mon, Nov 6, 2017 at 3:06 PM, Bjørn Mork  wrote:
> Andrey Konovalov  writes:
>
>> Hi!
>>
>> I've got the following report while fuzzing the kernel with syzkaller.
>
> Thanks.  It would have helped a lot of you said *what* you were fuzzing,
> though But based on where the bug is, I assume it is USB
> descriptors?

Yes, I was connecting USB devices with random descriptors. I though
it's obvious from the stack trace, but I'll include that in the
following reports. Thanks!

>
>
>> On commit 39dae59d66acd86d1de24294bd2f343fd5e7a625 (4.14-rc8).
>>
>> qmi_wwan 1-1:0.4: cdc-wdm0: USB WDM device
>> divide error:  [#1] PREEMPT SMP KASAN
>> Modules linked in:
>> CPU: 0 PID: 24 Comm: kworker/0:1 Not tainted 4.14.0-rc8-44453-g1fdc1a82c34f 
>> #56
>> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
>> Workqueue: usb_hub_wq hub_event
>> task: 88006bef5c00 task.stack: 88006bf6
>> RIP: 0010:usbnet_update_max_qlen+0x24d/0x390 drivers/net/usb/usbnet.c:355
>> RSP: 0018:88006bf67508 EFLAGS: 00010246
>> RAX: 000163c8 RBX: 8800621fce40 RCX: 8800621fcf34
>> RDX:  RSI: 837ecb7a RDI: 8800621fcf34
>> RBP: 88006bf67520 R08: 88006bef5c00 R09: ed000c43f881
>> R10: ed000c43f880 R11: 8800621fc406 R12: 0003
>> R13: 85c71de0 R14:  R15: 
>> FS:  () GS:88006ca0() knlGS:
>> CS:  0010 DS:  ES:  CR0: 80050033
>> CR2: 7ffe9c0d6dac CR3: 614f4000 CR4: 06f0
>> Call Trace:
>>  usbnet_probe+0x18b5/0x2790 drivers/net/usb/usbnet.c:1783
>>  qmi_wwan_probe+0x133/0x220 drivers/net/usb/qmi_wwan.c:1338
>
> So, looking over this again and again, the only obviously risky division
> I can see is the usbnet_update_max_qlen() call where we divide on both
> dev->rx_urb_size and dev->hard_mtu.
>
> I don't think dev->rx_urb_size can be 0 unless dev->hard_mtu is 0.  But
> the latter is possible, and this is a bug. In qmi_wwan_bind(), which is
> called by usbnet_probe() prior to the usbnet_update_max_qlen() call, we
> do
>
> if (cdc_ether) {
> dev->hard_mtu = le16_to_cpu(cdc_ether->wMaxSegmentSize);
> usbnet_get_ethernet_addr(dev, cdc_ether->iMACAddress);
> }
>
> which will be fatal if cdc_ether->wMaxSegmentSize is 0.  I assume that
> is what your fuzzer did?  Fix coming up RSN.  Thanks a lot for pointing
> this out.

I don't see the crash with your patches. Thanks for a quick fix!

>
>
> Bjørn

Re: usb/net/qmi_wwan: divide error in qmi_wwan_probe/usbnet_probe

[Bjørn Mork]

Andrey Konovalov  writes:

> Hi!
>
> I've got the following report while fuzzing the kernel with syzkaller.

Thanks.  It would have helped a lot of you said *what* you were fuzzing,
though But based on where the bug is, I assume it is USB
descriptors?


> On commit 39dae59d66acd86d1de24294bd2f343fd5e7a625 (4.14-rc8).
>
> qmi_wwan 1-1:0.4: cdc-wdm0: USB WDM device
> divide error:  [#1] PREEMPT SMP KASAN
> Modules linked in:
> CPU: 0 PID: 24 Comm: kworker/0:1 Not tainted 4.14.0-rc8-44453-g1fdc1a82c34f 
> #56
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
> Workqueue: usb_hub_wq hub_event
> task: 88006bef5c00 task.stack: 88006bf6
> RIP: 0010:usbnet_update_max_qlen+0x24d/0x390 drivers/net/usb/usbnet.c:355
> RSP: 0018:88006bf67508 EFLAGS: 00010246
> RAX: 000163c8 RBX: 8800621fce40 RCX: 8800621fcf34
> RDX:  RSI: 837ecb7a RDI: 8800621fcf34
> RBP: 88006bf67520 R08: 88006bef5c00 R09: ed000c43f881
> R10: ed000c43f880 R11: 8800621fc406 R12: 0003
> R13: 85c71de0 R14:  R15: 
> FS:  () GS:88006ca0() knlGS:
> CS:  0010 DS:  ES:  CR0: 80050033
> CR2: 7ffe9c0d6dac CR3: 614f4000 CR4: 06f0
> Call Trace:
>  usbnet_probe+0x18b5/0x2790 drivers/net/usb/usbnet.c:1783
>  qmi_wwan_probe+0x133/0x220 drivers/net/usb/qmi_wwan.c:1338

So, looking over this again and again, the only obviously risky division
I can see is the usbnet_update_max_qlen() call where we divide on both
dev->rx_urb_size and dev->hard_mtu.

I don't think dev->rx_urb_size can be 0 unless dev->hard_mtu is 0.  But
the latter is possible, and this is a bug. In qmi_wwan_bind(), which is
called by usbnet_probe() prior to the usbnet_update_max_qlen() call, we
do

if (cdc_ether) {
dev->hard_mtu = le16_to_cpu(cdc_ether->wMaxSegmentSize);
usbnet_get_ethernet_addr(dev, cdc_ether->iMACAddress);
}

which will be fatal if cdc_ether->wMaxSegmentSize is 0.  I assume that
is what your fuzzer did?  Fix coming up RSN.  Thanks a lot for pointing
this out.


Bjørn