Re: [PATCH net-next] net: usb: aqc111: fix error handling of usbnet read calls

[Nikita Zhandarovich <n.zhandarovich@fintech.ru>]

Hello,

On 5/21/25 15:34, Andrew Lunn wrote:
> On Tue, May 20, 2025 at 02:32:39PM +0300, Nikita Zhandarovich wrote:
>> Syzkaller, courtesy of syzbot, identified an error (see report [1]) in
>> aqc111 driver, caused by incomplete sanitation of usb read calls'
>> results. This problem is quite similar to the one fixed in commit
>> 920a9fa27e78 ("net: asix: add proper error handling of usb read errors").
>>
>> For instance, usbnet_read_cmd() may read fewer than 'size' bytes,
>> even if the caller expected the full amount, and aqc111_read_cmd()
>> will not check its result properly. As [1] shows, this may lead
>> to MAC address in aqc111_bind() being only partly initialized,
>> triggering KMSAN warnings.
> 
> It looks like __ax88179_read_cmd() has the same issue? Please could
> you have a look around and see if more of the same problem exists.
> 

Yes, you are correct, similar issue with ax88179. There was once an
attempt to enable similar sanity checks there, see
https://lore.kernel.org/all/20220514133234.33796-1-k.kahurani@gmail.com/,
but for some reason it did not pan out.

As for other places with the same issue, after a quick glance I see these:

__ax88179_read_cmd - drivers/net/usb/ax88179_178a.c
cdc_ncm_init - drivers/net/usb/cdc_ncm.c
nc_vendor_read - drivers/net/usb/net1080.c
pla_read_word - drivers/net/usb/r8153_ecm.c
pla_write_word - drivers/net/usb/r8153_ecm.c
sierra_net_get_fw_attr - drivers/net/usb/sierra_net.c

This covers all instances usbnet_read_cmd[_nopm] that do not check for
full 'size' reads, only for straightforward errors. Roughly half of all
usbnet_read_cmd() calls kernel-wide.

> Are there any use cases where usbnet_read_cmd() can actually return
> less than size and it not being an error? Maybe this check for ret !=
> size can be moved inside usbnet_read_cmd()?

I can't reliably state how normal it is when usbnet_read_cmd() ends up
reading less than size. Both aqc111 and syzkaller with its repros (and
ax88179/asix as well) tell us it does happen when it shouldn't.

Personally, while I see logic of moving this fix into
usbnet_read_cmd(), I am wary for the very reason you stated in your
question - sometimes it may be expected. Also, more often than not
functions that envelop usbnet_read_cmd() (like ax88179_read_cmd()) seem
to be deliberately ignoring return values, but even an early warning
may be helpful in such cases.

In other words, I'd rather leave the decision up to the maintainers. I
don't mind doing either option.

Regards,
Nikita

> 
> 	Andrew