Re: [syzbot] KMSAN: uninit-value in ax88179_led_setting

[David Kahurani <k.kahurani@gmail.com>]

On Sat, Mar 26, 2022 at 3:43 PM Pavel Skripkin <paskripkin@gmail.com> wrote:
>
> Hi David,
>
> On 3/26/22 14:47, David Kahurani wrote:
> >>
> >> Signed-off-by: David Kahurani <k.kahurani@gmail.com>
> >> Reported-by: syzbot+d3dbdf31fbe9d8f5f311@syzkaller.appspotmail.com
> >> ---
> >>  drivers/net/usb/ax88179_178a.c | 181 +++++++++++++++++++++++++++------
> >>  1 file changed, 152 insertions(+), 29 deletions(-)
> >>
> >> diff --git a/drivers/net/usb/ax88179_178a.c b/drivers/net/usb/ax88179_178a.c
> >> index a31098981..932e21a65 100644
> >> --- a/drivers/net/usb/ax88179_178a.c
> >> +++ b/drivers/net/usb/ax88179_178a.c
> >> @@ -224,9 +224,12 @@ static int __ax88179_write_cmd(struct usbnet *dev, u8 cmd, u16 value, u16 index,
> >>   ret = fn(dev, cmd, USB_DIR_OUT | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
> >>   value, index, data, size);
> >>
>

Yes, this breaks the logic of the entire patch. It was a mistake. Will
fix that alongside your other comments.

> You've changed __ax88179_write_cmd(), but not __ax88179_read_cmd(). I've
> missed it. Changing  __ax88179_write_cmd() does not help with uninit
> value bugs
>
> Also I believe, __ax88179_read_cmd() should have __must_check annotation
> too, since problem came from it in the first place (I mean after added
> sane error handling inside it)
>
> Next thing is ax88179_read_cmd_nopm() still prone to uninit value bugs,
> since it touches uninitialized `buf` in case of __ax88179_read_cmd()
> error...
>
>
>
> I remembered why I gave up on fixing this driver... I hope, you have
> more free time and motivation :)
>
>
>
>
> With regards,
> Pavel Skripkin