On Fri, Apr 15, 2022 at 05:47:05PM +0800, Wang Cheng wrote:
> > > diff --git a/drivers/staging/rtl8712/usb_ops_linux.c b/drivers/staging/rtl8712/usb_ops_linux.c
> > > index f984a5ab2c6f..e321ca4453ca 100644
> > > --- a/drivers/staging/rtl8712/usb_ops_linux.c
> > > +++ b/drivers/staging/rtl8712/usb_ops_linux.c
> > > @@ -495,12 +495,14 @@ int r8712_usbctrl_vendorreq(struct intf_priv *pintfpriv, u8 request, u16 value,
> > > }
> > > status = usb_control_msg(udev, pipe, request, reqtype, value, index,
> > > pIo_buf, len, 500);
> > > - if (status > 0) { /* Success this control transfer. */
> > > - if (requesttype == 0x01) {
> > > - /* For Control read transfer, we have to copy the read
> > > - * data from pIo_buf to pdata.
> > > - */
> > > - memcpy(pdata, pIo_buf, status);
> > > + /* For Control read transfer, copy the read data from pIo_buf to pdata
> > > + * when control transfer success; otherwise init *pdata with 0.
> > > + */
> > > + if (requesttype == 0x01) {
> > > + if (status > 0)
> > > + memcpy(pdata, pIo_buf, status);
> > > + else
> > > + *(u32 *)pdata = 0;
> > > }
> >
> > This isn't really correct. In many cases status is "len" is less than 4.
> > I'm slightly surprised that nothing complains about that as an
> > uninitialized access. But then another problem is that "status" can be
> > less than "len".
>
> Sorry, I should explain it clearly. If I did right, watching "status"
> with gdb while running syzkaller reproducer, "status" returns from
> usb_control_msg() is -71. In which case, *pdata won't be touched in
> r8712_usbctrl_vendorreq(). As a result, "data" in
> usb_read8()/usb_read16()/usb_read32() will be returned without
> initialization. I think that is why kmsan reports:
> Local variable data created at:
> usb_read8+0x5d/0x130 drivers/staging/rtl8712/usb_ops.c:33
> r8712_read8+0xa5/0xd0 drivers/staging/rtl8712/rtl8712_io.c:29
>
> >
> > A better fix instead of setting pdata to zero would be to add error
> > checking in the callers and then change this code to use
> > usb_control_msg_send/recv(). Probably just initialize "data" in the
> > callers as well.
>
> I tried something similar which also works fine, but I think this patch
> does't fix it at root.
> https://syzkaller.appspot.com/text?tag=Patch&x=15be2970f00000
Oh... I meant initialize data on the success path as well.
__le32 data = 0;
You're saying that kasan generates a warning for your updated patch?
Earlier I mentioned "I'm slightly surprised that nothing complains about
that as an uninitialized access." so I guess maybe it does complain?
Good.
regards,
dan carpenter
