Re: [PATCH v2] usb: dwc3: gadget: check that event count does not exceed event buffer length

[Thinh Nguyen <Thinh.Nguyen@synopsys.com>]

On Wed, Apr 02, 2025, Frode Isaksen wrote:
> On 4/2/25 6:42 AM, Krishna Kurapati wrote:
> > > I still wonder what's current behavior of the HW to properly respond
> > > here. If the device is dead, register read often returns all Fs, which
> > > may be the case you're seeing here. If so, we should properly prevent
> > > the driver from accessing the device and properly teardown the driver.
> > > 
> > > If this is a momentary bleep/lost of power in the device, perhaps your
> > > change here is sufficient and the driver can continue to access the
> > > device.
> > > 
> > > With the difficulty of reproducing this issue, can you confirm that the
> > > device still operates properly after this change?
> > 
> > Unfortunately, I did not test this particular change of returning when
> > ev count is invalid. I stress tested the change of copying only 4K [1],
> > but didn't see any issue. I suspect we didn't hit the issue later again
> > in the course of 14 day testing.
> > 
> > [1]: https://urldefense.com/v3/__https://lore.kernel.org/all/20230521100330.22478-1-quic_kriskura@quicinc.com/__;!!A4F2R9G_pg!alsTyn4jdAnTCH4s4h41hQJCncQjFukgGelwrLBxOt1ckeN8GRQk0DPMKSpjX1hSz2260P9VOfqpqHrNU_twCwSX$
> > 
> > 
> > Regards,
> > Krishna,
> > 
> I have been running my fix for over a year with millions of Quest 3 devices,
> and no strange bugs caused by this has been seen. Without the fix, there
> were 1 to 2 crashes per week.
> 
> So I think it's safe to just return with IRQ_NONE when the count exceeds the
> event length.
> 

Thanks for confirming.

Thinh