Re: [Qemu-devel] Segfault using qemu-system-arm in smc91c111

[Peter Maydell]

On 6 September 2015 at 19:37, Peter Crosthwaite
<address@hidden> wrote:
> This could still cause the same issue here though I think. The guest
> can release first (case 5) without a corresponding fifo pop (case 3),
> which actually means that the first patch idea might be the winner as
> it catches this issue too. What is supposed to happen if the guest
> does a case 5 while the rx_fifo is full without a matching case 3, and
> then a packet arrives on the wire?

The data sheet seems to be unclear. Definitely incoming data when
there's no packet available to be allocated means we drop the
incoming packet and signal an rx overrun. The data sheet doesn't
say anything about what happens when the rx fifo overflows, though...

My best guess is that fifo overflow also results in an overrun report,
but you'd probably have to actually write a test case for the real
hardware to find out for sure.

thanks
-- PMM