On Mon, Mar 13, 2017 at 12:38 PM, Sergiy Kibrik <
[email protected]> wrote:
> On 03/06/2017 05:24 PM, Sergiy Kibrik wrote:
>
> + return _interface->in_cons != _interface->in_prod;
>> }
>>
>> char XEN_Console::readch() {
>> - return '\0'; /*TODO: support input */
>> + XENCONS_RING_IDX cons;
>> + char c;
>> + assert(_interface);
>> + cons = _interface->in_cons;
>> + c = _interface->in[MASK_XENCONS_IDX(cons, _interface->in)];
>>
>
> I think you should gracefully handle the case where you had a false wakeup
> and called readch() when nothing is available, and return 0 in this case
> (as you can see, LineDiscipline::read_poll handles 0 being returned). I
> guess you need to check if cons is actually beyond prod?
>
>
>
> I will do that.
>
>
> Hi Nadav,
> I've looked through LineDiscipline::read_poll() once more and can't figure
> out the possible case of false wake up — it looks like
> driver->input_ready() predicate should always be checked when thread is
> woken, shouldn't it?
> Can you please describe how false readch() call can sneak in?
>
I think you're right, sorry - indeed read_poll() indeed first wait for
input_ready(), so we shouldn't have any false wakeups if input_ready()
doesn't have false positives.
The check for c == 0 was added in commit
4b18685daa5544e0766195fec9eddd454aac149b. Apparently the readkey()
implementation in kbd.cc (the x86 keyboard driver - not the more usual
serial port) sometimes is forced to return 0 after thinking we had an
event. So read_poll() had to ignore this case. But this is not relevant to
your case.
Nadav.
--
You received this message because you are subscribed to the Google Groups "OSv
Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
