Re: [Qemu-devel] [RFC PATCH] ati-vga: Implement dummy VBlank IRQ

[BALATON Zoltan <balaton@eik.bme.hu>]

On Thu, 15 Aug 2019, Gerd Hoffmann wrote:
> On Thu, Aug 15, 2019 at 02:25:07AM +0200, BALATON Zoltan wrote:
>> The MacOS driver exits if the card does not have an interrupt. If we
>> set PCI_INTERRUPT_PIN to 1 then it enables VBlank interrupts and it
>> boots but the mouse poniter can not be moved. This patch implements a
>> dummy VBlank interrupt by a timer triggered at 60 Hz to test if it
>> helps. Unfortunately it doesn't: MacOS with this patch hangs during
>> boot just polling interrupts and acknowledging them so maybe it needs
>> something else or there may be some other problem with this
>> implementation.
>>
>> This is posted for comments and to let others experiment with it but
>> probably should not be committed upstream yet.
>>
>> Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
>> ---
>>  hw/display/ati.c      | 41 +++++++++++++++++++++++++++++++++++++++++
>>  hw/display/ati_dbg.c  |  1 +
>>  hw/display/ati_int.h  |  4 ++++
>>  hw/display/ati_regs.h |  1 +
>>  4 files changed, 47 insertions(+)
>>
>> diff --git a/hw/display/ati.c b/hw/display/ati.c
>> index a365e2455d..e06cbf3e91 100644
>> --- a/hw/display/ati.c
>> +++ b/hw/display/ati.c
>> @@ -243,6 +243,21 @@ static uint64_t ati_i2c(bitbang_i2c_interface *i2c, uint64_t data, int base)
>>      return data;
>>  }
>>
>> +static void ati_vga_update_irq(ATIVGAState *s)
>> +{
>> +    pci_set_irq(&s->dev, s->regs.gen_int_status & 1);
>
> This should be "s->regs.gen_int_status & s->regs.gen_int_cntl" I guess?

Probably, but we only try to emulate VBlank yet so to avoid any problems 
due to raising irq for unknown bits I restricted it for that now.

>> +static void ati_vga_vblank_irq(void *opaque)
>> +{
>> +    ATIVGAState *s = opaque;
>> +
>> +    timer_mod(&s->vblank_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
>> +              NANOSECONDS_PER_SECOND / 60);
>> +    s->regs.gen_int_status |= 1;
>
> #defines for the irq status bits would be nice.

Yes, I thought about that but this was only for quick testing. I'll add 
constant for this in next version.

>> +    case GEN_INT_CNTL:
>> +        s->regs.gen_int_cntl = data;
>> +        if (data & 1) {
>> +            ati_vga_vblank_irq(s);
>> +        } else {
>> +            timer_del(&s->vblank_timer);
>> +        }
>
> ati_vga_update_irq() needed here.
>
>> +        break;
>> +    case GEN_INT_STATUS:
>> +        data &= (s->dev_id == PCI_DEVICE_ID_ATI_RAGE128_PF ?
>> +                 0x000f040fUL : 0xfc080effUL);
>
> Add IRQ_MASK #define ?

I'd leave these as constants because there are many of them (basically 
reserved bit mask for regs where we care or in this case writable bits) 
and one has to check docs to verify them and also in some cases we combine 
rage128p and rv100 so hiding them behind a constant would just make code 
less readable in my opinion. (This would become 3 lines for example with 
defines you'd have to look up in a different header so it's easier to see 
this way.)

>> +        s->regs.gen_int_status &= ~data;
>
> ati_vga_update_irq() needed here too.

Thanks. Indeed I forgot this. With that it works a bit better, mouse now 
can be moved but only vertically... No idea why, I'll have to check,

Regards,
BALATON Zoltan