Re: hw/i386/q35: Where go LPC irqs?

[Julia Suvorova]

On Sat, Nov 7, 2020 at 3:48 PM Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>
> Hi, I am confuse with the LPC/GSI code.
>
> In pc_q35_init() we connect the LPC outputs to
> GSI input:
>
> 116 static void pc_q35_init(MachineState *machine)
> 117 {
> ...
> 240     /* irq lines */
> 241     gsi_state = pc_gsi_create(&x86ms->gsi, pcmc->pci_enabled);
> 242
> 243     ich9_lpc = ICH9_LPC_DEVICE(lpc);
> 244     lpc_dev = DEVICE(lpc);
> 245     for (i = 0; i < GSI_NUM_PINS; i++) {
> 246         qdev_connect_gpio_out_named(lpc_dev, ICH9_GPIO_GSI, i,
> x86ms->gsi[i]);
> 247     }
> ...
> 268     /* init basic PC hardware */
> 269     pc_basic_device_init(pcms, isa_bus, x86ms->gsi, &rtc_state,
> !mc->no_floppy,
> 270                          0xff0104);
>
> But then we call pc_basic_device_init() which overwrite
> the GSI inputs with HPET outputs:
>
> 1118 void pc_basic_device_init(struct PCMachineState *pcms,
> 1119                           ISABus *isa_bus, qemu_irq *gsi,
> 1120                           ISADevice **rtc_state,
> 1121                           bool create_fdctrl,
> 1122                           uint32_t hpet_irqs)
> 1123 {
> ...
> 1139     /*
> 1140      * Check if an HPET shall be created.
> 1141      *
> 1142      * Without KVM_CAP_PIT_STATE2, we cannot switch off the
> in-kernel PIT
> 1143      * when the HPET wants to take over. Thus we have to disable
> the latter.
> 1144      */
> 1145     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
> 1146                                kvm_has_pit_state2())) {
> ...
> 1165         for (i = 0; i < GSI_NUM_PINS; i++) {
> 1166             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
> 1167         }
>
> Are LPC IRQ still delivered?

>From what I got, LPC IRQs and HPET IRQs do not usually overlap (IRQ
0,2,8 for HPET, IRQ 3-7,9+ for all ICH9), which means that connecting
them together will not create an issue. I don't know what will happen
if higher IRQ is chosen for HPET, but according to ICH9 spec, it seems
to be aware of HPET interrupts.

> Peter commented here:
> https://www.mail-archive.com/qemu-devel@nongnu.org/msg758178.html
>
> "Connecting two qemu_irqs outputs directly
> to the same input is not valid as it produces subtly wrong behaviour
> (for instance if both the IRQ lines are high, and then one goes
> low, the PIC input will see this as a high-to-low transition
> even though the second IRQ line should still be holding it high)."
>
> Are this IRQ OR'ed to the GSI?
>
> Thanks,
>
> Phil.
>
>