Re: [RFC PATCH V4 13/17] x86/sev: Add Check of #HV event in path

[Pankaj Gupta <pankaj.gupta.linux@xxxxxxxxx>]

> >> +void check_hv_pending_irq_enable(void)
> >> +{
> >> +       struct pt_regs regs;
> >> +
> >> +       if (!cc_platform_has(CC_ATTR_GUEST_SEV_SNP))
> >> +               return;
> >> +
> >> +       memset(&regs, 0, sizeof(struct pt_regs));
> >> +       asm volatile("movl %%cs, %%eax;" : "=a" (regs.cs));
> >> +       asm volatile("movl %%ss, %%eax;" : "=a" (regs.ss));
> >> +       regs.orig_ax = 0xffffffff;
> >> +       regs.flags = native_save_fl();
> >> +
> >> +       /*
> >> +        * Disable irq when handle pending #HV events after
> >> +        * re-enabling irq.
> >> +        */
> >> +       asm volatile("cli" : : : "memory");
> > Just curious, Does the hypervisor injects irqs via doorbell page when
> > interrupts are disabled with "cli" ? Trying to understand the need to
> > cli/sti covering on "do_exc_hv".
>
>
> Hi Pankaj:
>         Thanks for your review. Yes, Hypervisor still injects #HV exception
> when irq was disabled check_hv_pending() is called when
> there is a #HV exception. It checks irq flag and return back without
> handling irq event when irq was disabled.

o.k. Thanks for your reply! I am clear with this part.

But want to know if there is possibility when "do_exc_hv" would keep
handling irqs in the continuous while loop i.e from the update in the
hv doorbell page and that can result in DOS like scenario? Is there is already
a protection for this?

Thanks,
Pankaj