On Wed, Feb 12, 2025 at 4:54 AM Kirill A. Shutemov kirill@shutemov.name wrote:
On Wed, Feb 12, 2025 at 12:07:45AM +0000, Vishal Annapurve wrote:
Direct HLT instruction execution causes #VEs for TDX VMs which is routed to hypervisor via TDCALL. safe_halt() routines execute HLT in STI-shadow so IRQs need to remain disabled until the TDCALL to ensure that pending IRQs are correctly treated as wake events. So "sti;hlt" sequence needs to be replaced with "TDCALL; raw_local_irq_enable()" for TDX VMs.
The last sentence is somewhat confusing.
Maybe drop it and add explanation that #VE handler doesn't have info about STI shadow, enables interrupts before TDCALL which can lead to missed wakeup events.
Ack, will fix it in the next version.
@@ -409,6 +410,12 @@ void __cpuidle tdx_safe_halt(void) WARN_ONCE(1, "HLT instruction emulation failed\n"); }
+static void __cpuidle tdx_safe_halt(void) +{
tdx_halt();
raw_local_irq_enable();
What is justification for raw_? Why local_irq_enable() is not enough?
To very least, it has to be explained.
Let me replace it with a more suitable arch specific <>_irq_enable() function in the next version. Intention here is to just enable interrupts.
-- Kiryl Shutsemau / Kirill A. Shutemov