Andrew Jones <andrew.jones@xxxxxxxxx> writes:
On Thu, Aug 24, 2023 at 09:29:33AM +0200, Andrew Jones wrote:
No need to speculate, QEMU is open source :-) QEMU is setting on offset,
but not because it explicitly wants to. Simply reading the CNT register
and writing back the same value is enough to set an offset, since the
timer will have certainly moved past whatever value was read by the time
it's written. QEMU frequently saves and restores all registers in the
get-reg-list array, unless they've been explicitly filtered out (with
Linux commit 680232a94c12, KVM_REG_ARM_PTIMER_CNT is now in the array).
So, to restore trapless ptimer accesses, we need a QEMU patch like below
to filter out the register.
diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c
index 94bbd9661fd3..f89ea31f170d 100644
--- a/target/arm/kvm64.c
+++ b/target/arm/kvm64.c
@@ -674,6 +674,7 @@ typedef struct CPRegStateLevel {
*/
static const CPRegStateLevel non_runtime_cpregs[] = {
{ KVM_REG_ARM_TIMER_CNT, KVM_PUT_FULL_STATE },
+ { KVM_REG_ARM_PTIMER_CNT, KVM_PUT_FULL_STATE },
};
int kvm_arm_cpreg_level(uint64_t regidx)
Actually, the QEMU fix above is necessary for more than just trap
avoidance. The ptimer will lag more and more with respect to other time
sources, even with respect to the vtimer (QEMU only updates the vtimer
offset when the VM is "paused".)
I'm not sure when I'll have time to test and post the QEMU patch. It may
be better if somebody else picks it up.
I haven't posted any QEMU patches before, but I think I can handle this.
