| Message ID | 20220309143835.253911-1-dwmw2@infradead.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | KVM: Xen PV timer fixups | expand |
On Wed, 2022-03-09 at 14:38 +0000, David Woodhouse wrote: > I'll follow up separately with a patch to remove __kvm_migrate_apic_timer > because it seems to be utterly pointless now. No I won't, because that and the PIT timer can both be *periodic*, so do potentially still want migrating for the original reasons described in commit a3d7f85f471. But the Xen one really is just oneshot, so will expire soon and be restarted anyway, and will land on the right physical CPU then.
For squashing into the relevant commits already in kvm/queue, as discussed. Note that I've switched the immediate timer delivery to just call the xen_timer_callback() function, because I couldn't see *why* just setting the ->arch.xen.timer_pending flag actually worked, without also setting KVM_REQ_UNBLOCK. I suspect that the timer was only actually getting set after the vCPU exited for some *other* reason. This should be better. I'll follow up separately with a patch to remove __kvm_migrate_apic_timer because it seems to be utterly pointless now. David Woodhouse (2): KVM: x86/xen: PV oneshot timer fixes KVM: x86/xen: Update self test for Xen PV timers arch/x86/kvm/irq.c | 1 - arch/x86/kvm/xen.c | 37 +++++++++------------- arch/x86/kvm/xen.h | 1 - .../testing/selftests/kvm/x86_64/xen_shinfo_test.c | 35 ++++++++++++++++++-- 4 files changed, 47 insertions(+), 27 deletions(-)