Improve Decrementor Implementation
Subsystems (3)
cover Improve Decrementor Implementation
We currently have an ugly hack called AGGRESSIVE_DEC that makes the Decrementor either work well for PPC32 or PPC64 targets.
This patchset removes that hack, bringing the decrementor implementation closer to real hardware.
Alexander Graf (3):
Move vector to irqprio resolving to separate function Improve DEC handling
Remove AGGRESSIVE_DEC
arch/powerpc/include/asm/kvm_ppc.h | 1 + arch/powerpc/kvm/book3s.c | 45 ++++++++++++++++++++--------------- arch/powerpc/kvm/booke.c | 5 ++++ arch/powerpc/kvm/emulate.c | 1 + 4 files changed, 33 insertions(+), 19 deletions(-)
PATCH 1/3 Move vector to irqprio resolving to separate function
We're using a switch table to find the irqprio that belongs to a specific interrupt vector. This table is part of the interrupt inject logic.
Since we'll add a new function to stop interrupts, let's move this table out of the injection logic into a separate function.
--- arch/powerpc/kvm/book3s.c | 13 ++++++++++--- 1 files changed, 10 insertions(+), 3 deletions(-)
arch/powerpc/kvm/book3s.c +10 -3
@@ -125,11 +125,10 @@ void kvmppc_inject_interrupt(struct kvm_vcpu *vcpu, int vec, u64 flags) vcpu->arch.mmu.reset_msr (vcpu); } -void kvmppc_book3s_queue_irqprio (struct kvm_vcpu *vcpu,unsigned int vec) +static int kvmppc_book3s_vec2irqprio (unsigned int vec) {unsigned int prio; - vcpu->stat.queue_intr++;switch (vec) {case 0x100 : prio =BOOK3S_IRQPRIO_SYSTEM_RESET ;break ;case 0x200 : prio =BOOK3S_IRQPRIO_MACHINE_CHECK ;break ;
@@ -149,7 +148,15 @@ void kvmppc_book3s_queue_irqprio(struct kvm_vcpu *vcpu, unsigned int vec)default : prio =BOOK3S_IRQPRIO_MAX ;break ; } -set_bit (prio, &vcpu->arch.pending_exceptions); +return prio; +} + +void kvmppc_book3s_queue_irqprio (struct kvm_vcpu *vcpu,unsigned int vec) +{ + vcpu->stat.queue_intr++; + +set_bit (kvmppc_book3s_vec2irqprio (vec), + &vcpu->arch.pending_exceptions);#ifdef EXIT_DEBUG printk (KERN_INFO "Queueing interrupt %x\n" , vec);#endif
Signature · 1 line
1.6.0.2
PATCH 2/3 Improve DEC handling
In reply to Alexander GrafWe treated the DEC interrupt like an edge based one. This is not true for Book3s. The DEC keeps firing until mtdec is issued again and thus clears the interrupt line.
So let's implement this logic in KVM too. This patch moves the line clearing from the firing of the interrupt to the mtdec emulation.
This makes PPC64 guests work without AGGRESSIVE_DEC defined.
--- arch/powerpc/include/asm/kvm_ppc.h | 1 + arch/powerpc/kvm/book3s.c | 16 +++++++++++++++- arch/powerpc/kvm/booke.c | 5 +++++ arch/powerpc/kvm/emulate.c | 1 + 4 files changed, 22 insertions(+), 1 deletions(-)
arch/powerpc/include/asm/kvm_ppc.h +1 -0
@@ -80,6 +80,7 @@ extern void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu);extern void kvmppc_core_deliver_interrupts (struct kvm_vcpu *vcpu);extern int kvmppc_core_pending_dec (struct kvm_vcpu *vcpu); +extern void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu);extern void kvmppc_core_queue_program (struct kvm_vcpu *vcpu);extern void kvmppc_core_queue_dec (struct kvm_vcpu *vcpu); extern void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
arch/powerpc/kvm/book3s.c +15 -1
@@ -151,6 +151,13 @@ static int kvmppc_book3s_vec2irqprio(unsigned int vec)return prio; } +static void kvmppc_book3s_stop_interrupt (struct kvm_vcpu *vcpu, +unsigned int vec) +{ +clear_bit (kvmppc_book3s_vec2irqprio (vec), + &vcpu->arch.pending_exceptions); +} +void kvmppc_book3s_queue_irqprio (struct kvm_vcpu *vcpu,unsigned int vec) { vcpu->stat.queue_intr++;
@@ -178,6 +185,11 @@ int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)return test_bit (BOOK3S_INTERRUPT_DECREMENTER >>7 , &vcpu->arch.pending_exceptions); } +void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu) +{ +kvmppc_book3s_stop_interrupt (vcpu,BOOK3S_INTERRUPT_DECREMENTER ); +} +void kvmppc_core_queue_external (struct kvm_vcpu *vcpu,struct kvm_interrupt *irq) {
@@ -275,7 +287,9 @@ void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)#endif priority =__ffs (*pending);while (priority <= (sizeof (unsigned int ) *8 )) { -if (kvmppc_book3s_irqprio_deliver (vcpu, priority)) { +if (kvmppc_book3s_irqprio_deliver (vcpu, priority) && + (priority !=BOOK3S_IRQPRIO_DECREMENTER )) { +/* DEC interrupts get cleared by mtdec */ clear_bit (priority, &vcpu->arch.pending_exceptions);break ; }
arch/powerpc/kvm/booke.c +5 -0
@@ -97,6 +97,11 @@ int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)return test_bit (BOOKE_IRQPRIO_DECREMENTER , &vcpu->arch.pending_exceptions); } +void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu) +{ +clear_bit (BOOKE_IRQPRIO_DECREMENTER , &vcpu->arch.pending_exceptions); +} +void kvmppc_core_queue_external (struct kvm_vcpu *vcpu,struct kvm_interrupt *irq) {
arch/powerpc/kvm/emulate.c +1 -0
@@ -82,6 +82,7 @@ void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)unsigned long dec_nsec;pr_debug ("mtDEC: %x\n" , vcpu->arch.dec); +kvmppc_core_stop_dec (vcpu);#ifdef CONFIG_PPC64 /* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec &0x80000000 ) {
Signature · 1 line
1.6.0.2
PATCH 3/3 Remove AGGRESSIVE_DEC
In reply to Alexander GrafBecause we now emulate the DEC interrupt according to real life behavior, there's no need to keep the AGGRESSIVE_DEC hack around.
Let's just remove it.
--- arch/powerpc/kvm/book3s.c | 16 +--------------- 1 files changed, 1 insertions(+), 15 deletions(-)
arch/powerpc/kvm/book3s.c +1 -15
@@ -34,12 +34,6 @@/* #define EXIT_DEBUG */ /* #define EXIT_DEBUG_SIMPLE */ -/* Without AGGRESSIVE_DEC we only fire off a DEC interrupt when DEC turns 0. -* When set, we retrigger a DEC interrupt after that if DEC <= 0. -* PPC32 Linux runs faster without AGGRESSIVE_DEC, PPC64 Linux requires it. */ - -/* #define AGGRESSIVE_DEC */ -struct kvm_stats_debugfs_item debugfs_entries[] = { {"exits" ,VCPU_STAT (sum_exits) }, {"mmio" ,VCPU_STAT (mmio_exits) },
@@ -81,7 +75,7 @@ void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu)to_book3s (vcpu)->slb_shadow_max =get_paca ()->kvm_slb_max; } -#if defined(AGGRESSIVE_DEC ) || defined(EXIT_DEBUG ) +#if defined(EXIT_DEBUG )static u32 kvmppc_get_dec (struct kvm_vcpu *vcpu) {u64 jd =mftb () - vcpu->arch.dec_jiffies;
@@ -273,14 +267,6 @@ void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)unsigned long *pending = &vcpu->arch.pending_exceptions;unsigned int priority; -/* XXX be more clever here - no need to mftb() on every entry */ -/* Issue DEC again if it's still active */ -#ifdef AGGRESSIVE_DEC -if (vcpu->arch.msr &MSR_EE ) -if (kvmppc_get_dec (vcpu) &0x80000000 ) -kvmppc_core_queue_dec (vcpu); -#endif -#ifdef EXIT_DEBUG if (vcpu->arch.pending_exceptions)printk (KERN_EMERG "KVM: Check pending: %lx\n" , vcpu->arch.pending_exceptions);
Signature · 1 line
1.6.0.2
PATCH 2/3 Improve DEC handling
We treated the DEC interrupt like an edge based one. This is not true for Book3s. The DEC keeps firing until mtdec is issued again and thus clears the interrupt line.
So let's implement this logic in KVM too. This patch moves the line clearing from the firing of the interrupt to the mtdec emulation.
This makes PPC64 guests work without AGGRESSIVE_DEC defined.
--- arch/powerpc/include/asm/kvm_ppc.h | 1 + arch/powerpc/kvm/book3s.c | 16 +++++++++++++++- arch/powerpc/kvm/booke.c | 5 +++++ arch/powerpc/kvm/emulate.c | 1 + 4 files changed, 22 insertions(+), 1 deletions(-)
arch/powerpc/include/asm/kvm_ppc.h +1 -0
@@ -80,6 +80,7 @@ extern void kvmppc_core_vcpu_put(struct kvm_vcpu *vcpu);extern void kvmppc_core_deliver_interrupts (struct kvm_vcpu *vcpu);extern int kvmppc_core_pending_dec (struct kvm_vcpu *vcpu); +extern void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu);extern void kvmppc_core_queue_program (struct kvm_vcpu *vcpu);extern void kvmppc_core_queue_dec (struct kvm_vcpu *vcpu); extern void kvmppc_core_queue_external(struct kvm_vcpu *vcpu,
arch/powerpc/kvm/book3s.c +15 -1
@@ -151,6 +151,13 @@ static int kvmppc_book3s_vec2irqprio(unsigned int vec)return prio; } +static void kvmppc_book3s_stop_interrupt (struct kvm_vcpu *vcpu, +unsigned int vec) +{ +clear_bit (kvmppc_book3s_vec2irqprio (vec), + &vcpu->arch.pending_exceptions); +} +void kvmppc_book3s_queue_irqprio (struct kvm_vcpu *vcpu,unsigned int vec) { vcpu->stat.queue_intr++;
@@ -178,6 +185,11 @@ int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)return test_bit (BOOK3S_INTERRUPT_DECREMENTER >>7 , &vcpu->arch.pending_exceptions); } +void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu) +{ +kvmppc_book3s_stop_interrupt (vcpu,BOOK3S_INTERRUPT_DECREMENTER ); +} +void kvmppc_core_queue_external (struct kvm_vcpu *vcpu,struct kvm_interrupt *irq) {
@@ -275,7 +287,9 @@ void kvmppc_core_deliver_interrupts(struct kvm_vcpu *vcpu)#endif priority =__ffs (*pending);while (priority <= (sizeof (unsigned int ) *8 )) { -if (kvmppc_book3s_irqprio_deliver (vcpu, priority)) { +if (kvmppc_book3s_irqprio_deliver (vcpu, priority) && + (priority !=BOOK3S_IRQPRIO_DECREMENTER )) { +/* DEC interrupts get cleared by mtdec */ clear_bit (priority, &vcpu->arch.pending_exceptions);break ; }
arch/powerpc/kvm/booke.c +5 -0
@@ -97,6 +97,11 @@ int kvmppc_core_pending_dec(struct kvm_vcpu *vcpu)return test_bit (BOOKE_IRQPRIO_DECREMENTER , &vcpu->arch.pending_exceptions); } +void kvmppc_core_stop_dec (struct kvm_vcpu *vcpu) +{ +clear_bit (BOOKE_IRQPRIO_DECREMENTER , &vcpu->arch.pending_exceptions); +} +void kvmppc_core_queue_external (struct kvm_vcpu *vcpu,struct kvm_interrupt *irq) {
arch/powerpc/kvm/emulate.c +1 -0
@@ -82,6 +82,7 @@ void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)unsigned long dec_nsec;pr_debug ("mtDEC: %x\n" , vcpu->arch.dec); +kvmppc_core_stop_dec (vcpu);#ifdef CONFIG_PPC64 /* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec &0x80000000 ) {
Signature · 1 line
1.6.0.2
A message that is not in the archive (sent off-list or not yet received): 1261405373-8008-3-git-send-email-agraf-l3A5Bk7waGM@public.gmane.org
Re: [PATCH 2/3] Improve DEC handling
Quoted 3 lines · Alexander Graf
We treated the DEC interrupt like an edge based one. This is not true for Book3s. The DEC keeps firing until mtdec is issued again and thus clears the interrupt line.
That's not quite right. The decrementer keeps firing until the top bit is cleared, i.e. with mtdec. However, not *every* mtdec clears it. (Also, I'm pretty sure this varies between Book 3S implementations, e.g. 970 behaves differently than POWERn. I don't remember specific values of <n> though, and I could be misremembering...)
So is this the failure mode?
- a decrementer interrupt is delivered
- guest does *not* issue mtdec to clear it (ppc64's lazy interrupt disabling?)
- guest expects a second decrementer interrupt, but KVM doesn't deliver one
In that case, it seems like the real fix would be something like this:
void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
{
unsigned long dec_nsec;
pr_debug("mtDEC: %x\n", vcpu->arch.dec);
#ifdef CONFIG_PPC64
/* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec & 0x80000000) {
hrtimer_try_to_cancel(&vcpu->arch.dec_timer); kvmppc_core_queue_dec(vcpu);
+ /* keep queuing interrupts until guest clears high MSR bit */
+ hrtimer_start(&vcpu->arch.dec_timer, ktime_set(0, 100),
+ HRTIMER_MODE_REL);
return;
}
#endif
-Hollis
A message that is not in the archive (sent off-list or not yet received): fb412d760912211013w265d3d30i99498ab136a15e00-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org
Re: [PATCH 2/3] Improve DEC handling
Quoted 4 lines · Hollis Blanchard
void
kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
{
unsigned longdec_nsec;
Quoted 12 lines
pr_debug("mtDEC: %x\n",
vcpu->arch.dec);
#ifdefCONFIG_PPC64
/* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec&0x80000000) {hrtimer_try_to_cancel(&vcpu->arch.dec_timer);kvmppc_core_queue_dec(vcpu);
+ /* keep queuing interrupts until guest clears high MSR bit */
+ hrtimer_start(&vcpu->arch.dec_timer,ktime_set(0, 100),
+HRTIMER_MODE_REL);
return;
}
#endif
Of course, removing the hardcoded 100-ns timer would be better, and indeed we can do that. What we *really* want is to key off of MSR[EE] changes (there's no point in queuing anything until then). So why not move your "AGGRESSIVE_DEC" check into Book 3S's kvmppc_set_msr()?
-Hollis
A message that is not in the archive (sent off-list or not yet received): fb412d760912211017i3f47fce2oc3b67c9c5c645b6-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org
Re: [PATCH 2/3] Improve DEC handling
Quoted 21 lines · Hollis Blanchard
Quoted 11 lines · Hollis Blanchard
void
kvmppc_emulate_dec(struct kvm_vcpu *vcpu){ unsigned long dec_nsec; pr_debug("mtDEC: %x\n", vcpu->arch.dec); #ifdef CONFIG_PPC64 /* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec & 0x80000000) { hrtimer_try_to_cancel(&vcpu->arch.dec_timer); kvmppc_core_queue_dec(vcpu);Review
Quoted 6 lines
(quoted diff) +3
-0+ /* keep queuing interrupts until guest clears high MSR bit */ + hrtimer_start(&vcpu->arch.dec_timer, ktime_set(0, 100), + HRTIMER_MODE_REL); return; } #endif
Quoted 3 lines
Of course, removing the hardcoded 100-ns timer would be better, and indeed we can do that. What we *really* want is to key off of MSR[EE] changes (there's no point in queuing anything until then).
Yes. mtmsr with EE=1 triggers a #VMEXIT which then goes off into a VM entry which on entering checks that the DEC interrupt is still active and fires it.
Quoted 3 lines
So why not
move your "AGGRESSIVE_DEC" check into Book 3S'skvmppc_set_msr()?
The idea was to get rid of AGGRESSIVE_DEC :-).
Alex
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 2/3] Improve DEC handling
For the record, we've discussed more by IRC, and I think revised patches will be forthcoming.
-Hollis
Re: [PATCH 2/3] Improve DEC handling
In reply to a message not in the archiveQuoted 10 lines · Hollis Blanchard
Quoted 4 lines · Alexander Graf
We treated the DEC interrupt like an edge based one. This is not true for Book3s. The DEC keeps firing until mtdec is issued again and thus clears the interrupt line.
That's not quite right. The decrementer keeps firing until the top bit is cleared, i.e. with mtdec. However, not *every* mtdec clears it.
Right, that's we we fire a dec interrupt off whenever we mtdec with the top bit set.
Quoted 4 lines
(Also, I'm pretty sure this varies between Book 3S implementations, e.g. 970 behaves differently than POWERn. I don't remember specific values of <n> though, and I could be misremembering...)
IIRC only the embedded cores were different. But I could be wrong. How do I find out?
Quoted 11 lines
So is this the failure mode?
- a decrementer interrupt is delivered
- guest does *not* issue mtdec to clear it (ppc64's lazy interrupt disabling?)
- guest expects a second decrementer interrupt, but KVM doesn't deliver oneIn that case, it seems like the real fix would be something like this:
void
kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
{arch/powerpc/kvm/emulate.c +0
-0unsigned long dec_nsec;
arch/powerpc/kvm/emulate.c void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
Quoted 10 lines
arch/powerpc/kvm/emulate.c +3
-0pr_debug ("mtDEC: %x\n" , vcpu->arch.dec);#ifdef CONFIG_PPC64 /* POWER4+ triggers a dec interrupt if the value is < 0 */ if (vcpu->arch.dec &0x80000000 ) {hrtimer_try_to_cancel (&vcpu->arch.dec_timer);kvmppc_core_queue_dec (vcpu); +/* keep queuing interrupts until guest clears high MSR bit */ +hrtimer_start (&vcpu->arch.dec_timer,ktime_set (0 ,100 ), +HRTIMER_MODE_REL );
This code path is only triggered when the guest mtdecs with a negative value.
But I understand what you're trying to suggest and I think it's a bad idea. We don't want to poll the guest for interrupt enablement.
On a real CPU the DEC interrupt keeps being active when the DEC register is negative. And that's exactly what this patch implements, no? That way we are automatically event-based and everyone's happy.
Alex