KVM: arm64: PMU: Fixing chained events, and PMUv3p5 support
Marc Zyngier
maintainercover KVM: arm64: PMU: Fixing chained events, and PMUv3p5 support
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Allow ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Allow PMUv3p5 to be exposed to the guest
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 1/9 KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
Ricardo recently pointed out that the PMU chained counter emulation in KVM wasn't quite behaving like the one on actual hardware, in the sense that a chained counter would expose an overflow on both halves of a chained counter, while KVM would only expose the overflow on the top half.
The difference is subtle, but significant. What does the architecture say (DDI0087 H.a):
- Before PMUv3p4, all counters but the cycle counter are 32bit
- A 32bit counter that overflows generates a CHAIN event on the adjacent counter after exposing its own overflow status
- The CHAIN event is accounted if the counter is correctly configured (CHAIN event selected and counter enabled)
This all means that our current implementation (which uses 64bit perf events) prevents us from emulating this overflow on the lower half.
How to fix this? By implementing the above, to the letter.
This largly results in code deletion, removing the notions of "counter pair", "chained counters", and "canonical counter". The code is further restructured to make the CHAIN handling similar to SWINC, as the two are now extremely similar in behaviour.
--- arch/arm64/kvm/pmu-emul.c | 324 +++++++++++--------------------------- include/kvm/arm_pmu.h | 2 - 2 files changed, 91 insertions(+), 235 deletions(-)
arch/arm64/kvm/pmu-emul.c +91 -233
@@ -21,10 +21,6 @@ static LIST_HEAD(arm_pmus);static DEFINE_MUTEX (arm_pmus_lock);static void kvm_pmu_create_perf_event (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_update_pmc_chained (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_stop_counter (struct kvm_vcpu *vcpu,struct kvm_pmc *pmc); - -#define PERF_ATTR_CFG1_KVM_PMU_CHAINED 0x1static u32 kvm_pmu_event_mask (struct kvm *kvm) {
@@ -57,6 +53,11 @@ static bool kvm_pmu_idx_is_64bit(struct kvm_vcpu *vcpu, u64 select_idx) __vcpu_sys_reg(vcpu, PMCR_EL0) & ARMV8_PMU_PMCR_LC); } +static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx) +{ +return (!(idx &1 ) && (idx +1 ) <ARMV8_PMU_CYCLE_IDX ); +} +static struct kvm_vcpu *kvm_pmc_to_vcpu (struct kvm_pmc *pmc) {struct kvm_pmu *pmu;
@@ -69,91 +70,22 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) }/** -* kvm_pmu_pmc_is_chained - determine if the pmc is chained -* @pmc: The PMU counter pointer -*/ -static bool kvm_pmu_pmc_is_chained (struct kvm_pmc *pmc) -{ -struct kvm_vcpu *vcpu =kvm_pmc_to_vcpu (pmc); - -return test_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -} - -/** -* kvm_pmu_idx_is_high_counter - determine if select_idx is a high/low counter -* @select_idx: The counter index -*/ -static bool kvm_pmu_idx_is_high_counter (u64 select_idx) -{ -return select_idx &0x1 ; -} - -/** -* kvm_pmu_get_canonical_pmc - obtain the canonical pmc -* @pmc: The PMU counter pointer -* -* When a pair of PMCs are chained together we use the low counter (canonical) -* to hold the underlying perf event. -*/ -static struct kvm_pmc *kvm_pmu_get_canonical_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; - -return pmc; -} -static struct kvm_pmc *kvm_pmu_get_alternate_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; -else -return pmc +1 ; -} - -/** -* kvm_pmu_idx_has_chain_evtype - determine if the event type is chain +* kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer * @select_idx: The counter index */ -static bool kvm_pmu_idx_has_chain_evtype (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 eventsel, reg; - - select_idx |=0x1 ; - -if (select_idx ==ARMV8_PMU_CYCLE_IDX ) -return false; - - reg =PMEVTYPER0_EL0 + select_idx; - eventsel =__vcpu_sys_reg (vcpu, reg) &kvm_pmu_event_mask (vcpu->kvm); - -return eventsel ==ARMV8_PMUV3_PERFCTR_CHAIN ; -} - -/** -* kvm_pmu_get_pair_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @pmc: The PMU counter pointer -*/ -static u64 kvm_pmu_get_pair_counter_value (struct kvm_vcpu *vcpu, -struct kvm_pmc *pmc) +u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) { -u64 counter, counter_high, reg, enabled, running; - -if (kvm_pmu_pmc_is_chained (pmc)) { - pmc =kvm_pmu_get_canonical_pmc (pmc); - reg =PMEVCNTR0_EL0 + pmc->idx; +u64 counter, reg, enabled, running; +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - counter =__vcpu_sys_reg (vcpu, reg); - counter_high =__vcpu_sys_reg (vcpu, reg +1 ); +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; - counter =lower_32_bits (counter) | (counter_high <<32 ); - }else { - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; - counter =__vcpu_sys_reg (vcpu, reg); - } + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) + ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + counter =__vcpu_sys_reg (vcpu, reg);/* * The real counter value is equal to the value of counter register plus
@@ -163,29 +95,7 @@ static u64 kvm_pmu_get_pair_counter_value(struct kvm_vcpu *vcpu, counter +=perf_event_read_value (pmc->perf_event, &enabled, &running); -return counter; -} - -/** -* kvm_pmu_get_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @select_idx: The counter index -*/ -u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 counter; -struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - -if (!kvm_vcpu_has_pmu (vcpu)) -return 0 ; - - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); - -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (select_idx)) - counter =upper_32_bits (counter); -else if (select_idx !=ARMV8_PMU_CYCLE_IDX ) +if (select_idx !=ARMV8_PMU_CYCLE_IDX ) counter =lower_32_bits (counter);return counter;
@@ -218,7 +128,6 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)*/ static void kvm_pmu_release_perf_event (struct kvm_pmc *pmc) { - pmc =kvm_pmu_get_canonical_pmc (pmc);if (pmc->perf_event) {perf_event_disable (pmc->perf_event);perf_event_release_kernel (pmc->perf_event);
@@ -236,11 +145,10 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) {u64 counter, reg, val; - pmc =kvm_pmu_get_canonical_pmc (pmc);if (!pmc->perf_event)return ; - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); + counter =kvm_pmu_get_counter_value (vcpu, pmc->idx);if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) { reg =PMCCNTR_EL0 ;
@@ -252,9 +160,6 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc)__vcpu_sys_reg (vcpu, reg) = val; -if (kvm_pmu_pmc_is_chained (pmc)) -__vcpu_sys_reg (vcpu, reg +1 ) =upper_32_bits (counter); -kvm_pmu_release_perf_event (pmc); }
@@ -285,8 +190,6 @@ void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu)for_each_set_bit (i, &mask,32 )kvm_pmu_stop_counter (vcpu, &pmu->pmc[i]); - -bitmap_zero (vcpu->arch.pmu.chained,ARMV8_PMU_MAX_COUNTER_PAIRS ); }/**
@@ -340,11 +243,8 @@ void kvm_pmu_enable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event) {perf_event_enable (pmc->perf_event);if (pmc->perf_event->state !=PERF_EVENT_STATE_ACTIVE )
@@ -375,11 +275,8 @@ void kvm_pmu_disable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event)perf_event_disable (pmc->perf_event); }
@@ -484,6 +381,51 @@ static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work)kvm_vcpu_kick (vcpu); } +/* +* Perform an increment on any of the counters described in @mask, +* generating the overflow if required, and propagate it as a chained +* event if possible. +*/ +static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu)) +return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg; + +/* Filter on event type */ + type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); + type &=kvm_pmu_event_mask (vcpu->kvm); +if (type != event) +continue ; + +/* Increment this counter */ + reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; + reg =lower_32_bits (reg); +__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; + +if (reg)/* No overflow? move on */ +continue ; + +/* Mark overflow */ +__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i); + +if (kvm_pmu_counter_can_chain (vcpu, i)) +kvm_pmu_counter_increment (vcpu,BIT (i +1 ), +ARMV8_PMUV3_PERFCTR_CHAIN ); + } +} +/** * When the perf event overflows, set the overflow status and inform the vcpu. */
@@ -514,6 +456,10 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (idx); +if (kvm_pmu_counter_can_chain (vcpu, idx)) +kvm_pmu_counter_increment (vcpu,BIT (idx +1 ), +ARMV8_PMUV3_PERFCTR_CHAIN ); +if (kvm_pmu_overflow_status (vcpu)) {kvm_make_request (KVM_REQ_IRQ_PENDING , vcpu);
@@ -533,50 +479,7 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,*/ void kvm_pmu_software_increment (struct kvm_vcpu *vcpu,u64 val) { -struct kvm_pmu *pmu = &vcpu->arch.pmu; -int i; - -if (!kvm_vcpu_has_pmu (vcpu)) -return ; - -if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) -return ; - -/* Weed out disabled counters */ - val &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); - -for (i =0 ; i <ARMV8_PMU_CYCLE_IDX ; i++) { -u64 type, reg; - -if (!(val &BIT (i))) -continue ; - -/* PMSWINC only applies to ... SW_INC! */ - type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); - type &=kvm_pmu_event_mask (vcpu->kvm); -if (type !=ARMV8_PMUV3_PERFCTR_SW_INCR ) -continue ; - -/* increment this even SW_INC counter */ - reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; - reg =lower_32_bits (reg); -__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; - -if (reg)/* no overflow on the low part */ -continue ; - -if (kvm_pmu_pmc_is_chained (&pmu->pmc[i])) { -/* increment the high counter */ - reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i +1 ) +1 ; - reg =lower_32_bits (reg); -__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i +1 ) = reg; -if (!reg)/* mark overflow on the high counter */ -__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i +1 ); - }else { -/* mark overflow on low counter */ -__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i); - } - } +kvm_pmu_counter_increment (vcpu, val,ARMV8_PMUV3_PERFCTR_SW_INCR ); }/**
@@ -625,30 +528,27 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) {struct arm_pmu *arm_pmu = vcpu->kvm->arch.arm_pmu;struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];struct perf_event *event;struct perf_event_attr attr;u64 eventsel, counter, reg, data; -/* -* For chained counters the event type and filtering attributes are -* obtained from the low/even counter. We also use this counter to -* determine if the event is enabled/disabled. -*/ - pmc =kvm_pmu_get_canonical_pmc (&pmu->pmc[select_idx]); - - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx; data =__vcpu_sys_reg (vcpu, reg);kvm_pmu_stop_counter (vcpu, pmc); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) +if (select_idx ==ARMV8_PMU_CYCLE_IDX ) eventsel =ARMV8_PMUV3_PERFCTR_CPU_CYCLES ;else eventsel = data &kvm_pmu_event_mask (vcpu->kvm); -/* Software increment event doesn't need to be backed by a perf event */ -if (eventsel ==ARMV8_PMUV3_PERFCTR_SW_INCR ) +/* +* Neither SW increment nor chained events need to be backed +* by a perf event. +*/ +if (eventsel ==ARMV8_PMUV3_PERFCTR_SW_INCR || + eventsel ==ARMV8_PMUV3_PERFCTR_CHAIN )return ;/*
@@ -663,37 +563,27 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) attr.type = arm_pmu->pmu.type; attr.size =sizeof (attr); attr.pinned =1 ; - attr.disabled = !kvm_pmu_counter_is_enabled (vcpu, pmc->idx); + attr.disabled = !kvm_pmu_counter_is_enabled (vcpu, select_idx); attr.exclude_user = data &ARMV8_PMU_EXCLUDE_EL0 ?1 :0 ; attr.exclude_kernel = data &ARMV8_PMU_EXCLUDE_EL1 ?1 :0 ; attr.exclude_hv =1 ;/* Don't count EL2 events */ attr.exclude_host =1 ;/* Don't count host events */ attr.config = eventsel; - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); +/* If counting a 64bit event, advertise it to the perf code */ +if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) + attr.config1 |=1 ; -if (kvm_pmu_pmc_is_chained (pmc)) { -/** -* The initial sample period (overflow count) of an event. For -* chained counters we only support overflow interrupts on the -* high counter. -*/ - attr.sample_period = (-counter) &GENMASK (63 ,0 ); - attr.config1 |=PERF_ATTR_CFG1_KVM_PMU_CHAINED ; + counter =kvm_pmu_get_counter_value (vcpu, select_idx); - event =perf_event_create_kernel_counter (&attr,-1 , current, - kvm_pmu_perf_overflow, - pmc +1 ); - }else { -/* The initial sample period (overflow count) of an event. */ -if (kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) - attr.sample_period = (-counter) &GENMASK (63 ,0 ); -else - attr.sample_period = (-counter) &GENMASK (31 ,0 ); +/* The initial sample period (overflow count) of an event. */ +if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) + attr.sample_period = (-counter) &GENMASK (63 ,0 ); +else + attr.sample_period = (-counter) &GENMASK (31 ,0 ); - event =perf_event_create_kernel_counter (&attr,-1 , current, + event =perf_event_create_kernel_counter (&attr,-1 , current, kvm_pmu_perf_overflow, pmc); - }if (IS_ERR (event)) { pr_err_once("kvm: pmu event creation failed %ld\n",
@@ -704,41 +594,6 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) pmc->perf_event = event; } -/** -* kvm_pmu_update_pmc_chained - update chained bitmap -* @vcpu: The vcpu pointer -* @select_idx: The number of selected counter -* -* Update the chained bitmap based on the event type written in the -* typer register and the enable state of the odd register. -*/ -static void kvm_pmu_update_pmc_chained (struct kvm_vcpu *vcpu,u64 select_idx) -{ -struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc = &pmu->pmc[select_idx], *canonical_pmc; -bool new_state, old_state; - - old_state =kvm_pmu_pmc_is_chained (pmc); - new_state =kvm_pmu_idx_has_chain_evtype (vcpu, pmc->idx) && -kvm_pmu_counter_is_enabled (vcpu, pmc->idx |0x1 ); - -if (old_state == new_state) -return ; - - canonical_pmc =kvm_pmu_get_canonical_pmc (pmc); -kvm_pmu_stop_counter (vcpu, canonical_pmc); -if (new_state) { -/* -* During promotion from !chained to chained we must ensure -* the adjacent counter is stopped and its event destroyed -*/ -kvm_pmu_stop_counter (vcpu,kvm_pmu_get_alternate_pmc (pmc)); -set_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -return ; - } -clear_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -} -/** * kvm_pmu_set_counter_event_type - set selected counter to monitor some event * @vcpu: The vcpu pointer
@@ -752,11 +607,15 @@ static void kvm_pmu_update_pmc_chained(struct kvm_vcpu *vcpu, u64 select_idx)void kvm_pmu_set_counter_event_type (struct kvm_vcpu *vcpu,u64 data,u64 select_idx) { +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];u64 reg, mask;if (!kvm_vcpu_has_pmu (vcpu))return ; +kvm_pmu_stop_counter (vcpu, pmc); + mask =ARMV8_PMU_EVTYPE_MASK ; mask &= ~ARMV8_PMU_EVTYPE_EVENT ; mask |=kvm_pmu_event_mask (vcpu->kvm);
@@ -766,7 +625,6 @@ void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data,__vcpu_sys_reg (vcpu, reg) = data & mask; -kvm_pmu_update_pmc_chained (vcpu, select_idx);kvm_pmu_create_perf_event (vcpu, select_idx); }
include/kvm/arm_pmu.h +0 -2
@@ -11,7 +11,6 @@#include <asm/perf_event.h> #define ARMV8_PMU_CYCLE_IDX (ARMV8_PMU_MAX_COUNTERS - 1) -#define ARMV8_PMU_MAX_COUNTER_PAIRS ((ARMV8_PMU_MAX_COUNTERS + 1) >> 1)#ifdef CONFIG_HW_PERF_EVENTS
@@ -29,7 +28,6 @@ struct kvm_pmu {struct irq_work overflow_work;struct kvm_pmu_events events;struct kvm_pmc pmc[ARMV8_PMU_MAX_COUNTERS ]; -DECLARE_BITMAP (chained,ARMV8_PMU_MAX_COUNTER_PAIRS );int irq_num;bool created;bool irq_level;
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 1/9] KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
Hi Marc,
Quoted 37 lines · Marc Zyngier
Ricardo recently pointed out that the PMU chained counter emulation in KVM wasn't quite behaving like the one on actual hardware, in the sense that a chained counter would expose an overflow on both halves of a chained counter, while KVM would only expose the overflow on the top half.
The difference is subtle, but significant. What does the architecture say (DDI0087 H.a):
- Before PMUv3p4, all counters but the cycle counter are 32bit
- A 32bit counter that overflows generates a CHAIN event on the adjacent counter after exposing its own overflow status
- The CHAIN event is accounted if the counter is correctly configured (CHAIN event selected and counter enabled)This all means that our current implementation (which uses 64bit perf events) prevents us from emulating this overflow on the lower half.
How to fix this? By implementing the above, to the letter.
This largly results in code deletion, removing the notions of "counter pair", "chained counters", and "canonical counter". The code is further restructured to make the CHAIN handling similar to SWINC, as the two are now extremely similar in behaviour.
Reported-by Ricardo Koller <[email protected]>Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 324 +++++++++++--------------------------- include/kvm/arm_pmu.h | 2 - 2 files changed, 91 insertions(+), 235 deletions(-)arch/arm64/kvm/pmu-emul.c +0
-0
arch/arm64/kvm/pmu-emul.c static LIST_HEAD(arm_pmus); · line 27
Quoted 7 lines
arch/arm64/kvm/pmu-emul.c +0
-4@@ -21,10 +21,6 @@ static LIST_HEAD(arm_pmus);static DEFINE_MUTEX (arm_pmus_lock);static void kvm_pmu_create_perf_event (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_update_pmc_chained (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_stop_counter (struct kvm_vcpu *vcpu,struct kvm_pmc *pmc); - -#define PERF_ATTR_CFG1_KVM_PMU_CHAINED 0x1
nit: The name isn't a good fit for the config bit, but it might be nice to keep something around.
Quoted 226 lines
arch/arm64/kvm/pmu-emul.c +28
-116static u32 kvm_pmu_event_mask (struct kvm *kvm) {@@ -57,6 +53,11 @@ static bool kvm_pmu_idx_is_64bit(struct kvm_vcpu *vcpu, u64 select_idx) __vcpu_sys_reg(vcpu, PMCR_EL0) & ARMV8_PMU_PMCR_LC); } +static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx) +{ +return (!(idx &1 ) && (idx +1 ) <ARMV8_PMU_CYCLE_IDX ); +} +static struct kvm_vcpu *kvm_pmc_to_vcpu (struct kvm_pmc *pmc) {struct kvm_pmu *pmu;@@ -69,91 +70,22 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) }/** -* kvm_pmu_pmc_is_chained - determine if the pmc is chained -* @pmc: The PMU counter pointer -*/ -static bool kvm_pmu_pmc_is_chained (struct kvm_pmc *pmc) -{ -struct kvm_vcpu *vcpu =kvm_pmc_to_vcpu (pmc); - -return test_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -} - -/** -* kvm_pmu_idx_is_high_counter - determine if select_idx is a high/low counter -* @select_idx: The counter index -*/ -static bool kvm_pmu_idx_is_high_counter (u64 select_idx) -{ -return select_idx &0x1 ; -} - -/** -* kvm_pmu_get_canonical_pmc - obtain the canonical pmc -* @pmc: The PMU counter pointer -* -* When a pair of PMCs are chained together we use the low counter (canonical) -* to hold the underlying perf event. -*/ -static struct kvm_pmc *kvm_pmu_get_canonical_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; - -return pmc; -} -static struct kvm_pmc *kvm_pmu_get_alternate_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; -else -return pmc +1 ; -} - -/** -* kvm_pmu_idx_has_chain_evtype - determine if the event type is chain +* kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer * @select_idx: The counter index */ -static bool kvm_pmu_idx_has_chain_evtype (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 eventsel, reg; - - select_idx |=0x1 ; - -if (select_idx ==ARMV8_PMU_CYCLE_IDX ) -return false; - - reg =PMEVTYPER0_EL0 + select_idx; - eventsel =__vcpu_sys_reg (vcpu, reg) &kvm_pmu_event_mask (vcpu->kvm); - -return eventsel ==ARMV8_PMUV3_PERFCTR_CHAIN ; -} - -/** -* kvm_pmu_get_pair_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @pmc: The PMU counter pointer -*/ -static u64 kvm_pmu_get_pair_counter_value (struct kvm_vcpu *vcpu, -struct kvm_pmc *pmc) +u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) { -u64 counter, counter_high, reg, enabled, running; - -if (kvm_pmu_pmc_is_chained (pmc)) { - pmc =kvm_pmu_get_canonical_pmc (pmc); - reg =PMEVCNTR0_EL0 + pmc->idx; +u64 counter, reg, enabled, running; +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - counter =__vcpu_sys_reg (vcpu, reg); - counter_high =__vcpu_sys_reg (vcpu, reg +1 ); +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; - counter =lower_32_bits (counter) | (counter_high <<32 ); - }else { - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; - counter =__vcpu_sys_reg (vcpu, reg); - } + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) + ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + counter =__vcpu_sys_reg (vcpu, reg);/* * The real counter value is equal to the value of counter register plus @@ -163,29 +95,7 @@ static u64 kvm_pmu_get_pair_counter_value(struct kvm_vcpu *vcpu, counter +=perf_event_read_value (pmc->perf_event, &enabled, &running); -return counter; -} - -/** -* kvm_pmu_get_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @select_idx: The counter index -*/ -u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 counter; -struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - -if (!kvm_vcpu_has_pmu (vcpu)) -return 0 ; - - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); - -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (select_idx)) - counter =upper_32_bits (counter); -else if (select_idx !=ARMV8_PMU_CYCLE_IDX ) +if (select_idx !=ARMV8_PMU_CYCLE_IDX ) counter =lower_32_bits (counter);return counter;@@ -218,7 +128,6 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)*/ static void kvm_pmu_release_perf_event (struct kvm_pmc *pmc) { - pmc =kvm_pmu_get_canonical_pmc (pmc);if (pmc->perf_event) {perf_event_disable (pmc->perf_event);perf_event_release_kernel (pmc->perf_event);@@ -236,11 +145,10 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) {u64 counter, reg, val; - pmc =kvm_pmu_get_canonical_pmc (pmc);if (!pmc->perf_event)return ; - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); + counter =kvm_pmu_get_counter_value (vcpu, pmc->idx);if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) { reg =PMCCNTR_EL0 ;@@ -252,9 +160,6 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc)__vcpu_sys_reg (vcpu, reg) = val; -if (kvm_pmu_pmc_is_chained (pmc)) -__vcpu_sys_reg (vcpu, reg +1 ) =upper_32_bits (counter); -kvm_pmu_release_perf_event (pmc); }@@ -285,8 +190,6 @@ void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu)for_each_set_bit (i, &mask,32 )kvm_pmu_stop_counter (vcpu, &pmu->pmc[i]); - -bitmap_zero (vcpu->arch.pmu.chained,ARMV8_PMU_MAX_COUNTER_PAIRS ); }/** @@ -340,11 +243,8 @@ void kvm_pmu_enable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event) {perf_event_enable (pmc->perf_event);if (pmc->perf_event->state !=PERF_EVENT_STATE_ACTIVE )@@ -375,11 +275,8 @@ void kvm_pmu_disable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event)perf_event_disable (pmc->perf_event); }@@ -484,6 +381,51 @@ static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work)kvm_vcpu_kick (vcpu); } +/* +* Perform an increment on any of the counters described in @mask, +* generating the overflow if required, and propagate it as a chained +* event if possible. +*/ +static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu))
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work) · line 404
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +10
-0+return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg;
nit: replace 'reg' with 'counter' or 'val'. I think it might read better as it avoids a collision with counter_index_to_reg()
It feels like this patch could be broken down a bit as I found myself skipping around a bit. The s/pmc->idx/select_idx/ doesn't seem strictly necessary to bake in with this patch, either. Nonetheless, the end result looks good.
--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 1/9] KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
In reply to Marc ZyngierHi Marc,
Quoted 289 lines · Marc Zyngier
Ricardo recently pointed out that the PMU chained counter emulation in KVM wasn't quite behaving like the one on actual hardware, in the sense that a chained counter would expose an overflow on both halves of a chained counter, while KVM would only expose the overflow on the top half.
The difference is subtle, but significant. What does the architecture say (DDI0087 H.a):
- Before PMUv3p4, all counters but the cycle counter are 32bit
- A 32bit counter that overflows generates a CHAIN event on the adjacent counter after exposing its own overflow status
- The CHAIN event is accounted if the counter is correctly configured (CHAIN event selected and counter enabled)This all means that our current implementation (which uses 64bit perf events) prevents us from emulating this overflow on the lower half.
How to fix this? By implementing the above, to the letter.
This largly results in code deletion, removing the notions of "counter pair", "chained counters", and "canonical counter". The code is further restructured to make the CHAIN handling similar to SWINC, as the two are now extremely similar in behaviour.
Reported-by Ricardo Koller <[email protected]>Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 324 +++++++++++--------------------------- include/kvm/arm_pmu.h | 2 - 2 files changed, 91 insertions(+), 235 deletions(-)arch/arm64/kvm/pmu-emul.c +45
-120@@ -21,10 +21,6 @@ static LIST_HEAD(arm_pmus);static DEFINE_MUTEX (arm_pmus_lock);static void kvm_pmu_create_perf_event (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_update_pmc_chained (struct kvm_vcpu *vcpu,u64 select_idx); -static void kvm_pmu_stop_counter (struct kvm_vcpu *vcpu,struct kvm_pmc *pmc); - -#define PERF_ATTR_CFG1_KVM_PMU_CHAINED 0x1static u32 kvm_pmu_event_mask (struct kvm *kvm) {@@ -57,6 +53,11 @@ static bool kvm_pmu_idx_is_64bit(struct kvm_vcpu *vcpu, u64 select_idx) __vcpu_sys_reg(vcpu, PMCR_EL0) & ARMV8_PMU_PMCR_LC); } +static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx) +{ +return (!(idx &1 ) && (idx +1 ) <ARMV8_PMU_CYCLE_IDX ); +} +static struct kvm_vcpu *kvm_pmc_to_vcpu (struct kvm_pmc *pmc) {struct kvm_pmu *pmu;@@ -69,91 +70,22 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) }/** -* kvm_pmu_pmc_is_chained - determine if the pmc is chained -* @pmc: The PMU counter pointer -*/ -static bool kvm_pmu_pmc_is_chained (struct kvm_pmc *pmc) -{ -struct kvm_vcpu *vcpu =kvm_pmc_to_vcpu (pmc); - -return test_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -} - -/** -* kvm_pmu_idx_is_high_counter - determine if select_idx is a high/low counter -* @select_idx: The counter index -*/ -static bool kvm_pmu_idx_is_high_counter (u64 select_idx) -{ -return select_idx &0x1 ; -} - -/** -* kvm_pmu_get_canonical_pmc - obtain the canonical pmc -* @pmc: The PMU counter pointer -* -* When a pair of PMCs are chained together we use the low counter (canonical) -* to hold the underlying perf event. -*/ -static struct kvm_pmc *kvm_pmu_get_canonical_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; - -return pmc; -} -static struct kvm_pmc *kvm_pmu_get_alternate_pmc (struct kvm_pmc *pmc) -{ -if (kvm_pmu_idx_is_high_counter (pmc->idx)) -return pmc -1 ; -else -return pmc +1 ; -} - -/** -* kvm_pmu_idx_has_chain_evtype - determine if the event type is chain +* kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer * @select_idx: The counter index */ -static bool kvm_pmu_idx_has_chain_evtype (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 eventsel, reg; - - select_idx |=0x1 ; - -if (select_idx ==ARMV8_PMU_CYCLE_IDX ) -return false; - - reg =PMEVTYPER0_EL0 + select_idx; - eventsel =__vcpu_sys_reg (vcpu, reg) &kvm_pmu_event_mask (vcpu->kvm); - -return eventsel ==ARMV8_PMUV3_PERFCTR_CHAIN ; -} - -/** -* kvm_pmu_get_pair_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @pmc: The PMU counter pointer -*/ -static u64 kvm_pmu_get_pair_counter_value (struct kvm_vcpu *vcpu, -struct kvm_pmc *pmc) +u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) { -u64 counter, counter_high, reg, enabled, running; - -if (kvm_pmu_pmc_is_chained (pmc)) { - pmc =kvm_pmu_get_canonical_pmc (pmc); - reg =PMEVCNTR0_EL0 + pmc->idx; +u64 counter, reg, enabled, running; +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - counter =__vcpu_sys_reg (vcpu, reg); - counter_high =__vcpu_sys_reg (vcpu, reg +1 ); +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; - counter =lower_32_bits (counter) | (counter_high <<32 ); - }else { - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; - counter =__vcpu_sys_reg (vcpu, reg); - } + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) + ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + counter =__vcpu_sys_reg (vcpu, reg);/* * The real counter value is equal to the value of counter register plus @@ -163,29 +95,7 @@ static u64 kvm_pmu_get_pair_counter_value(struct kvm_vcpu *vcpu, counter +=perf_event_read_value (pmc->perf_event, &enabled, &running); -return counter; -} - -/** -* kvm_pmu_get_counter_value - get PMU counter value -* @vcpu: The vcpu pointer -* @select_idx: The counter index -*/ -u64 kvm_pmu_get_counter_value (struct kvm_vcpu *vcpu,u64 select_idx) -{ -u64 counter; -struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc = &pmu->pmc[select_idx]; - -if (!kvm_vcpu_has_pmu (vcpu)) -return 0 ; - - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); - -if (kvm_pmu_pmc_is_chained (pmc) && -kvm_pmu_idx_is_high_counter (select_idx)) - counter =upper_32_bits (counter); -else if (select_idx !=ARMV8_PMU_CYCLE_IDX ) +if (select_idx !=ARMV8_PMU_CYCLE_IDX ) counter =lower_32_bits (counter);return counter;@@ -218,7 +128,6 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)*/ static void kvm_pmu_release_perf_event (struct kvm_pmc *pmc) { - pmc =kvm_pmu_get_canonical_pmc (pmc);if (pmc->perf_event) {perf_event_disable (pmc->perf_event);perf_event_release_kernel (pmc->perf_event);@@ -236,11 +145,10 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) {u64 counter, reg, val; - pmc =kvm_pmu_get_canonical_pmc (pmc);if (!pmc->perf_event)return ; - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); + counter =kvm_pmu_get_counter_value (vcpu, pmc->idx);if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) { reg =PMCCNTR_EL0 ;@@ -252,9 +160,6 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc)__vcpu_sys_reg (vcpu, reg) = val; -if (kvm_pmu_pmc_is_chained (pmc)) -__vcpu_sys_reg (vcpu, reg +1 ) =upper_32_bits (counter); -kvm_pmu_release_perf_event (pmc); }@@ -285,8 +190,6 @@ void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu)for_each_set_bit (i, &mask,32 )kvm_pmu_stop_counter (vcpu, &pmu->pmc[i]); - -bitmap_zero (vcpu->arch.pmu.chained,ARMV8_PMU_MAX_COUNTER_PAIRS ); }/** @@ -340,11 +243,8 @@ void kvm_pmu_enable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event) {perf_event_enable (pmc->perf_event);if (pmc->perf_event->state !=PERF_EVENT_STATE_ACTIVE )@@ -375,11 +275,8 @@ void kvm_pmu_disable_counter_mask(struct kvm_vcpu *vcpu, u64 val) pmc = &pmu->pmc[i]; -/* A change in the enable state may affect the chain state */ -kvm_pmu_update_pmc_chained (vcpu, i);kvm_pmu_create_perf_event (vcpu, i); -/* At this point, pmc must be the canonical */ if (pmc->perf_event)perf_event_disable (pmc->perf_event); }@@ -484,6 +381,51 @@ static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work)kvm_vcpu_kick (vcpu); } +/* +* Perform an increment on any of the counters described in @mask, +* generating the overflow if required, and propagate it as a chained +* event if possible. +*/ +static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu)) +return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg; + +/* Filter on event type */ + type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); + type &=kvm_pmu_event_mask (vcpu->kvm); +if (type != event) +continue ; +
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work) · line 421
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +10
-0+/* Increment this counter */ + reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; + reg =lower_32_bits (reg); +__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; + +if (reg)/* No overflow? move on */ +continue ; + +/* Mark overflow */ +__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i);
Perhaps it might be useful to create another helper that takes care of just one counter (it would essentially do the code above in the loop). The helper could be used (in addition to the above loop) from the code below for the CHAIN event case and from kvm_pmu_perf_overflow(). Then unnecessary execution of for_each_set_bit() could be avoided for these two cases.
Quoted 83 lines
arch/arm64/kvm/pmu-emul.c +13
-45+ +if (kvm_pmu_counter_can_chain (vcpu, i)) +kvm_pmu_counter_increment (vcpu,BIT (i +1 ), +ARMV8_PMUV3_PERFCTR_CHAIN ); + } +} +/** * When the perf event overflows, set the overflow status and inform the vcpu. */ @@ -514,6 +456,10 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (idx); +if (kvm_pmu_counter_can_chain (vcpu, idx)) +kvm_pmu_counter_increment (vcpu,BIT (idx +1 ), +ARMV8_PMUV3_PERFCTR_CHAIN ); +if (kvm_pmu_overflow_status (vcpu)) {kvm_make_request (KVM_REQ_IRQ_PENDING , vcpu);@@ -533,50 +479,7 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,*/ void kvm_pmu_software_increment (struct kvm_vcpu *vcpu,u64 val) { -struct kvm_pmu *pmu = &vcpu->arch.pmu; -int i; - -if (!kvm_vcpu_has_pmu (vcpu)) -return ; - -if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) -return ; - -/* Weed out disabled counters */ - val &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); - -for (i =0 ; i <ARMV8_PMU_CYCLE_IDX ; i++) { -u64 type, reg; - -if (!(val &BIT (i))) -continue ; - -/* PMSWINC only applies to ... SW_INC! */ - type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); - type &=kvm_pmu_event_mask (vcpu->kvm); -if (type !=ARMV8_PMUV3_PERFCTR_SW_INCR ) -continue ; - -/* increment this even SW_INC counter */ - reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; - reg =lower_32_bits (reg); -__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; - -if (reg)/* no overflow on the low part */ -continue ; - -if (kvm_pmu_pmc_is_chained (&pmu->pmc[i])) { -/* increment the high counter */ - reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i +1 ) +1 ; - reg =lower_32_bits (reg); -__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i +1 ) = reg; -if (!reg)/* mark overflow on the high counter */ -__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i +1 ); - }else { -/* mark overflow on low counter */ -__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i); - } - } +kvm_pmu_counter_increment (vcpu, val,ARMV8_PMUV3_PERFCTR_SW_INCR ); }/** @@ -625,30 +528,27 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) {struct arm_pmu *arm_pmu = vcpu->kvm->arch.arm_pmu;struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];struct perf_event *event;struct perf_event_attr attr;u64 eventsel, counter, reg, data;
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) · line 537
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +1
-8-/* -* For chained counters the event type and filtering attributes are -* obtained from the low/even counter. We also use this counter to -* determine if the event is enabled/disabled. -*/ - pmc =kvm_pmu_get_canonical_pmc (&pmu->pmc[select_idx]); - - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx;
You may want to use select_idx instead of pmc->id for consistency ?
Quoted 113 lines
arch/arm64/kvm/pmu-emul.c +18
-59data =__vcpu_sys_reg (vcpu, reg);kvm_pmu_stop_counter (vcpu, pmc); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) +if (select_idx ==ARMV8_PMU_CYCLE_IDX ) eventsel =ARMV8_PMUV3_PERFCTR_CPU_CYCLES ;else eventsel = data &kvm_pmu_event_mask (vcpu->kvm); -/* Software increment event doesn't need to be backed by a perf event */ -if (eventsel ==ARMV8_PMUV3_PERFCTR_SW_INCR ) +/* +* Neither SW increment nor chained events need to be backed +* by a perf event. +*/ +if (eventsel ==ARMV8_PMUV3_PERFCTR_SW_INCR || + eventsel ==ARMV8_PMUV3_PERFCTR_CHAIN )return ;/* @@ -663,37 +563,27 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) attr.type = arm_pmu->pmu.type; attr.size =sizeof (attr); attr.pinned =1 ; - attr.disabled = !kvm_pmu_counter_is_enabled (vcpu, pmc->idx); + attr.disabled = !kvm_pmu_counter_is_enabled (vcpu, select_idx); attr.exclude_user = data &ARMV8_PMU_EXCLUDE_EL0 ?1 :0 ; attr.exclude_kernel = data &ARMV8_PMU_EXCLUDE_EL1 ?1 :0 ; attr.exclude_hv =1 ;/* Don't count EL2 events */ attr.exclude_host =1 ;/* Don't count host events */ attr.config = eventsel; - counter =kvm_pmu_get_pair_counter_value (vcpu, pmc); +/* If counting a 64bit event, advertise it to the perf code */ +if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) + attr.config1 |=1 ; -if (kvm_pmu_pmc_is_chained (pmc)) { -/** -* The initial sample period (overflow count) of an event. For -* chained counters we only support overflow interrupts on the -* high counter. -*/ - attr.sample_period = (-counter) &GENMASK (63 ,0 ); - attr.config1 |=PERF_ATTR_CFG1_KVM_PMU_CHAINED ; + counter =kvm_pmu_get_counter_value (vcpu, select_idx); - event =perf_event_create_kernel_counter (&attr,-1 , current, - kvm_pmu_perf_overflow, - pmc +1 ); - }else { -/* The initial sample period (overflow count) of an event. */ -if (kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) - attr.sample_period = (-counter) &GENMASK (63 ,0 ); -else - attr.sample_period = (-counter) &GENMASK (31 ,0 ); +/* The initial sample period (overflow count) of an event. */ +if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) + attr.sample_period = (-counter) &GENMASK (63 ,0 ); +else + attr.sample_period = (-counter) &GENMASK (31 ,0 ); - event =perf_event_create_kernel_counter (&attr,-1 , current, + event =perf_event_create_kernel_counter (&attr,-1 , current, kvm_pmu_perf_overflow, pmc); - }if (IS_ERR (event)) { pr_err_once("kvm: pmu event creation failed %ld\n",@@ -704,41 +594,6 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) pmc->perf_event = event; } -/** -* kvm_pmu_update_pmc_chained - update chained bitmap -* @vcpu: The vcpu pointer -* @select_idx: The number of selected counter -* -* Update the chained bitmap based on the event type written in the -* typer register and the enable state of the odd register. -*/ -static void kvm_pmu_update_pmc_chained (struct kvm_vcpu *vcpu,u64 select_idx) -{ -struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc = &pmu->pmc[select_idx], *canonical_pmc; -bool new_state, old_state; - - old_state =kvm_pmu_pmc_is_chained (pmc); - new_state =kvm_pmu_idx_has_chain_evtype (vcpu, pmc->idx) && -kvm_pmu_counter_is_enabled (vcpu, pmc->idx |0x1 ); - -if (old_state == new_state) -return ; - - canonical_pmc =kvm_pmu_get_canonical_pmc (pmc); -kvm_pmu_stop_counter (vcpu, canonical_pmc); -if (new_state) { -/* -* During promotion from !chained to chained we must ensure -* the adjacent counter is stopped and its event destroyed -*/ -kvm_pmu_stop_counter (vcpu,kvm_pmu_get_alternate_pmc (pmc)); -set_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -return ; - } -clear_bit (pmc->idx >>1 , vcpu->arch.pmu.chained); -} -/** * kvm_pmu_set_counter_event_type - set selected counter to monitor some event * @vcpu: The vcpu pointer @@ -752,11 +607,15 @@ static void kvm_pmu_update_pmc_chained(struct kvm_vcpu *vcpu, u64 select_idx) void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data,
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_update_pmc_chained(struct kvm_vcpu *vcpu, u64 select_idx) · line 617
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +3
-0u64 select_idx) { +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];u64 reg, mask;if (!kvm_vcpu_has_pmu (vcpu))return ; +kvm_pmu_stop_counter (vcpu, pmc);
It appears that kvm_pmu_stop_counter() doesn't have to be called here because it is called in the beginning of kvm_pmu_create_perf_event().
Thank you,
Reiji
Quoted 35 lines
arch/arm64/kvm/pmu-emul.c +1
-1+ mask =ARMV8_PMU_EVTYPE_MASK ; mask &= ~ARMV8_PMU_EVTYPE_EVENT ; mask |=kvm_pmu_event_mask (vcpu->kvm);@@ -766,7 +625,6 @@ void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data,__vcpu_sys_reg (vcpu, reg) = data & mask; -kvm_pmu_update_pmc_chained (vcpu, select_idx);kvm_pmu_create_perf_event (vcpu, select_idx); }include/kvm/arm_pmu.h +0
-3@@ -11,7 +11,6 @@#include <asm/perf_event.h> #define ARMV8_PMU_CYCLE_IDX (ARMV8_PMU_MAX_COUNTERS - 1) -#define ARMV8_PMU_MAX_COUNTER_PAIRS ((ARMV8_PMU_MAX_COUNTERS + 1) >> 1)#ifdef CONFIG_HW_PERF_EVENTS @@ -29,7 +28,6 @@ struct kvm_pmu {struct irq_work overflow_work;struct kvm_pmu_events events;struct kvm_pmc pmc[ARMV8_PMU_MAX_COUNTERS ]; -DECLARE_BITMAP (chained,ARMV8_PMU_MAX_COUNTER_PAIRS );int irq_num;bool created;bool irq_level; --2.34.1
Quoted 4 lines
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 1/9] KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
Hi Reiji,
Catching up on this.
Quoted 41 lines · Reiji Watanabe
Hi Marc,
Quoted 37 lines · Marc Zyngier
Ricardo recently pointed out that the PMU chained counter emulation in KVM wasn't quite behaving like the one on actual hardware, in the sense that a chained counter would expose an overflow on both halves of a chained counter, while KVM would only expose the overflow on the top half.
The difference is subtle, but significant. What does the architecture say (DDI0087 H.a):
- Before PMUv3p4, all counters but the cycle counter are 32bit
- A 32bit counter that overflows generates a CHAIN event on the adjacent counter after exposing its own overflow status
- The CHAIN event is accounted if the counter is correctly configured (CHAIN event selected and counter enabled)This all means that our current implementation (which uses 64bit perf events) prevents us from emulating this overflow on the lower half.
How to fix this? By implementing the above, to the letter.
This largly results in code deletion, removing the notions of "counter pair", "chained counters", and "canonical counter". The code is further restructured to make the CHAIN handling similar to SWINC, as the two are now extremely similar in behaviour.
Reported-by Ricardo Koller <[email protected]>Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 324 +++++++++++--------------------------- include/kvm/arm_pmu.h | 2 - 2 files changed, 91 insertions(+), 235 deletions(-)arch/arm64/kvm/pmu-emul.c +0
-0
[...]
Quoted 37 lines
Quoted 37 lines
arch/arm64/kvm/pmu-emul.c +37
-0+/* +* Perform an increment on any of the counters described in @mask, +* generating the overflow if required, and propagate it as a chained +* event if possible. +*/ +static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu)) +return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg; + +/* Filter on event type */ + type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); + type &=kvm_pmu_event_mask (vcpu->kvm); +if (type != event) +continue ; + +/* Increment this counter */ + reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; + reg =lower_32_bits (reg); +__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; + +if (reg)/* No overflow? move on */ +continue ; + +/* Mark overflow */
Quoted 8 lines
Quoted 1 line
+
__vcpu_sys_reg(vcpu, PMOVSSET_EL0)|= BIT(i);Perhaps it might be useful to create another helper that takes care of just one counter (it would essentially do the code above in the loop). The helper could be used (in addition to the above loop) from the code below for the CHAIN event case and from
kvm_pmu_perf_overflow(). Then unnecessary execution offor_each_set_bit()could be avoided for these two cases.
I'm not sure it really helps. We would still need to check whether the counter is enabled, and we'd need to bring that into the helper instead of keeping it outside of the loop.
[...]
Quoted 10 lines
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +1
-1@@ -625,30 +528,27 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) {struct arm_pmu *arm_pmu = vcpu->kvm->arch.arm_pmu;struct kvm_pmu *pmu = &vcpu->arch.pmu; -struct kvm_pmc *pmc; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];struct perf_event *event;struct perf_event_attr attr;u64 eventsel, counter, reg, data;
Quoted 12 lines
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) · line 537
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +1
-8-/* -* For chained counters the event type and filtering attributes are -* obtained from the low/even counter. We also use this counter to -* determine if the event is enabled/disabled. -*/ - pmc =kvm_pmu_get_canonical_pmc (&pmu->pmc[select_idx]); - - reg = (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) + reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx;You may want to use
select_idxinstead ofpmc->idfor consistency ?
Yes. Although Oliver had a point in saying that these pmc->idx vs select_idx conversions were not strictly necessary and cluttered the patch.
[...]
Quoted 8 lines
Quoted 8 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -752,11 +607,15 @@ static void kvm_pmu_update_pmc_chained(struct kvm_vcpu *vcpu, u64 select_idx)void kvm_pmu_set_counter_event_type (struct kvm_vcpu *vcpu,u64 data,u64 select_idx) { +struct kvm_pmu *pmu = &vcpu->arch.pmu; +struct kvm_pmc *pmc = &pmu->pmc[select_idx];u64 reg, mask;
Quoted 7 lines
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_update_pmc_chained(struct kvm_vcpu *vcpu, u64 select_idx) · line 617
Quoted 4 lines
arch/arm64/kvm/pmu-emul.c +1
-0if (!kvm_vcpu_has_pmu (vcpu))return ; +kvm_pmu_stop_counter (vcpu, pmc);It appears that
kvm_pmu_stop_counter()doesn't have to be called here because it is called in the beginning ofkvm_pmu_create_perf_event().
It feels a bit odd to change the event type without stopping the counter first, but I can't see anything going wrong if we omit it.
I'll drop it.
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 1/9] KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
Hi Marc,
Quoted 41 lines
Quoted 32 lines
Quoted 32 lines
arch/arm64/kvm/pmu-emul.c +32
-0+static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu)) +return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg; + +/* Filter on event type */ + type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); + type &=kvm_pmu_event_mask (vcpu->kvm); +if (type != event) +continue ; + +/* Increment this counter */ + reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; + reg =lower_32_bits (reg); +__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; + +if (reg)/* No overflow? move on */ +continue ; + +/* Mark overflow */ Quoted 8 lines
Quoted 1 line
+
__vcpu_sys_reg(vcpu, PMOVSSET_EL0)|= BIT(i);Perhaps it might be useful to create another helper that takes care of just one counter (it would essentially do the code above in the loop). The helper could be used (in addition to the above loop) from the code below for the CHAIN event case and from
kvm_pmu_perf_overflow(). Then unnecessary execution offor_each_set_bit()could be avoided for these two cases.
Quoted 3 lines
I'm not sure it really helps. We would still need to check whether the counter is enabled, and we'd need to bring that into the helper instead of keeping it outside of the loop.
That's true. It seems that I overlooked that.
Although it appears checking with kvm_vcpu_has_pmu() is unnecessary (redundant), the check with PMCR_EL0.E is necessary.
Thank you,
Reiji
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 1/9] KVM: arm64: PMU: Align chained counter implementation with architecture pseudocode
Hi Reiji,
Quoted 43 lines · Reiji Watanabe
Hi Marc,
Quoted 41 lines
Quoted 40 lines
Quoted 23 lines
arch/arm64/kvm/pmu-emul.c +23
-0+static void kvm_pmu_counter_increment (struct kvm_vcpu *vcpu, +unsigned long mask,u32 event) +{ +int i; + +if (!kvm_vcpu_has_pmu (vcpu)) +return ; + +if (!(__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_E )) +return ; + +/* Weed out disabled counters */ + mask &=__vcpu_sys_reg (vcpu,PMCNTENSET_EL0 ); + +for_each_set_bit (i, &mask,ARMV8_PMU_CYCLE_IDX ) { +u64 type, reg; + +/* Filter on event type */ + type =__vcpu_sys_reg (vcpu,PMEVTYPER0_EL0 + i); + type &=kvm_pmu_event_mask (vcpu->kvm); +if (type != event) +continue ; +arch/arm64/kvm/pmu-emul.c static void kvm_pmu_perf_overflow_notify_vcpu(struct irq_work *work) · line 421
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +10
-0+/* Increment this counter */ + reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; + reg =lower_32_bits (reg); +__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; + +if (reg)/* No overflow? move on */ +continue ; + +/* Mark overflow */ +__vcpu_sys_reg (vcpu,PMOVSSET_EL0 ) |=BIT (i);Perhaps it might be useful to create another helper that takes care of just one counter (it would essentially do the code above in the loop). The helper could be used (in addition to the above loop) from the code below for the CHAIN event case and from
kvm_pmu_perf_overflow(). Then unnecessary execution offor_each_set_bit()could be avoided for these two cases.
Quoted 7 lines
Quoted 3 lines
I'm not sure it really helps. We would still need to check whether the counter is enabled, and we'd need to bring that into the helper instead of keeping it outside of the loop.
That's true. It seems that I overlooked that.
Although it appears checking withkvm_vcpu_has_pmu()is unnecessary (redundant), the check with PMCR_EL0.E is necessary.
Ah indeed, I'll drop the kvm_vcpu_has_pmu() check.
Thanks,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 2/9 KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
In reply to Marc ZyngierThe PMU architecture makes a subtle difference between a 64bit counter and a counter that has a 64bit overflow. This is for example the case of the cycle counter, which can generate an overflow on a 32bit boundary if PMCR_EL0.LC==0 despite the accumulation being done on 64 bits.
Use this distinction in the few cases where it matters in the code, as we will reuse this with PMUv3p5 long counters.
--- arch/arm64/kvm/pmu-emul.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-)
arch/arm64/kvm/pmu-emul.c +10 -4
@@ -48,6 +48,11 @@ static u32 kvm_pmu_event_mask(struct kvm *kvm)* @select_idx: The counter index */ static bool kvm_pmu_idx_is_64bit (struct kvm_vcpu *vcpu,u64 select_idx) +{ +return (select_idx ==ARMV8_PMU_CYCLE_IDX ); +} + +static bool kvm_pmu_idx_has_64bit_overflow (struct kvm_vcpu *vcpu,u64 select_idx) {return (select_idx ==ARMV8_PMU_CYCLE_IDX &&__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_LC );
@@ -55,7 +60,8 @@ static bool kvm_pmu_idx_is_64bit(struct kvm_vcpu *vcpu, u64 select_idx)static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx) { -return (!(idx &1 ) && (idx +1 ) <ARMV8_PMU_CYCLE_IDX ); +return (!(idx &1 ) && (idx +1 ) <ARMV8_PMU_CYCLE_IDX && + !kvm_pmu_idx_has_64bit_overflow (vcpu, idx)); }static struct kvm_vcpu *kvm_pmc_to_vcpu (struct kvm_pmc *pmc)
@@ -95,7 +101,7 @@ u64 kvm_pmu_get_counter_value(struct kvm_vcpu *vcpu, u64 select_idx) counter +=perf_event_read_value (pmc->perf_event, &enabled, &running); -if (select_idx !=ARMV8_PMU_CYCLE_IDX ) +if (!kvm_pmu_idx_is_64bit (vcpu, select_idx)) counter =lower_32_bits (counter);return counter;
@@ -447,7 +453,7 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,*/ period = -(local64_read (&perf_event->count)); -if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) +if (!kvm_pmu_idx_has_64bit_overflow (vcpu, pmc->idx)) period &=GENMASK (31 ,0 );local64_set (&perf_event->hw.period_left,0 );
@@ -577,7 +583,7 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) counter =kvm_pmu_get_counter_value (vcpu, select_idx);/* The initial sample period (overflow count) of an event. */ -if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) +if (kvm_pmu_idx_has_64bit_overflow (vcpu, select_idx)) attr.sample_period = (-counter) &GENMASK (63 ,0 );else attr.sample_period = (-counter) &GENMASK (31 ,0 );
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 3/9 KVM: arm64: PMU: Only narrow counters that are not 64bit wide
In reply to Marc ZyngierThe current PMU emulation sometimes narrows counters to 32bit if the counter isn't the cycle counter. As this is going to change with PMUv3p5 where the counters are all 64bit, only perform the narrowing on 32bit counters.
--- arch/arm64/kvm/pmu-emul.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-)
arch/arm64/kvm/pmu-emul.c +9 -8
@@ -149,22 +149,22 @@ static void kvm_pmu_release_perf_event(struct kvm_pmc *pmc)*/ static void kvm_pmu_stop_counter (struct kvm_vcpu *vcpu,struct kvm_pmc *pmc) { -u64 counter, reg, val; +u64 counter, reg;if (!pmc->perf_event)return ; counter =kvm_pmu_get_counter_value (vcpu, pmc->idx); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) { +if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) reg =PMCCNTR_EL0 ; - val = counter; - }else { +else reg =PMEVCNTR0_EL0 + pmc->idx; - val =lower_32_bits (counter); - } -__vcpu_sys_reg (vcpu, reg) = val; +if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) + counter =lower_32_bits (counter); + +__vcpu_sys_reg (vcpu, reg) = counter;kvm_pmu_release_perf_event (pmc); }
@@ -417,7 +417,8 @@ static void kvm_pmu_counter_increment(struct kvm_vcpu *vcpu,/* Increment this counter */ reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; - reg =lower_32_bits (reg); +if (!kvm_pmu_idx_is_64bit (vcpu, i)) + reg =lower_32_bits (reg);__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg;if (reg)/* No overflow? move on */
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 3/9] KVM: arm64: PMU: Only narrow counters that are not 64bit wide
Hi Marc,
Quoted 30 lines · Marc Zyngier
The current PMU emulation sometimes narrows counters to 32bit if the counter isn't the cycle counter. As this is going to change with PMUv3p5 where the counters are all 64bit, only perform the narrowing on 32bit counters.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-)arch/arm64/kvm/pmu-emul.c +2
-2@@ -149,22 +149,22 @@ static void kvm_pmu_release_perf_event(struct kvm_pmc *pmc)*/ static void kvm_pmu_stop_counter (struct kvm_vcpu *vcpu,struct kvm_pmc *pmc) { -u64 counter, reg, val; +u64 counter, reg;if (!pmc->perf_event)return ; counter =kvm_pmu_get_counter_value (vcpu, pmc->idx); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) { +if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) reg =PMCCNTR_EL0 ;
arch/arm64/kvm/pmu-emul.c static void kvm_pmu_release_perf_event(struct kvm_pmc *pmc) · line 165
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +3
-5- val = counter; - }else { +else reg =PMEVCNTR0_EL0 + pmc->idx; - val =lower_32_bits (counter); - } -__vcpu_sys_reg (vcpu, reg) = val; +if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) + counter =lower_32_bits (counter);
It appears that narrowing the counter to 32bit here is unnecessary because it is already done by kvm_pmu_get_counter_value().
Thank you,
Reiji
Quoted 18 lines
arch/arm64/kvm/pmu-emul.c +4
-2+ +__vcpu_sys_reg (vcpu, reg) = counter;kvm_pmu_release_perf_event (pmc); }@@ -417,7 +417,8 @@ static void kvm_pmu_counter_increment(struct kvm_vcpu *vcpu,/* Increment this counter */ reg =__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) +1 ; - reg =lower_32_bits (reg); +if (!kvm_pmu_idx_is_64bit (vcpu, i)) + reg =lower_32_bits (reg);__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg;if (reg)/* No overflow? move on */ --2.34.1
Quoted 4 lines
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 4/9 KVM: arm64: PMU: Add counter_index_to_*reg() helpers
In reply to Marc ZyngierIn order to reduce the boilerplate code, add two helpers returning the counter register index (resp. the event register) in the vcpu register file from the counter index.
--- arch/arm64/kvm/pmu-emul.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-)
arch/arm64/kvm/pmu-emul.c +15 -12
@@ -75,6 +75,16 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) return container_of(vcpu_arch, struct kvm_vcpu, arch); } +static u32 counter_index_to_reg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + idx; +} + +static u32 counter_index_to_evtreg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + idx; +} +/** * kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer
@@ -89,8 +99,7 @@ u64 kvm_pmu_get_counter_value(struct kvm_vcpu *vcpu, u64 select_idx)if (!kvm_vcpu_has_pmu (vcpu))return 0 ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (select_idx); counter =__vcpu_sys_reg (vcpu, reg);/*
@@ -120,8 +129,7 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)if (!kvm_vcpu_has_pmu (vcpu))return ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + select_idx; + reg =counter_index_to_reg (select_idx);__vcpu_sys_reg (vcpu, reg) += (s64 )val -kvm_pmu_get_counter_value (vcpu, select_idx);/* Recreate the perf event to reflect the updated sample_period */
@@ -156,10 +164,7 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) counter =kvm_pmu_get_counter_value (vcpu, pmc->idx); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - reg =PMCCNTR_EL0 ; -else - reg =PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (pmc->idx);if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) counter =lower_32_bits (counter);
@@ -540,8 +545,7 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx)struct perf_event_attr attr;u64 eventsel, counter, reg, data; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx; + reg =counter_index_to_evtreg (select_idx); data =__vcpu_sys_reg (vcpu, reg);kvm_pmu_stop_counter (vcpu, pmc);
@@ -627,8 +631,7 @@ void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data, mask &= ~ARMV8_PMU_EVTYPE_EVENT ; mask |=kvm_pmu_event_mask (vcpu->kvm); - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + select_idx; + reg =counter_index_to_evtreg (select_idx);__vcpu_sys_reg (vcpu, reg) = data & mask;
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 4/9] KVM: arm64: PMU: Add counter_index_to_*reg() helpers
Quoted 5 lines · Marc Zyngier
In order to reduce the boilerplate code, add two helpers returning the counter register index (resp. the event register) in the vcpu register file from the counter index.
Signed-off-by Marc Zyngier <[email protected]>
Quoted 80 lines
--- arch/arm64/kvm/pmu-emul.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-)arch/arm64/kvm/pmu-emul.c +15
-12@@ -75,6 +75,16 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) return container_of(vcpu_arch, struct kvm_vcpu, arch); } +static u32 counter_index_to_reg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + idx; +} + +static u32 counter_index_to_evtreg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + idx; +} +/** * kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer @@ -89,8 +99,7 @@ u64 kvm_pmu_get_counter_value(struct kvm_vcpu *vcpu, u64 select_idx)if (!kvm_vcpu_has_pmu (vcpu))return 0 ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (select_idx); counter =__vcpu_sys_reg (vcpu, reg);/* @@ -120,8 +129,7 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)if (!kvm_vcpu_has_pmu (vcpu))return ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + select_idx; + reg =counter_index_to_reg (select_idx);__vcpu_sys_reg (vcpu, reg) += (s64 )val -kvm_pmu_get_counter_value (vcpu, select_idx);/* Recreate the perf event to reflect the updated sample_period */ @@ -156,10 +164,7 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) counter =kvm_pmu_get_counter_value (vcpu, pmc->idx); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - reg =PMCCNTR_EL0 ; -else - reg =PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (pmc->idx);if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) counter =lower_32_bits (counter);@@ -540,8 +545,7 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx)struct perf_event_attr attr;u64 eventsel, counter, reg, data; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx; + reg =counter_index_to_evtreg (select_idx); data =__vcpu_sys_reg (vcpu, reg);kvm_pmu_stop_counter (vcpu, pmc);@@ -627,8 +631,7 @@ void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data, mask &= ~ARMV8_PMU_EVTYPE_EVENT ; mask |=kvm_pmu_event_mask (vcpu->kvm); - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + select_idx; + reg =counter_index_to_evtreg (select_idx);__vcpu_sys_reg (vcpu, reg) = data & mask;Signature · 1 line
2.34.1
Re: [PATCH 4/9] KVM: arm64: PMU: Add counter_index_to_*reg() helpers
Quoted 21 lines · Oliver Upton
Quoted 5 lines · Marc Zyngier
In order to reduce the boilerplate code, add two helpers returning the counter register index (resp. the event register) in the vcpu register file from the counter index.
Signed-off-by Marc Zyngier <[email protected]>Reviewed-by Oliver Upton <[email protected]>Quoted 12 lines
--- arch/arm64/kvm/pmu-emul.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-)arch/arm64/kvm/pmu-emul.c +0
-0@@ -75,6 +75,16 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) return container_of(vcpu_arch, struct kvm_vcpu, arch); }
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) · line 87
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +10
-0+static u32 counter_index_to_reg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + idx; +} + +static u32 counter_index_to_evtreg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + idx; +} +
After reading the series, do you think these helpers could be applied to kvm_pmu_counter_increment() as well?
--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 4/9] KVM: arm64: PMU: Add counter_index_to_*reg() helpers
In reply to Marc ZyngierHi Marc,
Quoted 74 lines · Marc Zyngier
In order to reduce the boilerplate code, add two helpers returning the counter register index (resp. the event register) in the vcpu register file from the counter index.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-)arch/arm64/kvm/pmu-emul.c +14
-10@@ -75,6 +75,16 @@ static struct kvm_vcpu *kvm_pmc_to_vcpu(struct kvm_pmc *pmc) return container_of(vcpu_arch, struct kvm_vcpu, arch); } +static u32 counter_index_to_reg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + idx; +} + +static u32 counter_index_to_evtreg (u64 idx) +{ +return (idx ==ARMV8_PMU_CYCLE_IDX ) ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + idx; +} +/** * kvm_pmu_get_counter_value - get PMU counter value * @vcpu: The vcpu pointer @@ -89,8 +99,7 @@ u64 kvm_pmu_get_counter_value(struct kvm_vcpu *vcpu, u64 select_idx)if (!kvm_vcpu_has_pmu (vcpu))return 0 ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (select_idx); counter =__vcpu_sys_reg (vcpu, reg);/* @@ -120,8 +129,7 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)if (!kvm_vcpu_has_pmu (vcpu))return ; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCNTR_EL0 :PMEVCNTR0_EL0 + select_idx; + reg =counter_index_to_reg (select_idx);__vcpu_sys_reg (vcpu, reg) += (s64 )val -kvm_pmu_get_counter_value (vcpu, select_idx);/* Recreate the perf event to reflect the updated sample_period */ @@ -156,10 +164,7 @@ static void kvm_pmu_stop_counter(struct kvm_vcpu *vcpu, struct kvm_pmc *pmc) counter =kvm_pmu_get_counter_value (vcpu, pmc->idx); -if (pmc->idx ==ARMV8_PMU_CYCLE_IDX ) - reg =PMCCNTR_EL0 ; -else - reg =PMEVCNTR0_EL0 + pmc->idx; + reg =counter_index_to_reg (pmc->idx);if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) counter =lower_32_bits (counter);@@ -540,8 +545,7 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx)struct perf_event_attr attr;u64 eventsel, counter, reg, data; - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + pmc->idx; + reg =counter_index_to_evtreg (select_idx); data =__vcpu_sys_reg (vcpu, reg);kvm_pmu_stop_counter (vcpu, pmc);
arch/arm64/kvm/pmu-emul.c void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data,
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +1
-3@@ -627,8 +631,7 @@ void kvm_pmu_set_counter_event_type(struct kvm_vcpu *vcpu, u64 data, mask &= ~ARMV8_PMU_EVTYPE_EVENT ; mask |=kvm_pmu_event_mask (vcpu->kvm); - reg = (select_idx ==ARMV8_PMU_CYCLE_IDX ) - ?PMCCFILTR_EL0 :PMEVTYPER0_EL0 + select_idx; + reg =counter_index_to_evtreg (select_idx);__vcpu_sys_reg (vcpu, reg) = data & mask; --
counter_index_to_evtreg() could also be used in access_pmu_evtyper() in sys_regs.c (counter_index_to_evtreg() is currently defined as static in pmu-emul.c though).
---
static bool access_pmu_evtyper(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
const struct sys_reg_desc *r)
{
<...>
} else if (r->CRn == 14 && (r->CRm & 12) == 12) {
idx = ((r->CRm & 3) << 3) | (r->Op2 & 7);
if (idx == ARMV8_PMU_CYCLE_IDX)
reg = PMCCFILTR_EL0;
else
/* PMEVTYPERn_EL0 */
reg = PMEVTYPER0_EL0 + idx;
<...>---
Thank you,
Reiji
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 5/9 KVM: arm64: PMU: Simplify setting a counter to a specific value
In reply to Marc Zyngierkvm_pmu_set_counter_value() is pretty odd, as it tries to update the counter value while taking into account the value that is currently held by the running perf counter.
This is not only complicated, this is quite wrong. Nowhere in the architecture is it said that the counter would be offset by something that is pending. The counter should be updated with the value set by SW, and start counting from there if required.
Remove the odd computation and just assign the provided value after having released the perf event (which is then restarted).
--- arch/arm64/kvm/pmu-emul.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
arch/arm64/kvm/pmu-emul.c +4 -1
@@ -21,6 +21,7 @@ static LIST_HEAD(arm_pmus);static DEFINE_MUTEX (arm_pmus_lock);static void kvm_pmu_create_perf_event (struct kvm_vcpu *vcpu,u64 select_idx); +static void kvm_pmu_release_perf_event (struct kvm_pmc *pmc);static u32 kvm_pmu_event_mask (struct kvm *kvm) {
@@ -129,8 +130,10 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)if (!kvm_vcpu_has_pmu (vcpu))return ; +kvm_pmu_release_perf_event (&vcpu->arch.pmu.pmc[select_idx]); + reg =counter_index_to_reg (select_idx); -__vcpu_sys_reg (vcpu, reg) += (s64 )val -kvm_pmu_get_counter_value (vcpu, select_idx); +__vcpu_sys_reg (vcpu, reg) = val;/* Recreate the perf event to reflect the updated sample_period */ kvm_pmu_create_perf_event (vcpu, select_idx);
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 5/9] KVM: arm64: PMU: Simplify setting a counter to a specific value
Quoted 10 lines · Marc Zyngier
kvm_pmu_set_counter_value()is pretty odd, as it tries to update the counter value while taking into account the value that is currently held by the running perf counter.This is not only complicated, this is quite wrong. Nowhere in the architecture is it said that the counter would be offset by something that is pending. The counter should be updated with the value set by SW, and start counting from there if required.
Quoted 4 lines
Remove the odd computation and just assign the provided value after having released the perf event (which is then restarted).
Signed-off-by Marc Zyngier <[email protected]>
Looks much better.
Quoted 17 lines
--- arch/arm64/kvm/pmu-emul.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)arch/arm64/kvm/pmu-emul.c +1
-0@@ -21,6 +21,7 @@ static LIST_HEAD(arm_pmus);static DEFINE_MUTEX (arm_pmus_lock);static void kvm_pmu_create_perf_event (struct kvm_vcpu *vcpu,u64 select_idx); +static void kvm_pmu_release_perf_event (struct kvm_pmc *pmc);static u32 kvm_pmu_event_mask (struct kvm *kvm) {
arch/arm64/kvm/pmu-emul.c void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val) · line 134
Quoted 5 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -129,8 +130,10 @@ void kvm_pmu_set_counter_value(struct kvm_vcpu *vcpu, u64 select_idx, u64 val)if (!kvm_vcpu_has_pmu (vcpu))return ; +kvm_pmu_release_perf_event (&vcpu->arch.pmu.pmc[select_idx]); +
<bikeshed>
Not your code, but since we're here: it seems as though there is some inconsistency in parameterization as some functions take an index and others take a kvm_pmc pointer. For example, kvm_pmu_{create,release}_perf_event() are inconsistent.
It might be nice to pick a scheme and apply consistently throughout.
</bikeshed>
--
Thanks,
Oliver
Quoted 10 lines
arch/arm64/kvm/pmu-emul.c +1
-1reg =counter_index_to_reg (select_idx); -__vcpu_sys_reg (vcpu, reg) += (s64 )val -kvm_pmu_get_counter_value (vcpu, select_idx); +__vcpu_sys_reg (vcpu, reg) = val;/* Recreate the perf event to reflect the updated sample_period */ kvm_pmu_create_perf_event (vcpu, select_idx);Signature · 1 line
2.34.1
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 6/9 KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
In reply to Marc ZyngierAs further patches will enable the selection of a PMU revision from userspace, sample the supported PMU revision at VM creation time, rather than building each time the ID_AA64DFR0_EL1 register is accessed.
This shouldn't result in any change in behaviour.
--- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 ++++++ arch/arm64/kvm/pmu-emul.c | 11 +++++++++++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++----- include/kvm/arm_pmu.h | 6 ++++++ 5 files changed, 45 insertions(+), 5 deletions(-)
arch/arm64/include/asm/kvm_host.h +1 -0
@@ -163,6 +163,7 @@ struct kvm_arch {u8 pfr0_csv2;u8 pfr0_csv3; +u8 dfr0_pmuver;/* Hypercall features firmware registers' descriptor */ struct kvm_smccc_features smccc_feat;
arch/arm64/kvm/arm.c +6 -0
@@ -164,6 +164,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)set_default_spectre (kvm);kvm_arm_init_hypercalls (kvm); +/* +* Initialise the default PMUver before there is a chance to +* create an actual PMU. +*/ + kvm->arch.dfr0_pmuver =kvm_arm_pmu_get_host_pmuver (); +return ret; out_free_stage2_pgd:kvm_free_stage2_pgd (&kvm->arch.mmu);
arch/arm64/kvm/pmu-emul.c +11 -0
@@ -1021,3 +1021,14 @@ int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return -ENXIO ; } + +u8 kvm_arm_pmu_get_host_pmuver (void ) +{ +u64 tmp; + + tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); + tmp =cpuid_feature_cap_perfmon_field (tmp, +ID_AA64DFR0_PMUVER_SHIFT , +ID_AA64DFR0_PMUVER_8_4 ); +return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); +}
arch/arm64/kvm/sys_regs.c +21 -5
@@ -1062,6 +1062,22 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,return true; } +static u8 pmuver_to_perfmon (const struct kvm_vcpu *vcpu) +{ +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; + +switch (vcpu->kvm->arch.dfr0_pmuver) { +case ID_AA64DFR0_PMUVER_8_0 : +return ID_DFR0_PERFMON_8_0 ; +case ID_AA64DFR0_PMUVER_IMP_DEF : +return 0 ; +default : +/* Anything ARMv8.4+ has the same value. For now. */ +return vcpu->kvm->arch.dfr0_pmuver; + } +} +/* Read a sanitised cpufeature ID register by sys_reg_desc */ static u64 read_id_reg (const struct kvm_vcpu *vcpu,struct sys_reg_desc const *r,bool raz)
@@ -1112,10 +1128,10 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit debug to ARMv8.0 */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ); val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ),6 ); -/* Limit guests to PMUv3 for ARMv8.4 */ - val =cpuid_feature_cap_perfmon_field (val, -ID_AA64DFR0_PMUVER_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_AA64DFR0_PMUVER_8_4 :0 ); +/* Set PMUver to the required version */ + val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ); + val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), +kvm_vcpu_has_pmu (vcpu) ? vcpu->kvm->arch.dfr0_pmuver :0 );/* Hide SPE from guests */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMSVER );break ;
@@ -1123,7 +1139,7 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit guests to PMUv3 for ARMv8.4 */ val =cpuid_feature_cap_perfmon_field (val,ID_DFR0_PERFMON_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_DFR0_PERFMON_8_4 :0 ); +pmuver_to_perfmon (vcpu));break ; }
include/kvm/arm_pmu.h +6 -0
@@ -89,6 +89,8 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ } while (0) +u8 kvm_arm_pmu_get_host_pmuver(void); +#else struct kvm_pmu { };
@@ -154,6 +156,10 @@ static inline u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1)static inline void kvm_pmu_update_vcpu_events (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_guest (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_host (struct kvm_vcpu *vcpu) {} +static inline u8 kvm_arm_pmu_get_host_pmuver (void ) +{ +return 0 ; +}#endif
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 6/9] KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
Hi Marc,
Quoted 51 lines · Marc Zyngier
As further patches will enable the selection of a PMU revision from userspace, sample the supported PMU revision at VM creation time, rather than building each time the
ID_AA64DFR0_EL1register is accessed.This shouldn't result in any change in behaviour.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 ++++++ arch/arm64/kvm/pmu-emul.c | 11 +++++++++++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++----- include/kvm/arm_pmu.h | 6 ++++++ 5 files changed, 45 insertions(+), 5 deletions(-)arch/arm64/include/asm/kvm_host.h +1
-0@@ -163,6 +163,7 @@ struct kvm_arch {u8 pfr0_csv2;u8 pfr0_csv3; +u8 dfr0_pmuver;/* Hypercall features firmware registers' descriptor */ struct kvm_smccc_features smccc_feat;arch/arm64/kvm/arm.c +6
-0@@ -164,6 +164,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)set_default_spectre (kvm);kvm_arm_init_hypercalls (kvm); +/* +* Initialise the default PMUver before there is a chance to +* create an actual PMU. +*/ + kvm->arch.dfr0_pmuver =kvm_arm_pmu_get_host_pmuver (); +return ret; out_free_stage2_pgd:kvm_free_stage2_pgd (&kvm->arch.mmu);arch/arm64/kvm/pmu-emul.c +0
-0
arch/arm64/kvm/pmu-emul.c int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1025
Quoted 5 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -1021,3 +1021,14 @@ int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return -ENXIO ; } + +u8 kvm_arm_pmu_get_host_pmuver (void )
Nit: Since this function doesn't simply return the host's pmuver, but the pmuver limit for guests, perhaps "kvm_arm_pmu_get_guest_pmuver_limit" might be more clear (closer to what it does) ?
Quoted 52 lines
arch/arm64/kvm/pmu-emul.c +9
-0+{ +u64 tmp; + + tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); + tmp =cpuid_feature_cap_perfmon_field (tmp, +ID_AA64DFR0_PMUVER_SHIFT , +ID_AA64DFR0_PMUVER_8_4 ); +return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); +}arch/arm64/kvm/sys_regs.c +20
-4@@ -1062,6 +1062,22 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,return true; } +static u8 pmuver_to_perfmon (const struct kvm_vcpu *vcpu) +{ +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; + +switch (vcpu->kvm->arch.dfr0_pmuver) { +case ID_AA64DFR0_PMUVER_8_0 : +return ID_DFR0_PERFMON_8_0 ; +case ID_AA64DFR0_PMUVER_IMP_DEF : +return 0 ; +default : +/* Anything ARMv8.4+ has the same value. For now. */ +return vcpu->kvm->arch.dfr0_pmuver; + } +} +/* Read a sanitised cpufeature ID register by sys_reg_desc */ static u64 read_id_reg (const struct kvm_vcpu *vcpu,struct sys_reg_desc const *r,bool raz)@@ -1112,10 +1128,10 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit debug to ARMv8.0 */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ); val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ),6 ); -/* Limit guests to PMUv3 for ARMv8.4 */ - val =cpuid_feature_cap_perfmon_field (val, -ID_AA64DFR0_PMUVER_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_AA64DFR0_PMUVER_8_4 :0 ); +/* Set PMUver to the required version */ + val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ); + val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), +kvm_vcpu_has_pmu (vcpu) ? vcpu->kvm->arch.dfr0_pmuver :0 );/* Hide SPE from guests */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMSVER );break ;
arch/arm64/kvm/sys_regs.c static u64 read_id_reg(const struct kvm_vcpu *vcpu, · line 1139
Quoted 1 line
arch/arm64/kvm/sys_regs.c +0
-0@@ -1123,7 +1139,7 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit guests to PMUv3 for ARMv8.4 */
Nit: I think the comment above should be removed like you did for ID_AA64DFR0_EL1 (or move it to kvm_arm_pmu_get_host_pmuver()?).
Thank you,
Reiji
Quoted 38 lines
arch/arm64/kvm/sys_regs.c +1
-1val =cpuid_feature_cap_perfmon_field (val,ID_DFR0_PERFMON_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_DFR0_PERFMON_8_4 :0 ); +pmuver_to_perfmon (vcpu));break ; }include/kvm/arm_pmu.h +6
-1@@ -89,6 +89,8 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ } while (0) +u8 kvm_arm_pmu_get_host_pmuver(void); +#else struct kvm_pmu { };@@ -154,6 +156,10 @@ static inline u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1)static inline void kvm_pmu_update_vcpu_events (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_guest (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_host (struct kvm_vcpu *vcpu) {} +static inline u8 kvm_arm_pmu_get_host_pmuver (void ) +{ +return 0 ; +}#endif --2.34.1
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 6/9] KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
Hi Marc,
Quoted 105 lines · Reiji Watanabe
Hi Marc,
Quoted 51 lines · Marc Zyngier
As further patches will enable the selection of a PMU revision from userspace, sample the supported PMU revision at VM creation time, rather than building each time the
ID_AA64DFR0_EL1register is accessed.This shouldn't result in any change in behaviour.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 ++++++ arch/arm64/kvm/pmu-emul.c | 11 +++++++++++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++----- include/kvm/arm_pmu.h | 6 ++++++ 5 files changed, 45 insertions(+), 5 deletions(-)arch/arm64/include/asm/kvm_host.h +1
-0@@ -163,6 +163,7 @@ struct kvm_arch {u8 pfr0_csv2;u8 pfr0_csv3; +u8 dfr0_pmuver;/* Hypercall features firmware registers' descriptor */ struct kvm_smccc_features smccc_feat;arch/arm64/kvm/arm.c +6
-0@@ -164,6 +164,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)set_default_spectre (kvm);kvm_arm_init_hypercalls (kvm); +/* +* Initialise the default PMUver before there is a chance to +* create an actual PMU. +*/ + kvm->arch.dfr0_pmuver =kvm_arm_pmu_get_host_pmuver (); +return ret; out_free_stage2_pgd:kvm_free_stage2_pgd (&kvm->arch.mmu);arch/arm64/kvm/pmu-emul.c +0
-0arch/arm64/kvm/pmu-emul.c int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1025
Quoted 5 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -1021,3 +1021,14 @@ int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return -ENXIO ; } + +u8 kvm_arm_pmu_get_host_pmuver (void )Nit: Since this function doesn't simply return the host's pmuver, but the pmuver limit for guests, perhaps "
kvm_arm_pmu_get_guest_pmuver_limit" might be more clear (closer to what it does) ?Quoted 40 lines
arch/arm64/kvm/pmu-emul.c +9
-0+{ +u64 tmp; + + tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); + tmp =cpuid_feature_cap_perfmon_field (tmp, +ID_AA64DFR0_PMUVER_SHIFT , +ID_AA64DFR0_PMUVER_8_4 ); +return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); +}arch/arm64/kvm/sys_regs.c +16
-0@@ -1062,6 +1062,22 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,return true; } +static u8 pmuver_to_perfmon (const struct kvm_vcpu *vcpu) +{ +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; + +switch (vcpu->kvm->arch.dfr0_pmuver) { +case ID_AA64DFR0_PMUVER_8_0 : +return ID_DFR0_PERFMON_8_0 ; +case ID_AA64DFR0_PMUVER_IMP_DEF : +return 0 ; +default : +/* Anything ARMv8.4+ has the same value. For now. */ +return vcpu->kvm->arch.dfr0_pmuver; + } +} +/* Read a sanitised cpufeature ID register by sys_reg_desc */ static u64 read_id_reg (const struct kvm_vcpu *vcpu,struct sys_reg_desc const *r,bool raz)@@ -1112,10 +1128,10 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit debug to ARMv8.0 */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ); val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ),6 );
Quoted 8 lines
arch/arm64/kvm/sys_regs.c static u64 read_id_reg(const struct kvm_vcpu *vcpu, · line 1134
Quoted 8 lines
arch/arm64/kvm/sys_regs.c +4
-4-/* Limit guests to PMUv3 for ARMv8.4 */ - val =cpuid_feature_cap_perfmon_field (val, -ID_AA64DFR0_PMUVER_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_AA64DFR0_PMUVER_8_4 :0 ); +/* Set PMUver to the required version */ + val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ); + val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), +kvm_vcpu_has_pmu (vcpu) ? vcpu->kvm->arch.dfr0_pmuver :0 );
I've just noticed one issue in this patch while I'm reviewing patch-7.
I would think that this patch makes PMUVER and PERFMON inconsistent when PMU is not enabled for the vCPU, and the host's sanitised PMUVER is IMP_DEF.
Previously, when PMU is not enabled for the vCPU and the host's sanitized value of PMUVER is IMP_DEF(0xf), the vCPU's PMUVER and PERFMON are set to IMP_DEF due to a bug of cpuid_feature_cap_perfmon_field().
(https://lore.kernel.org/all/[email protected]/)
With this patch, the vCPU's PMUVER will be 0 for the same case, while the vCPU's PERFMON will stay the same (IMP_DEF).
I guess you unintentionally corrected only the PMUVER value of the VCPU.
Thank you,
Reiji
Quoted 54 lines
arch/arm64/kvm/sys_regs.c static u64 read_id_reg(const struct kvm_vcpu *vcpu, · line 1139
Quoted 5 lines
arch/arm64/kvm/sys_regs.c +0
-0/* Hide SPE from guests */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMSVER );break ;@@ -1123,7 +1139,7 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit guests to PMUv3 for ARMv8.4 */ Nit: I think the comment above should be removed like you did for
ID_AA64DFR0_EL1(or move it tokvm_arm_pmu_get_host_pmuver()?).Reviewed-by Reiji Watanabe <[email protected]>Thank you,
ReijiQuoted 38 lines
arch/arm64/kvm/sys_regs.c +1
-1val =cpuid_feature_cap_perfmon_field (val,ID_DFR0_PERFMON_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_DFR0_PERFMON_8_4 :0 ); +pmuver_to_perfmon (vcpu));break ; }include/kvm/arm_pmu.h +6
-1@@ -89,6 +89,8 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ } while (0) +u8 kvm_arm_pmu_get_host_pmuver(void); +#else struct kvm_pmu { };@@ -154,6 +156,10 @@ static inline u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1)static inline void kvm_pmu_update_vcpu_events (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_guest (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_host (struct kvm_vcpu *vcpu) {} +static inline u8 kvm_arm_pmu_get_host_pmuver (void ) +{ +return 0 ; +}#endif --2.34.1
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 6/9] KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
Hi Reiji,
Sorry it took so long to get back to this.
Quoted 68 lines · Reiji Watanabe
Hi Marc,
Quoted 64 lines · Reiji Watanabe
Hi Marc,
Quoted 51 lines · Marc Zyngier
As further patches will enable the selection of a PMU revision from userspace, sample the supported PMU revision at VM creation time, rather than building each time the
ID_AA64DFR0_EL1register is accessed.This shouldn't result in any change in behaviour.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 ++++++ arch/arm64/kvm/pmu-emul.c | 11 +++++++++++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++----- include/kvm/arm_pmu.h | 6 ++++++ 5 files changed, 45 insertions(+), 5 deletions(-)arch/arm64/include/asm/kvm_host.h +1
-0@@ -163,6 +163,7 @@ struct kvm_arch {u8 pfr0_csv2;u8 pfr0_csv3; +u8 dfr0_pmuver;/* Hypercall features firmware registers' descriptor */ struct kvm_smccc_features smccc_feat;arch/arm64/kvm/arm.c +6
-0@@ -164,6 +164,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)set_default_spectre (kvm);kvm_arm_init_hypercalls (kvm); +/* +* Initialise the default PMUver before there is a chance to +* create an actual PMU. +*/ + kvm->arch.dfr0_pmuver =kvm_arm_pmu_get_host_pmuver (); +return ret; out_free_stage2_pgd:kvm_free_stage2_pgd (&kvm->arch.mmu);arch/arm64/kvm/pmu-emul.c +0
-0arch/arm64/kvm/pmu-emul.c int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1025
Quoted 5 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -1021,3 +1021,14 @@ int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return -ENXIO ; } + +u8 kvm_arm_pmu_get_host_pmuver (void )Nit: Since this function doesn't simply return the host's pmuver, but the pmuver limit for guests, perhaps "
kvm_arm_pmu_get_guest_pmuver_limit" might be more clear (closer to what it does) ?
Maybe a bit verbose, but I'll work something out.
Quoted 56 lines
Quoted 41 lines
Quoted 41 lines
arch/arm64/kvm/pmu-emul.c +9
-0+{ +u64 tmp; + + tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); + tmp =cpuid_feature_cap_perfmon_field (tmp, +ID_AA64DFR0_PMUVER_SHIFT , +ID_AA64DFR0_PMUVER_8_4 ); +return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); +}arch/arm64/kvm/sys_regs.c +16
-0@@ -1062,6 +1062,22 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,return true; } +static u8 pmuver_to_perfmon (const struct kvm_vcpu *vcpu) +{ +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; + +switch (vcpu->kvm->arch.dfr0_pmuver) { +case ID_AA64DFR0_PMUVER_8_0 : +return ID_DFR0_PERFMON_8_0 ; +case ID_AA64DFR0_PMUVER_IMP_DEF : +return 0 ; +default : +/* Anything ARMv8.4+ has the same value. For now. */ +return vcpu->kvm->arch.dfr0_pmuver; + } +} +/* Read a sanitised cpufeature ID register by sys_reg_desc */ static u64 read_id_reg (const struct kvm_vcpu *vcpu,struct sys_reg_desc const *r,bool raz)@@ -1112,10 +1128,10 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit debug to ARMv8.0 */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ); val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ),6 );Quoted 8 lines
arch/arm64/kvm/sys_regs.c static u64 read_id_reg(const struct kvm_vcpu *vcpu, · line 1134
Quoted 8 lines
arch/arm64/kvm/sys_regs.c +4
-4-/* Limit guests to PMUv3 for ARMv8.4 */ - val =cpuid_feature_cap_perfmon_field (val, -ID_AA64DFR0_PMUVER_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_AA64DFR0_PMUVER_8_4 :0 ); +/* Set PMUver to the required version */ + val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ); + val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), +kvm_vcpu_has_pmu (vcpu) ? vcpu->kvm->arch.dfr0_pmuver :0 );I've just noticed one issue in this patch while I'm reviewing patch-7.
I would think that this patch makes PMUVER and PERFMON inconsistent when PMU is not enabled for the vCPU, and the host's sanitised PMUVER is
IMP_DEF.
Quoted 8 lines
Previously, when PMU is not enabled for the vCPU and the host's sanitized value of PMUVER is
IMP_DEF(0xf), the vCPU's PMUVER and PERFMON are set toIMP_DEFdue to a bug ofcpuid_feature_cap_perfmon_field().
(https://lore.kernel.org/all/[email protected]/)With this patch, the vCPU's PMUVER will be 0 for the same case, while the vCPU's PERFMON will stay the same (
IMP_DEF).
I guess you unintentionally corrected only the PMUVER value of the VCPU.
I think that with this patch both PMUVer and Perfmon values get set to 0 (pmuver_to_perfmon() returns 0 for both ID_AA64DFR0_PMUVER_IMP_DEF and no PMU at all). Am I missing anything here?
However, you definitely have a point that we should handle a guest being restored with an IMPDEF PMU. Which means I need to revisit this patch and the userspace accessors. Oh well...
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 6/9] KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
Hi Marc,
Quoted 1 line
Sorry it took so long to get back to this.
No problem!
Quoted 143 lines
Quoted 68 lines · Reiji Watanabe
Hi Marc,
Quoted 64 lines · Reiji Watanabe
Hi Marc,
Quoted 51 lines · Marc Zyngier
As further patches will enable the selection of a PMU revision from userspace, sample the supported PMU revision at VM creation time, rather than building each time the
ID_AA64DFR0_EL1register is accessed.This shouldn't result in any change in behaviour.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 ++++++ arch/arm64/kvm/pmu-emul.c | 11 +++++++++++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++----- include/kvm/arm_pmu.h | 6 ++++++ 5 files changed, 45 insertions(+), 5 deletions(-)arch/arm64/include/asm/kvm_host.h +1
-0@@ -163,6 +163,7 @@ struct kvm_arch {u8 pfr0_csv2;u8 pfr0_csv3; +u8 dfr0_pmuver;/* Hypercall features firmware registers' descriptor */ struct kvm_smccc_features smccc_feat;arch/arm64/kvm/arm.c +6
-0@@ -164,6 +164,12 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)set_default_spectre (kvm);kvm_arm_init_hypercalls (kvm); +/* +* Initialise the default PMUver before there is a chance to +* create an actual PMU. +*/ + kvm->arch.dfr0_pmuver =kvm_arm_pmu_get_host_pmuver (); +return ret; out_free_stage2_pgd:kvm_free_stage2_pgd (&kvm->arch.mmu);arch/arm64/kvm/pmu-emul.c +0
-0arch/arm64/kvm/pmu-emul.c int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1025
Quoted 5 lines
arch/arm64/kvm/pmu-emul.c +2
-0@@ -1021,3 +1021,14 @@ int kvm_arm_pmu_v3_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return -ENXIO ; } + +u8 kvm_arm_pmu_get_host_pmuver (void )Nit: Since this function doesn't simply return the host's pmuver, but the pmuver limit for guests, perhaps "
kvm_arm_pmu_get_guest_pmuver_limit" might be more clear (closer to what it does) ?Maybe a bit verbose, but I'll work something out.
Quoted 56 lines
Quoted 41 lines
Quoted 41 lines
arch/arm64/kvm/pmu-emul.c +9
-0+{ +u64 tmp; + + tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); + tmp =cpuid_feature_cap_perfmon_field (tmp, +ID_AA64DFR0_PMUVER_SHIFT , +ID_AA64DFR0_PMUVER_8_4 ); +return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); +}arch/arm64/kvm/sys_regs.c +16
-0@@ -1062,6 +1062,22 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,return true; } +static u8 pmuver_to_perfmon (const struct kvm_vcpu *vcpu) +{ +if (!kvm_vcpu_has_pmu (vcpu)) +return 0 ; + +switch (vcpu->kvm->arch.dfr0_pmuver) { +case ID_AA64DFR0_PMUVER_8_0 : +return ID_DFR0_PERFMON_8_0 ; +case ID_AA64DFR0_PMUVER_IMP_DEF : +return 0 ; +default : +/* Anything ARMv8.4+ has the same value. For now. */ +return vcpu->kvm->arch.dfr0_pmuver; + } +} +/* Read a sanitised cpufeature ID register by sys_reg_desc */ static u64 read_id_reg (const struct kvm_vcpu *vcpu,struct sys_reg_desc const *r,bool raz)@@ -1112,10 +1128,10 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu,/* Limit debug to ARMv8.0 */ val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ); val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_DEBUGVER ),6 );Quoted 8 lines
arch/arm64/kvm/sys_regs.c static u64 read_id_reg(const struct kvm_vcpu *vcpu, · line 1134
Quoted 8 lines
arch/arm64/kvm/sys_regs.c +4
-4-/* Limit guests to PMUv3 for ARMv8.4 */ - val =cpuid_feature_cap_perfmon_field (val, -ID_AA64DFR0_PMUVER_SHIFT , -kvm_vcpu_has_pmu (vcpu) ?ID_AA64DFR0_PMUVER_8_4 :0 ); +/* Set PMUver to the required version */ + val &= ~ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ); + val |=FIELD_PREP (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), +kvm_vcpu_has_pmu (vcpu) ? vcpu->kvm->arch.dfr0_pmuver :0 );I've just noticed one issue in this patch while I'm reviewing patch-7.
I would think that this patch makes PMUVER and PERFMON inconsistent when PMU is not enabled for the vCPU, and the host's sanitised PMUVER is
IMP_DEF.Quoted 8 lines
Previously, when PMU is not enabled for the vCPU and the host's sanitized value of PMUVER is
IMP_DEF(0xf), the vCPU's PMUVER and PERFMON are set toIMP_DEFdue to a bug ofcpuid_feature_cap_perfmon_field().
(https://lore.kernel.org/all/[email protected]/)With this patch, the vCPU's PMUVER will be 0 for the same case, while the vCPU's PERFMON will stay the same (
IMP_DEF).
I guess you unintentionally corrected only the PMUVER value of the VCPU.I think that with this patch both PMUVer and Perfmon values get set to 0 (
pmuver_to_perfmon()returns 0 for bothID_AA64DFR0_PMUVER_IMP_DEFand no PMU at all). Am I missing anything here?
Quoted 7 lines
Quoted 3 lines
With this patch, the vCPU's PMUVER will be 0 for the same case, while the vCPU's PERFMON will stay the same (
IMP_DEF).
I guess you unintentionally corrected only the PMUVER value of the VCPU.I think that with this patch both PMUVer and Perfmon values get set to 0 (
pmuver_to_perfmon()returns 0 for bothID_AA64DFR0_PMUVER_IMP_DEFand no PMU at all). Am I missing anything here?
When pmuver_to_perfmon() returns 0 for ID_AA64DFR0_PMUVER_IMP_DEF, cpuid_feature_cap_perfmon_field() is called with 'cap' == 0. Then, the code in cpuid_feature_cap_perfmon_field() updates the 'val' with 0 if the given 'features' (sanitized) value is ID_AA64DFR0_PMUVER_IMP_DEF. So, now the val(== 0) is not larger than the cap (== 0), and cpuid_feature_cap_perfmon_field() ends up returning the given 'features' value as it is without updating the PERFMON field.
Or am I missing anything here?
Thank you,
Reiji
Quoted 11 lines
However, you definitely have a point that we should handle a guest being restored with an IMPDEF PMU. Which means I need to revisit this patch and the userspace accessors. Oh well...
Thanks,
M.--
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH 6/9] KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
Quoted 11 lines · Reiji Watanabe
Quoted 3 lines
I think that with this patch both PMUVer and Perfmon values get set to 0 (
pmuver_to_perfmon()returns 0 for bothID_AA64DFR0_PMUVER_IMP_DEFand no PMU at all). Am I missing anything here?When
pmuver_to_perfmon()returns 0 forID_AA64DFR0_PMUVER_IMP_DEF,cpuid_feature_cap_perfmon_field()is called with 'cap' == 0. Then, the code incpuid_feature_cap_perfmon_field()updates the 'val' with 0 if the given 'features' (sanitized) value isID_AA64DFR0_PMUVER_IMP_DEF. So, now the val(== 0) is not larger than the cap (== 0), andcpuid_feature_cap_perfmon_field()ends up returning the given 'features' value as it is without updating the PERFMON field.
Ah, thanks for spelling it out for me, I was definitely looking at the wrong side of things. You're absolutely right. The code I have now makes sure to:
(1) preserve the IMP_DEF view of the PMU if userspace provides such setting
(2) directly places the emulated PMU revision in the feature set without calling cpuid_feature_cap_perfmon_field(), which indeed does the wrong thing.
Hopefully I got it right this time! ;-)
Thanks again,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 7/9 KVM: arm64: PMU: Allow ID_AA64DFR0_EL1.PMUver to be set from userspace
In reply to Marc ZyngierAllow userspace to write ID_AA64DFR0_EL1, on the condition that only the PMUver field can be altered and be at most the one that was initially computed for the guest.
--- arch/arm64/kvm/sys_regs.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)
arch/arm64/kvm/sys_regs.c +34 -1
@@ -1236,6 +1236,38 @@ static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,return 0 ; } +static int set_id_aa64dfr0_el1 (struct kvm_vcpu *vcpu, +const struct sys_reg_desc *rd, +u64 val) +{ +u8 pmuver, host_pmuver; + + host_pmuver =kvm_arm_pmu_get_host_pmuver (); + +/* +* Allow AA64DFR0_EL1.PMUver to be set from userspace as long +* as it doesn't promise more than what the HW gives us. We +* don't allow an IMPDEF PMU though. +*/ + pmuver =FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), val); +if (pmuver ==ID_AA64DFR0_PMUVER_IMP_DEF || pmuver > host_pmuver) +return -EINVAL ; + +/* We already have a PMU, don't try to disable it... */ +if (kvm_vcpu_has_pmu (vcpu) && pmuver ==0 ) +return -EINVAL ; + +/* We can only differ with PMUver, and anything else is an error */ + val ^=read_id_reg (vcpu, rd, false); + val &= ~(0xFUL <<ID_AA64DFR0_PMUVER_SHIFT ); +if (val) +return -EINVAL ; + + vcpu->kvm->arch.dfr0_pmuver = pmuver; + +return 0 ; +} +/* * cpufeature ID register user accessors *
@@ -1510,7 +1542,8 @@ static const struct sys_reg_desc sys_reg_descs[] = {ID_UNALLOCATED (4 ,7 ),/* CRm=5 */ -ID_SANITISED (ID_AA64DFR0_EL1 ), + { SYS_DESC(SYS_ID_AA64DFR0_EL1), .access = access_id_reg, + .get_user = get_id_reg, .set_user = set_id_aa64dfr0_el1, },ID_SANITISED (ID_AA64DFR1_EL1 ),ID_UNALLOCATED (5 ,2 ), ID_UNALLOCATED(5,3),
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 7/9] KVM: arm64: PMU: Allow ID_AA64DFR0_EL1.PMUver to be set from userspace
Hi Marc,
Quoted 3 lines · Marc Zyngier
Allow userspace to write
ID_AA64DFR0_EL1, on the condition that only the PMUver field can be altered and be at most the one that was initially computed for the guest.
As DFR0_EL1 is exposed to userspace, isn't a ->set_user() hook required for it as well?
Quoted 27 lines
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/sys_regs.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +14
-0@@ -1236,6 +1236,38 @@ static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,return 0 ; } +static int set_id_aa64dfr0_el1 (struct kvm_vcpu *vcpu, +const struct sys_reg_desc *rd, +u64 val) +{ +u8 pmuver, host_pmuver; + + host_pmuver =kvm_arm_pmu_get_host_pmuver (); + +/* +* Allow AA64DFR0_EL1.PMUver to be set from userspace as long +* as it doesn't promise more than what the HW gives us. We +* don't allow an IMPDEF PMU though. +*/ + pmuver =FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), val);
arch/arm64/kvm/sys_regs.c static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu, · line 1262
Quoted 10 lines
arch/arm64/kvm/sys_regs.c +10
-0+if (pmuver ==ID_AA64DFR0_PMUVER_IMP_DEF || pmuver > host_pmuver) +return -EINVAL ; + +/* We already have a PMU, don't try to disable it... */ +if (kvm_vcpu_has_pmu (vcpu) && pmuver ==0 ) +return -EINVAL ; + +/* We can only differ with PMUver, and anything else is an error */ + val ^=read_id_reg (vcpu, rd, false); + val &= ~(0xFUL <<ID_AA64DFR0_PMUVER_SHIFT );
nit: ~ARM64_FEATURE_MASK(ID_AA64DFR0_PMUVER)
--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH 7/9] KVM: arm64: PMU: Allow ID_AA64DFR0_EL1.PMUver to be set from userspace
Quoted 10 lines · Oliver Upton
Hi Marc,
Quoted 3 lines · Marc Zyngier
Allow userspace to write
ID_AA64DFR0_EL1, on the condition that only the PMUver field can be altered and be at most the one that was initially computed for the guest.As
DFR0_EL1is exposed to userspace, isn't a ->set_user() hook required for it as well?
Hmm. Yes, absolutely. Which is really annoying. It also pushed me to have a look at what PMUv3p5 means for AArch32, and it is utter nonsense...
Here's what the spec says about PMEVCNTR<n>:
<quote>
If FEAT_PMUv3p5 is implemented, the event counter is 64 bits and only the least-significant part of the event counter is accessible in AArch32 state:
- Reads from PMEVCNTR<n> return bits [31:0] of the counter.
- Writes to PMEVCNTR<n> update bits [31:0] and leave bits [63:32] unchanged.
- There is no means to access bits [63:32] directly from AArch32 state </quote>
But PMCR.LP does exist! You just can't make any reasonable use of it. Yet another reason to want AArch32 dead.
Quoted 32 lines
Quoted 32 lines
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/sys_regs.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +18
-0@@ -1236,6 +1236,38 @@ static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,return 0 ; } +static int set_id_aa64dfr0_el1 (struct kvm_vcpu *vcpu, +const struct sys_reg_desc *rd, +u64 val) +{ +u8 pmuver, host_pmuver; + + host_pmuver =kvm_arm_pmu_get_host_pmuver (); + +/* +* Allow AA64DFR0_EL1.PMUver to be set from userspace as long +* as it doesn't promise more than what the HW gives us. We +* don't allow an IMPDEF PMU though. +*/ + pmuver =FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), val); +if (pmuver ==ID_AA64DFR0_PMUVER_IMP_DEF || pmuver > host_pmuver) +return -EINVAL ; + +/* We already have a PMU, don't try to disable it... */
Quoted 8 lines
arch/arm64/kvm/sys_regs.c static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu, · line 1262
Quoted 6 lines
arch/arm64/kvm/sys_regs.c +6
-0+if (kvm_vcpu_has_pmu (vcpu) && pmuver ==0 ) +return -EINVAL ; + +/* We can only differ with PMUver, and anything else is an error */ + val ^=read_id_reg (vcpu, rd, false); + val &= ~(0xFUL <<ID_AA64DFR0_PMUVER_SHIFT );nit: ~ARM64_FEATURE_MASK(ID_AA64DFR0_PMUVER)
Good point.
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 7/9] KVM: arm64: PMU: Allow ID_AA64DFR0_EL1.PMUver to be set from userspace
In reply to Marc ZyngierHi Marc,
Quoted 15 lines · Marc Zyngier
Allow userspace to write
ID_AA64DFR0_EL1, on the condition that only the PMUver field can be altered and be at most the one that was initially computed for the guest.Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/sys_regs.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-)arch/arm64/kvm/sys_regs.c +0
-0
arch/arm64/kvm/sys_regs.c static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu, · line 1241
Quoted 6 lines
arch/arm64/kvm/sys_regs.c +3
-0@@ -1236,6 +1236,38 @@ static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,return 0 ; } +static int set_id_aa64dfr0_el1 (struct kvm_vcpu *vcpu, +const struct sys_reg_desc *rd, +u64 val)
The function prototype doesn't appear to be right as theset_user of sys_reg_desc().
---
[From sys_regs.h]
[sys_regs.h]
int (*set_user)(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
const struct kvm_one_reg *reg, void __user *uaddr);
---
arch/arm64/kvm/sys_regs.c static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu, · line 1254
Quoted 13 lines
arch/arm64/kvm/sys_regs.c +13
-0+{ +u8 pmuver, host_pmuver; + + host_pmuver =kvm_arm_pmu_get_host_pmuver (); + +/* +* Allow AA64DFR0_EL1.PMUver to be set from userspace as long +* as it doesn't promise more than what the HW gives us. We +* don't allow an IMPDEF PMU though. +*/ + pmuver =FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), val); +if (pmuver ==ID_AA64DFR0_PMUVER_IMP_DEF || pmuver > host_pmuver) +return -EINVAL ;
As mentioned in my comments for the patch-6, the vCPU's PMUVER could currently be IMP_DEF. So, with this IMP_DEF checking, a guest with IMP_DEF PMU cannot be migrated to a newer KVM host.
Do we need to tolerate writes of IMP_DEF for compatibility ?
Oliver originally point this out for my ID register series, and my V6 or newer series tried to not return an error for this by ignoring the user requested IMP_DEF when PMU is not enabled for the vCPU (Instead, the field is set to 0x0).
https://lore.kernel.org/all/[email protected]/
Thank you,
Reiji
Quoted 32 lines
arch/arm64/kvm/sys_regs.c +18
-2+ +/* We already have a PMU, don't try to disable it... */ +if (kvm_vcpu_has_pmu (vcpu) && pmuver ==0 ) +return -EINVAL ; + +/* We can only differ with PMUver, and anything else is an error */ + val ^=read_id_reg (vcpu, rd, false); + val &= ~(0xFUL <<ID_AA64DFR0_PMUVER_SHIFT ); +if (val) +return -EINVAL ; + + vcpu->kvm->arch.dfr0_pmuver = pmuver; + +return 0 ; +} +/* * cpufeature ID register user accessors * @@ -1510,7 +1542,8 @@ static const struct sys_reg_desc sys_reg_descs[] = {ID_UNALLOCATED (4 ,7 ),/* CRm=5 */ -ID_SANITISED (ID_AA64DFR0_EL1 ), + { SYS_DESC(SYS_ID_AA64DFR0_EL1), .access = access_id_reg, + .get_user = get_id_reg, .set_user = set_id_aa64dfr0_el1, },ID_SANITISED (ID_AA64DFR1_EL1 ),ID_UNALLOCATED (5 ,2 ), ID_UNALLOCATED(5,3), --2.34.1
Quoted 4 lines
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 8/9 KVM: arm64: PMU: Implement PMUv3p5 long counter support
In reply to Marc ZyngierPMUv3p5 (which is mandatory with ARMv8.5) comes with some extra features:
- All counters are 64bit
- The overflow point is controlled by the PMCR_EL0.LP bit
Add the required checks in the helpers that control counter width and overflow, as well as the sysreg handling for the LP bit. A new kvm_pmu_is_3p5() helper makes it easy to spot the PMUv3p5 specific handling.
--- arch/arm64/kvm/pmu-emul.c | 8 +++++--- arch/arm64/kvm/sys_regs.c | 4 ++++ include/kvm/arm_pmu.h | 8 ++++++++ 3 files changed, 17 insertions(+), 3 deletions(-)
arch/arm64/kvm/pmu-emul.c +5 -3
@@ -50,13 +50,15 @@ static u32 kvm_pmu_event_mask(struct kvm *kvm)*/ static bool kvm_pmu_idx_is_64bit (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX ); +return (select_idx ==ARMV8_PMU_CYCLE_IDX ||kvm_pmu_is_3p5 (vcpu)); }static bool kvm_pmu_idx_has_64bit_overflow (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX && -__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_LC ); +u64 val =__vcpu_sys_reg (vcpu,PMCR_EL0 ); + +return (select_idx <ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LP )) || + (select_idx ==ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LC )); }static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx)
arch/arm64/kvm/sys_regs.c +4 -0
@@ -654,6 +654,8 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r) | (ARMV8_PMU_PMCR_MASK & 0xdecafbad)) & (~ARMV8_PMU_PMCR_E);if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu, r->reg) = val; }
@@ -703,6 +705,8 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, val |= p->regval &ARMV8_PMU_PMCR_MASK ;if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu,PMCR_EL0 ) = val;kvm_pmu_handle_pmcr (vcpu, val);kvm_vcpu_pmu_restore_guest (vcpu);
include/kvm/arm_pmu.h +8 -0
@@ -89,6 +89,13 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ } while (0) +/* +* Evaluates as true when emulating PMUv3p5, and false otherwise. + */ +#define kvm_pmu_is_3p5 (vcpu) \ + (vcpu->kvm->arch.dfr0_pmuver >= ID_AA64DFR0_PMUVER_8_5 && \ + vcpu->kvm->arch.dfr0_pmuver != ID_AA64DFR0_PMUVER_IMP_DEF) +u8 kvm_arm_pmu_get_host_pmuver (void );#else
@@ -153,6 +160,7 @@ static inline u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1) }#define kvm_vcpu_has_pmu (vcpu) ({ false; }) +#define kvm_pmu_is_3p5 (vcpu) ({ false; })static inline void kvm_pmu_update_vcpu_events (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_guest (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_host (struct kvm_vcpu *vcpu) {}
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 8/9] KVM: arm64: PMU: Implement PMUv3p5 long counter support
Hi Marc,
Quoted 69 lines · Marc Zyngier
PMUv3p5 (which is mandatory with ARMv8.5) comes with some extra features:
- All counters are 64bit
- The overflow point is controlled by the PMCR_EL0.LP bit
Add the required checks in the helpers that control counter width and overflow, as well as the sysreg handling for the LP bit. A new
kvm_pmu_is_3p5()helper makes it easy to spot the PMUv3p5 specific handling.Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 8 +++++--- arch/arm64/kvm/sys_regs.c | 4 ++++ include/kvm/arm_pmu.h | 8 ++++++++ 3 files changed, 17 insertions(+), 3 deletions(-)arch/arm64/kvm/pmu-emul.c +5
-3@@ -50,13 +50,15 @@ static u32 kvm_pmu_event_mask(struct kvm *kvm)*/ static bool kvm_pmu_idx_is_64bit (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX ); +return (select_idx ==ARMV8_PMU_CYCLE_IDX ||kvm_pmu_is_3p5 (vcpu)); }static bool kvm_pmu_idx_has_64bit_overflow (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX && -__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_LC ); +u64 val =__vcpu_sys_reg (vcpu,PMCR_EL0 ); + +return (select_idx <ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LP )) || + (select_idx ==ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LC )); }static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx)arch/arm64/kvm/sys_regs.c +4
-0@@ -654,6 +654,8 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r) | (ARMV8_PMU_PMCR_MASK & 0xdecafbad)) & (~ARMV8_PMU_PMCR_E);if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu, r->reg) = val; }@@ -703,6 +705,8 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, val |= p->regval &ARMV8_PMU_PMCR_MASK ;if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu,PMCR_EL0 ) = val;kvm_pmu_handle_pmcr (vcpu, val);kvm_vcpu_pmu_restore_guest (vcpu);include/kvm/arm_pmu.h +0
-0
include/kvm/arm_pmu.h void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); · line 97
Quoted 9 lines
include/kvm/arm_pmu.h +6
-0@@ -89,6 +89,13 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ }while (0 ) +/* +* Evaluates as true when emulating PMUv3p5, and false otherwise. +*/ +#define kvm_pmu_is_3p5(vcpu) \ + (vcpu->kvm->arch.dfr0_pmuver >=ID_AA64DFR0_PMUVER_8_5 && \ + vcpu->kvm->arch.dfr0_pmuver !=ID_AA64DFR0_PMUVER_IMP_DEF )
I don't believe the IMP_DEF condition will ever evaluate to false as dfr0_pmuver is sanitized at initialization and writes from userspace.
Quoted 14 lines
include/kvm/arm_pmu.h +1
-0u8 kvm_arm_pmu_get_host_pmuver (void );#else @@ -153,6 +160,7 @@ static inline u64 kvm_pmu_get_pmceid(struct kvm_vcpu *vcpu, bool pmceid1) }#define kvm_vcpu_has_pmu (vcpu) ({ false; }) +#define kvm_pmu_is_3p5 (vcpu) ({ false; })static inline void kvm_pmu_update_vcpu_events (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_guest (struct kvm_vcpu *vcpu) {}static inline void kvm_vcpu_pmu_restore_host (struct kvm_vcpu *vcpu) {}Signature · 1 line
2.34.1
Marc Zyngier
maintainerRe: [PATCH 8/9] KVM: arm64: PMU: Implement PMUv3p5 long counter support
Quoted 76 lines · Oliver Upton
Hi Marc,
Quoted 72 lines · Marc Zyngier
PMUv3p5 (which is mandatory with ARMv8.5) comes with some extra features:
- All counters are 64bit
- The overflow point is controlled by the PMCR_EL0.LP bit
Add the required checks in the helpers that control counter width and overflow, as well as the sysreg handling for the LP bit. A new
kvm_pmu_is_3p5()helper makes it easy to spot the PMUv3p5 specific handling.Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 8 +++++--- arch/arm64/kvm/sys_regs.c | 4 ++++ include/kvm/arm_pmu.h | 8 ++++++++ 3 files changed, 17 insertions(+), 3 deletions(-)arch/arm64/kvm/pmu-emul.c +5
-3@@ -50,13 +50,15 @@ static u32 kvm_pmu_event_mask(struct kvm *kvm)*/ static bool kvm_pmu_idx_is_64bit (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX ); +return (select_idx ==ARMV8_PMU_CYCLE_IDX ||kvm_pmu_is_3p5 (vcpu)); }static bool kvm_pmu_idx_has_64bit_overflow (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX && -__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_LC ); +u64 val =__vcpu_sys_reg (vcpu,PMCR_EL0 ); + +return (select_idx <ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LP )) || + (select_idx ==ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LC )); }static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx)arch/arm64/kvm/sys_regs.c +4
-0@@ -654,6 +654,8 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r) | (ARMV8_PMU_PMCR_MASK & 0xdecafbad)) & (~ARMV8_PMU_PMCR_E);if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu, r->reg) = val; }@@ -703,6 +705,8 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, val |= p->regval &ARMV8_PMU_PMCR_MASK ;if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu,PMCR_EL0 ) = val;kvm_pmu_handle_pmcr (vcpu, val);kvm_vcpu_pmu_restore_guest (vcpu);include/kvm/arm_pmu.h +0
-0@@ -89,6 +89,13 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ }while (0 )
Quoted 9 lines
include/kvm/arm_pmu.h void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); · line 97
Quoted 6 lines
include/kvm/arm_pmu.h +6
-0+/* +* Evaluates as true when emulating PMUv3p5, and false otherwise. +*/ +#define kvm_pmu_is_3p5 (vcpu) \ + (vcpu->kvm->arch.dfr0_pmuver >= ID_AA64DFR0_PMUVER_8_5 && \ + vcpu->kvm->arch.dfr0_pmuver != ID_AA64DFR0_PMUVER_IMP_DEF)I don't believe the
IMP_DEFcondition will ever evaluate to false asdfr0_pmuveris sanitized at initialization and writes from userspace.
Good point. That's a leftover from a previous version. I'll fix that.
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 8/9] KVM: arm64: PMU: Implement PMUv3p5 long counter support
Quoted 48 lines · Marc Zyngier
Quoted 45 lines · Oliver Upton
Hi Marc,
Quoted 41 lines · Marc Zyngier
PMUv3p5 (which is mandatory with ARMv8.5) comes with some extra features:
- All counters are 64bit
- The overflow point is controlled by the PMCR_EL0.LP bit
Add the required checks in the helpers that control counter width and overflow, as well as the sysreg handling for the LP bit. A new
kvm_pmu_is_3p5()helper makes it easy to spot the PMUv3p5 specific handling.Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 8 +++++--- arch/arm64/kvm/sys_regs.c | 4 ++++ include/kvm/arm_pmu.h | 8 ++++++++ 3 files changed, 17 insertions(+), 3 deletions(-)arch/arm64/kvm/pmu-emul.c +5
-3@@ -50,13 +50,15 @@ static u32 kvm_pmu_event_mask(struct kvm *kvm)*/ static bool kvm_pmu_idx_is_64bit (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX ); +return (select_idx ==ARMV8_PMU_CYCLE_IDX ||kvm_pmu_is_3p5 (vcpu)); }static bool kvm_pmu_idx_has_64bit_overflow (struct kvm_vcpu *vcpu,u64 select_idx) { -return (select_idx ==ARMV8_PMU_CYCLE_IDX && -__vcpu_sys_reg (vcpu,PMCR_EL0 ) &ARMV8_PMU_PMCR_LC ); +u64 val =__vcpu_sys_reg (vcpu,PMCR_EL0 ); + +return (select_idx <ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LP )) || + (select_idx ==ARMV8_PMU_CYCLE_IDX && (val &ARMV8_PMU_PMCR_LC )); }
Quoted 7 lines
Quoted 7 lines
arch/arm64/kvm/sys_regs.c static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r) · line 654
Quoted 7 lines
arch/arm64/kvm/pmu-emul.c +0
-0static bool kvm_pmu_counter_can_chain (struct kvm_vcpu *vcpu,u64 idx)arch/arm64/kvm/sys_regs.c +0
-0@@ -654,6 +654,8 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r) | (ARMV8_PMU_PMCR_MASK & 0xdecafbad)) & (~ARMV8_PMU_PMCR_E);
Not directly related to this series, but using 0xdecafbad above appears to be odd. I think that would lead the bit 3 and 5 to be unconditionally set in the register's reset value that the guest will initially see even on the configuration where those should be RES0.
Quoted 34 lines
Quoted 24 lines
Quoted 24 lines
arch/arm64/kvm/sys_regs.c +4
-0if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu, r->reg) = val; }@@ -703,6 +705,8 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p, val |= p->regval &ARMV8_PMU_PMCR_MASK ;if (!system_supports_32bit_el0 ()) val |=ARMV8_PMU_PMCR_LC ; +if (!kvm_pmu_is_3p5 (vcpu)) + val &= ~ARMV8_PMU_PMCR_LP ;__vcpu_sys_reg (vcpu,PMCR_EL0 ) = val;kvm_pmu_handle_pmcr (vcpu, val);kvm_vcpu_pmu_restore_guest (vcpu);include/kvm/arm_pmu.h +0
-0@@ -89,6 +89,13 @@ void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); vcpu->arch.pmu.events = *kvm_get_pmu_events (); \ }while (0 )Quoted 9 lines
include/kvm/arm_pmu.h void kvm_vcpu_pmu_restore_host(struct kvm_vcpu *vcpu); · line 97
Quoted 6 lines
include/kvm/arm_pmu.h +6
-0+/* +* Evaluates as true when emulating PMUv3p5, and false otherwise. +*/ +#define kvm_pmu_is_3p5 (vcpu) \ + (vcpu->kvm->arch.dfr0_pmuver >= ID_AA64DFR0_PMUVER_8_5 && \ + vcpu->kvm->arch.dfr0_pmuver != ID_AA64DFR0_PMUVER_IMP_DEF)I don't believe the
IMP_DEFcondition will ever evaluate to false asdfr0_pmuveris sanitized at initialization and writes from userspace.
Quoted 1 line
Good point. That's a leftover from a previous version. I'll fix that.
With the current series, I think the dfr0_pmuver could be IMP_DEF due to the same bug that I mentioned for the patch-6.
(https://lore.kernel.org/all/[email protected]/)
Thank you,
Reiji
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerPATCH 9/9 KVM: arm64: PMU: Allow PMUv3p5 to be exposed to the guest
In reply to Marc ZyngierNow that the infrastructure is in place, bummp the PMU support up to PMUv3p5.
--- arch/arm64/kvm/pmu-emul.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kvm/pmu-emul.c +1 -1
@@ -1031,6 +1031,6 @@ u8 kvm_arm_pmu_get_host_pmuver(void) tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); tmp =cpuid_feature_cap_perfmon_field (tmp,ID_AA64DFR0_PMUVER_SHIFT , -ID_AA64DFR0_PMUVER_8_4 ); +ID_AA64DFR0_PMUVER_8_5 );return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); }
Signature · 6 lines
2.34.1 _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH 9/9] KVM: arm64: PMU: Allow PMUv3p5 to be exposed to the guest
Quoted 1 line · Marc Zyngier
Now that the infrastructure is in place, bummp the PMU support up
typo: bump
Quoted 22 lines
to PMUv3p5.
Signed-off-by Marc Zyngier <[email protected]>--- arch/arm64/kvm/pmu-emul.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)arch/arm64/kvm/pmu-emul.c +1
-1@@ -1031,6 +1031,6 @@ u8 kvm_arm_pmu_get_host_pmuver(void) tmp =read_sanitised_ftr_reg (SYS_ID_AA64DFR0_EL1 ); tmp =cpuid_feature_cap_perfmon_field (tmp,ID_AA64DFR0_PMUVER_SHIFT , -ID_AA64DFR0_PMUVER_8_4 ); +ID_AA64DFR0_PMUVER_8_5 );return FIELD_GET (ARM64_FEATURE_MASK (ID_AA64DFR0_PMUVER ), tmp); }Signature · 1 line
2.34.1
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 24 lines · Marc Zyngier
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Aleven ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Aleven PMUv3p5 to be exposed to the guest
Quoted 10 lines
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)Signature · 1 line
2.34.1
Hi Marc,
There is one extra potential issue with exposing PMUv3p5. I see this weird behavior when doing passthrough ("bare metal") on the fast-model configured to emulate PMUv3p5: the [63:32] half of the counters overflowing at 32-bits is still incremented.
Fast model - ARMv8.5:
Assuming the initial state is even=0xFFFFFFFF and odd=0x0, incrementing the even counter leads to: 0x00000001_00000000 0x00000000_00000001 0x1 even counter odd counter PMOVSET Assuming the initial state is even=0xFFFFFFFF and odd=0xFFFFFFFF, incrementing the even counter leads to: 0x00000001_00000000 0x00000001_00000000 0x3 even counter odd counter PMOVSET
More specifically, the pmu-chained-sw-incr kvm-unit-test fails when doing passthrough of PMUv3p5 (fast model - ARMv8.5):
INFO: PMU version: 0x5
INFO: PMU implementer/ID code: 0x41("A")/0
INFO: Implements 8 event counters
PASS: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x1, #0=4294967380 #1=1
^^^^^^^^^^^^^
#0=0x00000001_00000054
#1=0x00000000_00000001
FAIL: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x3, #0=4294967380 #1=4294967296
^^^^^^^^^^^^^^^^^^^^^^^^^^^
#0=0x00000001_00000054
#1=0x00000001_00000000There's really no good use for this behavior, and not sure if it's worth emulating it. Can't find any reference in the ARM ARM.
Thanks,
Ricardo
Hi Ricardo,
Quoted 45 lines · Ricardo Koller
Quoted 24 lines · Marc Zyngier
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Aleven ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Aleven PMUv3p5 to be exposed to the guestQuoted 10 lines
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)Signature · 1 line
2.34.1Hi Marc,
There is one extra potential issue with exposing PMUv3p5. I see this weird behavior when doing passthrough ("bare metal") on the fast-model configured to emulate PMUv3p5: the [63:32] half of the counters overflowing at 32-bits is still incremented.
Fast model - ARMv8.5:
Quoted 11 lines
Assuming the initial state is even=0xFFFFFFFF and odd=0x0, incrementing the even counter leads to: 0x00000001_00000000 0x00000000_00000001 0x1 even counter odd counter PMOVSET Assuming the initial state is even=0xFFFFFFFF and odd=0xFFFFFFFF, incrementing the even counter leads to: 0x00000001_00000000 0x00000001_00000000 0x3 even counter odd counter PMOVSET
This is to be expected, actually. PMUv8p5 counters are always 64 bit, regardless of the configured overflow.
DDI 0487H D8.3 Behavior on overflow
If FEAT_PMUv3p5 is implemented, 64-bit event counters are implemented, HDCR.HPMN is not 0, and either n is in the range [0 .. (HDCR.HPMN-1)] or EL2 is not implemented, then event counter overflow is configured by PMCR.LP:
— When PMCR.LP is set to 0, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [31:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
— When PMCR.LP is set to 1, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [63:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
[...]
For all 64-bit counters, incrementing the counter is the same whether an unsigned overflow occurs at [31:0] or [63:0]. If the counter increments for an event, bits [63:0] are always incremented.
Do you see this same (expected) failure w/ Marc's series?
--
Thanks,
Oliver
Quoted 75 lines · Oliver Upton
Hi Ricardo,
Quoted 45 lines · Ricardo Koller
Quoted 24 lines · Marc Zyngier
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Aleven ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Aleven PMUv3p5 to be exposed to the guestQuoted 10 lines
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)Signature · 1 line
2.34.1Hi Marc,
There is one extra potential issue with exposing PMUv3p5. I see this weird behavior when doing passthrough ("bare metal") on the fast-model configured to emulate PMUv3p5: the [63:32] half of the counters overflowing at 32-bits is still incremented.
Fast model - ARMv8.5:
Quoted 11 lines
Assuming the initial state is even=0xFFFFFFFF and odd=0x0, incrementing the even counter leads to: 0x00000001_00000000 0x00000000_00000001 0x1 even counter odd counter PMOVSET Assuming the initial state is even=0xFFFFFFFF and odd=0xFFFFFFFF, incrementing the even counter leads to: 0x00000001_00000000 0x00000001_00000000 0x3 even counter odd counter PMOVSETThis is to be expected, actually. PMUv8p5 counters are always 64 bit, regardless of the configured overflow.
DDI 0487H D8.3 Behavior on overflow
If
FEAT_PMUv3p5is implemented, 64-bit event counters are implemented, HDCR.HPMN is not 0, and either n is in the range [0 .. (HDCR.HPMN-1)] or EL2 is not implemented, then event counter overflow is configured by PMCR.LP:— When PMCR.LP is set to 0, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [31:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
— When PMCR.LP is set to 1, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [63:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
Quoted 7 lines
[...]
For all 64-bit counters, incrementing the counter is the same whether an unsigned overflow occurs at [31:0] or [63:0]. If the counter increments for an event, bits [63:0] are always incremented.
Do you see this same (expected) failure w/ Marc's series?
I don't know, I'm hitting another bug it seems.
Just realized that KVM does not offer PMUv3p5 (with this series applied) when the real hardware is only Armv8.2 (the setup I originally tried). So, tried these other two setups on the fast model:
has_arm_v8-5=1
# ./lkvm-static run --nodefaults --pmu pmu.flat -p pmu-chained-sw-incr # lkvm run -k pmu.flat -m 704 -c 8 --name guest-135
INFO: PMU version: 0x6
^^^
PMUv3 for Armv8.5
INFO: PMU implementer/ID code: 0x41("A")/0
INFO: Implements 8 event counters
FAIL: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=0
^^^
no overflows
FAIL: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=84 #1=-1
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=4294967295
SUMMARY: 2 tests, 2 unexpected failureshas_arm_v8-2=1
# ./lkvm-static run --nodefaults --pmu pmu.flat -p pmu-chained-sw-incr # lkvm run -k pmu.flat -m 704 -c 8 --name guest-134
INFO: PMU version: 0x4
^^^
PMUv3 for Armv8.1
INFO: PMU implementer/ID code: 0x41("A")/0
INFO: Implements 8 event counters
PASS: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x1, #0=84 #1=1
PASS: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x3, #0=84 #1=0
SUMMARY: 2 testsThe Armv8.2 case is working as expected, but the Armv8.5 one is not.
Quoted 4 lines
--
Thanks,
Oliver
Marc Zyngier
maintainerQuoted 96 lines · Ricardo Koller
Quoted 75 lines · Oliver Upton
Hi Ricardo,
Quoted 45 lines · Ricardo Koller
Quoted 24 lines · Marc Zyngier
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Aleven ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Aleven PMUv3p5 to be exposed to the guestQuoted 10 lines
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)Signature · 1 line
2.34.1Hi Marc,
There is one extra potential issue with exposing PMUv3p5. I see this weird behavior when doing passthrough ("bare metal") on the fast-model configured to emulate PMUv3p5: the [63:32] half of the counters overflowing at 32-bits is still incremented.
Fast model - ARMv8.5:
Quoted 11 lines
Assuming the initial state is even=0xFFFFFFFF and odd=0x0, incrementing the even counter leads to: 0x00000001_00000000 0x00000000_00000001 0x1 even counter odd counter PMOVSET Assuming the initial state is even=0xFFFFFFFF and odd=0xFFFFFFFF, incrementing the even counter leads to: 0x00000001_00000000 0x00000001_00000000 0x3 even counter odd counter PMOVSETThis is to be expected, actually. PMUv8p5 counters are always 64 bit, regardless of the configured overflow.
DDI 0487H D8.3 Behavior on overflow
If
FEAT_PMUv3p5is implemented, 64-bit event counters are implemented, HDCR.HPMN is not 0, and either n is in the range [0 .. (HDCR.HPMN-1)] or EL2 is not implemented, then event counter overflow is configured by PMCR.LP:— When PMCR.LP is set to 0, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [31:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
— When PMCR.LP is set to 1, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [63:0] of the event counter, the PE sets PMOVSCLR[n] to 1.Quoted 7 lines
[...]
For all 64-bit counters, incrementing the counter is the same whether an unsigned overflow occurs at [31:0] or [63:0]. If the counter increments for an event, bits [63:0] are always incremented.
Do you see this same (expected) failure w/ Marc's series?
I don't know, I'm hitting another bug it seems.
Just realized that KVM does not offer PMUv3p5 (with this series applied) when the real hardware is only Armv8.2 (the setup I originally tried). So, tried these other two setups on the fast model:
has_arm_v8-5=1
# ./lkvm-static run --nodefaults --pmu pmu.flat -p pmu-chained-sw-incr # lkvm run -k pmu.flat -m 704 -c 8 --name guest-135
Quoted 13 lines
INFO: PMU version: 0x6
^^^
PMUv3 for Armv8.5
INFO: PMU implementer/ID code: 0x41("A")/0
INFO: Implements 8 event counters
FAIL: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=0
^^^
no overflows
FAIL: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=84 #1=-1
INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=4294967295
SUMMARY: 2 tests, 2 unexpected failuresHmm. I think I see what's wrong. In kvm_pmu_create_perf_event(), we have this:
if (kvm_pmu_idx_is_64bit(vcpu, select_idx)) attr.config1 |= 1; counter = kvm_pmu_get_counter_value(vcpu, select_idx); /* The initial sample period (overflow count) of an event. */ if (kvm_pmu_idx_has_64bit_overflow(vcpu, select_idx)) attr.sample_period = (-counter) & GENMASK(63, 0); else attr.sample_period = (-counter) & GENMASK(31, 0);
but the initial sampling period shouldn't be based on the *guest* counter overflow. It really is about the getting to an overflow on the *host*, so the initial code was correct, and only the width of the counter matters here.
/me goes back to running the FVP...
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Quoted 126 lines · Marc Zyngier
Quoted 96 lines · Ricardo Koller
Quoted 75 lines · Oliver Upton
Hi Ricardo,
Quoted 45 lines · Ricardo Koller
Quoted 24 lines · Marc Zyngier
Ricardo recently reported[1] that our PMU emulation was busted when it comes to chained events, as we cannot expose the overflow on a 32bit boundary (which the architecture requires).
This series aims at fixing this (by deleting a lot of code), and as a bonus adds support for PMUv3p5, as this requires us to fix a few more things.
Tested on A53 (PMUv3) and FVP (PMUv3p5).
[1] https://lore.kernel.org/r/[email protected]
Marc Zyngier (9):
KVM: arm64: PMU: Align chained counter implementation with
architecture pseudocode
KVM: arm64: PMU: Distinguish between 64bit counter and 64bit overflow
KVM: arm64: PMU: Only narrow counters that are not 64bit wide
KVM: arm64: PMU: Add counter_index_to_*reg() helpers
KVM: arm64: PMU: Simplify setting a counter to a specific value
KVM: arm64: PMU: Move the ID_AA64DFR0_EL1.PMUver limit to VM creation
KVM: arm64: PMU: Aleven ID_AA64DFR0_EL1.PMUver to be set from userspace
KVM: arm64: PMU: Implement PMUv3p5 long counter support
KVM: arm64: PMU: Aleven PMUv3p5 to be exposed to the guestQuoted 10 lines
arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 6 + arch/arm64/kvm/pmu-emul.c | 372 ++++++++++-------------------- arch/arm64/kvm/sys_regs.c | 65 +++++- include/kvm/arm_pmu.h | 16 +- 5 files changed, 208 insertions(+), 252 deletions(-)Signature · 1 line
2.34.1Hi Marc,
There is one extra potential issue with exposing PMUv3p5. I see this weird behavior when doing passthrough ("bare metal") on the fast-model configured to emulate PMUv3p5: the [63:32] half of the counters overflowing at 32-bits is still incremented.
Fast model - ARMv8.5:
Quoted 11 lines
Assuming the initial state is even=0xFFFFFFFF and odd=0x0, incrementing the even counter leads to: 0x00000001_00000000 0x00000000_00000001 0x1 even counter odd counter PMOVSET Assuming the initial state is even=0xFFFFFFFF and odd=0xFFFFFFFF, incrementing the even counter leads to: 0x00000001_00000000 0x00000001_00000000 0x3 even counter odd counter PMOVSETThis is to be expected, actually. PMUv8p5 counters are always 64 bit, regardless of the configured overflow.
DDI 0487H D8.3 Behavior on overflow
If
FEAT_PMUv3p5is implemented, 64-bit event counters are implemented, HDCR.HPMN is not 0, and either n is in the range [0 .. (HDCR.HPMN-1)] or EL2 is not implemented, then event counter overflow is configured by PMCR.LP:— When PMCR.LP is set to 0, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [31:0] of the event counter, the PE sets PMOVSCLR[n] to 1.
— When PMCR.LP is set to 1, if incrementing PMEVCNTR<n> causes an unsigned
overflow of bits [63:0] of the event counter, the PE sets PMOVSCLR[n] to 1.Quoted 7 lines
[...]
For all 64-bit counters, incrementing the counter is the same whether an unsigned overflow occurs at [31:0] or [63:0]. If the counter increments for an event, bits [63:0] are always incremented.
Do you see this same (expected) failure w/ Marc's series?
I don't know, I'm hitting another bug it seems.
Just realized that KVM does not offer PMUv3p5 (with this series applied) when the real hardware is only Armv8.2 (the setup I originally tried). So, tried these other two setups on the fast model:
has_arm_v8-5=1
# ./lkvm-static run --nodefaults --pmu pmu.flat -p pmu-chained-sw-incr # lkvm run -k pmu.flat -m 704 -c 8 --name guest-135Quoted 13 lines
INFO: PMU version: 0x6 ^^^ PMUv3 for Armv8.5 INFO: PMU implementer/ID code: 0x41("A")/0 INFO: Implements 8 event counters FAIL: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=0 ^^^ no overflows FAIL: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=84 #1=-1 INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=4294967295 SUMMARY: 2 tests, 2 unexpected failuresHmm. I think I see what's wrong. In
kvm_pmu_create_perf_event(), we have this:if (kvm_pmu_idx_is_64bit(vcpu, select_idx)) attr.config1 |= 1; counter = kvm_pmu_get_counter_value(vcpu, select_idx); /* The initial sample period (overflow count) of an event. */ if (kvm_pmu_idx_has_64bit_overflow(vcpu, select_idx)) attr.sample_period = (-counter) & GENMASK(63, 0); else attr.sample_period = (-counter) & GENMASK(31, 0);
Quoted 4 lines
but the initial sampling period shouldn't be based on the *guest* counter overflow. It really is about the getting to an overflow on the *host*, so the initial code was correct, and only the width of the counter matters here.
Right, I think this requires bringing back some of the chained related code (like update_pmc_chained() and pmc_is_chained()), because
attr.sample_period = (-counter) & GENMASK(31, 0);
should also be used when the counter is chained.
Thanks,
Ricardo
Quoted 9 lines
/me goes back to running the FVP...
Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerHi Ricardo,
Quoted 47 lines · Ricardo Koller
Quoted 40 lines · Marc Zyngier
Quoted 10 lines · Ricardo Koller
Just realized that KVM does not offer PMUv3p5 (with this series applied) when the real hardware is only Armv8.2 (the setup I originally tried). So, tried these other two setups on the fast model:
has_arm_v8-5=1
# ./lkvm-static run --nodefaults --pmu pmu.flat -p pmu-chained-sw-incr # lkvm run -k pmu.flat -m 704 -c 8 --name guest-135Quoted 13 lines
INFO: PMU version: 0x6 ^^^ PMUv3 for Armv8.5 INFO: PMU implementer/ID code: 0x41("A")/0 INFO: Implements 8 event counters FAIL: pmu: pmu-chained-sw-incr: overflow and chain counter incremented after 100 SW_INCR/CHAIN INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=0 ^^^ no overflows FAIL: pmu: pmu-chained-sw-incr: expected overflows and values after 100 SW_INCR/CHAIN INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=84 #1=-1 INFO: pmu: pmu-chained-sw-incr: overflow=0x0, #0=4294967380 #1=4294967295 SUMMARY: 2 tests, 2 unexpected failuresHmm. I think I see what's wrong. In
kvm_pmu_create_perf_event(), we have this:if (kvm_pmu_idx_is_64bit(vcpu, select_idx)) attr.config1 |= 1; counter = kvm_pmu_get_counter_value(vcpu, select_idx); /* The initial sample period (overflow count) of an event. */ if (kvm_pmu_idx_has_64bit_overflow(vcpu, select_idx)) attr.sample_period = (-counter) & GENMASK(63, 0); else attr.sample_period = (-counter) & GENMASK(31, 0);Quoted 4 lines
but the initial sampling period shouldn't be based on the *guest* counter overflow. It really is about the getting to an overflow on the *host*, so the initial code was correct, and only the width of the counter matters here.
Quoted 6 lines
Right, I think this requires bringing back some of the chained related code (like
update_pmc_chained()andpmc_is_chained()), becauseattr.sample_period = (-counter) & GENMASK(31, 0);should also be used when the counter is chained.
Almost, but not quite. I came up with the following hack (not everything is relevant, but you'll get my drift):
arch/arm64/kvm/pmu-emul.c +19 -9
@@ -450,6 +450,9 @@ static void kvm_pmu_counter_increment(struct kvm_vcpu *vcpu, reg =lower_32_bits (reg);__vcpu_sys_reg (vcpu,PMEVCNTR0_EL0 + i) = reg; +if (!kvm_pmu_idx_has_64bit_overflow (vcpu, i)) + reg =lower_32_bits (reg); +if (reg)/* No overflow? move on */ continue ;
@@ -483,7 +486,7 @@ static void kvm_pmu_perf_overflow(struct perf_event *perf_event,*/ period = -(local64_read (&perf_event->count)); -if (!kvm_pmu_idx_has_64bit_overflow (vcpu, pmc->idx)) +if (!kvm_pmu_idx_is_64bit (vcpu, pmc->idx)) period &=GENMASK (31 ,0 );local64_set (&perf_event->hw.period_left,0 );
@@ -605,17 +608,24 @@ static void kvm_pmu_create_perf_event(struct kvm_vcpu *vcpu, u64 select_idx) attr.exclude_host =1 ;/* Don't count host events */ attr.config = eventsel; -/* If counting a 64bit event, advertise it to the perf code */ -if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) - attr.config1 |=1 ; - counter =kvm_pmu_get_counter_value (vcpu, select_idx); -/* The initial sample period (overflow count) of an event. */ -if (kvm_pmu_idx_has_64bit_overflow (vcpu, select_idx)) - attr.sample_period = (-counter) &GENMASK (63 ,0 ); -else +/* +* If counting with a 64bit counter, advertise it to the perf +* code, carefully dealing with the initial sample period +* which also depends on the overflow. +*/ +if (kvm_pmu_idx_is_64bit (vcpu, select_idx)) { + attr.config1 |=1 ; + +if (!kvm_pmu_idx_has_64bit_overflow (vcpu, select_idx)) { + attr.sample_period = -(counter &GENMASK (31 ,0 )); + }else { + attr.sample_period = (-counter) &GENMASK (63 ,0 ); + } + }else { attr.sample_period = (-counter) &GENMASK (31 ,0 ); + } event =perf_event_create_kernel_counter (&attr,-1 , current, kvm_pmu_perf_overflow, pmc);
With this, I'm back in business (in QEMU, as I *still* cannot get ARM to give me a model that runs natively on arm64...):
root@debian:~/kvm-unit-tests# ../kvmtool/lkvm run --nodefaults --pmu --firmware arm/pmu.flat -p pmu-chained-sw-incr
# lkvm run --firmware arm/pmu.flat -m 448 -c 4 --name guest-400
Info: Removed ghost socket file "/root/.lkvm//guest-400.sock".
WARNING: early print support may not work. Found uart at 0x1000000, but early base is 0x9000000.
INFO: PMU version: 0x6
INFO: PMU implementer/ID code: 0x41("A")/0x1
INFO: Implements 6 event counters
FAIL: pmu: pmu-chained-sw-incr: no overflow and chain counter incremented after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x1, #0=4294967380 #1=1
FAIL: pmu: pmu-chained-sw-incr: overflow on chain counter and expected values after 100 SW_INCR/CHAIN
INFO: pmu: pmu-chained-sw-incr: overflow=0x3, #0=4294967380 #1=4294967296
SUMMARY: 2 tests, 2 unexpected failuresThe tests themselves need some extra love to account for the fact that the counters are always 64bit irrespective of the overflow, but at least I'm now correctly seeing the odd counter incrementing.
I'll try to continue addressing the comments tomorrow.
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm