arm/kvm: Enable support for KVM_ARM_VCPU_PMU_V3_FILTER
PATCH v2 1/1 arm/kvm: Enable support for KVM_ARM_VCPU_PMU_V3_FILTER
The KVM_ARM_VCPU_PMU_V3_FILTER provide the ability to let the VMM decide which PMU events are provided to the guest. Add a new option pmu-filter as -accel sub-option to set the PMU Event Filtering.
The pmu-filter has such format:
pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"
The A means "allow" and D means "deny", start is the first event of the range and the end is the last one. The first filter action defines if the whole event list is an allow or deny list, if the first filter action is "allow", all other events are denied except start-end; if the first filter action is "deny", all other events are allowed except start-end. For example:
pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"
This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is also allowed except the event 0x30 is denied, and all the other events are disallowed.
Here is an real example shows how to use the PMU Event Filtering, when we launch a guest by use kvm, add such command line:
# qemu-system-aarch64 \ -accel kvm,pmu-filter="D:0x11-0x11"
And then in guest, use the perf to count the cycle:
# perf stat sleep 1
Performance counter stats for 'sleep 1':
1.22 msec task-clock # 0.001 CPUs utilized
1 context-switches # 820.695 /sec
0 cpu-migrations # 0.000 /sec
55 page-faults # 45.138 K/sec
<not supported> cycles
1128954 instructions
227031 branches # 186.323 M/sec
8686 branch-misses # 3.83% of all branches 1.002492480 seconds time elapsed
0.001752000 seconds user
0.000000000 seconds sysAs we can see, the cycle counter has been disabled in the guest, but other pmu events are still work.
---v1->v2:
- Add more description for allow and deny meaning in
commit message. [Sebastian]
- Small improvement. [Sebastian]
v1: https://lore.kernel.org/all/[email protected]/
--- include/sysemu/kvm_int.h | 1 + qemu-options.hx | 16 +++++++++++++ target/arm/kvm.c | 22 +++++++++++++++++ target/arm/kvm64.c | 51 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 90 insertions(+)
include/sysemu/kvm_int.h +1 -0
@@ -120,6 +120,7 @@ struct KVMStateuint32_t xen_caps;uint16_t xen_gnttab_max_frames;uint16_t xen_evtchn_max_pirq; +char *kvm_pmu_filter; }; void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,
qemu-options.hx +16 -0
@@ -187,6 +187,7 @@ DEF("accel", HAS_ARG, QEMU_OPTION_accel, " tb-size=n (TCG translation block cache size)\n" " dirty-ring-size=n (KVM dirty ring GFN count, default 0)\n" " eager-split-size=n (KVM Eager Page Split chunk size, default 0, disabled. ARM only)\n" + " pmu-filter={A,D}:start-end[;...] (KVM PMU Event Filter, default no filter. ARM only)\n" " notify-vmexit=run|internal-error|disable,notify-window=n (enable notify VM exit and set notify window, x86 only)\n" " thread=single|multi (enable multi-threaded TCG)\n", QEMU_ARCH_ALL) SRST
@@ -259,6 +260,21 @@ SRST impact on the memory. By default, this feature is disabled (eager-split-size=0). + ``pmu-filter={A,D}:start-end[;...]`` + KVM implements pmu event filtering to prevent a guest from being able to + sample certain events. It has the following format: + + pmu-filter="{A,D}:start-end[;{A,D}:start-end...]" + + The A means "allow" and D means "deny", start if the first event of the + range and the end is the last one. For example: + + pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30" + + This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is + also allowed except the event 0x30 is denied, and all the other events + are disallowed. + ``notify-vmexit=run|internal-error|disable,notify-window=n`` Enables or disables notify VM exit support on x86 host and specify the corresponding notify window to trigger the VM exit if enabled.
target/arm/kvm.c +22 -0
@@ -1108,6 +1108,21 @@ static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, s->kvm_eager_split_size = value; } +static char *kvm_arch_get_pmu_filter (Object *obj,Error **errp) +{ +KVMState *s =KVM_STATE (obj); + +return g_strdup (s->kvm_pmu_filter); +} + +static void kvm_arch_set_pmu_filter (Object *obj,const char *pmu_filter, +Error **errp) +{ +KVMState *s =KVM_STATE (obj); + + s->kvm_pmu_filter =g_strdup (pmu_filter); +} +void kvm_arch_accel_class_init (ObjectClass *oc) { object_class_property_add(oc, "eager-split-size", "size",
@@ -1116,4 +1131,11 @@ void kvm_arch_accel_class_init(ObjectClass *oc)object_class_property_set_description (oc,"eager-split-size" ,"Eager Page Split chunk size for hugepages. (default: 0, disabled)" ); + +object_class_property_add_str (oc,"pmu-filter" , + kvm_arch_get_pmu_filter, + kvm_arch_set_pmu_filter); + +object_class_property_set_description (oc,"pmu-filter" , +"PMU Event Filtering description for guest pmu. (default: NULL, disabled)" ); }
target/arm/kvm64.c +51 -0
69 lines not on this page · raw message
Signature · 1 line
2.40.1
Hi Shaoqin,
Quoted 3 lines · Shaoqin Huang
The
KVM_ARM_VCPU_PMU_V3_FILTERprovide the ability to let the VMM decide which PMU events are provided to the guest. Add a new optionpmu-filteras -accel sub-option to set the PMU Event Filtering.
you remind the reader the default policy without filter (ie. expose all events from the hots)
Quoted 10 lines
The
pmu-filterhas such format:pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"
The A means "allow" and D means "deny", start is the first event of the range and the end is the last one. The first filter action defines if the whole event list is an allow or deny list, if the first filter action is "allow", all other events are denied except start-end; if the first filter action is "deny", all other events are allowed except start-end. For example:
I prefer the kernel doc wording
The first registered range defines the global policy (global ALLOW if the first @action is DENY, global DENY if the first @action is ALLOW).
Quoted 2 lines
pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"
shoudn't the "," be replaced by a ";"?
I would add: since the first action is allow, we have a global deny policy.
Quoted 10 lines
This will allow event
0x11(The cycle counter), events0x23to0x3ais also allowed except the event0x30is denied, and all the other events are disallowed.Here is an real example shows how to use the PMU Event Filtering, when we launch a guest by use kvm, add such command line:
# qemu-system-aarch64 \ -accel kvm,pmu-filter="D:0x11-0x11"
Since the first filter action is deny, we have a global allow policy. this disables the filtering of the cycle counter (event 0x11 being CPU_CYCLES)
kernel doc says that the ranges should match the PMU arch (10 bits on ARMv8.0, 16 bits from ARMv8.1 onwards). How do you handle that?
Quoted 16 lines
And then in guest, use the perf to count the cycle:
# perf stat sleep 1Performance counter stats for 'sleep 1':
1.22 msec task-clock # 0.001 CPUs utilized 1 context-switches # 820.695 /sec 0 cpu-migrations # 0.000 /sec 55 page-faults # 45.138 K/sec <not supported> cycles 1128954 instructions 227031 branches # 186.323 M/sec 8686 branch-misses # 3.83% of all branches
Quoted 7 lines
1.002492480 seconds time elapsed 0.001752000 seconds user 0.000000000 seconds sysAs we can see, the cycle counter has been disabled in the guest, but other pmu events are still work.
perf list should work as well
Quoted 41 lines
Signed-off-by Shaoqin Huang <[email protected]>---
v1->v2:
- Add more description for allow and deny meaning in
commit message. [Sebastian]
- Small improvement. [Sebastian]v1: https://lore.kernel.org/all/[email protected]/
--- include/sysemu/kvm_int.h | 1 + qemu-options.hx | 16 +++++++++++++ target/arm/kvm.c | 22 +++++++++++++++++ target/arm/kvm64.c | 51 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 90 insertions(+)include/sysemu/kvm_int.h +1
-0@@ -120,6 +120,7 @@ struct KVMStateuint32_t xen_caps;uint16_t xen_gnttab_max_frames;uint16_t xen_evtchn_max_pirq; +char *kvm_pmu_filter; }; void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,qemu-options.hx +1
-0@@ -187,6 +187,7 @@ DEF("accel", HAS_ARG, QEMU_OPTION_accel, " tb-size=n (TCG translation block cache size)\n" " dirty-ring-size=n (KVM dirty ring GFN count, default 0)\n" " eager-split-size=n (KVM Eager Page Split chunk size, default 0, disabled. ARM only)\n" + " pmu-filter={A,D}:start-end[;...] (KVM PMU Event Filter, default no filter. ARM only)\n" " notify-vmexit=run|internal-error|disable,notify-window=n (enable notify VM exit and set notify window, x86 only)\n" " thread=single|multi (enable multi-threaded TCG)\n", QEMU_ARCH_ALL) SRST
qemu-options.hx SRST · line 267
Quoted 8 lines
qemu-options.hx +5
-0@@ -259,6 +260,21 @@ SRST impact on the memory. By default, this feature is disabled (eager-split-size=0). + ``pmu-filter={A,D}:start-end[;...]`` + KVM implements pmu event filtering to prevent a guest from being able to + sample certain events. It has the following format: + + pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"
you may add []* to express that you have any number of ranges
qemu-options.hx SRST · line 272
Quoted 5 lines
qemu-options.hx +5
-0+ + The A means "allow" and D means "deny", start if the first event of the + range and the end is the last one. For example: + + pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"
is is hex format only?
qemu-options.hx SRST · line 276
Quoted 4 lines
qemu-options.hx +4
-0+ + This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is + also allowed except the event 0x30 is denied, and all the other events + are disallowed.
s/disallowed/hidden?
Quoted 15 lines
qemu-options.hx +1
-0+ ``notify-vmexit=run|internal-error|disable,notify-window=n`` Enables or disables notify VM exit support on x86 host and specify the corresponding notify window to trigger the VM exit if enabled.target/arm/kvm.c +3
-0@@ -1108,6 +1108,21 @@ static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, s->kvm_eager_split_size = value; } +static char *kvm_arch_get_pmu_filter (Object *obj,Error **errp) +{ +KVMState *s =KVM_STATE (obj);
target/arm/kvm.c static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, · line 1123
Quoted 10 lines
target/arm/kvm.c +10
-0+ +return g_strdup (s->kvm_pmu_filter); +} + +static void kvm_arch_set_pmu_filter (Object *obj,const char *pmu_filter, +Error **errp) +{ +KVMState *s =KVM_STATE (obj); + + s->kvm_pmu_filter =g_strdup (pmu_filter);
can the user specify the option several times in which case we would leak here?
Quoted 42 lines
target/arm/kvm.c +9
-0+} +void kvm_arch_accel_class_init (ObjectClass *oc) { object_class_property_add(oc, "eager-split-size", "size",@@ -1116,4 +1131,11 @@ void kvm_arch_accel_class_init(ObjectClass *oc)object_class_property_set_description (oc,"eager-split-size" ,"Eager Page Split chunk size for hugepages. (default: 0, disabled)" ); + +object_class_property_add_str (oc,"pmu-filter" , + kvm_arch_get_pmu_filter, + kvm_arch_set_pmu_filter); + +object_class_property_set_description (oc,"pmu-filter" , +"PMU Event Filtering description for guest pmu. (default: NULL, disabled)" ); }target/arm/kvm64.c +10
-0@@ -10,6 +10,7 @@*/ #include "qemu/osdep.h" +#include <asm-arm64/kvm.h> #include <sys/ioctl.h> #include <sys/ptrace.h> @@ -131,6 +132,53 @@ static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr,return true; } +static void kvm_arm_pmu_filter_init (CPUState *cs) +{ +static bool pmu_filter_init = false; +struct kvm_pmu_event_filter filter; +struct kvm_device_attr attr = { + .group =KVM_ARM_VCPU_PMU_V3_CTRL , + .attr =KVM_ARM_VCPU_PMU_V3_FILTER , + .addr = (uint64_t )&filter, + };
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 153
Quoted 10 lines
target/arm/kvm64.c +10
-0+KVMState *kvm_state = cs->kvm_state; +char *tmp; +char *str, act; + +if (!kvm_state->kvm_pmu_filter) +return ; + +/* This only needs to be called for 1 vcpu. */ +if (!pmu_filter_init) + pmu_filter_init = true;
where is it used?
Quoted 11 lines
target/arm/kvm64.c +11
-0+ + tmp =g_strdup (kvm_state->kvm_pmu_filter); + +for (str =strtok (tmp,";" ); str !=NULL ; str =strtok (NULL ,";" )) { +unsigned short start =0 , end =0 ; + +sscanf (str,"%c:%hx-%hx" , &act, &start, &end); +if ((act !='A' && act !='D' ) || (!start && !end)) { +error_report ("skipping invalid filter %s\n" , str); +continue ; + }
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 174
Quoted 10 lines
target/arm/kvm64.c +10
-0+ + filter = (struct kvm_pmu_event_filter ) { + .base_event = start, + .nevents = end - start +1 , + .action = act =='A' ?KVM_PMU_EVENT_ALLOW : +KVM_PMU_EVENT_DENY , + }; + +if (!kvm_arm_set_device_attr (cs, &attr,"PMU Event Filter" )) { +error_report ("Failed to init PMU Event Filter\n" );
you may add some hints about why this failed.
Quoted 11 lines
target/arm/kvm64.c +7
-0+abort (); + } + } + +g_free (tmp); +} +void kvm_arm_pmu_init (CPUState *cs) {struct kvm_device_attr attr = {
target/arm/kvm64.c void kvm_arm_pmu_init(CPUState *cs) · line 197
Quoted 9 lines
target/arm/kvm64.c +3
-0@@ -141,6 +189,9 @@ void kvm_arm_pmu_init(CPUState *cs)if (!ARM_CPU (cs)->has_pmu) {return ; } + +kvm_arm_pmu_filter_init (cs); +if (!kvm_arm_set_device_attr (cs, &attr,"PMU" )) {error_report ("failed to init PMU" );abort ();
I see x86 seems to have a similar capability (see KVM_CAP_PMU_EVENT_FILTER). But I am not sure this is integrated in qemu?
Thanks
Eric
Hi Eric,
Quoted 8 lines · Eric Auger
Hi Shaoqin,
Quoted 3 lines · Shaoqin Huang
The
KVM_ARM_VCPU_PMU_V3_FILTERprovide the ability to let the VMM decide which PMU events are provided to the guest. Add a new optionpmu-filteras -accel sub-option to set the PMU Event Filtering.you remind the reader the default policy without filter (ie. expose all events from the hots)
Yes. I will add this description to the default policy.
Quoted 11 lines
Quoted 10 lines
The
pmu-filterhas such format:pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"The A means "allow" and D means "deny", start is the first event of the range and the end is the last one. The first filter action defines if the whole event list is an allow or deny list, if the first filter action is "allow", all other events are denied except start-end; if the first filter action is "deny", all other events are allowed except start-end. For example:
Quoted 3 lines
I prefer the kernel doc wording
The first registered range defines the global policy (global ALLOW if the first @action is DENY, global DENY if the first @action is ALLOW).
I can replace it this by kernel doc description in next version.
Quoted 3 lines
Quoted 2 lines
pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"shoudn't the "," be replaced by a ";"?
Yes. Thanks for catching this. It should be ";".
Quoted 3 lines
I would add: since the first action is allow, we have a global deny policy.
That makes the example more clear, will add it.
Quoted 8 lines
Quoted 8 lines
This will allow event
0x11(The cycle counter), events0x23to0x3ais also allowed except the event0x30is denied, and all the other events are disallowed.Here is an real example shows how to use the PMU Event Filtering, when we launch a guest by use kvm, add such command line:
Quoted 8 lines
Quoted 2 lines
# qemu-system-aarch64 \ -accel kvm,pmu-filter="D:0x11-0x11"Since the first filter action is deny, we have a global allow policy. this disables the filtering of the cycle counter (event
0x11beingCPU_CYCLES)kernel doc says that the ranges should match the PMU arch (10 bits on ARMv8.0, 16 bits from ARMv8.1 onwards). How do you handle that?
Currently I think I can rely on the SET_DEVICE_ATTR, when set the KVM_ARM_VCPU_PMU_V3_FILTER, check the errno number, if it equals to EINVAL, then report the error to the use it's an invalid filter.
Or another way is to detect the ARM version, and do more check in the userspace?
Do you have any good suggestions on handle the two different event space in QEMU?
Quoted 18 lines
Quoted 18 lines
And then in guest, use the perf to count the cycle:
# perf stat sleep 1Performance counter stats for 'sleep 1':
1.22 msec task-clock # 0.001 CPUs utilized
1 context-switches # 820.695 /sec
0 cpu-migrations # 0.000 /sec
55 page-faults # 45.138 K/sec
<not supported> cycles
1128954 instructions
227031 branches # 186.323 M/sec
8686 branch-misses # 3.83% of all branches1.002492480 seconds time elapsed
Quoted 7 lines
Quoted 5 lines
0.001752000 seconds user 0.000000000 seconds sysAs we can see, the cycle counter has been disabled in the guest, but other pmu events are still work.
perf list should work as well
It works, should I post it output at here?
Quoted 50 lines
Quoted 41 lines
Signed-off-by Shaoqin Huang <[email protected]>---
v1->v2:
- Add more description for allow and deny meaning in
commit message. [Sebastian]
- Small improvement. [Sebastian]v1: https://lore.kernel.org/all/[email protected]/
--- include/sysemu/kvm_int.h | 1 + qemu-options.hx | 16 +++++++++++++ target/arm/kvm.c | 22 +++++++++++++++++ target/arm/kvm64.c | 51 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 90 insertions(+)include/sysemu/kvm_int.h +1
-0@@ -120,6 +120,7 @@ struct KVMStateuint32_t xen_caps;uint16_t xen_gnttab_max_frames;uint16_t xen_evtchn_max_pirq; +char *kvm_pmu_filter; }; void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,qemu-options.hx +1
-0@@ -187,6 +187,7 @@ DEF("accel", HAS_ARG, QEMU_OPTION_accel, " tb-size=n (TCG translation block cache size)\n" " dirty-ring-size=n (KVM dirty ring GFN count, default 0)\n" " eager-split-size=n (KVM Eager Page Split chunk size, default 0, disabled. ARM only)\n" + " pmu-filter={A,D}:start-end[;...] (KVM PMU Event Filter, default no filter. ARM only)\n" " notify-vmexit=run|internal-error|disable,notify-window=n (enable notify VM exit and set notify window, x86 only)\n" " thread=single|multi (enable multi-threaded TCG)\n", QEMU_ARCH_ALL) SRSTqemu-options.hx SRST · line 267
Quoted 8 lines
qemu-options.hx +5
-0@@ -259,6 +260,21 @@ SRST impact on the memory. By default, this feature is disabled (eager-split-size=0). + ``pmu-filter={A,D}:start-end[;...]`` + KVM implements pmu event filtering to prevent a guest from being able to + sample certain events. It has the following format: + + pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"you may add []* to express that you have any number of ranges
Quoted 6 lines
qemu-options.hx SRST · line 272
Quoted 5 lines
qemu-options.hx +5
-0+ + The A means "allow" and D means "deny", start if the first event of the + range and the end is the last one. For example: + + pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"is is hex format only?
Yes, only support hex format now. Because we need to handle the input string by ourself, handle different format at the input is not interesting. If it really need to support different format, we can consider it in next version?
Quoted 5 lines
qemu-options.hx SRST · line 276
Quoted 4 lines
qemu-options.hx +4
-0+ + This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is + also allowed except the event 0x30 is denied, and all the other events + are disallowed.s/disallowed/hidden?
I think disallowed is properly.
Quoted 12 lines
Quoted 12 lines
qemu-options.hx +1
-0+ ``notify-vmexit=run|internal-error|disable,notify-window=n`` Enables or disables notify VM exit support on x86 host and specify the corresponding notify window to trigger the VM exit if enabled.target/arm/kvm.c +0
-0@@ -1108,6 +1108,21 @@ static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, s->kvm_eager_split_size = value; }
Quoted 15 lines
target/arm/kvm.c static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, · line 1123
Quoted 13 lines
target/arm/kvm.c +13
-0+static char *kvm_arch_get_pmu_filter (Object *obj,Error **errp) +{ +KVMState *s =KVM_STATE (obj); + +return g_strdup (s->kvm_pmu_filter); +} + +static void kvm_arch_set_pmu_filter (Object *obj,const char *pmu_filter, +Error **errp) +{ +KVMState *s =KVM_STATE (obj); + + s->kvm_pmu_filter =g_strdup (pmu_filter);can the user specify the option several times in which case we would leak here?
It seems will leak. I should add more code to handle the case here.
Quoted 45 lines
Quoted 45 lines
target/arm/kvm.c +9
-0+} +void kvm_arch_accel_class_init (ObjectClass *oc) { object_class_property_add(oc, "eager-split-size", "size",@@ -1116,4 +1131,11 @@ void kvm_arch_accel_class_init(ObjectClass *oc)object_class_property_set_description (oc,"eager-split-size" ,"Eager Page Split chunk size for hugepages. (default: 0, disabled)" ); + +object_class_property_add_str (oc,"pmu-filter" , + kvm_arch_get_pmu_filter, + kvm_arch_set_pmu_filter); + +object_class_property_set_description (oc,"pmu-filter" , +"PMU Event Filtering description for guest pmu. (default: NULL, disabled)" ); }target/arm/kvm64.c +13
-0@@ -10,6 +10,7 @@*/ #include "qemu/osdep.h" +#include <asm-arm64/kvm.h> #include <sys/ioctl.h> #include <sys/ptrace.h> @@ -131,6 +132,53 @@ static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr,return true; } +static void kvm_arm_pmu_filter_init (CPUState *cs) +{ +static bool pmu_filter_init = false; +struct kvm_pmu_event_filter filter; +struct kvm_device_attr attr = { + .group =KVM_ARM_VCPU_PMU_V3_CTRL , + .attr =KVM_ARM_VCPU_PMU_V3_FILTER , + .addr = (uint64_t )&filter, + }; +KVMState *kvm_state = cs->kvm_state; +char *tmp; +char *str, act;
Quoted 8 lines
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 153
Quoted 7 lines
target/arm/kvm64.c +7
-0+ +if (!kvm_state->kvm_pmu_filter) +return ; + +/* This only needs to be called for 1 vcpu. */ +if (!pmu_filter_init) + pmu_filter_init = true;where is it used?
It's only used in this function. It's a static variable, because the pmu_filter only need to be init once from the vcpu ioctl. But the kvm_arm_pmu_init() will call the kvm_arm_pmu_filter_init() multiple times since every cpu needs to init pmu.
Quoted 14 lines
Quoted 14 lines
target/arm/kvm64.c +14
-0+ + tmp =g_strdup (kvm_state->kvm_pmu_filter); + +for (str =strtok (tmp,";" ); str !=NULL ; str =strtok (NULL ,";" )) { +unsigned short start =0 , end =0 ; + +sscanf (str,"%c:%hx-%hx" , &act, &start, &end); +if ((act !='A' && act !='D' ) || (!start && !end)) { +error_report ("skipping invalid filter %s\n" , str); +continue ; + } + + filter = (struct kvm_pmu_event_filter ) { + .base_event = start,
Quoted 8 lines
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 174
Quoted 7 lines
target/arm/kvm64.c +7
-0+ .nevents = end - start + 1, + .action = act == 'A' ? KVM_PMU_EVENT_ALLOW : + KVM_PMU_EVENT_DENY, + }; + +if (!kvm_arm_set_device_attr (cs, &attr,"PMU Event Filter" )) { +error_report ("Failed to init PMU Event Filter\n" );you may add some hints about why this failed.
I think I can report it depends on the errno, and gives more detailed information for the user.
Quoted 15 lines
Quoted 15 lines
target/arm/kvm64.c +8
-0+abort (); + } + } + +g_free (tmp); +} +void kvm_arm_pmu_init (CPUState *cs) {struct kvm_device_attr attr = {@@ -141,6 +189,9 @@ void kvm_arm_pmu_init(CPUState *cs)if (!ARM_CPU (cs)->has_pmu) {return ; } +
Quoted 9 lines
target/arm/kvm64.c void kvm_arm_pmu_init(CPUState *cs) · line 197
Quoted 5 lines
target/arm/kvm64.c +2
-0+kvm_arm_pmu_filter_init (cs); +if (!kvm_arm_set_device_attr (cs, &attr,"PMU" )) {error_report ("failed to init PMU" );abort ();I see x86 seems to have a similar capability (see
KVM_CAP_PMU_EVENT_FILTER). But I am not sure this is integrated in qemu?
I don't see the x86 capability has been integrated into qemu. But I know there is a kselftest for this capability on x86.
Thanks,
Shaoqin
Quoted 4 lines
Thanks
Eric
Signature · 1 line
Shaoqin
Hi,
Quoted 170 lines · Shaoqin Huang
The
KVM_ARM_VCPU_PMU_V3_FILTERprovide the ability to let the VMM decide which PMU events are provided to the guest. Add a new optionpmu-filteras -accel sub-option to set the PMU Event Filtering.The
pmu-filterhas such format:pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"
The A means "allow" and D means "deny", start is the first event of the range and the end is the last one. The first filter action defines if the whole event list is an allow or deny list, if the first filter action is "allow", all other events are denied except start-end; if the first filter action is "deny", all other events are allowed except start-end. For example:
pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"
This will allow event
0x11(The cycle counter), events0x23to0x3ais also allowed except the event0x30is denied, and all the other events are disallowed.Here is an real example shows how to use the PMU Event Filtering, when we launch a guest by use kvm, add such command line:
# qemu-system-aarch64 \ -accel kvm,pmu-filter="D:0x11-0x11"And then in guest, use the perf to count the cycle:
# perf stat sleep 1Performance counter stats for 'sleep 1':
1.22 msec task-clock # 0.001 CPUs utilized 1 context-switches # 820.695 /sec 0 cpu-migrations # 0.000 /sec 55 page-faults # 45.138 K/sec <not supported> cycles 1128954 instructions 227031 branches # 186.323 M/sec 8686 branch-misses # 3.83% of all branches1.002492480 seconds time elapsed 0.001752000 seconds user 0.000000000 seconds sysAs we can see, the cycle counter has been disabled in the guest, but other pmu events are still work.
Signed-off-by Shaoqin Huang <[email protected]>---
v1->v2:
- Add more description for allow and deny meaning in
commit message. [Sebastian]
- Small improvement. [Sebastian]v1: https://lore.kernel.org/all/[email protected]/
--- include/sysemu/kvm_int.h | 1 + qemu-options.hx | 16 +++++++++++++ target/arm/kvm.c | 22 +++++++++++++++++ target/arm/kvm64.c | 51 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 90 insertions(+)include/sysemu/kvm_int.h +1
-0@@ -120,6 +120,7 @@ struct KVMStateuint32_t xen_caps;uint16_t xen_gnttab_max_frames;uint16_t xen_evtchn_max_pirq; +char *kvm_pmu_filter; }; void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,qemu-options.hx +16
-0@@ -187,6 +187,7 @@ DEF("accel", HAS_ARG, QEMU_OPTION_accel, " tb-size=n (TCG translation block cache size)\n" " dirty-ring-size=n (KVM dirty ring GFN count, default 0)\n" " eager-split-size=n (KVM Eager Page Split chunk size, default 0, disabled. ARM only)\n" + " pmu-filter={A,D}:start-end[;...] (KVM PMU Event Filter, default no filter. ARM only)\n" " notify-vmexit=run|internal-error|disable,notify-window=n (enable notify VM exit and set notify window, x86 only)\n" " thread=single|multi (enable multi-threaded TCG)\n", QEMU_ARCH_ALL) SRST@@ -259,6 +260,21 @@ SRST impact on the memory. By default, this feature is disabled (eager-split-size=0). + ``pmu-filter={A,D}:start-end[;...]`` + KVM implements pmu event filtering to prevent a guest from being able to + sample certain events. It has the following format: + + pmu-filter="{A,D}:start-end[;{A,D}:start-end...]" + + The A means "allow" and D means "deny", start if the first event of the + range and the end is the last one. For example: + + pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30" + + This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is + also allowed except the event 0x30 is denied, and all the other events + are disallowed. + ``notify-vmexit=run|internal-error|disable,notify-window=n`` Enables or disables notify VM exit support on x86 host and specify the corresponding notify window to trigger the VM exit if enabled.target/arm/kvm.c +22
-0@@ -1108,6 +1108,21 @@ static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, s->kvm_eager_split_size = value; } +static char *kvm_arch_get_pmu_filter (Object *obj,Error **errp) +{ +KVMState *s =KVM_STATE (obj); + +return g_strdup (s->kvm_pmu_filter); +} + +static void kvm_arch_set_pmu_filter (Object *obj,const char *pmu_filter, +Error **errp) +{ +KVMState *s =KVM_STATE (obj); + + s->kvm_pmu_filter =g_strdup (pmu_filter); +} +void kvm_arch_accel_class_init (ObjectClass *oc) { object_class_property_add(oc, "eager-split-size", "size",@@ -1116,4 +1131,11 @@ void kvm_arch_accel_class_init(ObjectClass *oc)object_class_property_set_description (oc,"eager-split-size" ,"Eager Page Split chunk size for hugepages. (default: 0, disabled)" ); + +object_class_property_add_str (oc,"pmu-filter" , + kvm_arch_get_pmu_filter, + kvm_arch_set_pmu_filter); + +object_class_property_set_description (oc,"pmu-filter" , +"PMU Event Filtering description for guest pmu. (default: NULL, disabled)" ); }target/arm/kvm64.c +7
-0@@ -10,6 +10,7 @@*/ #include "qemu/osdep.h" +#include <asm-arm64/kvm.h> #include <sys/ioctl.h> #include <sys/ptrace.h> @@ -131,6 +132,53 @@ static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr,return true; } +static void kvm_arm_pmu_filter_init (CPUState *cs) +{ +static bool pmu_filter_init = false; +struct kvm_pmu_event_filter filter; +struct kvm_device_attr attr = { + .group =KVM_ARM_VCPU_PMU_V3_CTRL ,
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 150
Quoted 10 lines
target/arm/kvm64.c +10
-0+ .attr = KVM_ARM_VCPU_PMU_V3_FILTER, + .addr = (uint64_t)&filter, + }; +KVMState *kvm_state = cs->kvm_state; +char *tmp; +char *str, act; + +if (!kvm_state->kvm_pmu_filter) +return ; +
usually we check the kernel capability (here KVM_CAP_ARM_PMU_V3) before doing further actions. It allows you to give an inidication to the user that the kernel does not allow it. Also you should precise in the doc that this accel option requires host kernel caps I think.
Quoted 14 lines
target/arm/kvm64.c +14
-0+/* This only needs to be called for 1 vcpu. */ +if (!pmu_filter_init) + pmu_filter_init = true; + + tmp =g_strdup (kvm_state->kvm_pmu_filter); + +for (str =strtok (tmp,";" ); str !=NULL ; str =strtok (NULL ,";" )) { +unsigned short start =0 , end =0 ; + +sscanf (str,"%c:%hx-%hx" , &act, &start, &end); +if ((act !='A' && act !='D' ) || (!start && !end)) { +error_report ("skipping invalid filter %s\n" , str); +continue ; + }
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 174
Quoted 10 lines
target/arm/kvm64.c +10
-0+ + filter = (struct kvm_pmu_event_filter ) { + .base_event = start, + .nevents = end - start +1 , + .action = act =='A' ?KVM_PMU_EVENT_ALLOW : +KVM_PMU_EVENT_DENY , + }; + +if (!kvm_arm_set_device_attr (cs, &attr,"PMU Event Filter" )) { +error_report ("Failed to init PMU Event Filter\n" );
if you do the above, here you know that the host allows to set filters but that the user input is incorrect.
Thanks
Eric
Quoted 20 lines
target/arm/kvm64.c +10
-0+abort (); + } + } + +g_free (tmp); +} +void kvm_arm_pmu_init (CPUState *cs) {struct kvm_device_attr attr = {@@ -141,6 +189,9 @@ void kvm_arm_pmu_init(CPUState *cs)if (!ARM_CPU (cs)->has_pmu) {return ; } + +kvm_arm_pmu_filter_init (cs); +if (!kvm_arm_set_device_attr (cs, &attr,"PMU" )) {error_report ("failed to init PMU" );abort ();
Hi Eric,
Quoted 179 lines · Eric Auger
Hi,
Quoted 176 lines · Shaoqin Huang
The
KVM_ARM_VCPU_PMU_V3_FILTERprovide the ability to let the VMM decide which PMU events are provided to the guest. Add a new optionpmu-filteras -accel sub-option to set the PMU Event Filtering.The
pmu-filterhas such format:pmu-filter="{A,D}:start-end[;{A,D}:start-end...]"The A means "allow" and D means "deny", start is the first event of the range and the end is the last one. The first filter action defines if the whole event list is an allow or deny list, if the first filter action is "allow", all other events are denied except start-end; if the first filter action is "deny", all other events are allowed except start-end. For example:
pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30"This will allow event
0x11(The cycle counter), events0x23to0x3ais also allowed except the event0x30is denied, and all the other events are disallowed.Here is an real example shows how to use the PMU Event Filtering, when we launch a guest by use kvm, add such command line:
# qemu-system-aarch64 \ -accel kvm,pmu-filter="D:0x11-0x11"And then in guest, use the perf to count the cycle:
# perf stat sleep 1Performance counter stats for 'sleep 1':
1.22 msec task-clock # 0.001 CPUs utilized
1 context-switches # 820.695 /sec
0 cpu-migrations # 0.000 /sec
55 page-faults # 45.138 K/sec
<not supported> cycles
1128954 instructions
227031 branches # 186.323 M/sec
8686 branch-misses # 3.83% of all branches1.002492480 seconds time elapsed
0.001752000 seconds user
0.000000000 seconds sysAs we can see, the cycle counter has been disabled in the guest, but other pmu events are still work.
Signed-off-by Shaoqin Huang <[email protected]>---
v1->v2:
- Add more description for allow and deny meaning in
commit message. [Sebastian]
- Small improvement. [Sebastian]v1: https://lore.kernel.org/all/[email protected]/
--- include/sysemu/kvm_int.h | 1 + qemu-options.hx | 16 +++++++++++++ target/arm/kvm.c | 22 +++++++++++++++++ target/arm/kvm64.c | 51 ++++++++++++++++++++++++++++++++++++++++ 4 files changed, 90 insertions(+)include/sysemu/kvm_int.h +1
-0@@ -120,6 +120,7 @@ struct KVMStateuint32_t xen_caps;uint16_t xen_gnttab_max_frames;uint16_t xen_evtchn_max_pirq; +char *kvm_pmu_filter; }; void kvm_memory_listener_register(KVMState *s, KVMMemoryListener *kml,qemu-options.hx +16
-0@@ -187,6 +187,7 @@ DEF("accel", HAS_ARG, QEMU_OPTION_accel, " tb-size=n (TCG translation block cache size)\n" " dirty-ring-size=n (KVM dirty ring GFN count, default 0)\n" " eager-split-size=n (KVM Eager Page Split chunk size, default 0, disabled. ARM only)\n" + " pmu-filter={A,D}:start-end[;...] (KVM PMU Event Filter, default no filter. ARM only)\n" " notify-vmexit=run|internal-error|disable,notify-window=n (enable notify VM exit and set notify window, x86 only)\n" " thread=single|multi (enable multi-threaded TCG)\n", QEMU_ARCH_ALL) SRST@@ -259,6 +260,21 @@ SRST impact on the memory. By default, this feature is disabled (eager-split-size=0). + ``pmu-filter={A,D}:start-end[;...]`` + KVM implements pmu event filtering to prevent a guest from being able to + sample certain events. It has the following format: + + pmu-filter="{A,D}:start-end[;{A,D}:start-end...]" + + The A means "allow" and D means "deny", start if the first event of the + range and the end is the last one. For example: + + pmu-filter="A:0x11-0x11;A:0x23-0x3a,D:0x30-0x30" + + This will allow event 0x11 (The cycle counter), events 0x23 to 0x3a is + also allowed except the event 0x30 is denied, and all the other events + are disallowed. + ``notify-vmexit=run|internal-error|disable,notify-window=n`` Enables or disables notify VM exit support on x86 host and specify the corresponding notify window to trigger the VM exit if enabled.target/arm/kvm.c +22
-0@@ -1108,6 +1108,21 @@ static void kvm_arch_set_eager_split_size(Object *obj, Visitor *v, s->kvm_eager_split_size = value; } +static char *kvm_arch_get_pmu_filter (Object *obj,Error **errp) +{ +KVMState *s =KVM_STATE (obj); + +return g_strdup (s->kvm_pmu_filter); +} + +static void kvm_arch_set_pmu_filter (Object *obj,const char *pmu_filter, +Error **errp) +{ +KVMState *s =KVM_STATE (obj); + + s->kvm_pmu_filter =g_strdup (pmu_filter); +} +void kvm_arch_accel_class_init (ObjectClass *oc) { object_class_property_add(oc, "eager-split-size", "size",@@ -1116,4 +1131,11 @@ void kvm_arch_accel_class_init(ObjectClass *oc)object_class_property_set_description (oc,"eager-split-size" ,"Eager Page Split chunk size for hugepages. (default: 0, disabled)" ); + +object_class_property_add_str (oc,"pmu-filter" , + kvm_arch_get_pmu_filter, + kvm_arch_set_pmu_filter); + +object_class_property_set_description (oc,"pmu-filter" , +"PMU Event Filtering description for guest pmu. (default: NULL, disabled)" ); }target/arm/kvm64.c +13
-0@@ -10,6 +10,7 @@*/ #include "qemu/osdep.h" +#include <asm-arm64/kvm.h> #include <sys/ioctl.h> #include <sys/ptrace.h> @@ -131,6 +132,53 @@ static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr,return true; } +static void kvm_arm_pmu_filter_init (CPUState *cs) +{ +static bool pmu_filter_init = false; +struct kvm_pmu_event_filter filter; +struct kvm_device_attr attr = { + .group =KVM_ARM_VCPU_PMU_V3_CTRL , + .attr =KVM_ARM_VCPU_PMU_V3_FILTER , + .addr = (uint64_t )&filter, + }; +KVMState *kvm_state = cs->kvm_state; +char *tmp; +char *str, act;
Quoted 8 lines
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 150
Quoted 4 lines
target/arm/kvm64.c +4
-0+ +if (!kvm_state->kvm_pmu_filter) +return ; +usually we check the kernel capability (here
KVM_CAP_ARM_PMU_V3) before doing further actions. It allows you to give an inidication to the user that the kernel does not allow it. Also you should precise in the doc that this accel option requires host kernel caps I think.
I get it.
Quoted 18 lines
Quoted 18 lines
target/arm/kvm64.c +18
-0+/* This only needs to be called for 1 vcpu. */ +if (!pmu_filter_init) + pmu_filter_init = true; + + tmp =g_strdup (kvm_state->kvm_pmu_filter); + +for (str =strtok (tmp,";" ); str !=NULL ; str =strtok (NULL ,";" )) { +unsigned short start =0 , end =0 ; + +sscanf (str,"%c:%hx-%hx" , &act, &start, &end); +if ((act !='A' && act !='D' ) || (!start && !end)) { +error_report ("skipping invalid filter %s\n" , str); +continue ; + } + + filter = (struct kvm_pmu_event_filter ) { + .base_event = start, + .nevents = end - start +1 ,
Quoted 8 lines
target/arm/kvm64.c static bool kvm_arm_set_device_attr(CPUState *cs, struct kvm_device_attr *attr, · line 174
Quoted 6 lines
target/arm/kvm64.c +6
-0+ .action = act == 'A' ? KVM_PMU_EVENT_ALLOW : + KVM_PMU_EVENT_DENY, + }; + +if (!kvm_arm_set_device_attr (cs, &attr,"PMU Event Filter" )) { +error_report ("Failed to init PMU Event Filter\n" );if you do the above, here you know that the host allows to set filters but that the user input is incorrect.
I see. Thanks for your additional information.
Thanks,
Shaoqin
Quoted 25 lines
Thanks
Eric
Quoted 20 lines
target/arm/kvm64.c +10
-0+abort (); + } + } + +g_free (tmp); +} +void kvm_arm_pmu_init (CPUState *cs) {struct kvm_device_attr attr = {@@ -141,6 +189,9 @@ void kvm_arm_pmu_init(CPUState *cs)if (!ARM_CPU (cs)->has_pmu) {return ; } + +kvm_arm_pmu_filter_init (cs); +if (!kvm_arm_set_device_attr (cs, &attr,"PMU" )) {error_report ("failed to init PMU" );abort ();
Signature · 1 line
Shaoqin