KVM: Add idempotent controls for migrating system counter state
Subsystems (7)
v6 cover KVM: Add idempotent controls for migrating system counter state
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (cntvct_el0, ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.
Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in KVM_GET_CLOCK where use_master_clock is read outside of the pvclock_gtod_sync_lock.
Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.
Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
Patches 11-12 lay the groundwork for patch 13, which exposes CNTVOFF_EL2 through the ONE_REG interface.
Patches 14-15 add test cases for userspace manipulation of the virtual counter-timer.
Patches 16-17 add a vCPU attribute to adjust the host-guest offset of an ARM vCPU, but only implements support for ECV hosts. Patches 18-19 add support for non-ECV hosts by emulating physical counter offsetting.
Patch 20 adds test cases for adjusting the host-guest offset, and finally patch 21 adds a test to measure the emulation overhead of CNTPCT_EL2.
This series was tested on both an Ampere Mt. Jade and Haswell systems. Unfortunately, the ECV portions of this series are untested, as there is no ECV-capable hardware and the ARM fast models only partially implement ECV.
Physical counter benchmark
--------------------------
The following data was collected by running 10000 iterations of the benchmark test from Patch 21 on an Ampere Mt. Jade reference server, A 2S machine with 2 80-core Ampere Altra SoCs. Measurements were collected for both VHE and nVHE operation using the kvm-arm.mode= command-line parameter.
nVHE
----
+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 54ns | 148ns | | Standard Deviation | 124ns | 122ns | | 95th Percentile | 258ns | 348ns | +--------------------+--------+---------+
VHE
---
+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 53ns | 152ns | | Standard Deviation | 92ns | 94ns | | 95th Percentile | 204ns | 307ns | +--------------------+--------+---------+
This series applies cleanly to kvm/queue at the following commit:
6cd974485e25 ("KVM: selftests: Add a test of an unbacked nested PI descriptor")
v1 -> v2:
- Reimplemented as vCPU device attributes instead of a distinct ioctl.
- Added the (realtime, host_tsc) instant support to KVM_{GET,SET}_CLOCK
- Changed the arm64 implementation to broadcast counter offset values to all vCPUs in a guest. This upholds the architectural expectations of a consistent counter-timer across CPUs.
- Fixed a bug with traps in VHE mode. We now configure traps on every transition into a guest to handle differing VMs (trapped, emulated).
v2 -> v3:
- Added documentation for additions to KVM_{GET,SET}_CLOCK
- Added documentation for all new vCPU attributes
- Added documentation for suggested algorithm to migrate a guest's
TSC(s)
- Bug fixes throughout series
- Rename KVM_CLOCK_REAL_TIME -> KVM_CLOCK_REALTIME
v3 -> v4:
- Added patch to address incorrect device helper assertions (Drew)
- Carried Drew's r-b tags where appropriate
- x86 selftest cleanup
- Removed stale kvm_timer_init_vhe() function
- Removed unnecessary GUEST_DONE() from selftests
v4 -> v5:
- Fix typo in TSC migration algorithm
- Carry more of Drew's r-b tags
- clean up run loop logic in counter emulation benchmark (missed from Drew's comments on v3)
v5 -> v6:
- Add fix for race in KVM_GET_CLOCK (Sean)
- Fix 32-bit build issues in series + use of uninitialized host tsc value (Sean)
- General style cleanups
- Rework ARM virtual counter offsetting to match guest behavior. Use the ONE_REG interface instead of a VM attribute (Marc)
- Maintain a single host-guest counter offset, which applies to both physical and virtual counters
- Dropped some of Drew's r-b tags due to nontrivial patch changes (sorry for the churn!)
v1: https://lore.kernel.org/kvm/[email protected]/
v2: https://lore.kernel.org/r/[email protected]
v3: https://lore.kernel.org/r/[email protected]
v4: https://lore.kernel.org/r/[email protected]
v5: https://lore.kernel.org/r/[email protected]
Oliver Upton (21):
KVM: x86: Fix potential race in KVM_GET_CLOCK
KVM: x86: Report host tsc and realtime values in KVM_GET_CLOCK
KVM: x86: Take the pvclock sync lock behind the tsc_write_lock
KVM: x86: Refactor tsc synchronization code
KVM: x86: Expose TSC offset controls to userspace
tools: arch: x86: pull in pvclock headers
selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
selftests: KVM: Fix kvm device helper ioctl assertions
selftests: KVM: Add helpers for vCPU device attributes
selftests: KVM: Introduce system counter offset test
KVM: arm64: Refactor update_vtimer_cntvoff()
KVM: arm64: Separate guest/host counter offset values
KVM: arm64: Allow userspace to configure a vCPU's virtual offset
selftests: KVM: Add helper to check for register presence
selftests: KVM: Add support for aarch64 to system_counter_offset_test
arm64: cpufeature: Enumerate support for Enhanced Counter
Virtualization
KVM: arm64: Allow userspace to configure a guest's counter-timer
offset
KVM: arm64: Configure timer traps in vcpu_load() for VHE
KVM: arm64: Emulate physical counter offsetting on non-ECV systems
selftests: KVM: Test physical counter offsetting
selftests: KVM: Add counter emulation benchmark
Documentation/virt/kvm/api.rst | 52 ++- Documentation/virt/kvm/devices/vcpu.rst | 85 ++++ Documentation/virt/kvm/locking.rst | 11 + arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 5 + arch/arm64/include/uapi/asm/kvm.h | 2 + arch/arm64/kernel/cpufeature.c | 10 + arch/arm64/kvm/arch_timer.c | 224 ++++++++++- arch/arm64/kvm/arm.c | 4 +- arch/arm64/kvm/guest.c | 6 +- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 + arch/arm64/kvm/hyp/nvhe/timer-sr.c | 16 +- arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + arch/arm64/tools/cpucaps | 1 + arch/x86/include/asm/kvm_host.h | 4 + arch/x86/include/uapi/asm/kvm.h | 4 + arch/x86/kvm/x86.c | 364 +++++++++++++----- include/clocksource/arm_arch_timer.h | 1 + include/kvm/arm_arch_timer.h | 6 +- include/uapi/linux/kvm.h | 7 +- tools/arch/x86/include/asm/pvclock-abi.h | 48 +++ tools/arch/x86/include/asm/pvclock.h | 103 +++++ tools/testing/selftests/kvm/.gitignore | 3 + tools/testing/selftests/kvm/Makefile | 4 + .../kvm/aarch64/counter_emulation_benchmark.c | 207 ++++++++++ .../selftests/kvm/include/aarch64/processor.h | 24 ++ .../testing/selftests/kvm/include/kvm_util.h | 13 + tools/testing/selftests/kvm/lib/kvm_util.c | 63 ++- .../kvm/system_counter_offset_test.c | 211 ++++++++++ .../selftests/kvm/x86_64/kvm_clock_test.c | 204 ++++++++++ 31 files changed, 1581 insertions(+), 143 deletions(-) create mode 100644 tools/arch/x86/include/asm/pvclock-abi.h create mode 100644 tools/arch/x86/include/asm/pvclock.h create mode 100644 tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c create mode 100644 tools/testing/selftests/kvm/system_counter_offset_test.c create mode 100644 tools/testing/selftests/kvm/x86_64/kvm_clock_test.c
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 1/21 KVM: x86: Fix potential race in KVM_GET_CLOCK
Sean noticed that KVM_GET_CLOCK was checking kvm_arch.use_master_clock outside of the pvclock sync lock. This is problematic, as the clock value written to the user may or may not actually correspond to a stable TSC.
Fix the race by populating the entire kvm_clock_data structure behind the pvclock_gtod_sync_lock.
--- arch/x86/kvm/x86.c | 39 ++++++++++++++++++++++++++++----------- 1 file changed, 28 insertions(+), 11 deletions(-)
arch/x86/kvm/x86.c +28 -11
@@ -2782,19 +2782,20 @@ static void kvm_gen_update_masterclock(struct kvm *kvm)#endif } -u64 get_kvmclock_ns (struct kvm *kvm) +static void get_kvmclock (struct kvm *kvm,struct kvm_clock_data *data) {struct kvm_arch *ka = &kvm->arch;struct pvclock_vcpu_time_info hv_clock;unsigned long flags; -u64 ret;spin_lock_irqsave (&ka->pvclock_gtod_sync_lock, flags);if (!ka->use_master_clock) {spin_unlock_irqrestore (&ka->pvclock_gtod_sync_lock, flags); -return get_kvmclock_base_ns () + ka->kvmclock_offset; + data->clock =get_kvmclock_base_ns () + ka->kvmclock_offset; +return ; } + data->flags |=KVM_CLOCK_TSC_STABLE ; hv_clock.tsc_timestamp = ka->master_cycle_now; hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset;spin_unlock_irqrestore (&ka->pvclock_gtod_sync_lock, flags);
@@ -2806,13 +2807,26 @@ u64 get_kvmclock_ns(struct kvm *kvm)kvm_get_time_scale (NSEC_PER_SEC ,__this_cpu_read (cpu_tsc_khz) *1000LL , &hv_clock.tsc_shift, &hv_clock.tsc_to_system_mul); - ret =__pvclock_read_cycles (&hv_clock,rdtsc ()); - }else - ret =get_kvmclock_base_ns () + ka->kvmclock_offset; + data->clock =__pvclock_read_cycles (&hv_clock,rdtsc ()); + }else { + data->clock =get_kvmclock_base_ns () + ka->kvmclock_offset; + }put_cpu (); +} -return ret; +u64 get_kvmclock_ns (struct kvm *kvm) +{ +struct kvm_clock_data data; + +/* +* Zero flags as it's accessed RMW, leave everything else uninitialized +* as clock is always written and no other fields are consumed. +*/ + data.flags =0 ; + +get_kvmclock (kvm, &data); +return data.clock; } static void kvm_setup_pvclock_page(struct kvm_vcpu *v,
@@ -6104,11 +6118,14 @@ long kvm_arch_vm_ioctl(struct file *filp, }case KVM_GET_CLOCK : {struct kvm_clock_data user_ns; -u64 now_ns; - now_ns =get_kvmclock_ns (kvm); - user_ns.clock = now_ns; - user_ns.flags = kvm->arch.use_master_clock ?KVM_CLOCK_TSC_STABLE :0 ; +/* +* Zero flags as it is accessed RMW, leave everything else +* uninitialized as clock is always written and no other fields +* are consumed. +*/ + user_ns.flags =0 ; +get_kvmclock (kvm, &user_ns);memset (&user_ns.pad,0 ,sizeof (user_ns.pad)); r = -EFAULT ;
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Paolo Bonzini
maintainerRe: [PATCH v6 01/21] KVM: x86: Fix potential race in KVM_GET_CLOCK
Quoted 5 lines · Oliver Upton
Sean noticed that
KVM_GET_CLOCKwas checking kvm_arch.use_master_clock outside of the pvclock sync lock. This is problematic, as the clock value written to the user may or may not actually correspond to a stable TSC.
Quoted 8 lines
Fix the race by populating the entire
kvm_clock_datastructure behind thepvclock_gtod_sync_lock.Suggested-by Sean Christopherson <[email protected]>Signed-off-by Oliver Upton <[email protected]>--- arch/x86/kvm/x86.c | 39 ++++++++++++++++++++++++++++----------- 1 file changed, 28 insertions(+), 11 deletions(-)
I had a completely independent patch that fixed the same race. It unifies the read sides of tsc_write_lock and pvclock_gtod_sync_lock into a seqcount (and replaces pvclock_gtod_sync_lock with kvm->lock on the write side).
I attach it now (based on https://lore.kernel.org/kvm/[email protected]/T/#t),
but the testing was extremely light so I'm not sure I will be able to include
it in 5.15.
Paolo
-------------- 8< -------------
From: Paolo Bonzini <[email protected]>
Date: Thu, 8 Apr 2021 05:03:44 -0400
Subject: [PATCH] kvm: x86: protect masterclock with a seqcount
Protect the reference point for kvmclock with a seqcount, so that kvmclock updates for all vCPUs can proceed in parallel. Xen runstate updates will also run in parallel and not bounce the kvmclock cacheline.
This also makes it possible to use KVM_REQ_CLOCK_UPDATE (which will block on the seqcount) to prevent entering in the guests until pvclock_update_vm_gtod_copy is complete, and thus to get rid of KVM_REQ_MCLOCK_INPROGRESS.
nr_vcpus_matched_tsc is updated outside pvclock_update_vm_gtod_copy though, so a spinlock must be kept for that one.
Documentation/virt/kvm/locking.rst +2 -0
@@ -29,6 +29,8 @@ The acquisition orders for mutexes are as follows: On x86: +- the seqcount kvm->arch.pvclock_sc is written under kvm->lock. + - vcpu->mutex is taken outside kvm->arch.hyperv.hv_lock - kvm->arch.mmu_lock is an rwlock. kvm->arch.tdp_mmu_pages_lock is
arch/x86/include/asm/kvm_host.h +7 -3
@@ -68,8 +68,7 @@#define KVM_REQ_PMI KVM_ARCH_REQ(11)#define KVM_REQ_SMI KVM_ARCH_REQ(12)#define KVM_REQ_MASTERCLOCK_UPDATE KVM_ARCH_REQ(13) -#define KVM_REQ_MCLOCK_INPROGRESS \ - KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) +/* 14 unused */ #define KVM_REQ_SCAN_IOAPIC \ KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)#define KVM_REQ_GLOBAL_CLOCK_UPDATE KVM_ARCH_REQ(16)
@@ -1067,6 +1066,11 @@ struct kvm_arch {unsigned long irq_sources_bitmap;s64 kvmclock_offset; + +/* +* This also protects nr_vcpus_matched_tsc which is read from a +* preemption-disabled region, so it must be a raw spinlock. +*/ raw_spinlock_t tsc_write_lock;u64 last_tsc_nsec;u64 last_tsc_write;
@@ -1077,7 +1081,7 @@ struct kvm_arch {u64 cur_tsc_generation;int nr_vcpus_matched_tsc; -spinlock_t pvclock_gtod_sync_lock; +seqcount_mutex_t pvclock_sc;bool use_master_clock;u64 master_kernel_ns;u64 master_cycle_now;
arch/x86/kvm/x86.c +53 -41
183 lines not on this page · raw message
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Oliver Upton
Reviewed-byRe: [PATCH v6 01/21] KVM: x86: Fix potential race in KVM_GET_CLOCK
Hi Paolo,
Quoted 16 lines · Paolo Bonzini
Quoted 5 lines · Oliver Upton
Sean noticed that
KVM_GET_CLOCKwas checking kvm_arch.use_master_clock outside of the pvclock sync lock. This is problematic, as the clock value written to the user may or may not actually correspond to a stable TSC.Quoted 8 lines
Fix the race by populating the entire
kvm_clock_datastructure behind thepvclock_gtod_sync_lock.Suggested-by Sean Christopherson <[email protected]>Signed-off-by Oliver Upton <[email protected]>--- arch/x86/kvm/x86.c | 39 ++++++++++++++++++++++++++++----------- 1 file changed, 28 insertions(+), 11 deletions(-)
Quoted 3 lines
I had a completely independent patch that fixed the same race. It unifies the read sides of
tsc_write_lockandpvclock_gtod_sync_lockinto a seqcount (and replacespvclock_gtod_sync_lockwithkvm->lockon the write side).
Might it make sense to fix this issue under the existing locking scheme, then shift to what you're proposing? I say that, but the locking change in 03/21 would most certainly have a short lifetime until this patch supersedes it.
Quoted 263 lines
I attach it now (based on https://lore.kernel.org/kvm/[email protected]/T/#t),
but the testing was extremely light so I'm not sure I will be able to include
it in 5.15.Paolo
-------------- 8< -------------
From: Paolo Bonzini <[email protected]>
Date: Thu, 8 Apr 2021 05:03:44 -0400
Subject: [PATCH] kvm: x86: protect masterclock with a seqcountProtect the reference point for kvmclock with a seqcount, so that kvmclock updates for all vCPUs can proceed in parallel. Xen runstate updates will also run in parallel and not bounce the kvmclock cacheline.
This also makes it possible to use
KVM_REQ_CLOCK_UPDATE(which will block on the seqcount) to prevent entering in the guests untilpvclock_update_vm_gtod_copyis complete, and thus to get rid ofKVM_REQ_MCLOCK_INPROGRESS.
nr_vcpus_matched_tscis updated outsidepvclock_update_vm_gtod_copythough, so a spinlock must be kept for that one.Signed-off-by Paolo Bonzini <[email protected]>Documentation/virt/kvm/locking.rst +2
-0@@ -29,6 +29,8 @@ The acquisition orders for mutexes are as follows: On x86: +- the seqcount kvm->arch.pvclock_sc is written under kvm->lock. + - vcpu->mutex is taken outside kvm->arch.hyperv.hv_lock - kvm->arch.mmu_lock is an rwlock. kvm->arch.tdp_mmu_pages_lock isarch/x86/include/asm/kvm_host.h +7
-3@@ -68,8 +68,7 @@#define KVM_REQ_PMI KVM_ARCH_REQ(11)#define KVM_REQ_SMI KVM_ARCH_REQ(12)#define KVM_REQ_MASTERCLOCK_UPDATE KVM_ARCH_REQ(13) -#define KVM_REQ_MCLOCK_INPROGRESS \ - KVM_ARCH_REQ_FLAGS(14, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP) +/* 14 unused */ #define KVM_REQ_SCAN_IOAPIC \ KVM_ARCH_REQ_FLAGS(15, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)#define KVM_REQ_GLOBAL_CLOCK_UPDATE KVM_ARCH_REQ(16)@@ -1067,6 +1066,11 @@ struct kvm_arch {unsigned long irq_sources_bitmap;s64 kvmclock_offset; + +/* +* This also protects nr_vcpus_matched_tsc which is read from a +* preemption-disabled region, so it must be a raw spinlock. +*/ raw_spinlock_t tsc_write_lock;u64 last_tsc_nsec;u64 last_tsc_write;@@ -1077,7 +1081,7 @@ struct kvm_arch {u64 cur_tsc_generation;int nr_vcpus_matched_tsc; -spinlock_t pvclock_gtod_sync_lock; +seqcount_mutex_t pvclock_sc;bool use_master_clock;u64 master_kernel_ns;u64 master_cycle_now;arch/x86/kvm/x86.c +52
-39@@ -2533,9 +2533,7 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write;kvm_vcpu_write_tsc_offset (vcpu, offset); -raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags); -spin_lock_irqsave (&kvm->arch.pvclock_gtod_sync_lock, flags);if (!matched) { kvm->arch.nr_vcpus_matched_tsc =0 ; }else if (!already_matched) {@@ -2543,7 +2541,7 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) }kvm_track_tsc_matching (vcpu); -spin_unlock_irqrestore (&kvm->arch.pvclock_gtod_sync_lock, flags); +raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags); } static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,@@ -2730,9 +2728,7 @@ static void pvclock_update_vm_gtod_copy(struct kvm *kvm)struct kvm_arch *ka = &kvm->arch;int vclock_mode;bool host_tsc_clocksource, vcpus_matched; - - vcpus_matched = (ka->nr_vcpus_matched_tsc +1 == -atomic_read (&kvm->online_vcpus)); +unsigned long flags;/* * If the host uses TSC clock, then passthrough TSC as stable @@ -2742,9 +2738,14 @@ static void pvclock_update_vm_gtod_copy(struct kvm *kvm) &ka->master_kernel_ns, &ka->master_cycle_now); +raw_spin_lock_irqsave (&kvm->arch.tsc_write_lock, flags); + vcpus_matched = (ka->nr_vcpus_matched_tsc +1 == +atomic_read (&kvm->online_vcpus)); + ka->use_master_clock = host_tsc_clocksource && vcpus_matched && !ka->backwards_tsc_observed && !ka->boot_vcpu_runs_old_kvmclock; +raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags);if (ka->use_master_clock)atomic_set (&kvm_guest_has_master_clock,1 );@@ -2758,25 +2759,26 @@ static void pvclock_update_vm_gtod_copy(struct kvm *kvm)static void kvm_start_pvclock_update (struct kvm *kvm) {struct kvm_arch *ka = &kvm->arch; -kvm_make_all_cpus_request (kvm,KVM_REQ_MCLOCK_INPROGRESS ); -/* no guest entries from this point */ -spin_lock_irq (&ka->pvclock_gtod_sync_lock); +mutex_lock (&kvm->lock); +/* +* write_seqcount_begin disables preemption. This is needed not just +* to avoid livelock, but also because the preempt notifier is a reader +* for ka->pvclock_sc. +*/ +write_seqcount_begin (&ka->pvclock_sc); +kvm_make_all_cpus_request (kvm, +KVM_REQ_CLOCK_UPDATE |KVM_REQUEST_WAIT |KVM_REQUEST_NO_WAKEUP ); + +/* no guest entries from this point until write_seqcount_end */ }static void kvm_end_pvclock_update (struct kvm *kvm) {struct kvm_arch *ka = &kvm->arch; -struct kvm_vcpu *vcpu; -int i; -spin_unlock_irq (&ka->pvclock_gtod_sync_lock); -kvm_for_each_vcpu (i, vcpu, kvm) -kvm_make_request (KVM_REQ_CLOCK_UPDATE , vcpu); - -/* guest entries allowed */ -kvm_for_each_vcpu (i, vcpu, kvm) -kvm_clear_request (KVM_REQ_MCLOCK_INPROGRESS , vcpu); +write_seqcount_end (&ka->pvclock_sc); +mutex_unlock (&kvm->lock); }static void kvm_update_masterclock (struct kvm *kvm)@@ -2787,27 +2789,21 @@ static void kvm_update_masterclock(struct kvm *kvm)kvm_end_pvclock_update (kvm); } -u64 get_kvmclock_ns (struct kvm *kvm) +static u64 __get_kvmclock_ns (struct kvm *kvm) {struct kvm_arch *ka = &kvm->arch;struct pvclock_vcpu_time_info hv_clock; -unsigned long flags;u64 ret; -spin_lock_irqsave (&ka->pvclock_gtod_sync_lock, flags); -if (!ka->use_master_clock) { -spin_unlock_irqrestore (&ka->pvclock_gtod_sync_lock, flags); +if (!ka->use_master_clock)return get_kvmclock_base_ns () + ka->kvmclock_offset; - } - - hv_clock.tsc_timestamp = ka->master_cycle_now; - hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; -spin_unlock_irqrestore (&ka->pvclock_gtod_sync_lock, flags);/* both __this_cpu_read() and rdtsc() should be on the same cpu */ get_cpu ();if (__this_cpu_read (cpu_tsc_khz)) { + hv_clock.tsc_timestamp = ka->master_cycle_now; + hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset;kvm_get_time_scale (NSEC_PER_SEC ,__this_cpu_read (cpu_tsc_khz) *1000LL , &hv_clock.tsc_shift, &hv_clock.tsc_to_system_mul);@@ -2816,6 +2812,19 @@ u64 get_kvmclock_ns(struct kvm *kvm) ret =get_kvmclock_base_ns () + ka->kvmclock_offset;put_cpu (); +return ret; +} + +u64 get_kvmclock_ns (struct kvm *kvm) +{ +struct kvm_arch *ka = &kvm->arch; +unsigned seq; +u64 ret; + +do { + seq =read_seqcount_begin (&ka->pvclock_sc); + ret =__get_kvmclock_ns (kvm); + }while (read_seqcount_retry (&ka->pvclock_sc, seq));return ret; }@@ -2882,6 +2891,7 @@ static void kvm_setup_pvclock_page(struct kvm_vcpu *v,static int kvm_guest_time_update (struct kvm_vcpu *v) {unsigned long flags, tgt_tsc_khz; +unsigned seq;struct kvm_vcpu_arch *vcpu = &v->arch;struct kvm_arch *ka = &v->kvm->arch;s64 kernel_ns;@@ -2896,13 +2906,14 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)* If the host uses TSC clock, then passthrough TSC as stable * to the guest. */ -spin_lock_irqsave (&ka->pvclock_gtod_sync_lock, flags); - use_master_clock = ka->use_master_clock; -if (use_master_clock) { - host_tsc = ka->master_cycle_now; - kernel_ns = ka->master_kernel_ns; - } -spin_unlock_irqrestore (&ka->pvclock_gtod_sync_lock, flags); + seq =read_seqcount_begin (&ka->pvclock_sc); +do { + use_master_clock = ka->use_master_clock; +if (use_master_clock) { + host_tsc = ka->master_cycle_now; + kernel_ns = ka->master_kernel_ns; + } + }while (read_seqcount_retry (&ka->pvclock_sc, seq));/* Keep irq disabled to prevent changes to the clock */ local_irq_save (flags);@@ -6098,11 +6109,13 @@ long kvm_arch_vm_ioctl(struct file *filp, }case KVM_GET_CLOCK : {struct kvm_clock_data user_ns; -u64 now_ns; +unsigned seq; - now_ns =get_kvmclock_ns (kvm); - user_ns.clock = now_ns; - user_ns.flags = kvm->arch.use_master_clock ?KVM_CLOCK_TSC_STABLE :0 ; +do { + seq =read_seqcount_begin (&kvm->arch.pvclock_sc); + user_ns.clock =__get_kvmclock_ns (kvm); + user_ns.flags = kvm->arch.use_master_clock ?KVM_CLOCK_TSC_STABLE :0 ; + }while (read_seqcount_retry (&kvm->arch.pvclock_sc, seq));memset (&user_ns.pad,0 ,sizeof (user_ns.pad)); r = -EFAULT ;
arch/x86/kvm/x86.c int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) · line 11163
Quoted 9 lines
arch/x86/kvm/x86.c +1
-2@@ -11144,8 +11157,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)raw_spin_lock_init (&kvm->arch.tsc_write_lock);mutex_init (&kvm->arch.apic_map_lock); -spin_lock_init (&kvm->arch.pvclock_gtod_sync_lock); - +seqcount_mutex_init (&kvm->arch.pvclock_sc, &kvm->lock); kvm->arch.kvmclock_offset = -get_kvmclock_base_ns ();pvclock_update_vm_gtod_copy (kvm);
This all looks good to me, so:
Definitely supplants 03/21 from my series. If you'd rather take your own for this entire series then I can rework around this patch and resend.
--
Thanks,
Oliver
Paolo Bonzini
maintainerRe: [PATCH v6 01/21] KVM: x86: Fix potential race in KVM_GET_CLOCK
Quoted 4 lines · Oliver Upton
Might it make sense to fix this issue under the existing locking scheme, then shift to what you're proposing? I say that, but the locking change in 03/21 would most certainly have a short lifetime until this patch supersedes it.
Yes, definitely. The seqcount change would definitely go in much later.
Extracting KVM_{GET,SET}_CLOCK to separate function would also be a
patch of its own. Give me a few more days of frantic KVM Forum preparation. :)
Paolo
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 01/21] KVM: x86: Fix potential race in KVM_GET_CLOCK
Quoted 11 lines · Paolo Bonzini
Quoted 4 lines · Oliver Upton
Might it make sense to fix this issue under the existing locking scheme, then shift to what you're proposing? I say that, but the locking change in 03/21 would most certainly have a short lifetime until this patch supersedes it.
Yes, definitely. The seqcount change would definitely go in much later.
Extracting KVM_{GET,SET}_CLOCK to separate function would also be a
patch of its own. Give me a few more days of frantic KVM Forum preparation. :)
Sounds good :-) I'm probably going to send this out once more, in three separate series:
- x86 (no changes, just rebasing)
- arm64 (address some comments, bugs)
- selftests (no changes)
--
Thanks,
Oliver
Quoted 2 lines
Paolo
PATCH v6 2/21 KVM: x86: Report host tsc and realtime values in KVM_GET_CLOCK
In reply to Oliver UptonHandling the migration of TSCs correctly is difficult, in part because Linux does not provide userspace with the ability to retrieve a (TSC, realtime) clock pair for a single instant in time. In lieu of a more convenient facility, KVM can report similar information in the kvm_clock structure.
Provide userspace with a host TSC & realtime pair iff the realtime clock is based on the TSC. If userspace provides KVM_SET_CLOCK with a valid realtime value, advance the KVM clock by the amount of elapsed time. Do not step the KVM clock backwards, though, as it is a monotonic oscillator.
--- Documentation/virt/kvm/api.rst | 42 ++++++++--- arch/x86/include/asm/kvm_host.h | 3 + arch/x86/kvm/x86.c | 127 ++++++++++++++++++-------------- include/uapi/linux/kvm.h | 7 +- 4 files changed, 112 insertions(+), 67 deletions(-)
Documentation/virt/kvm/api.rst +33 -9
@@ -993,20 +993,34 @@ such as migration. When KVM_CAP_ADJUST_CLOCK is passed to KVM_CHECK_EXTENSION, it returns the set of bits that KVM can return in struct kvm_clock_data's flag member. -The only flag defined now is KVM_CLOCK_TSC_STABLE. If set, the returned -value is the exact kvmclock value seen by all VCPUs at the instant -when KVM_GET_CLOCK was called. If clear, the returned value is simply -CLOCK_MONOTONIC plus a constant offset; the offset can be modified -with KVM_SET_CLOCK. KVM will try to make all VCPUs follow this clock, -but the exact value read by each VCPU could differ, because the host -TSC is not stable. +FLAGS: + +KVM_CLOCK_TSC_STABLE. If set, the returned value is the exact kvmclock +value seen by all VCPUs at the instant when KVM_GET_CLOCK was called. +If clear, the returned value is simply CLOCK_MONOTONIC plus a constant +offset; the offset can be modified with KVM_SET_CLOCK. KVM will try +to make all VCPUs follow this clock, but the exact value read by each +VCPU could differ, because the host TSC is not stable. + +KVM_CLOCK_REALTIME. If set, the `realtime` field in the kvm_clock_data +structure is populated with the value of the host's real time +clocksource at the instant when KVM_GET_CLOCK was called. If clear, +the `realtime` field does not contain a value. + +KVM_CLOCK_HOST_TSC. If set, the `host_tsc` field in the kvm_clock_data +structure is populated with the value of the host's timestamp counter (TSC) +at the instant when KVM_GET_CLOCK was called. If clear, the `host_tsc` field +does not contain a value. :: struct kvm_clock_data { __u64 clock; /* kvmclock current value */ __u32 flags; - __u32 pad[9]; + __u32 pad0; + __u64 realtime; + __u64 host_tsc; + __u32 pad[4]; };
@@ -1023,12 +1037,22 @@ Sets the current timestamp of kvmclock to the value specified in its parameter. In conjunction with KVM_GET_CLOCK, it is used to ensure monotonicity on scenarios such as migration. +FLAGS: + +KVM_CLOCK_REALTIME. If set, KVM will compare the value of the `realtime` field +with the value of the host's real time clocksource at the instant when +KVM_SET_CLOCK was called. The difference in elapsed time is added to the final +kvmclock value that will be provided to guests. + :: struct kvm_clock_data { __u64 clock; /* kvmclock current value */ __u32 flags; - __u32 pad[9]; + __u32 pad0; + __u64 realtime; + __u64 host_tsc; + __u32 pad[4]; };
arch/x86/include/asm/kvm_host.h +3 -0
@@ -1926,4 +1926,7 @@ int kvm_cpu_dirty_log_size(void);int alloc_all_memslots_rmaps (struct kvm *kvm); +#define KVM_CLOCK_VALID_FLAGS \ + (KVM_CLOCK_TSC_STABLE | KVM_CLOCK_REALTIME | KVM_CLOCK_HOST_TSC) +#endif /* _ASM_X86_KVM_HOST_H */
arch/x86/kvm/x86.c +70 -57
156 lines not on this page · raw message
include/uapi/linux/kvm.h +6 -1
@@ -1223,11 +1223,16 @@ struct kvm_irqfd {/* Do not use 1, KVM_CHECK_EXTENSION returned it before we had flags. */ #define KVM_CLOCK_TSC_STABLE 2 +#define KVM_CLOCK_REALTIME (1 << 2) +#define KVM_CLOCK_HOST_TSC (1 << 3)struct kvm_clock_data {__u64 clock;__u32 flags; -__u32 pad[9 ]; +__u32 pad0; +__u64 realtime; +__u64 host_tsc; +__u32 pad[4 ]; };/* For KVM_CAP_SW_TLB */
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 3/21 KVM: x86: Take the pvclock sync lock behind the tsc_write_lock
In reply to Oliver UptonA later change requires that the pvclock sync lock be taken while holding the tsc_write_lock. Change the locking in kvm_synchronize_tsc() to align with the requirement to isolate the locking change to its own commit.
--- Documentation/virt/kvm/locking.rst | 11 +++++++++++ arch/x86/kvm/x86.c | 2 +- 2 files changed, 12 insertions(+), 1 deletion(-)
Documentation/virt/kvm/locking.rst +11 -0
@@ -36,6 +36,9 @@ On x86: holding kvm->arch.mmu_lock (typically with ``read_lock``, otherwise there's no need to take kvm->arch.tdp_mmu_pages_lock at all). +- kvm->arch.tsc_write_lock is taken outside + kvm->arch.pvclock_gtod_sync_lock + Everything else is a leaf: no other lock is taken inside the critical sections.
@@ -222,6 +225,14 @@ time it will be set using the Dirty tracking mechanism described above. :Comment: 'raw' because hardware enabling/disabling must be atomic /wrt migration. +:Name: kvm_arch::pvclock_gtod_sync_lock +:Type: raw_spinlock_t +:Arch: x86 +:Protects: kvm_arch::{cur_tsc_generation,cur_tsc_nsec,cur_tsc_write, + cur_tsc_offset,nr_vcpus_matched_tsc} +:Comment: 'raw' because updating the kvm master clock must not be + preempted. + :Name: kvm_arch::tsc_write_lock :Type: raw_spinlock :Arch: x86
arch/x86/kvm/x86.c +1 -1
@@ -2529,7 +2529,6 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write;kvm_vcpu_write_tsc_offset (vcpu, offset); -raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags);spin_lock_irqsave (&kvm->arch.pvclock_gtod_sync_lock, flags);if (!matched) {
@@ -2540,6 +2539,7 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data)kvm_track_tsc_matching (vcpu);spin_unlock_irqrestore (&kvm->arch.pvclock_gtod_sync_lock, flags); +raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags); } static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 4/21 KVM: x86: Refactor tsc synchronization code
In reply to Oliver UptonRefactor kvm_synchronize_tsc to make a new function that allows callers to specify TSC parameters (offset, value, nanoseconds, etc.) explicitly for the sake of participating in TSC synchronization.
--- arch/x86/kvm/x86.c | 105 ++++++++++++++++++++++++++------------------- 1 file changed, 61 insertions(+), 44 deletions(-)
arch/x86/kvm/x86.c +61 -44
@@ -2443,13 +2443,71 @@ static inline bool kvm_check_tsc_unstable(void)return check_tsc_unstable (); } +/* +* Infers attempts to synchronize the guest's tsc from host writes. Sets the +* offset for the vcpu and tracks the TSC matching generation that the vcpu +* participates in. +*/ +static void __kvm_synchronize_tsc (struct kvm_vcpu *vcpu,u64 offset,u64 tsc, +u64 ns,bool matched) +{ +struct kvm *kvm = vcpu->kvm; +bool already_matched; + +lockdep_assert_held (&kvm->arch.tsc_write_lock); + + already_matched = + (vcpu->arch.this_tsc_generation == kvm->arch.cur_tsc_generation); + +/* +* We track the most recent recorded KHZ, write and time to +* allow the matching interval to be extended at each write. +*/ + kvm->arch.last_tsc_nsec = ns; + kvm->arch.last_tsc_write = tsc; + kvm->arch.last_tsc_khz = vcpu->arch.virtual_tsc_khz; + + vcpu->arch.last_guest_tsc = tsc; + +/* Keep track of which generation this VCPU has synchronized to */ + vcpu->arch.this_tsc_generation = kvm->arch.cur_tsc_generation; + vcpu->arch.this_tsc_nsec = kvm->arch.cur_tsc_nsec; + vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write; + +kvm_vcpu_write_tsc_offset (vcpu, offset); + +if (!matched) { +/* +* We split periods of matched TSC writes into generations. +* For each generation, we track the original measured +* nanosecond time, offset, and write, so if TSCs are in +* sync, we can match exact offset, and if not, we can match +* exact software computation in compute_guest_tsc() +* +* These values are tracked in kvm->arch.cur_xxx variables. +*/ + kvm->arch.cur_tsc_generation++; + kvm->arch.cur_tsc_nsec = ns; + kvm->arch.cur_tsc_write = tsc; + kvm->arch.cur_tsc_offset = offset; + +spin_lock (&kvm->arch.pvclock_gtod_sync_lock); + kvm->arch.nr_vcpus_matched_tsc =0 ; + }else if (!already_matched) { +spin_lock (&kvm->arch.pvclock_gtod_sync_lock); + kvm->arch.nr_vcpus_matched_tsc++; + } + +kvm_track_tsc_matching (vcpu); +spin_unlock (&kvm->arch.pvclock_gtod_sync_lock); +} +static void kvm_synchronize_tsc (struct kvm_vcpu *vcpu,u64 data) {struct kvm *kvm = vcpu->kvm;u64 offset, ns, elapsed;unsigned long flags; -bool matched; -bool already_matched; +bool matched = false;bool synchronizing = false;raw_spin_lock_irqsave (&kvm->arch.tsc_write_lock, flags);
@@ -2495,50 +2553,9 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) offset =kvm_compute_l1_tsc_offset (vcpu, data); } matched = true; - already_matched = (vcpu->arch.this_tsc_generation == kvm->arch.cur_tsc_generation); - }else { -/* -* We split periods of matched TSC writes into generations. -* For each generation, we track the original measured -* nanosecond time, offset, and write, so if TSCs are in -* sync, we can match exact offset, and if not, we can match -* exact software computation in compute_guest_tsc() -* -* These values are tracked in kvm->arch.cur_xxx variables. -*/ - kvm->arch.cur_tsc_generation++; - kvm->arch.cur_tsc_nsec = ns; - kvm->arch.cur_tsc_write = data; - kvm->arch.cur_tsc_offset = offset; - matched = false; } -/* -* We also track th most recent recorded KHZ, write and time to -* allow the matching interval to be extended at each write. -*/ - kvm->arch.last_tsc_nsec = ns; - kvm->arch.last_tsc_write = data; - kvm->arch.last_tsc_khz = vcpu->arch.virtual_tsc_khz; - - vcpu->arch.last_guest_tsc = data; - -/* Keep track of which generation this VCPU has synchronized to */ - vcpu->arch.this_tsc_generation = kvm->arch.cur_tsc_generation; - vcpu->arch.this_tsc_nsec = kvm->arch.cur_tsc_nsec; - vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write; - -kvm_vcpu_write_tsc_offset (vcpu, offset); - -spin_lock_irqsave (&kvm->arch.pvclock_gtod_sync_lock, flags); -if (!matched) { - kvm->arch.nr_vcpus_matched_tsc =0 ; - }else if (!already_matched) { - kvm->arch.nr_vcpus_matched_tsc++; - } - -kvm_track_tsc_matching (vcpu); -spin_unlock_irqrestore (&kvm->arch.pvclock_gtod_sync_lock, flags); +__kvm_synchronize_tsc (vcpu, offset, data, ns, matched);raw_spin_unlock_irqrestore (&kvm->arch.tsc_write_lock, flags); }
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 5/21 KVM: x86: Expose TSC offset controls to userspace
In reply to Oliver UptonTo date, VMM-directed TSC synchronization and migration has been a bit messy. KVM has some baked-in heuristics around TSC writes to infer if the VMM is attempting to synchronize. This is problematic, as it depends on host userspace writing to the guest's TSC within 1 second of the last write.
A much cleaner approach to configuring the guest's views of the TSC is to simply migrate the TSC offset for every vCPU. Offsets are idempotent, and thus not subject to change depending on when the VMM actually reads/writes values from/to KVM. The VMM can then read the TSC once with KVM_GET_CLOCK to capture a (realtime, host_tsc) pair at the instant when the guest is paused.
--- Documentation/virt/kvm/devices/vcpu.rst | 57 +++++++++++++ arch/x86/include/asm/kvm_host.h | 1 + arch/x86/include/uapi/asm/kvm.h | 4 + arch/x86/kvm/x86.c | 109 ++++++++++++++++++++++++ 4 files changed, 171 insertions(+)
Documentation/virt/kvm/devices/vcpu.rst +57 -0
@@ -161,3 +161,60 @@ Specifies the base address of the stolen time structure for this VCPU. The base address must be 64 byte aligned and exist within a valid guest memory region. See Documentation/virt/kvm/arm/pvtime.rst for more information including the layout of the stolen time structure. + +4. GROUP: KVM_VCPU_TSC_CTRL +=========================== + +:Architectures: x86 + +4.1 ATTRIBUTE: KVM_VCPU_TSC_OFFSET + +:Parameters: 64-bit unsigned TSC offset + +Returns: + + ======= ====================================== + -EFAULT Error reading/writing the provided + parameter address. + -ENXIO Attribute not supported + ======= ====================================== + +Specifies the guest's TSC offset relative to the host's TSC. The guest's +TSC is then derived by the following equation: + + guest_tsc = host_tsc + KVM_VCPU_TSC_OFFSET + +This attribute is useful for the precise migration of a guest's TSC. The +following describes a possible algorithm to use for the migration of a +guest's TSC: + +From the source VMM process: + +1. Invoke the KVM_GET_CLOCK ioctl to record the host TSC (t_0), + kvmclock nanoseconds (k_0), and realtime nanoseconds (r_0). + +2. Read the KVM_VCPU_TSC_OFFSET attribute for every vCPU to record the + guest TSC offset (off_n). + +3. Invoke the KVM_GET_TSC_KHZ ioctl to record the frequency of the + guest's TSC (freq). + +From the destination VMM process: + +4. Invoke the KVM_SET_CLOCK ioctl, providing the kvmclock nanoseconds + (k_0) and realtime nanoseconds (r_0) in their respective fields. + Ensure that the KVM_CLOCK_REALTIME flag is set in the provided + structure. KVM will advance the VM's kvmclock to account for elapsed + time since recording the clock values. + +5. Invoke the KVM_GET_CLOCK ioctl to record the host TSC (t_1) and + kvmclock nanoseconds (k_1). + +6. Adjust the guest TSC offsets for every vCPU to account for (1) time + elapsed since recording state and (2) difference in TSCs between the + source and destination machine: + + new_off_n = t_0 + off_n + (k_1 - k_0) * freq - t_1 + +7. Write the KVM_VCPU_TSC_OFFSET attribute for every vCPU with the + respective value derived in the previous step.
arch/x86/include/asm/kvm_host.h +1 -0
@@ -1064,6 +1064,7 @@ struct kvm_arch {u64 last_tsc_nsec;u64 last_tsc_write;u32 last_tsc_khz; +u64 last_tsc_offset;u64 cur_tsc_nsec;u64 cur_tsc_write;u64 cur_tsc_offset;
arch/x86/include/uapi/asm/kvm.h +4 -0
@@ -503,4 +503,8 @@ struct kvm_pmu_event_filter {#define KVM_PMU_EVENT_ALLOW 0#define KVM_PMU_EVENT_DENY 1 +/* for KVM_{GET,SET,HAS}_DEVICE_ATTR */ +#define KVM_VCPU_TSC_CTRL 0/* control group for the timestamp counter (TSC) */ +#define KVM_VCPU_TSC_OFFSET 0/* attribute for the TSC offset */ +#endif /* _ASM_X86_KVM_H */
arch/x86/kvm/x86.c +109 -0
127 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 6/21 tools: arch: x86: pull in pvclock headers
In reply to Oliver UptonCopy over approximately clean versions of the pvclock headers into tools. Reconcile headers/symbols missing in tools that are unneeded.
--- tools/arch/x86/include/asm/pvclock-abi.h | 48 +++++++++++ tools/arch/x86/include/asm/pvclock.h | 103 +++++++++++++++++++++++ 2 files changed, 151 insertions(+) create mode 100644 tools/arch/x86/include/asm/pvclock-abi.h create mode 100644 tools/arch/x86/include/asm/pvclock.h
tools/arch/x86/include/asm/pvclock-abi.h new file +48 -0
@@ -0,0 +1,48 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _ASM_X86_PVCLOCK_ABI_H +#define _ASM_X86_PVCLOCK_ABI_H +#ifndef __ASSEMBLY__ + +/* +* These structs MUST NOT be changed. +* They are the ABI between hypervisor and guest OS. +* Both Xen and KVM are using this. +* +* pvclock_vcpu_time_info holds the system time and the tsc timestamp +* of the last update. So the guest can use the tsc delta to get a +* more precise system time. There is one per virtual cpu. +* +* pvclock_wall_clock references the point in time when the system +* time was zero (usually boot time), thus the guest calculates the +* current wall clock by adding the system time. +* +* Protocol for the "version" fields is: hypervisor raises it (making +* it uneven) before it starts updating the fields and raises it again +* (making it even) when it is done. Thus the guest can make sure the +* time values it got are consistent by checking the version before +* and after reading them. +*/ + +struct pvclock_vcpu_time_info { +u32 version; +u32 pad0; +u64 tsc_timestamp; +u64 system_time; +u32 tsc_to_system_mul; +s8 tsc_shift; +u8 flags; +u8 pad[2 ]; +} __attribute__((__packed__ ));/* 32 bytes */ + +struct pvclock_wall_clock { +u32 version; +u32 sec; +u32 nsec; +} __attribute__((__packed__ )); + +#define PVCLOCK_TSC_STABLE_BIT (1 << 0) +#define PVCLOCK_GUEST_STOPPED (1 << 1) +/* PVCLOCK_COUNTS_FROM_ZERO broke ABI and can't be used anymore. */ +#define PVCLOCK_COUNTS_FROM_ZERO (1 << 2) +#endif /* __ASSEMBLY__ */ +#endif /* _ASM_X86_PVCLOCK_ABI_H */
tools/arch/x86/include/asm/pvclock.h new file +103 -0
103 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 7/21 selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
In reply to Oliver UptonAdd a selftest for the new KVM clock UAPI that was introduced. Ensure that the KVM clock is consistent between userspace and the guest, and that the difference in realtime will only ever cause the KVM clock to advance forward.
--- tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 1 + .../testing/selftests/kvm/include/kvm_util.h | 2 + .../selftests/kvm/x86_64/kvm_clock_test.c | 204 ++++++++++++++++++ 4 files changed, 208 insertions(+) create mode 100644 tools/testing/selftests/kvm/x86_64/kvm_clock_test.c
tools/testing/selftests/kvm/.gitignore +1 -0
@@ -11,6 +11,7 @@ /x86_64/emulator_error_test /x86_64/get_cpuid_test /x86_64/get_msr_index_features +/x86_64/kvm_clock_test /x86_64/kvm_pv_test /x86_64/hyperv_clock /x86_64/hyperv_cpuid
tools/testing/selftests/kvm/Makefile +1 -0
@@ -46,6 +46,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/get_cpuid_test TEST_GEN_PROGS_x86_64 += x86_64/hyperv_clock TEST_GEN_PROGS_x86_64 += x86_64/hyperv_cpuid TEST_GEN_PROGS_x86_64 += x86_64/hyperv_features +TEST_GEN_PROGS_x86_64 += x86_64/kvm_clock_test TEST_GEN_PROGS_x86_64 += x86_64/kvm_pv_test TEST_GEN_PROGS_x86_64 += x86_64/mmio_warning_test TEST_GEN_PROGS_x86_64 += x86_64/mmu_role_test
tools/testing/selftests/kvm/include/kvm_util.h +2 -0
@@ -19,6 +19,8 @@#define KVM_DEV_PATH "/dev/kvm"#define KVM_MAX_VCPUS 512 +#define NSEC_PER_SEC 1000000000L +/* * Callers of kvm_util only have an incomplete/opaque description of the * structure kvm_util is using to maintain the state of a VM.
tools/testing/selftests/kvm/x86_64/kvm_clock_test.c new file +204 -0
204 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 8/21 selftests: KVM: Fix kvm device helper ioctl assertions
In reply to Oliver UptonThe KVM_CREATE_DEVICE and KVM_{GET,SET}_DEVICE_ATTR ioctls are defined to return a value of zero on success. As such, tighten the assertions in the helper functions to only pass if the return code is zero.
--- tools/testing/selftests/kvm/lib/kvm_util.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
tools/testing/selftests/kvm/lib/kvm_util.c +3 -3
@@ -1984,7 +1984,7 @@ int kvm_device_check_attr(int dev_fd, uint32_t group, uint64_t attr) {int ret =_kvm_device_check_attr (dev_fd, group, attr); -TEST_ASSERT (ret >=0 ,"KVM_HAS_DEVICE_ATTR failed, rc: %i errno: %i" , ret, errno); +TEST_ASSERT (!ret,"KVM_HAS_DEVICE_ATTR failed, rc: %i errno: %i" , ret, errno);return ret; }
@@ -2008,7 +2008,7 @@ int kvm_create_device(struct kvm_vm *vm, uint64_t type, bool test) ret =_kvm_create_device (vm, type, test, &fd);if (!test) { -TEST_ASSERT (ret >=0 , +TEST_ASSERT (!ret,"KVM_CREATE_DEVICE IOCTL failed, rc: %i errno: %i" , ret, errno);return fd; }
@@ -2036,7 +2036,7 @@ int kvm_device_access(int dev_fd, uint32_t group, uint64_t attr, {int ret =_kvm_device_access (dev_fd, group, attr, val, write); -TEST_ASSERT (ret >=0 ,"KVM_SET|GET_DEVICE_ATTR IOCTL failed, rc: %i errno: %i" , ret, errno); +TEST_ASSERT (!ret,"KVM_SET|GET_DEVICE_ATTR IOCTL failed, rc: %i errno: %i" , ret, errno);return ret; }
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 9/21 selftests: KVM: Add helpers for vCPU device attributes
In reply to Oliver UptonvCPU file descriptors are abstracted away from test code in KVM selftests, meaning that tests cannot directly access a vCPU's device attributes. Add helpers that tests can use to get at vCPU device attributes.
--- .../testing/selftests/kvm/include/kvm_util.h | 9 +++++ tools/testing/selftests/kvm/lib/kvm_util.c | 38 +++++++++++++++++++ 2 files changed, 47 insertions(+)
tools/testing/selftests/kvm/include/kvm_util.h +9 -0
@@ -240,6 +240,15 @@ int _kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,int kvm_device_access (int dev_fd,uint32_t group,uint64_t attr,void *val,bool write); +int _vcpu_has_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr); +int vcpu_has_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr); +int _vcpu_access_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr,void *val,bool write); +int vcpu_access_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr,void *val,bool write); +const char *exit_reason_str (unsigned int exit_reason);void virt_pgd_alloc (struct kvm_vm *vm);
tools/testing/selftests/kvm/lib/kvm_util.c +38 -0
@@ -2040,6 +2040,44 @@ int kvm_device_access(int dev_fd, uint32_t group, uint64_t attr,return ret; } +int _vcpu_has_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr) +{ +struct vcpu *vcpu =vcpu_find (vm, vcpuid); + +TEST_ASSERT (vcpu,"nonexistent vcpu id: %d" , vcpuid); + +return _kvm_device_check_attr (vcpu->fd, group, attr); +} + +int vcpu_has_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr) +{ +int ret =_vcpu_has_device_attr (vm, vcpuid, group, attr); + +TEST_ASSERT (!ret,"KVM_HAS_DEVICE_ATTR IOCTL failed, rc: %i errno: %i" , ret, errno); +return ret; +} + +int _vcpu_access_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr,void *val,bool write) +{ +struct vcpu *vcpu =vcpu_find (vm, vcpuid); + +TEST_ASSERT (vcpu,"nonexistent vcpu id: %d" , vcpuid); + +return _kvm_device_access (vcpu->fd, group, attr, val, write); +} + +int vcpu_access_device_attr (struct kvm_vm *vm,uint32_t vcpuid,uint32_t group, +uint64_t attr,void *val,bool write) +{ +int ret =_vcpu_access_device_attr (vm, vcpuid, group, attr, val, write); + +TEST_ASSERT (!ret,"KVM_SET|GET_DEVICE_ATTR IOCTL failed, rc: %i errno: %i" , ret, errno); +return ret; +} +/* * VM Dump *
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 10/21 selftests: KVM: Introduce system counter offset test
In reply to Oliver UptonIntroduce a KVM selftest to verify that userspace manipulation of the TSC (via the new vCPU attribute) results in the correct behavior within the guest.
--- tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 1 + .../kvm/system_counter_offset_test.c | 132 ++++++++++++++++++ 3 files changed, 134 insertions(+) create mode 100644 tools/testing/selftests/kvm/system_counter_offset_test.c
tools/testing/selftests/kvm/.gitignore +1 -0
@@ -52,3 +52,4 @@ /set_memory_region_test /steal_time /kvm_binary_stats_test +/system_counter_offset_test
tools/testing/selftests/kvm/Makefile +1 -0
@@ -85,6 +85,7 @@ TEST_GEN_PROGS_x86_64 += memslot_perf_test TEST_GEN_PROGS_x86_64 += set_memory_region_test TEST_GEN_PROGS_x86_64 += steal_time TEST_GEN_PROGS_x86_64 += kvm_binary_stats_test +TEST_GEN_PROGS_x86_64 += system_counter_offset_test TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list
tools/testing/selftests/kvm/system_counter_offset_test.c new file +132 -0
132 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 11/21 KVM: arm64: Refactor update_vtimer_cntvoff()
In reply to Oliver UptonMake the implementation of update_vtimer_cntvoff() generic w.r.t. guest timer context and spin off into a new helper method for later use. Require callers of this new helper method to grab the kvm lock beforehand.
No functional change intended.
--- arch/arm64/kvm/arch_timer.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)
arch/arm64/kvm/arch_timer.c +15 -5
@@ -747,22 +747,32 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)return 0 ; } -/* Make the updates of cntvoff for all vtimer contexts atomic */ -static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff) +/* Make offset updates for all timer contexts atomic */ +static void update_timer_offset (struct kvm_vcpu *vcpu, +enum kvm_arch_timers timer,u64 offset) {int i;struct kvm *kvm = vcpu->kvm;struct kvm_vcpu *tmp; -mutex_lock (&kvm->lock); +lockdep_assert_held (&kvm->lock); +kvm_for_each_vcpu (i, tmp, kvm) -timer_set_offset (vcpu_vtimer (tmp), cntvoff); +timer_set_offset (vcpu_get_timer (tmp, timer), offset);/* * When called from the vcpu create path, the CPU being created is not * included in the loop above, so we just set it here as well. */ -timer_set_offset (vcpu_vtimer (vcpu), cntvoff); +timer_set_offset (vcpu_get_timer (vcpu, timer), offset); +} + +static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff) +{ +struct kvm *kvm = vcpu->kvm; + +mutex_lock (&kvm->lock); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff);mutex_unlock (&kvm->lock); }
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 11/21] KVM: arm64: Refactor update_vtimer_cntvoff()
Quoted 57 lines · Oliver Upton
Make the implementation of
update_vtimer_cntvoff()generic w.r.t. guest timer context and spin off into a new helper method for later use. Require callers of this new helper method to grab the kvm lock beforehand.No functional change intended.
Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/kvm/arch_timer.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-)arch/arm64/kvm/arch_timer.c +15
-5@@ -747,22 +747,32 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)return 0 ; } -/* Make the updates of cntvoff for all vtimer contexts atomic */ -static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff) +/* Make offset updates for all timer contexts atomic */ +static void update_timer_offset (struct kvm_vcpu *vcpu, +enum kvm_arch_timers timer,u64 offset) {int i;struct kvm *kvm = vcpu->kvm;struct kvm_vcpu *tmp; -mutex_lock (&kvm->lock); +lockdep_assert_held (&kvm->lock); +kvm_for_each_vcpu (i, tmp, kvm) -timer_set_offset (vcpu_vtimer (tmp), cntvoff); +timer_set_offset (vcpu_get_timer (tmp, timer), offset);/* * When called from the vcpu create path, the CPU being created is not * included in the loop above, so we just set it here as well. */ -timer_set_offset (vcpu_vtimer (vcpu), cntvoff); +timer_set_offset (vcpu_get_timer (vcpu, timer), offset); +} + +static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff) +{ +struct kvm *kvm = vcpu->kvm; + +mutex_lock (&kvm->lock); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff);mutex_unlock (&kvm->lock); }Signature · 1 line
2.32.0.605.g8dce9f2422-goog
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
PATCH v6 12/21 KVM: arm64: Separate guest/host counter offset values
In reply to Oliver UptonIn some instances, a VMM may want to update the guest's counter-timer offset in a transparent manner, meaning that changes to the hardware value do not affect the synthetic register presented to the guest or the VMM through said guest's architectural state. Lay the groundwork to separate guest offset register writes from the hardware values utilized by KVM.
--- arch/arm64/kvm/arch_timer.c | 48 ++++++++++++++++++++++++++++++++---- include/kvm/arm_arch_timer.h | 3 +++ 2 files changed, 46 insertions(+), 5 deletions(-)
arch/arm64/kvm/arch_timer.c +43 -5
@@ -87,6 +87,18 @@ static u64 timer_get_offset(struct arch_timer_context *ctxt)struct kvm_vcpu *vcpu = ctxt->vcpu;switch (arch_timer_ctx_index (ctxt)) { +case TIMER_VTIMER : +return ctxt->host_offset; +default : +return 0 ; + } +} + +static u64 timer_get_guest_offset (struct arch_timer_context *ctxt) +{ +struct kvm_vcpu *vcpu = ctxt->vcpu; + +switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER :return __vcpu_sys_reg (vcpu,CNTVOFF_EL2 );default :
@@ -132,13 +144,31 @@ static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : -__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ) = offset; + ctxt->host_offset = offset;break ;default :WARN (offset,"timer %ld\n" ,arch_timer_ctx_index (ctxt)); } } +static void timer_set_guest_offset (struct arch_timer_context *ctxt,u64 offset) +{ +struct kvm_vcpu *vcpu = ctxt->vcpu; + +switch (arch_timer_ctx_index (ctxt)) { +case TIMER_VTIMER : { +u64 host_offset =timer_get_offset (ctxt); + + host_offset += offset -__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ); +__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ) = offset; +timer_set_offset (ctxt, host_offset); +break ; + } +default : +WARN_ONCE (offset,"timer %ld\n" ,arch_timer_ctx_index (ctxt)); + } +} +u64 kvm_phys_timer_read (void ) {return timecounter->cc->read (timecounter->cc);
@@ -749,7 +779,8 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)/* Make offset updates for all timer contexts atomic */ static void update_timer_offset (struct kvm_vcpu *vcpu, -enum kvm_arch_timers timer,u64 offset) +enum kvm_arch_timers timer,u64 offset, +bool guest_visible) {int i;struct kvm *kvm = vcpu->kvm;
@@ -758,13 +789,20 @@ static void update_timer_offset(struct kvm_vcpu *vcpu,lockdep_assert_held (&kvm->lock);kvm_for_each_vcpu (i, tmp, kvm) -timer_set_offset (vcpu_get_timer (tmp, timer), offset); +if (guest_visible) +timer_set_guest_offset (vcpu_get_timer (tmp, timer), + offset); +else +timer_set_offset (vcpu_get_timer (tmp, timer), offset);/* * When called from the vcpu create path, the CPU being created is not * included in the loop above, so we just set it here as well. */ -timer_set_offset (vcpu_get_timer (vcpu, timer), offset); +if (guest_visible) +timer_set_guest_offset (vcpu_get_timer (vcpu, timer), offset); +else +timer_set_offset (vcpu_get_timer (vcpu, timer), offset); }static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff)
@@ -772,7 +810,7 @@ static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)struct kvm *kvm = vcpu->kvm;mutex_lock (&kvm->lock); -update_timer_offset (vcpu,TIMER_VTIMER , cntvoff); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff, true);mutex_unlock (&kvm->lock); }
include/kvm/arm_arch_timer.h +3 -0
@@ -42,6 +42,9 @@ struct arch_timer_context {/* Duplicated state from arch_timer.c for convenience */ u32 host_timer_irq;u32 host_timer_irq_flags; + +/* offset relative to the host's physical counter-timer */ +u64 host_offset; };struct timer_map {
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 12/21] KVM: arm64: Separate guest/host counter offset values
Quoted 128 lines · Oliver Upton
In some instances, a VMM may want to update the guest's counter-timer offset in a transparent manner, meaning that changes to the hardware value do not affect the synthetic register presented to the guest or the VMM through said guest's architectural state. Lay the groundwork to separate guest offset register writes from the hardware values utilized by KVM.
Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/kvm/arch_timer.c | 48 ++++++++++++++++++++++++++++++++---- include/kvm/arm_arch_timer.h | 3 +++ 2 files changed, 46 insertions(+), 5 deletions(-)arch/arm64/kvm/arch_timer.c +43
-5@@ -87,6 +87,18 @@ static u64 timer_get_offset(struct arch_timer_context *ctxt)struct kvm_vcpu *vcpu = ctxt->vcpu;switch (arch_timer_ctx_index (ctxt)) { +case TIMER_VTIMER : +return ctxt->host_offset; +default : +return 0 ; + } +} + +static u64 timer_get_guest_offset (struct arch_timer_context *ctxt) +{ +struct kvm_vcpu *vcpu = ctxt->vcpu; + +switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER :return __vcpu_sys_reg (vcpu,CNTVOFF_EL2 );default :@@ -132,13 +144,31 @@ static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : -__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ) = offset; + ctxt->host_offset = offset;break ;default :WARN (offset,"timer %ld\n" ,arch_timer_ctx_index (ctxt)); } } +static void timer_set_guest_offset (struct arch_timer_context *ctxt,u64 offset) +{ +struct kvm_vcpu *vcpu = ctxt->vcpu; + +switch (arch_timer_ctx_index (ctxt)) { +case TIMER_VTIMER : { +u64 host_offset =timer_get_offset (ctxt); + + host_offset += offset -__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ); +__vcpu_sys_reg (vcpu,CNTVOFF_EL2 ) = offset; +timer_set_offset (ctxt, host_offset); +break ; + } +default : +WARN_ONCE (offset,"timer %ld\n" ,arch_timer_ctx_index (ctxt)); + } +} +u64 kvm_phys_timer_read (void ) {return timecounter->cc->read (timecounter->cc);@@ -749,7 +779,8 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)/* Make offset updates for all timer contexts atomic */ static void update_timer_offset (struct kvm_vcpu *vcpu, -enum kvm_arch_timers timer,u64 offset) +enum kvm_arch_timers timer,u64 offset, +bool guest_visible) {int i;struct kvm *kvm = vcpu->kvm;@@ -758,13 +789,20 @@ static void update_timer_offset(struct kvm_vcpu *vcpu,lockdep_assert_held (&kvm->lock);kvm_for_each_vcpu (i, tmp, kvm) -timer_set_offset (vcpu_get_timer (tmp, timer), offset); +if (guest_visible) +timer_set_guest_offset (vcpu_get_timer (tmp, timer), + offset); +else +timer_set_offset (vcpu_get_timer (tmp, timer), offset);/* * When called from the vcpu create path, the CPU being created is not * included in the loop above, so we just set it here as well. */ -timer_set_offset (vcpu_get_timer (vcpu, timer), offset); +if (guest_visible) +timer_set_guest_offset (vcpu_get_timer (vcpu, timer), offset); +else +timer_set_offset (vcpu_get_timer (vcpu, timer), offset); }static void update_vtimer_cntvoff (struct kvm_vcpu *vcpu,u64 cntvoff)@@ -772,7 +810,7 @@ static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)struct kvm *kvm = vcpu->kvm;mutex_lock (&kvm->lock); -update_timer_offset (vcpu,TIMER_VTIMER , cntvoff); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff, true);mutex_unlock (&kvm->lock); }include/kvm/arm_arch_timer.h +3
-0@@ -42,6 +42,9 @@ struct arch_timer_context {/* Duplicated state from arch_timer.c for convenience */ u32 host_timer_irq;u32 host_timer_irq_flags; + +/* offset relative to the host's physical counter-timer */ +u64 host_offset; };struct timer_map {Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 13/21 KVM: arm64: Allow userspace to configure a vCPU's virtual offset
In reply to Oliver UptonAllow userspace to access the guest's virtual counter-timer offset through the ONE_REG interface. The value read or written is defined to be an offset from the guest's physical counter-timer. Add some documentation to clarify how a VMM should use this and the existing CNTVCT_EL0.
--- Documentation/virt/kvm/api.rst | 10 ++++++++++ arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 11 +++++++++++ arch/arm64/kvm/guest.c | 6 +++++- include/kvm/arm_arch_timer.h | 1 + 5 files changed, 28 insertions(+), 1 deletion(-)
Documentation/virt/kvm/api.rst +10 -0
@@ -2487,6 +2487,16 @@ arm64 system registers have the following id bit patterns:: derived from the register encoding for CNTV_CVAL_EL0. As this is API, it must remain this way. +.. warning:: + + The value of KVM_REG_ARM_TIMER_OFFSET is defined as an offset from + the guest's view of the physical counter-timer. + + Userspace should use either KVM_REG_ARM_TIMER_OFFSET or + KVM_REG_ARM_TIMER_CVAL to pause and resume a guest's virtual + counter-timer. Mixed use of these registers could result in an + unpredictable guest counter value. + arm64 firmware pseudo-registers have the following bit pattern:: 0x6030 0000 0014 <regno:16>
arch/arm64/include/uapi/asm/kvm.h +1 -0
@@ -255,6 +255,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_REG_ARM_TIMER_CTL ARM64_SYS_REG(3, 3, 14, 3, 1)#define KVM_REG_ARM_TIMER_CVAL ARM64_SYS_REG(3, 3, 14, 0, 2)#define KVM_REG_ARM_TIMER_CNT ARM64_SYS_REG(3, 3, 14, 3, 2) +#define KVM_REG_ARM_TIMER_OFFSET ARM64_SYS_REG(3, 4, 14, 0, 3)/* KVM-as-firmware specific pseudo-registers */ #define KVM_REG_ARM_FW (0x0014 << KVM_REG_ARM_COPROC_SHIFT)
arch/arm64/kvm/arch_timer.c +11 -0
@@ -868,6 +868,10 @@ int kvm_arm_timer_set_reg(struct kvm_vcpu *vcpu, u64 regid, u64 value) timer =vcpu_vtimer (vcpu);kvm_arm_timer_write (vcpu, timer,TIMER_REG_CVAL , value);break ; +case KVM_REG_ARM_TIMER_OFFSET : + timer =vcpu_vtimer (vcpu); +update_vtimer_cntvoff (vcpu, value); +break ;case KVM_REG_ARM_PTIMER_CTL : timer =vcpu_ptimer (vcpu);kvm_arm_timer_write (vcpu, timer,TIMER_REG_CTL , value);
@@ -912,6 +916,9 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid)case KVM_REG_ARM_TIMER_CVAL :return kvm_arm_timer_read (vcpu,vcpu_vtimer (vcpu),TIMER_REG_CVAL ); +case KVM_REG_ARM_TIMER_OFFSET : +return kvm_arm_timer_read (vcpu, +vcpu_vtimer (vcpu),TIMER_REG_OFFSET );case KVM_REG_ARM_PTIMER_CTL :return kvm_arm_timer_read (vcpu,vcpu_ptimer (vcpu),TIMER_REG_CTL );
@@ -949,6 +956,10 @@ static u64 kvm_arm_timer_read(struct kvm_vcpu *vcpu, val =kvm_phys_timer_read () -timer_get_offset (timer);break ; +case TIMER_REG_OFFSET : + val =timer_get_offset (timer); +break ; +default :BUG (); }
arch/arm64/kvm/guest.c +5 -1
@@ -591,7 +591,7 @@ static unsigned long num_core_regs(const struct kvm_vcpu *vcpu)* ARM64 versions of the TIMER registers, always available on arm64 */ -#define NUM_TIMER_REGS 3 +#define NUM_TIMER_REGS 4static bool is_timer_reg (u64 index) {
@@ -599,6 +599,7 @@ static bool is_timer_reg(u64 index)case KVM_REG_ARM_TIMER_CTL :case KVM_REG_ARM_TIMER_CNT :case KVM_REG_ARM_TIMER_CVAL : +case KVM_REG_ARM_TIMER_OFFSET :return true; }return false;
@@ -614,6 +615,9 @@ static int copy_timer_indices(struct kvm_vcpu *vcpu, u64 __user *uindices) uindices++;if (put_user (KVM_REG_ARM_TIMER_CVAL , uindices))return -EFAULT ; + uindices++; +if (put_user (KVM_REG_ARM_TIMER_OFFSET , uindices)) +return -EFAULT ;return 0 ; }
include/kvm/arm_arch_timer.h +1 -0
@@ -21,6 +21,7 @@ enum kvm_arch_timer_regs {TIMER_REG_CVAL ,TIMER_REG_TVAL ,TIMER_REG_CTL , + TIMER_REG_OFFSET, }; struct arch_timer_context {
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 13/21] KVM: arm64: Allow userspace to configure a vCPU's virtual offset
Quoted 130 lines · Oliver Upton
Allow userspace to access the guest's virtual counter-timer offset through the
ONE_REGinterface. The value read or written is defined to be an offset from the guest's physical counter-timer. Add some documentation to clarify how a VMM should use this and the existingCNTVCT_EL0.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/api.rst | 10 ++++++++++ arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 11 +++++++++++ arch/arm64/kvm/guest.c | 6 +++++- include/kvm/arm_arch_timer.h | 1 + 5 files changed, 28 insertions(+), 1 deletion(-)Documentation/virt/kvm/api.rst +10
-0@@ -2487,6 +2487,16 @@ arm64 system registers have the following id bit patterns:: derived from the register encoding for CNTV_CVAL_EL0. As this is API, it must remain this way. +.. warning:: + + The value of KVM_REG_ARM_TIMER_OFFSET is defined as an offset from + the guest's view of the physical counter-timer. + + Userspace should use either KVM_REG_ARM_TIMER_OFFSET or + KVM_REG_ARM_TIMER_CVAL to pause and resume a guest's virtual + counter-timer. Mixed use of these registers could result in an + unpredictable guest counter value. + arm64 firmware pseudo-registers have the following bit pattern:: 0x6030 0000 0014 <regno:16>arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -255,6 +255,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_REG_ARM_TIMER_CTL ARM64_SYS_REG(3, 3, 14, 3, 1)#define KVM_REG_ARM_TIMER_CVAL ARM64_SYS_REG(3, 3, 14, 0, 2)#define KVM_REG_ARM_TIMER_CNT ARM64_SYS_REG(3, 3, 14, 3, 2) +#define KVM_REG_ARM_TIMER_OFFSET ARM64_SYS_REG(3, 4, 14, 0, 3)/* KVM-as-firmware specific pseudo-registers */ #define KVM_REG_ARM_FW (0x0014 << KVM_REG_ARM_COPROC_SHIFT)arch/arm64/kvm/arch_timer.c +11
-0@@ -868,6 +868,10 @@ int kvm_arm_timer_set_reg(struct kvm_vcpu *vcpu, u64 regid, u64 value) timer =vcpu_vtimer (vcpu);kvm_arm_timer_write (vcpu, timer,TIMER_REG_CVAL , value);break ; +case KVM_REG_ARM_TIMER_OFFSET : + timer =vcpu_vtimer (vcpu); +update_vtimer_cntvoff (vcpu, value); +break ;case KVM_REG_ARM_PTIMER_CTL : timer =vcpu_ptimer (vcpu);kvm_arm_timer_write (vcpu, timer,TIMER_REG_CTL , value);@@ -912,6 +916,9 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid)case KVM_REG_ARM_TIMER_CVAL :return kvm_arm_timer_read (vcpu,vcpu_vtimer (vcpu),TIMER_REG_CVAL ); +case KVM_REG_ARM_TIMER_OFFSET : +return kvm_arm_timer_read (vcpu, +vcpu_vtimer (vcpu),TIMER_REG_OFFSET );case KVM_REG_ARM_PTIMER_CTL :return kvm_arm_timer_read (vcpu,vcpu_ptimer (vcpu),TIMER_REG_CTL );@@ -949,6 +956,10 @@ static u64 kvm_arm_timer_read(struct kvm_vcpu *vcpu, val =kvm_phys_timer_read () -timer_get_offset (timer);break ; +case TIMER_REG_OFFSET : + val =timer_get_offset (timer); +break ; +default :BUG (); }arch/arm64/kvm/guest.c +5
-1@@ -591,7 +591,7 @@ static unsigned long num_core_regs(const struct kvm_vcpu *vcpu)* ARM64 versions of the TIMER registers, always available on arm64 */ -#define NUM_TIMER_REGS 3 +#define NUM_TIMER_REGS 4static bool is_timer_reg (u64 index) {@@ -599,6 +599,7 @@ static bool is_timer_reg(u64 index)case KVM_REG_ARM_TIMER_CTL :case KVM_REG_ARM_TIMER_CNT :case KVM_REG_ARM_TIMER_CVAL : +case KVM_REG_ARM_TIMER_OFFSET :return true; }return false;@@ -614,6 +615,9 @@ static int copy_timer_indices(struct kvm_vcpu *vcpu, u64 __user *uindices) uindices++;if (put_user (KVM_REG_ARM_TIMER_CVAL , uindices))return -EFAULT ; + uindices++; +if (put_user (KVM_REG_ARM_TIMER_OFFSET , uindices)) +return -EFAULT ;return 0 ; }include/kvm/arm_arch_timer.h +1
-0@@ -21,6 +21,7 @@ enum kvm_arch_timer_regs {TIMER_REG_CVAL ,TIMER_REG_TVAL ,TIMER_REG_CTL , + TIMER_REG_OFFSET, }; struct arch_timer_context {Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Marc Zyngier
maintainerRe: [PATCH v6 13/21] KVM: arm64: Allow userspace to configure a vCPU's virtual offset
In reply to Oliver UptonQuoted 21 lines · Oliver Upton
Allow userspace to access the guest's virtual counter-timer offset through the
ONE_REGinterface. The value read or written is defined to be an offset from the guest's physical counter-timer. Add some documentation to clarify how a VMM should use this and the existingCNTVCT_EL0.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/api.rst | 10 ++++++++++ arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 11 +++++++++++ arch/arm64/kvm/guest.c | 6 +++++- include/kvm/arm_arch_timer.h | 1 + 5 files changed, 28 insertions(+), 1 deletion(-)Documentation/virt/kvm/api.rst +0
-0
Documentation/virt/kvm/api.rst arm64 system registers have the following id bit patterns:: · line 2496
Quoted 10 lines
Documentation/virt/kvm/api.rst +7
-0@@ -2487,6 +2487,16 @@ arm64 system registers have the following id bit patterns:: derived from the register encoding for CNTV_CVAL_EL0. As this is API, it must remain this way. +.. warning:: + + The value of KVM_REG_ARM_TIMER_OFFSET is defined as an offset from + the guest's view of the physical counter-timer. + + Userspace should use either KVM_REG_ARM_TIMER_OFFSET or + KVM_REG_ARM_TIMER_CVAL to pause and resume a guest's virtual
You probably mean KVM_REG_ARM_TIMER_CNT here, despite the broken encoding.
Quoted 11 lines
Documentation/virt/kvm/api.rst +3
-0+ counter-timer. Mixed use of these registers could result in an + unpredictable guest counter value. + arm64 firmware pseudo-registers have the following bit pattern:: 0x6030 0000 0014 <regno:16>arch/arm64/include/uapi/asm/kvm.h +0
-0
arch/arm64/include/uapi/asm/kvm.h struct kvm_arm_copy_mte_tags { · line 258
Quoted 4 lines
arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -255,6 +255,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_REG_ARM_TIMER_CTL ARM64_SYS_REG(3, 3, 14, 3, 1)#define KVM_REG_ARM_TIMER_CVAL ARM64_SYS_REG(3, 3, 14, 0, 2)#define KVM_REG_ARM_TIMER_CNT ARM64_SYS_REG(3, 3, 14, 3, 2) +#define KVM_REG_ARM_TIMER_OFFSET ARM64_SYS_REG(3, 4, 14, 0, 3)
I don't think we can use the encoding for CNTPOFF_EL2 here, as it will eventually clash with a NV guest using the same feature for its own purpose. We don't want this offset to overlap with any of the existing features.
I actually liked your previous proposal of controlling the physical offset via a device property, as it clearly indicated that you were dealing with non-architectural state.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v6 13/21] KVM: arm64: Allow userspace to configure a vCPU's virtual offset
Quoted 27 lines · Marc Zyngier
Quoted 24 lines · Oliver Upton
Allow userspace to access the guest's virtual counter-timer offset through the
ONE_REGinterface. The value read or written is defined to be an offset from the guest's physical counter-timer. Add some documentation to clarify how a VMM should use this and the existingCNTVCT_EL0.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/api.rst | 10 ++++++++++ arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 11 +++++++++++ arch/arm64/kvm/guest.c | 6 +++++- include/kvm/arm_arch_timer.h | 1 + 5 files changed, 28 insertions(+), 1 deletion(-)Documentation/virt/kvm/api.rst +0
-0@@ -2487,6 +2487,16 @@ arm64 system registers have the following id bit patterns:: derived from the register encoding for CNTV_CVAL_EL0. As this is API, it must remain this way.
Quoted 10 lines
Documentation/virt/kvm/api.rst arm64 system registers have the following id bit patterns:: · line 2496
Quoted 7 lines
Documentation/virt/kvm/api.rst +7
-0+.. warning:: + + The value of KVM_REG_ARM_TIMER_OFFSET is defined as an offset from + the guest's view of the physical counter-timer. + + Userspace should use either KVM_REG_ARM_TIMER_OFFSET or + KVM_REG_ARM_TIMER_CVAL to pause and resume a guest's virtualYou probably mean
KVM_REG_ARM_TIMER_CNThere, despite the broken encoding.
Indeed I do!
Quoted 22 lines
Quoted 12 lines
Documentation/virt/kvm/api.rst +3
-0+ counter-timer. Mixed use of these registers could result in an + unpredictable guest counter value. + arm64 firmware pseudo-registers have the following bit pattern:: 0x6030 0000 0014 <regno:16>arch/arm64/include/uapi/asm/kvm.h +0
-0arch/arm64/include/uapi/asm/kvm.h struct kvm_arm_copy_mte_tags { · line 258
Quoted 4 lines
arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -255,6 +255,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_REG_ARM_TIMER_CTL ARM64_SYS_REG(3, 3, 14, 3, 1)#define KVM_REG_ARM_TIMER_CVAL ARM64_SYS_REG(3, 3, 14, 0, 2)#define KVM_REG_ARM_TIMER_CNT ARM64_SYS_REG(3, 3, 14, 3, 2) +#define KVM_REG_ARM_TIMER_OFFSET ARM64_SYS_REG(3, 4, 14, 0, 3)I don't think we can use the encoding for
CNTPOFF_EL2here, as it will eventually clash with a NV guest using the same feature for its own purpose. We don't want this offset to overlap with any of the existing features.
Quoted 3 lines
I actually liked your previous proposal of controlling the physical offset via a device property, as it clearly indicated that you were dealing with non-architectural state.
That's actually exactly what I did here :) That said, the macro name is horribly obfuscated from CNTVOFF_EL2. I did this for the sake of symmetry with other virtual counter-timer registers above, though this may warrant special casing given the fact that we have a similarly named device attribute to handle the physical offset.
--
Thanks,
Oliver
Marc Zyngier
maintainerRe: [PATCH v6 13/21] KVM: arm64: Allow userspace to configure a vCPU's virtual offset
Quoted 68 lines · Oliver Upton
Quoted 27 lines · Marc Zyngier
Quoted 24 lines · Oliver Upton
Allow userspace to access the guest's virtual counter-timer offset through the
ONE_REGinterface. The value read or written is defined to be an offset from the guest's physical counter-timer. Add some documentation to clarify how a VMM should use this and the existingCNTVCT_EL0.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/api.rst | 10 ++++++++++ arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 11 +++++++++++ arch/arm64/kvm/guest.c | 6 +++++- include/kvm/arm_arch_timer.h | 1 + 5 files changed, 28 insertions(+), 1 deletion(-)Documentation/virt/kvm/api.rst +0
-0@@ -2487,6 +2487,16 @@ arm64 system registers have the following id bit patterns:: derived from the register encoding for CNTV_CVAL_EL0. As this is API, it must remain this way.Quoted 10 lines
Documentation/virt/kvm/api.rst arm64 system registers have the following id bit patterns:: · line 2496
Quoted 7 lines
Documentation/virt/kvm/api.rst +7
-0+.. warning:: + + The value of KVM_REG_ARM_TIMER_OFFSET is defined as an offset from + the guest's view of the physical counter-timer. + + Userspace should use either KVM_REG_ARM_TIMER_OFFSET or + KVM_REG_ARM_TIMER_CVAL to pause and resume a guest's virtualYou probably mean
KVM_REG_ARM_TIMER_CNThere, despite the broken encoding.Indeed I do!
Quoted 22 lines
Quoted 12 lines
Documentation/virt/kvm/api.rst +3
-0+ counter-timer. Mixed use of these registers could result in an + unpredictable guest counter value. + arm64 firmware pseudo-registers have the following bit pattern:: 0x6030 0000 0014 <regno:16>arch/arm64/include/uapi/asm/kvm.h +0
-0arch/arm64/include/uapi/asm/kvm.h struct kvm_arm_copy_mte_tags { · line 258
Quoted 4 lines
arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -255,6 +255,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_REG_ARM_TIMER_CTL ARM64_SYS_REG(3, 3, 14, 3, 1)#define KVM_REG_ARM_TIMER_CVAL ARM64_SYS_REG(3, 3, 14, 0, 2)#define KVM_REG_ARM_TIMER_CNT ARM64_SYS_REG(3, 3, 14, 3, 2) +#define KVM_REG_ARM_TIMER_OFFSET ARM64_SYS_REG(3, 4, 14, 0, 3)I don't think we can use the encoding for
CNTPOFF_EL2here, as it will eventually clash with a NV guest using the same feature for its own purpose. We don't want this offset to overlap with any of the existing features.Quoted 3 lines
I actually liked your previous proposal of controlling the physical offset via a device property, as it clearly indicated that you were dealing with non-architectural state.
Quoted 5 lines
That's actually exactly what I did here :) That said, the macro name is horribly obfuscated from
CNTVOFF_EL2. I did this for the sake of symmetry with other virtual counter-timer registers above, though this may warrant special casing given the fact that we have a similarly named device attribute to handle the physical offset.
Gah, you are of course right. Ignore my rambling. The name is fine (or at least in keeping with existing quality level of the making).
For the physical offset, something along the lines of KVM_ARM_VCPU_TIMER_PHYS_OFFSET is probably right (but feel free to be creative, I'm terrible at this stuff [1]).
Thanks,
M.
[1] https://twitter.com/codinghorror/status/506010907021828096?lang=en
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v6 14/21 selftests: KVM: Add helper to check for register presence
In reply to Oliver UptonThe KVM_GET_REG_LIST vCPU ioctl returns a list of supported registers for a given vCPU. Add a helper to check if a register exists in the list of supported registers.
--- .../testing/selftests/kvm/include/kvm_util.h | 2 ++ tools/testing/selftests/kvm/lib/kvm_util.c | 19 +++++++++++++++++++ 2 files changed, 21 insertions(+)
tools/testing/selftests/kvm/include/kvm_util.h +2 -0
@@ -215,6 +215,8 @@ void vcpu_fpu_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu);void vcpu_fpu_set (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_fpu *fpu); + +bool vcpu_has_reg (struct kvm_vm *vm,uint32_t vcpuid,uint64_t reg_id);void vcpu_get_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg);void vcpu_set_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg);#ifdef __KVM_HAVE_VCPU_EVENTS
tools/testing/selftests/kvm/lib/kvm_util.c +19 -0
@@ -1823,6 +1823,25 @@ void vcpu_fpu_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu) ret, errno, strerror(errno)); } +bool vcpu_has_reg (struct kvm_vm *vm,uint32_t vcpuid,uint64_t reg_id) +{ +struct kvm_reg_list *list; +bool ret = false; +uint64_t i; + + list =vcpu_get_reg_list (vm, vcpuid); + +for (i =0 ; i < list->n; i++) { +if (list->reg[i] == reg_id) { + ret = true; +break ; + } + } + +free (list); +return ret; +} +void vcpu_get_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg) {int ret;
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 14/21] selftests: KVM: Add helper to check for register presence
Quoted 56 lines · Oliver Upton
The
KVM_GET_REG_LISTvCPU ioctl returns a list of supported registers for a given vCPU. Add a helper to check if a register exists in the list of supported registers.Signed-off-by Oliver Upton <[email protected]>--- .../testing/selftests/kvm/include/kvm_util.h | 2 ++ tools/testing/selftests/kvm/lib/kvm_util.c | 19 +++++++++++++++++++ 2 files changed, 21 insertions(+)tools/testing/selftests/kvm/include/kvm_util.h +2
-0@@ -215,6 +215,8 @@ void vcpu_fpu_get(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu);void vcpu_fpu_set (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_fpu *fpu); + +bool vcpu_has_reg (struct kvm_vm *vm,uint32_t vcpuid,uint64_t reg_id);void vcpu_get_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg);void vcpu_set_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg);#ifdef __KVM_HAVE_VCPU_EVENTStools/testing/selftests/kvm/lib/kvm_util.c +19
-0@@ -1823,6 +1823,25 @@ void vcpu_fpu_set(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_fpu *fpu) ret, errno, strerror(errno)); } +bool vcpu_has_reg (struct kvm_vm *vm,uint32_t vcpuid,uint64_t reg_id) +{ +struct kvm_reg_list *list; +bool ret = false; +uint64_t i; + + list =vcpu_get_reg_list (vm, vcpuid); + +for (i =0 ; i < list->n; i++) { +if (list->reg[i] == reg_id) { + ret = true; +break ; + } + } + +free (list); +return ret; +} +void vcpu_get_reg (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_one_reg *reg) {int ret;Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 15/21 selftests: KVM: Add support for aarch64 to system_counter_offset_test
In reply to Oliver UptonKVM/arm64 now allows userspace to adjust the guest virtual counter-timer via a vCPU register. Test that changes to the virtual counter-timer offset result in the correct view being presented to the guest.
--- tools/testing/selftests/kvm/Makefile | 1 + .../selftests/kvm/include/aarch64/processor.h | 12 ++++ .../kvm/system_counter_offset_test.c | 56 ++++++++++++++++++- 3 files changed, 68 insertions(+), 1 deletion(-)
tools/testing/selftests/kvm/Makefile +1 -0
@@ -98,6 +98,7 @@ TEST_GEN_PROGS_aarch64 += kvm_page_table_test TEST_GEN_PROGS_aarch64 += set_memory_region_test TEST_GEN_PROGS_aarch64 += steal_time TEST_GEN_PROGS_aarch64 += kvm_binary_stats_test +TEST_GEN_PROGS_aarch64 += system_counter_offset_test TEST_GEN_PROGS_s390x = s390x/memop TEST_GEN_PROGS_s390x += s390x/resets
tools/testing/selftests/kvm/include/aarch64/processor.h +12 -0
@@ -129,4 +129,16 @@ void vm_install_sync_handler(struct kvm_vm *vm,#define isb () asm volatile("isb" : : : "memory") +static inline uint64_t read_cntvct_ordered (void ) +{ +uint64_t r; + + __asm__ __volatile__("isb\n\t" +"mrs %0, cntvct_el0\n\t" +"isb\n\t" + :"=r" (r)); + +return r; +} +#endif /* SELFTEST_KVM_PROCESSOR_H */
tools/testing/selftests/kvm/system_counter_offset_test.c +55 -1
62 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 16/21 arm64: cpufeature: Enumerate support for Enhanced Counter Virtualization
In reply to Oliver UptonIntroduce a new cpucap to indicate if the system supports full enhanced counter virtualization (i.e. ID_AA64MMFR0_EL1.ECV==0x2).
--- arch/arm64/include/asm/sysreg.h | 2 ++ arch/arm64/kernel/cpufeature.c | 10 ++++++++++ arch/arm64/tools/cpucaps | 1 + 3 files changed, 13 insertions(+)
arch/arm64/include/asm/sysreg.h +2 -0
@@ -847,6 +847,8 @@#define ID_AA64MMFR0_ASID_SHIFT 4#define ID_AA64MMFR0_PARANGE_SHIFT 0 +#define ID_AA64MMFR0_ECV_VIRT 0x1 +#define ID_AA64MMFR0_ECV_PHYS 0x2#define ID_AA64MMFR0_TGRAN4_NI 0xf#define ID_AA64MMFR0_TGRAN4_SUPPORTED 0x0#define ID_AA64MMFR0_TGRAN64_NI 0xf
arch/arm64/kernel/cpufeature.c +10 -0
@@ -2301,6 +2301,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .matches = has_cpuid_feature, .min_field_value = 1, }, + { + .desc = "Enhanced Counter Virtualization (Physical)", + .capability = ARM64_ECV, + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .sys_reg = SYS_ID_AA64MMFR0_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64MMFR0_ECV_SHIFT, + .matches = has_cpuid_feature, + .min_field_value = ID_AA64MMFR0_ECV_PHYS, + }, {}, };
arch/arm64/tools/cpucaps +1 -0
@@ -3,6 +3,7 @@ # Internal CPU capabilities constants, keep this list sorted BTI +ECV # Unreliable: use system_supports_32bit_el0() instead. HAS_32BIT_EL0_DO_NOT_USE HAS_32BIT_EL1
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Marc Zyngier
maintainerRe: [PATCH v6 16/21] arm64: cpufeature: Enumerate support for Enhanced Counter Virtualization
Quoted 50 lines · Oliver Upton
Introduce a new cpucap to indicate if the system supports full enhanced counter virtualization (i.e. ID_AA64MMFR0_EL1.ECV==0x2).
Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/include/asm/sysreg.h | 2 ++ arch/arm64/kernel/cpufeature.c | 10 ++++++++++ arch/arm64/tools/cpucaps | 1 + 3 files changed, 13 insertions(+)arch/arm64/include/asm/sysreg.h +2
-0@@ -847,6 +847,8 @@#define ID_AA64MMFR0_ASID_SHIFT 4#define ID_AA64MMFR0_PARANGE_SHIFT 0 +#define ID_AA64MMFR0_ECV_VIRT 0x1 +#define ID_AA64MMFR0_ECV_PHYS 0x2#define ID_AA64MMFR0_TGRAN4_NI 0xf#define ID_AA64MMFR0_TGRAN4_SUPPORTED 0x0#define ID_AA64MMFR0_TGRAN64_NI 0xfarch/arm64/kernel/cpufeature.c +10
-0@@ -2301,6 +2301,16 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .matches = has_cpuid_feature, .min_field_value = 1, }, + { + .desc = "Enhanced Counter Virtualization (Physical)", + .capability = ARM64_ECV, + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .sys_reg = SYS_ID_AA64MMFR0_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64MMFR0_ECV_SHIFT, + .matches = has_cpuid_feature, + .min_field_value = ID_AA64MMFR0_ECV_PHYS, + }, {}, };arch/arm64/tools/cpucaps +0
-0
arch/arm64/tools/cpucaps line 9
Quoted 7 lines
arch/arm64/tools/cpucaps +1
-0@@ -3,6 +3,7 @@ # Internal CPU capabilities constants, keep this list sorted BTI +ECV # Unreliable: use system_supports_32bit_el0() instead. HAS_32BIT_EL0_DO_NOT_USE HAS_32BIT_EL1
As discussed in another context, we probably want both ECV and ECV2 to distinguish the two feature sets that ECV has so far.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v6 17/21 KVM: arm64: Allow userspace to configure a guest's counter-timer offset
In reply to Oliver UptonPresently, KVM provides no facilities for correctly migrating a guest that depends on the physical counter-timer. Whie most guests (barring NV, of course) should not depend on the physical counter-timer, an operator may wish to provide a consistent view of the physical counter-timer across migrations.
Provide userspace with a new vCPU attribute to modify the guest counter-timer offset. Unlike KVM_REG_ARM_TIMER_OFFSET, this attribute is hidden from the guest's architectural state. The value offsets *both* the virtual and physical counter-timer views for the guest. Only support this attribute on ECV systems as ECV is required for hardware offsetting of the physical counter-timer.
--- Documentation/virt/kvm/devices/vcpu.rst | 28 ++++++ arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 2 + arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 122 +++++++++++++++++++++++- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 ++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 5 + arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + include/clocksource/arm_arch_timer.h | 1 + 9 files changed, 169 insertions(+), 3 deletions(-)
Documentation/virt/kvm/devices/vcpu.rst +28 -0
@@ -139,6 +139,34 @@ configured values on other VCPUs. Userspace should configure the interrupt numbers on at least one VCPU after creating all VCPUs and before running any VCPUs. +2.2. ATTRIBUTE: KVM_ARM_VCPU_TIMER_OFFSET +----------------------------------------- + +:Parameters: in kvm_device_attr.addr the address for the timer offset is a + pointer to a __u64 + +Returns: + + ======= ================================== + -EFAULT Error reading/writing the provided + parameter address + -ENXIO Timer offsetting not implemented + ======= ================================== + +Specifies the guest's counter-timer offset from the host's virtual counter. +The guest's physical counter value is then derived by the following +equation: + + guest_cntpct = host_cntvct - KVM_ARM_VCPU_TIMER_OFFSET + +The guest's virtual counter value is derived by the following equation: + + guest_cntvct = host_cntvct - KVM_REG_ARM_TIMER_OFFSET + - KVM_ARM_VCPU_TIMER_OFFSET + +KVM does not allow the use of varying offset values for different vCPUs; +the last written offset value will be broadcasted to all vCPUs in a VM. + 3. GROUP: KVM_ARM_VCPU_PVTIME_CTRL ==================================
arch/arm64/include/asm/kvm_asm.h +2 -0
@@ -65,6 +65,7 @@#define __KVM_HOST_SMCCC_FUNC___pkvm_prot_finalize 19#define __KVM_HOST_SMCCC_FUNC___pkvm_mark_hyp 20#define __KVM_HOST_SMCCC_FUNC___kvm_adjust_pc 21 +#define __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntpoff 22#ifndef __ASSEMBLY__
@@ -200,6 +201,7 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff); +extern void __kvm_timer_set_cntpoff (u64 cntpoff);extern int __kvm_vcpu_run (struct kvm_vcpu *vcpu);
arch/arm64/include/asm/sysreg.h +2 -0
@@ -586,6 +586,8 @@#define SYS_ICH_LR14_EL2 __SYS__LR8_EL2(6)#define SYS_ICH_LR15_EL2 __SYS__LR8_EL2(7) +#define SYS_CNTPOFF_EL2 sys_reg(3, 4, 14, 0, 6) +/* VHE encodings for architectural EL0/1 system registers */ #define SYS_SCTLR_EL12 sys_reg(3, 5, 1, 0, 0)#define SYS_CPACR_EL12 sys_reg(3, 5, 1, 0, 2)
arch/arm64/include/uapi/asm/kvm.h +1 -0
@@ -366,6 +366,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_ARM_VCPU_TIMER_CTRL 1#define KVM_ARM_VCPU_TIMER_IRQ_VTIMER 0#define KVM_ARM_VCPU_TIMER_IRQ_PTIMER 1 +#define KVM_ARM_VCPU_TIMER_OFFSET 2#define KVM_ARM_VCPU_PVTIME_CTRL 2#define KVM_ARM_VCPU_PVTIME_IPA 0
arch/arm64/kvm/arch_timer.c +119 -3
199 lines not on this page · raw message
arch/arm64/kvm/hyp/nvhe/hyp-main.c +6 -0
@@ -68,6 +68,11 @@ static void handle___kvm_timer_set_cntvoff(struct kvm_cpu_context *host_ctxt)__kvm_timer_set_cntvoff (cpu_reg (host_ctxt,1 )); } +static void handle___kvm_timer_set_cntpoff (struct kvm_cpu_context *host_ctxt) +{ +__kvm_timer_set_cntpoff (cpu_reg (host_ctxt,1 )); +} +static void handle___kvm_enable_ssbs (struct kvm_cpu_context *host_ctxt) {u64 tmp;
@@ -197,6 +202,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__pkvm_create_private_mapping),HANDLE_FUNC (__pkvm_prot_finalize),HANDLE_FUNC (__pkvm_mark_hyp), + HANDLE_FUNC(__kvm_timer_set_cntpoff), };static void handle_host_hcall (struct kvm_cpu_context *host_ctxt)
arch/arm64/kvm/hyp/nvhe/timer-sr.c +5 -0
@@ -15,6 +15,11 @@ void __kvm_timer_set_cntvoff(u64 cntvoff)write_sysreg (cntvoff, cntvoff_el2); } +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +} +/* * Should only be called on non-VHE systems. * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe().
arch/arm64/kvm/hyp/vhe/timer-sr.c +5 -0
@@ -10,3 +10,8 @@ void __kvm_timer_set_cntvoff(u64 cntvoff) {write_sysreg (cntvoff, cntvoff_el2); } + +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +}
include/clocksource/arm_arch_timer.h +1 -0
@@ -21,6 +21,7 @@#define CNTHCTL_EVNTEN (1 << 2)#define CNTHCTL_EVNTDIR (1 << 3)#define CNTHCTL_EVNTI (0xF << 4) +#define CNTHCTL_ECV (1 << 12)enum arch_timer_reg {ARCH_TIMER_REG_CTRL ,
Signature · 6 lines
2.32.0.605.g8dce9f2422-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 17/21] KVM: arm64: Allow userspace to configure a guest's counter-timer offset
Quoted 115 lines · Oliver Upton
Presently, KVM provides no facilities for correctly migrating a guest that depends on the physical counter-timer. Whie most guests (barring NV, of course) should not depend on the physical counter-timer, an operator may wish to provide a consistent view of the physical counter-timer across migrations.
Provide userspace with a new vCPU attribute to modify the guest counter-timer offset. Unlike
KVM_REG_ARM_TIMER_OFFSET, this attribute is hidden from the guest's architectural state. The value offsets *both* the virtual and physical counter-timer views for the guest. Only support this attribute on ECV systems as ECV is required for hardware offsetting of the physical counter-timer.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/devices/vcpu.rst | 28 ++++++ arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 2 + arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 122 +++++++++++++++++++++++- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 ++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 5 + arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + include/clocksource/arm_arch_timer.h | 1 + 9 files changed, 169 insertions(+), 3 deletions(-)Documentation/virt/kvm/devices/vcpu.rst +28
-0@@ -139,6 +139,34 @@ configured values on other VCPUs. Userspace should configure the interrupt numbers on at least one VCPU after creating all VCPUs and before running any VCPUs. +2.2. ATTRIBUTE: KVM_ARM_VCPU_TIMER_OFFSET +----------------------------------------- + +:Parameters: in kvm_device_attr.addr the address for the timer offset is a + pointer to a __u64 + +Returns: + + ======= ================================== + -EFAULT Error reading/writing the provided + parameter address + -ENXIO Timer offsetting not implemented + ======= ================================== + +Specifies the guest's counter-timer offset from the host's virtual counter. +The guest's physical counter value is then derived by the following +equation: + + guest_cntpct = host_cntvct - KVM_ARM_VCPU_TIMER_OFFSET + +The guest's virtual counter value is derived by the following equation: + + guest_cntvct = host_cntvct - KVM_REG_ARM_TIMER_OFFSET + - KVM_ARM_VCPU_TIMER_OFFSET + +KVM does not allow the use of varying offset values for different vCPUs; +the last written offset value will be broadcasted to all vCPUs in a VM. + 3. GROUP: KVM_ARM_VCPU_PVTIME_CTRL ==================================arch/arm64/include/asm/kvm_asm.h +2
-0@@ -65,6 +65,7 @@#define __KVM_HOST_SMCCC_FUNC___pkvm_prot_finalize 19#define __KVM_HOST_SMCCC_FUNC___pkvm_mark_hyp 20#define __KVM_HOST_SMCCC_FUNC___kvm_adjust_pc 21 +#define __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntpoff 22#ifndef __ASSEMBLY__@@ -200,6 +201,7 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff); +extern void __kvm_timer_set_cntpoff (u64 cntpoff);extern int __kvm_vcpu_run (struct kvm_vcpu *vcpu);arch/arm64/include/asm/sysreg.h +2
-0@@ -586,6 +586,8 @@#define SYS_ICH_LR14_EL2 __SYS__LR8_EL2(6)#define SYS_ICH_LR15_EL2 __SYS__LR8_EL2(7) +#define SYS_CNTPOFF_EL2 sys_reg(3, 4, 14, 0, 6) +/* VHE encodings for architectural EL0/1 system registers */ #define SYS_SCTLR_EL12 sys_reg(3, 5, 1, 0, 0)#define SYS_CPACR_EL12 sys_reg(3, 5, 1, 0, 2)arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -366,6 +366,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_ARM_VCPU_TIMER_CTRL 1#define KVM_ARM_VCPU_TIMER_IRQ_VTIMER 0#define KVM_ARM_VCPU_TIMER_IRQ_PTIMER 1 +#define KVM_ARM_VCPU_TIMER_OFFSET 2#define KVM_ARM_VCPU_PVTIME_CTRL 2#define KVM_ARM_VCPU_PVTIME_IPA 0arch/arm64/kvm/arch_timer.c +0
-0
arch/arm64/kvm/arch_timer.c u64 timer_get_cval(struct arch_timer_context *ctxt) · line 88
Quoted 4 lines
arch/arm64/kvm/arch_timer.c +1
-0@@ -85,11 +85,15 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)static u64 timer_get_offset (struct arch_timer_context *ctxt) {struct kvm_vcpu *vcpu = ctxt->vcpu; +struct arch_timer_cpu *timer =vcpu_timer (vcpu);
This new timer variable doesn't appear to get used.
Quoted 273 lines
arch/arm64/kvm/arch_timer.c +118
-3switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER :return ctxt->host_offset;default : +WARN_ONCE (1 ,"unrecognized timer %ld\n" , +arch_timer_ctx_index (ctxt));return 0 ; } }@@ -144,6 +148,7 @@ static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER : ctxt->host_offset = offset;break ;default :@@ -572,6 +577,11 @@ static void set_cntvoff(u64 cntvoff)kvm_call_hyp (__kvm_timer_set_cntvoff, cntvoff); } +static void set_cntpoff (u64 cntpoff) +{ +kvm_call_hyp (__kvm_timer_set_cntpoff, cntpoff); +} +static inline void set_timer_irq_phys_active (struct arch_timer_context *ctx,bool active) {int r;@@ -647,6 +657,8 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) }set_cntvoff (timer_get_offset (map.direct_vtimer)); +if (cpus_have_const_cap (ARM64_ECV )) +set_cntpoff (timer_get_offset (vcpu_ptimer (vcpu)));kvm_timer_unblocking (vcpu);@@ -814,6 +826,22 @@ static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)mutex_unlock (&kvm->lock); } +static void update_ptimer_cntpoff (struct kvm_vcpu *vcpu,u64 cntpoff) +{ +struct kvm *kvm = vcpu->kvm; +u64 cntvoff; + +mutex_lock (&kvm->lock); + +/* adjustments to the physical offset also affect vtimer */ + cntvoff =timer_get_offset (vcpu_vtimer (vcpu)); + cntvoff += cntpoff -timer_get_offset (vcpu_ptimer (vcpu)); + +update_timer_offset (vcpu,TIMER_PTIMER , cntpoff, false); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff, false); +mutex_unlock (&kvm->lock); +} +void kvm_timer_vcpu_init (struct kvm_vcpu *vcpu) {struct arch_timer_cpu *timer =vcpu_timer (vcpu);@@ -932,6 +960,29 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid)return (u64)-1 ; } +/** +* kvm_arm_timer_read_offset - returns the guest value of CNTVOFF_EL2. +* @vcpu: the vcpu pointer +* +* Computes the guest value of CNTVOFF_EL2 by subtracting the physical +* counter offset. Note that KVM defines CNTVOFF_EL2 as the offset from the +* guest's physical counter-timer, not the host's. +* +* Returns: the guest value for CNTVOFF_EL2 +*/ +static u64 kvm_arm_timer_read_offset (struct kvm_vcpu *vcpu) +{ +struct kvm *kvm = vcpu->kvm; +u64 offset; + +mutex_lock (&kvm->lock); + offset =timer_get_offset (vcpu_vtimer (vcpu)) - +timer_get_offset (vcpu_ptimer (vcpu)); +mutex_unlock (&kvm->lock); + +return offset; +} +static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg)@@ -957,7 +1008,7 @@ static u64 kvm_arm_timer_read(struct kvm_vcpu *vcpu,break ;case TIMER_REG_OFFSET : - val =timer_get_offset (timer); + val =kvm_arm_timer_read_offset (vcpu);break ;default :@@ -1350,6 +1401,9 @@ void kvm_timer_init_vhe(void) val =read_sysreg (cnthctl_el2); val |= (CNTHCTL_EL1PCEN << cnthctl_shift); val |= (CNTHCTL_EL1PCTEN << cnthctl_shift); + +if (cpus_have_const_cap (ARM64_ECV )) + val |=CNTHCTL_ECV ;write_sysreg (val, cnthctl_el2); }@@ -1364,7 +1418,8 @@ static void set_timer_irqs(struct kvm *kvm, int vtimer_irq, int ptimer_irq) } } -int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *vtimer =vcpu_vtimer (vcpu);@@ -1397,7 +1452,37 @@ int kvm_arm_timer_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return 0 ; } -int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + +if (get_user (offset, uaddr)) +return -EFAULT ; + +update_ptimer_cntpoff (vcpu, offset); +return 0 ; +} + +int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_set_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_set_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} + +static int kvm_arm_timer_get_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *timer;@@ -1418,12 +1503,43 @@ int kvm_arm_timer_get_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return put_user (irq, uaddr); } +static int kvm_arm_timer_get_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + + offset =timer_get_offset (vcpu_ptimer (vcpu)); +return put_user (offset, uaddr); +} + +int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_get_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_get_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} +int kvm_arm_timer_has_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) {switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER :return 0 ; +case KVM_ARM_VCPU_TIMER_OFFSET : +if (cpus_have_const_cap (ARM64_ECV )) +return 0 ; +break ; }return -ENXIO ;arch/arm64/kvm/hyp/nvhe/hyp-main.c +6
-0@@ -68,6 +68,11 @@ static void handle___kvm_timer_set_cntvoff(struct kvm_cpu_context *host_ctxt)__kvm_timer_set_cntvoff (cpu_reg (host_ctxt,1 )); } +static void handle___kvm_timer_set_cntpoff (struct kvm_cpu_context *host_ctxt) +{ +__kvm_timer_set_cntpoff (cpu_reg (host_ctxt,1 )); +} +static void handle___kvm_enable_ssbs (struct kvm_cpu_context *host_ctxt) {u64 tmp;@@ -197,6 +202,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__pkvm_create_private_mapping),HANDLE_FUNC (__pkvm_prot_finalize),HANDLE_FUNC (__pkvm_mark_hyp), + HANDLE_FUNC(__kvm_timer_set_cntpoff), };static void handle_host_hcall (struct kvm_cpu_context *host_ctxt)arch/arm64/kvm/hyp/nvhe/timer-sr.c +5
-0@@ -15,6 +15,11 @@ void __kvm_timer_set_cntvoff(u64 cntvoff)write_sysreg (cntvoff, cntvoff_el2); } +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +} +/* * Should only be called on non-VHE systems. * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe(). arch/arm64/kvm/hyp/vhe/timer-sr.c +5
-0@@ -10,3 +10,8 @@ void __kvm_timer_set_cntvoff(u64 cntvoff) {write_sysreg (cntvoff, cntvoff_el2); } + +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +}include/clocksource/arm_arch_timer.h +1
-0@@ -21,6 +21,7 @@#define CNTHCTL_EVNTEN (1 << 2)#define CNTHCTL_EVNTDIR (1 << 3)#define CNTHCTL_EVNTI (0xF << 4) +#define CNTHCTL_ECV (1 << 12)enum arch_timer_reg {ARCH_TIMER_REG_CTRL ,Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 17/21] KVM: arm64: Allow userspace to configure a guest's counter-timer offset
Quoted 11 lines · Andrew Jones
arch/arm64/kvm/arch_timer.c u64 timer_get_cval(struct arch_timer_context *ctxt) · line 88
Quoted 9 lines
arch/arm64/kvm/arch_timer.c +1
-0@@ -85,11 +85,15 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)static u64 timer_get_offset (struct arch_timer_context *ctxt) {struct kvm_vcpu *vcpu = ctxt->vcpu; +struct arch_timer_cpu *timer =vcpu_timer (vcpu);This new timer variable doesn't appear to get used.
Ooops, this is stale. Thanks for catching that.
Quoted 3 lines
Reviewed-by Andrew Jones <[email protected]>
Thanks for the quick review!
--
Oliver
Marc Zyngier
maintainerRe: [PATCH v6 17/21] KVM: arm64: Allow userspace to configure a guest's counter-timer offset
In reply to Oliver UptonQuoted 3 lines · Oliver Upton
Presently, KVM provides no facilities for correctly migrating a guest that depends on the physical counter-timer. Whie most guests (barring
nit: While
Quoted 113 lines
NV, of course) should not depend on the physical counter-timer, an operator may wish to provide a consistent view of the physical counter-timer across migrations.
Provide userspace with a new vCPU attribute to modify the guest counter-timer offset. Unlike
KVM_REG_ARM_TIMER_OFFSET, this attribute is hidden from the guest's architectural state. The value offsets *both* the virtual and physical counter-timer views for the guest. Only support this attribute on ECV systems as ECV is required for hardware offsetting of the physical counter-timer.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/devices/vcpu.rst | 28 ++++++ arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 2 + arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 122 +++++++++++++++++++++++- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 ++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 5 + arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + include/clocksource/arm_arch_timer.h | 1 + 9 files changed, 169 insertions(+), 3 deletions(-)Documentation/virt/kvm/devices/vcpu.rst +28
-0@@ -139,6 +139,34 @@ configured values on other VCPUs. Userspace should configure the interrupt numbers on at least one VCPU after creating all VCPUs and before running any VCPUs. +2.2. ATTRIBUTE: KVM_ARM_VCPU_TIMER_OFFSET +----------------------------------------- + +:Parameters: in kvm_device_attr.addr the address for the timer offset is a + pointer to a __u64 + +Returns: + + ======= ================================== + -EFAULT Error reading/writing the provided + parameter address + -ENXIO Timer offsetting not implemented + ======= ================================== + +Specifies the guest's counter-timer offset from the host's virtual counter. +The guest's physical counter value is then derived by the following +equation: + + guest_cntpct = host_cntvct - KVM_ARM_VCPU_TIMER_OFFSET + +The guest's virtual counter value is derived by the following equation: + + guest_cntvct = host_cntvct - KVM_REG_ARM_TIMER_OFFSET + - KVM_ARM_VCPU_TIMER_OFFSET + +KVM does not allow the use of varying offset values for different vCPUs; +the last written offset value will be broadcasted to all vCPUs in a VM. + 3. GROUP: KVM_ARM_VCPU_PVTIME_CTRL ==================================arch/arm64/include/asm/kvm_asm.h +2
-0@@ -65,6 +65,7 @@#define __KVM_HOST_SMCCC_FUNC___pkvm_prot_finalize 19#define __KVM_HOST_SMCCC_FUNC___pkvm_mark_hyp 20#define __KVM_HOST_SMCCC_FUNC___kvm_adjust_pc 21 +#define __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntpoff 22#ifndef __ASSEMBLY__@@ -200,6 +201,7 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff); +extern void __kvm_timer_set_cntpoff (u64 cntpoff);extern int __kvm_vcpu_run (struct kvm_vcpu *vcpu);arch/arm64/include/asm/sysreg.h +2
-0@@ -586,6 +586,8 @@#define SYS_ICH_LR14_EL2 __SYS__LR8_EL2(6)#define SYS_ICH_LR15_EL2 __SYS__LR8_EL2(7) +#define SYS_CNTPOFF_EL2 sys_reg(3, 4, 14, 0, 6) +/* VHE encodings for architectural EL0/1 system registers */ #define SYS_SCTLR_EL12 sys_reg(3, 5, 1, 0, 0)#define SYS_CPACR_EL12 sys_reg(3, 5, 1, 0, 2)arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -366,6 +366,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_ARM_VCPU_TIMER_CTRL 1#define KVM_ARM_VCPU_TIMER_IRQ_VTIMER 0#define KVM_ARM_VCPU_TIMER_IRQ_PTIMER 1 +#define KVM_ARM_VCPU_TIMER_OFFSET 2#define KVM_ARM_VCPU_PVTIME_CTRL 2#define KVM_ARM_VCPU_PVTIME_IPA 0arch/arm64/kvm/arch_timer.c +0
-0
arch/arm64/kvm/arch_timer.c u64 timer_get_cval(struct arch_timer_context *ctxt) · line 88
Quoted 4 lines
arch/arm64/kvm/arch_timer.c +1
-0@@ -85,11 +85,15 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)static u64 timer_get_offset (struct arch_timer_context *ctxt) {struct kvm_vcpu *vcpu = ctxt->vcpu; +struct arch_timer_cpu *timer =vcpu_timer (vcpu);
Unused variable?
Quoted 32 lines
arch/arm64/kvm/arch_timer.c +9
-0switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER :return ctxt->host_offset;default : +WARN_ONCE (1 ,"unrecognized timer %ld\n" , +arch_timer_ctx_index (ctxt));return 0 ; } }@@ -144,6 +148,7 @@ static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER : ctxt->host_offset = offset;break ;default :@@ -572,6 +577,11 @@ static void set_cntvoff(u64 cntvoff)kvm_call_hyp (__kvm_timer_set_cntvoff, cntvoff); } +static void set_cntpoff (u64 cntpoff) +{ +kvm_call_hyp (__kvm_timer_set_cntpoff, cntpoff); +} +static inline void set_timer_irq_phys_active (struct arch_timer_context *ctx,bool active) {int r;
arch/arm64/kvm/arch_timer.c void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) · line 660
Quoted 4 lines
arch/arm64/kvm/arch_timer.c +1
-0@@ -647,6 +657,8 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) }set_cntvoff (timer_get_offset (map.direct_vtimer)); +if (cpus_have_const_cap (ARM64_ECV ))
This really should be a final cap instead (same for all the other use cases).
arch/arm64/kvm/arch_timer.c void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) · line 661
Quoted 1 line
arch/arm64/kvm/arch_timer.c +1
-0+set_cntpoff (timer_get_offset (vcpu_ptimer (vcpu)));
However, tripping to EL2 for each offset on nVHE may prove to be an unnecessary overhead. Not a problem for now anyway.
Quoted 38 lines
arch/arm64/kvm/arch_timer.c +24
-0kvm_timer_unblocking (vcpu);@@ -814,6 +826,22 @@ static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)mutex_unlock (&kvm->lock); } +static void update_ptimer_cntpoff (struct kvm_vcpu *vcpu,u64 cntpoff) +{ +struct kvm *kvm = vcpu->kvm; +u64 cntvoff; + +mutex_lock (&kvm->lock); + +/* adjustments to the physical offset also affect vtimer */ + cntvoff =timer_get_offset (vcpu_vtimer (vcpu)); + cntvoff += cntpoff -timer_get_offset (vcpu_ptimer (vcpu)); + +update_timer_offset (vcpu,TIMER_PTIMER , cntpoff, false); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff, false); +mutex_unlock (&kvm->lock); +} +void kvm_timer_vcpu_init (struct kvm_vcpu *vcpu) {struct arch_timer_cpu *timer =vcpu_timer (vcpu);@@ -932,6 +960,29 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid)return (u64)-1 ; } +/** +* kvm_arm_timer_read_offset - returns the guest value of CNTVOFF_EL2. +* @vcpu: the vcpu pointer +* +* Computes the guest value of CNTVOFF_EL2 by subtracting the physical +* counter offset. Note that KVM defines CNTVOFF_EL2 as the offset from the +* guest's physical counter-timer, not the host's. +*
arch/arm64/kvm/arch_timer.c u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid) · line 980
Quoted 10 lines
arch/arm64/kvm/arch_timer.c +10
-0+* Returns: the guest value for CNTVOFF_EL2 +*/ +static u64 kvm_arm_timer_read_offset (struct kvm_vcpu *vcpu) +{ +struct kvm *kvm = vcpu->kvm; +u64 offset; + +mutex_lock (&kvm->lock); + offset =timer_get_offset (vcpu_vtimer (vcpu)) - +timer_get_offset (vcpu_ptimer (vcpu));
nit: please keep this on a single line.
Quoted 18 lines
arch/arm64/kvm/arch_timer.c +6
-1+mutex_unlock (&kvm->lock); + +return offset; +} +static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg)@@ -957,7 +1008,7 @@ static u64 kvm_arm_timer_read(struct kvm_vcpu *vcpu,break ;case TIMER_REG_OFFSET : - val =timer_get_offset (timer); + val =kvm_arm_timer_read_offset (vcpu);break ;default :
arch/arm64/kvm/arch_timer.c void kvm_timer_init_vhe(void) · line 1406
Quoted 6 lines
arch/arm64/kvm/arch_timer.c +3
-0@@ -1350,6 +1401,9 @@ void kvm_timer_init_vhe(void) val =read_sysreg (cnthctl_el2); val |= (CNTHCTL_EL1PCEN << cnthctl_shift); val |= (CNTHCTL_EL1PCTEN << cnthctl_shift); + +if (cpus_have_const_cap (ARM64_ECV )) + val |=CNTHCTL_ECV ;
I cannot immediately see where you are doing the equivalent enablement of ECV on the nVHE path. Obviously, it has to be done eagerly from EL2, together with the rest of the EL1 timer setup. Something like:
arch/arm64/kvm/hyp/nvhe/timer-sr.c +4 -0
@@ -26,6 +26,8 @@ void __timer_disable_traps(struct kvm_vcpu *vcpu)/* Allow physical timer/counter access for the host */ val =read_sysreg (cnthctl_el2); val |=CNTHCTL_EL1PCTEN |CNTHCTL_EL1PCEN ; +if (cpus_have_final_cap (ARM64_ECV )) + val |=CNTHCTL_ECV ;write_sysreg (val, cnthctl_el2); }
@@ -42,6 +44,8 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)* Physical counter access is allowed */ val =read_sysreg (cnthctl_el2); +if (cpus_have_final_cap (ARM64_ECV )) + val &= ~CNTHCTL_ECV ; val &= ~CNTHCTL_EL1PCEN ; val |=CNTHCTL_EL1PCTEN ;write_sysreg (val, cnthctl_el2);
This will ensure that only the guest sees the physical offset.
Quoted 154 lines
arch/arm64/kvm/arch_timer.c +64
-2write_sysreg (val, cnthctl_el2); }@@ -1364,7 +1418,8 @@ static void set_timer_irqs(struct kvm *kvm, int vtimer_irq, int ptimer_irq) } } -int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *vtimer =vcpu_vtimer (vcpu);@@ -1397,7 +1452,37 @@ int kvm_arm_timer_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return 0 ; } -int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + +if (get_user (offset, uaddr)) +return -EFAULT ; + +update_ptimer_cntpoff (vcpu, offset); +return 0 ; +} + +int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_set_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_set_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} + +static int kvm_arm_timer_get_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *timer;@@ -1418,12 +1503,43 @@ int kvm_arm_timer_get_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return put_user (irq, uaddr); } +static int kvm_arm_timer_get_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + + offset =timer_get_offset (vcpu_ptimer (vcpu)); +return put_user (offset, uaddr); +} + +int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_get_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_get_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} +int kvm_arm_timer_has_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) {switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER :return 0 ; +case KVM_ARM_VCPU_TIMER_OFFSET : +if (cpus_have_const_cap (ARM64_ECV )) +return 0 ; +break ; }return -ENXIO ;arch/arm64/kvm/hyp/nvhe/hyp-main.c +6
-0@@ -68,6 +68,11 @@ static void handle___kvm_timer_set_cntvoff(struct kvm_cpu_context *host_ctxt)__kvm_timer_set_cntvoff (cpu_reg (host_ctxt,1 )); } +static void handle___kvm_timer_set_cntpoff (struct kvm_cpu_context *host_ctxt) +{ +__kvm_timer_set_cntpoff (cpu_reg (host_ctxt,1 )); +} +static void handle___kvm_enable_ssbs (struct kvm_cpu_context *host_ctxt) {u64 tmp;@@ -197,6 +202,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__pkvm_create_private_mapping),HANDLE_FUNC (__pkvm_prot_finalize),HANDLE_FUNC (__pkvm_mark_hyp), + HANDLE_FUNC(__kvm_timer_set_cntpoff), };static void handle_host_hcall (struct kvm_cpu_context *host_ctxt)arch/arm64/kvm/hyp/nvhe/timer-sr.c +5
-0@@ -15,6 +15,11 @@ void __kvm_timer_set_cntvoff(u64 cntvoff)write_sysreg (cntvoff, cntvoff_el2); } +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +} +/* * Should only be called on non-VHE systems. * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe(). arch/arm64/kvm/hyp/vhe/timer-sr.c +5
-0@@ -10,3 +10,8 @@ void __kvm_timer_set_cntvoff(u64 cntvoff) {write_sysreg (cntvoff, cntvoff_el2); } + +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +}include/clocksource/arm_arch_timer.h +0
-0
include/clocksource/arm_arch_timer.h line 27
Quoted 7 lines
include/clocksource/arm_arch_timer.h +1
-0@@ -21,6 +21,7 @@#define CNTHCTL_EVNTEN (1 << 2)#define CNTHCTL_EVNTDIR (1 << 3)#define CNTHCTL_EVNTI (0xF << 4) +#define CNTHCTL_ECV (1 << 12)enum arch_timer_reg {ARCH_TIMER_REG_CTRL ,
You also want to document that SCR_EL3.ECVEn has to be set to 1 for this to work (see Documentation/arm64/booting.txt). And if it isn't, the firmware better handle the CNTPOFF_EL2 traps correctly...
What firmware did you use for this? I think we need to update the boot wrapper, but that's something that can be done in parallel.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v6 17/21] KVM: arm64: Allow userspace to configure a guest's counter-timer offset
Quoted 8 lines · Marc Zyngier
Quoted 3 lines · Oliver Upton
Presently, KVM provides no facilities for correctly migrating a guest that depends on the physical counter-timer. Whie most guests (barring
nit: While
Ack.
Quoted 109 lines
Quoted 109 lines
NV, of course) should not depend on the physical counter-timer, an operator may wish to provide a consistent view of the physical counter-timer across migrations.
Provide userspace with a new vCPU attribute to modify the guest counter-timer offset. Unlike
KVM_REG_ARM_TIMER_OFFSET, this attribute is hidden from the guest's architectural state. The value offsets *both* the virtual and physical counter-timer views for the guest. Only support this attribute on ECV systems as ECV is required for hardware offsetting of the physical counter-timer.Signed-off-by Oliver Upton <[email protected]>--- Documentation/virt/kvm/devices/vcpu.rst | 28 ++++++ arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 2 + arch/arm64/include/uapi/asm/kvm.h | 1 + arch/arm64/kvm/arch_timer.c | 122 +++++++++++++++++++++++- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 ++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 5 + arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + include/clocksource/arm_arch_timer.h | 1 + 9 files changed, 169 insertions(+), 3 deletions(-)Documentation/virt/kvm/devices/vcpu.rst +28
-0@@ -139,6 +139,34 @@ configured values on other VCPUs. Userspace should configure the interrupt numbers on at least one VCPU after creating all VCPUs and before running any VCPUs. +2.2. ATTRIBUTE: KVM_ARM_VCPU_TIMER_OFFSET +----------------------------------------- + +:Parameters: in kvm_device_attr.addr the address for the timer offset is a + pointer to a __u64 + +Returns: + + ======= ================================== + -EFAULT Error reading/writing the provided + parameter address + -ENXIO Timer offsetting not implemented + ======= ================================== + +Specifies the guest's counter-timer offset from the host's virtual counter. +The guest's physical counter value is then derived by the following +equation: + + guest_cntpct = host_cntvct - KVM_ARM_VCPU_TIMER_OFFSET + +The guest's virtual counter value is derived by the following equation: + + guest_cntvct = host_cntvct - KVM_REG_ARM_TIMER_OFFSET + - KVM_ARM_VCPU_TIMER_OFFSET + +KVM does not allow the use of varying offset values for different vCPUs; +the last written offset value will be broadcasted to all vCPUs in a VM. + 3. GROUP: KVM_ARM_VCPU_PVTIME_CTRL ==================================arch/arm64/include/asm/kvm_asm.h +2
-0@@ -65,6 +65,7 @@#define __KVM_HOST_SMCCC_FUNC___pkvm_prot_finalize 19#define __KVM_HOST_SMCCC_FUNC___pkvm_mark_hyp 20#define __KVM_HOST_SMCCC_FUNC___kvm_adjust_pc 21 +#define __KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntpoff 22#ifndef __ASSEMBLY__@@ -200,6 +201,7 @@ extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,extern void __kvm_tlb_flush_vmid (struct kvm_s2_mmu *mmu);extern void __kvm_timer_set_cntvoff (u64 cntvoff); +extern void __kvm_timer_set_cntpoff (u64 cntpoff);extern int __kvm_vcpu_run (struct kvm_vcpu *vcpu);arch/arm64/include/asm/sysreg.h +2
-0@@ -586,6 +586,8 @@#define SYS_ICH_LR14_EL2 __SYS__LR8_EL2(6)#define SYS_ICH_LR15_EL2 __SYS__LR8_EL2(7) +#define SYS_CNTPOFF_EL2 sys_reg(3, 4, 14, 0, 6) +/* VHE encodings for architectural EL0/1 system registers */ #define SYS_SCTLR_EL12 sys_reg(3, 5, 1, 0, 0)#define SYS_CPACR_EL12 sys_reg(3, 5, 1, 0, 2)arch/arm64/include/uapi/asm/kvm.h +1
-0@@ -366,6 +366,7 @@ struct kvm_arm_copy_mte_tags {#define KVM_ARM_VCPU_TIMER_CTRL 1#define KVM_ARM_VCPU_TIMER_IRQ_VTIMER 0#define KVM_ARM_VCPU_TIMER_IRQ_PTIMER 1 +#define KVM_ARM_VCPU_TIMER_OFFSET 2#define KVM_ARM_VCPU_PVTIME_CTRL 2#define KVM_ARM_VCPU_PVTIME_IPA 0
Quoted 11 lines
arch/arm64/kvm/arch_timer.c u64 timer_get_cval(struct arch_timer_context *ctxt) · line 88
Quoted 9 lines
arch/arm64/kvm/arch_timer.c +1
-0@@ -85,11 +85,15 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)static u64 timer_get_offset (struct arch_timer_context *ctxt) {struct kvm_vcpu *vcpu = ctxt->vcpu; +struct arch_timer_cpu *timer =vcpu_timer (vcpu);Unused variable?
Yep!
Quoted 34 lines
Quoted 34 lines
arch/arm64/kvm/arch_timer.c +9
-0switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER :return ctxt->host_offset;default : +WARN_ONCE (1 ,"unrecognized timer %ld\n" , +arch_timer_ctx_index (ctxt));return 0 ; } }@@ -144,6 +148,7 @@ static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)switch (arch_timer_ctx_index (ctxt)) {case TIMER_VTIMER : +case TIMER_PTIMER : ctxt->host_offset = offset;break ;default :@@ -572,6 +577,11 @@ static void set_cntvoff(u64 cntvoff)kvm_call_hyp (__kvm_timer_set_cntvoff, cntvoff); } +static void set_cntpoff (u64 cntpoff) +{ +kvm_call_hyp (__kvm_timer_set_cntpoff, cntpoff); +} +static inline void set_timer_irq_phys_active (struct arch_timer_context *ctx,bool active) {int r;@@ -647,6 +657,8 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) }
Quoted 5 lines
arch/arm64/kvm/arch_timer.c void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) · line 660
Quoted 2 lines
arch/arm64/kvm/arch_timer.c +1
-0set_cntvoff (timer_get_offset (map.direct_vtimer)); +if (cpus_have_const_cap (ARM64_ECV ))This really should be a final cap instead (same for all the other use cases).
Ack.
Quoted 72 lines
arch/arm64/kvm/arch_timer.c void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu) · line 661
Quoted 1 line
arch/arm64/kvm/arch_timer.c +1
-0+set_cntpoff (timer_get_offset (vcpu_ptimer (vcpu)));However, tripping to EL2 for each offset on nVHE may prove to be an unnecessary overhead. Not a problem for now anyway.
Quoted 38 lines
arch/arm64/kvm/arch_timer.c +24
-0kvm_timer_unblocking (vcpu);@@ -814,6 +826,22 @@ static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)mutex_unlock (&kvm->lock); } +static void update_ptimer_cntpoff (struct kvm_vcpu *vcpu,u64 cntpoff) +{ +struct kvm *kvm = vcpu->kvm; +u64 cntvoff; + +mutex_lock (&kvm->lock); + +/* adjustments to the physical offset also affect vtimer */ + cntvoff =timer_get_offset (vcpu_vtimer (vcpu)); + cntvoff += cntpoff -timer_get_offset (vcpu_ptimer (vcpu)); + +update_timer_offset (vcpu,TIMER_PTIMER , cntpoff, false); +update_timer_offset (vcpu,TIMER_VTIMER , cntvoff, false); +mutex_unlock (&kvm->lock); +} +void kvm_timer_vcpu_init (struct kvm_vcpu *vcpu) {struct arch_timer_cpu *timer =vcpu_timer (vcpu);@@ -932,6 +960,29 @@ u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid)return (u64)-1 ; } +/** +* kvm_arm_timer_read_offset - returns the guest value of CNTVOFF_EL2. +* @vcpu: the vcpu pointer +* +* Computes the guest value of CNTVOFF_EL2 by subtracting the physical +* counter offset. Note that KVM defines CNTVOFF_EL2 as the offset from the +* guest's physical counter-timer, not the host's. +* arch/arm64/kvm/arch_timer.c u64 kvm_arm_timer_get_reg(struct kvm_vcpu *vcpu, u64 regid) · line 980
Quoted 10 lines
arch/arm64/kvm/arch_timer.c +10
-0+* Returns: the guest value for CNTVOFF_EL2 +*/ +static u64 kvm_arm_timer_read_offset (struct kvm_vcpu *vcpu) +{ +struct kvm *kvm = vcpu->kvm; +u64 offset; + +mutex_lock (&kvm->lock); + offset =timer_get_offset (vcpu_vtimer (vcpu)) - +timer_get_offset (vcpu_ptimer (vcpu));nit: please keep this on a single line.
Quoted 16 lines
arch/arm64/kvm/arch_timer.c +6
-1+mutex_unlock (&kvm->lock); + +return offset; +} +static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg)@@ -957,7 +1008,7 @@ static u64 kvm_arm_timer_read(struct kvm_vcpu *vcpu,break ;case TIMER_REG_OFFSET : - val =timer_get_offset (timer); + val =kvm_arm_timer_read_offset (vcpu);break ;
Quoted 12 lines
arch/arm64/kvm/arch_timer.c void kvm_timer_init_vhe(void) · line 1406
Quoted 8 lines
arch/arm64/kvm/arch_timer.c +3
-0default :@@ -1350,6 +1401,9 @@ void kvm_timer_init_vhe(void) val =read_sysreg (cnthctl_el2); val |= (CNTHCTL_EL1PCEN << cnthctl_shift); val |= (CNTHCTL_EL1PCTEN << cnthctl_shift); + +if (cpus_have_const_cap (ARM64_ECV )) + val |=CNTHCTL_ECV ;I cannot immediately see where you are doing the equivalent enablement of ECV on the nVHE path. Obviously, it has to be done eagerly from EL2, together with the rest of the EL1 timer setup. Something like:
Yep, I had caught this when I was actually able to run the Base FVP. I have a fix (same as yours, basically) but held back until you reviewed to avoid storming your inbox :)
Quoted 184 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c +4
-0@@ -26,6 +26,8 @@ void __timer_disable_traps(struct kvm_vcpu *vcpu)/* Allow physical timer/counter access for the host */ val =read_sysreg (cnthctl_el2); val |=CNTHCTL_EL1PCTEN |CNTHCTL_EL1PCEN ; +if (cpus_have_final_cap (ARM64_ECV )) + val |=CNTHCTL_ECV ;write_sysreg (val, cnthctl_el2); }@@ -42,6 +44,8 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)* Physical counter access is allowed */ val =read_sysreg (cnthctl_el2); +if (cpus_have_final_cap (ARM64_ECV )) + val &= ~CNTHCTL_ECV ; val &= ~CNTHCTL_EL1PCEN ; val |=CNTHCTL_EL1PCTEN ;write_sysreg (val, cnthctl_el2);This will ensure that only the guest sees the physical offset.
Quoted 159 lines
arch/arm64/kvm/arch_timer.c +64
-2write_sysreg (val, cnthctl_el2); }@@ -1364,7 +1418,8 @@ static void set_timer_irqs(struct kvm *kvm, int vtimer_irq, int ptimer_irq) } } -int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *vtimer =vcpu_vtimer (vcpu);@@ -1397,7 +1452,37 @@ int kvm_arm_timer_set_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return 0 ; } -int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +static int kvm_arm_timer_set_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + +if (get_user (offset, uaddr)) +return -EFAULT ; + +update_ptimer_cntpoff (vcpu, offset); +return 0 ; +} + +int kvm_arm_timer_set_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_set_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_set_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} + +static int kvm_arm_timer_get_attr_irq (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) {int __user *uaddr = (int __user *)(long )attr->addr;struct arch_timer_context *timer;@@ -1418,12 +1503,43 @@ int kvm_arm_timer_get_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)return put_user (irq, uaddr); } +static int kvm_arm_timer_get_attr_offset (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +u64 __user *uaddr = (u64 __user *)(long )attr->addr; +u64 offset; + +if (!cpus_have_const_cap (ARM64_ECV )) +return -ENXIO ; + + offset =timer_get_offset (vcpu_ptimer (vcpu)); +return put_user (offset, uaddr); +} + +int kvm_arm_timer_get_attr (struct kvm_vcpu *vcpu, +struct kvm_device_attr *attr) +{ +switch (attr->attr) { +case KVM_ARM_VCPU_TIMER_IRQ_VTIMER : +case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : +return kvm_arm_timer_get_attr_irq (vcpu, attr); +case KVM_ARM_VCPU_TIMER_OFFSET : +return kvm_arm_timer_get_attr_offset (vcpu, attr); +default : +return -ENXIO ; + } +} +int kvm_arm_timer_has_attr (struct kvm_vcpu *vcpu,struct kvm_device_attr *attr) {switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER :return 0 ; +case KVM_ARM_VCPU_TIMER_OFFSET : +if (cpus_have_const_cap (ARM64_ECV )) +return 0 ; +break ; }return -ENXIO ;arch/arm64/kvm/hyp/nvhe/hyp-main.c +6
-0@@ -68,6 +68,11 @@ static void handle___kvm_timer_set_cntvoff(struct kvm_cpu_context *host_ctxt)__kvm_timer_set_cntvoff (cpu_reg (host_ctxt,1 )); } +static void handle___kvm_timer_set_cntpoff (struct kvm_cpu_context *host_ctxt) +{ +__kvm_timer_set_cntpoff (cpu_reg (host_ctxt,1 )); +} +static void handle___kvm_enable_ssbs (struct kvm_cpu_context *host_ctxt) {u64 tmp;@@ -197,6 +202,7 @@ static const hcall_t host_hcall[] = {HANDLE_FUNC (__pkvm_create_private_mapping),HANDLE_FUNC (__pkvm_prot_finalize),HANDLE_FUNC (__pkvm_mark_hyp), + HANDLE_FUNC(__kvm_timer_set_cntpoff), };static void handle_host_hcall (struct kvm_cpu_context *host_ctxt)arch/arm64/kvm/hyp/nvhe/timer-sr.c +5
-0@@ -15,6 +15,11 @@ void __kvm_timer_set_cntvoff(u64 cntvoff)write_sysreg (cntvoff, cntvoff_el2); } +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +} +/* * Should only be called on non-VHE systems. * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe(). arch/arm64/kvm/hyp/vhe/timer-sr.c +5
-0@@ -10,3 +10,8 @@ void __kvm_timer_set_cntvoff(u64 cntvoff) {write_sysreg (cntvoff, cntvoff_el2); } + +void __kvm_timer_set_cntpoff (u64 cntpoff) +{ +write_sysreg_s (cntpoff,SYS_CNTPOFF_EL2 ); +}include/clocksource/arm_arch_timer.h +1
-0@@ -21,6 +21,7 @@#define CNTHCTL_EVNTEN (1 << 2)#define CNTHCTL_EVNTDIR (1 << 3)#define CNTHCTL_EVNTI (0xF << 4) +#define CNTHCTL_ECV (1 << 12)
Quoted 6 lines
include/clocksource/arm_arch_timer.h line 27
Quoted 2 lines
include/clocksource/arm_arch_timer.h +0
-0enum arch_timer_reg {ARCH_TIMER_REG_CTRL ,You also want to document that SCR_EL3.ECVEn has to be set to 1 for this to work (see
Documentation/arm64/booting.txt). And if it isn't, the firmware better handle theCNTPOFF_EL2traps correctly...
I'll grab the popcorn now ;-) Adding docs for this, good idea.
Quoted 2 lines
What firmware did you use for this? I think we need to update the boot wrapper, but that's something that can be done in parallel.
I had actually just done a direct boot from ARM-TF -> Linux, nothing else in between.
--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v6 17/21] KVM: arm64: Allow userspace to configure a guest's counter-timer offset
Quoted 5 lines · Oliver Upton
Quoted 3 lines · Marc Zyngier
On Wed, 04 Aug 2021 09:58:15 +0100,
Oliver Upton <[email protected]> wrote:
[...]
Quoted 17 lines
Quoted 10 lines
Quoted 10 lines
include/clocksource/arm_arch_timer.h +1
-0@@ -21,6 +21,7 @@#define CNTHCTL_EVNTEN (1 << 2)#define CNTHCTL_EVNTDIR (1 << 3)#define CNTHCTL_EVNTI (0xF << 4) +#define CNTHCTL_ECV (1 << 12)Quoted 6 lines
include/clocksource/arm_arch_timer.h line 27
Quoted 2 lines
include/clocksource/arm_arch_timer.h +0
-0enum arch_timer_reg {ARCH_TIMER_REG_CTRL ,You also want to document that SCR_EL3.ECVEn has to be set to 1 for this to work (see
Documentation/arm64/booting.txt). And if it isn't, the firmware better handle theCNTPOFF_EL2traps correctly...
Quoted 7 lines
I'll grab the popcorn now ;-) Adding docs for this, good idea.
Quoted 2 lines
What firmware did you use for this? I think we need to update the boot wrapper, but that's something that can be done in parallel.
I had actually just done a direct boot from ARM-TF -> Linux, nothing else in between.
Ah, right. I tend to use the boot-wrapper[1] to build a single binary that contains the 'boot loader', DT and kernel. Using ATF is probably more representative of the final thing, but the boot-wrapper is dead easy to hack on...
Thanks,
M.
[1] git://git.kernel.org/pub/scm/linux/kernel/git/mark/boot-wrapper-aarch64.git
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v6 18/21 KVM: arm64: Configure timer traps in vcpu_load() for VHE
In reply to Oliver UptonIn preparation for emulated physical counter-timer offsetting, configure traps on every vcpu_load() for VHE systems. As before, these trap settings do not affect host userspace, and are only active for the guest.
--- arch/arm64/kvm/arch_timer.c | 10 +++++++--- arch/arm64/kvm/arm.c | 4 +--- include/kvm/arm_arch_timer.h | 2 -- 3 files changed, 8 insertions(+), 8 deletions(-)
arch/arm64/kvm/arch_timer.c +7 -3
@@ -51,6 +51,7 @@ static void kvm_arm_timer_write(struct kvm_vcpu *vcpu,static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg); +static void kvm_timer_enable_traps_vhe (void );u32 timer_get_ctl (struct arch_timer_context *ctxt) {
@@ -668,6 +669,9 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)if (map.emul_ptimer)timer_emulate (map.emul_ptimer); + +if (has_vhe ()) +kvm_timer_enable_traps_vhe (); }bool kvm_timer_should_notify_user (struct kvm_vcpu *vcpu)
@@ -1383,12 +1387,12 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu) }/* -* On VHE system, we only need to configure the EL2 timer trap register once, -* not for every world switch. +* On VHE system, we only need to configure the EL2 timer trap register on +* vcpu_load(), but not every world switch into the guest. * The host kernel runs at EL2 with HCR_EL2.TGE == 1, * and this makes those bits have no effect for the host kernel execution. */ -void kvm_timer_init_vhe (void ) +static void kvm_timer_enable_traps_vhe (void ) {/* When HCR_EL2.E2H ==1, EL1PCEN and EL1PCTEN are shifted by 10 */ u32 cnthctl_shift =10 ;
arch/arm64/kvm/arm.c +1 -3
@@ -1558,9 +1558,7 @@ static void cpu_hyp_reinit(void)cpu_hyp_reset (); -if (is_kernel_in_hyp_mode ()) -kvm_timer_init_vhe (); -else +if (!is_kernel_in_hyp_mode ())cpu_init_hyp_mode ();cpu_set_hyp_vector ();
include/kvm/arm_arch_timer.h +0 -2
@@ -87,8 +87,6 @@ u64 kvm_phys_timer_read(void);void kvm_timer_vcpu_load (struct kvm_vcpu *vcpu);void kvm_timer_vcpu_put (struct kvm_vcpu *vcpu); -void kvm_timer_init_vhe (void ); -bool kvm_arch_timer_get_input_level (int vintid);#define vcpu_timer (v) (&(v)->arch.timer_cpu)
Signature · 6 lines
2.32.0.605.g8dce9f2422-goog _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v6 18/21] KVM: arm64: Configure timer traps in vcpu_load() for VHE
Quoted 82 lines · Oliver Upton
In preparation for emulated physical counter-timer offsetting, configure traps on every
vcpu_load()for VHE systems. As before, these trap settings do not affect host userspace, and are only active for the guest.Suggested-by Marc Zyngier <[email protected]>Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/kvm/arch_timer.c | 10 +++++++--- arch/arm64/kvm/arm.c | 4 +--- include/kvm/arm_arch_timer.h | 2 -- 3 files changed, 8 insertions(+), 8 deletions(-)arch/arm64/kvm/arch_timer.c +7
-3@@ -51,6 +51,7 @@ static void kvm_arm_timer_write(struct kvm_vcpu *vcpu,static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg); +static void kvm_timer_enable_traps_vhe (void );u32 timer_get_ctl (struct arch_timer_context *ctxt) {@@ -668,6 +669,9 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)if (map.emul_ptimer)timer_emulate (map.emul_ptimer); + +if (has_vhe ()) +kvm_timer_enable_traps_vhe (); }bool kvm_timer_should_notify_user (struct kvm_vcpu *vcpu)@@ -1383,12 +1387,12 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu) }/* -* On VHE system, we only need to configure the EL2 timer trap register once, -* not for every world switch. +* On VHE system, we only need to configure the EL2 timer trap register on +* vcpu_load(), but not every world switch into the guest. * The host kernel runs at EL2 with HCR_EL2.TGE == 1, * and this makes those bits have no effect for the host kernel execution. */ -void kvm_timer_init_vhe (void ) +static void kvm_timer_enable_traps_vhe (void ) {/* When HCR_EL2.E2H ==1, EL1PCEN and EL1PCTEN are shifted by 10 */ u32 cnthctl_shift =10 ;arch/arm64/kvm/arm.c +1
-3@@ -1558,9 +1558,7 @@ static void cpu_hyp_reinit(void)cpu_hyp_reset (); -if (is_kernel_in_hyp_mode ()) -kvm_timer_init_vhe (); -else +if (!is_kernel_in_hyp_mode ())cpu_init_hyp_mode ();cpu_set_hyp_vector ();include/kvm/arm_arch_timer.h +0
-2@@ -87,8 +87,6 @@ u64 kvm_phys_timer_read(void);void kvm_timer_vcpu_load (struct kvm_vcpu *vcpu);void kvm_timer_vcpu_put (struct kvm_vcpu *vcpu); -void kvm_timer_init_vhe (void ); -bool kvm_arch_timer_get_input_level (int vintid);#define vcpu_timer (v) (&(v)->arch.timer_cpu)Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 19/21 KVM: arm64: Emulate physical counter offsetting on non-ECV systems
In reply to Oliver UptonUnfortunately, ECV hasn't yet arrived in any tangible hardware. At the same time, controlling the guest view of the physical counter-timer is useful. Support guest counter-timer offsetting on non-ECV systems by trapping guest accesses to the physical counter-timer. Emulate reads of the physical counter in the fast exit path.
--- arch/arm64/include/asm/sysreg.h | 1 + arch/arm64/kvm/arch_timer.c | 53 +++++++++++++++---------- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++++++++++++++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 11 ++++- 4 files changed, 70 insertions(+), 24 deletions(-)
arch/arm64/include/asm/sysreg.h +1 -0
@@ -505,6 +505,7 @@#define SYS_AMEVCNTR0_MEM_STALL SYS_AMEVCNTR0_EL0(3)#define SYS_CNTFRQ_EL0 sys_reg(3, 3, 14, 0, 0) +#define SYS_CNTPCT_EL0 sys_reg(3, 3, 14, 0, 1)#define SYS_CNTP_TVAL_EL0 sys_reg(3, 3, 14, 2, 0)#define SYS_CNTP_CTL_EL0 sys_reg(3, 3, 14, 2, 1)
arch/arm64/kvm/arch_timer.c +31 -22
110 lines not on this page · raw message
arch/arm64/kvm/hyp/include/hyp/switch.h +29 -0
@@ -15,6 +15,7 @@#include <linux/jump_label.h> #include <uapi/linux/psci.h> +#include <kvm/arm_arch_timer.h> #include <kvm/arm_psci.h> #include <asm/barrier.h>
@@ -405,6 +406,31 @@ static inline bool __hyp_handle_ptrauth(struct kvm_vcpu *vcpu)return true; } +static inline u64 __timer_read_cntpct (struct kvm_vcpu *vcpu) +{ +return __arch_counter_get_cntpct () -vcpu_ptimer (vcpu)->host_offset; +} + +static inline bool __hyp_handle_counter (struct kvm_vcpu *vcpu) +{ +u32 sysreg; +int rt; +u64 rv; + +if (kvm_vcpu_trap_get_class (vcpu) != ESR_ELx_EC_SYS64) +return false; + + sysreg =esr_sys64_to_sysreg (kvm_vcpu_get_esr (vcpu)); +if (sysreg !=SYS_CNTPCT_EL0 ) +return false; + + rt =kvm_vcpu_sys_get_rt (vcpu); + rv =__timer_read_cntpct (vcpu); +vcpu_set_reg (vcpu, rt, rv); +__kvm_skip_instr (vcpu); +return true; +} +/* * Return true when we were able to fixup the guest exit and should return to * the guest, false when we should restore the host state and return to the
@@ -439,6 +465,9 @@ static inline bool fixup_guest_exit(struct kvm_vcpu *vcpu, u64 *exit_code)if (*exit_code !=ARM_EXCEPTION_TRAP ) goto exit; +if (__hyp_handle_counter (vcpu)) + goto guest; +if (cpus_have_final_cap (ARM64_WORKAROUND_CAVIUM_TX2_219_TVM ) &&kvm_vcpu_trap_get_class (vcpu) == ESR_ELx_EC_SYS64 &&handle_tx2_tvm (vcpu))
arch/arm64/kvm/hyp/nvhe/timer-sr.c +9 -2
@@ -44,10 +44,17 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)/* * Disallow physical timer access for the guest -* Physical counter access is allowed */ val =read_sysreg (cnthctl_el2); val &= ~CNTHCTL_EL1PCEN ; - val |=CNTHCTL_EL1PCTEN ; + +/* +* Disallow physical counter access for the guest if offsetting is +* requested on a non-ECV system. +*/ +if (vcpu_ptimer (vcpu)->host_offset && !cpus_have_const_cap (ARM64_ECV )) + val &= ~CNTHCTL_EL1PCTEN ; +else + val |=CNTHCTL_EL1PCTEN ;write_sysreg (val, cnthctl_el2); }
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 19/21] KVM: arm64: Emulate physical counter offsetting on non-ECV systems
Quoted 40 lines · Oliver Upton
Unfortunately, ECV hasn't yet arrived in any tangible hardware. At the same time, controlling the guest view of the physical counter-timer is useful. Support guest counter-timer offsetting on non-ECV systems by trapping guest accesses to the physical counter-timer. Emulate reads of the physical counter in the fast exit path.
Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/include/asm/sysreg.h | 1 + arch/arm64/kvm/arch_timer.c | 53 +++++++++++++++---------- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++++++++++++++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 11 ++++- 4 files changed, 70 insertions(+), 24 deletions(-)arch/arm64/include/asm/sysreg.h +1
-0@@ -505,6 +505,7 @@#define SYS_AMEVCNTR0_MEM_STALL SYS_AMEVCNTR0_EL0(3)#define SYS_CNTFRQ_EL0 sys_reg(3, 3, 14, 0, 0) +#define SYS_CNTPCT_EL0 sys_reg(3, 3, 14, 0, 1)#define SYS_CNTP_TVAL_EL0 sys_reg(3, 3, 14, 2, 0)#define SYS_CNTP_CTL_EL0 sys_reg(3, 3, 14, 2, 1)arch/arm64/kvm/arch_timer.c +1
-1@@ -51,7 +51,7 @@ static void kvm_arm_timer_write(struct kvm_vcpu *vcpu,static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg); -static void kvm_timer_enable_traps_vhe (void ); +static void kvm_timer_enable_traps_vhe (struct kvm_vcpu *vcpu);u32 timer_get_ctl (struct arch_timer_context *ctxt) {
arch/arm64/kvm/arch_timer.c static void timer_set_guest_offset(struct arch_timer_context *ctxt, u64 offset) · line 181
Quoted 7 lines
arch/arm64/kvm/arch_timer.c +4
-0@@ -175,6 +175,12 @@ static void timer_set_guest_offset(struct arch_timer_context *ctxt, u64 offset) } } +static bool ptimer_emulation_required (struct kvm_vcpu *vcpu) +{ +return timer_get_offset (vcpu_ptimer (vcpu)) && + !cpus_have_const_cap (ARM64_ECV );
Whenever I see a static branch check and something else in the same condition, I always wonder if we could trim a few instructions for the static branch is false case by testing it first.
Quoted 89 lines
arch/arm64/kvm/arch_timer.c +25
-17+} +u64 kvm_phys_timer_read (void ) {return timecounter->cc->read (timecounter->cc);@@ -184,8 +190,13 @@ static void get_timer_map(struct kvm_vcpu *vcpu, struct timer_map *map) {if (has_vhe ()) { map->direct_vtimer =vcpu_vtimer (vcpu); - map->direct_ptimer =vcpu_ptimer (vcpu); - map->emul_ptimer =NULL ; +if (!ptimer_emulation_required (vcpu)) { + map->direct_ptimer =vcpu_ptimer (vcpu); + map->emul_ptimer =NULL ; + }else { + map->direct_ptimer =NULL ; + map->emul_ptimer =vcpu_ptimer (vcpu); + } }else { map->direct_vtimer =vcpu_vtimer (vcpu); map->direct_ptimer =NULL ;@@ -671,7 +682,7 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)timer_emulate (map.emul_ptimer);if (has_vhe ()) -kvm_timer_enable_traps_vhe (); +kvm_timer_enable_traps_vhe (vcpu); }bool kvm_timer_should_notify_user (struct kvm_vcpu *vcpu)@@ -1392,22 +1403,29 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu)* The host kernel runs at EL2 with HCR_EL2.TGE == 1, * and this makes those bits have no effect for the host kernel execution. */ -static void kvm_timer_enable_traps_vhe (void ) +static void kvm_timer_enable_traps_vhe (struct kvm_vcpu *vcpu) {/* When HCR_EL2.E2H ==1, EL1PCEN and EL1PCTEN are shifted by 10 */ u32 cnthctl_shift =10 ; -u64 val; +u64 val, mask; + + mask =CNTHCTL_EL1PCEN << cnthctl_shift; + mask |=CNTHCTL_EL1PCTEN << cnthctl_shift; -/* -* VHE systems allow the guest direct access to the EL1 physical -* timer/counter. -*/ val =read_sysreg (cnthctl_el2); - val |= (CNTHCTL_EL1PCEN << cnthctl_shift); - val |= (CNTHCTL_EL1PCTEN << cnthctl_shift);if (cpus_have_const_cap (ARM64_ECV )) val |=CNTHCTL_ECV ; + +/* +* VHE systems allow the guest direct access to the EL1 physical +* timer/counter if offsetting isn't requested on a non-ECV system. +*/ +if (ptimer_emulation_required (vcpu)) + val &= ~mask; +else + val |= mask; +write_sysreg (val, cnthctl_el2); }@@ -1462,9 +1480,6 @@ static int kvm_arm_timer_set_attr_offset(struct kvm_vcpu *vcpu,u64 __user *uaddr = (u64 __user *)(long )attr->addr;u64 offset; -if (!cpus_have_const_cap (ARM64_ECV )) -return -ENXIO ; -if (get_user (offset, uaddr))return -EFAULT ;@@ -1513,9 +1528,6 @@ static int kvm_arm_timer_get_attr_offset(struct kvm_vcpu *vcpu,u64 __user *uaddr = (u64 __user *)(long )attr->addr;u64 offset; -if (!cpus_have_const_cap (ARM64_ECV )) -return -ENXIO ; - offset =timer_get_offset (vcpu_ptimer (vcpu));return put_user (offset, uaddr); }
arch/arm64/kvm/arch_timer.c int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1555
Quoted 9 lines
arch/arm64/kvm/arch_timer.c +1
-4@@ -1539,11 +1551,8 @@ int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : -return 0 ;case KVM_ARM_VCPU_TIMER_OFFSET : -if (cpus_have_const_cap (ARM64_ECV )) -return 0 ; -break ; +return 0 ;
So now, if userspace wants to know when they're using an emulated TIMER_OFFSET vs. ECV, then they'll need to check the HWCAP. I guess that's fair. We should update the selftest to report what it's testing when the HWCAP is available.
Quoted 66 lines
arch/arm64/kvm/arch_timer.c +0
-0}return -ENXIO ;arch/arm64/kvm/hyp/include/hyp/switch.h +29
-0@@ -15,6 +15,7 @@#include <linux/jump_label.h> #include <uapi/linux/psci.h> +#include <kvm/arm_arch_timer.h> #include <kvm/arm_psci.h> #include <asm/barrier.h> @@ -405,6 +406,31 @@ static inline bool __hyp_handle_ptrauth(struct kvm_vcpu *vcpu)return true; } +static inline u64 __timer_read_cntpct (struct kvm_vcpu *vcpu) +{ +return __arch_counter_get_cntpct () -vcpu_ptimer (vcpu)->host_offset; +} + +static inline bool __hyp_handle_counter (struct kvm_vcpu *vcpu) +{ +u32 sysreg; +int rt; +u64 rv; + +if (kvm_vcpu_trap_get_class (vcpu) != ESR_ELx_EC_SYS64) +return false; + + sysreg =esr_sys64_to_sysreg (kvm_vcpu_get_esr (vcpu)); +if (sysreg !=SYS_CNTPCT_EL0 ) +return false; + + rt =kvm_vcpu_sys_get_rt (vcpu); + rv =__timer_read_cntpct (vcpu); +vcpu_set_reg (vcpu, rt, rv); +__kvm_skip_instr (vcpu); +return true; +} +/* * Return true when we were able to fixup the guest exit and should return to * the guest, false when we should restore the host state and return to the @@ -439,6 +465,9 @@ static inline bool fixup_guest_exit(struct kvm_vcpu *vcpu, u64 *exit_code)if (*exit_code !=ARM_EXCEPTION_TRAP ) goto exit; +if (__hyp_handle_counter (vcpu)) + goto guest; +if (cpus_have_final_cap (ARM64_WORKAROUND_CAVIUM_TX2_219_TVM ) &&kvm_vcpu_trap_get_class (vcpu) == ESR_ELx_EC_SYS64 &&handle_tx2_tvm (vcpu))arch/arm64/kvm/hyp/nvhe/timer-sr.c +0
-1@@ -44,10 +44,17 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)/* * Disallow physical timer access for the guest -* Physical counter access is allowed
arch/arm64/kvm/hyp/nvhe/timer-sr.c void __timer_enable_traps(struct kvm_vcpu *vcpu) · line 55
Quoted 10 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c +6
-1*/ val =read_sysreg (cnthctl_el2); val &= ~CNTHCTL_EL1PCEN ; - val |=CNTHCTL_EL1PCTEN ; + +/* +* Disallow physical counter access for the guest if offsetting is +* requested on a non-ECV system. +*/ +if (vcpu_ptimer (vcpu)->host_offset && !cpus_have_const_cap (ARM64_ECV ))
Shouldn't we expose and reuse ptimer_emulation_required() here?
Quoted 8 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c +3
-0+ val &= ~CNTHCTL_EL1PCTEN ; + else + val |= CNTHCTL_EL1PCTEN;write_sysreg (val, cnthctl_el2); }Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Otherwise,
Re: [PATCH v6 19/21] KVM: arm64: Emulate physical counter offsetting on non-ECV systems
Hi Drew,
Quoted 8 lines · Andrew Jones
arch/arm64/kvm/arch_timer.c static void timer_set_guest_offset(struct arch_timer_context *ctxt, u64 offset) · line 181
Quoted 4 lines
arch/arm64/kvm/arch_timer.c +4
-0+static bool ptimer_emulation_required (struct kvm_vcpu *vcpu) +{ +return timer_get_offset (vcpu_ptimer (vcpu)) && + !cpus_have_const_cap (ARM64_ECV );Whenever I see a static branch check and something else in the same condition, I always wonder if we could trim a few instructions for the static branch is false case by testing it first.
Good point, I'll reclaim those few cycles in the next spin ;-)
Quoted 7 lines
Quoted 7 lines
arch/arm64/kvm/arch_timer.c +0
-2@@ -1539,11 +1551,8 @@ int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : -return 0 ;case KVM_ARM_VCPU_TIMER_OFFSET : -if (cpus_have_const_cap (ARM64_ECV ))
Quoted 9 lines
arch/arm64/kvm/arch_timer.c int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr) · line 1555
Quoted 3 lines
arch/arm64/kvm/arch_timer.c +1
-2-return 0 ; -break ; +return 0 ;So now, if userspace wants to know when they're using an emulated
TIMER_OFFSETvs. ECV, then they'll need to check the HWCAP. I guess that's fair. We should update the selftest to report what it's testing when the HWCAP is available.
Hmm...
I hadn't yet wired up the ECV cpufeature bits to an ELF HWCAP, but this point is a bit interesting. I can see the argument being made that we shouldn't have two ELF HWCAP bits for ECV (depending on partial or full ECV support). ECV=0x1 is most certainly of interest to userspace, since self-synchronized views of the counter are then available. However, ECV=0x2 is purely of interest to EL2.
What if we only had only one ELF HWCAP bit for ECV >= 0x1? We could let userspace read ID_AA64MMFR0_EL1.ECV if it really needs to know about ECV = 0x2.
Quoted 4 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c void __timer_enable_traps(struct kvm_vcpu *vcpu) · line 55
Quoted 1 line
arch/arm64/kvm/hyp/nvhe/timer-sr.c +1
-0+if (vcpu_ptimer (vcpu)->host_offset && !cpus_have_const_cap (ARM64_ECV ))Shouldn't we expose and reuse
ptimer_emulation_required()here?
Agreed, makes it much cleaner.
Quoted 9 lines
Quoted 8 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c +3
-1+ val &= ~CNTHCTL_EL1PCTEN ; + else + val |= CNTHCTL_EL1PCTEN;write_sysreg (val, cnthctl_el2); } --2.32.0.605.g8dce9f2422-goog
Quoted 4 lines
Otherwise,
Reviewed-by Andrew Jones <[email protected]>
Thanks!
--
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v6 19/21] KVM: arm64: Emulate physical counter offsetting on non-ECV systems
In reply to Oliver UptonQuoted 41 lines · Oliver Upton
Unfortunately, ECV hasn't yet arrived in any tangible hardware. At the same time, controlling the guest view of the physical counter-timer is useful. Support guest counter-timer offsetting on non-ECV systems by trapping guest accesses to the physical counter-timer. Emulate reads of the physical counter in the fast exit path.
Signed-off-by Oliver Upton <[email protected]>--- arch/arm64/include/asm/sysreg.h | 1 + arch/arm64/kvm/arch_timer.c | 53 +++++++++++++++---------- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++++++++++++++ arch/arm64/kvm/hyp/nvhe/timer-sr.c | 11 ++++- 4 files changed, 70 insertions(+), 24 deletions(-)arch/arm64/include/asm/sysreg.h +1
-0@@ -505,6 +505,7 @@#define SYS_AMEVCNTR0_MEM_STALL SYS_AMEVCNTR0_EL0(3)#define SYS_CNTFRQ_EL0 sys_reg(3, 3, 14, 0, 0) +#define SYS_CNTPCT_EL0 sys_reg(3, 3, 14, 0, 1)#define SYS_CNTP_TVAL_EL0 sys_reg(3, 3, 14, 2, 0)#define SYS_CNTP_CTL_EL0 sys_reg(3, 3, 14, 2, 1)arch/arm64/kvm/arch_timer.c +1
-1@@ -51,7 +51,7 @@ static void kvm_arm_timer_write(struct kvm_vcpu *vcpu,static u64 kvm_arm_timer_read (struct kvm_vcpu *vcpu,struct arch_timer_context *timer,enum kvm_arch_timer_regs treg); -static void kvm_timer_enable_traps_vhe (void ); +static void kvm_timer_enable_traps_vhe (struct kvm_vcpu *vcpu);u32 timer_get_ctl (struct arch_timer_context *ctxt) {
arch/arm64/kvm/arch_timer.c static void timer_set_guest_offset(struct arch_timer_context *ctxt, u64 offset) · line 181
Quoted 7 lines
arch/arm64/kvm/arch_timer.c +4
-0@@ -175,6 +175,12 @@ static void timer_set_guest_offset(struct arch_timer_context *ctxt, u64 offset) } } +static bool ptimer_emulation_required (struct kvm_vcpu *vcpu) +{ +return timer_get_offset (vcpu_ptimer (vcpu)) && + !cpus_have_const_cap (ARM64_ECV );
What Andrew said! :-)
Quoted 117 lines
arch/arm64/kvm/arch_timer.c +26
-21+} +u64 kvm_phys_timer_read (void ) {return timecounter->cc->read (timecounter->cc);@@ -184,8 +190,13 @@ static void get_timer_map(struct kvm_vcpu *vcpu, struct timer_map *map) {if (has_vhe ()) { map->direct_vtimer =vcpu_vtimer (vcpu); - map->direct_ptimer =vcpu_ptimer (vcpu); - map->emul_ptimer =NULL ; +if (!ptimer_emulation_required (vcpu)) { + map->direct_ptimer =vcpu_ptimer (vcpu); + map->emul_ptimer =NULL ; + }else { + map->direct_ptimer =NULL ; + map->emul_ptimer =vcpu_ptimer (vcpu); + } }else { map->direct_vtimer =vcpu_vtimer (vcpu); map->direct_ptimer =NULL ;@@ -671,7 +682,7 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)timer_emulate (map.emul_ptimer);if (has_vhe ()) -kvm_timer_enable_traps_vhe (); +kvm_timer_enable_traps_vhe (vcpu); }bool kvm_timer_should_notify_user (struct kvm_vcpu *vcpu)@@ -1392,22 +1403,29 @@ int kvm_timer_enable(struct kvm_vcpu *vcpu)* The host kernel runs at EL2 with HCR_EL2.TGE == 1, * and this makes those bits have no effect for the host kernel execution. */ -static void kvm_timer_enable_traps_vhe (void ) +static void kvm_timer_enable_traps_vhe (struct kvm_vcpu *vcpu) {/* When HCR_EL2.E2H ==1, EL1PCEN and EL1PCTEN are shifted by 10 */ u32 cnthctl_shift =10 ; -u64 val; +u64 val, mask; + + mask =CNTHCTL_EL1PCEN << cnthctl_shift; + mask |=CNTHCTL_EL1PCTEN << cnthctl_shift; -/* -* VHE systems allow the guest direct access to the EL1 physical -* timer/counter. -*/ val =read_sysreg (cnthctl_el2); - val |= (CNTHCTL_EL1PCEN << cnthctl_shift); - val |= (CNTHCTL_EL1PCTEN << cnthctl_shift);if (cpus_have_const_cap (ARM64_ECV )) val |=CNTHCTL_ECV ; + +/* +* VHE systems allow the guest direct access to the EL1 physical +* timer/counter if offsetting isn't requested on a non-ECV system. +*/ +if (ptimer_emulation_required (vcpu)) + val &= ~mask; +else + val |= mask; +write_sysreg (val, cnthctl_el2); }@@ -1462,9 +1480,6 @@ static int kvm_arm_timer_set_attr_offset(struct kvm_vcpu *vcpu,u64 __user *uaddr = (u64 __user *)(long )attr->addr;u64 offset; -if (!cpus_have_const_cap (ARM64_ECV )) -return -ENXIO ; -if (get_user (offset, uaddr))return -EFAULT ;@@ -1513,9 +1528,6 @@ static int kvm_arm_timer_get_attr_offset(struct kvm_vcpu *vcpu,u64 __user *uaddr = (u64 __user *)(long )attr->addr;u64 offset; -if (!cpus_have_const_cap (ARM64_ECV )) -return -ENXIO ; - offset =timer_get_offset (vcpu_ptimer (vcpu));return put_user (offset, uaddr); }@@ -1539,11 +1551,8 @@ int kvm_arm_timer_has_attr(struct kvm_vcpu *vcpu, struct kvm_device_attr *attr)switch (attr->attr) {case KVM_ARM_VCPU_TIMER_IRQ_VTIMER :case KVM_ARM_VCPU_TIMER_IRQ_PTIMER : -return 0 ;case KVM_ARM_VCPU_TIMER_OFFSET : -if (cpus_have_const_cap (ARM64_ECV )) -return 0 ; -break ; +return 0 ; }return -ENXIO ;arch/arm64/kvm/hyp/include/hyp/switch.h +1
-0@@ -15,6 +15,7 @@#include <linux/jump_label.h> #include <uapi/linux/psci.h> +#include <kvm/arm_arch_timer.h> #include <kvm/arm_psci.h> #include <asm/barrier.h> @@ -405,6 +406,31 @@ static inline bool __hyp_handle_ptrauth(struct kvm_vcpu *vcpu)return true; }
arch/arm64/kvm/hyp/include/hyp/switch.h static inline bool __hyp_handle_ptrauth(struct kvm_vcpu *vcpu) · line 418
Quoted 10 lines
arch/arm64/kvm/hyp/include/hyp/switch.h +10
-0+static inline u64 __timer_read_cntpct (struct kvm_vcpu *vcpu) +{ +return __arch_counter_get_cntpct () -vcpu_ptimer (vcpu)->host_offset; +} + +static inline bool __hyp_handle_counter (struct kvm_vcpu *vcpu) +{ +u32 sysreg; +int rt; +u64 rv;
You could start with a
if (cpus_have_final_cap(ARM64_ECV)) return false;
which will speed-up the exit on ECV-capable systems.
arch/arm64/kvm/hyp/include/hyp/switch.h static inline bool __hyp_handle_ptrauth(struct kvm_vcpu *vcpu) · line 425
Quoted 7 lines
arch/arm64/kvm/hyp/include/hyp/switch.h +7
-0+ +if (kvm_vcpu_trap_get_class (vcpu) != ESR_ELx_EC_SYS64) +return false; + + sysreg =esr_sys64_to_sysreg (kvm_vcpu_get_esr (vcpu)); +if (sysreg !=SYS_CNTPCT_EL0 ) +return false;
You also want to check for CNTPCTSS_EL0 which will also be caught by this trap.
Quoted 35 lines
arch/arm64/kvm/hyp/include/hyp/switch.h +11
-0+ + rt =kvm_vcpu_sys_get_rt (vcpu); + rv =__timer_read_cntpct (vcpu); +vcpu_set_reg (vcpu, rt, rv); +__kvm_skip_instr (vcpu); +return true; +} +/* * Return true when we were able to fixup the guest exit and should return to * the guest, false when we should restore the host state and return to the @@ -439,6 +465,9 @@ static inline bool fixup_guest_exit(struct kvm_vcpu *vcpu, u64 *exit_code)if (*exit_code !=ARM_EXCEPTION_TRAP ) goto exit; +if (__hyp_handle_counter (vcpu)) + goto guest; +if (cpus_have_final_cap (ARM64_WORKAROUND_CAVIUM_TX2_219_TVM ) &&kvm_vcpu_trap_get_class (vcpu) == ESR_ELx_EC_SYS64 &&handle_tx2_tvm (vcpu))arch/arm64/kvm/hyp/nvhe/timer-sr.c +1
-2@@ -44,10 +44,17 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)/* * Disallow physical timer access for the guest -* Physical counter access is allowed */ val =read_sysreg (cnthctl_el2); val &= ~CNTHCTL_EL1PCEN ; - val |=CNTHCTL_EL1PCTEN ; +
arch/arm64/kvm/hyp/nvhe/timer-sr.c void __timer_enable_traps(struct kvm_vcpu *vcpu) · line 60
Quoted 10 lines
arch/arm64/kvm/hyp/nvhe/timer-sr.c +8
-0+/* +* Disallow physical counter access for the guest if offsetting is +* requested on a non-ECV system. +*/ +if (vcpu_ptimer (vcpu)->host_offset && !cpus_have_const_cap (ARM64_ECV )) + val &= ~CNTHCTL_EL1PCTEN ; +else + val |=CNTHCTL_EL1PCTEN ;write_sysreg (val, cnthctl_el2); }
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
PATCH v6 20/21 selftests: KVM: Test physical counter offsetting
In reply to Oliver UptonTest that userspace adjustment of the guest physical counter-timer results in the correct view within the guest.
--- .../selftests/kvm/include/aarch64/processor.h | 12 +++++++ .../kvm/system_counter_offset_test.c | 31 +++++++++++++++++-- 2 files changed, 40 insertions(+), 3 deletions(-)
tools/testing/selftests/kvm/include/aarch64/processor.h +12 -0
@@ -141,4 +141,16 @@ static inline uint64_t read_cntvct_ordered(void) return r; } +static inline uint64_t read_cntpct_ordered (void ) +{ +uint64_t r; + + __asm__ __volatile__("isb\n\t" +"mrs %0, cntpct_el0\n\t" +"isb\n\t" + :"=r" (r)); + +return r; +} +#endif /* SELFTEST_KVM_PROCESSOR_H */
tools/testing/selftests/kvm/system_counter_offset_test.c +28 -3
@@ -57,6 +57,9 @@ static uint64_t host_read_guest_system_counter(struct test_case *test)enum arch_counter {VIRTUAL , +PHYSICAL , +/* offset physical, read virtual */ +PHYSICAL_READ_VIRTUAL , };struct test_case {
@@ -68,32 +71,54 @@ static struct test_case test_cases[] = { { .counter = VIRTUAL, .offset = 0 }, { .counter = VIRTUAL, .offset = 180 * NSEC_PER_SEC }, { .counter = VIRTUAL, .offset = -180 * NSEC_PER_SEC }, + { .counter = PHYSICAL, .offset = 0 }, + { .counter = PHYSICAL, .offset = 180 * NSEC_PER_SEC }, + { .counter = PHYSICAL, .offset = -180 * NSEC_PER_SEC }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = 0 }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = 180 * NSEC_PER_SEC }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = -180 * NSEC_PER_SEC }, };static void check_preconditions (struct kvm_vm *vm) { -if (vcpu_has_reg (vm,VCPU_ID ,KVM_REG_ARM_TIMER_OFFSET )) +if (vcpu_has_reg (vm,VCPU_ID ,KVM_REG_ARM_TIMER_OFFSET ) && + !_vcpu_has_device_attr (vm,VCPU_ID ,KVM_ARM_VCPU_TIMER_CTRL , +KVM_ARM_VCPU_TIMER_OFFSET ))return ; -print_skip ("KVM_REG_ARM_TIMER_OFFSET not supported; skipping test" ); +print_skip ("KVM_REG_ARM_TIMER_OFFSET|KVM_ARM_VCPU_TIMER_OFFSET not supported; skipping test" );exit (KSFT_SKIP ); }static void setup_system_counter (struct kvm_vm *vm,struct test_case *test) { +uint64_t cntvoff, cntpoff;struct kvm_one_reg reg = { .id =KVM_REG_ARM_TIMER_OFFSET , - .addr = (__u64)&test->offset, + .addr = (__u64)&cntvoff, }; +if (test->counter ==VIRTUAL ) { + cntvoff = test->offset; + cntpoff =0 ; + }else { + cntvoff =0 ; + cntpoff = test->offset; + } +vcpu_set_reg (vm,VCPU_ID , ®); +vcpu_access_device_attr (vm,VCPU_ID ,KVM_ARM_VCPU_TIMER_CTRL , +KVM_ARM_VCPU_TIMER_OFFSET , &cntpoff, true); }static uint64_t guest_read_system_counter (struct test_case *test) {switch (test->counter) {case VIRTUAL : +case PHYSICAL_READ_VIRTUAL :return read_cntvct_ordered (); +case PHYSICAL : +return read_cntpct_ordered ();default :GUEST_ASSERT (0 ); }
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Re: [PATCH v6 20/21] selftests: KVM: Test physical counter offsetting
Quoted 106 lines · Oliver Upton
Test that userspace adjustment of the guest physical counter-timer results in the correct view within the guest.
Cc Andrew Jones <[email protected]>Signed-off-by Oliver Upton <[email protected]>--- .../selftests/kvm/include/aarch64/processor.h | 12 +++++++ .../kvm/system_counter_offset_test.c | 31 +++++++++++++++++-- 2 files changed, 40 insertions(+), 3 deletions(-)tools/testing/selftests/kvm/include/aarch64/processor.h +12
-0@@ -141,4 +141,16 @@ static inline uint64_t read_cntvct_ordered(void) return r; } +static inline uint64_t read_cntpct_ordered (void ) +{ +uint64_t r; + + __asm__ __volatile__("isb\n\t" +"mrs %0, cntpct_el0\n\t" +"isb\n\t" + :"=r" (r)); + +return r; +} +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/system_counter_offset_test.c +28
-3@@ -57,6 +57,9 @@ static uint64_t host_read_guest_system_counter(struct test_case *test)enum arch_counter {VIRTUAL , +PHYSICAL , +/* offset physical, read virtual */ +PHYSICAL_READ_VIRTUAL , };struct test_case {@@ -68,32 +71,54 @@ static struct test_case test_cases[] = { { .counter = VIRTUAL, .offset = 0 }, { .counter = VIRTUAL, .offset = 180 * NSEC_PER_SEC }, { .counter = VIRTUAL, .offset = -180 * NSEC_PER_SEC }, + { .counter = PHYSICAL, .offset = 0 }, + { .counter = PHYSICAL, .offset = 180 * NSEC_PER_SEC }, + { .counter = PHYSICAL, .offset = -180 * NSEC_PER_SEC }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = 0 }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = 180 * NSEC_PER_SEC }, + { .counter = PHYSICAL_READ_VIRTUAL, .offset = -180 * NSEC_PER_SEC }, };static void check_preconditions (struct kvm_vm *vm) { -if (vcpu_has_reg (vm,VCPU_ID ,KVM_REG_ARM_TIMER_OFFSET )) +if (vcpu_has_reg (vm,VCPU_ID ,KVM_REG_ARM_TIMER_OFFSET ) && + !_vcpu_has_device_attr (vm,VCPU_ID ,KVM_ARM_VCPU_TIMER_CTRL , +KVM_ARM_VCPU_TIMER_OFFSET ))return ; -print_skip ("KVM_REG_ARM_TIMER_OFFSET not supported; skipping test" ); +print_skip ("KVM_REG_ARM_TIMER_OFFSET|KVM_ARM_VCPU_TIMER_OFFSET not supported; skipping test" );exit (KSFT_SKIP ); }static void setup_system_counter (struct kvm_vm *vm,struct test_case *test) { +uint64_t cntvoff, cntpoff;struct kvm_one_reg reg = { .id =KVM_REG_ARM_TIMER_OFFSET , - .addr = (__u64)&test->offset, + .addr = (__u64)&cntvoff, }; +if (test->counter ==VIRTUAL ) { + cntvoff = test->offset; + cntpoff =0 ; + }else { + cntvoff =0 ; + cntpoff = test->offset; + } +vcpu_set_reg (vm,VCPU_ID , ®); +vcpu_access_device_attr (vm,VCPU_ID ,KVM_ARM_VCPU_TIMER_CTRL , +KVM_ARM_VCPU_TIMER_OFFSET , &cntpoff, true); }static uint64_t guest_read_system_counter (struct test_case *test) {switch (test->counter) {case VIRTUAL : +case PHYSICAL_READ_VIRTUAL :return read_cntvct_ordered (); +case PHYSICAL : +return read_cntpct_ordered ();default :GUEST_ASSERT (0 ); }Signature · 1 line
2.32.0.605.g8dce9f2422-goog
PATCH v6 21/21 selftests: KVM: Add counter emulation benchmark
In reply to Oliver UptonAdd a test case for counter emulation on arm64. A side effect of how KVM handles physical counter offsetting on non-ECV systems is that the virtual counter will always hit hardware and the physical could be emulated. Force emulation by writing a nonzero offset to the physical counter and compare the elapsed cycles to a direct read of the hardware register.
--- tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 1 + .../kvm/aarch64/counter_emulation_benchmark.c | 207 ++++++++++++++++++ 3 files changed, 209 insertions(+) create mode 100644 tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c
tools/testing/selftests/kvm/.gitignore +1 -0
@@ -1,4 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only +/aarch64/counter_emulation_benchmark /aarch64/debug-exceptions /aarch64/get-reg-list /aarch64/vgic_init
tools/testing/selftests/kvm/Makefile +1 -0
@@ -87,6 +87,7 @@ TEST_GEN_PROGS_x86_64 += steal_time TEST_GEN_PROGS_x86_64 += kvm_binary_stats_test TEST_GEN_PROGS_x86_64 += system_counter_offset_test +TEST_GEN_PROGS_aarch64 += aarch64/counter_emulation_benchmark TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list TEST_GEN_PROGS_aarch64 += aarch64/vgic_init
tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c new file +207 -0
207 lines not on this page · raw message
Signature · 1 line
2.32.0.605.g8dce9f2422-goog
Quoted 51 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
Patches 11-12 lay the groundwork for patch 13, which exposes
CNTVOFF_EL2through theONE_REGinterface.Patches 14-15 add test cases for userspace manipulation of the virtual counter-timer.
Patches 16-17 add a vCPU attribute to adjust the host-guest offset of an ARM vCPU, but only implements support for ECV hosts. Patches 18-19 add support for non-ECV hosts by emulating physical counter offsetting.
Quoted 8 lines
Patch 20 adds test cases for adjusting the host-guest offset, and finally patch 21 adds a test to measure the emulation overhead of
CNTPCT_EL2.This series was tested on both an Ampere Mt. Jade and Haswell systems. Unfortunately, the ECV portions of this series are untested, as there is no ECV-capable hardware and the ARM fast models only partially implement ECV.
Small correction: I was only using the foundation model. Apparently the AEM FVP provides full ECV support.
Quoted 150 lines
Physical counter benchmark
--------------------------The following data was collected by running 10000 iterations of the benchmark test from Patch 21 on an Ampere Mt. Jade reference server, A 2S machine with 2 80-core Ampere Altra SoCs. Measurements were collected for both VHE and nVHE operation using the
kvm-arm.mode=command-line parameter.nVHE
----+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 54ns | 148ns | | Standard Deviation | 124ns | 122ns | | 95th Percentile | 258ns | 348ns | +--------------------+--------+---------+VHE
---+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 53ns | 152ns | | Standard Deviation | 92ns | 94ns | | 95th Percentile | 204ns | 307ns | +--------------------+--------+---------+This series applies cleanly to kvm/queue at the following commit:
6cd974485e25("KVM: selftests: Add a test of an unbacked nested PI descriptor")v1 -> v2:
- Reimplemented as vCPU device attributes instead of a distinct ioctl.
- Added the (realtime,host_tsc) instant support to KVM_{GET,SET}_CLOCK
- Changed the arm64 implementation to broadcast counter offset values to all vCPUs in a guest. This upholds the architectural expectations of a consistent counter-timer across CPUs.
- Fixed a bug with traps in VHE mode. We now configure traps on every transition into a guest to handle differing VMs (trapped, emulated).v2 -> v3:
- Added documentation for additions to KVM_{GET,SET}_CLOCK
- Added documentation for all new vCPU attributes
- Added documentation for suggested algorithm to migrate a guest'sTSC(s)- Bug fixes throughout series
- RenameKVM_CLOCK_REAL_TIME->KVM_CLOCK_REALTIMEv3 -> v4:
- Added patch to address incorrect device helper assertions (Drew)
- Carried Drew's r-b tags where appropriate
- x86 selftest cleanup
- Removed stalekvm_timer_init_vhe()function
- Removed unnecessaryGUEST_DONE()from selftestsv4 -> v5:
- Fix typo in TSC migration algorithm
- Carry more of Drew's r-b tags
- clean up run loop logic in counter emulation benchmark (missed from Drew's comments on v3)v5 -> v6:
- Add fix for race inKVM_GET_CLOCK(Sean)
- Fix 32-bit build issues in series + use of uninitialized host tsc value (Sean)
- General style cleanups
- Rework ARM virtual counter offsetting to match guest behavior. Use theONE_REGinterface instead of a VM attribute (Marc)
- Maintain a single host-guest counter offset, which applies to both physical and virtual counters
- Dropped some of Drew's r-b tags due to nontrivial patch changes (sorry for the churn!)v1: https://lore.kernel.org/kvm/[email protected]/
v2: https://lore.kernel.org/r/[email protected]
v3: https://lore.kernel.org/r/[email protected]
v4: https://lore.kernel.org/r/[email protected]
v5: https://lore.kernel.org/r/[email protected]Oliver Upton (21):
KVM: x86: Fix potential race inKVM_GET_CLOCK
KVM: x86: Report host tsc and realtime values inKVM_GET_CLOCK
KVM: x86: Take the pvclock sync lock behind thetsc_write_lock
KVM: x86: Refactor tsc synchronization code
KVM: x86: Expose TSC offset controls to userspace
tools: arch: x86: pull in pvclock headers
selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
selftests: KVM: Fix kvm device helper ioctl assertions
selftests: KVM: Add helpers for vCPU device attributes
selftests: KVM: Introduce system counter offset test
KVM: arm64: Refactorupdate_vtimer_cntvoff()
KVM: arm64: Separate guest/host counter offset values
KVM: arm64: Allow userspace to configure a vCPU's virtual offset
selftests: KVM: Add helper to check for register presence
selftests: KVM: Add support for aarch64 tosystem_counter_offset_test
arm64: cpufeature: Enumerate support for Enhanced Counter
Virtualization
KVM: arm64: Allow userspace to configure a guest's counter-timer
offset
KVM: arm64: Configure timer traps invcpu_load()for VHE
KVM: arm64: Emulate physical counter offsetting on non-ECV systems
selftests: KVM: Test physical counter offsetting
selftests: KVM: Add counter emulation benchmarkDocumentation/virt/kvm/api.rst | 52 ++- Documentation/virt/kvm/devices/vcpu.rst | 85 ++++ Documentation/virt/kvm/locking.rst | 11 + arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 5 + arch/arm64/include/uapi/asm/kvm.h | 2 + arch/arm64/kernel/cpufeature.c | 10 + arch/arm64/kvm/arch_timer.c | 224 ++++++++++- arch/arm64/kvm/arm.c | 4 +- arch/arm64/kvm/guest.c | 6 +- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 + arch/arm64/kvm/hyp/nvhe/timer-sr.c | 16 +- arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + arch/arm64/tools/cpucaps | 1 + arch/x86/include/asm/kvm_host.h | 4 + arch/x86/include/uapi/asm/kvm.h | 4 + arch/x86/kvm/x86.c | 364 +++++++++++++----- include/clocksource/arm_arch_timer.h | 1 + include/kvm/arm_arch_timer.h | 6 +- include/uapi/linux/kvm.h | 7 +- tools/arch/x86/include/asm/pvclock-abi.h | 48 +++ tools/arch/x86/include/asm/pvclock.h | 103 +++++ tools/testing/selftests/kvm/.gitignore | 3 + tools/testing/selftests/kvm/Makefile | 4 + .../kvm/aarch64/counter_emulation_benchmark.c | 207 ++++++++++ .../selftests/kvm/include/aarch64/processor.h | 24 ++ .../testing/selftests/kvm/include/kvm_util.h | 13 + tools/testing/selftests/kvm/lib/kvm_util.c | 63 ++- .../kvm/system_counter_offset_test.c | 211 ++++++++++ .../selftests/kvm/x86_64/kvm_clock_test.c | 204 ++++++++++ 31 files changed, 1581 insertions(+), 143 deletions(-) create mode 100644 tools/arch/x86/include/asm/pvclock-abi.h create mode 100644 tools/arch/x86/include/asm/pvclock.h create mode 100644 tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c create mode 100644 tools/testing/selftests/kvm/system_counter_offset_test.c create mode 100644 tools/testing/selftests/kvm/x86_64/kvm_clock_test.c--
2.32.0.605.g8dce9f2422-goog
Quoted 57 lines · Oliver Upton
Quoted 55 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
Patches 11-12 lay the groundwork for patch 13, which exposes
CNTVOFF_EL2through theONE_REGinterface.Patches 14-15 add test cases for userspace manipulation of the virtual counter-timer.
Patches 16-17 add a vCPU attribute to adjust the host-guest offset of an ARM vCPU, but only implements support for ECV hosts. Patches 18-19 add support for non-ECV hosts by emulating physical counter offsetting.
Patch 20 adds test cases for adjusting the host-guest offset, and finally patch 21 adds a test to measure the emulation overhead of
CNTPCT_EL2.
Quoted 7 lines
Quoted 4 lines
This series was tested on both an Ampere Mt. Jade and Haswell systems. Unfortunately, the ECV portions of this series are untested, as there is no ECV-capable hardware and the ARM fast models only partially implement ECV.
Small correction: I was only using the foundation model. Apparently the AEM FVP provides full ECV support.
Ok. I've now tested this series on the FVP Base RevC fast [email protected] + ECV=2. Passes on VHE, fails on nVHE.
I'll respin this series with the fix for nVHE+ECV soon.
--
Thanks,
Oliver
Quoted 151 lines
Quoted 151 lines
Physical counter benchmark
--------------------------The following data was collected by running 10000 iterations of the benchmark test from Patch 21 on an Ampere Mt. Jade reference server, A 2S machine with 2 80-core Ampere Altra SoCs. Measurements were collected for both VHE and nVHE operation using the
kvm-arm.mode=command-line parameter.nVHE
----+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 54ns | 148ns | | Standard Deviation | 124ns | 122ns | | 95th Percentile | 258ns | 348ns | +--------------------+--------+---------+VHE
---+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 53ns | 152ns | | Standard Deviation | 92ns | 94ns | | 95th Percentile | 204ns | 307ns | +--------------------+--------+---------+This series applies cleanly to kvm/queue at the following commit:
6cd974485e25("KVM: selftests: Add a test of an unbacked nested PI descriptor")v1 -> v2:
- Reimplemented as vCPU device attributes instead of a distinct ioctl.
- Added the (realtime,host_tsc) instant support to KVM_{GET,SET}_CLOCK
- Changed the arm64 implementation to broadcast counter offset values to all vCPUs in a guest. This upholds the architectural expectations of a consistent counter-timer across CPUs.
- Fixed a bug with traps in VHE mode. We now configure traps on every transition into a guest to handle differing VMs (trapped, emulated).v2 -> v3:
- Added documentation for additions to KVM_{GET,SET}_CLOCK
- Added documentation for all new vCPU attributes
- Added documentation for suggested algorithm to migrate a guest'sTSC(s)- Bug fixes throughout series
- RenameKVM_CLOCK_REAL_TIME->KVM_CLOCK_REALTIMEv3 -> v4:
- Added patch to address incorrect device helper assertions (Drew)
- Carried Drew's r-b tags where appropriate
- x86 selftest cleanup
- Removed stalekvm_timer_init_vhe()function
- Removed unnecessaryGUEST_DONE()from selftestsv4 -> v5:
- Fix typo in TSC migration algorithm
- Carry more of Drew's r-b tags
- clean up run loop logic in counter emulation benchmark (missed from Drew's comments on v3)v5 -> v6:
- Add fix for race inKVM_GET_CLOCK(Sean)
- Fix 32-bit build issues in series + use of uninitialized host tsc value (Sean)
- General style cleanups
- Rework ARM virtual counter offsetting to match guest behavior. Use theONE_REGinterface instead of a VM attribute (Marc)
- Maintain a single host-guest counter offset, which applies to both physical and virtual counters
- Dropped some of Drew's r-b tags due to nontrivial patch changes (sorry for the churn!)v1: https://lore.kernel.org/kvm/[email protected]/
v2: https://lore.kernel.org/r/[email protected]
v3: https://lore.kernel.org/r/[email protected]
v4: https://lore.kernel.org/r/[email protected]
v5: https://lore.kernel.org/r/[email protected]Oliver Upton (21):
KVM: x86: Fix potential race inKVM_GET_CLOCK
KVM: x86: Report host tsc and realtime values inKVM_GET_CLOCK
KVM: x86: Take the pvclock sync lock behind thetsc_write_lock
KVM: x86: Refactor tsc synchronization code
KVM: x86: Expose TSC offset controls to userspace
tools: arch: x86: pull in pvclock headers
selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
selftests: KVM: Fix kvm device helper ioctl assertions
selftests: KVM: Add helpers for vCPU device attributes
selftests: KVM: Introduce system counter offset test
KVM: arm64: Refactorupdate_vtimer_cntvoff()
KVM: arm64: Separate guest/host counter offset values
KVM: arm64: Allow userspace to configure a vCPU's virtual offset
selftests: KVM: Add helper to check for register presence
selftests: KVM: Add support for aarch64 tosystem_counter_offset_test
arm64: cpufeature: Enumerate support for Enhanced Counter
Virtualization
KVM: arm64: Allow userspace to configure a guest's counter-timer
offset
KVM: arm64: Configure timer traps invcpu_load()for VHE
KVM: arm64: Emulate physical counter offsetting on non-ECV systems
selftests: KVM: Test physical counter offsetting
selftests: KVM: Add counter emulation benchmarkDocumentation/virt/kvm/api.rst | 52 ++- Documentation/virt/kvm/devices/vcpu.rst | 85 ++++ Documentation/virt/kvm/locking.rst | 11 + arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 5 + arch/arm64/include/uapi/asm/kvm.h | 2 + arch/arm64/kernel/cpufeature.c | 10 + arch/arm64/kvm/arch_timer.c | 224 ++++++++++- arch/arm64/kvm/arm.c | 4 +- arch/arm64/kvm/guest.c | 6 +- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 + arch/arm64/kvm/hyp/nvhe/timer-sr.c | 16 +- arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + arch/arm64/tools/cpucaps | 1 + arch/x86/include/asm/kvm_host.h | 4 + arch/x86/include/uapi/asm/kvm.h | 4 + arch/x86/kvm/x86.c | 364 +++++++++++++----- include/clocksource/arm_arch_timer.h | 1 + include/kvm/arm_arch_timer.h | 6 +- include/uapi/linux/kvm.h | 7 +- tools/arch/x86/include/asm/pvclock-abi.h | 48 +++ tools/arch/x86/include/asm/pvclock.h | 103 +++++ tools/testing/selftests/kvm/.gitignore | 3 + tools/testing/selftests/kvm/Makefile | 4 + .../kvm/aarch64/counter_emulation_benchmark.c | 207 ++++++++++ .../selftests/kvm/include/aarch64/processor.h | 24 ++ .../testing/selftests/kvm/include/kvm_util.h | 13 + tools/testing/selftests/kvm/lib/kvm_util.c | 63 ++- .../kvm/system_counter_offset_test.c | 211 ++++++++++ .../selftests/kvm/x86_64/kvm_clock_test.c | 204 ++++++++++ 31 files changed, 1581 insertions(+), 143 deletions(-) create mode 100644 tools/arch/x86/include/asm/pvclock-abi.h create mode 100644 tools/arch/x86/include/asm/pvclock.h create mode 100644 tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c create mode 100644 tools/testing/selftests/kvm/system_counter_offset_test.c create mode 100644 tools/testing/selftests/kvm/x86_64/kvm_clock_test.c--
2.32.0.605.g8dce9f2422-goog
Quoted 37 lines · Oliver Upton
Quoted 35 lines · Oliver Upton
Quoted 33 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
Paolo,
Is there anything else you're waiting to see for the x86 portion of this series after addressing Sean's comments? There's some work remaining on the arm64 side, though I believe the two architectures are now disjoint for this series.
--
Thanks,
Oliver
Quoted 182 lines
Quoted 14 lines
Quoted 14 lines
Patches 11-12 lay the groundwork for patch 13, which exposes
CNTVOFF_EL2through theONE_REGinterface.Patches 14-15 add test cases for userspace manipulation of the virtual counter-timer.
Patches 16-17 add a vCPU attribute to adjust the host-guest offset of an ARM vCPU, but only implements support for ECV hosts. Patches 18-19 add support for non-ECV hosts by emulating physical counter offsetting.
Patch 20 adds test cases for adjusting the host-guest offset, and finally patch 21 adds a test to measure the emulation overhead of
CNTPCT_EL2.Quoted 7 lines
Quoted 4 lines
This series was tested on both an Ampere Mt. Jade and Haswell systems. Unfortunately, the ECV portions of this series are untested, as there is no ECV-capable hardware and the ARM fast models only partially implement ECV.
Small correction: I was only using the foundation model. Apparently the AEM FVP provides full ECV support.
Ok. I've now tested this series on the FVP Base RevC fast [email protected] + ECV=2. Passes on VHE, fails on nVHE.
I'll respin this series with the fix for nVHE+ECV soon.
--
Thanks,
OliverQuoted 151 lines
Quoted 151 lines
Physical counter benchmark
--------------------------The following data was collected by running 10000 iterations of the benchmark test from Patch 21 on an Ampere Mt. Jade reference server, A 2S machine with 2 80-core Ampere Altra SoCs. Measurements were collected for both VHE and nVHE operation using the
kvm-arm.mode=command-line parameter.nVHE
----+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 54ns | 148ns | | Standard Deviation | 124ns | 122ns | | 95th Percentile | 258ns | 348ns | +--------------------+--------+---------+VHE
---+--------------------+--------+---------+ | Metric | Native | Trapped | +--------------------+--------+---------+ | Average | 53ns | 152ns | | Standard Deviation | 92ns | 94ns | | 95th Percentile | 204ns | 307ns | +--------------------+--------+---------+This series applies cleanly to kvm/queue at the following commit:
6cd974485e25("KVM: selftests: Add a test of an unbacked nested PI descriptor")v1 -> v2:
- Reimplemented as vCPU device attributes instead of a distinct ioctl.
- Added the (realtime,host_tsc) instant support to KVM_{GET,SET}_CLOCK
- Changed the arm64 implementation to broadcast counter offset values to all vCPUs in a guest. This upholds the architectural expectations of a consistent counter-timer across CPUs.
- Fixed a bug with traps in VHE mode. We now configure traps on every transition into a guest to handle differing VMs (trapped, emulated).v2 -> v3:
- Added documentation for additions to KVM_{GET,SET}_CLOCK
- Added documentation for all new vCPU attributes
- Added documentation for suggested algorithm to migrate a guest'sTSC(s)- Bug fixes throughout series
- RenameKVM_CLOCK_REAL_TIME->KVM_CLOCK_REALTIMEv3 -> v4:
- Added patch to address incorrect device helper assertions (Drew)
- Carried Drew's r-b tags where appropriate
- x86 selftest cleanup
- Removed stalekvm_timer_init_vhe()function
- Removed unnecessaryGUEST_DONE()from selftestsv4 -> v5:
- Fix typo in TSC migration algorithm
- Carry more of Drew's r-b tags
- clean up run loop logic in counter emulation benchmark (missed from Drew's comments on v3)v5 -> v6:
- Add fix for race inKVM_GET_CLOCK(Sean)
- Fix 32-bit build issues in series + use of uninitialized host tsc value (Sean)
- General style cleanups
- Rework ARM virtual counter offsetting to match guest behavior. Use theONE_REGinterface instead of a VM attribute (Marc)
- Maintain a single host-guest counter offset, which applies to both physical and virtual counters
- Dropped some of Drew's r-b tags due to nontrivial patch changes (sorry for the churn!)v1: https://lore.kernel.org/kvm/[email protected]/
v2: https://lore.kernel.org/r/[email protected]
v3: https://lore.kernel.org/r/[email protected]
v4: https://lore.kernel.org/r/[email protected]
v5: https://lore.kernel.org/r/[email protected]Oliver Upton (21):
KVM: x86: Fix potential race inKVM_GET_CLOCK
KVM: x86: Report host tsc and realtime values inKVM_GET_CLOCK
KVM: x86: Take the pvclock sync lock behind thetsc_write_lock
KVM: x86: Refactor tsc synchronization code
KVM: x86: Expose TSC offset controls to userspace
tools: arch: x86: pull in pvclock headers
selftests: KVM: Add test for KVM_{GET,SET}_CLOCK
selftests: KVM: Fix kvm device helper ioctl assertions
selftests: KVM: Add helpers for vCPU device attributes
selftests: KVM: Introduce system counter offset test
KVM: arm64: Refactorupdate_vtimer_cntvoff()
KVM: arm64: Separate guest/host counter offset values
KVM: arm64: Allow userspace to configure a vCPU's virtual offset
selftests: KVM: Add helper to check for register presence
selftests: KVM: Add support for aarch64 tosystem_counter_offset_test
arm64: cpufeature: Enumerate support for Enhanced Counter
Virtualization
KVM: arm64: Allow userspace to configure a guest's counter-timer
offset
KVM: arm64: Configure timer traps invcpu_load()for VHE
KVM: arm64: Emulate physical counter offsetting on non-ECV systems
selftests: KVM: Test physical counter offsetting
selftests: KVM: Add counter emulation benchmarkDocumentation/virt/kvm/api.rst | 52 ++- Documentation/virt/kvm/devices/vcpu.rst | 85 ++++ Documentation/virt/kvm/locking.rst | 11 + arch/arm64/include/asm/kvm_asm.h | 2 + arch/arm64/include/asm/sysreg.h | 5 + arch/arm64/include/uapi/asm/kvm.h | 2 + arch/arm64/kernel/cpufeature.c | 10 + arch/arm64/kvm/arch_timer.c | 224 ++++++++++- arch/arm64/kvm/arm.c | 4 +- arch/arm64/kvm/guest.c | 6 +- arch/arm64/kvm/hyp/include/hyp/switch.h | 29 ++ arch/arm64/kvm/hyp/nvhe/hyp-main.c | 6 + arch/arm64/kvm/hyp/nvhe/timer-sr.c | 16 +- arch/arm64/kvm/hyp/vhe/timer-sr.c | 5 + arch/arm64/tools/cpucaps | 1 + arch/x86/include/asm/kvm_host.h | 4 + arch/x86/include/uapi/asm/kvm.h | 4 + arch/x86/kvm/x86.c | 364 +++++++++++++----- include/clocksource/arm_arch_timer.h | 1 + include/kvm/arm_arch_timer.h | 6 +- include/uapi/linux/kvm.h | 7 +- tools/arch/x86/include/asm/pvclock-abi.h | 48 +++ tools/arch/x86/include/asm/pvclock.h | 103 +++++ tools/testing/selftests/kvm/.gitignore | 3 + tools/testing/selftests/kvm/Makefile | 4 + .../kvm/aarch64/counter_emulation_benchmark.c | 207 ++++++++++ .../selftests/kvm/include/aarch64/processor.h | 24 ++ .../testing/selftests/kvm/include/kvm_util.h | 13 + tools/testing/selftests/kvm/lib/kvm_util.c | 63 ++- .../kvm/system_counter_offset_test.c | 211 ++++++++++ .../selftests/kvm/x86_64/kvm_clock_test.c | 204 ++++++++++ 31 files changed, 1581 insertions(+), 143 deletions(-) create mode 100644 tools/arch/x86/include/asm/pvclock-abi.h create mode 100644 tools/arch/x86/include/asm/pvclock.h create mode 100644 tools/testing/selftests/kvm/aarch64/counter_emulation_benchmark.c create mode 100644 tools/testing/selftests/kvm/system_counter_offset_test.c create mode 100644 tools/testing/selftests/kvm/x86_64/kvm_clock_test.c--
2.32.0.605.g8dce9f2422-goog
Marc Zyngier
maintainerQuoted 47 lines · Oliver Upton
Quoted 37 lines · Oliver Upton
Quoted 35 lines · Oliver Upton
Quoted 33 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
Quoted 6 lines
Paolo,
Is there anything else you're waiting to see for the x86 portion of this series after addressing Sean's comments? There's some work remaining on the arm64 side, though I believe the two architectures are now disjoint for this series.
I think at this stage it would make sense to split the series in two and drive them independently.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Paolo Bonzini
maintainerQuoted 32 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Quoted 7 lines
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
The x86 parts look good, except that patch 3 is a bit redundant with my idea of altogether getting rid of the pvclock_gtod_sync_lock. That said I agree that patches 1 and 2 (and extracting kvm_vm_ioctl_get_clock and kvm_vm_ioctl_set_clock) should be done before whatever locking changes have to be done.
Time is ticking for 5.15 due to my vacation, I'll see if I have some time to look at it further next week.
I agree that arm64 can be done separately from x86.
Paolo
Quoted 39 lines · Paolo Bonzini
Quoted 37 lines · Oliver Upton
KVM's current means of saving/restoring system counters is plagued with temporal issues. At least on ARM64 and x86, we migrate the guest's system counter by-value through the respective guest system register values (
cntvct_el0,ia32_tsc). Restoring system counters by-value is brittle as the state is not idempotent: the host system counter is still oscillating between the attempted save and restore. Furthermore, VMMs may wish to transparently live migrate guest VMs, meaning that they include the elapsed time due to live migration blackout in the guest system counter view. The VMM thread could be preempted for any number of reasons (scheduler, L0 hypervisor under nested) between the time that it calculates the desired guest counter value and when KVM actually sets this counter state.Despite the value-based interface that we present to userspace, KVM actually has idempotent guest controls by way of system counter offsets. We can avoid all of the issues associated with a value-based interface by abstracting these offset controls in new ioctls. This series introduces new vCPU device attributes to provide userspace access to the vCPU's system counter offset.
Patch 1 addresses a possible race in
KVM_GET_CLOCKwhereuse_master_clockis read outside of thepvclock_gtod_sync_lock.Patch 2 adopts Paolo's suggestion, augmenting the KVM_{GET,SET}_CLOCK ioctls to provide userspace with a (
host_tsc, realtime) instant. This is essential for a VMM to perform precise migration of the guest's system counters.Patches 3-4 are some preparatory changes for exposing the TSC offset to userspace. Patch 5 provides a vCPU attribute to provide userspace access to the TSC offset.
Patches 6-7 implement a test for the new additions to KVM_{GET,SET}_CLOCK.
Patch 8 fixes some assertions in the kvm device attribute helpers.
Quoted 8 lines
Quoted 2 lines
Patches 9-10 implement at test for the tsc offset attribute introduced in patch 5.
The x86 parts look good, except that patch 3 is a bit redundant with my idea of altogether getting rid of the
pvclock_gtod_sync_lock. That said I agree that patches 1 and 2 (and extractingkvm_vm_ioctl_get_clockandkvm_vm_ioctl_set_clock) should be done before whatever locking changes have to be done.
Following up on patch 3.
Quoted 4 lines
Time is ticking for 5.15 due to my vacation, I'll see if I have some time to look at it further next week.
I agree that arm64 can be done separately from x86.
Marc, just a disclaimer:
I'm going to separate these two series, although there will still exist dependencies in the selftests changes. Otherwise, kernel changes are disjoint.
--
Thanks,
Oliver
Marc Zyngier
maintainerOn Wed, 11 Aug 2021 19:56:22 +0100,
Oliver Upton <[email protected]> wrote:
[...]
Quoted 10 lines
Quoted 4 lines
Time is ticking for 5.15 due to my vacation, I'll see if I have some time to look at it further next week.
I agree that arm64 can be done separately from x86.
Marc, just a disclaimer:
I'm going to separate these two series, although there will still exist dependencies in the selftests changes. Otherwise, kernel changes are disjoint.
No problem. The selftests can even sit in a third series if that makes it easier.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.