leandrof/linux
1
0
Fork 0
forked from Minki/linux
Code Issues Pull Requests Packages Projects Releases Wiki Activity
5dcd0fdbb4
linux/arch/arm64/kvm/fpsimd.c

140 lines
4.0 KiB
C
Raw Normal View History

KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
// SPDX-License-Identifier: GPL-2.0
/*
* arch/arm64/kvm/fpsimd.c: Guest/host FPSIMD context coordination helpers
*
* Copyright 2018 Arm Limited
* Author: Dave Martin <Dave.Martin@arm.com>
*/
KVM: arm64: Don't mask softirq with IRQs disabled in vcpu_put() Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") introduces a specific helper kvm_arch_vcpu_put_fp() for saving the vcpu FPSIMD state during vcpu_put(). This function uses local_bh_disable()/_enable() to protect the FPSIMD context manipulation from interruption by softirqs. This approach is not correct, because vcpu_put() can be invoked either from the KVM host vcpu thread (when exiting the vcpu run loop), or via a preempt notifier. In the former case, only preemption is disabled. In the latter case, the function is called from inside __schedule(), which means that IRQs are disabled. Use of local_bh_disable()/_enable() with IRQs disabled is considerd an error, resulting in lockdep splats while running VMs if lockdep is enabled. This patch disables IRQs instead of attempting to disable softirqs, avoiding the problem of calling local_bh_enable() with IRQs disabled in the __schedule() path. This creates an additional interrupt blackout during vcpu run loop exit, but this is the rare case and the blackout latency is still less than that of __schedule(). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Reported-by: Andre Przywara <andre.przywara@arm.com> Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:24 +00:00
#include <linux/irqflags.h>
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
#include <linux/sched.h>
#include <linux/thread_info.h>
#include <linux/kvm_host.h>
arm64/sve: Enable SVE state tracking for non-task contexts The current FPSIMD/SVE context handling support for non-task (i.e., KVM vcpu) contexts does not take SVE into account. This means that only task contexts can safely use SVE at present. In preparation for enabling KVM guests to use SVE, it is necessary to keep track of SVE state for non-task contexts too. This patch adds the necessary support, removing assumptions from the context switch code about the location of the SVE context storage. When binding a vcpu context, its vector length is arbitrarily specified as SVE_VL_MIN for now. In any case, because TIF_SVE is presently cleared at vcpu context bind time, the specified vector length will not be used for anything yet. In later patches TIF_SVE will be set here as appropriate, and the appropriate maximum vector length for the vcpu will be passed when binding. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Julien Grall <julien.grall@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:11 +00:00
#include <asm/fpsimd.h>
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
#include <asm/kvm_asm.h>
#include <asm/kvm_host.h>
#include <asm/kvm_mmu.h>
KVM: arm64/sve: Fix SVE trap restoration for non-current tasks Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") attempts to restore the configuration of userspace SVE trapping via a call to fpsimd_bind_task_to_cpu(), but the logic for determining when to do this is not correct. The patch makes the errnoenous assumption that the only task that may try to enter userspace with the currently loaded FPSIMD/SVE register content is current. This may not be the case however: if some other user task T is scheduled on the CPU during the execution of the KVM run loop, and the vcpu does not try to use the registers in the meantime, then T's state may be left there intact. If T happens to be the next task to enter userspace on this CPU then the hooks for reloading the register state and configuring traps will be skipped. (Also, current never has SVE state at this point anyway and should always have the trap enabled, as a side-effect of the ioctl() syscall needed to reach the KVM run loop in the first place.) This patch instead restores the state of the EL0 trap from the state observed at the most recent vcpu_load(), ensuring that the trap is set correctly for the loaded context (if any). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:25 +00:00
#include <asm/sysreg.h>
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
/*
* Called on entry to KVM_RUN unless this vcpu previously ran at least
* once and the most recent prior KVM_RUN for this vcpu was called from
* the same task as current (highly likely).
*
* This is guaranteed to execute before kvm_arch_vcpu_load_fp(vcpu),
* such that on entering hyp the relevant parts of current are already
* mapped.
*/
int kvm_arch_vcpu_run_map_fp(struct kvm_vcpu *vcpu)
{
int ret;
struct thread_info *ti = &current->thread_info;
struct user_fpsimd_state *fpsimd = &current->thread.uw.fpsimd_state;
/*
* Make sure the host task thread flags and fpsimd state are
* visible to hyp:
*/
ret = create_hyp_mappings(ti, ti + 1, PAGE_HYP);
if (ret)
goto error;
ret = create_hyp_mappings(fpsimd, fpsimd + 1, PAGE_HYP);
if (ret)
goto error;
vcpu->arch.host_thread_info = kern_hyp_va(ti);
vcpu->arch.host_fpsimd_state = kern_hyp_va(fpsimd);
error:
return ret;
}
/*
* Prepare vcpu for saving the host's FPSIMD state and loading the guest's.
* The actual loading is done by the FPSIMD access trap taken to hyp.
*
* Here, we just set the correct metadata to indicate that the FPSIMD
* state in the cpu regs (if any) belongs to current on the host.
*
* TIF_SVE is backed up here, since it may get clobbered with guest state.
* This flag is restored by kvm_arch_vcpu_put_fp(vcpu).
*/
void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu)
{
BUG_ON(!current->mm);
KVM: arm64/sve: Fix SVE trap restoration for non-current tasks Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") attempts to restore the configuration of userspace SVE trapping via a call to fpsimd_bind_task_to_cpu(), but the logic for determining when to do this is not correct. The patch makes the errnoenous assumption that the only task that may try to enter userspace with the currently loaded FPSIMD/SVE register content is current. This may not be the case however: if some other user task T is scheduled on the CPU during the execution of the KVM run loop, and the vcpu does not try to use the registers in the meantime, then T's state may be left there intact. If T happens to be the next task to enter userspace on this CPU then the hooks for reloading the register state and configuring traps will be skipped. (Also, current never has SVE state at this point anyway and should always have the trap enabled, as a side-effect of the ioctl() syscall needed to reach the KVM run loop in the first place.) This patch instead restores the state of the EL0 trap from the state observed at the most recent vcpu_load(), ensuring that the trap is set correctly for the loaded context (if any). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:25 +00:00
vcpu->arch.flags &= ~(KVM_ARM64_FP_ENABLED |
KVM_ARM64_HOST_SVE_IN_USE |
KVM_ARM64_HOST_SVE_ENABLED);
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
vcpu->arch.flags |= KVM_ARM64_FP_HOST;
KVM: arm64/sve: Fix SVE trap restoration for non-current tasks Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") attempts to restore the configuration of userspace SVE trapping via a call to fpsimd_bind_task_to_cpu(), but the logic for determining when to do this is not correct. The patch makes the errnoenous assumption that the only task that may try to enter userspace with the currently loaded FPSIMD/SVE register content is current. This may not be the case however: if some other user task T is scheduled on the CPU during the execution of the KVM run loop, and the vcpu does not try to use the registers in the meantime, then T's state may be left there intact. If T happens to be the next task to enter userspace on this CPU then the hooks for reloading the register state and configuring traps will be skipped. (Also, current never has SVE state at this point anyway and should always have the trap enabled, as a side-effect of the ioctl() syscall needed to reach the KVM run loop in the first place.) This patch instead restores the state of the EL0 trap from the state observed at the most recent vcpu_load(), ensuring that the trap is set correctly for the loaded context (if any). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:25 +00:00
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
if (test_thread_flag(TIF_SVE))
vcpu->arch.flags |= KVM_ARM64_HOST_SVE_IN_USE;
KVM: arm64/sve: Fix SVE trap restoration for non-current tasks Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") attempts to restore the configuration of userspace SVE trapping via a call to fpsimd_bind_task_to_cpu(), but the logic for determining when to do this is not correct. The patch makes the errnoenous assumption that the only task that may try to enter userspace with the currently loaded FPSIMD/SVE register content is current. This may not be the case however: if some other user task T is scheduled on the CPU during the execution of the KVM run loop, and the vcpu does not try to use the registers in the meantime, then T's state may be left there intact. If T happens to be the next task to enter userspace on this CPU then the hooks for reloading the register state and configuring traps will be skipped. (Also, current never has SVE state at this point anyway and should always have the trap enabled, as a side-effect of the ioctl() syscall needed to reach the KVM run loop in the first place.) This patch instead restores the state of the EL0 trap from the state observed at the most recent vcpu_load(), ensuring that the trap is set correctly for the loaded context (if any). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:25 +00:00
if (read_sysreg(cpacr_el1) & CPACR_EL1_ZEN_EL0EN)
vcpu->arch.flags |= KVM_ARM64_HOST_SVE_ENABLED;
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
}
/*
* If the guest FPSIMD state was loaded, update the host's context
* tracking data mark the CPU FPSIMD regs as dirty and belonging to vcpu
* so that they will be written back if the kernel clobbers them due to
* kernel-mode NEON before re-entry into the guest.
*/
void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu)
{
WARN_ON_ONCE(!irqs_disabled());
if (vcpu->arch.flags & KVM_ARM64_FP_ENABLED) {
arm64/sve: Enable SVE state tracking for non-task contexts The current FPSIMD/SVE context handling support for non-task (i.e., KVM vcpu) contexts does not take SVE into account. This means that only task contexts can safely use SVE at present. In preparation for enabling KVM guests to use SVE, it is necessary to keep track of SVE state for non-task contexts too. This patch adds the necessary support, removing assumptions from the context switch code about the location of the SVE context storage. When binding a vcpu context, its vector length is arbitrarily specified as SVE_VL_MIN for now. In any case, because TIF_SVE is presently cleared at vcpu context bind time, the specified vector length will not be used for anything yet. In later patches TIF_SVE will be set here as appropriate, and the appropriate maximum vector length for the vcpu will be passed when binding. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Julien Grall <julien.grall@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:11 +00:00
fpsimd_bind_state_to_cpu(&vcpu->arch.ctxt.gp_regs.fp_regs,
KVM: arm64/sve: Context switch the SVE registers In order to give each vcpu its own view of the SVE registers, this patch adds context storage via a new sve_state pointer in struct vcpu_arch. An additional member sve_max_vl is also added for each vcpu, to determine the maximum vector length visible to the guest and thus the value to be configured in ZCR_EL2.LEN while the vcpu is active. This also determines the layout and size of the storage in sve_state, which is read and written by the same backend functions that are used for context-switching the SVE state for host tasks. On SVE-enabled vcpus, SVE access traps are now handled by switching in the vcpu's SVE context and disabling the trap before returning to the guest. On other vcpus, the trap is not handled and an exit back to the host occurs, where the handle_sve() fallback path reflects an undefined instruction exception back to the guest, consistently with the behaviour of non-SVE-capable hardware (as was done unconditionally prior to this patch). No SVE handling is added on non-VHE-only paths, since VHE is an architectural and Kconfig prerequisite of SVE. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Julien Thierry <julien.thierry@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:17 +00:00
vcpu->arch.sve_state,
vcpu->arch.sve_max_vl);
arm64/sve: Enable SVE state tracking for non-task contexts The current FPSIMD/SVE context handling support for non-task (i.e., KVM vcpu) contexts does not take SVE into account. This means that only task contexts can safely use SVE at present. In preparation for enabling KVM guests to use SVE, it is necessary to keep track of SVE state for non-task contexts too. This patch adds the necessary support, removing assumptions from the context switch code about the location of the SVE context storage. When binding a vcpu context, its vector length is arbitrarily specified as SVE_VL_MIN for now. In any case, because TIF_SVE is presently cleared at vcpu context bind time, the specified vector length will not be used for anything yet. In later patches TIF_SVE will be set here as appropriate, and the appropriate maximum vector length for the vcpu will be passed when binding. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Julien Grall <julien.grall@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:11 +00:00
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
clear_thread_flag(TIF_FOREIGN_FPSTATE);
KVM: arm64/sve: Context switch the SVE registers In order to give each vcpu its own view of the SVE registers, this patch adds context storage via a new sve_state pointer in struct vcpu_arch. An additional member sve_max_vl is also added for each vcpu, to determine the maximum vector length visible to the guest and thus the value to be configured in ZCR_EL2.LEN while the vcpu is active. This also determines the layout and size of the storage in sve_state, which is read and written by the same backend functions that are used for context-switching the SVE state for host tasks. On SVE-enabled vcpus, SVE access traps are now handled by switching in the vcpu's SVE context and disabling the trap before returning to the guest. On other vcpus, the trap is not handled and an exit back to the host occurs, where the handle_sve() fallback path reflects an undefined instruction exception back to the guest, consistently with the behaviour of non-SVE-capable hardware (as was done unconditionally prior to this patch). No SVE handling is added on non-VHE-only paths, since VHE is an architectural and Kconfig prerequisite of SVE. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Julien Thierry <julien.thierry@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:17 +00:00
update_thread_flag(TIF_SVE, vcpu_has_sve(vcpu));
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
}
}
/*
* Write back the vcpu FPSIMD regs if they are dirty, and invalidate the
* cpu FPSIMD regs so that they can't be spuriously reused if this vcpu
* disappears and another task or vcpu appears that recycles the same
* struct fpsimd_state.
*/
void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu)
{
KVM: arm64: Don't mask softirq with IRQs disabled in vcpu_put() Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") introduces a specific helper kvm_arch_vcpu_put_fp() for saving the vcpu FPSIMD state during vcpu_put(). This function uses local_bh_disable()/_enable() to protect the FPSIMD context manipulation from interruption by softirqs. This approach is not correct, because vcpu_put() can be invoked either from the KVM host vcpu thread (when exiting the vcpu run loop), or via a preempt notifier. In the former case, only preemption is disabled. In the latter case, the function is called from inside __schedule(), which means that IRQs are disabled. Use of local_bh_disable()/_enable() with IRQs disabled is considerd an error, resulting in lockdep splats while running VMs if lockdep is enabled. This patch disables IRQs instead of attempting to disable softirqs, avoiding the problem of calling local_bh_enable() with IRQs disabled in the __schedule() path. This creates an additional interrupt blackout during vcpu run loop exit, but this is the rare case and the blackout latency is still less than that of __schedule(). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Reported-by: Andre Przywara <andre.przywara@arm.com> Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:24 +00:00
unsigned long flags;
KVM: arm64/sve: System register context switch and access support This patch adds the necessary support for context switching ZCR_EL1 for each vcpu. ZCR_EL1 is trapped alongside the FPSIMD/SVE registers, so it makes sense for it to be handled as part of the guest FPSIMD/SVE context for context switch purposes instead of handling it as a general system register. This means that it can be switched in lazily at the appropriate time. No effort is made to track host context for this register, since SVE requires VHE: thus the hosts's value for this register lives permanently in ZCR_EL2 and does not alias the guest's value at any time. The Hyp switch and fpsimd context handling code is extended appropriately. Accessors are added in sys_regs.c to expose the SVE system registers and ID register fields. Because these need to be conditionally visible based on the guest configuration, they are implemented separately for now rather than by use of the generic system register helpers. This may be abstracted better later on when/if there are more features requiring this model. ID_AA64ZFR0_EL1 is RO-RAZ for MRS/MSR when SVE is disabled for the guest, but for compatibility with non-SVE aware KVM implementations the register should not be enumerated at all for KVM_GET_REG_LIST in this case. For consistency we also reject ioctl access to the register. This ensures that a non-SVE-enabled guest looks the same to userspace, irrespective of whether the kernel KVM implementation supports SVE. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Julien Thierry <julien.thierry@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:16 +00:00
bool host_has_sve = system_supports_sve();
bool guest_has_sve = vcpu_has_sve(vcpu);
KVM: arm64: Don't mask softirq with IRQs disabled in vcpu_put() Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") introduces a specific helper kvm_arch_vcpu_put_fp() for saving the vcpu FPSIMD state during vcpu_put(). This function uses local_bh_disable()/_enable() to protect the FPSIMD context manipulation from interruption by softirqs. This approach is not correct, because vcpu_put() can be invoked either from the KVM host vcpu thread (when exiting the vcpu run loop), or via a preempt notifier. In the former case, only preemption is disabled. In the latter case, the function is called from inside __schedule(), which means that IRQs are disabled. Use of local_bh_disable()/_enable() with IRQs disabled is considerd an error, resulting in lockdep splats while running VMs if lockdep is enabled. This patch disables IRQs instead of attempting to disable softirqs, avoiding the problem of calling local_bh_enable() with IRQs disabled in the __schedule() path. This creates an additional interrupt blackout during vcpu run loop exit, but this is the rare case and the blackout latency is still less than that of __schedule(). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Reported-by: Andre Przywara <andre.przywara@arm.com> Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:24 +00:00
local_irq_save(flags);
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
if (vcpu->arch.flags & KVM_ARM64_FP_ENABLED) {
KVM: arm64/sve: System register context switch and access support This patch adds the necessary support for context switching ZCR_EL1 for each vcpu. ZCR_EL1 is trapped alongside the FPSIMD/SVE registers, so it makes sense for it to be handled as part of the guest FPSIMD/SVE context for context switch purposes instead of handling it as a general system register. This means that it can be switched in lazily at the appropriate time. No effort is made to track host context for this register, since SVE requires VHE: thus the hosts's value for this register lives permanently in ZCR_EL2 and does not alias the guest's value at any time. The Hyp switch and fpsimd context handling code is extended appropriately. Accessors are added in sys_regs.c to expose the SVE system registers and ID register fields. Because these need to be conditionally visible based on the guest configuration, they are implemented separately for now rather than by use of the generic system register helpers. This may be abstracted better later on when/if there are more features requiring this model. ID_AA64ZFR0_EL1 is RO-RAZ for MRS/MSR when SVE is disabled for the guest, but for compatibility with non-SVE aware KVM implementations the register should not be enumerated at all for KVM_GET_REG_LIST in this case. For consistency we also reject ioctl access to the register. This ensures that a non-SVE-enabled guest looks the same to userspace, irrespective of whether the kernel KVM implementation supports SVE. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Julien Thierry <julien.thierry@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:16 +00:00
u64 *guest_zcr = &vcpu->arch.ctxt.sys_regs[ZCR_EL1];
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
/* Clean guest FP state to memory and invalidate cpu view */
fpsimd_save();
fpsimd_flush_cpu_state();
KVM: arm64/sve: System register context switch and access support This patch adds the necessary support for context switching ZCR_EL1 for each vcpu. ZCR_EL1 is trapped alongside the FPSIMD/SVE registers, so it makes sense for it to be handled as part of the guest FPSIMD/SVE context for context switch purposes instead of handling it as a general system register. This means that it can be switched in lazily at the appropriate time. No effort is made to track host context for this register, since SVE requires VHE: thus the hosts's value for this register lives permanently in ZCR_EL2 and does not alias the guest's value at any time. The Hyp switch and fpsimd context handling code is extended appropriately. Accessors are added in sys_regs.c to expose the SVE system registers and ID register fields. Because these need to be conditionally visible based on the guest configuration, they are implemented separately for now rather than by use of the generic system register helpers. This may be abstracted better later on when/if there are more features requiring this model. ID_AA64ZFR0_EL1 is RO-RAZ for MRS/MSR when SVE is disabled for the guest, but for compatibility with non-SVE aware KVM implementations the register should not be enumerated at all for KVM_GET_REG_LIST in this case. For consistency we also reject ioctl access to the register. This ensures that a non-SVE-enabled guest looks the same to userspace, irrespective of whether the kernel KVM implementation supports SVE. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Julien Thierry <julien.thierry@arm.com> Tested-by: zhang.lei <zhang.lei@jp.fujitsu.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-09-28 13:39:16 +00:00
if (guest_has_sve)
*guest_zcr = read_sysreg_s(SYS_ZCR_EL12);
} else if (host_has_sve) {
KVM: arm64/sve: Fix SVE trap restoration for non-current tasks Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") attempts to restore the configuration of userspace SVE trapping via a call to fpsimd_bind_task_to_cpu(), but the logic for determining when to do this is not correct. The patch makes the errnoenous assumption that the only task that may try to enter userspace with the currently loaded FPSIMD/SVE register content is current. This may not be the case however: if some other user task T is scheduled on the CPU during the execution of the KVM run loop, and the vcpu does not try to use the registers in the meantime, then T's state may be left there intact. If T happens to be the next task to enter userspace on this CPU then the hooks for reloading the register state and configuring traps will be skipped. (Also, current never has SVE state at this point anyway and should always have the trap enabled, as a side-effect of the ioctl() syscall needed to reach the KVM run loop in the first place.) This patch instead restores the state of the EL0 trap from the state observed at the most recent vcpu_load(), ensuring that the trap is set correctly for the loaded context (if any). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:25 +00:00
/*
* The FPSIMD/SVE state in the CPU has not been touched, and we
* have SVE (and VHE): CPACR_EL1 (alias CPTR_EL2) has been
* reset to CPACR_EL1_DEFAULT by the Hyp code, disabling SVE
* for EL0. To avoid spurious traps, restore the trap state
* seen by kvm_arch_vcpu_load_fp():
*/
if (vcpu->arch.flags & KVM_ARM64_HOST_SVE_ENABLED)
sysreg_clear_set(CPACR_EL1, 0, CPACR_EL1_ZEN_EL0EN);
else
sysreg_clear_set(CPACR_EL1, CPACR_EL1_ZEN_EL0EN, 0);
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
}
KVM: arm64: Avoid mistaken attempts to save SVE state for vcpus Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") uses fpsimd_save() to save the FPSIMD state for a vcpu when scheduling the vcpu out. However, currently current's value of TIF_SVE is restored before calling fpsimd_save() which means that fpsimd_save() may erroneously attempt to save SVE state from the vcpu. This enables current's vector state to be polluted with guest data. current->thread.sve_state may be unallocated or not large enough, so this can also trigger a NULL dereference or buffer overrun. Instead of this, TIF_SVE should be configured properly for the guest when calling fpsimd_save() with the vcpu context loaded. This patch ensures this by delaying restoration of current's TIF_SVE until after the call to fpsimd_save(). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:26 +00:00
update_thread_flag(TIF_SVE,
vcpu->arch.flags & KVM_ARM64_HOST_SVE_IN_USE);
KVM: arm64: Don't mask softirq with IRQs disabled in vcpu_put() Commit e6b673b ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") introduces a specific helper kvm_arch_vcpu_put_fp() for saving the vcpu FPSIMD state during vcpu_put(). This function uses local_bh_disable()/_enable() to protect the FPSIMD context manipulation from interruption by softirqs. This approach is not correct, because vcpu_put() can be invoked either from the KVM host vcpu thread (when exiting the vcpu run loop), or via a preempt notifier. In the former case, only preemption is disabled. In the latter case, the function is called from inside __schedule(), which means that IRQs are disabled. Use of local_bh_disable()/_enable() with IRQs disabled is considerd an error, resulting in lockdep splats while running VMs if lockdep is enabled. This patch disables IRQs instead of attempting to disable softirqs, avoiding the problem of calling local_bh_enable() with IRQs disabled in the __schedule() path. This creates an additional interrupt blackout during vcpu run loop exit, but this is the rare case and the blackout latency is still less than that of __schedule(). Fixes: e6b673b741ea ("KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing") Reported-by: Andre Przywara <andre.przywara@arm.com> Signed-off-by: Dave Martin <Dave.Martin@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-06-15 15:47:24 +00:00
local_irq_restore(flags);
KVM: arm64: Optimise FPSIMD handling to reduce guest/host thrashing This patch refactors KVM to align the host and guest FPSIMD save/restore logic with each other for arm64. This reduces the number of redundant save/restore operations that must occur, and reduces the common-case IRQ blackout time during guest exit storms by saving the host state lazily and optimising away the need to restore the host state before returning to the run loop. Four hooks are defined in order to enable this: * kvm_arch_vcpu_run_map_fp(): Called on PID change to map necessary bits of current to Hyp. * kvm_arch_vcpu_load_fp(): Set up FP/SIMD for entering the KVM run loop (parse as "vcpu_load fp"). * kvm_arch_vcpu_ctxsync_fp(): Get FP/SIMD into a safe state for re-enabling interrupts after a guest exit back to the run loop. For arm64 specifically, this involves updating the host kernel's FPSIMD context tracking metadata so that kernel-mode NEON use will cause the vcpu's FPSIMD state to be saved back correctly into the vcpu struct. This must be done before re-enabling interrupts because kernel-mode NEON may be used by softirqs. * kvm_arch_vcpu_put_fp(): Save guest FP/SIMD state back to memory and dissociate from the CPU ("vcpu_put fp"). Also, the arm64 FPSIMD context switch code is updated to enable it to save back FPSIMD state for a vcpu, not just current. A few helpers drive this: * fpsimd_bind_state_to_cpu(struct user_fpsimd_state *fp): mark this CPU as having context fp (which may belong to a vcpu) currently loaded in its registers. This is the non-task equivalent of the static function fpsimd_bind_to_cpu() in fpsimd.c. * task_fpsimd_save(): exported to allow KVM to save the guest's FPSIMD state back to memory on exit from the run loop. * fpsimd_flush_state(): invalidate any context's FPSIMD state that is currently loaded. Used to disassociate the vcpu from the CPU regs on run loop exit. These changes allow the run loop to enable interrupts (and thus softirqs that may use kernel-mode NEON) without having to save the guest's FPSIMD state eagerly. Some new vcpu_arch fields are added to make all this work. Because host FPSIMD state can now be saved back directly into current's thread_struct as appropriate, host_cpu_context is no longer used for preserving the FPSIMD state. However, it is still needed for preserving other things such as the host's system registers. To avoid ABI churn, the redundant storage space in host_cpu_context is not removed for now. arch/arm is not addressed by this patch and continues to use its current save/restore logic. It could provide implementations of the helpers later if desired. Signed-off-by: Dave Martin <Dave.Martin@arm.com> Reviewed-by: Marc Zyngier <marc.zyngier@arm.com> Reviewed-by: Christoffer Dall <christoffer.dall@arm.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
2018-04-06 13:55:59 +00:00
}
Reference in New Issue Copy Permalink