[PATCH OLK-6.6 v2 00/15] Support NMI in the virtual machine
From: caijian <caijian11@h-partners.com> The patches support for nmi in the virtual machine. https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/log/?h... Marc Zyngier (15): KVM: arm64: vgic-v3: Upgrade AP1Rn to 64bit. KVM: arm64: vgic-v3: Allow the NMI state to make it into the LRs KVM: arm64: vgic-v3: Make NMI priority RES0 KVM: arm64: vgic-v4: Propagate the NMI state into the GICv4.1 VSGI configuration KVM: arm64: vgic-v3: Use the NMI attribute as part of the AP-list sorting KVM: arm64: vgic-v3: Add support for GIC{D,R}_INMIR registers KVM: arm64: vgic-v3: Add userspace selection for GICv3.3 NMI KVM: arm64: vgic-debug: Add the NMI field to the debug output KVM: arm64: Allow userspace to control ID_AA64PFR1_EL1.NMI KVM: arm64: Don't trap ALLINT accesses if the vcpu has FEAT_NMI KVM: arm64: vgic-v3: Don't inject an NMI if the vcpu doesn't have FEAT_NMI KVM: arm64: Allow GICv3.3 NMI if the host supports it KVM: arm64: Handle traps of ALLINT arm64: Decouple KVM from CONFIG_ARM64_NMI KVM: arm64: vgic-v3: Handle traps of ICV_NMIAR1_EL1 arch/arm64/include/asm/cpufeature.h | 3 +- arch/arm64/include/asm/kvm_host.h | 2 + arch/arm64/include/asm/sysreg.h | 3 ++ arch/arm64/kernel/cpufeature.c | 13 +++-- arch/arm64/kvm/arm.c | 3 ++ arch/arm64/kvm/hyp/include/hyp/switch.h | 6 ++- arch/arm64/kvm/hyp/vgic-v3-sr.c | 14 +++-- arch/arm64/kvm/sys_regs.c | 33 +++++++++++- arch/arm64/kvm/vgic/vgic-debug.c | 9 ++-- arch/arm64/kvm/vgic/vgic-init.c | 10 +++- arch/arm64/kvm/vgic/vgic-mmio-v2.c | 1 + arch/arm64/kvm/vgic/vgic-mmio-v3.c | 70 ++++++++++++++++++++++++- arch/arm64/kvm/vgic/vgic-mmio.c | 25 ++++++--- arch/arm64/kvm/vgic/vgic-v3.c | 12 ++++- arch/arm64/kvm/vgic/vgic-v4.c | 1 + arch/arm64/kvm/vgic/vgic.c | 8 ++- arch/arm64/kvm/vgic/vgic.h | 1 + drivers/irqchip/irq-gic-v3-its.c | 9 ++++ drivers/irqchip/irq-gic-v3.c | 12 +++++ drivers/irqchip/irq-gic-v4.c | 3 +- include/kvm/arm_vgic.h | 10 +++- include/linux/irqchip/arm-gic-v4.h | 4 +- include/linux/irqchip/arm-vgic-info.h | 2 + 23 files changed, 224 insertions(+), 30 deletions(-) -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- With the advent of FEAT_GIC_NMI, ICH_AP1R0_EL2 is now a 64bit register, as the NMI priority is encoded in bit 63. Upgrade the whole of the AP1Rn array to 64bit, leaving the Group0 equivalent to 32bit. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/hyp/vgic-v3-sr.c | 11 ++++++++--- include/kvm/arm_vgic.h | 2 +- 2 files changed, 9 insertions(+), 4 deletions(-) diff --git a/arch/arm64/kvm/hyp/vgic-v3-sr.c b/arch/arm64/kvm/hyp/vgic-v3-sr.c index 6cb638b184b1..05955e369c51 100644 --- a/arch/arm64/kvm/hyp/vgic-v3-sr.c +++ b/arch/arm64/kvm/hyp/vgic-v3-sr.c @@ -130,7 +130,11 @@ static void __vgic_v3_write_ap0rn(u32 val, int n) } } -static void __vgic_v3_write_ap1rn(u32 val, int n) +/* + * Contrary to ICH_AP0Rn_EL2, ICH_AP1R0_EL2 is 64bit, thanks to the + * NMI bit stuck at [63]. Isn't that fun? + */ +static void __vgic_v3_write_ap1rn(u64 val, int n) { switch (n) { case 0: @@ -172,9 +176,10 @@ static u32 __vgic_v3_read_ap0rn(int n) return val; } -static u32 __vgic_v3_read_ap1rn(int n) +/* Same remark about the 64bit-ness of AP1R0 */ +static u64 __vgic_v3_read_ap1rn(int n) { - u32 val; + u64 val; switch (n) { case 0: diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h index 664d4a16970b..8d41e8ad7cf7 100644 --- a/include/kvm/arm_vgic.h +++ b/include/kvm/arm_vgic.h @@ -375,7 +375,7 @@ struct vgic_v3_cpu_if { u32 vgic_vmcr; u32 vgic_sre; /* Restored only, change ignored */ u32 vgic_ap0r[4]; - u32 vgic_ap1r[4]; + u64 vgic_ap1r[4]; u64 vgic_lr[VGIC_V3_MAX_LRS]; /* -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- Add the new NMI state to the vgic IRQ state, and allow it to make it into the LRs. Nothing can set it yet, so no impact is expected from this change. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/include/asm/sysreg.h | 1 + arch/arm64/kvm/vgic/vgic-v3.c | 5 ++++- include/kvm/arm_vgic.h | 1 + 3 files changed, 6 insertions(+), 1 deletion(-) diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h index 0086a617dfa1..4783bd2eb819 100644 --- a/arch/arm64/include/asm/sysreg.h +++ b/arch/arm64/include/asm/sysreg.h @@ -932,6 +932,7 @@ #define ICH_LR_VIRTUAL_ID_MASK ((1ULL << 32) - 1) #define ICH_LR_EOI (1ULL << 41) +#define ICH_LR_NMI (1ULL << 59) #define ICH_LR_GROUP (1ULL << 60) #define ICH_LR_HW (1ULL << 61) #define ICH_LR_STATE (3ULL << 62) diff --git a/arch/arm64/kvm/vgic/vgic-v3.c b/arch/arm64/kvm/vgic/vgic-v3.c index c71fc1a70e24..ba6c907ac1f4 100644 --- a/arch/arm64/kvm/vgic/vgic-v3.c +++ b/arch/arm64/kvm/vgic/vgic-v3.c @@ -181,7 +181,10 @@ void vgic_v3_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr) if (irq->group) val |= ICH_LR_GROUP; - val |= (u64)irq->priority << ICH_LR_PRIORITY_SHIFT; + if (irq->nmi) + val |= ICH_LR_NMI; + else + val |= (u64)irq->priority << ICH_LR_PRIORITY_SHIFT; vcpu->arch.vgic_cpu.vgic_v3.vgic_lr[lr] = val; } diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h index 8d41e8ad7cf7..1b39e6221ecf 100644 --- a/include/kvm/arm_vgic.h +++ b/include/kvm/arm_vgic.h @@ -180,6 +180,7 @@ struct vgic_irq { bool active; /* not used for LPIs */ bool enabled; bool hw; /* Tied to HW IRQ */ + bool nmi; /* Configured as NMI */ struct kref refcount; /* Used for LPIs */ u32 hwintid; /* HW INTID number */ unsigned int host_irq; /* linux irq corresponding to hwintid */ -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- The priority of an NMI is always RES0. Let's enforce it when the guest accesses the priority MMIO range. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-mmio.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) diff --git a/arch/arm64/kvm/vgic/vgic-mmio.c b/arch/arm64/kvm/vgic/vgic-mmio.c index 7aca296f9b3f..b7ba5b73cc0c 100644 --- a/arch/arm64/kvm/vgic/vgic-mmio.c +++ b/arch/arm64/kvm/vgic/vgic-mmio.c @@ -699,13 +699,17 @@ unsigned long vgic_mmio_read_priority(struct kvm_vcpu *vcpu, gpa_t addr, unsigned int len) { u32 intid = VGIC_ADDR_TO_INTID(addr, 8); + unsigned long flags; int i; u64 val = 0; for (i = 0; i < len; i++) { struct vgic_irq *irq = vgic_get_irq(vcpu->kvm, vcpu, intid + i); - val |= (u64)irq->priority << (i * 8); + raw_spin_lock_irqsave(&irq->irq_lock, flags); + if (!irq->nmi) + val |= (u64)irq->priority << (i * 8); + raw_spin_unlock_irqrestore(&irq->irq_lock, flags); vgic_put_irq(vcpu->kvm, irq); } @@ -732,10 +736,15 @@ void vgic_mmio_write_priority(struct kvm_vcpu *vcpu, struct vgic_irq *irq = vgic_get_irq(vcpu->kvm, vcpu, intid + i); raw_spin_lock_irqsave(&irq->irq_lock, flags); - /* Narrow the priority range to what we actually support */ - irq->priority = (val >> (i * 8)) & GENMASK(7, 8 - VGIC_PRI_BITS); - if (vgic_direct_sgi_or_ppi(irq)) - vgic_update_vsgi(irq); + if (!irq->nmi) { + /* + * Narrow the priority range to what we + * actually support + */ + irq->priority = (val >> (i * 8)) & GENMASK(7, 8 - VGIC_PRI_BITS); + if (vgic_direct_sgi_or_ppi(irq)) + vgic_update_vsgi(irq); + } raw_spin_unlock_irqrestore(&irq->irq_lock, flags); vgic_put_irq(vcpu->kvm, irq); -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- Just as we now allow the NMI state to make it into the LRs, allow the same state to be propagated into the VSGI configuration. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-mmio.c | 6 ++++-- arch/arm64/kvm/vgic/vgic-v4.c | 1 + arch/arm64/kvm/vgic/vgic.h | 1 + drivers/irqchip/irq-gic-v3-its.c | 9 +++++++++ drivers/irqchip/irq-gic-v4.c | 3 ++- include/linux/irqchip/arm-gic-v4.h | 4 +++- 6 files changed, 20 insertions(+), 4 deletions(-) diff --git a/arch/arm64/kvm/vgic/vgic-mmio.c b/arch/arm64/kvm/vgic/vgic-mmio.c index b7ba5b73cc0c..59b81e5ce1d8 100644 --- a/arch/arm64/kvm/vgic/vgic-mmio.c +++ b/arch/arm64/kvm/vgic/vgic-mmio.c @@ -61,9 +61,11 @@ unsigned long vgic_mmio_read_group(struct kvm_vcpu *vcpu, return value; } -static void vgic_update_vsgi(struct vgic_irq *irq) +void vgic_update_vsgi(struct vgic_irq *irq) { - WARN_ON(its_prop_update_vsgi(irq->host_irq, irq->priority, irq->group)); + WARN_ON(its_prop_update_vsgi(irq->host_irq, + irq->nmi ? 0 : irq->priority, + irq->group, irq->nmi)); } void vgic_mmio_write_group(struct kvm_vcpu *vcpu, gpa_t addr, diff --git a/arch/arm64/kvm/vgic/vgic-v4.c b/arch/arm64/kvm/vgic/vgic-v4.c index 1d02c50593d9..06fec5cea3d7 100644 --- a/arch/arm64/kvm/vgic/vgic-v4.c +++ b/arch/arm64/kvm/vgic/vgic-v4.c @@ -110,6 +110,7 @@ static void vgic_v4_sync_sgi_config(struct its_vpe *vpe, struct vgic_irq *irq) vpe->sgi_config[irq->intid].enabled = irq->enabled; vpe->sgi_config[irq->intid].group = irq->group; vpe->sgi_config[irq->intid].priority = irq->priority; + vpe->sgi_config[irq->intid].nmi = irq->nmi; } static void vgic_v4_enable_vsgis(struct kvm_vcpu *vcpu) diff --git a/arch/arm64/kvm/vgic/vgic.h b/arch/arm64/kvm/vgic/vgic.h index 764038fadf3f..e38bf7d94d3b 100644 --- a/arch/arm64/kvm/vgic/vgic.h +++ b/arch/arm64/kvm/vgic/vgic.h @@ -352,6 +352,7 @@ bool vgic_supports_direct_msis(struct kvm *kvm); int vgic_v4_init(struct kvm *kvm); void vgic_v4_teardown(struct kvm *kvm); void vgic_v4_configure_vsgis(struct kvm *kvm); +void vgic_update_vsgi(struct vgic_irq *irq); void vgic_v4_get_vlpi_state(struct vgic_irq *irq, bool *val); int vgic_v4_request_vpe_irq(struct kvm_vcpu *vcpu, int irq); #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c index c66676a21ad0..72a27a018e5c 100644 --- a/drivers/irqchip/irq-gic-v3-its.c +++ b/drivers/irqchip/irq-gic-v3-its.c @@ -627,6 +627,7 @@ struct its_cmd_desc { u8 sgi; u8 priority; bool enable; + bool nmi; bool group; bool clear; } its_vsgi_cmd; @@ -800,6 +801,11 @@ static void its_encode_sgi_priority(struct its_cmd_block *cmd, u8 prio) its_mask_encode(&cmd->raw_cmd[0], prio >> 4, 23, 20); } +static void its_encode_sgi_nmi(struct its_cmd_block *cmd, bool nmi) +{ + its_mask_encode(&cmd->raw_cmd[0], nmi, 11, 11); +} + static void its_encode_sgi_group(struct its_cmd_block *cmd, bool grp) { its_mask_encode(&cmd->raw_cmd[0], grp, 10, 10); @@ -1203,6 +1209,7 @@ static struct its_vpe *its_build_vsgi_cmd(struct its_node *its, its_encode_sgi_intid(cmd, desc->its_vsgi_cmd.sgi); #endif its_encode_sgi_priority(cmd, desc->its_vsgi_cmd.priority); + its_encode_sgi_nmi(cmd, desc->its_vsgi_cmd.nmi); its_encode_sgi_group(cmd, desc->its_vsgi_cmd.group); its_encode_sgi_clear(cmd, desc->its_vsgi_cmd.clear); its_encode_sgi_enable(cmd, desc->its_vsgi_cmd.enable); @@ -4495,6 +4502,7 @@ static void its_configure_sgi(struct irq_data *d, bool clear) desc.its_vsgi_cmd.priority = vpe->sgi_config[d->hwirq].priority; desc.its_vsgi_cmd.enable = vpe->sgi_config[d->hwirq].enabled; desc.its_vsgi_cmd.group = vpe->sgi_config[d->hwirq].group; + desc.its_vsgi_cmd.nmi = vpe->sgi_config[d->hwirq].nmi; desc.its_vsgi_cmd.clear = clear; /* @@ -4653,6 +4661,7 @@ static int its_sgi_set_vcpu_affinity(struct irq_data *d, void *vcpu_info) case PROP_UPDATE_VSGI: vpe->sgi_config[d->hwirq].priority = info->priority; vpe->sgi_config[d->hwirq].group = info->group; + vpe->sgi_config[d->hwirq].nmi = info->nmi; its_configure_sgi(d, false); return 0; diff --git a/drivers/irqchip/irq-gic-v4.c b/drivers/irqchip/irq-gic-v4.c index 22a9a7f1739f..494001500261 100644 --- a/drivers/irqchip/irq-gic-v4.c +++ b/drivers/irqchip/irq-gic-v4.c @@ -393,13 +393,14 @@ int its_prop_update_vlpi(int irq, u8 config, bool inv) return irq_set_vcpu_affinity(irq, &info); } -int its_prop_update_vsgi(int irq, u8 priority, bool group) +int its_prop_update_vsgi(int irq, u8 priority, bool group, bool nmi) { struct its_cmd_info info = { .cmd_type = PROP_UPDATE_VSGI, { .priority = priority, .group = group, + .nmi = nmi, }, }; diff --git a/include/linux/irqchip/arm-gic-v4.h b/include/linux/irqchip/arm-gic-v4.h index 90b23d8ebab9..f489defdd6e1 100644 --- a/include/linux/irqchip/arm-gic-v4.h +++ b/include/linux/irqchip/arm-gic-v4.h @@ -57,6 +57,7 @@ struct its_vpe { u8 priority; bool enabled; bool group; + bool nmi; #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS } sgi_config[32]; int nr_irqs; @@ -129,6 +130,7 @@ struct its_cmd_info { struct { u8 priority; bool group; + bool nmi; }; }; }; @@ -143,7 +145,7 @@ int its_map_vlpi(int irq, struct its_vlpi_map *map); int its_get_vlpi(int irq, struct its_vlpi_map *map); int its_unmap_vlpi(int irq); int its_prop_update_vlpi(int irq, u8 config, bool inv); -int its_prop_update_vsgi(int irq, u8 priority, bool group); +int its_prop_update_vsgi(int irq, u8 priority, bool group, bool nmi); struct irq_domain_ops; int its_init_v4(struct irq_domain *domain, -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- Since we want NMIs to make it quicker into the LRs, add them to the priority sorting. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/arch/arm64/kvm/vgic/vgic.c b/arch/arm64/kvm/vgic/vgic.c index 6e1fa6f2918f..2459b0adea08 100644 --- a/arch/arm64/kvm/vgic/vgic.c +++ b/arch/arm64/kvm/vgic/vgic.c @@ -252,6 +252,7 @@ static struct kvm_vcpu *vgic_target_oracle(struct vgic_irq *irq) * * Otherwise things should be sorted by the priority field and the GIC * hardware support will take care of preemption of priority groups etc. + * NMI acts as a super-priority. * * Return negative if "a" sorts before "b", 0 to preserve order, and positive * to sort "b" before "a". @@ -287,7 +288,12 @@ static int vgic_irq_cmp(void *priv, const struct list_head *a, goto out; } - /* Both pending and enabled, sort by priority */ + /* Both pending and enabled, sort by NMI and then priority */ + if (irqa->nmi != irqb->nmi) { + ret = (int)irqb->nmi - (int)irqa->nmi; + goto out; + } + ret = irqa->priority - irqb->priority; out: raw_spin_unlock(&irqb->irq_lock); -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- Plumb the distributor and redistributor NMI configuration registerse into the MMIO spaces. The update to the registers are gated by a distributor flag which is never set, so there is still no observable behaviour change... Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-mmio-v3.c | 55 ++++++++++++++++++++++++++++++ include/kvm/arm_vgic.h | 1 + 2 files changed, 56 insertions(+) diff --git a/arch/arm64/kvm/vgic/vgic-mmio-v3.c b/arch/arm64/kvm/vgic/vgic-mmio-v3.c index dde858daa7d7..ca14daf8c13a 100644 --- a/arch/arm64/kvm/vgic/vgic-mmio-v3.c +++ b/arch/arm64/kvm/vgic/vgic-mmio-v3.c @@ -598,6 +598,55 @@ static void vgic_mmio_write_invall(struct kvm_vcpu *vcpu, vgic_set_rdist_busy(vcpu, false); } +static unsigned long vgic_mmio_read_nmi(struct kvm_vcpu *vcpu, + gpa_t addr, unsigned int len) +{ + u32 intid = VGIC_ADDR_TO_INTID(addr, 1); + u32 value = 0; + int i; + + /* Loop over all IRQs affected by this read */ + for (i = 0; i < len * 8; i++) { + struct vgic_irq *irq = vgic_get_irq(vcpu->kvm, vcpu, intid + i); + + if (irq->nmi) + value |= (1U << i); + + vgic_put_irq(vcpu->kvm, irq); + } + + return value; +} + +static void vgic_mmio_write_nmi(struct kvm_vcpu *vcpu, gpa_t addr, + unsigned int len, unsigned long val) +{ + u32 intid = VGIC_ADDR_TO_INTID(addr, 1); + unsigned long flags; + int i; + + if (!vcpu->kvm->arch.vgic.has_nmi) + return; + + for (i = 0; i < len * 8; i++) { + struct vgic_irq *irq = vgic_get_irq(vcpu->kvm, vcpu, intid + i); + bool was_nmi; + + raw_spin_lock_irqsave(&irq->irq_lock, flags); + + was_nmi = irq->nmi; + irq->nmi = (val & BIT(i)); + + if (irq->hw && vgic_irq_is_sgi(irq->intid) && + was_nmi != irq->nmi) + vgic_update_vsgi(irq); + + raw_spin_unlock_irqrestore(&irq->irq_lock, flags); + + vgic_put_irq(vcpu->kvm, irq); + } +} + /* * The GICv3 per-IRQ registers are split to control PPIs and SGIs in the * redistributors, while SPIs are covered by registers in the distributor @@ -671,6 +720,9 @@ static const struct vgic_register_region vgic_v3_dist_registers[] = { REGISTER_DESC_WITH_BITS_PER_IRQ_SHARED(GICD_IGRPMODR, vgic_mmio_read_raz, vgic_mmio_write_wi, NULL, NULL, 1, VGIC_ACCESS_32bit), + REGISTER_DESC_WITH_BITS_PER_IRQ_SHARED(GICD_INMIR, + vgic_mmio_read_nmi, vgic_mmio_write_nmi, NULL, NULL, 1, + VGIC_ACCESS_32bit), REGISTER_DESC_WITH_BITS_PER_IRQ_SHARED(GICD_IROUTER, vgic_mmio_read_irouter, vgic_mmio_write_irouter, NULL, NULL, 64, VGIC_ACCESS_64bit | VGIC_ACCESS_32bit), @@ -755,6 +807,9 @@ static const struct vgic_register_region vgic_v3_rd_registers[] = { REGISTER_DESC_WITH_LENGTH(SZ_64K + GICR_NSACR, vgic_mmio_read_raz, vgic_mmio_write_wi, 4, VGIC_ACCESS_32bit), + REGISTER_DESC_WITH_LENGTH(SZ_64K + GICR_INMIR0, + vgic_mmio_read_nmi, vgic_mmio_write_nmi, 4, + VGIC_ACCESS_32bit), }; unsigned int vgic_v3_init_dist_iodev(struct vgic_io_device *dev) diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h index 1b39e6221ecf..a90776388130 100644 --- a/include/kvm/arm_vgic.h +++ b/include/kvm/arm_vgic.h @@ -321,6 +321,7 @@ struct vgic_dist { struct vgic_io_device dist_iodev; + bool has_nmi; bool has_its; bool table_write_in_progress; -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ---------------------------------------------------------------- In order to allow the GIC NMI support to be selected from userspace, add userspace access to GICD_TYPER, which allows the NMI support bit to be set. This is gated by a global capability that nobody can set yet, so no observable change is expected. Again. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-init.c | 10 +++++++++- arch/arm64/kvm/vgic/vgic-mmio-v2.c | 1 + arch/arm64/kvm/vgic/vgic-mmio-v3.c | 15 ++++++++++++++- include/kvm/arm_vgic.h | 6 +++++- 4 files changed, 29 insertions(+), 3 deletions(-) diff --git a/arch/arm64/kvm/vgic/vgic-init.c b/arch/arm64/kvm/vgic/vgic-init.c index 5ca9d945e3a2..0d9d71dc338e 100644 --- a/arch/arm64/kvm/vgic/vgic-init.c +++ b/arch/arm64/kvm/vgic/vgic-init.c @@ -353,8 +353,16 @@ int vgic_init(struct kvm *kvm) * If userspace didn't set the GIC implementation revision, * default to the latest and greatest. You know want it. */ - if (!dist->implementation_rev) + if (!dist->implementation_rev) { dist->implementation_rev = KVM_VGIC_IMP_REV_LATEST; + /* + * Advertise NMI if available. Userspace that explicitly + * doesn't want NMI will have written to GICD_{IIDR,TYPER} + * to set the implementation and the NMI support status. + */ + dist->has_nmi = kvm_vgic_global_state.has_nmi; + } + dist->initialized = true; out: diff --git a/arch/arm64/kvm/vgic/vgic-mmio-v2.c b/arch/arm64/kvm/vgic/vgic-mmio-v2.c index e070cda86e12..365d83b0f263 100644 --- a/arch/arm64/kvm/vgic/vgic-mmio-v2.c +++ b/arch/arm64/kvm/vgic/vgic-mmio-v2.c @@ -95,6 +95,7 @@ static int vgic_mmio_uaccess_write_v2_misc(struct kvm_vcpu *vcpu, switch (reg) { case KVM_VGIC_IMP_REV_2: case KVM_VGIC_IMP_REV_3: + case KVM_VGIC_IMP_REV_4: vcpu->kvm->arch.vgic.v2_groups_user_writable = true; dist->implementation_rev = reg; return 0; diff --git a/arch/arm64/kvm/vgic/vgic-mmio-v3.c b/arch/arm64/kvm/vgic/vgic-mmio-v3.c index ca14daf8c13a..3dd779cae09f 100644 --- a/arch/arm64/kvm/vgic/vgic-mmio-v3.c +++ b/arch/arm64/kvm/vgic/vgic-mmio-v3.c @@ -78,6 +78,8 @@ static unsigned long vgic_mmio_read_v3_misc(struct kvm_vcpu *vcpu, case GICD_TYPER: value = vgic->nr_spis + VGIC_NR_PRIVATE_IRQS; value = (value >> 5) - 1; + if (vgic->has_nmi) + value |= GICD_TYPER_NMI; if (vgic_has_its(vcpu->kvm)) { value |= (INTERRUPT_ID_BITS_ITS - 1) << 19; value |= GICD_TYPER_LPIS; @@ -158,6 +160,13 @@ static int vgic_mmio_uaccess_write_v3_misc(struct kvm_vcpu *vcpu, u32 reg; switch (addr & 0x0c) { + case GICD_TYPER: + if (dist->implementation_rev >= KVM_VGIC_IMP_REV_4 && + kvm_vgic_global_state.has_nmi) + dist->has_nmi = val & GICD_TYPER_NMI; + else if (val & GICD_TYPER_NMI) + return -EINVAL; + return 0; case GICD_TYPER2: if (val != vgic_mmio_read_v3_misc(vcpu, addr, len)) return -EINVAL; @@ -171,6 +180,10 @@ static int vgic_mmio_uaccess_write_v3_misc(struct kvm_vcpu *vcpu, switch (reg) { case KVM_VGIC_IMP_REV_2: case KVM_VGIC_IMP_REV_3: + /* Disable NMI on selecting an older revision */ + dist->has_nmi = false; + fallthrough; + case KVM_VGIC_IMP_REV_4: dist->implementation_rev = reg; return 0; default: @@ -186,7 +199,7 @@ static int vgic_mmio_uaccess_write_v3_misc(struct kvm_vcpu *vcpu, return 0; } - vgic_mmio_write_v3_misc(vcpu, addr, len, val); + /* Not reachable... */ return 0; } diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h index a90776388130..0cbcf4e06761 100644 --- a/include/kvm/arm_vgic.h +++ b/include/kvm/arm_vgic.h @@ -108,6 +108,9 @@ struct vgic_global { bool has_gicv4; bool has_gicv4_1; + /* NMI */ + bool has_nmi; + /* Pseudo GICv3 from outer space */ bool no_hw_deactivation; @@ -288,7 +291,8 @@ struct vgic_dist { u32 implementation_rev; #define KVM_VGIC_IMP_REV_2 2 /* GICv2 restorable groups */ #define KVM_VGIC_IMP_REV_3 3 /* GICv3 GICR_CTLR.{IW,CES,RWP} */ -#define KVM_VGIC_IMP_REV_LATEST KVM_VGIC_IMP_REV_3 +#define KVM_VGIC_IMP_REV_4 4 /* GICv3 NMI */ +#define KVM_VGIC_IMP_REV_LATEST KVM_VGIC_IMP_REV_4 /* Userspace can write to GICv2 IGROUPR */ bool v2_groups_user_writable; -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Add the per-INTID NMI state to the vgic-state file. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-debug.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/arm64/kvm/vgic/vgic-debug.c b/arch/arm64/kvm/vgic/vgic-debug.c index 20fed597531b..340f960a1154 100644 --- a/arch/arm64/kvm/vgic/vgic-debug.c +++ b/arch/arm64/kvm/vgic/vgic-debug.c @@ -155,7 +155,7 @@ static void print_dist_state(struct seq_file *s, struct vgic_dist *dist) seq_printf(s, "P=pending_latch, L=line_level, A=active\n"); seq_printf(s, "E=enabled, H=hw, C=config (level=1, edge=0)\n"); - seq_printf(s, "G=group\n"); + seq_printf(s, "G=group, N=NMI\n"); } static void print_header(struct seq_file *s, struct vgic_irq *irq, @@ -170,8 +170,8 @@ static void print_header(struct seq_file *s, struct vgic_irq *irq, } seq_printf(s, "\n"); - seq_printf(s, "%s%2d TYP ID TGT_ID PLAEHCG HWID TARGET SRC PRI VCPU_ID\n", hdr, id); - seq_printf(s, "----------------------------------------------------------------\n"); + seq_printf(s, "%s%2d TYP ID TGT_ID PLAEHCGN HWID TARGET SRC PRI VCPU_ID\n", hdr, id); + seq_printf(s, "-----------------------------------------------------------------\n"); } static void print_irq_state(struct seq_file *s, struct vgic_irq *irq, @@ -204,7 +204,7 @@ static void print_irq_state(struct seq_file *s, struct vgic_irq *irq, seq_printf(s, " %s %4d " " %2d " - "%d%d%d%d%d%d%d " + "%d%d%d%d%d%d%d%d " "%8d " "%8x " " %2x " @@ -220,6 +220,7 @@ static void print_irq_state(struct seq_file *s, struct vgic_irq *irq, irq->hw, irq->config == VGIC_CONFIG_LEVEL, irq->group, + irq->nmi, irq->hwintid, irq->mpidr, irq->source, -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- In order for userspace to be able to control the CPU side of the NMI distribution (just like we have it on the GIC side), allow it to set/clear ID_AA64PFR1_EL1.NMI. This relies on a per-VM property that defaults to the host value. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/include/asm/kvm_host.h | 2 ++ arch/arm64/kvm/arm.c | 3 +++ arch/arm64/kvm/sys_regs.c | 26 +++++++++++++++++++++++++- 3 files changed, 30 insertions(+), 1 deletion(-) diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h index 8c64fa87279b..abe581982a75 100644 --- a/arch/arm64/include/asm/kvm_host.h +++ b/arch/arm64/include/asm/kvm_host.h @@ -209,6 +209,8 @@ struct kvm_arch { /* VTCR_EL2 value for this VM */ u64 vtcr; + u8 pfr1_nmi; + /* Interrupt controller */ struct vgic_dist vgic; diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c index 7f6de7585894..cdfdd08f4c10 100644 --- a/arch/arm64/kvm/arm.c +++ b/arch/arm64/kvm/arm.c @@ -192,6 +192,9 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) /* The maximum number of VCPUs is limited by the host's GIC model */ kvm->max_vcpus = kvm_arm_default_max_vcpus(); + if (system_uses_nmi()) + kvm->arch.pfr1_nmi = ID_AA64PFR1_EL1_NMI_IMP; + kvm_arm_init_hypercalls(kvm); bitmap_zero(kvm->arch.vcpu_features, KVM_VCPU_MAX_FEATURES); diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index efc8fbefb434..5a6e64670feb 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -1342,6 +1342,7 @@ static u64 __kvm_read_sanitised_id_reg(const struct kvm_vcpu *vcpu, val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_SME); val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_NMI); + val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_NMI), vcpu->kvm->arch.pfr1_nmi); break; case SYS_ID_AA64ISAR1_EL1: if (!vcpu_has_ptrauth(vcpu)) @@ -1572,6 +1573,27 @@ static int set_id_dfr0_el1(struct kvm_vcpu *vcpu, return set_id_reg(vcpu, rd, val); } +static int set_id_aa64pfr1_el1(struct kvm_vcpu *vcpu, + const struct sys_reg_desc *rd, + u64 val) +{ + u8 nmi; + + nmi = cpuid_feature_extract_unsigned_field(val, ID_AA64PFR1_EL1_NMI_SHIFT); + if (nmi > ID_AA64PFR1_EL1_NMI_IMP || (nmi && !system_uses_nmi())) + return -EINVAL; + + /* We can only differ with NMI, and anything else is an error */ + val ^= read_id_reg(vcpu, rd); + val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_NMI); + if (val) + return -EINVAL; + + vcpu->kvm->arch.pfr1_nmi = nmi; + + return 0; +} + /* * cpufeature ID register user accessors * @@ -2033,7 +2055,9 @@ static const struct sys_reg_desc sys_reg_descs[] = { .set_user = set_id_reg, .reset = read_sanitised_id_aa64pfr0_el1, .val = ID_AA64PFR0_EL1_CSV2_MASK | ID_AA64PFR0_EL1_CSV3_MASK, }, - ID_SANITISED(ID_AA64PFR1_EL1), + { SYS_DESC(SYS_ID_AA64PFR1_EL1), .access = access_id_reg, + .get_user = get_id_reg, .set_user = set_id_aa64pfr1_el1, + .reset = kvm_read_sanitised_id_reg, }, ID_UNALLOCATED(4,2), ID_UNALLOCATED(4,3), ID_SANITISED(ID_AA64ZFR0_EL1), -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- This is counter-productive, so let it rip. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/hyp/include/hyp/switch.h | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/arch/arm64/kvm/hyp/include/hyp/switch.h b/arch/arm64/kvm/hyp/include/hyp/switch.h index 0a7f049bcf91..821d803b395d 100644 --- a/arch/arm64/kvm/hyp/include/hyp/switch.h +++ b/arch/arm64/kvm/hyp/include/hyp/switch.h @@ -224,7 +224,8 @@ static inline void __activate_traps_common(struct kvm_vcpu *vcpu) vcpu_set_flag(vcpu, PMUSERENR_ON_CPU); } - if (cpus_have_final_cap(ARM64_HAS_NMI)) + if (cpus_have_final_cap(ARM64_HAS_NMI) && + !kern_hyp_va(vcpu->kvm)->arch.pfr1_nmi) sysreg_clear_set_s(SYS_HCRX_EL2, 0, HCRX_EL2_TALLINT); vcpu->arch.mdcr_el2_host = read_sysreg(mdcr_el2); @@ -252,7 +253,8 @@ static inline void __deactivate_traps_common(struct kvm_vcpu *vcpu) { write_sysreg(vcpu->arch.mdcr_el2_host, mdcr_el2); - if (cpus_have_final_cap(ARM64_HAS_NMI)) + if (cpus_have_final_cap(ARM64_HAS_NMI) && + !kern_hyp_va(vcpu->kvm)->arch.pfr1_nmi) sysreg_clear_set_s(SYS_HCRX_EL2, HCRX_EL2_TALLINT, 0); write_sysreg(0, hstr_el2); -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Since it is allowed to have any combination of CPU and GIC supporting NMIs or not, let's drop the NMI feature at the point where it is injected in the LR if the vcpu doesn't have FEAT_NMI. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/vgic/vgic-v3.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/arch/arm64/kvm/vgic/vgic-v3.c b/arch/arm64/kvm/vgic/vgic-v3.c index ba6c907ac1f4..e82402dada77 100644 --- a/arch/arm64/kvm/vgic/vgic-v3.c +++ b/arch/arm64/kvm/vgic/vgic-v3.c @@ -181,7 +181,8 @@ void vgic_v3_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr) if (irq->group) val |= ICH_LR_GROUP; - if (irq->nmi) + if (vcpu->kvm->arch.pfr1_nmi == ID_AA64PFR1_EL1_NMI_IMP && + irq->nmi) val |= ICH_LR_NMI; else val |= (u64)irq->priority << ICH_LR_PRIORITY_SHIFT; -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Let the GICv3 driver populate a has_nmi flag in the kvm_gic_info structure, and communicate it to the rest of KVM. We disallow the use of NMI if trapping of the CPU interface is enabled, because life is definitely too short to write more emulation. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/arm.c | 2 +- arch/arm64/kvm/sys_regs.c | 3 ++- arch/arm64/kvm/vgic/vgic-v3.c | 6 ++++++ drivers/irqchip/irq-gic-v3.c | 1 + include/linux/irqchip/arm-vgic-info.h | 2 ++ 5 files changed, 12 insertions(+), 2 deletions(-) diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c index cdfdd08f4c10..9c6b4accb10b 100644 --- a/arch/arm64/kvm/arm.c +++ b/arch/arm64/kvm/arm.c @@ -192,7 +192,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) /* The maximum number of VCPUs is limited by the host's GIC model */ kvm->max_vcpus = kvm_arm_default_max_vcpus(); - if (system_uses_nmi()) + if (system_uses_nmi() && !static_branch_unlikely(&vgic_v3_cpuif_trap)) kvm->arch.pfr1_nmi = ID_AA64PFR1_EL1_NMI_IMP; kvm_arm_init_hypercalls(kvm); diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 5a6e64670feb..c92fd3aac0ba 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -1580,7 +1580,8 @@ static int set_id_aa64pfr1_el1(struct kvm_vcpu *vcpu, u8 nmi; nmi = cpuid_feature_extract_unsigned_field(val, ID_AA64PFR1_EL1_NMI_SHIFT); - if (nmi > ID_AA64PFR1_EL1_NMI_IMP || (nmi && !system_uses_nmi())) + if (nmi > ID_AA64PFR1_EL1_NMI_IMP || + (nmi && (!system_uses_nmi() || static_branch_unlikely(&vgic_v3_cpuif_trap)))) return -EINVAL; /* We can only differ with NMI, and anything else is an error */ diff --git a/arch/arm64/kvm/vgic/vgic-v3.c b/arch/arm64/kvm/vgic/vgic-v3.c index e82402dada77..69ca111e349d 100644 --- a/arch/arm64/kvm/vgic/vgic-v3.c +++ b/arch/arm64/kvm/vgic/vgic-v3.c @@ -724,6 +724,12 @@ int vgic_v3_probe(const struct gic_kvm_info *info) static_branch_enable(&vgic_v3_cpuif_trap); } + if (info->has_nmi) { + kvm_vgic_global_state.has_nmi = !static_branch_unlikely(&vgic_v3_cpuif_trap); + kvm_info("GICv3 NMI support %s\n", + kvm_vgic_global_state.has_nmi ? "enabled" : "disabled due to trapping"); + } + kvm_vgic_global_state.vctrl_base = NULL; kvm_vgic_global_state.type = VGIC_V3; kvm_vgic_global_state.max_gic_vcpus = VGIC_V3_MAX_CPUS; diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c index a9fda74a198f..7785faa6c7f7 100644 --- a/drivers/irqchip/irq-gic-v3.c +++ b/drivers/irqchip/irq-gic-v3.c @@ -2681,6 +2681,7 @@ static void __init gic_acpi_setup_kvm_info(void) gic_v3_kvm_info.has_v4 = gic_data.rdists.has_vlpis; gic_v3_kvm_info.has_v4_1 = gic_data.rdists.has_rvpeid; + gic_v3_kvm_info.has_nmi = has_v3_3_nmi(); #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS gic_v3_kvm_info.has_vtimer = gic_data.rdists.has_vtimer; #endif diff --git a/include/linux/irqchip/arm-vgic-info.h b/include/linux/irqchip/arm-vgic-info.h index ac362587d2fb..aa2e9eaa9ad9 100644 --- a/include/linux/irqchip/arm-vgic-info.h +++ b/include/linux/irqchip/arm-vgic-info.h @@ -32,6 +32,8 @@ struct gic_kvm_info { bool has_v4; /* rvpeid support */ bool has_v4_1; + /* NMI support */ + bool has_nmi; #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS /* vtimer irqbypass support */ bool has_vtimer; -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Although we do force a trap of the ALLINT system register when FEAT_NMI isn't exposed to the guest, we don't provide any handler. That's bad. Let's fix that. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/include/asm/sysreg.h | 2 ++ arch/arm64/kvm/sys_regs.c | 5 +++++ 2 files changed, 7 insertions(+) diff --git a/arch/arm64/include/asm/sysreg.h b/arch/arm64/include/asm/sysreg.h index 4783bd2eb819..42358b8d678e 100644 --- a/arch/arm64/include/asm/sysreg.h +++ b/arch/arm64/include/asm/sysreg.h @@ -300,6 +300,8 @@ #define SYS_SPSR_EL1 sys_reg(3, 0, 4, 0, 0) #define SYS_ELR_EL1 sys_reg(3, 0, 4, 0, 1) +#define SYS_ALLINT sys_reg(3, 0, 4, 3, 0) + #define SYS_ICC_PMR_EL1 sys_reg(3, 0, 4, 6, 0) #define SYS_AFSR0_EL1 sys_reg(3, 0, 5, 1, 0) diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index c92fd3aac0ba..aea39eab513f 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -1958,6 +1958,9 @@ static bool access_spsr(struct kvm_vcpu *vcpu, * guest... */ static const struct sys_reg_desc sys_reg_descs[] = { + { SYS_DESC(SYS_ALLINT_CLR), undef_access }, + { SYS_DESC(SYS_ALLINT_SET), undef_access }, + { SYS_DESC(SYS_DC_ISW), access_dcsw }, { SYS_DESC(SYS_DC_IGSW), access_dcgsw }, { SYS_DESC(SYS_DC_IGDSW), access_dcgsw }, @@ -2126,6 +2129,8 @@ static const struct sys_reg_desc sys_reg_descs[] = { { SYS_DESC(SYS_SPSR_EL1), access_spsr}, { SYS_DESC(SYS_ELR_EL1), access_elr}, + { SYS_DESC(SYS_ALLINT), undef_access }, + { SYS_DESC(SYS_AFSR0_EL1), access_vm_reg, reset_unknown, AFSR0_EL1 }, { SYS_DESC(SYS_AFSR1_EL1), access_vm_reg, reset_unknown, AFSR1_EL1 }, { SYS_DESC(SYS_ESR_EL1), access_vm_reg, reset_unknown, ESR_EL1 }, -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Even if the host isn't using NMIs, that isn't a reason for preventing the feature being available to guests, Decouple the two and let the capability be available irrespective of the host NMI level of support. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/include/asm/cpufeature.h | 3 ++- arch/arm64/kernel/cpufeature.c | 13 +++++++++---- arch/arm64/kvm/arm.c | 2 +- arch/arm64/kvm/sys_regs.c | 2 +- drivers/irqchip/irq-gic-v3.c | 13 ++++++++++++- 5 files changed, 25 insertions(+), 8 deletions(-) diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h index 396ad4bcd8cf..450124238563 100644 --- a/arch/arm64/include/asm/cpufeature.h +++ b/arch/arm64/include/asm/cpufeature.h @@ -818,7 +818,8 @@ static __always_inline bool system_uses_irq_prio_masking(void) static __always_inline bool system_uses_nmi(void) { return IS_ENABLED(CONFIG_ARM64_NMI) && - cpus_have_const_cap(ARM64_USES_NMI); + cpus_have_const_cap(ARM64_USES_NMI) && + !system_uses_irq_prio_masking(); } static inline bool system_supports_mte(void) diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c index 316c57b19c93..dccf8c3a23bb 100644 --- a/arch/arm64/kernel/cpufeature.c +++ b/arch/arm64/kernel/cpufeature.c @@ -2189,20 +2189,24 @@ static bool has_gic_prio_relaxed_sync(const struct arm64_cpu_capabilities *entry } #endif -#ifdef CONFIG_ARM64_NMI static bool use_nmi(const struct arm64_cpu_capabilities *entry, int scope) { if (!has_cpuid_feature(entry, scope)) return false; /* + * NMI support was not enabled in the kernel, but can still be + * used by guests. Let the world know. + * * Having both real and pseudo NMIs enabled simultaneously is * likely to cause confusion. Since pseudo NMIs must be * enabled with an explicit command line option, if the user * has set that option on a system with real NMIs for some * reason assume they know what they're doing. */ - if (IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI) && enable_pseudo_nmi) { + if (!IS_ENABLED(CONFIG_ARM64_NMI)) + pr_info("CONFIG_ARM64_NMI disabled, using NMIs for guests only\n"); + else if (IS_ENABLED(CONFIG_ARM64_PSEUDO_NMI) && enable_pseudo_nmi) { pr_info("Pseudo NMI enabled, not using architected NMI\n"); return false; } @@ -2210,6 +2214,7 @@ static bool use_nmi(const struct arm64_cpu_capabilities *entry, int scope) return true; } +#ifdef CONFIG_ARM64_NMI static void nmi_enable(const struct arm64_cpu_capabilities *__unused) { /* @@ -2821,7 +2826,6 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .matches = has_cpuid_feature, ARM64_CPUID_FIELDS(ID_AA64MMFR2_EL1, EVT, IMP) }, -#ifdef CONFIG_ARM64_NMI { .desc = "Non-maskable Interrupts present", .capability = ARM64_HAS_NMI, @@ -2843,9 +2847,10 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .field_width = 4, .min_field_value = ID_AA64PFR1_EL1_NMI_IMP, .matches = use_nmi, +#ifdef CONFIG_ARM64_NMI .cpu_enable = nmi_enable, - }, #endif + }, #ifdef CONFIG_ARM64_MPAM { .desc = "Memory Partitioning And Monitoring", diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c index 9c6b4accb10b..c71ef028dde6 100644 --- a/arch/arm64/kvm/arm.c +++ b/arch/arm64/kvm/arm.c @@ -192,7 +192,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) /* The maximum number of VCPUs is limited by the host's GIC model */ kvm->max_vcpus = kvm_arm_default_max_vcpus(); - if (system_uses_nmi() && !static_branch_unlikely(&vgic_v3_cpuif_trap)) + if (cpus_have_const_cap(ARM64_HAS_NMI) && !static_branch_unlikely(&vgic_v3_cpuif_trap)) kvm->arch.pfr1_nmi = ID_AA64PFR1_EL1_NMI_IMP; kvm_arm_init_hypercalls(kvm); diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index aea39eab513f..2c05984415a4 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -1581,7 +1581,7 @@ static int set_id_aa64pfr1_el1(struct kvm_vcpu *vcpu, nmi = cpuid_feature_extract_unsigned_field(val, ID_AA64PFR1_EL1_NMI_SHIFT); if (nmi > ID_AA64PFR1_EL1_NMI_IMP || - (nmi && (!system_uses_nmi() || static_branch_unlikely(&vgic_v3_cpuif_trap)))) + (nmi && (!cpus_have_const_cap(ARM64_HAS_NMI) || static_branch_unlikely(&vgic_v3_cpuif_trap)))) return -EINVAL; /* We can only differ with NMI, and anything else is an error */ diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c index 7785faa6c7f7..8bb42ac8fb2a 100644 --- a/drivers/irqchip/irq-gic-v3.c +++ b/drivers/irqchip/irq-gic-v3.c @@ -148,11 +148,21 @@ static inline bool has_v3_3_nmi(void) { return gic_data.has_nmi && system_uses_nmi(); } + +static bool system_is_nmi_capable(void) +{ + return gic_data.has_nmi && cpus_have_const_cap(ARM64_HAS_NMI); +} #else static inline bool has_v3_3_nmi(void) { return false; } + +static bool system_is_nmi_capable(void) +{ + return false; +} #endif #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS @@ -2330,6 +2340,7 @@ static void __init gic_of_setup_kvm_info(struct device_node *node) gic_v3_kvm_info.has_v4 = gic_data.rdists.has_vlpis; gic_v3_kvm_info.has_v4_1 = gic_data.rdists.has_rvpeid; + gic_v3_kvm_info.has_nmi = system_is_nmi_capable(); #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS gic_v3_kvm_info.has_vtimer = gic_data.rdists.has_vtimer; #endif @@ -2681,7 +2692,7 @@ static void __init gic_acpi_setup_kvm_info(void) gic_v3_kvm_info.has_v4 = gic_data.rdists.has_vlpis; gic_v3_kvm_info.has_v4_1 = gic_data.rdists.has_rvpeid; - gic_v3_kvm_info.has_nmi = has_v3_3_nmi(); + gic_v3_kvm_info.has_nmi = system_is_nmi_capable(); #ifdef CONFIG_VIRT_VTIMER_IRQ_BYPASS gic_v3_kvm_info.has_vtimer = gic_data.rdists.has_vtimer; #endif -- 2.30.0
From: Marc Zyngier <maz@kernel.org> virt inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I97WGU Reference: https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit... ----------------------------------------------------------- Even if we do not plan to deal with the GICv3.3 NMI feature in our in-kernel emulation of the CPU interface for terminally broken systems, the actual system register may still exist and raise its ugly head. Hit it with an UNDEF-sized hammer. Signed-off-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com> Signed-off-by: caijian <caijian11@h-partners.com> --- arch/arm64/kvm/hyp/vgic-v3-sr.c | 3 +++ arch/arm64/kvm/sys_regs.c | 1 + 2 files changed, 4 insertions(+) diff --git a/arch/arm64/kvm/hyp/vgic-v3-sr.c b/arch/arm64/kvm/hyp/vgic-v3-sr.c index 05955e369c51..7a9e132f68db 100644 --- a/arch/arm64/kvm/hyp/vgic-v3-sr.c +++ b/arch/arm64/kvm/hyp/vgic-v3-sr.c @@ -1048,6 +1048,9 @@ int __vgic_v3_perform_cpuif_access(struct kvm_vcpu *vcpu) return 0; fn = __vgic_v3_read_iar; break; + case SYS_ICC_NMIAR1_EL1: + /* Here's an UNDEF for you */ + return 0; case SYS_ICC_EOIR0_EL1: case SYS_ICC_EOIR1_EL1: if (unlikely(is_read)) diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c index 2c05984415a4..dc8663759b0e 100644 --- a/arch/arm64/kvm/sys_regs.c +++ b/arch/arm64/kvm/sys_regs.c @@ -2189,6 +2189,7 @@ static const struct sys_reg_desc sys_reg_descs[] = { { SYS_DESC(SYS_ICC_IAR0_EL1), write_to_read_only }, { SYS_DESC(SYS_ICC_EOIR0_EL1), read_from_write_only }, { SYS_DESC(SYS_ICC_HPPIR0_EL1), write_to_read_only }, + { SYS_DESC(SYS_ICC_NMIAR1_EL1), undef_access }, { SYS_DESC(SYS_ICC_DIR_EL1), read_from_write_only }, { SYS_DESC(SYS_ICC_RPR_EL1), write_to_read_only }, { SYS_DESC(SYS_ICC_SGI1R_EL1), access_gic_sgi }, -- 2.30.0
反馈： 您发送到kernel@openeuler.org的补丁/补丁集，已成功转换为PR！ PR链接地址： https://gitee.com/openeuler/kernel/pulls/5815 邮件列表地址：https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/Z... FeedBack: The patch(es) which you have sent to kernel@openeuler.org mailing list has been converted to a pull request successfully! Pull request link: https://gitee.com/openeuler/kernel/pulls/5815 Mailing list address: https://mailweb.openeuler.org/hyperkitty/list/kernel@openeuler.org/message/Z...
participants (2)
-
chenxiang -
patchwork bot