| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <3b41286f-d2b0-5fdf-88ef-1e63f302f9c8@arm.com>
Date: Thu, 5 Oct 2023 19:05:01 +0100
From: Suzuki K Poulose <suzuki.poulose@....com>
To: James Clark <james.clark@....com>, coresight@...ts.linaro.org,
linux-arm-kernel@...ts.infradead.org, kvmarm@...ts.linux.dev,
broonie@...nel.org, maz@...nel.org
Cc: Oliver Upton <oliver.upton@...ux.dev>,
James Morse <james.morse@....com>,
Zenghui Yu <yuzenghui@...wei.com>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Mike Leach <mike.leach@...aro.org>,
Leo Yan <leo.yan@...aro.org>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Anshuman Khandual <anshuman.khandual@....com>,
Rob Herring <robh@...nel.org>,
Jintack Lim <jintack.lim@...aro.org>,
Akihiko Odaki <akihiko.odaki@...nix.com>,
Fuad Tabba <tabba@...gle.com>, Joey Gouly <joey.gouly@....com>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 5/6] arm64: KVM: Write TRFCR value on guest switch with
nVHE
On 05/10/2023 13:57, James Clark wrote:
> The guest value for TRFCR requested by the Coresight driver is saved
> in sysregs[TRFCR_EL1]. On guest switch this value needs to be written to
> the register. Currently TRFCR is only modified when we want to disable
> trace completely in guests due to an issue with TRBE. Expand the
> __debug_save_trace() function to always write to the register if a
> different value for guests is required, but also keep the existing TRBE
> disable behavior if that's required.
>
> The TRFCR restore function remains functionally the same, except a value
> of 0 doesn't mean "don't restore" anymore. Now that we save both guest
> and host values the register is restored any time the guest and host
> values differ.
>
> Signed-off-by: James Clark <james.clark@....com>
> ---
> arch/arm64/include/asm/kvm_hyp.h | 6 ++-
> arch/arm64/kvm/debug.c | 13 +++++-
> arch/arm64/kvm/hyp/nvhe/debug-sr.c | 63 ++++++++++++++++++------------
> arch/arm64/kvm/hyp/nvhe/switch.c | 4 +-
> 4 files changed, 57 insertions(+), 29 deletions(-)
>
> diff --git a/arch/arm64/include/asm/kvm_hyp.h b/arch/arm64/include/asm/kvm_hyp.h
> index 37e238f526d7..0383fd3d60b5 100644
> --- a/arch/arm64/include/asm/kvm_hyp.h
> +++ b/arch/arm64/include/asm/kvm_hyp.h
> @@ -103,8 +103,10 @@ void __debug_switch_to_guest(struct kvm_vcpu *vcpu);
> void __debug_switch_to_host(struct kvm_vcpu *vcpu);
>
> #ifdef __KVM_NVHE_HYPERVISOR__
> -void __debug_save_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt);
> -void __debug_restore_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt);
> +void __debug_save_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt);
> +void __debug_restore_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt);
> #endif
>
> void __fpsimd_save_state(struct user_fpsimd_state *fp_regs);
> diff --git a/arch/arm64/kvm/debug.c b/arch/arm64/kvm/debug.c
> index 19e722359154..d949dd354464 100644
> --- a/arch/arm64/kvm/debug.c
> +++ b/arch/arm64/kvm/debug.c
> @@ -337,10 +337,21 @@ void kvm_arch_vcpu_load_debug_state_flags(struct kvm_vcpu *vcpu)
> !(read_sysreg_s(SYS_PMBIDR_EL1) & BIT(PMBIDR_EL1_P_SHIFT)))
> vcpu_set_flag(vcpu, DEBUG_STATE_SAVE_SPE);
>
> - /* Check if we have TRBE implemented and available at the host */
> + /*
> + * Check if we have TRBE implemented and available at the host. If it's
> + * in use at the time of guest switch it will need to be disabled and
> + * then restored.
> + */
> if (cpuid_feature_extract_unsigned_field(dfr0, ID_AA64DFR0_EL1_TraceBuffer_SHIFT) &&
> !(read_sysreg_s(SYS_TRBIDR_EL1) & TRBIDR_EL1_P))
> vcpu_set_flag(vcpu, DEBUG_STATE_SAVE_TRFCR);
As per A3.1 Armv9-A architecture extensions (DDI 0487J.a), FEAT_TRBE
mandates FEAT_TRF. So, we could check FEAT_TRF and if we have a hit,
skip the TRBE checks. But, having read the code below, it looks like
we need separate flags for TRFCR and TRBE.
> + /*
> + * Also save TRFCR on nVHE if FEAT_TRF (TraceFilt) exists. This will be
> + * done in cases where use of TRBE doesn't completely disable trace and
> + * handles the exclude_host/exclude_guest rules of the trace session.
> + */
> + if (cpuid_feature_extract_unsigned_field(dfr0, ID_AA64DFR0_EL1_TraceFilt_SHIFT))
> + vcpu_set_flag(vcpu, DEBUG_STATE_SAVE_TRFCR);
> }
>
> void kvm_arch_vcpu_put_debug_state_flags(struct kvm_vcpu *vcpu)
> diff --git a/arch/arm64/kvm/hyp/nvhe/debug-sr.c b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> index 128a57dddabf..c6252029c277 100644
> --- a/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> +++ b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> @@ -51,42 +51,56 @@ static void __debug_restore_spe(struct kvm_cpu_context *host_ctxt)
> write_sysreg_s(ctxt_sys_reg(host_ctxt, PMSCR_EL1), SYS_PMSCR_EL1);
> }
>
> -static void __debug_save_trace(struct kvm_cpu_context *host_ctxt)
> +/*
> + * Save TRFCR and disable trace completely if TRBE is being used, otherwise
> + * apply required guest TRFCR value.
> + */
> +static void __debug_save_trace(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt)
> {
> - ctxt_sys_reg(host_ctxt, TRFCR_EL1) = 0;
> + ctxt_sys_reg(host_ctxt, TRFCR_EL1) = read_sysreg_s(SYS_TRFCR_EL1);
>
> /* Check if the TRBE is enabled */
> - if (!(read_sysreg_s(SYS_TRBLIMITR_EL1) & TRBLIMITR_EL1_E))
> - return;
> - /*
> - * Prohibit trace generation while we are in guest.
> - * Since access to TRFCR_EL1 is trapped, the guest can't
> - * modify the filtering set by the host.
> - */
> - ctxt_sys_reg(host_ctxt, TRFCR_EL1) = read_sysreg_s(SYS_TRFCR_EL1);
> - write_sysreg_s(0, SYS_TRFCR_EL1);
> - isb();
> - /* Drain the trace buffer to memory */
> - tsb_csync();
> + if (read_sysreg_s(SYS_TRBLIMITR_EL1) & TRBLIMITR_EL1_E) {
This is problematic. At this point, we are not sure if TRBE is available
or not (e.g. we could be on a v8.4 CPU or a v9.0 with TRBE disabled by
higher EL). May be we need to add a separate flag to indicate the
presence of TRBE.
Suzuki
> + /*
> + * Prohibit trace generation while we are in guest. Since access
> + * to TRFCR_EL1 is trapped, the guest can't modify the filtering
> + * set by the host.
> + */
> + ctxt_sys_reg(guest_ctxt, TRFCR_EL1) = 0;
> + write_sysreg_s(0, SYS_TRFCR_EL1);
> + isb();
> + /* Drain the trace buffer to memory */
> + tsb_csync();
> + } else {
> + /*
> + * Not using TRBE, so guest trace works. Apply the guest filters
> + * provided by the Coresight driver, if different.
> + */
> + if (ctxt_sys_reg(host_ctxt, TRFCR_EL1) !=
> + ctxt_sys_reg(guest_ctxt, TRFCR_EL1))
> + write_sysreg_s(ctxt_sys_reg(guest_ctxt, TRFCR_EL1),
> + SYS_TRFCR_EL1);
> + }
> }
>
> -static void __debug_restore_trace(struct kvm_cpu_context *host_ctxt)
> +static void __debug_restore_trace(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt)
> {
> - if (!ctxt_sys_reg(host_ctxt, TRFCR_EL1))
> - return;
> -
> /* Restore trace filter controls */
> - write_sysreg_s(ctxt_sys_reg(host_ctxt, TRFCR_EL1), SYS_TRFCR_EL1);
> + if (ctxt_sys_reg(host_ctxt, TRFCR_EL1) != ctxt_sys_reg(guest_ctxt, TRFCR_EL1))
> + write_sysreg_s(ctxt_sys_reg(host_ctxt, TRFCR_EL1), SYS_TRFCR_EL1);
> }
>
> -void __debug_save_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt)
> +void __debug_save_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt)
> {
> /* Disable and flush SPE data generation */
> if (vcpu_get_flag(host_ctxt->__hyp_running_vcpu, DEBUG_STATE_SAVE_SPE))
> __debug_save_spe(host_ctxt);
> - /* Disable and flush Self-Hosted Trace generation */
> +
> if (vcpu_get_flag(host_ctxt->__hyp_running_vcpu, DEBUG_STATE_SAVE_TRFCR))
> - __debug_save_trace(host_ctxt);
> + __debug_save_trace(host_ctxt, guest_ctxt);
> }
>
> void __debug_switch_to_guest(struct kvm_vcpu *vcpu)
> @@ -94,12 +108,13 @@ void __debug_switch_to_guest(struct kvm_vcpu *vcpu)
> __debug_switch_to_guest_common(vcpu);
> }
>
> -void __debug_restore_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt)
> +void __debug_restore_host_buffers_nvhe(struct kvm_cpu_context *host_ctxt,
> + struct kvm_cpu_context *guest_ctxt)
> {
> if (vcpu_get_flag(host_ctxt->__hyp_running_vcpu, DEBUG_STATE_SAVE_SPE))
> __debug_restore_spe(host_ctxt);
> if (vcpu_get_flag(host_ctxt->__hyp_running_vcpu, DEBUG_STATE_SAVE_TRFCR))
> - __debug_restore_trace(host_ctxt);
> + __debug_restore_trace(host_ctxt, guest_ctxt);
> }
>
> void __debug_switch_to_host(struct kvm_vcpu *vcpu)
> diff --git a/arch/arm64/kvm/hyp/nvhe/switch.c b/arch/arm64/kvm/hyp/nvhe/switch.c
> index c8f15e4dab19..55207ec31bd3 100644
> --- a/arch/arm64/kvm/hyp/nvhe/switch.c
> +++ b/arch/arm64/kvm/hyp/nvhe/switch.c
> @@ -276,7 +276,7 @@ int __kvm_vcpu_run(struct kvm_vcpu *vcpu)
> * translation regime to EL2 (via MDCR_EL2_E2PB == 0) and
> * before we load guest Stage1.
> */
> - __debug_save_host_buffers_nvhe(host_ctxt);
> + __debug_save_host_buffers_nvhe(host_ctxt, guest_ctxt);
>
> /*
> * We're about to restore some new MMU state. Make sure
> @@ -343,7 +343,7 @@ int __kvm_vcpu_run(struct kvm_vcpu *vcpu)
> * This must come after restoring the host sysregs, since a non-VHE
> * system may enable SPE here and make use of the TTBRs.
> */
> - __debug_restore_host_buffers_nvhe(host_ctxt);
> + __debug_restore_host_buffers_nvhe(host_ctxt, guest_ctxt);
>
> if (pmu_switch_needed)
> __pmu_switch_to_host(vcpu);
Powered by blists - more mailing lists