From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-9.0 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_PASS,URIBL_BLOCKED, USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id A2AEEC43387 for ; Mon, 7 Jan 2019 22:35:20 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 704A72173C for ; Mon, 7 Jan 2019 22:35:20 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727145AbfAGWfJ (ORCPT ); Mon, 7 Jan 2019 17:35:09 -0500 Received: from mga17.intel.com ([192.55.52.151]:34117 "EHLO mga17.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726643AbfAGWfI (ORCPT ); Mon, 7 Jan 2019 17:35:08 -0500 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga001.jf.intel.com ([10.7.209.18]) by fmsmga107.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 07 Jan 2019 14:35:07 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.56,451,1539673200"; d="scan'208";a="124985856" Received: from otc-lr-04.jf.intel.com ([10.54.39.135]) by orsmga001.jf.intel.com with ESMTP; 07 Jan 2019 14:35:07 -0800 From: kan.liang@linux.intel.com To: x86@kernel.org, linux-kernel@vger.kernel.org, tglx@linutronix.de, bp@alien8.de, peterz@infradead.org, mingo@redhat.com Cc: ak@linux.intel.com, eranian@google.com Subject: [PATCH V5 2/4] perf/x86/kvm: Avoid unnecessary work in guest filtering Date: Mon, 7 Jan 2019 14:34:23 -0800 Message-Id: <1546900465-5121-2-git-send-email-kan.liang@linux.intel.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1546900465-5121-1-git-send-email-kan.liang@linux.intel.com> References: <1546900465-5121-1-git-send-email-kan.liang@linux.intel.com> Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Andi Kleen KVM added a workaround for PEBS events leaking into guests with commit 26a4f3c08de4 ("perf/x86: disable PEBS on a guest entry.") This uses the VT entry/exit list to add an extra disable of the PEBS_ENABLE MSR. Intel also added a fix for this issue to microcode updates on Haswell/Broadwell/Skylake. It turns out using the MSR entry/exit list makes VM exits significantly slower. The list is only needed for disabling PEBS, because the GLOBAL_CTRL change gets optimized by KVM into changing the VMCS. Check for the microcode updates that have the microcode fix for leaking PEBS, and disable the extra entry/exit list entry for PEBS_ENABLE. In addition we always clear the GLOBAL_CTRL for the PEBS counter while running in the guest, which is enough to make them never fire at the wrong side of the host/guest transition. We see significantly reduced overhead for VM exits with the filtering active with the patch from 8% to 4%. Signed-off-by: Andi Kleen --- Changes since V4: - Use new name x86_cpu_has_min_microcode_rev() and INTEL_CHECK_UCODE arch/x86/events/intel/core.c | 80 +++++++++++++++++++++++++++++++++++++++----- arch/x86/events/perf_event.h | 3 +- 2 files changed, 73 insertions(+), 10 deletions(-) diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c index ecc3e34..587d83e 100644 --- a/arch/x86/events/intel/core.c +++ b/arch/x86/events/intel/core.c @@ -18,6 +18,7 @@ #include #include #include +#include #include "../perf_event.h" @@ -3206,16 +3207,27 @@ static struct perf_guest_switch_msr *intel_guest_get_msrs(int *nr) arr[0].msr = MSR_CORE_PERF_GLOBAL_CTRL; arr[0].host = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_guest_mask; arr[0].guest = x86_pmu.intel_ctrl & ~cpuc->intel_ctrl_host_mask; - /* - * If PMU counter has PEBS enabled it is not enough to disable counter - * on a guest entry since PEBS memory write can overshoot guest entry - * and corrupt guest memory. Disabling PEBS solves the problem. - */ - arr[1].msr = MSR_IA32_PEBS_ENABLE; - arr[1].host = cpuc->pebs_enabled; - arr[1].guest = 0; + if (x86_pmu.flags & PMU_FL_PEBS_ALL) + arr[0].guest &= ~cpuc->pebs_enabled; + else + arr[0].guest &= ~(cpuc->pebs_enabled & PEBS_COUNTER_MASK); + *nr = 1; + + if (!x86_pmu.pebs_isolated) { + /* + * If PMU counter has PEBS enabled it is not enough to + * disable counter on a guest entry since PEBS memory + * write can overshoot guest entry and corrupt guest + * memory. Disabling PEBS solves the problem. + * + * Don't do this if the CPU already enforces it. + */ + arr[1].msr = MSR_IA32_PEBS_ENABLE; + arr[1].host = cpuc->pebs_enabled; + arr[1].guest = 0; + *nr = 2; + } - *nr = 2; return arr; } @@ -3733,6 +3745,45 @@ static __init void intel_clovertown_quirk(void) x86_pmu.pebs_constraints = NULL; } +static const struct x86_cpu_check isolation_ucodes[] = { + INTEL_CHECK_UCODE(INTEL_FAM6_HASWELL_CORE, 3, 0x0000001f), + INTEL_CHECK_UCODE(INTEL_FAM6_HASWELL_ULT, 1, 0x0000001e), + INTEL_CHECK_UCODE(INTEL_FAM6_HASWELL_GT3E, 1, 0x00000015), + INTEL_CHECK_UCODE(INTEL_FAM6_HASWELL_X, 2, 0x00000037), + INTEL_CHECK_UCODE(INTEL_FAM6_HASWELL_X, 4, 0x0000000a), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_CORE, 4, 0x00000023), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_GT3E, 1, 0x00000014), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_XEON_D, 2, 0x00000010), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_XEON_D, 3, 0x07000009), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_XEON_D, 4, 0x0f000009), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_XEON_D, 5, 0x0e000002), + INTEL_CHECK_UCODE(INTEL_FAM6_BROADWELL_X, 2, 0x0b000014), + INTEL_CHECK_UCODE(INTEL_FAM6_SKYLAKE_X, 3, 0x00000021), + INTEL_CHECK_UCODE(INTEL_FAM6_SKYLAKE_X, 4, 0x00000000), + INTEL_CHECK_UCODE(INTEL_FAM6_SKYLAKE_MOBILE, 3, 0x0000007c), + INTEL_CHECK_UCODE(INTEL_FAM6_SKYLAKE_DESKTOP, 3, 0x0000007c), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_DESKTOP, 9, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_MOBILE, 9, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_MOBILE, 10, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_MOBILE, 11, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_MOBILE, 12, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_DESKTOP, 10, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_DESKTOP, 11, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_DESKTOP, 12, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_KABYLAKE_DESKTOP, 13, 0x0000004e), + INTEL_CHECK_UCODE(INTEL_FAM6_CANNONLAKE_MOBILE, 3, 0x00000000), + {} +}; + +static void intel_check_isolation(void) +{ + if (!x86_cpu_has_min_microcode_rev(isolation_ucodes)) { + x86_pmu.pebs_isolated = 0; + return; + } + x86_pmu.pebs_isolated = 1; +} + static int intel_snb_pebs_broken(int cpu) { u32 rev = UINT_MAX; /* default to broken for unknown models */ @@ -3757,6 +3808,8 @@ static void intel_snb_check_microcode(void) int pebs_broken = 0; int cpu; + intel_check_isolation(); + for_each_online_cpu(cpu) { if ((pebs_broken = intel_snb_pebs_broken(cpu))) break; @@ -3838,6 +3891,12 @@ static __init void intel_sandybridge_quirk(void) cpus_read_unlock(); } +static __init void intel_isolation_quirk(void) +{ + x86_pmu.check_microcode = intel_check_isolation; + intel_check_isolation(); +} + static const struct { int id; char *name; } intel_arch_events_map[] __initconst = { { PERF_COUNT_HW_CPU_CYCLES, "cpu cycles" }, { PERF_COUNT_HW_INSTRUCTIONS, "instructions" }, @@ -4424,6 +4483,7 @@ __init int intel_pmu_init(void) case INTEL_FAM6_HASWELL_X: case INTEL_FAM6_HASWELL_ULT: case INTEL_FAM6_HASWELL_GT3E: + x86_add_quirk(intel_isolation_quirk); x86_add_quirk(intel_ht_bug); x86_pmu.late_ack = true; memcpy(hw_cache_event_ids, hsw_hw_cache_event_ids, sizeof(hw_cache_event_ids)); @@ -4456,6 +4516,7 @@ __init int intel_pmu_init(void) case INTEL_FAM6_BROADWELL_XEON_D: case INTEL_FAM6_BROADWELL_GT3E: case INTEL_FAM6_BROADWELL_X: + x86_add_quirk(intel_isolation_quirk); x86_pmu.late_ack = true; memcpy(hw_cache_event_ids, hsw_hw_cache_event_ids, sizeof(hw_cache_event_ids)); memcpy(hw_cache_extra_regs, hsw_hw_cache_extra_regs, sizeof(hw_cache_extra_regs)); @@ -4518,6 +4579,7 @@ __init int intel_pmu_init(void) case INTEL_FAM6_SKYLAKE_X: case INTEL_FAM6_KABYLAKE_MOBILE: case INTEL_FAM6_KABYLAKE_DESKTOP: + x86_add_quirk(intel_isolation_quirk); x86_pmu.late_ack = true; memcpy(hw_cache_event_ids, skl_hw_cache_event_ids, sizeof(hw_cache_event_ids)); memcpy(hw_cache_extra_regs, skl_hw_cache_extra_regs, sizeof(hw_cache_extra_regs)); diff --git a/arch/x86/events/perf_event.h b/arch/x86/events/perf_event.h index 78d7b70..c2d5a51 100644 --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -607,7 +607,8 @@ struct x86_pmu { pebs_active :1, pebs_broken :1, pebs_prec_dist :1, - pebs_no_tlb :1; + pebs_no_tlb :1, + pebs_isolated :1; int pebs_record_size; int pebs_buffer_size; void (*drain_pebs)(struct pt_regs *regs); -- 2.7.4