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 Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 997E2ECAAD4 for ; Mon, 29 Aug 2022 12:57:01 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230215AbiH2M47 (ORCPT ); Mon, 29 Aug 2022 08:56:59 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:56726 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229965AbiH2M42 (ORCPT ); Mon, 29 Aug 2022 08:56:28 -0400 Received: from mail-ej1-x64a.google.com (mail-ej1-x64a.google.com [IPv6:2a00:1450:4864:20::64a]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id DD42B76950 for ; Mon, 29 Aug 2022 05:48:10 -0700 (PDT) Received: by mail-ej1-x64a.google.com with SMTP id gs35-20020a1709072d2300b00730e14fd76eso2266620ejc.15 for ; Mon, 29 Aug 2022 05:48:10 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20210112; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:from:to:cc; bh=Qud8pWzf3g4FVwsYBG/lvnDK9jpEJXDArWePZHukaec=; b=QxY/7D4H+6Wm8ODbCoYIX4jcTSNanjtdoc1j3Iz7xu5mMl6TUJICrlbXaUNTWvcEjy 4c4cUTvwCtdAs5ab7bucTDOk1tTEHaWIDXORtNXuC3re0PrHVUefmShgyFj5B7b5OgZZ t4Yu15YoJLhe0U4wAbtL+f1gdUTACvONv/nwoJ486+p9lcXDNJ/FXaDMKxXkqZ/DcNnQ bVit8hcj5YbEy8u3iVI/NE3APWnlTlxjyJmpCk3g5re4516DamYP1dTQXUKRZojE3wJk ujWYpvxhsiZO2EWxUD8bVayH6zQpzOCVZkKsYEHOuPbJxM2l99CpGmiBI40NQ85R4yS8 pTRQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=cc:to:from:subject:message-id:references:mime-version:in-reply-to :date:x-gm-message-state:from:to:cc; bh=Qud8pWzf3g4FVwsYBG/lvnDK9jpEJXDArWePZHukaec=; b=fB67n8gm5ZWdgRLzPgXYarDYw/mm5XDfMBilk0Z6wVYpuArpyxELGpUM1t7+sfCwOy u1BZXZeMDTIkm+c5L6prSjuKBJgQw7gROCT45n/YuhaW2ReMC7ANZnKuukXVa+4XaQOO FNj6MF/QiBgxURvMO91RC6nkovwthBR4u+jLN4J9Rmq5MGG+jRoNzIQcCHEqJoOe2AyR JFNsR3Tlmbp3Z/Rm4loqdZ8K3f7QmJCl6ITy2yP1Jk8MtTGo928syPAx5YJLPokxbVz1 u3a45JQLIHSG6LjW8x5UrF81m8F1CRYrcENEp0aVavBLz6BEIc6jMRq5JtERr80QDfAx 46Tw== X-Gm-Message-State: ACgBeo2bJ5Ps2VD6i1o5bUaJ23SWGoAWe4POZBfDxQKEkcrdSKhpb2dF FJ5hKKIjtxBON7jSeLGmdj/jp3elPw== X-Google-Smtp-Source: AA6agR5PUlzv5XQworgNqpqqg/1h1X/W0LAjuxF8N7CtRrxKcV4spvajGtYVvIbKkvOHFON/rN0rja2wcQ== X-Received: from elver.muc.corp.google.com ([2a00:79e0:9c:201:196d:4fc7:fa9c:62e3]) (user=elver job=sendgmr) by 2002:a17:906:cc13:b0:73d:d22d:63cd with SMTP id ml19-20020a170906cc1300b0073dd22d63cdmr13213673ejb.741.1661777289495; Mon, 29 Aug 2022 05:48:09 -0700 (PDT) Date: Mon, 29 Aug 2022 14:47:09 +0200 In-Reply-To: <20220829124719.675715-1-elver@google.com> Mime-Version: 1.0 References: <20220829124719.675715-1-elver@google.com> X-Mailer: git-send-email 2.37.2.672.g94769d06f0-goog Message-ID: <20220829124719.675715-5-elver@google.com> Subject: [PATCH v4 04/14] perf/hw_breakpoint: Optimize list of per-task breakpoints From: Marco Elver To: elver@google.com, Peter Zijlstra , Frederic Weisbecker , Ingo Molnar Cc: Thomas Gleixner , Arnaldo Carvalho de Melo , Mark Rutland , Alexander Shishkin , Jiri Olsa , Namhyung Kim , Dmitry Vyukov , Michael Ellerman , linuxppc-dev@lists.ozlabs.org, linux-perf-users@vger.kernel.org, x86@kernel.org, linux-sh@vger.kernel.org, kasan-dev@googlegroups.com, linux-kernel@vger.kernel.org, Ian Rogers Content-Type: text/plain; charset="UTF-8" Precedence: bulk List-ID: X-Mailing-List: linux-perf-users@vger.kernel.org On a machine with 256 CPUs, running the recently added perf breakpoint benchmark results in: | $> perf bench -r 30 breakpoint thread -b 4 -p 64 -t 64 | # Running 'breakpoint/thread' benchmark: | # Created/joined 30 threads with 4 breakpoints and 64 parallelism | Total time: 236.418 [sec] | | 123134.794271 usecs/op | 7880626.833333 usecs/op/cpu The benchmark tests inherited breakpoint perf events across many threads. Looking at a perf profile, we can see that the majority of the time is spent in various hw_breakpoint.c functions, which execute within the 'nr_bp_mutex' critical sections which then results in contention on that mutex as well: 37.27% [kernel] [k] osq_lock 34.92% [kernel] [k] mutex_spin_on_owner 12.15% [kernel] [k] toggle_bp_slot 11.90% [kernel] [k] __reserve_bp_slot The culprit here is task_bp_pinned(), which has a runtime complexity of O(#tasks) due to storing all task breakpoints in the same list and iterating through that list looking for a matching task. Clearly, this does not scale to thousands of tasks. Instead, make use of the "rhashtable" variant "rhltable" which stores multiple items with the same key in a list. This results in average runtime complexity of O(1) for task_bp_pinned(). With the optimization, the benchmark shows: | $> perf bench -r 30 breakpoint thread -b 4 -p 64 -t 64 | # Running 'breakpoint/thread' benchmark: | # Created/joined 30 threads with 4 breakpoints and 64 parallelism | Total time: 0.208 [sec] | | 108.422396 usecs/op | 6939.033333 usecs/op/cpu On this particular setup that's a speedup of ~1135x. While one option would be to make task_struct a breakpoint list node, this would only further bloat task_struct for infrequently used data. Furthermore, after all optimizations in this series, there's no evidence it would result in better performance: later optimizations make the time spent looking up entries in the hash table negligible (we'll reach the theoretical ideal performance i.e. no constraints). Signed-off-by: Marco Elver Reviewed-by: Dmitry Vyukov Acked-by: Ian Rogers --- v2: * Commit message tweaks. --- include/linux/perf_event.h | 3 +- kernel/events/hw_breakpoint.c | 56 ++++++++++++++++++++++------------- 2 files changed, 37 insertions(+), 22 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index ee8b9ecdc03b..a784e055002e 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -36,6 +36,7 @@ struct perf_guest_info_callbacks { }; #ifdef CONFIG_HAVE_HW_BREAKPOINT +#include #include #endif @@ -178,7 +179,7 @@ struct hw_perf_event { * creation and event initalization. */ struct arch_hw_breakpoint info; - struct list_head bp_list; + struct rhlist_head bp_list; }; #endif struct { /* amd_iommu */ diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c index 6076c6346291..6d09edc80d19 100644 --- a/kernel/events/hw_breakpoint.c +++ b/kernel/events/hw_breakpoint.c @@ -26,10 +26,10 @@ #include #include #include -#include #include #include #include +#include #include #include @@ -54,7 +54,13 @@ static struct bp_cpuinfo *get_bp_info(int cpu, enum bp_type_idx type) } /* Keep track of the breakpoints attached to tasks */ -static LIST_HEAD(bp_task_head); +static struct rhltable task_bps_ht; +static const struct rhashtable_params task_bps_ht_params = { + .head_offset = offsetof(struct hw_perf_event, bp_list), + .key_offset = offsetof(struct hw_perf_event, target), + .key_len = sizeof_field(struct hw_perf_event, target), + .automatic_shrinking = true, +}; static int constraints_initialized; @@ -103,17 +109,23 @@ static unsigned int max_task_bp_pinned(int cpu, enum bp_type_idx type) */ static int task_bp_pinned(int cpu, struct perf_event *bp, enum bp_type_idx type) { - struct task_struct *tsk = bp->hw.target; + struct rhlist_head *head, *pos; struct perf_event *iter; int count = 0; - list_for_each_entry(iter, &bp_task_head, hw.bp_list) { - if (iter->hw.target == tsk && - find_slot_idx(iter->attr.bp_type) == type && + rcu_read_lock(); + head = rhltable_lookup(&task_bps_ht, &bp->hw.target, task_bps_ht_params); + if (!head) + goto out; + + rhl_for_each_entry_rcu(iter, pos, head, hw.bp_list) { + if (find_slot_idx(iter->attr.bp_type) == type && (iter->cpu < 0 || cpu == iter->cpu)) count += hw_breakpoint_weight(iter); } +out: + rcu_read_unlock(); return count; } @@ -186,7 +198,7 @@ static void toggle_bp_task_slot(struct perf_event *bp, int cpu, /* * Add/remove the given breakpoint in our constraint table */ -static void +static int toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type, int weight) { @@ -199,7 +211,7 @@ toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type, /* Pinned counter cpu profiling */ if (!bp->hw.target) { get_bp_info(bp->cpu, type)->cpu_pinned += weight; - return; + return 0; } /* Pinned counter task profiling */ @@ -207,9 +219,9 @@ toggle_bp_slot(struct perf_event *bp, bool enable, enum bp_type_idx type, toggle_bp_task_slot(bp, cpu, type, weight); if (enable) - list_add_tail(&bp->hw.bp_list, &bp_task_head); + return rhltable_insert(&task_bps_ht, &bp->hw.bp_list, task_bps_ht_params); else - list_del(&bp->hw.bp_list); + return rhltable_remove(&task_bps_ht, &bp->hw.bp_list, task_bps_ht_params); } __weak int arch_reserve_bp_slot(struct perf_event *bp) @@ -307,9 +319,7 @@ static int __reserve_bp_slot(struct perf_event *bp, u64 bp_type) if (ret) return ret; - toggle_bp_slot(bp, true, type, weight); - - return 0; + return toggle_bp_slot(bp, true, type, weight); } int reserve_bp_slot(struct perf_event *bp) @@ -334,7 +344,7 @@ static void __release_bp_slot(struct perf_event *bp, u64 bp_type) type = find_slot_idx(bp_type); weight = hw_breakpoint_weight(bp); - toggle_bp_slot(bp, false, type, weight); + WARN_ON(toggle_bp_slot(bp, false, type, weight)); } void release_bp_slot(struct perf_event *bp) @@ -707,7 +717,7 @@ static struct pmu perf_breakpoint = { int __init init_hw_breakpoint(void) { int cpu, err_cpu; - int i; + int i, ret; for (i = 0; i < TYPE_MAX; i++) nr_slots[i] = hw_breakpoint_slots(i); @@ -718,18 +728,24 @@ int __init init_hw_breakpoint(void) info->tsk_pinned = kcalloc(nr_slots[i], sizeof(int), GFP_KERNEL); - if (!info->tsk_pinned) - goto err_alloc; + if (!info->tsk_pinned) { + ret = -ENOMEM; + goto err; + } } } + ret = rhltable_init(&task_bps_ht, &task_bps_ht_params); + if (ret) + goto err; + constraints_initialized = 1; perf_pmu_register(&perf_breakpoint, "breakpoint", PERF_TYPE_BREAKPOINT); return register_die_notifier(&hw_breakpoint_exceptions_nb); - err_alloc: +err: for_each_possible_cpu(err_cpu) { for (i = 0; i < TYPE_MAX; i++) kfree(get_bp_info(err_cpu, i)->tsk_pinned); @@ -737,7 +753,5 @@ int __init init_hw_breakpoint(void) break; } - return -ENOMEM; + return ret; } - - -- 2.37.2.672.g94769d06f0-goog