From patchwork Thu Mar 12 13:44:20 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dario Faggioli X-Patchwork-Id: 11434307 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id D65D91668 for ; Thu, 12 Mar 2020 13:45:32 +0000 (UTC) Received: from lists.xenproject.org (lists.xenproject.org [192.237.175.120]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id A8D2E2067C for ; Thu, 12 Mar 2020 13:45:32 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org A8D2E2067C Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=suse.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=xen-devel-bounces@lists.xenproject.org Received: from localhost ([127.0.0.1] helo=lists.xenproject.org) by lists.xenproject.org with esmtp (Exim 4.89) (envelope-from ) id 1jCO8P-0002UX-Fa; Thu, 12 Mar 2020 13:44:25 +0000 Received: from us1-rack-iad1.inumbo.com ([172.99.69.81]) by lists.xenproject.org with esmtp (Exim 4.89) (envelope-from ) id 1jCO8O-0002UJ-CA for xen-devel@lists.xenproject.org; Thu, 12 Mar 2020 13:44:24 +0000 X-Inumbo-ID: 94ed6498-6467-11ea-b34e-bc764e2007e4 Received: from mail-wm1-f65.google.com (unknown [209.85.128.65]) by us1-rack-iad1.inumbo.com (Halon) with ESMTPS id 94ed6498-6467-11ea-b34e-bc764e2007e4; Thu, 12 Mar 2020 13:44:23 +0000 (UTC) Received: by mail-wm1-f65.google.com with SMTP id n8so6149863wmc.4 for ; Thu, 12 Mar 2020 06:44:23 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:subject:from:to:cc:date:message-id:in-reply-to :references:user-agent:mime-version:content-transfer-encoding; bh=CkT7bc8hamegT/PZ3fRPIRZB1s85fsBvvkAcU7U7HTU=; b=s2nhzikcUv/QQc3BHHaIkZa/uVOrkSj2DGYonlwe+m/4exKtL1tLPNLUezFPLUaynM sNRzdUWLSHQq4v6Udu4kG311Gi3mugg7y6ADn2WJPyT5RYsw5ZxlAJYY2Q93MRxAhzWu 9CkBN9Hj08a6wtL1LhZNEs+jb+ZZzLy0VYikVq3m76Y/HiA+NfLox5AZ4GJwZsnludD0 x4BznRjj1lAgt8tU+AxMCZcSWqpCN5/A5m2vNHCF13dmdfwkZO0F+lNJGagm1P9PColx 5LnRdMc7tDH+jQbr1gr6LpTZbdDb8CK21kOZiKLn6YMsRHLRWbWC6HnP0zjj7XkIBPWY FqmQ== X-Gm-Message-State: ANhLgQ2Ghgy5uWwXBeEwGALjntsN8c23T0kmJ5dTb/VBczk9N2uUjiF2 FYqK54OghwyhlOLjQirGVh0= X-Google-Smtp-Source: ADFU+vvhJSmZLXjC7AIqwFTK6xpllQptyVlgFX1avV8kOqk5oFPn2TznN+Y51yVmutA8/2tc/V1LBA== X-Received: by 2002:a1c:ab04:: with SMTP id u4mr4918728wme.88.1584020662657; Thu, 12 Mar 2020 06:44:22 -0700 (PDT) Received: from [192.168.0.36] (87.78.186.89.cust.ip.kpnqwest.it. [89.186.78.87]) by smtp.gmail.com with ESMTPSA id w16sm42049520wrp.8.2020.03.12.06.44.21 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 12 Mar 2020 06:44:21 -0700 (PDT) From: Dario Faggioli To: xen-devel@lists.xenproject.org Date: Thu, 12 Mar 2020 14:44:20 +0100 Message-ID: <158402066079.753.15216909367478917806.stgit@Palanthas> In-Reply-To: <158402056376.753.7091379488590272336.stgit@Palanthas> References: <158402056376.753.7091379488590272336.stgit@Palanthas> User-Agent: StGit/0.21 MIME-Version: 1.0 Subject: [Xen-devel] [PATCH 2/2] xen: credit2: fix credit reset happening too few times X-BeenThere: xen-devel@lists.xenproject.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Xen developer discussion List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Cc: Juergen Gross , Charles Arnold , Jan Beulich , Glen , George Dunlap , Tomas Mozes , Sarah Newman Errors-To: xen-devel-bounces@lists.xenproject.org Sender: "Xen-devel" There is a bug in commit 5e4b4199667b9 ("xen: credit2: only reset credit on reset condition"). In fact, the aim of that commit was to make sure that we do not perform too many credit reset operations (which are not super cheap, and in an hot-path). But the check used to determine whether a reset is necessary was the wrong one. In fact, knowing just that some vCPUs have been skipped, while traversing the runqueue (in runq_candidate()), is not enough. We need to check explicitly whether the first vCPU in the runqueue has a negative amount of credit. Since a trace record is changed, this patch updates xentrace format file and xenalyze as well This should be backported. Signed-off-by: Dario Faggioli --- Cc: George Dunlap Cc: Juergen Gross Cc: Jan Beulich Cc: Charles Arnold Cc: Glen Cc: Tomas Mozes Cc: Sarah Newman --- About the Credit2 stall issue reported recently, and mentioned in patch 1 of this series. This second patch, alone, was already mitigating the issue quite substantially. Still, the proper fix for the issue itself is patch 1, while this is a fix for a bug in the code, introduced with a previous change, which happens to help to cure the sympthoms of the problem at hand. --- tools/xentrace/formats | 2 +- tools/xentrace/xenalyze.c | 8 +++----- xen/common/sched/credit2.c | 30 +++++++++++++----------------- 3 files changed, 17 insertions(+), 23 deletions(-) diff --git a/tools/xentrace/formats b/tools/xentrace/formats index d6e7e3f800..8142d8889b 100644 --- a/tools/xentrace/formats +++ b/tools/xentrace/formats @@ -67,7 +67,7 @@ 0x00022210 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:load_check [ lrq_id[16]:orq_id[16] = 0x%(1)08x, delta = %(2)d ] 0x00022211 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:load_balance [ l_bavgload = 0x%(2)08x%(1)08x, o_bavgload = 0x%(4)08x%(3)08x, lrq_id[16]:orq_id[16] = 0x%(5)08x ] 0x00022212 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:pick_cpu [ b_avgload = 0x%(2)08x%(1)08x, dom:vcpu = 0x%(3)08x, rq_id[16]:new_cpu[16] = %(4)d ] -0x00022213 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:runq_candidate [ dom:vcpu = 0x%(1)08x, credit = %(4)d, skipped_vcpus = %(3)d, tickled_cpu = %(2)d ] +0x00022213 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:runq_candidate [ dom:vcpu = 0x%(1)08x, credit = %(3)d, tickled_cpu = %(2)d ] 0x00022214 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:schedule [ rq:cpu = 0x%(1)08x, tasklet[8]:idle[8]:smt_idle[8]:tickled[8] = %(2)08x ] 0x00022215 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:ratelimit [ dom:vcpu = 0x%(1)08x, runtime = %(2)d ] 0x00022216 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) csched2:runq_cand_chk [ dom:vcpu = 0x%(1)08x ] diff --git a/tools/xentrace/xenalyze.c b/tools/xentrace/xenalyze.c index aa894673ad..e555c572f7 100644 --- a/tools/xentrace/xenalyze.c +++ b/tools/xentrace/xenalyze.c @@ -7856,14 +7856,12 @@ void sched_process(struct pcpu_info *p) if (opt.dump_all) { struct { unsigned vcpuid:16, domid:16; - unsigned tickled_cpu, skipped; + unsigned tickled_cpu; int credit; } *r = (typeof(r))ri->d; - printf(" %s csched2:runq_candidate d%uv%u, credit = %d, " - "%u vcpus skipped, ", - ri->dump_header, r->domid, r->vcpuid, - r->credit, r->skipped); + printf(" %s csched2:runq_candidate d%uv%u, credit = %d, ", + ri->dump_header, r->domid, r->vcpuid, r->credit); if (r->tickled_cpu == (unsigned)-1) printf("no cpu was tickled\n"); else diff --git a/xen/common/sched/credit2.c b/xen/common/sched/credit2.c index 5c0ab9cd05..20692cf8ef 100644 --- a/xen/common/sched/credit2.c +++ b/xen/common/sched/credit2.c @@ -3221,8 +3221,7 @@ csched2_runtime(const struct scheduler *ops, int cpu, static struct csched2_unit * runq_candidate(struct csched2_runqueue_data *rqd, struct csched2_unit *scurr, - int cpu, s_time_t now, - unsigned int *skipped) + int cpu, s_time_t now) { struct list_head *iter, *temp; const struct sched_resource *sr = get_sched_res(cpu); @@ -3230,8 +3229,6 @@ runq_candidate(struct csched2_runqueue_data *rqd, struct csched2_private *prv = csched2_priv(sr->scheduler); bool yield = false, soft_aff_preempt = false; - *skipped = 0; - if ( unlikely(is_idle_unit(scurr->unit)) ) { snext = scurr; @@ -3325,12 +3322,9 @@ runq_candidate(struct csched2_runqueue_data *rqd, (unsigned char *)&d); } - /* Only consider units that are allowed to run on this processor. */ + /* Only consider vcpus that are allowed to run on this processor. */ if ( !cpumask_test_cpu(cpu, svc->unit->cpu_hard_affinity) ) - { - (*skipped)++; continue; - } /* * If an unit is meant to be picked up by another processor, and such @@ -3339,7 +3333,6 @@ runq_candidate(struct csched2_runqueue_data *rqd, if ( svc->tickled_cpu != -1 && svc->tickled_cpu != cpu && cpumask_test_cpu(svc->tickled_cpu, &rqd->tickled) ) { - (*skipped)++; SCHED_STAT_CRANK(deferred_to_tickled_cpu); continue; } @@ -3351,7 +3344,6 @@ runq_candidate(struct csched2_runqueue_data *rqd, if ( sched_unit_master(svc->unit) != cpu && snext->credit + CSCHED2_MIGRATE_RESIST > svc->credit ) { - (*skipped)++; SCHED_STAT_CRANK(migrate_resisted); continue; } @@ -3375,14 +3367,13 @@ runq_candidate(struct csched2_runqueue_data *rqd, { struct { unsigned unit:16, dom:16; - unsigned tickled_cpu, skipped; + unsigned tickled_cpu; int credit; } d; d.dom = snext->unit->domain->domain_id; d.unit = snext->unit->unit_id; d.credit = snext->credit; d.tickled_cpu = snext->tickled_cpu; - d.skipped = *skipped; __trace_var(TRC_CSCHED2_RUNQ_CANDIDATE, 1, sizeof(d), (unsigned char *)&d); @@ -3414,7 +3405,6 @@ static void csched2_schedule( struct csched2_runqueue_data *rqd; struct csched2_unit * const scurr = csched2_unit(currunit); struct csched2_unit *snext = NULL; - unsigned int skipped_units = 0; bool tickled; bool migrated = false; @@ -3492,7 +3482,7 @@ static void csched2_schedule( snext = csched2_unit(sched_idle_unit(sched_cpu)); } else - snext = runq_candidate(rqd, scurr, sched_cpu, now, &skipped_units); + snext = runq_candidate(rqd, scurr, sched_cpu, now); /* If switching from a non-idle runnable unit, put it * back on the runqueue. */ @@ -3504,6 +3494,8 @@ static void csched2_schedule( /* Accounting for non-idle tasks */ if ( !is_idle_unit(snext->unit) ) { + int top_credit; + /* If switching, remove this from the runqueue and mark it scheduled */ if ( snext != scurr ) { @@ -3531,11 +3523,15 @@ static void csched2_schedule( * 2) no other unit with higher credits wants to run. * * Here, where we want to check for reset, we need to make sure the - * proper unit is being used. In fact, runqueue_candidate() may have - * not returned the first unit in the runqueue, for various reasons + * proper unit is being used. In fact, runq_candidate() may have not + * returned the first unit in the runqueue, for various reasons * (e.g., affinity). Only trigger a reset when it does. */ - if ( skipped_units == 0 && snext->credit <= CSCHED2_CREDIT_RESET ) + if ( list_empty(&rqd->runq) ) + top_credit = snext->credit; + else + top_credit = max(snext->credit, runq_elem(rqd->runq.next)->credit); + if ( top_credit <= CSCHED2_CREDIT_RESET ) { reset_credit(sched_cpu, now, snext); balance_load(ops, sched_cpu, now);