Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
Received-SPF: Pass (protection.outlook.com: domain of amd.com designates
 165.204.84.17 as permitted sender) receiver=protection.outlook.com;
 client-ip=165.204.84.17; helo=SATLEXMB04.amd.com; pr=C
From:   Ravi Bangoria <ravi.bangoria@amd.com>
To:     <acme@kernel.org>, <namhyung@kernel.org>
CC:     <ravi.bangoria@amd.com>, <kim.phillips@amd.com>,
        <peterz@infradead.org>, <mingo@redhat.com>, <mark.rutland@arm.com>,
        <alexander.shishkin@linux.intel.com>, <jolsa@kernel.org>,
        <irogers@google.com>, <adrian.hunter@intel.com>,
        <kan.liang@linux.intel.com>, <changbin.du@huawei.com>,
        <yangjihong1@huawei.com>, <zwisler@chromium.org>,
        <wangming01@loongson.cn>, <chenhuacai@kernel.org>,
        <kprateek.nayak@amd.com>, <linux-perf-users@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <sandipan.das@amd.com>,
        <ananth.narayan@amd.com>, <santosh.shukla@amd.com>
Subject: [PATCH 1/2] perf tool AMD: Use non-precise cycles as default event on certain Zen2 processors
Date:   Tue, 7 Nov 2023 14:03:30 +0530
Message-ID: <20231107083331.901-1-ravi.bangoria@amd.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Content-Type: text/plain
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 07 Nov 2023 08:34:06.2310
 (UTC)
X-MS-Exchange-CrossTenant-Network-Message-Id: 34ae9064-6986-449b-4a7a-08dbdf6c4e27
X-MS-Exchange-CrossTenant-Id: 3dd8961f-e488-4e60-8e11-a82d994e183d
X-MS-Exchange-CrossTenant-OriginalAttributedTenantConnectingIp: TenantId=3dd8961f-e488-4e60-8e11-a82d994e183d;Ip=[165.204.84.17];Helo=[SATLEXMB04.amd.com]
X-MS-Exchange-CrossTenant-AuthSource: SA2PEPF00001507.namprd04.prod.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Anonymous
X-MS-Exchange-CrossTenant-FromEntityHeader: HybridOnPrem
X-MS-Exchange-Transport-CrossTenantHeadersStamped: SN7PR12MB7785
X-Spam-Status: No, score=-0.9 required=5.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED,
	DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,
	MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,T_SCC_BODY_TEXT_LINE
	autolearn=unavailable autolearn_force=no version=3.4.6
X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lipwig.vger.email
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
X-Greylist: Sender passed SPF test, not delayed by milter-greylist-4.6.4 (lipwig.vger.email [0.0.0.0]); Tue, 07 Nov 2023 00:34:29 -0800 (PST)

By default, Perf uses precise cycles event when no explicit event is
specified by user. Precise cycles event is forwarded to ibs_op// pmu
on AMD. However, IBS has hw issue on certain Zen2 processors where
it might raise NMI without sample_valid bit set, which causes Unknown
NMI warnings. (Erratum #1215: IBS (Instruction Based Sampling) Counter
Valid Value May be Incorrect After Exit From Core C6 (CC6) State.) So,
use non-precise cycles as default event on affected processors.

This does not prevent user to use explicit precise cycles event or
ibs_op// pmu directly.

Suggested-by: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
---
 tools/perf/arch/x86/util/evlist.c | 34 +++++++++++++++++++++++++++++++
 tools/perf/builtin-record.c       |  2 +-
 tools/perf/builtin-top.c          |  2 +-
 tools/perf/util/evlist.c          | 12 ++++++++++-
 tools/perf/util/evlist.h          |  2 ++
 5 files changed, 49 insertions(+), 3 deletions(-)

diff --git a/tools/perf/arch/x86/util/evlist.c b/tools/perf/arch/x86/util/evlist.c
index b1ce0c52d88d..f4478179c91b 100644
--- a/tools/perf/arch/x86/util/evlist.c
+++ b/tools/perf/arch/x86/util/evlist.c
@@ -5,6 +5,8 @@
 #include "util/evlist.h"
 #include "util/parse-events.h"
 #include "util/event.h"
+#include "util/env.h"
+#include "linux/string.h"
 #include "topdown.h"
 #include "evsel.h"
 
@@ -92,3 +94,35 @@ int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs)
 	/* Default ordering by insertion index. */
 	return lhs->core.idx - rhs->core.idx;
 }
+
+/*
+ * Precise cycles event is forwarded to ibs_op// pmu on AMD. However, IBS
+ * has hw issue on certain Zen2 processors where it might raise NMI without
+ * sample_valid bit set, which causes Unknown NMI warnings. So default to
+ * non-precise cycles event on affected processors.
+ */
+const char *arch_evlist__default_cycles_event(bool can_profile_kernel)
+{
+	struct perf_env env = { .total_mem = 0, };
+	unsigned int family, model, stepping;
+	bool is_amd;
+	int ret;
+
+	perf_env__cpuid(&env);
+	is_amd = env.cpuid && strstarts(env.cpuid, "AuthenticAMD");
+	if (!is_amd)
+		goto out;
+
+	ret = sscanf(env.cpuid, "%*[^,],%u,%u,%u", &family, &model, &stepping);
+	if (ret == 3 && family == 0x17 && (
+	    (model >= 0x30 && model <= 0x3f) ||
+	    (model >= 0x60 && model <= 0x7f) ||
+	    (model >= 0x90 && model <= 0x9f))) {
+		perf_env__exit(&env);
+		return can_profile_kernel ? "cycles" : "cycles:u";
+	}
+
+out:
+	perf_env__exit(&env);
+	return can_profile_kernel ? "cycles:P" : "cycles:Pu";
+}
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index dcf288a4fb9a..e58d8ac8a77b 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -4150,7 +4150,7 @@ int cmd_record(int argc, const char **argv)
 	if (rec->evlist->core.nr_entries == 0) {
 		bool can_profile_kernel = perf_event_paranoid_check(1);
 
-		err = parse_event(rec->evlist, can_profile_kernel ? "cycles:P" : "cycles:Pu");
+		err = parse_event(rec->evlist, evlist__default_cycles_event(can_profile_kernel));
 		if (err)
 			goto out;
 	}
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index ea8c7eca5eee..21368421eddd 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -1666,7 +1666,7 @@ int cmd_top(int argc, const char **argv)
 
 	if (!top.evlist->core.nr_entries) {
 		bool can_profile_kernel = perf_event_paranoid_check(1);
-		int err = parse_event(top.evlist, can_profile_kernel ? "cycles:P" : "cycles:Pu");
+		int err = parse_event(top.evlist, evlist__default_cycles_event(can_profile_kernel));
 
 		if (err)
 			goto out_delete_evlist;
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index e36da58522ef..406ed851cafc 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -90,6 +90,16 @@ struct evlist *evlist__new(void)
 	return evlist;
 }
 
+const char * __weak arch_evlist__default_cycles_event(bool can_profile_kernel)
+{
+	return can_profile_kernel ? "cycles:P" : "cycles:Pu";
+}
+
+const char *evlist__default_cycles_event(bool can_profile_kernel)
+{
+	return arch_evlist__default_cycles_event(can_profile_kernel);
+}
+
 struct evlist *evlist__new_default(void)
 {
 	struct evlist *evlist = evlist__new();
@@ -100,7 +110,7 @@ struct evlist *evlist__new_default(void)
 		return NULL;
 
 	can_profile_kernel = perf_event_paranoid_check(1);
-	err = parse_event(evlist, can_profile_kernel ? "cycles:P" : "cycles:Pu");
+	err = parse_event(evlist, evlist__default_cycles_event(can_profile_kernel));
 	if (err) {
 		evlist__delete(evlist);
 		evlist = NULL;
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 98e7ddb2bd30..7267b4fb1981 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -91,6 +91,8 @@ struct evsel_str_handler {
 
 struct evlist *evlist__new(void);
 struct evlist *evlist__new_default(void);
+const char *arch_evlist__default_cycles_event(bool can_profile_kernel);
+const char *evlist__default_cycles_event(bool can_profile_kernel);
 struct evlist *evlist__new_dummy(void);
 void evlist__init(struct evlist *evlist, struct perf_cpu_map *cpus,
 		  struct perf_thread_map *threads);
-- 
2.41.0