linux-perf-users.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Ian Rogers <irogers@google.com>
To: perry.taylor@intel.com, caleb.biggers@intel.com,
	kshipra.bopardikar@intel.com,
	Kan Liang <kan.liang@linux.intel.com>,
	Zhengjun Xing <zhengjun.xing@linux.intel.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Jiri Olsa <jolsa@redhat.com>, Namhyung Kim <namhyung@kernel.org>,
	Maxime Coquelin <mcoquelin.stm32@gmail.com>,
	Alexandre Torgue <alexandre.torgue@foss.st.com>,
	Andi Kleen <ak@linux.intel.com>,
	James Clark <james.clark@arm.com>,
	John Garry <john.garry@huawei.com>,
	linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org
Cc: Stephane Eranian <eranian@google.com>, Ian Rogers <irogers@google.com>
Subject: [PATCH v3 2/2] perf tsc: Add arch TSC frequency information
Date: Fri, 15 Jul 2022 15:35:21 -0700	[thread overview]
Message-ID: <20220715223521.3389971-3-irogers@google.com> (raw)
In-Reply-To: <20220715223521.3389971-1-irogers@google.com>

From: Kan Liang <kan.liang@linux.intel.com>

The TSC frequency information is required for the event metrics with
the literal, system_tsc_freq. For the newer Intel platform, the TSC
frequency information can be retrieved from the CPUID leaf 0x15.
If the TSC frequency information isn't present the /proc/cpuinfo
approach is used.

Refactor cpuid for this use. Note, the previous stack pushing/popping
approach was broken on x86-64 that has stack red zones that would be
clobbered.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/arch/x86/util/cpuid.h  | 34 +++++++++++++++++++++++++++++++
 tools/perf/arch/x86/util/header.c | 27 ++++++++++--------------
 tools/perf/arch/x86/util/tsc.c    | 33 ++++++++++++++++++++++++++++++
 tools/perf/util/expr.c            | 15 +++++++++++++-
 tools/perf/util/tsc.h             |  1 +
 5 files changed, 93 insertions(+), 17 deletions(-)
 create mode 100644 tools/perf/arch/x86/util/cpuid.h

diff --git a/tools/perf/arch/x86/util/cpuid.h b/tools/perf/arch/x86/util/cpuid.h
new file mode 100644
index 000000000000..0a3ae0ace7e9
--- /dev/null
+++ b/tools/perf/arch/x86/util/cpuid.h
@@ -0,0 +1,34 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef PERF_CPUID_H
+#define PERF_CPUID_H 1
+
+
+static inline void
+cpuid(unsigned int op, unsigned int op2, unsigned int *a, unsigned int *b,
+	unsigned int *c, unsigned int *d)
+{
+	/*
+	 * Preserve %ebx/%rbx register by either placing it in %rdi or saving it
+	 * on the stack - x86-64 needs to avoid the stack red zone. In PIC
+	 * compilations %ebx contains the address of the global offset
+	 * table. %rbx is occasionally used to address stack variables in
+	 * presence of dynamic allocas.
+	 */
+	asm(
+#if defined(__x86_64__)
+		"mov %%rbx, %%rdi\n"
+		"cpuid\n"
+		"xchg %%rdi, %%rbx\n"
+#else
+		"pushl %%ebx\n"
+		"cpuid\n"
+		"movl %%ebx, %%edi\n"
+		"popl %%ebx\n"
+#endif
+		: "=a"(*a), "=D"(*b), "=c"(*c), "=d"(*d)
+		: "a"(op), "2"(op2));
+}
+
+void get_cpuid_0(char *vendor, unsigned int *lvl);
+
+#endif
diff --git a/tools/perf/arch/x86/util/header.c b/tools/perf/arch/x86/util/header.c
index 578c8c568ffd..a51444a77a5f 100644
--- a/tools/perf/arch/x86/util/header.c
+++ b/tools/perf/arch/x86/util/header.c
@@ -9,18 +9,17 @@
 
 #include "../../../util/debug.h"
 #include "../../../util/header.h"
+#include "cpuid.h"
 
-static inline void
-cpuid(unsigned int op, unsigned int *a, unsigned int *b, unsigned int *c,
-      unsigned int *d)
+void get_cpuid_0(char *vendor, unsigned int *lvl)
 {
-	__asm__ __volatile__ (".byte 0x53\n\tcpuid\n\t"
-			      "movl %%ebx, %%esi\n\t.byte 0x5b"
-			: "=a" (*a),
-			"=S" (*b),
-			"=c" (*c),
-			"=d" (*d)
-			: "a" (op));
+	unsigned int b, c, d;
+
+	cpuid(0, 0, lvl, &b, &c, &d);
+	strncpy(&vendor[0], (char *)(&b), 4);
+	strncpy(&vendor[4], (char *)(&d), 4);
+	strncpy(&vendor[8], (char *)(&c), 4);
+	vendor[12] = '\0';
 }
 
 static int
@@ -31,14 +30,10 @@ __get_cpuid(char *buffer, size_t sz, const char *fmt)
 	int nb;
 	char vendor[16];
 
-	cpuid(0, &lvl, &b, &c, &d);
-	strncpy(&vendor[0], (char *)(&b), 4);
-	strncpy(&vendor[4], (char *)(&d), 4);
-	strncpy(&vendor[8], (char *)(&c), 4);
-	vendor[12] = '\0';
+	get_cpuid_0(vendor, &lvl);
 
 	if (lvl >= 1) {
-		cpuid(1, &a, &b, &c, &d);
+		cpuid(1, 0, &a, &b, &c, &d);
 
 		family = (a >> 8) & 0xf;  /* bits 11 - 8 */
 		model  = (a >> 4) & 0xf;  /* Bits  7 - 4 */
diff --git a/tools/perf/arch/x86/util/tsc.c b/tools/perf/arch/x86/util/tsc.c
index 559365f8fe52..b69144f22489 100644
--- a/tools/perf/arch/x86/util/tsc.c
+++ b/tools/perf/arch/x86/util/tsc.c
@@ -1,7 +1,9 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <linux/types.h>
+#include <string.h>
 
 #include "../../../util/tsc.h"
+#include "cpuid.h"
 
 u64 rdtsc(void)
 {
@@ -11,3 +13,34 @@ u64 rdtsc(void)
 
 	return low | ((u64)high) << 32;
 }
+
+double arch_get_tsc_freq(void)
+{
+	unsigned int a, b, c, d, lvl;
+	static bool cached;
+	static double tsc;
+	char vendor[16];
+
+	if (cached)
+		return tsc;
+
+	cached = true;
+	get_cpuid_0(vendor, &lvl);
+	if (!strstr(vendor, "Intel"))
+		return 0;
+
+	/*
+	 * Don't support Time Stamp Counter and
+	 * Nominal Core Crystal Clock Information Leaf.
+	 */
+	if (lvl < 0x15)
+		return 0;
+
+	cpuid(0x15, 0, &a, &b, &c, &d);
+	/* TSC frequency is not enumerated */
+	if (!a || !b || !c)
+		return 0;
+
+	tsc = (double)c * (double)b / (double)a;
+	return tsc;
+}
diff --git a/tools/perf/util/expr.c b/tools/perf/util/expr.c
index 4c81533e4b43..16f10e6d5ca5 100644
--- a/tools/perf/util/expr.c
+++ b/tools/perf/util/expr.c
@@ -12,6 +12,7 @@
 #include "expr-bison.h"
 #include "expr-flex.h"
 #include "smt.h"
+#include "tsc.h"
 #include <linux/err.h>
 #include <linux/kernel.h>
 #include <linux/zalloc.h>
@@ -443,9 +444,19 @@ static double system_tsc_freq(void)
 
 	free(line);
 	fclose(cpuinfo);
+	if (isnan(result))
+		pr_err("Error reading system_tsc_freq");
+
 	return result;
 }
 
+#if !defined(__i386__) && !defined(__x86_64__)
+double arch_get_tsc_freq(void)
+{
+	return 0.0;
+}
+#endif
+
 double expr__get_literal(const char *literal)
 {
 	static struct cpu_topology *topology;
@@ -462,7 +473,9 @@ double expr__get_literal(const char *literal)
 	}
 
 	if (!strcasecmp("#system_tsc_freq", literal)) {
-		result = system_tsc_freq();
+		result = arch_get_tsc_freq();
+		if (fpclassify(result) == FP_ZERO)
+			result = system_tsc_freq();
 		goto out;
 	}
 
diff --git a/tools/perf/util/tsc.h b/tools/perf/util/tsc.h
index 7d83a31732a7..88fd1c4c1cb8 100644
--- a/tools/perf/util/tsc.h
+++ b/tools/perf/util/tsc.h
@@ -25,6 +25,7 @@ int perf_read_tsc_conversion(const struct perf_event_mmap_page *pc,
 u64 perf_time_to_tsc(u64 ns, struct perf_tsc_conversion *tc);
 u64 tsc_to_perf_time(u64 cyc, struct perf_tsc_conversion *tc);
 u64 rdtsc(void);
+double arch_get_tsc_freq(void);
 
 size_t perf_event__fprintf_time_conv(union perf_event *event, FILE *fp);
 
-- 
2.37.0.170.g444d1eabd0-goog


  parent reply	other threads:[~2022-07-15 22:35 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-07-15 22:35 [PATCH v3 0/2] Add arch TSC frequency information Ian Rogers
2022-07-15 22:35 ` [PATCH v3 1/2] perf metrics: Add literal for system TSC frequency Ian Rogers
2022-07-18 12:14   ` Liang, Kan
2022-07-15 22:35 ` Ian Rogers [this message]
2022-07-18 12:49   ` [PATCH v3 2/2] perf tsc: Add arch TSC frequency information Liang, Kan
2022-07-18 14:47     ` Ian Rogers

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220715223521.3389971-3-irogers@google.com \
    --to=irogers@google.com \
    --cc=acme@kernel.org \
    --cc=ak@linux.intel.com \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=alexandre.torgue@foss.st.com \
    --cc=caleb.biggers@intel.com \
    --cc=eranian@google.com \
    --cc=james.clark@arm.com \
    --cc=john.garry@huawei.com \
    --cc=jolsa@redhat.com \
    --cc=kan.liang@linux.intel.com \
    --cc=kshipra.bopardikar@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=mcoquelin.stm32@gmail.com \
    --cc=mingo@redhat.com \
    --cc=namhyung@kernel.org \
    --cc=perry.taylor@intel.com \
    --cc=peterz@infradead.org \
    --cc=zhengjun.xing@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).