From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org,
Adrian Hunter <adrian.hunter@intel.com>,
Alexander Shishkin <alexander.shishkin@linux.intel.com>,
He Kuang <hekuang@huawei.com>, Jiri Olsa <jolsa@redhat.com>,
Josh Poimboeuf <jpoimboe@redhat.com>,
Peter Zijlstra <peterz@infradead.org>,
Stephane Eranian <eranian@google.com>,
Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>,
Wang Nan <wangnan0@huawei.com>,
Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 05/10] perf jit: Add support for using TSC as a timestamp
Date: Fri, 1 Apr 2016 18:57:01 -0300 [thread overview]
Message-ID: <1459547826-20100-6-git-send-email-acme@kernel.org> (raw)
In-Reply-To: <1459547826-20100-1-git-send-email-acme@kernel.org>
From: Adrian Hunter <adrian.hunter@intel.com>
Intel PT uses TSC as a timestamp, so add support for using TSC instead
of the monotonic clock. Use of TSC is selected by an environment
variable "JITDUMP_USE_ARCH_TIMESTAMP" and flagged in the jitdump file
with flag JITDUMP_FLAGS_ARCH_TIMESTAMP.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1457426330-30226-1-git-send-email-adrian.hunter@intel.com
[ Added the fixup from He Kuang to make it build on other arches, ]
[ such as aarch64, to avoid inserting this bisectiong breakage upstream ]
Link: http://lkml.kernel.org/r/1459482572-129494-1-git-send-email-hekuang@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/arch/x86/util/tsc.c | 1 -
tools/perf/arch/x86/util/tsc.h | 17 -----------------
tools/perf/jvmti/jvmti_agent.c | 43 ++++++++++++++++++++++++++++++++++++++++--
tools/perf/util/Build | 3 +--
tools/perf/util/jitdump.c | 37 +++++++++++++++++++++++++++++++-----
tools/perf/util/jitdump.h | 3 +++
tools/perf/util/tsc.h | 11 ++++++++++-
7 files changed, 87 insertions(+), 28 deletions(-)
delete mode 100644 tools/perf/arch/x86/util/tsc.h
diff --git a/tools/perf/arch/x86/util/tsc.c b/tools/perf/arch/x86/util/tsc.c
index 70ff7c14bea6..357f1b13b5ae 100644
--- a/tools/perf/arch/x86/util/tsc.c
+++ b/tools/perf/arch/x86/util/tsc.c
@@ -7,7 +7,6 @@
#include <linux/types.h>
#include "../../util/debug.h"
#include "../../util/tsc.h"
-#include "tsc.h"
int perf_read_tsc_conversion(const struct perf_event_mmap_page *pc,
struct perf_tsc_conversion *tc)
diff --git a/tools/perf/arch/x86/util/tsc.h b/tools/perf/arch/x86/util/tsc.h
deleted file mode 100644
index 2edc4d31065c..000000000000
--- a/tools/perf/arch/x86/util/tsc.h
+++ /dev/null
@@ -1,17 +0,0 @@
-#ifndef TOOLS_PERF_ARCH_X86_UTIL_TSC_H__
-#define TOOLS_PERF_ARCH_X86_UTIL_TSC_H__
-
-#include <linux/types.h>
-
-struct perf_tsc_conversion {
- u16 time_shift;
- u32 time_mult;
- u64 time_zero;
-};
-
-struct perf_event_mmap_page;
-
-int perf_read_tsc_conversion(const struct perf_event_mmap_page *pc,
- struct perf_tsc_conversion *tc);
-
-#endif /* TOOLS_PERF_ARCH_X86_UTIL_TSC_H__ */
diff --git a/tools/perf/jvmti/jvmti_agent.c b/tools/perf/jvmti/jvmti_agent.c
index 6461e02ab940..3573f315f955 100644
--- a/tools/perf/jvmti/jvmti_agent.c
+++ b/tools/perf/jvmti/jvmti_agent.c
@@ -92,6 +92,22 @@ error:
return ret;
}
+static int use_arch_timestamp;
+
+static inline uint64_t
+get_arch_timestamp(void)
+{
+#if defined(__i386__) || defined(__x86_64__)
+ unsigned int low, high;
+
+ asm volatile("rdtsc" : "=a" (low), "=d" (high));
+
+ return low | ((uint64_t)high) << 32;
+#else
+ return 0;
+#endif
+}
+
#define NSEC_PER_SEC 1000000000
static int perf_clk_id = CLOCK_MONOTONIC;
@@ -107,6 +123,9 @@ perf_get_timestamp(void)
struct timespec ts;
int ret;
+ if (use_arch_timestamp)
+ return get_arch_timestamp();
+
ret = clock_gettime(perf_clk_id, &ts);
if (ret)
return 0;
@@ -203,6 +222,17 @@ perf_close_marker_file(void)
munmap(marker_addr, pgsz);
}
+static void
+init_arch_timestamp(void)
+{
+ char *str = getenv("JITDUMP_USE_ARCH_TIMESTAMP");
+
+ if (!str || !*str || !strcmp(str, "0"))
+ return;
+
+ use_arch_timestamp = 1;
+}
+
void *jvmti_open(void)
{
int pad_cnt;
@@ -211,11 +241,17 @@ void *jvmti_open(void)
int fd;
FILE *fp;
+ init_arch_timestamp();
+
/*
* check if clockid is supported
*/
- if (!perf_get_timestamp())
- warnx("jvmti: kernel does not support %d clock id", perf_clk_id);
+ if (!perf_get_timestamp()) {
+ if (use_arch_timestamp)
+ warnx("jvmti: arch timestamp not supported");
+ else
+ warnx("jvmti: kernel does not support %d clock id", perf_clk_id);
+ }
memset(&header, 0, sizeof(header));
@@ -263,6 +299,9 @@ void *jvmti_open(void)
header.timestamp = perf_get_timestamp();
+ if (use_arch_timestamp)
+ header.flags |= JITDUMP_FLAGS_ARCH_TIMESTAMP;
+
if (!fwrite(&header, sizeof(header), 1, fp)) {
warn("jvmti: cannot write dumpfile header");
goto error;
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index da48fd843438..85ceff357769 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -69,8 +69,7 @@ libperf-y += stat-shadow.o
libperf-y += record.o
libperf-y += srcline.o
libperf-y += data.o
-libperf-$(CONFIG_X86) += tsc.o
-libperf-$(CONFIG_AUXTRACE) += tsc.o
+libperf-y += tsc.o
libperf-y += cloexec.o
libperf-y += thread-stack.o
libperf-$(CONFIG_AUXTRACE) += auxtrace.o
diff --git a/tools/perf/util/jitdump.c b/tools/perf/util/jitdump.c
index ad0c0bb1fbc7..52fcef3074fe 100644
--- a/tools/perf/util/jitdump.c
+++ b/tools/perf/util/jitdump.c
@@ -17,6 +17,7 @@
#include "strlist.h"
#include <elf.h>
+#include "tsc.h"
#include "session.h"
#include "jit.h"
#include "jitdump.h"
@@ -33,6 +34,7 @@ struct jit_buf_desc {
size_t bufsize;
FILE *in;
bool needs_bswap; /* handles cross-endianess */
+ bool use_arch_timestamp;
void *debug_data;
size_t nr_debug_entries;
uint32_t code_load_count;
@@ -158,13 +160,16 @@ jit_open(struct jit_buf_desc *jd, const char *name)
header.flags = bswap_64(header.flags);
}
+ jd->use_arch_timestamp = header.flags & JITDUMP_FLAGS_ARCH_TIMESTAMP;
+
if (verbose > 2)
- pr_debug("version=%u\nhdr.size=%u\nts=0x%llx\npid=%d\nelf_mach=%d\n",
+ pr_debug("version=%u\nhdr.size=%u\nts=0x%llx\npid=%d\nelf_mach=%d\nuse_arch_timestamp=%d\n",
header.version,
header.total_size,
(unsigned long long)header.timestamp,
header.pid,
- header.elf_mach);
+ header.elf_mach,
+ jd->use_arch_timestamp);
if (header.flags & JITDUMP_FLAGS_RESERVED) {
pr_err("jitdump file contains invalid or unsupported flags 0x%llx\n",
@@ -172,10 +177,15 @@ jit_open(struct jit_buf_desc *jd, const char *name)
goto error;
}
+ if (jd->use_arch_timestamp && !jd->session->time_conv.time_mult) {
+ pr_err("jitdump file uses arch timestamps but there is no timestamp conversion\n");
+ goto error;
+ }
+
/*
* validate event is using the correct clockid
*/
- if (jit_validate_events(jd->session)) {
+ if (!jd->use_arch_timestamp && jit_validate_events(jd->session)) {
pr_err("error, jitted code must be sampled with perf record -k 1\n");
goto error;
}
@@ -329,6 +339,23 @@ jit_inject_event(struct jit_buf_desc *jd, union perf_event *event)
return 0;
}
+static uint64_t convert_timestamp(struct jit_buf_desc *jd, uint64_t timestamp)
+{
+ struct perf_tsc_conversion tc;
+
+ if (!jd->use_arch_timestamp)
+ return timestamp;
+
+ tc.time_shift = jd->session->time_conv.time_shift;
+ tc.time_mult = jd->session->time_conv.time_mult;
+ tc.time_zero = jd->session->time_conv.time_zero;
+
+ if (!tc.time_mult)
+ return 0;
+
+ return tsc_to_perf_time(timestamp, &tc);
+}
+
static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
{
struct perf_sample sample;
@@ -410,7 +437,7 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
id->tid = tid;
}
if (jd->sample_type & PERF_SAMPLE_TIME)
- id->time = jr->load.p.timestamp;
+ id->time = convert_timestamp(jd, jr->load.p.timestamp);
/*
* create pseudo sample to induce dso hit increment
@@ -499,7 +526,7 @@ static int jit_repipe_code_move(struct jit_buf_desc *jd, union jr_entry *jr)
id->tid = tid;
}
if (jd->sample_type & PERF_SAMPLE_TIME)
- id->time = jr->load.p.timestamp;
+ id->time = convert_timestamp(jd, jr->load.p.timestamp);
/*
* create pseudo sample to induce dso hit increment
diff --git a/tools/perf/util/jitdump.h b/tools/perf/util/jitdump.h
index b66c1f503d9e..bcacd20d0c1c 100644
--- a/tools/perf/util/jitdump.h
+++ b/tools/perf/util/jitdump.h
@@ -23,9 +23,12 @@
#define JITHEADER_VERSION 1
enum jitdump_flags_bits {
+ JITDUMP_FLAGS_ARCH_TIMESTAMP_BIT,
JITDUMP_FLAGS_MAX_BIT,
};
+#define JITDUMP_FLAGS_ARCH_TIMESTAMP (1ULL << JITDUMP_FLAGS_ARCH_TIMESTAMP_BIT)
+
#define JITDUMP_FLAGS_RESERVED (JITDUMP_FLAGS_MAX_BIT < 64 ? \
(~((1ULL << JITDUMP_FLAGS_MAX_BIT) - 1)) : 0)
diff --git a/tools/perf/util/tsc.h b/tools/perf/util/tsc.h
index 280ddc067556..d5b11e2b85e0 100644
--- a/tools/perf/util/tsc.h
+++ b/tools/perf/util/tsc.h
@@ -4,7 +4,16 @@
#include <linux/types.h>
#include "event.h"
-#include "../arch/x86/util/tsc.h"
+
+struct perf_tsc_conversion {
+ u16 time_shift;
+ u32 time_mult;
+ u64 time_zero;
+};
+struct perf_event_mmap_page;
+
+int perf_read_tsc_conversion(const struct perf_event_mmap_page *pc,
+ struct perf_tsc_conversion *tc);
u64 perf_time_to_tsc(u64 ns, struct perf_tsc_conversion *tc);
u64 tsc_to_perf_time(u64 cyc, struct perf_tsc_conversion *tc);
--
2.5.5
next prev parent reply other threads:[~2016-04-01 21:57 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-01 21:56 [GIT PULL 00/10] perf/core improvements and fixes Arnaldo Carvalho de Melo
2016-04-01 21:56 ` [PATCH 01/10] perf trace: Do not process PERF_RECORD_LOST twice Arnaldo Carvalho de Melo
2016-04-01 21:56 ` [PATCH 02/10] perf trace: Pretty print seccomp() args Arnaldo Carvalho de Melo
2016-04-01 21:56 ` [PATCH 03/10] perf trace: Pretty print getrandom() args Arnaldo Carvalho de Melo
2016-04-01 21:57 ` [PATCH 04/10] perf tools: Add time conversion event Arnaldo Carvalho de Melo
2016-04-01 21:57 ` Arnaldo Carvalho de Melo [this message]
2016-04-01 21:57 ` [PATCH 06/10] perf intel-pt/bts: Define JITDUMP_USE_ARCH_TIMESTAMP Arnaldo Carvalho de Melo
2016-04-01 21:57 ` [PATCH 07/10] perf tools: Fix PMU term format max value calculation Arnaldo Carvalho de Melo
2016-04-01 21:57 ` [PATCH 08/10] perf trace: Introduce function to set the base timestamp Arnaldo Carvalho de Melo
2016-04-01 21:57 ` [PATCH 09/10] perf trace: Don't set the base timestamp using events without PERF_SAMPLE_TIME Arnaldo Carvalho de Melo
2016-04-01 21:57 ` [PATCH 10/10] perf bpf: Add sample types for 'bpf-output' event Arnaldo Carvalho de Melo
2016-04-06 6:47 ` [GIT PULL 00/10] perf/core improvements and fixes Ingo Molnar
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=1459547826-20100-6-git-send-email-acme@kernel.org \
--to=acme@kernel.org \
--cc=acme@redhat.com \
--cc=adrian.hunter@intel.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=eranian@google.com \
--cc=hekuang@huawei.com \
--cc=jolsa@redhat.com \
--cc=jpoimboe@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=peterz@infradead.org \
--cc=sukadev@linux.vnet.ibm.com \
--cc=wangnan0@huawei.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).