Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1237039 > unrolled thread

[PATCH v7 0/4] perf: add support for profiling jitted code

Started byStephane Eranian <eranian@google.com>
First post2015-10-01 08:50 +0200
Last post2015-10-04 22:10 +0200
Articles 9 — 5 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH v7 0/4] perf: add support for profiling jitted code Stephane Eranian <eranian@google.com> - 2015-10-01 08:50 +0200
    [PATCH v7 5/5] perf/jit: add source line info support Stephane Eranian <eranian@google.com> - 2015-10-01 08:50 +0200
    [PATCH v7 4/5] perf tools: add JVMTI agent library Stephane Eranian <eranian@google.com> - 2015-10-01 08:50 +0200
    Re: [PATCH v7 0/4] perf: add support for profiling jitted code Peter Zijlstra <peterz@infradead.org> - 2015-10-01 11:20 +0200
      Re: [PATCH v7 0/4] perf: add support for profiling jitted code Ingo Molnar <mingo@kernel.org> - 2015-10-01 11:40 +0200
        Re: [PATCH v7 0/4] perf: add support for profiling jitted code Arnaldo Carvalho de Melo <acme@redhat.com> - 2015-10-01 18:20 +0200
      Re: [PATCH v7 0/4] perf: add support for profiling jitted code Stephane Eranian <eranian@google.com> - 2015-10-04 22:00 +0200
    Re: [PATCH v7 0/4] perf: add support for profiling jitted code Brendan Gregg <brendan.d.gregg@gmail.com> - 2015-10-02 00:50 +0200
      Re: [PATCH v7 0/4] perf: add support for profiling jitted code Stephane Eranian <eranian@google.com> - 2015-10-04 22:10 +0200

#1237039 — [PATCH v7 0/4] perf: add support for profiling jitted code

FromStephane Eranian <eranian@google.com>
Date2015-10-01 08:50 +0200
Subject[PATCH v7 0/4] perf: add support for profiling jitted code
Message-ID<qeFQK-4lA-3@gated-at.bofh.it>
This patch series extends perf record/report/annotate to enable
profiling of jitted (just-in-time compiled) code. The current
perf tool provides very limited support for profiling jitted
code for some runtime environments. But the support is experimental
and cannot be used in complex environments. It relies on files
in /tmp, for instance. It does not support annotate mode or
rejitted code.

This patch series adds a better way of profiling jitted code
with the following advantages:
   - support any jitted code environment (some with modifications)
   - support Java runtime with JVMTI interface with no modifications
   - provides a portable JVMTI agent library
   - known to support V8 runtime
   - known to support DART runtime
   - supports code rejitting and code movements
   - no files in /tmp
   - meta-data file is unique to each run
   - no changes to perf report/annotate
   - support per-thread and system-wide profiling
   - support monitoring of multiple simultaneous Jit runtimes
   - source level view in perf annotate

The support is based on cooperation with the runtime. For Java runtimes,
supporting the JVMTI interface, there is no change necessary. For other
runtimes, modifications are necessary to emit the meta-data to support
symbolization, annotation, source lines correlation of the samples.
Those modifications are relatively straighforward, some have been 
implemented in V8 and DART.

The jit environment emits a binary dump file which contains the jitted
code (in raw format) and meta-data describing the mapping of functions.
The binary format is documented in the jitdump.h header file. It is
adapted from the OProfile jitdump format.

To enable synchronization of the runtime MMAPs with those recorded by
the kernel on behalf of the perf tool, the runtime needs to timestamp
any record in the dump file using the same time source. This is possible
since Linux 4.1 where the kernel supports per event timestamp clock source.
In the case of the JVMTI agent, the clock used is CLOCK_MONOTONIC, thus
perf record is invoked with -k mono such that it matches the agent.

The current support only works when the runtime is monitored from
start to finish: perf record java --agentpath:libpfmjvmti.so my_class.

Once the run is completed, the jitdump file needs to be injected into
the perf.data file. This is accomplished by using the perf inject command.
This will also generate an ELF image for each jitted function. The
injected MMAP records will point to these ELF images. The reasoning
behind using ELF images is that it makes processing for perf report
and annotate automatic and transparent. It also makes it easier to
package and analyze on a remote machine. Binutils tools can decode
the ELF images easily.

The reporting is unchanged, simply invoke perf report or perf annotate
on the modified perf.data file. The jitted code will appear symbolized
and the assembly view will display the instruction level profile and
source level profile.

As an added bonus, the series includes support for demangling function
signature from OpenJDK.

Furthermore, we believe there is a way to skip the perf inject phase
and have perf report/annotate directly inject the MMAP records 
on the fly during processing of the perf.data file. Perf report would
also generate the ELF files if necessary. Such optimization, would
make using this extension seamless in system-wide mode and larger
environments. This will be added in a later update as well.

In V2, we have switched to Pawell Moll and David Ahern posix
clock kernel module instead. We have dropped the patch which
modified the arguments to map_init() because the change was
not used. We are not printing the return type of Java methods
anymore and have made the Java demangler a separate module.
We also rebased to 3.19.0+ from tip.git.

In V3, we switched to Pawel Moll's CLOCK_MONOTONIC perf
clock patches. This patch switch perf_events from sched_clock
to CLOCK_MONOTONIC, a clock source which is available to users.

In V4, we rebased to 4.0-rc5. We also simplified the process by
getting rid of the requirement to pass the jitdump file name to
perf inject. Now, perf injects automtically detects if jitdumps
were generated and it merges the relevant meta-data. This is
accomplished by having the jit runtime mmap the jitdump file
for the purpose of creating a MMAP record in the perf.data file.
That MMAP contains all the info to locate the jitdump file and
generate the ELF images for jitted functions.

In V5, we rebase to acme's perf/core branch (instead of tip.git).
We fixed some bswap issues, switched to using scnprintf() and fixed
formatting issues. Also made sure all the files were included in the
patches. We also fix one error message in the JVMTI agent.


In V6, we switched back to using tip.git to leverage PeterZ's clockid
patch for perf_events in 4.0.0-rc6. Clock source can now be specified
per event and they are connected with the MONOTONIC Posix clock. We
leverage this extension to timestamp samples in the jit runtime and
correlate them with perf samples. Notice the -k mono option in perf
record example below.

In V7, we rebased to 4.3.0-rc3 using tip.git (at commit 0dc7757).
We fixed several issues in the agent. We also added source line
information in the jitdump file from the JVMTI agent. This is
still experimental and probably has some issues. The source
line info is encoded in DWARF2 format in each ELF image. The
code to do this is leveraged from Oprofile with some fixes
and cleanups.

To use the new feature:
  - need to run with 4.1 or later
  - compile perf
  - cd tools/perf/jvmti; make; install wherever is appropriate

  Example using openJDK:
   $ perf record -k mono java -agentpath:libjvmti.so my_class
   $ perf inject -i perf.data -jit -o perf.data.jitted
   $ perf report -i perf.data.jitted

Thanks to all the contributors and testers. Special thanks
to PeterZ for adding the clock source to perf_events and solving
the problem of common timesource for user and kernel level samples.
Thanks to the Oprofile authors for the DWARF2 source line code
generation.

Enjoy,

Stephane Eranian (5):
  perf tools: add Java demangling support
  perf tools: pass session to mmap processing code
  perf inject: add jitdump mmap injection support
  perf tools: add JVMTI agent library
  perf/jit: add source line info support

 tools/build/Makefile.feature             |   2 +
 tools/build/feature/Makefile             |   4 +
 tools/perf/Documentation/perf-inject.txt |   7 +
 tools/perf/builtin-inject.c              |  97 ++++-
 tools/perf/builtin-script.c              |  14 +-
 tools/perf/config/Makefile               |  11 +
 tools/perf/jvmti/Makefile                |  70 ++++
 tools/perf/jvmti/jvmti_agent.c           | 464 +++++++++++++++++++++
 tools/perf/jvmti/jvmti_agent.h           |  36 ++
 tools/perf/jvmti/libjvmti.c              | 304 ++++++++++++++
 tools/perf/util/Build                    |   4 +
 tools/perf/util/demangle-java.c          | 199 +++++++++
 tools/perf/util/demangle-java.h          |  10 +
 tools/perf/util/event.c                  |   6 +-
 tools/perf/util/event.h                  |   6 +-
 tools/perf/util/genelf.c                 | 448 +++++++++++++++++++++
 tools/perf/util/genelf.h                 |  60 +++
 tools/perf/util/genelf_debug.c           | 610 ++++++++++++++++++++++++++++
 tools/perf/util/jitdump.c                | 666 +++++++++++++++++++++++++++++++
 tools/perf/util/jitdump.h                | 116 ++++++
 tools/perf/util/session.c                |  43 +-
 tools/perf/util/symbol-elf.c             |   3 +
 tools/perf/util/tool.h                   |  11 +-
 23 files changed, 3152 insertions(+), 39 deletions(-)
 create mode 100644 tools/perf/jvmti/Makefile
 create mode 100644 tools/perf/jvmti/jvmti_agent.c
 create mode 100644 tools/perf/jvmti/jvmti_agent.h
 create mode 100644 tools/perf/jvmti/libjvmti.c
 create mode 100644 tools/perf/util/demangle-java.c
 create mode 100644 tools/perf/util/demangle-java.h
 create mode 100644 tools/perf/util/genelf.c
 create mode 100644 tools/perf/util/genelf.h
 create mode 100644 tools/perf/util/genelf_debug.c
 create mode 100644 tools/perf/util/jitdump.c
 create mode 100644 tools/perf/util/jitdump.h

-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]

#1237041 — [PATCH v7 5/5] perf/jit: add source line info support

FromStephane Eranian <eranian@google.com>
Date2015-10-01 08:50 +0200
Subject[PATCH v7 5/5] perf/jit: add source line info support
Message-ID<qeFQK-4lA-13@gated-at.bofh.it>
In reply to#1237039
This patch adds source line information support to perf for jitted code.
The source line info must be emitted by the runtime, such as JVMTI.
Perf injects extract the source line info from the jitdump file and
adds the corresponding .debug_lines section in the ELF image generated
for each jitted function. The source line enables matching any address
in the profile with a source file and line number. The improvement is
visible in perf annotate with the source code displayed alongside
the assembly code.

The dwarf code leverages the support from OProfile which is also released
under GPLv2.  Copyright 2007 OProfile authors.

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 tools/perf/jvmti/jvmti_agent.c |  32 +--
 tools/perf/jvmti/jvmti_agent.h |  11 +-
 tools/perf/jvmti/libjvmti.c    | 122 ++++++++-
 tools/perf/util/Build          |   1 +
 tools/perf/util/genelf.c       |  15 +-
 tools/perf/util/genelf.h       |   6 +-
 tools/perf/util/genelf_debug.c | 610 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/jitdump.c      |   8 +-
 8 files changed, 766 insertions(+), 39 deletions(-)
 create mode 100644 tools/perf/util/genelf_debug.c

diff --git a/tools/perf/jvmti/jvmti_agent.c b/tools/perf/jvmti/jvmti_agent.c
index 7d96581..ca064be 100644
--- a/tools/perf/jvmti/jvmti_agent.c
+++ b/tools/perf/jvmti/jvmti_agent.c
@@ -373,20 +373,20 @@ jvmti_write_code(void *agent, char const *sym,
 
 int
 jvmti_write_debug_info(void *agent, uint64_t code, const char *file,
-		       jvmtiAddrLocationMap const *map,
-		       jvmtiLineNumberEntry *li, jint num)
+		       jvmti_line_info_t *li, int nr_lines)
 {
-	static const char *prev_str = "\xff";
 	struct jr_code_debug_info rec;
 	size_t sret, len, size, flen;
 	size_t padding_count;
+	uint64_t addr;
+	const char *fn = file;
 	FILE *fp = agent;
 	int i;
 
 	/*
 	 * no entry to write
 	 */
-	if (!num)
+	if (!nr_lines)
 		return 0;
 
 	if (!fp) {
@@ -400,17 +400,18 @@ jvmti_write_debug_info(void *agent, uint64_t code, const char *file,
 	size            = sizeof(rec);
 	rec.p.timestamp = perf_get_timestamp();
 	rec.code_addr   = (uint64_t)(uintptr_t)code;
-	rec.nr_entry    = num;
+	rec.nr_entry    = nr_lines;
 
 	/*
 	 * on disk source line info layout:
 	 * uint64_t : addr
 	 * int      : line number
+	 * int      : column discriminator
 	 * file[]   : source file name
 	 * padding  : pad to multiple of 8 bytes
 	 */
-	size += num * (sizeof(uint64_t) + sizeof(int));
-	size += flen + (num - 1) * 2;
+	size += nr_lines * sizeof(struct debug_entry);
+	size += flen * nr_lines;
 	/*
 	 * pad to 8 bytes
 	 */
@@ -428,28 +429,27 @@ jvmti_write_debug_info(void *agent, uint64_t code, const char *file,
 	if (sret != 1)
 		goto error;
 
-	for (i = 0; i < num; i++) {
-		uint64_t addr;
+	for (i = 0; i < nr_lines; i++) {
 
-		addr = (uint64_t)map[i].start_address;
+		addr = (uint64_t)li[i].pc;
 		len  = sizeof(addr);
 		sret = fwrite_unlocked(&addr, len, 1, fp);
 		if (sret != 1)
 			goto error;
 
-		len  = sizeof(int);
+		len  = sizeof(li[0].line_number);
 		sret = fwrite_unlocked(&li[i].line_number, len, 1, fp);
 		if (sret != 1)
 			goto error;
 
-		if (i == 0) {
-			sret = fwrite_unlocked(file, flen, 1, fp);
-		} else {
-			sret = fwrite_unlocked(prev_str, 2, 1, fp);
-		}
+		len  = sizeof(li[0].discrim);
+		sret = fwrite_unlocked(&li[i].discrim, len, 1, fp);
 		if (sret != 1)
 			goto error;
 
+		sret = fwrite_unlocked(fn, flen, 1, fp);
+		if (sret != 1)
+			goto error;
 	}
 	if (padding_count)
 		sret = fwrite_unlocked(pad_bytes, padding_count, 1, fp);
diff --git a/tools/perf/jvmti/jvmti_agent.h b/tools/perf/jvmti/jvmti_agent.h
index 8251a1c..bedf5d0 100644
--- a/tools/perf/jvmti/jvmti_agent.h
+++ b/tools/perf/jvmti/jvmti_agent.h
@@ -11,16 +11,23 @@
 extern "C" {
 #endif
 
+typedef struct {
+	unsigned long	pc;
+	int		line_number;
+	int		discrim; /* discriminator -- 0 for now */
+} jvmti_line_info_t;
+
 void *jvmti_open(void);
 int   jvmti_close(void *agent);
 int   jvmti_write_code(void *agent, char const *symbol_name,
 		       uint64_t vma, void const *code,
 		       const unsigned int code_size);
+
 int   jvmti_write_debug_info(void *agent,
 		             uint64_t code,
 			     const char *file,
-			     jvmtiAddrLocationMap const *map,
-			     jvmtiLineNumberEntry *tab, jint nr);
+			     jvmti_line_info_t *li,
+			     int nr_lines);
 
 #if defined(__cplusplus)
 }
diff --git a/tools/perf/jvmti/libjvmti.c b/tools/perf/jvmti/libjvmti.c
index 745f20c..6ee98b0 100644
--- a/tools/perf/jvmti/libjvmti.c
+++ b/tools/perf/jvmti/libjvmti.c
@@ -4,6 +4,7 @@
 #include <stdlib.h>
 #include <err.h>
 #include <jvmti.h>
+#include <jvmticmlr.h>
 #include <limits.h>
 
 #include "jvmti_agent.h"
@@ -11,6 +12,100 @@
 static int has_line_numbers;
 void *jvmti_agent;
 
+static jvmtiError
+do_get_line_numbers(jvmtiEnv *jvmti, void *pc, jmethodID m, jint bci,
+		    jvmti_line_info_t *tab, jint *nr)
+{
+	jint i, lines = 0;
+	jint nr_lines = 0;
+	jvmtiLineNumberEntry *loc_tab = NULL;
+	jvmtiError ret;
+
+	ret = (*jvmti)->GetLineNumberTable(jvmti, m, &nr_lines, &loc_tab);
+	if (ret != JVMTI_ERROR_NONE)
+		return ret;
+
+	for (i = 0; i < nr_lines; i++) {
+		if (loc_tab[i].start_location < bci) {
+			tab[lines].pc = (unsigned long)pc;
+			tab[lines].line_number = loc_tab[i].line_number;
+			tab[lines].discrim = 0; /* not yet used */
+			lines++;
+		} else {
+			break;
+		}
+	}
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)loc_tab);
+	*nr = lines;
+	return JVMTI_ERROR_NONE;
+}
+
+static jvmtiError
+get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **tab, int *nr_lines)
+{
+	const jvmtiCompiledMethodLoadRecordHeader *hdr;
+	jvmtiCompiledMethodLoadInlineRecord *rec;
+	jvmtiLineNumberEntry *lne = NULL;
+	PCStackInfo *c;
+	jint nr, ret;
+	int nr_total = 0;
+	int i, lines_total = 0;
+
+	if (!(tab && nr_lines))
+		return JVMTI_ERROR_NULL_POINTER;
+
+	/*
+	 * Phase 1 -- get the number of lines necessary
+	 */
+	for (hdr = compile_info; hdr != NULL; hdr = hdr->next) {
+		if (hdr->kind == JVMTI_CMLR_INLINE_INFO) {
+			rec = (jvmtiCompiledMethodLoadInlineRecord *)hdr;
+			for (i = 0; i < rec->numpcs; i++) {
+				c = rec->pcinfo + i;
+				nr = 0;
+				/*
+				 * unfortunately, need a tab to get the number of lines!
+				 */
+				ret = (*jvmti)->GetLineNumberTable(jvmti, c->methods[0], &nr, &lne);
+				if (ret == JVMTI_ERROR_NONE) {
+					/* free what was allocated for nothing */
+					(*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
+					nr_total += (int)nr;
+				}
+			}
+		}
+	}
+
+	if (nr_total == 0)
+		return JVMTI_ERROR_NOT_FOUND;
+
+	/*
+	 * Phase 2 -- allocate big enough line table
+	 */
+	*tab = malloc(nr_total * sizeof(**tab));
+	if (!*tab)
+		return JVMTI_ERROR_OUT_OF_MEMORY;
+
+	for (hdr = compile_info; hdr != NULL; hdr = hdr->next) {
+		if (hdr->kind == JVMTI_CMLR_INLINE_INFO) {
+			rec = (jvmtiCompiledMethodLoadInlineRecord *)hdr;
+			for (i = 0; i < rec->numpcs; i++) {
+				c = rec->pcinfo + i;
+				nr = 0;
+				ret = do_get_line_numbers(jvmti, c->pc,
+							  c->methods[0],
+							  c->bcis[0],
+							  *tab + lines_total,
+							  &nr);
+				if (ret == JVMTI_ERROR_NONE)
+					lines_total += nr;
+			}
+		}
+	}
+	*nr_lines = lines_total;
+	return JVMTI_ERROR_NONE;
+}
+
 static void JNICALL
 compiled_method_load_cb(jvmtiEnv *jvmti,
 			jmethodID method,
@@ -18,9 +113,9 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 			void const *code_addr,
 			jint map_length,
 			jvmtiAddrLocationMap const *map,
-			void const *compile_info __unused)
+			const void *compile_info)
 {
-	jvmtiLineNumberEntry *tab = NULL;
+	jvmti_line_info_t *line_tab = NULL;
 	jclass decl_class;
 	char *class_sign = NULL;
 	char *func_name = NULL;
@@ -29,7 +124,7 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	char fn[PATH_MAX];
 	uint64_t addr = (uint64_t)(uintptr_t)code_addr;
 	jvmtiError ret;
-	jint nr_lines = 0;
+	int nr_lines = 0; /* in line_tab[] */
 	size_t len;
 
 	ret = (*jvmti)->GetMethodDeclaringClass(jvmti, method,
@@ -40,19 +135,19 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	}
 
 	if (has_line_numbers && map && map_length) {
-
-		ret = (*jvmti)->GetLineNumberTable(jvmti, method, &nr_lines, &tab);
+		ret = get_line_numbers(jvmti, compile_info, &line_tab, &nr_lines);
 		if (ret != JVMTI_ERROR_NONE) {
 			warnx("jvmti: cannot get line table for method");
-		} else {
-			ret = (*jvmti)->GetSourceFileName(jvmti, decl_class, &file_name);
-			if (ret != JVMTI_ERROR_NONE) {
-				warnx("jvmti: cannot get source filename ret=%d", ret);
-				nr_lines = 0;
-			}
+			nr_lines = 0;
 		}
 	}
 
+	ret = (*jvmti)->GetSourceFileName(jvmti, decl_class, &file_name);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: cannot get source filename ret=%d", ret);
+		goto error;
+	}
+
 	ret = (*jvmti)->GetClassSignature(jvmti, decl_class,
 					  &class_sign, NULL);
 	if (ret != JVMTI_ERROR_NONE) {
@@ -92,13 +187,14 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	/*
 	 * write source line info record if we have it
 	 */
-	if (jvmti_write_debug_info(jvmti_agent, addr, fn, map, tab, nr_lines))
+	if (jvmti_write_debug_info(jvmti_agent, addr, fn, line_tab, nr_lines))
 		warnx("jvmti: write_debug_info() failed");
 
 	len = strlen(func_name) + strlen(class_sign) + strlen(func_sign) + 2;
 	{
 		char str[len];
 		snprintf(str, len, "%s%s%s", class_sign, func_name, func_sign);
+
 		if (jvmti_write_code(jvmti_agent, str, addr, code_addr, code_size))
 			warnx("jvmti: write_code() failed");
 	}
@@ -106,8 +202,8 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	(*jvmti)->Deallocate(jvmti, (unsigned char *)func_name);
 	(*jvmti)->Deallocate(jvmti, (unsigned char *)func_sign);
 	(*jvmti)->Deallocate(jvmti, (unsigned char *)class_sign);
-	(*jvmti)->Deallocate(jvmti, (unsigned char *)tab);
 	(*jvmti)->Deallocate(jvmti, (unsigned char *)file_name);
+	free(line_tab);
 }
 
 static void JNICALL
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 9fd4906..b6f8c96 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -110,6 +110,7 @@ libperf-$(CONFIG_LZMA) += lzma.o
 libperf-y += demangle-java.o
 libperf-y += jitdump.o
 libperf-y += genelf.o
+libperf-y += genelf_debug.o
 
 CFLAGS_config.o   += -DETC_PERFCONFIG="BUILD_STR($(ETC_PERFCONFIG_SQ))"
 CFLAGS_exec_cmd.o += -DPERF_EXEC_PATH="BUILD_STR($(perfexecdir_SQ))" -DPREFIX="BUILD_STR($(prefix_SQ))"
diff --git a/tools/perf/util/genelf.c b/tools/perf/util/genelf.c
index 3832cd0..6ce8cf3 100644
--- a/tools/perf/util/genelf.c
+++ b/tools/perf/util/genelf.c
@@ -156,7 +156,8 @@ gen_build_id(struct buildid_note *note, unsigned long load_addr, const void *cod
  */
 int
 jit_write_elf(int fd, uint64_t load_addr, const char *sym,
-	      const void *code, int csize)
+	      const void *code, int csize,
+	      void *debug, int nr_debug_entries)
 {
 	Elf *e;
 	Elf_Data *d;
@@ -384,9 +385,15 @@ jit_write_elf(int fd, uint64_t load_addr, const char *sym,
 	shdr->sh_size = sizeof(bnote);
 	shdr->sh_entsize = 0;
 
-	if (elf_update(e, ELF_C_WRITE) < 0) {
-		warnx("elf_update 4 failed");
-		goto error;
+	if (debug && nr_debug_entries) {
+		retval = jit_add_debug_info(e, load_addr, debug, nr_debug_entries);
+		if (retval)
+			goto error;
+	} else {
+		if (elf_update(e, ELF_C_WRITE) < 0) {
+			warnx("elf_update 4 failed");
+			goto error;
+		}
 	}
 
 	retval = 0;
diff --git a/tools/perf/util/genelf.h b/tools/perf/util/genelf.h
index 79b89e2..957e3de 100644
--- a/tools/perf/util/genelf.h
+++ b/tools/perf/util/genelf.h
@@ -3,7 +3,11 @@
 
 /* genelf.c */
 extern int jit_write_elf(int fd, uint64_t code_addr, const char *sym,
-			 const void *code, int csize);
+			 const void *code, int csize,
+			 void *debug, int nr_debug_entries);
+/* genelf_debug.c */
+extern int jit_add_debug_info(Elf *e, uint64_t code_addr,
+			      void *debug, int nr_debug_entries);
 
 #if   defined(__arm__)
 #define GEN_ELF_ARCH	EM_ARM
diff --git a/tools/perf/util/genelf_debug.c b/tools/perf/util/genelf_debug.c
new file mode 100644
index 0000000..1ff5a46
--- /dev/null
+++ b/tools/perf/util/genelf_debug.c
@@ -0,0 +1,610 @@
+/*
+ * genelf_debug.c
+ * Copyright (C) 2015, Google, Inc
+ *
+ * Contributed by:
+ * 	Stephane Eranian <eranian@google.com>
+ *
+ * Released under the GPL v2.
+ *
+ * based on GPLv2 source code from Oprofile
+ * @remark Copyright 2007 OProfile authors
+ * @author Philippe Elie
+ */
+#include <sys/types.h>
+#include <stdio.h>
+#include <getopt.h>
+#include <stddef.h>
+#include <libelf.h>
+#include <string.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <fcntl.h>
+#include <err.h>
+#include <dwarf.h>
+
+#include "perf.h"
+#include "genelf.h"
+#include "../util/jitdump.h"
+
+#define BUFFER_EXT_DFL_SIZE	(4 * 1024)
+
+typedef uint32_t uword;
+typedef uint16_t uhalf;
+typedef int32_t  sword;
+typedef int16_t  shalf;
+typedef uint8_t  ubyte;
+typedef int8_t   sbyte;
+
+struct buffer_ext {
+	size_t cur_pos;
+	size_t max_sz;
+	void *data;
+};
+
+static void
+buffer_ext_dump(struct buffer_ext *be, const char *msg)
+{
+	size_t i;
+	warnx("DUMP for %s", msg);
+	for (i = 0 ; i < be->cur_pos; i++)
+		warnx("%4zu 0x%02x", i, (((char *)be->data)[i]) & 0xff);
+}
+
+static inline int
+buffer_ext_add(struct buffer_ext *be, void *addr, size_t sz)
+{
+	void *tmp;
+	size_t be_sz = be->max_sz;
+
+retry:
+	if ((be->cur_pos + sz) < be_sz) {
+		memcpy(be->data + be->cur_pos, addr, sz);
+		be->cur_pos += sz;
+		return 0;
+	}
+
+	if (!be_sz)
+		be_sz = BUFFER_EXT_DFL_SIZE;
+	else
+		be_sz <<= 1;
+
+	tmp = realloc(be->data, be_sz);
+	if (!tmp)
+		return -1;
+
+	be->data   = tmp;
+	be->max_sz = be_sz;
+
+	goto retry;
+}
+
+static void
+buffer_ext_init(struct buffer_ext *be)
+{
+	be->data = NULL;
+	be->cur_pos = 0;
+	be->max_sz = 0;
+}
+
+static inline size_t
+buffer_ext_size(struct buffer_ext *be)
+{
+	return be->cur_pos;
+}
+
+static inline void *
+buffer_ext_addr(struct buffer_ext *be)
+{
+	return be->data;
+}
+
+struct debug_line_header {
+	// Not counting this field
+	uword total_length;
+	// version number (2 currently)
+	uhalf version;
+	// relative offset from next field to
+	// program statement
+	uword prolog_length;
+	ubyte minimum_instruction_length;
+	ubyte default_is_stmt;
+	// line_base - see DWARF 2 specs
+	sbyte line_base;
+	// line_range - see DWARF 2 specs
+	ubyte line_range;
+	// number of opcode + 1
+	ubyte opcode_base;
+	/* follow the array of opcode args nr: ubytes [nr_opcode_base] */
+	/* follow the search directories index, zero terminated string
+	 * terminated by an empty string.
+	 */
+	/* follow an array of { filename, LEB128, LEB128, LEB128 }, first is
+	 * the directory index entry, 0 means current directory, then mtime
+	 * and filesize, last entry is followed by en empty string.
+	 */
+	/* follow the first program statement */
+} __attribute__((packed));
+
+/* DWARF 2 spec talk only about one possible compilation unit header while
+ * binutils can handle two flavours of dwarf 2, 32 and 64 bits, this is not
+ * related to the used arch, an ELF 32 can hold more than 4 Go of debug
+ * information. For now we handle only DWARF 2 32 bits comp unit. It'll only
+ * become a problem if we generate more than 4GB of debug information.
+ */
+struct compilation_unit_header {
+	uword total_length;
+	uhalf version;
+	uword debug_abbrev_offset;
+	ubyte pointer_size;
+} __attribute__((packed));
+
+#define DW_LNS_num_opcode (DW_LNS_set_isa + 1)
+
+/* field filled at run time are marked with -1 */
+static struct debug_line_header const default_debug_line_header = {
+	.total_length = -1,
+	.version = 2,
+	.prolog_length = -1,
+	.minimum_instruction_length = 1,	/* could be better when min instruction size != 1 */
+	.default_is_stmt = 1,	/* we don't take care about basic block */
+	.line_base = -5,	/* sensible value for line base ... */
+	.line_range = -14,     /* ... and line range are guessed statically */
+	.opcode_base = DW_LNS_num_opcode
+};
+
+static ubyte standard_opcode_length[] =
+{
+	0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 1
+};
+#if 0
+{
+	[DW_LNS_advance_pc]   = 1,
+	[DW_LNS_advance_line] = 1,
+	[DW_LNS_set_file] =  1,
+	[DW_LNS_set_column] = 1,
+	[DW_LNS_fixed_advance_pc] = 1,
+	[DW_LNS_set_isa] = 1,
+};
+#endif
+
+/* field filled at run time are marked with -1 */
+static struct compilation_unit_header default_comp_unit_header = {
+	.total_length = -1,
+	.version = 2,
+	.debug_abbrev_offset = 0,     /* we reuse the same abbrev entries for all comp unit */
+	.pointer_size = sizeof(void *)
+};
+
+static void emit_uword(struct buffer_ext *be, uword data)
+{
+	buffer_ext_add(be, &data, sizeof(uword));
+}
+
+static void emit_string(struct buffer_ext *be, const char *s)
+{
+	buffer_ext_add(be, (void *)s, strlen(s) + 1);
+}
+
+static void emit_unsigned_LEB128(struct buffer_ext *be,
+				 unsigned long data)
+{
+	do {
+		ubyte cur = data & 0x7F;
+		data >>= 7;
+		if (data)
+			cur |= 0x80;
+		buffer_ext_add(be, &cur, 1);
+	} while (data);
+}
+
+static void emit_signed_LEB128(struct buffer_ext *be, long data)
+{
+	int more = 1;
+	int negative = data < 0;
+	int size = sizeof(long) * CHAR_BIT;
+	while (more) {
+		ubyte cur = data & 0x7F;
+		data >>= 7;
+		if (negative)
+			data |= - (1 << (size - 7));
+		if ((data == 0 && !(cur & 0x40)) ||
+		    (data == -1l && (cur & 0x40)))
+			more = 0;
+		else
+			cur |= 0x80;
+		buffer_ext_add(be, &cur, 1);
+	}
+}
+
+static void emit_extended_opcode(struct buffer_ext *be, ubyte opcode,
+				 void *data, size_t data_len)
+{
+	buffer_ext_add(be, (char *)"", 1);
+
+	emit_unsigned_LEB128(be, data_len + 1);
+
+	buffer_ext_add(be, &opcode, 1);
+	buffer_ext_add(be, data, data_len);
+}
+
+static void emit_opcode(struct buffer_ext *be, ubyte opcode)
+{
+	buffer_ext_add(be, &opcode, 1);
+}
+
+static void emit_opcode_signed(struct buffer_ext  *be,
+			       ubyte opcode, long data)
+{
+	buffer_ext_add(be, &opcode, 1);
+	emit_signed_LEB128(be, data);
+}
+
+static void emit_opcode_unsigned(struct buffer_ext *be, ubyte opcode, 
+				 unsigned long data)
+{
+	buffer_ext_add(be, &opcode, 1);
+	emit_unsigned_LEB128(be, data);
+}
+
+static void emit_advance_pc(struct buffer_ext *be, unsigned long delta_pc)
+{
+	emit_opcode_unsigned(be, DW_LNS_advance_pc, delta_pc);
+}
+
+static void emit_advance_lineno(struct buffer_ext  *be, long delta_lineno)
+{
+	emit_opcode_signed(be, DW_LNS_advance_line, delta_lineno);
+}
+
+static void emit_lne_end_of_sequence(struct buffer_ext *be)
+{
+	emit_extended_opcode(be, DW_LNE_end_sequence, NULL, 0);
+}
+
+static void emit_set_file(struct buffer_ext *be, unsigned long index)
+{
+	emit_opcode_unsigned(be, DW_LNS_set_file, index);
+}
+
+static void emit_lne_define_filename(struct buffer_ext *be,
+				     const char *filename)
+{
+	buffer_ext_add(be, (void *)"", 1);
+
+	/* LNE field, strlen(filename) + zero termination, 3 bytes for: the dir entry, timestamp, filesize */
+	emit_unsigned_LEB128(be, strlen(filename) + 5);
+	emit_opcode(be, DW_LNE_define_file);
+	emit_string(be, filename);
+	/* directory index 0=do not know */
+        emit_unsigned_LEB128(be, 0);
+	/* last modification date on file 0=do not know */
+        emit_unsigned_LEB128(be, 0);
+	/* filesize 0=do not know */
+        emit_unsigned_LEB128(be, 0);
+}
+
+static void emit_lne_set_address(struct buffer_ext *be,
+				 void *address)
+{
+	emit_extended_opcode(be, DW_LNE_set_address, &address, sizeof(unsigned long));
+}
+
+static ubyte get_special_opcode(struct debug_entry *ent,
+				unsigned int last_line,
+				unsigned long last_vma)
+{
+	unsigned int temp;
+	unsigned long delta_addr;
+
+	/*
+	 * delta from line_base
+	 */
+	temp = (ent->lineno - last_line) - default_debug_line_header.line_base;
+
+	if (temp >= default_debug_line_header.line_range)
+		return 0;
+
+	/*
+	 * delta of addresses
+	 */
+	delta_addr = (ent->addr - last_vma) / default_debug_line_header.minimum_instruction_length;
+
+	/* This is not sufficient to ensure opcode will be in [0-256] but
+	 * sufficient to ensure when summing with the delta lineno we will
+	 * not overflow the unsigned long opcode */
+
+	if (delta_addr <= 256 / default_debug_line_header.line_range) {
+		unsigned long opcode = temp +
+			(delta_addr * default_debug_line_header.line_range) +
+			default_debug_line_header.opcode_base;
+
+		return opcode <= 255 ? opcode : 0;
+	}
+	return 0;
+}
+
+static void emit_lineno_info(struct buffer_ext *be,
+			     struct debug_entry *ent, size_t nr_entry,
+			     unsigned long code_addr)
+{
+	size_t i;
+
+	/*
+	 * Machine state at start of a statement program
+	 * address = 0
+	 * file    = 1
+	 * line    = 1
+	 * column  = 0
+	 * is_stmt = default_is_stmt as given in the debug_line_header
+	 * basic block = 0
+	 * end sequence = 0
+	 */
+
+	/* start state of the state machine we take care of */
+	unsigned long last_vma = code_addr;
+	char const  *cur_filename = NULL;
+	unsigned long cur_file_idx = 0;
+	int last_line = 1;
+
+	emit_lne_set_address(be, (void *)code_addr);
+
+	for (i = 0; i < nr_entry; i++, ent = debug_entry_next(ent)) {
+		int need_copy = 0;
+		ubyte special_opcode;
+
+		/*
+		 * check if filename changed, if so add it
+		 */
+		if (!cur_filename || strcmp(cur_filename, ent->name)) {
+			emit_lne_define_filename(be, ent->name);
+			cur_filename = ent->name;
+			emit_set_file(be, ++cur_file_idx);
+			need_copy = 1;
+		}
+
+		special_opcode = get_special_opcode(ent, last_line, last_vma);
+		if (special_opcode != 0) {
+			last_line = ent->lineno;
+			last_vma  = ent->addr;
+			emit_opcode(be, special_opcode);
+		} else {
+			/*
+			 * lines differ, emit line delta
+			 */
+			if (last_line != ent->lineno) {
+				emit_advance_lineno(be, ent->lineno - last_line);
+				last_line = ent->lineno;
+				need_copy = 1;
+			}
+			/*
+			 * addresses differ, emit address delta
+			 */
+			if (last_vma != ent->addr) {
+				emit_advance_pc(be, ent->addr - last_vma);
+				last_vma = ent->addr;
+				need_copy = 1;
+			}
+			/*
+			 * add new row to matrix
+			 */
+			if (need_copy)
+				emit_opcode(be, DW_LNS_copy);
+		}
+	}
+}
+
+static void add_debug_line(struct buffer_ext *be,
+	struct debug_entry *ent, size_t nr_entry,
+	unsigned long code_addr)
+{
+	struct debug_line_header * dbg_header;
+	size_t old_size;
+
+	old_size = buffer_ext_size(be);
+
+	buffer_ext_add(be, (void *)&default_debug_line_header,
+		 sizeof(default_debug_line_header));
+
+	buffer_ext_add(be, &standard_opcode_length,  sizeof(standard_opcode_length));
+
+	// empty directory entry
+	buffer_ext_add(be, (void *)"", 1);
+
+	// empty filename directory
+	buffer_ext_add(be, (void *)"", 1);
+
+	dbg_header = buffer_ext_addr(be) + old_size;
+	dbg_header->prolog_length = (buffer_ext_size(be) - old_size) -
+		offsetof(struct debug_line_header, minimum_instruction_length);
+
+	emit_lineno_info(be, ent, nr_entry, code_addr);
+
+	emit_lne_end_of_sequence(be);
+
+	dbg_header = buffer_ext_addr(be) + old_size;
+	dbg_header->total_length = (buffer_ext_size(be) - old_size) -
+		offsetof(struct debug_line_header, version);
+}
+
+static void
+add_debug_abbrev(struct buffer_ext *be)
+{
+        emit_unsigned_LEB128(be, 1);
+        emit_unsigned_LEB128(be, DW_TAG_compile_unit);
+        emit_unsigned_LEB128(be, DW_CHILDREN_yes);
+        emit_unsigned_LEB128(be, DW_AT_stmt_list);
+        emit_unsigned_LEB128(be, DW_FORM_data4);
+        emit_unsigned_LEB128(be, 0);
+        emit_unsigned_LEB128(be, 0);
+        emit_unsigned_LEB128(be, 0);
+}
+
+static void
+add_compilation_unit(struct buffer_ext *be,
+		     size_t offset_debug_line)
+{
+	struct compilation_unit_header *comp_unit_header;
+	size_t old_size = buffer_ext_size(be);
+
+	buffer_ext_add(be, &default_comp_unit_header,
+		       sizeof(default_comp_unit_header));
+
+	emit_unsigned_LEB128(be, 1);
+	emit_uword(be, offset_debug_line);
+
+	comp_unit_header = buffer_ext_addr(be) + old_size;
+	comp_unit_header->total_length = (buffer_ext_size(be) - old_size) -
+		offsetof(struct compilation_unit_header, version);
+}
+
+static int
+jit_process_debug_info(uint64_t code_addr,
+		       void *debug, int nr_debug_entries,
+		       struct buffer_ext *dl,
+		       struct buffer_ext *da,
+		       struct buffer_ext *di)
+{
+	struct debug_entry *ent = debug;
+	int i;
+
+	for (i = 0; i < nr_debug_entries; i++) {
+		ent->addr = ent->addr - code_addr;
+		ent = debug_entry_next(ent);
+	}
+	add_compilation_unit(di, buffer_ext_size(dl));
+	add_debug_line(dl, debug, nr_debug_entries, 0);
+	add_debug_abbrev(da);
+	if (0) buffer_ext_dump(da, "abbrev");
+
+	return 0;
+}
+
+int
+jit_add_debug_info(Elf *e, uint64_t code_addr, void *debug, int nr_debug_entries)
+{
+	Elf_Data *d;
+	Elf_Scn *scn;
+	Elf_Shdr *shdr;
+	struct buffer_ext dl, di, da;
+	int ret;
+
+	buffer_ext_init(&dl);
+	buffer_ext_init(&di);
+	buffer_ext_init(&da);
+	
+	ret = jit_process_debug_info(code_addr, debug, nr_debug_entries, &dl, &da, &di);
+	if (ret)
+		return -1;
+	/*
+	 * setup .debug_line section
+	 */
+	scn = elf_newscn(e);
+	if (!scn) {
+		warnx("cannot create section");
+		return -1;
+	}
+
+	d = elf_newdata(scn);
+	if (!d) {
+		warnx("cannot get new data");
+		return -1;
+	}
+
+	d->d_align = 1;
+	d->d_off = 0LL;
+	d->d_buf = buffer_ext_addr(&dl);
+	d->d_type = ELF_T_BYTE;
+	d->d_size = buffer_ext_size(&dl);
+	d->d_version = EV_CURRENT;
+
+	shdr = elf_getshdr(scn);
+	if (!shdr) {
+		warnx("cannot get section header");
+		return -1;
+	}
+
+	shdr->sh_name = 52; /* .debug_line */
+	shdr->sh_type = SHT_PROGBITS;
+	shdr->sh_addr = 0; /* must be zero or == sh_offset -> dynamic object */
+	shdr->sh_flags = 0;
+	shdr->sh_entsize = 0;
+
+	/*
+	 * setup .debug_info section
+	 */
+	scn = elf_newscn(e);
+	if (!scn) {
+		warnx("cannot create section");
+		return -1;
+	}
+
+	d = elf_newdata(scn);
+	if (!d) {
+		warnx("cannot get new data");
+		return -1;
+	}
+
+	d->d_align = 1;
+	d->d_off = 0LL;
+	d->d_buf = buffer_ext_addr(&di);
+	d->d_type = ELF_T_BYTE;
+	d->d_size = buffer_ext_size(&di);
+	d->d_version = EV_CURRENT;
+
+	shdr = elf_getshdr(scn);
+	if (!shdr) {
+		warnx("cannot get section header");
+		return -1;
+	}
+
+	shdr->sh_name = 64; /* .debug_info */
+	shdr->sh_type = SHT_PROGBITS;
+	shdr->sh_addr = 0; /* must be zero or == sh_offset -> dynamic object */
+	shdr->sh_flags = 0;
+	shdr->sh_entsize = 0;
+
+	/*
+	 * setup .debug_abbrev section
+	 */
+	scn = elf_newscn(e);
+	if (!scn) {
+		warnx("cannot create section");
+		return -1;
+	}
+
+	d = elf_newdata(scn);
+	if (!d) {
+		warnx("cannot get new data");
+		return -1;
+	}
+
+	d->d_align = 1;
+	d->d_off = 0LL;
+	d->d_buf = buffer_ext_addr(&da);
+	d->d_type = ELF_T_BYTE;
+	d->d_size = buffer_ext_size(&da);
+	d->d_version = EV_CURRENT;
+
+	shdr = elf_getshdr(scn);
+	if (!shdr) {
+		warnx("cannot get section header");
+		return -1;
+	}
+
+	shdr->sh_name = 76; /* .debug_info */
+	shdr->sh_type = SHT_PROGBITS;
+	shdr->sh_addr = 0; /* must be zero or == sh_offset -> dynamic object */
+	shdr->sh_flags = 0;
+	shdr->sh_entsize = 0;
+
+	/*
+	 * now we update the ELF image with all the sections
+	 */
+	if (elf_update(e, ELF_C_WRITE) < 0) {
+		warnx("elf_update debug failed");
+		return -1;
+	}
+	return 0;
+}
diff --git a/tools/perf/util/jitdump.c b/tools/perf/util/jitdump.c
index e910130..42b967b 100644
--- a/tools/perf/util/jitdump.c
+++ b/tools/perf/util/jitdump.c
@@ -63,7 +63,9 @@ jit_emit_elf(char *filename,
 	     const char *sym,
 	     uint64_t code_addr,
 	     const void *code,
-	     int csize)
+	     int csize,
+	     void *debug,
+	     int nr_debug_entries)
 {
 	int ret, fd;
 
@@ -76,7 +78,7 @@ jit_emit_elf(char *filename,
 		return -1;
 	}
 
-        ret = jit_write_elf(fd, code_addr, sym, (const void *)code, csize);
+        ret = jit_write_elf(fd, code_addr, sym, (const void *)code, csize, debug, nr_debug_entries);
 
         close(fd);
 
@@ -341,7 +343,7 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
 
 	size = PERF_ALIGN(size, sizeof(u64));
 
-	ret = jit_emit_elf(filename, sym, addr, (const void *)code, csize);
+	ret = jit_emit_elf(filename, sym, addr, (const void *)code, csize, jd->debug_data, jd->nr_debug_entries);
 
 	if (jd->debug_data && jd->nr_debug_entries) {
 		free(jd->debug_data);
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1237042 — [PATCH v7 4/5] perf tools: add JVMTI agent library

FromStephane Eranian <eranian@google.com>
Date2015-10-01 08:50 +0200
Subject[PATCH v7 4/5] perf tools: add JVMTI agent library
Message-ID<qeFQK-4lA-15@gated-at.bofh.it>
In reply to#1237039
This is a standalone JVMTI library to help  profile Java jitted
code with perf record/perf report. The library is not installed
or compiled automatically by perf Makefile. It is not used
directly by perf. It is arch agnostic and has been tested on
X86 and ARM. It needs to be used with a Java runtime, such
as OpenJDK, as follows:

$ java -agentpath:libjvmti.so .......

When used this way, java will generate a jitdump binary file in
$HOME/.debug/java/jit/java-jit-*

This binary dump file contains information to help symbolize and
annotate jitted code.

The jitdump information must be injected into the perf.data file
using:
$ perf inject --jit -i perf.data -o perf.data.jitted

This injects the MMAP records to cover the jitted code and also generates
one ELF image for each jitted function. The ELF images are created in the
same subdir as the jitdump file. The MMAP records point there too.

Then, to visualize the function or asm profile, simply use the regular
perf commands:
$ perf report -i perf.data.jitted
or
$ perf annotate -i perf.data.jitted

JVMTI agent code adapted from the OProfile's opagent code.

This version of the JVMTI agent is using the CLOCK_MONOTIC
as the time source to timestamp jit samples. To correlate
with perf_events samples, it needs to run on kernel 4.0.0-rc5+
or later with the following commit from Peter Zijlstra:
   34f4392 perf: Add per event clockid support

With this patch recording jitted code is done as follows:
   $ perf record -k mono -- java -agentpath:libjvmti.so .......

Signed-off-by: Stephane Eranian <eranian@google.com>
---
 tools/perf/jvmti/Makefile      |  70 +++++++
 tools/perf/jvmti/jvmti_agent.c | 464 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/jvmti/jvmti_agent.h |  29 +++
 tools/perf/jvmti/libjvmti.c    | 208 ++++++++++++++++++
 4 files changed, 771 insertions(+)
 create mode 100644 tools/perf/jvmti/Makefile
 create mode 100644 tools/perf/jvmti/jvmti_agent.c
 create mode 100644 tools/perf/jvmti/jvmti_agent.h
 create mode 100644 tools/perf/jvmti/libjvmti.c

diff --git a/tools/perf/jvmti/Makefile b/tools/perf/jvmti/Makefile
new file mode 100644
index 0000000..22efd13
--- /dev/null
+++ b/tools/perf/jvmti/Makefile
@@ -0,0 +1,70 @@
+ARCH=$(shell uname -m)
+
+ifeq ($(ARCH), x86_64)
+JARCH=amd64
+endif
+ifeq ($(ARCH), armv7l)
+JARCH=armhf
+endif
+ifeq ($(ARCH), armv6l)
+JARCH=armhf
+endif
+ifeq ($(ARCH), ppc64)
+JARCH=powerpc
+endif
+ifeq ($(ARCH), ppc64le)
+JARCH=powerpc
+endif
+
+DESTDIR=/usr/local
+
+VERSION=1
+REVISION=0
+AGE=0
+
+LN=ln -sf
+RM=rm
+
+SLIBJVMTI=libjvmti.so.$(VERSION).$(REVISION).$(AGE)
+VLIBJVMTI=libjvmti.so.$(VERSION)
+SLDFLAGS=-shared -Wl,-soname -Wl,$(VLIBJVMTI)
+SOLIBEXT=so
+
+JDIR=$(shell /usr/sbin/update-java-alternatives -l | head -1 | cut -d ' ' -f 3)
+# -lrt required in 32-bit mode for clock_gettime()
+LIBS=-lelf -lrt
+INCDIR=-I $(JDIR)/include -I $(JDIR)/include/linux
+
+TARGETS=$(SLIBJVMTI)
+
+SRCS=libjvmti.c jvmti_agent.c
+OBJS=$(SRCS:.c=.o)
+SOBJS=$(OBJS:.o=.lo)
+OPT=-O2 -g -Werror -Wall
+
+CFLAGS=$(INCDIR) $(OPT)
+
+all: $(TARGETS)
+
+.c.o:
+	$(CC) $(CFLAGS) -c $*.c
+.c.lo:
+	$(CC) -fPIC -DPIC $(CFLAGS) -c $*.c -o $*.lo
+
+$(OBJS) $(SOBJS): Makefile jvmti_agent.h ../util/jitdump.h
+
+$(SLIBJVMTI):  $(SOBJS)
+	$(CC) $(CFLAGS) $(SLDFLAGS)  -o $@ $(SOBJS) $(LIBS)
+	$(LN) $@ libjvmti.$(SOLIBEXT)
+
+clean:
+	$(RM) -f *.o *.so.* *.so *.lo
+
+install:
+	-mkdir -p $(DESTDIR)/lib
+	install -m 755 $(SLIBJVMTI) $(DESTDIR)/lib/
+	(cd $(DESTDIR)/lib; $(LN) $(SLIBJVMTI) $(VLIBJVMTI))
+	(cd $(DESTDIR)/lib; $(LN) $(SLIBJVMTI) libjvmti.$(SOLIBEXT))
+	ldconfig
+
+.SUFFIXES: .c .S .o .lo
diff --git a/tools/perf/jvmti/jvmti_agent.c b/tools/perf/jvmti/jvmti_agent.c
new file mode 100644
index 0000000..7d96581
--- /dev/null
+++ b/tools/perf/jvmti/jvmti_agent.c
@@ -0,0 +1,464 @@
+/*
+ * jvmti_agent.c: JVMTI agent interface
+ *
+ * Adapted from the Oprofile code in opagent.c:
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ *
+ * Copyright 2007 OProfile authors
+ * Jens Wilke
+ * Daniel Hansel
+ * Copyright IBM Corporation 2007
+ */
+#include <sys/types.h>
+#include <sys/stat.h> /* for mkdir() */
+#include <stdio.h>
+#include <errno.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <limits.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <syscall.h> /* for gettid() */
+#include <err.h>
+
+#include "jvmti_agent.h"
+#include "../util/jitdump.h"
+
+#define JIT_LANG "java"
+
+static char jit_path[PATH_MAX];
+static void *marker_addr;
+
+/*
+ * padding buffer
+ */
+static const char pad_bytes[7];
+
+static inline pid_t gettid(void)
+{
+	return (pid_t)syscall(__NR_gettid);
+}
+
+static int get_e_machine(struct jitheader *hdr)
+{
+	ssize_t sret;
+	char id[16];
+	int fd, ret = -1;
+	int m = -1;
+	struct {
+		uint16_t e_type;
+		uint16_t e_machine;
+	} info;
+
+	fd = open("/proc/self/exe", O_RDONLY);
+	if (fd == -1)
+		return -1;
+
+	sret = read(fd, id, sizeof(id));
+	if (sret != sizeof(id))
+		goto error;
+
+	/* check ELF signature */
+	if (id[0] != 0x7f || id[1] != 'E' || id[2] != 'L' || id[3] != 'F')
+		goto error;
+
+	sret = read(fd, &info, sizeof(info));
+	if (sret != sizeof(info))
+		goto error;
+
+	m = info.e_machine;
+	if (m < 0)
+		m = 0; /* ELF EM_NONE */
+
+	hdr->elf_mach = m;
+	ret = 0;
+error:
+	close(fd);
+	return ret;
+}
+
+#define NSEC_PER_SEC	1000000000
+static int perf_clk_id = CLOCK_MONOTONIC;
+
+static inline uint64_t
+timespec_to_ns(const struct timespec *ts)
+{
+        return ((uint64_t) ts->tv_sec * NSEC_PER_SEC) + ts->tv_nsec;
+}
+
+static inline uint64_t
+perf_get_timestamp(void)
+{
+	struct timespec ts;
+	int ret;
+
+	ret = clock_gettime(perf_clk_id, &ts);
+	if (ret)
+		return 0;
+
+	return timespec_to_ns(&ts);
+}
+
+static int
+debug_cache_init(void)
+{
+	char str[32];
+	char *base, *p;
+	struct tm tm;
+	time_t t;
+	int ret;
+
+	time(&t);
+	localtime_r(&t, &tm);
+
+	base = getenv("JITDUMPDIR");
+	if (!base)
+		base = getenv("HOME");
+	if (!base)
+		base = ".";
+
+	strftime(str, sizeof(str), JIT_LANG"-jit-%Y%m%d", &tm);
+
+	snprintf(jit_path, PATH_MAX - 1, "%s/.debug/", base);
+
+	ret = mkdir(jit_path, 0755);
+	if (ret == -1) {
+		if (errno != EEXIST) {
+			warn("jvmti: cannot create jit cache dir %s", jit_path);
+			return -1;
+		}
+	}
+
+	snprintf(jit_path, PATH_MAX - 1, "%s/.debug/jit", base);
+	ret = mkdir(jit_path, 0755);
+	if (ret == -1) {
+		if (errno != EEXIST) {
+			warn("cannot create jit cache dir %s", jit_path);
+			return -1;
+		}
+	}
+
+	snprintf(jit_path, PATH_MAX - 1, "%s/.debug/jit/%s.XXXXXXXX", base, str);
+
+	p = mkdtemp(jit_path);
+	if (p != jit_path) {
+		warn("cannot create jit cache dir %s", jit_path);
+		return -1;
+	}
+
+	return 0;
+}
+
+static int
+perf_open_marker_file(int fd)
+{
+	long pgsz;
+
+	pgsz = sysconf(_SC_PAGESIZE);
+	if (pgsz == -1)
+		return -1;
+
+	/*
+	 * we mmap the jitdump to create an MMAP RECORD in perf.data file.
+	 * The mmap is captured either live (perf record running when we mmap)
+	 * or  in deferred mode, via /proc/PID/maps
+	 * the MMAP record is used as a marker of a jitdump file for more meta
+	 * data info about the jitted code. Perf report/annotate detect this
+	 * special filename and process the jitdump file.
+	 *
+	 * mapping must be PROT_EXEC to ensure it is captured by perf record
+	 * even when not using -d option
+	 */
+	marker_addr = mmap(NULL, pgsz, PROT_READ|PROT_EXEC, MAP_PRIVATE, fd, 0);
+	return (marker_addr == MAP_FAILED) ? -1 : 0;
+}
+
+static void
+perf_close_marker_file(void)
+{
+	long pgsz;
+
+	if (!marker_addr)
+		return;
+
+	pgsz = sysconf(_SC_PAGESIZE);
+	if (pgsz == -1)
+		return;
+
+	munmap(marker_addr, pgsz);
+}
+
+void *jvmti_open(void)
+{
+	int pad_cnt;
+	char dump_path[PATH_MAX];
+	struct jitheader header;
+	int fd;
+	FILE *fp;
+
+	/*
+	 * check if clockid is supported
+	 */
+	if (!perf_get_timestamp())
+		warnx("jvmti: kernel does not support %d clock id", perf_clk_id);
+
+	memset(&header, 0, sizeof(header));
+
+	debug_cache_init();
+
+	/*
+	 * jitdump file name
+	 */
+	snprintf(dump_path, PATH_MAX, "%s/jit-%i.dump", jit_path, getpid());
+
+	fd = open(dump_path, O_CREAT|O_TRUNC|O_RDWR, 0666);
+	if (fd == -1)
+		return NULL;
+
+	/*
+	 * create perf.data maker for the jitdump file
+	 */
+	if (perf_open_marker_file(fd)) {
+		warnx("jvmti: failed to create marker file");
+		return NULL;
+	}
+
+	fp = fdopen(fd, "w+");
+	if (!fp) {
+		warn("jvmti: cannot create %s", dump_path);
+		close(fd);
+		goto error;
+	}
+
+	warnx("jvmti: jitdump in %s", dump_path);
+
+	if (get_e_machine(&header)) {
+		warn("get_e_machine failed\n");
+		goto error;
+	}
+
+	header.magic      = JITHEADER_MAGIC;
+	header.version    = JITHEADER_VERSION;
+	header.total_size = sizeof(header);
+	header.pid        = getpid();
+
+	/* calculate amount of padding '\0' */
+	pad_cnt = PADDING_8ALIGNED(header.total_size);
+	header.total_size += pad_cnt;
+
+	header.timestamp = perf_get_timestamp();
+
+	if (!fwrite(&header, sizeof(header), 1, fp)) {
+		warn("jvmti: cannot write dumpfile header");
+		goto error;
+	}
+
+	/* write padding '\0' if necessary */
+	if (pad_cnt && !fwrite(pad_bytes, pad_cnt, 1, fp)) {
+		warn("jvmti: cannot write dumpfile header padding");
+		goto error;
+	}
+
+	return fp;
+error:
+	fclose(fp);
+	return NULL;
+}
+
+int
+jvmti_close(void *agent)
+{
+	struct jr_code_close rec;
+	FILE *fp = agent;
+
+	if (!fp) {
+		warnx("jvmti: incalid fd in close_agent");
+		return -1;
+	}
+
+	rec.p.id = JIT_CODE_CLOSE;
+	rec.p.total_size = sizeof(rec);
+
+	rec.p.timestamp = perf_get_timestamp();
+
+	if (!fwrite(&rec, sizeof(rec), 1, fp))
+		return -1;
+
+	fclose(fp);
+
+	fp = NULL;
+
+	perf_close_marker_file();
+
+	return 0;
+}
+
+int
+jvmti_write_code(void *agent, char const *sym,
+	uint64_t vma, void const *code, unsigned int const size)
+{
+	static int code_generation = 1;
+	struct jr_code_load rec;
+	size_t sym_len;
+	size_t padding_count;
+	FILE *fp = agent;
+	int ret = -1;
+
+	/* don't care about 0 length function, no samples */
+	if (size == 0)
+		return 0;
+
+	if (!fp) {
+		warnx("jvmti: invalid fd in write_native_code");
+		return -1;
+	}
+
+	sym_len = strlen(sym) + 1;
+
+	rec.p.id           = JIT_CODE_LOAD;
+	rec.p.total_size   = sizeof(rec) + sym_len;
+	padding_count      = PADDING_8ALIGNED(rec.p.total_size);
+	rec.p. total_size += padding_count;
+	rec.p.timestamp    = perf_get_timestamp();
+
+	rec.code_size  = size;
+	rec.vma        = vma;
+	rec.code_addr  = vma;
+	rec.pid	       = getpid();
+	rec.tid	       = gettid();
+
+	if (code)
+		rec.p.total_size += size;
+
+	/*
+	 * If JVM is multi-threaded, nultiple concurrent calls to agent
+	 * may be possible, so protect file writes
+	 */
+	flockfile(fp);
+
+	/*
+	 * get code index inside lock to avoid race condition
+	 */
+	rec.code_index = code_generation++;
+
+	ret = fwrite_unlocked(&rec, sizeof(rec), 1, fp);
+	fwrite_unlocked(sym, sym_len, 1, fp);
+	if (code)
+		fwrite_unlocked(code, size, 1, fp);
+
+	if (padding_count)
+		fwrite_unlocked(pad_bytes, padding_count, 1, fp);
+
+	funlockfile(fp);
+
+	ret = 0;
+
+	return ret;
+}
+
+int
+jvmti_write_debug_info(void *agent, uint64_t code, const char *file,
+		       jvmtiAddrLocationMap const *map,
+		       jvmtiLineNumberEntry *li, jint num)
+{
+	static const char *prev_str = "\xff";
+	struct jr_code_debug_info rec;
+	size_t sret, len, size, flen;
+	size_t padding_count;
+	FILE *fp = agent;
+	int i;
+
+	/*
+	 * no entry to write
+	 */
+	if (!num)
+		return 0;
+
+	if (!fp) {
+		warnx("jvmti: invalid fd in write_debug_info");
+		return -1;
+	}
+
+	flen = strlen(file) + 1;
+
+	rec.p.id        = JIT_CODE_DEBUG_INFO;
+	size            = sizeof(rec);
+	rec.p.timestamp = perf_get_timestamp();
+	rec.code_addr   = (uint64_t)(uintptr_t)code;
+	rec.nr_entry    = num;
+
+	/*
+	 * on disk source line info layout:
+	 * uint64_t : addr
+	 * int      : line number
+	 * file[]   : source file name
+	 * padding  : pad to multiple of 8 bytes
+	 */
+	size += num * (sizeof(uint64_t) + sizeof(int));
+	size += flen + (num - 1) * 2;
+	/*
+	 * pad to 8 bytes
+	 */
+	padding_count = PADDING_8ALIGNED(size);
+
+	rec.p.total_size = size + padding_count;
+
+	/*
+	 * If JVM is multi-threaded, nultiple concurrent calls to agent
+	 * may be possible, so protect file writes
+	 */
+	flockfile(fp);
+
+	sret = fwrite_unlocked(&rec, sizeof(rec), 1, fp);
+	if (sret != 1)
+		goto error;
+
+	for (i = 0; i < num; i++) {
+		uint64_t addr;
+
+		addr = (uint64_t)map[i].start_address;
+		len  = sizeof(addr);
+		sret = fwrite_unlocked(&addr, len, 1, fp);
+		if (sret != 1)
+			goto error;
+
+		len  = sizeof(int);
+		sret = fwrite_unlocked(&li[i].line_number, len, 1, fp);
+		if (sret != 1)
+			goto error;
+
+		if (i == 0) {
+			sret = fwrite_unlocked(file, flen, 1, fp);
+		} else {
+			sret = fwrite_unlocked(prev_str, 2, 1, fp);
+		}
+		if (sret != 1)
+			goto error;
+
+	}
+	if (padding_count)
+		sret = fwrite_unlocked(pad_bytes, padding_count, 1, fp);
+		if (sret != 1)
+			goto error;
+
+	funlockfile(fp);
+	return 0;
+error:
+	funlockfile(fp);
+	return -1;
+}
diff --git a/tools/perf/jvmti/jvmti_agent.h b/tools/perf/jvmti/jvmti_agent.h
new file mode 100644
index 0000000..8251a1c
--- /dev/null
+++ b/tools/perf/jvmti/jvmti_agent.h
@@ -0,0 +1,29 @@
+#ifndef __JVMTI_AGENT_H__
+#define __JVMTI_AGENT_H__
+
+#include <sys/types.h>
+#include <stdint.h>
+#include <jvmti.h>
+
+#define __unused __attribute__((unused))
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+void *jvmti_open(void);
+int   jvmti_close(void *agent);
+int   jvmti_write_code(void *agent, char const *symbol_name,
+		       uint64_t vma, void const *code,
+		       const unsigned int code_size);
+int   jvmti_write_debug_info(void *agent,
+		             uint64_t code,
+			     const char *file,
+			     jvmtiAddrLocationMap const *map,
+			     jvmtiLineNumberEntry *tab, jint nr);
+
+#if defined(__cplusplus)
+}
+
+#endif
+#endif /* __JVMTI_H__ */
diff --git a/tools/perf/jvmti/libjvmti.c b/tools/perf/jvmti/libjvmti.c
new file mode 100644
index 0000000..745f20c
--- /dev/null
+++ b/tools/perf/jvmti/libjvmti.c
@@ -0,0 +1,208 @@
+#include <sys/types.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <err.h>
+#include <jvmti.h>
+#include <limits.h>
+
+#include "jvmti_agent.h"
+
+static int has_line_numbers;
+void *jvmti_agent;
+
+static void JNICALL
+compiled_method_load_cb(jvmtiEnv *jvmti,
+			jmethodID method,
+			jint code_size,
+			void const *code_addr,
+			jint map_length,
+			jvmtiAddrLocationMap const *map,
+			void const *compile_info __unused)
+{
+	jvmtiLineNumberEntry *tab = NULL;
+	jclass decl_class;
+	char *class_sign = NULL;
+	char *func_name = NULL;
+	char *func_sign = NULL;
+	char *file_name= NULL;
+	char fn[PATH_MAX];
+	uint64_t addr = (uint64_t)(uintptr_t)code_addr;
+	jvmtiError ret;
+	jint nr_lines = 0;
+	size_t len;
+
+	ret = (*jvmti)->GetMethodDeclaringClass(jvmti, method,
+						&decl_class);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: cannot get declaring class");
+		return;
+	}
+
+	if (has_line_numbers && map && map_length) {
+
+		ret = (*jvmti)->GetLineNumberTable(jvmti, method, &nr_lines, &tab);
+		if (ret != JVMTI_ERROR_NONE) {
+			warnx("jvmti: cannot get line table for method");
+		} else {
+			ret = (*jvmti)->GetSourceFileName(jvmti, decl_class, &file_name);
+			if (ret != JVMTI_ERROR_NONE) {
+				warnx("jvmti: cannot get source filename ret=%d", ret);
+				nr_lines = 0;
+			}
+		}
+	}
+
+	ret = (*jvmti)->GetClassSignature(jvmti, decl_class,
+					  &class_sign, NULL);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: getclassignature failed");
+		goto error;
+	}
+
+	ret = (*jvmti)->GetMethodName(jvmti, method, &func_name,
+				      &func_sign, NULL);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: failed getmethodname");
+		goto error;
+	}
+
+	/*
+	 * Assume path name is class hierarchy, this is a common practice with Java programs
+	 */
+	if (*class_sign == 'L') {
+		int j, i = 0;
+		char *p = strrchr(class_sign, '/');
+		if (p) {
+			/* drop the 'L' prefix and copy up to the final '/' */
+			for (i = 0; i < (p - class_sign); i++)
+				fn[i] = class_sign[i+1];
+		}
+		/*
+		 * append file name, we use loops and not string ops to avoid modifying
+		 * class_sign which is used later for the symbol name
+		 */
+		for (j = 0; i < (PATH_MAX - 1) && j < strlen(file_name); j++, i++)
+			fn[i] = file_name[j];
+		fn[i] = '\0';
+	} else {
+		/* fallback case */
+		strcpy(fn, file_name);
+	}
+	/*
+	 * write source line info record if we have it
+	 */
+	if (jvmti_write_debug_info(jvmti_agent, addr, fn, map, tab, nr_lines))
+		warnx("jvmti: write_debug_info() failed");
+
+	len = strlen(func_name) + strlen(class_sign) + strlen(func_sign) + 2;
+	{
+		char str[len];
+		snprintf(str, len, "%s%s%s", class_sign, func_name, func_sign);
+		if (jvmti_write_code(jvmti_agent, str, addr, code_addr, code_size))
+			warnx("jvmti: write_code() failed");
+	}
+error:
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)func_name);
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)func_sign);
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)class_sign);
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)tab);
+	(*jvmti)->Deallocate(jvmti, (unsigned char *)file_name);
+}
+
+static void JNICALL
+code_generated_cb(jvmtiEnv *jvmti,
+		  char const *name,
+		  void const *code_addr,
+		  jint code_size)
+{
+	uint64_t addr = (uint64_t)(unsigned long)code_addr;
+	int ret;
+
+	ret = jvmti_write_code(jvmti_agent, name, addr, code_addr, code_size);
+	if (ret)
+		warnx("jvmti: write_code() failed for code_generated");
+}
+
+JNIEXPORT jint JNICALL
+Agent_OnLoad(JavaVM *jvm, char *options, void *reserved __unused)
+{
+	jvmtiEventCallbacks cb;
+	jvmtiCapabilities caps1;
+	jvmtiJlocationFormat format;
+	jvmtiEnv *jvmti = NULL;
+	jint ret;
+
+	jvmti_agent = jvmti_open();
+	if (!jvmti_agent) {
+		warnx("jvmti: open_agent failed");
+		return -1;
+	}
+
+	/*
+	 * Request a JVMTI interface version 1 environment
+	 */
+	ret = (*jvm)->GetEnv(jvm, (void *)&jvmti, JVMTI_VERSION_1);
+	if (ret != JNI_OK) {
+		warnx("jvmti: jvmti version 1 not supported");
+		return -1;
+	}
+
+	/*
+	 * acquire method_load capability, we require it
+	 * request line numbers (optional)
+	 */
+	memset(&caps1, 0, sizeof(caps1));
+	caps1.can_generate_compiled_method_load_events = 1;
+
+	ret = (*jvmti)->AddCapabilities(jvmti, &caps1);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: acquire compiled_method capability failed");
+		return -1;
+	}
+	ret = (*jvmti)->GetJLocationFormat(jvmti, &format);
+        if (ret == JVMTI_ERROR_NONE && format == JVMTI_JLOCATION_JVMBCI) {
+                memset(&caps1, 0, sizeof(caps1));
+                caps1.can_get_line_numbers = 1;
+                caps1.can_get_source_file_name = 1;
+		ret = (*jvmti)->AddCapabilities(jvmti, &caps1);
+                if (ret == JVMTI_ERROR_NONE)
+                        has_line_numbers = 1;
+        }
+
+	memset(&cb, 0, sizeof(cb));
+
+	cb.CompiledMethodLoad   = compiled_method_load_cb;
+	cb.DynamicCodeGenerated = code_generated_cb;
+
+	ret = (*jvmti)->SetEventCallbacks(jvmti, &cb, sizeof(cb));
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: cannot set event callbacks");
+		return -1;
+	}
+
+	ret = (*jvmti)->SetEventNotificationMode(jvmti, JVMTI_ENABLE,
+			JVMTI_EVENT_COMPILED_METHOD_LOAD, NULL);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: setnotification failed for method_load");
+		return -1;
+	}
+
+	ret = (*jvmti)->SetEventNotificationMode(jvmti, JVMTI_ENABLE,
+			JVMTI_EVENT_DYNAMIC_CODE_GENERATED, NULL);
+	if (ret != JVMTI_ERROR_NONE) {
+		warnx("jvmti: setnotification failed on code_generated");
+		return -1;
+	}
+	return 0;
+}
+
+JNIEXPORT void JNICALL
+Agent_OnUnload(JavaVM *jvm __unused)
+{
+	int ret;
+
+	ret = jvmti_close(jvmti_agent);
+	if (ret)
+		errx(1, "Error: op_close_agent()");
+}
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1237167

FromPeter Zijlstra <peterz@infradead.org>
Date2015-10-01 11:20 +0200
Message-ID<qeIbU-8aj-13@gated-at.bofh.it>
In reply to#1237039
On Thu, Oct 01, 2015 at 08:45:44AM +0200, Stephane Eranian wrote:
> This patch series extends perf record/report/annotate to enable
> profiling of jitted (just-in-time compiled) code. The current
> perf tool provides very limited support for profiling jitted
> code for some runtime environments. But the support is experimental
> and cannot be used in complex environments. It relies on files
> in /tmp, for instance. It does not support annotate mode or
> rejitted code.
> 
> This patch series adds a better way of profiling jitted code
> with the following advantages:
>    - support any jitted code environment (some with modifications)
>    - support Java runtime with JVMTI interface with no modifications
>    - provides a portable JVMTI agent library
>    - known to support V8 runtime
>    - known to support DART runtime
>    - supports code rejitting and code movements
>    - no files in /tmp
>    - meta-data file is unique to each run
>    - no changes to perf report/annotate
>    - support per-thread and system-wide profiling
>    - support monitoring of multiple simultaneous Jit runtimes
>    - source level view in perf annotate
> 
> The support is based on cooperation with the runtime. For Java runtimes,
> supporting the JVMTI interface, there is no change necessary. For other
> runtimes, modifications are necessary to emit the meta-data to support
> symbolization, annotation, source lines correlation of the samples.
> Those modifications are relatively straighforward, some have been 
> implemented in V8 and DART.

Do V8 and DART come with these bits or will that be a future
contribution?

> This will also generate an ELF image for each jitted function. The
> injected MMAP records will point to these ELF images. The reasoning
> behind using ELF images is that it makes processing for perf report
> and annotate automatic and transparent. It also makes it easier to
> package and analyze on a remote machine. Binutils tools can decode
> the ELF images easily.

The generation of ELF files is really nice!

All in all this looks really nice.

Thanks for doing this.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1237213

FromIngo Molnar <mingo@kernel.org>
Date2015-10-01 11:40 +0200
Message-ID<qeIvh-5L-51@gated-at.bofh.it>
In reply to#1237167
* Peter Zijlstra <peterz@infradead.org> wrote:

> > This will also generate an ELF image for each jitted function. The injected 
> > MMAP records will point to these ELF images. The reasoning behind using ELF 
> > images is that it makes processing for perf report and annotate automatic and 
> > transparent. It also makes it easier to package and analyze on a remote 
> > machine. Binutils tools can decode the ELF images easily.
> 
> The generation of ELF files is really nice!
> 
> All in all this looks really nice.
> 
> Thanks for doing this.

Seconded, nice stuff!

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1237551

FromArnaldo Carvalho de Melo <acme@redhat.com>
Date2015-10-01 18:20 +0200
Message-ID<qeOKm-1gw-19@gated-at.bofh.it>
In reply to#1237213
Em Thu, Oct 01, 2015 at 11:39:14AM +0200, Ingo Molnar escreveu:
> * Peter Zijlstra <peterz@infradead.org> wrote:
> > > This will also generate an ELF image for each jitted function. The injected 
> > > MMAP records will point to these ELF images. The reasoning behind using ELF 
> > > images is that it makes processing for perf report and annotate automatic and 
> > > transparent. It also makes it easier to package and analyze on a remote 
> > > machine. Binutils tools can decode the ELF images easily.

> > The generation of ELF files is really nice!

> > All in all this looks really nice.

> > Thanks for doing this.
 
> Seconded, nice stuff!

Thirded! ;-)

Gave it a quick look, the ELF generation thing could later be librarized
somehow for other uses, the passing of perf_session for sample_type can
maybe use evsel->attr.sample_type instead, no?

Will try to try it after lunch :-)

- Arnaldo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1239201

FromStephane Eranian <eranian@google.com>
Date2015-10-04 22:00 +0200
Message-ID<qfXBU-1IF-11@gated-at.bofh.it>
In reply to#1237167
Hi Peter,

On Thu, Oct 1, 2015 at 2:17 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> On Thu, Oct 01, 2015 at 08:45:44AM +0200, Stephane Eranian wrote:
>> This patch series extends perf record/report/annotate to enable
>> profiling of jitted (just-in-time compiled) code. The current
>> perf tool provides very limited support for profiling jitted
>> code for some runtime environments. But the support is experimental
>> and cannot be used in complex environments. It relies on files
>> in /tmp, for instance. It does not support annotate mode or
>> rejitted code.
>>
>> This patch series adds a better way of profiling jitted code
>> with the following advantages:
>>    - support any jitted code environment (some with modifications)
>>    - support Java runtime with JVMTI interface with no modifications
>>    - provides a portable JVMTI agent library
>>    - known to support V8 runtime
>>    - known to support DART runtime
>>    - supports code rejitting and code movements
>>    - no files in /tmp
>>    - meta-data file is unique to each run
>>    - no changes to perf report/annotate
>>    - support per-thread and system-wide profiling
>>    - support monitoring of multiple simultaneous Jit runtimes
>>    - source level view in perf annotate
>>
>> The support is based on cooperation with the runtime. For Java runtimes,
>> supporting the JVMTI interface, there is no change necessary. For other
>> runtimes, modifications are necessary to emit the meta-data to support
>> symbolization, annotation, source lines correlation of the samples.
>> Those modifications are relatively straighforward, some have been
>> implemented in V8 and DART.
>
> Do V8 and DART come with these bits or will that be a future
> contribution?
>
I think both had this working some time ago. I have not had a chance to
verify if this new version still works. I did not change the format of
the jitdump
file, so if they use the same clock source for their timestamp as perf
record, it
should still work. All of this hinges on the fact that V8 & DART developers run
with 4.1 to get your clock changes.


>> This will also generate an ELF image for each jitted function. The
>> injected MMAP records will point to these ELF images. The reasoning
>> behind using ELF images is that it makes processing for perf report
>> and annotate automatic and transparent. It also makes it easier to
>> package and analyze on a remote machine. Binutils tools can decode
>> the ELF images easily.
>
> The generation of ELF files is really nice!
>
> All in all this looks really nice.
>
Thanks!

> Thanks for doing this.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1237808

FromBrendan Gregg <brendan.d.gregg@gmail.com>
Date2015-10-02 00:50 +0200
Message-ID<qeUPL-1B8-7@gated-at.bofh.it>
In reply to#1237039
G'Day,

On Wed, Sep 30, 2015 at 11:45 PM, Stephane Eranian <eranian@google.com> wrote:
>
> This patch series extends perf record/report/annotate to enable
> profiling of jitted (just-in-time compiled) code. The current
> perf tool provides very limited support for profiling jitted
> code for some runtime environments. But the support is experimental
> and cannot be used in complex environments. It relies on files
> in /tmp, for instance. It does not support annotate mode or
> rejitted code.
>
> This patch series adds a better way of profiling jitted code
> with the following advantages:
>    - support any jitted code environment (some with modifications)
>    - support Java runtime with JVMTI interface with no modifications
>    - provides a portable JVMTI agent library
>    - known to support V8 runtime
>    - known to support DART runtime
>    - supports code rejitting and code movements
>    - no files in /tmp
>    - meta-data file is unique to each run
>    - no changes to perf report/annotate
>    - support per-thread and system-wide profiling
>    - support monitoring of multiple simultaneous Jit runtimes
>    - source level view in perf annotate
>
> The support is based on cooperation with the runtime. For Java runtimes,
> supporting the JVMTI interface, there is no change necessary. For other
> runtimes, modifications are necessary to emit the meta-data to support
> symbolization, annotation, source lines correlation of the samples.
> Those modifications are relatively straighforward, some have been
> implemented in V8 and DART.
>
> The jit environment emits a binary dump file which contains the jitted
> code (in raw format) and meta-data describing the mapping of functions.
> The binary format is documented in the jitdump.h header file. It is
> adapted from the OProfile jitdump format.

While this is impressive work, I don't think I'd use it very much, and
I wouldn't encourage others too either. Is it right that this approach
needs to be turned on from runtime start, and will constantly emit
timestamped JIT records? I'd use that as a backup for existing
techniques for perf_events and jitted runtimes.

Right now we (Netflix) can profile Java in production using
perf_events, using an on-demand symbol dump: the perf-map-agent
JVMTI[1]. This agent used to be loaded on startup, like this patchset,
which cost overhead: CPU, filesystem I/O, storage. The problem was
that we didn't know which of the tens of thousands of instances we'd
want to profile, so loading an always-on JIT symbol dumper on all
instances would cost resources. With on-demand, we only dump symbols
on the few instances needed. On-demand has caveats, of course, and
symbols can change during the profile. But so far that's not been a
problem.

So this seems like a lot of changes for what won't be our primary way
of using perf_events on Java or Node.js, which we currently can
profile with perf_events effectively. So long as we're aware of that.

Brendan

[1] http://techblog.netflix.com/2015/07/java-in-flames.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]

#1239202

FromStephane Eranian <eranian@google.com>
Date2015-10-04 22:10 +0200
Message-ID<qfXLA-29h-7@gated-at.bofh.it>
In reply to#1237808
Brendan,

On Thu, Oct 1, 2015 at 3:45 PM, Brendan Gregg <brendan.d.gregg@gmail.com> wrote:
> G'Day,
>
> On Wed, Sep 30, 2015 at 11:45 PM, Stephane Eranian <eranian@google.com> wrote:
>>
>> This patch series extends perf record/report/annotate to enable
>> profiling of jitted (just-in-time compiled) code. The current
>> perf tool provides very limited support for profiling jitted
>> code for some runtime environments. But the support is experimental
>> and cannot be used in complex environments. It relies on files
>> in /tmp, for instance. It does not support annotate mode or
>> rejitted code.
>>
>> This patch series adds a better way of profiling jitted code
>> with the following advantages:
>>    - support any jitted code environment (some with modifications)
>>    - support Java runtime with JVMTI interface with no modifications
>>    - provides a portable JVMTI agent library
>>    - known to support V8 runtime
>>    - known to support DART runtime
>>    - supports code rejitting and code movements
>>    - no files in /tmp
>>    - meta-data file is unique to each run
>>    - no changes to perf report/annotate
>>    - support per-thread and system-wide profiling
>>    - support monitoring of multiple simultaneous Jit runtimes
>>    - source level view in perf annotate
>>
>> The support is based on cooperation with the runtime. For Java runtimes,
>> supporting the JVMTI interface, there is no change necessary. For other
>> runtimes, modifications are necessary to emit the meta-data to support
>> symbolization, annotation, source lines correlation of the samples.
>> Those modifications are relatively straighforward, some have been
>> implemented in V8 and DART.
>>
>> The jit environment emits a binary dump file which contains the jitted
>> code (in raw format) and meta-data describing the mapping of functions.
>> The binary format is documented in the jitdump.h header file. It is
>> adapted from the OProfile jitdump format.
>
> While this is impressive work, I don't think I'd use it very much, and
> I wouldn't encourage others too either. Is it right that this approach
> needs to be turned on from runtime start, and will constantly emit
> timestamped JIT records? I'd use that as a backup for existing
> techniques for perf_events and jitted runtimes.
>
This boils down that when does the JIT runtime emit the jitdump data?
In the case of openJDK and given how I wrote the agent, it needs to
run from start to finish.  In order for perf to have full visibility, it needs
to get info about all the code that has been jitted so far. This could be
done after start if there was somehow a protocol to handle this in the
runtime, like a signal. It would just need to dump the current status, including
all the code. The rest of the  support would work. In other words, if there was
a way to signal to the runtime that it is being monitored and that it needs to
dump its state, then everything would work. To avoid generating more dumps,
the runtime would also have to be informed that monitoring has stopped.


> Right now we (Netflix) can profile Java in production using
> perf_events, using an on-demand symbol dump: the perf-map-agent
> JVMTI[1]. This agent used to be loaded on startup, like this patchset,
> which cost overhead: CPU, filesystem I/O, storage. The problem was
> that we didn't know which of the tens of thousands of instances we'd
> want to profile, so loading an always-on JIT symbol dumper on all
> instances would cost resources. With on-demand, we only dump symbols
> on the few instances needed. On-demand has caveats, of course, and
> symbols can change during the profile. But so far that's not been a
> problem.
>
I am guessing that for what you are doing getting the jitted code symbol names
are sufficient. This patch series is not really needed for this, you
already have the
/tmp/perf-map approach. This is mostly for the user program
developers. This series
addresses the  needs of the other category, namely the jit compiler and runtime
developers who need to study the quality of the generated code. And there, they
need to see the profile of the jitted native code. This is the major
value-add of this
series. I would also add the support for re-jitting and code movements.

> So this seems like a lot of changes for what won't be our primary way
> of using perf_events on Java or Node.js, which we currently can
> profile with perf_events effectively. So long as we're aware of that.
>
The changes to perf are small. There are no change to perf record, report,
annotate. just the injection code with the jitdump reader and elf creation.

> Brendan
>
> [1] http://techblog.netflix.com/2015/07/java-in-flames.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web