Skip to contentSkip to conversation

perf jvmti: Various fixes to JVMTI agent

PATCH Applied Nick Gasson 5 participants 19 messages Apr 27, 2020 – May 28, 2020 perf events

lore
Share & tools
lore
#0 lore raw

cover perf jvmti: Various fixes to JVMTI agent

Hi,

These three patches fix a couple of issues I ran into while using the jitdump JVMTI agent to profile the SPECjbb benchmark.

Tested with OpenJDK 8 and 14.

Thanks,
Nick

Nick Gasson (3):
perf jvmti: Fix jitdump for methods without debug info perf jvmti: Do not report error when missing debug information perf jvmti: Fix demangling Java symbols

 tools/perf/jvmti/libjvmti.c           | 24 +++++++--------
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 6 files changed, 66 insertions(+), 19 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
Signature · 1 line
2.26.1
#1 lore raw

PATCH 1/3 perf jvmti: Fix jitdump for methods without debug info

If a Java class is compiled with -g:none to omit debug information, the JVMTI plugin won't write jitdump entries for any method in this class and prints a lot of errors like:

java: GetSourceFileName failed with JVMTI_ERROR_ABSENT_INFORMATION

The call to GetSourceFileName is used to derive the file name fn, but this value is not actually used since commit ca58d7e64bdf ("perf jvmti:
Generate correct debug information for inlined code") which moved the file name lookup into fill_source_filenames(). So the call to GetSourceFileName and related code can be safely removed.

---
 tools/perf/jvmti/libjvmti.c | 11 -----------
 1 file changed, 11 deletions(-)
tools/perf/jvmti/libjvmti.c +0 -11
index c441a34cb1c0..50ef524b5cd4 100644
@@ -246,8 +246,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	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;
 	int nr_lines = 0; /* in line_tab[] */
@@ -282,12 +280,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 		}
 	}
 
-	ret = (*jvmti)->GetSourceFileName(jvmti, decl_class, &file_name);
-	if (ret != JVMTI_ERROR_NONE) {
-		print_error(jvmti, "GetSourceFileName", ret);
-		goto error;
-	}
-
 	ret = (*jvmti)->GetClassSignature(jvmti, decl_class,
 					  &class_sign, NULL);
 	if (ret != JVMTI_ERROR_NONE) {
@@ -302,8 +294,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 		goto error;
 	}
 
-	copy_class_filename(class_sign, file_name, fn, PATH_MAX);
-
 	/*
 	 * write source line info record if we have it
 	 */
@@ -323,7 +313,6 @@ 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 *)file_name);
 	free(line_tab);
 	while (line_file_names && (nr_lines > 0)) {
 	    if (line_file_names[nr_lines - 1]) {
Signature · 1 line
2.26.1
reviewer Reviewed-by Tested-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] perf jvmti: Fix jitdump for methods without debug info

Quoted 7 lines · Nick Gasson

If a Java class is compiled with -g:none to omit debug information, the JVMTI plugin won't write jitdump entries for any method in this class and prints a lot of errors like:

java: GetSourceFileName failed with JVMTI_ERROR_ABSENT_INFORMATION

Quoted 7 lines

The call to GetSourceFileName is used to derive the file name fn, but this value is not actually used since commit ca58d7e64bdf ("perf jvmti:
Generate correct debug information for inlined code") which moved the file name lookup into fill_source_filenames(). So the call to GetSourceFileName and related code can be safely removed.

Thanks!
Ian

Quoted 50 lines
---
 tools/perf/jvmti/libjvmti.c | 11 -----------
 1 file changed, 11 deletions(-)
tools/perf/jvmti/libjvmti.c +0 -12
index c441a34cb1c0..50ef524b5cd4 100644
@@ -246,8 +246,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
        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;
        int nr_lines = 0; /* in line_tab[] */
@@ -282,12 +280,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
                }
        }
 
-       ret = (*jvmti)->GetSourceFileName(jvmti, decl_class, &file_name);
-       if (ret != JVMTI_ERROR_NONE) {
-               print_error(jvmti, "GetSourceFileName", ret);
-               goto error;
-       }
-
        ret = (*jvmti)->GetClassSignature(jvmti, decl_class,
                                          &class_sign, NULL);
        if (ret != JVMTI_ERROR_NONE) {
@@ -302,8 +294,6 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
                goto error;
        }
 
-       copy_class_filename(class_sign, file_name, fn, PATH_MAX);
-
        /*
         * write source line info record if we have it
         */
@@ -323,7 +313,6 @@ 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 *)file_name);
        free(line_tab);
        while (line_file_names && (nr_lines > 0)) {
            if (line_file_names[nr_lines - 1]) {
--

2.26.1

#3 lore raw

PATCH 2/3 perf jvmti: Do not report error when missing debug information

In reply to Nick Gasson

If the Java sources are compiled with -g:none to disable debug information the perf JVMTI plugin reports a lot of errors like:

java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION

Instead if GetLineNumberTable returns JVMTI_ERROR_ABSENT_INFORMATION simply skip emitting line number information for that method. Unlike the previous patch these errors don't affect the jitdump generation, they just generate a lot of noise.

Similarly for native methods which also don't have line tables.

---
 tools/perf/jvmti/libjvmti.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/perf/jvmti/libjvmti.c +11 -2
index 50ef524b5cd4..a9a056d68416 100644
@@ -41,7 +41,11 @@ do_get_line_numbers(jvmtiEnv *jvmti, void *pc, jmethodID m, jint bci,
 	jvmtiError ret;
 
 	ret = (*jvmti)->GetLineNumberTable(jvmti, m, &nr_lines, &loc_tab);
-	if (ret != JVMTI_ERROR_NONE) {
+	if (ret == JVMTI_ERROR_ABSENT_INFORMATION || ret == JVMTI_ERROR_NATIVE_METHOD) {
+		/* No debug information for this method */
+		*nr = 0;
+		return JVMTI_ERROR_NONE;
+	} else if (ret != JVMTI_ERROR_NONE) {
 		print_error(jvmti, "GetLineNumberTable", ret);
 		return ret;
 	}
@@ -93,6 +97,9 @@ get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **
 					/* free what was allocated for nothing */
 					(*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
 					nr_total += (int)nr;
+				} else if (ret == JVMTI_ERROR_ABSENT_INFORMATION
+					   || ret == JVMTI_ERROR_NATIVE_METHOD) {
+					/* No debug information for this method */
 				} else {
 					print_error(jvmti, "GetLineNumberTable", ret);
 				}
@@ -262,7 +269,9 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	if (has_line_numbers && map && map_length) {
 		ret = get_line_numbers(jvmti, compile_info, &line_tab, &nr_lines);
 		if (ret != JVMTI_ERROR_NONE) {
-			warnx("jvmti: cannot get line table for method");
+			if (ret != JVMTI_ERROR_NOT_FOUND) {
+				warnx("jvmti: cannot get line table for method");
+			}
 			nr_lines = 0;
 		} else if (nr_lines > 0) {
 			line_file_names = malloc(sizeof(char*) * nr_lines);
Signature · 1 line
2.26.1
reviewer Reviewed-by Tested-by on v1 2/3
#4 lore raw

Re: [PATCH 2/3] perf jvmti: Do not report error when missing debug information

Quoted 10 lines · Nick Gasson

If the Java sources are compiled with -g:none to disable debug information the perf JVMTI plugin reports a lot of errors like:

java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION

Quoted 8 lines

Instead if GetLineNumberTable returns JVMTI_ERROR_ABSENT_INFORMATION simply skip emitting line number information for that method. Unlike the previous patch these errors don't affect the jitdump generation, they just generate a lot of noise.

Similarly for native methods which also don't have line tables.

Thanks!
Ian

Quoted 45 lines
---
 tools/perf/jvmti/libjvmti.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/perf/jvmti/libjvmti.c +11 -3
index 50ef524b5cd4..a9a056d68416 100644
@@ -41,7 +41,11 @@ do_get_line_numbers(jvmtiEnv *jvmti, void *pc, jmethodID m, jint bci,
        jvmtiError ret;
 
        ret = (*jvmti)->GetLineNumberTable(jvmti, m, &nr_lines, &loc_tab);
-       if (ret != JVMTI_ERROR_NONE) {
+       if (ret == JVMTI_ERROR_ABSENT_INFORMATION || ret == JVMTI_ERROR_NATIVE_METHOD) {
+               /* No debug information for this method */
+               *nr = 0;
+               return JVMTI_ERROR_NONE;
+       } else if (ret != JVMTI_ERROR_NONE) {
                print_error(jvmti, "GetLineNumberTable", ret);
                return ret;
        }
@@ -93,6 +97,9 @@ get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **
                                        /* free what was allocated for nothing */
                                        (*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
                                        nr_total += (int)nr;
+                               } else if (ret == JVMTI_ERROR_ABSENT_INFORMATION
+                                          || ret == JVMTI_ERROR_NATIVE_METHOD) {
+                                       /* No debug information for this method */
                                } else {
                                        print_error(jvmti, "GetLineNumberTable", ret);
                                }
@@ -262,7 +269,9 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
        if (has_line_numbers && map && map_length) {
                ret = get_line_numbers(jvmti, compile_info, &line_tab, &nr_lines);
                if (ret != JVMTI_ERROR_NONE) {
-                       warnx("jvmti: cannot get line table for method");
+                       if (ret != JVMTI_ERROR_NOT_FOUND) {
+                               warnx("jvmti: cannot get line table for method");
+                       }
                        nr_lines = 0;
                } else if (nr_lines > 0) {
                        line_file_names = malloc(sizeof(char*) * nr_lines);
--

2.26.1

maintainer Applied
#5 lore raw

Re: [PATCH 2/3] perf jvmti: Do not report error when missing debug information

In reply to Nick Gasson

Em Mon, Apr 27, 2020 at 02:15:15PM +0800, Nick Gasson escreveu:

Quoted 39 lines

If the Java sources are compiled with -g:none to disable debug information the perf JVMTI plugin reports a lot of errors like:

java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION

Instead if GetLineNumberTable returns JVMTI_ERROR_ABSENT_INFORMATION simply skip emitting line number information for that method. Unlike the previous patch these errors don't affect the jitdump generation, they just generate a lot of noise.

Similarly for native methods which also don't have line tables.

---
 tools/perf/jvmti/libjvmti.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/perf/jvmti/libjvmti.c +5 -1
index 50ef524b5cd4..a9a056d68416 100644
@@ -41,7 +41,11 @@ do_get_line_numbers(jvmtiEnv *jvmti, void *pc, jmethodID m, jint bci,
 	jvmtiError ret;
 
 	ret = (*jvmti)->GetLineNumberTable(jvmti, m, &nr_lines, &loc_tab);
-	if (ret != JVMTI_ERROR_NONE) {
+	if (ret == JVMTI_ERROR_ABSENT_INFORMATION || ret == JVMTI_ERROR_NATIVE_METHOD) {
+		/* No debug information for this method */
+		*nr = 0;
+		return JVMTI_ERROR_NONE;
+	} else if (ret != JVMTI_ERROR_NONE) {
 		print_error(jvmti, "GetLineNumberTable", ret);
 		return ret;
 	}

tools/perf/jvmti/libjvmti.c get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t ** · line 101

Quoted 5 lines
tools/perf/jvmti/libjvmti.c +2 -0
@@ -93,6 +97,9 @@ get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **
 					/* free what was allocated for nothing */
 					(*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
 					nr_total += (int)nr;
+				} else if (ret == JVMTI_ERROR_ABSENT_INFORMATION
+					   || ret == JVMTI_ERROR_NATIVE_METHOD) {

Please add the || operator at the end of the line next time, I'm fixing this up this time,

Applied,

- Arnaldo

Quoted 18 lines
tools/perf/jvmti/libjvmti.c +4 -1
+					/* No debug information for this method */
 				} else {
 					print_error(jvmti, "GetLineNumberTable", ret);
 				}
@@ -262,7 +269,9 @@ compiled_method_load_cb(jvmtiEnv *jvmti,
 	if (has_line_numbers && map && map_length) {
 		ret = get_line_numbers(jvmti, compile_info, &line_tab, &nr_lines);
 		if (ret != JVMTI_ERROR_NONE) {
-			warnx("jvmti: cannot get line table for method");
+			if (ret != JVMTI_ERROR_NOT_FOUND) {
+				warnx("jvmti: cannot get line table for method");
+			}
 			nr_lines = 0;
 		} else if (nr_lines > 0) {
 			line_file_names = malloc(sizeof(char*) * nr_lines);
Signature · 1 line
2.26.1
Signature · 1 line
- Arnaldo
#6 lore raw

Re: [PATCH 2/3] perf jvmti: Do not report error when missing debug information

Em Wed, May 27, 2020 at 11:07:53AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 49 lines

Em Mon, Apr 27, 2020 at 02:15:15PM +0800, Nick Gasson escreveu:

Quoted 39 lines

If the Java sources are compiled with -g:none to disable debug information the perf JVMTI plugin reports a lot of errors like:

java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION
java: GetLineNumberTable failed with JVMTI_ERROR_ABSENT_INFORMATION

Instead if GetLineNumberTable returns JVMTI_ERROR_ABSENT_INFORMATION simply skip emitting line number information for that method. Unlike the previous patch these errors don't affect the jitdump generation, they just generate a lot of noise.

Similarly for native methods which also don't have line tables.

---
 tools/perf/jvmti/libjvmti.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
tools/perf/jvmti/libjvmti.c +5 -1
index 50ef524b5cd4..a9a056d68416 100644
@@ -41,7 +41,11 @@ do_get_line_numbers(jvmtiEnv *jvmti, void *pc, jmethodID m, jint bci,
 	jvmtiError ret;
 
 	ret = (*jvmti)->GetLineNumberTable(jvmti, m, &nr_lines, &loc_tab);
-	if (ret != JVMTI_ERROR_NONE) {
+	if (ret == JVMTI_ERROR_ABSENT_INFORMATION || ret == JVMTI_ERROR_NATIVE_METHOD) {
+		/* No debug information for this method */
+		*nr = 0;
+		return JVMTI_ERROR_NONE;
+	} else if (ret != JVMTI_ERROR_NONE) {
 		print_error(jvmti, "GetLineNumberTable", ret);
 		return ret;
 	}

tools/perf/jvmti/libjvmti.c get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t ** · line 101

Quoted 5 lines
tools/perf/jvmti/libjvmti.c +2 -0
@@ -93,6 +97,9 @@ get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **
 					/* free what was allocated for nothing */
 					(*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
 					nr_total += (int)nr;
+				} else if (ret == JVMTI_ERROR_ABSENT_INFORMATION
+					   || ret == JVMTI_ERROR_NATIVE_METHOD) {

Please add the || operator at the end of the line next time, I'm fixing this up this time,

Quoted 1 line

Applied,

This:

tools/perf/jvmti/libjvmti.c +2 -2
index a9a056d68416..c5d30834a64c 100644
@@ -97,8 +97,8 @@ get_line_numbers(jvmtiEnv *jvmti, const void *compile_info, jvmti_line_info_t **
 					/* free what was allocated for nothing */
 					(*jvmti)->Deallocate(jvmti, (unsigned char *)lne);
 					nr_total += (int)nr;
-				} else if (ret == JVMTI_ERROR_ABSENT_INFORMATION
-					   || ret == JVMTI_ERROR_NATIVE_METHOD) {
+				} else if (ret == JVMTI_ERROR_ABSENT_INFORMATION ||
+					   ret == JVMTI_ERROR_NATIVE_METHOD) {
 					/* No debug information for this method */
 				} else {
 					print_error(jvmti, "GetLineNumberTable", ret);
#7 lore raw

PATCH 3/3 perf jvmti: Fix demangling Java symbols

In reply to Nick Gasson

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Also added a test case.

---
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 5 files changed, 55 insertions(+), 6 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
tools/perf/tests/Build +1 -0
index 1692529639b0..2c45ac4a9581 100644
@@ -55,6 +55,7 @@ perf-y += mem2node.o
 perf-y += maps.o
 perf-y += time-utils-test.o
 perf-y += genelf.o
+perf-y += demangle-java-test.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
tools/perf/tests/builtin-test.c +4 -0
index 54d9516c9839..03b362b37f97 100644
@@ -309,6 +309,10 @@ static struct test generic_tests[] = {
 		.desc = "maps__merge_in",
 		.func = test__maps__merge_in,
 	},
+	{
+		.desc = "Demangle Java",
+		.func = test__demangle_java,
+	},
 	{
 		.func = NULL,
 	},
tools/perf/tests/demangle-java-test.c new file +42 -0

42 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index 9a160fef47c9..49b791d978f6 100644
@@ -111,6 +111,7 @@ int test__mem2node(struct test *t, int subtest);
 int test__maps__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 int test__jit_write_elf(struct test *test, int subtest);
+int test__demangle_java(struct test *test, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__bp_account_is_supported(void);
tools/perf/util/demangle-java.c +7 -6
index 6fb7f34c0814..39c05200ed65 100644
@@ -15,7 +15,7 @@ enum {
 	MODE_CLASS  = 1,
 	MODE_FUNC   = 2,
 	MODE_TYPE   = 3,
-	MODE_CTYPE  = 3, /* class arg */
+	MODE_CTYPE  = 4, /* class arg */
 };
 
 #define BASE_ENT(c, n)	[c - 'A']=n
@@ -27,7 +27,7 @@ static const char *base_types['Z' - 'A' + 1] = {
 	BASE_ENT('I', "int" ),
 	BASE_ENT('J', "long" ),
 	BASE_ENT('S', "short" ),
-	BASE_ENT('Z', "bool" ),
+	BASE_ENT('Z', "boolean" ),
 };
 
 /*
@@ -59,15 +59,16 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 
 		switch (*q) {
 		case 'L':
-			if (mode == MODE_PREFIX || mode == MODE_CTYPE) {
-				if (mode == MODE_CTYPE) {
+			if (mode == MODE_PREFIX || mode == MODE_TYPE) {
+				if (mode == MODE_TYPE) {
 					if (narg)
 						rlen += scnprintf(buf + rlen, maxlen - rlen, ", ");
 					narg++;
 				}
-				rlen += scnprintf(buf + rlen, maxlen - rlen, "class ");
 				if (mode == MODE_PREFIX)
 					mode = MODE_CLASS;
+				else
+					mode = MODE_CTYPE;
 			} else
 				buf[rlen++] = *q;
 			break;
@@ -120,7 +121,7 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 			if (mode != MODE_CLASS && mode != MODE_CTYPE)
 				goto error;
 			/* safe because at least one other char to process */
-			if (isalpha(*(q + 1)))
+			if (isalpha(*(q + 1)) && mode == MODE_CLASS)
 				rlen += scnprintf(buf + rlen, maxlen - rlen, ".");
 			if (mode == MODE_CLASS)
 				mode = MODE_FUNC;
Signature · 1 line
2.26.1
reviewer Reviewed-by Tested-by on v1 3/3
#8 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

Quoted 15 lines · Nick Gasson

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

Quoted 8 lines

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Also added a test case.

Thanks!
Ian

Quoted 150 lines
---
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 5 files changed, 55 insertions(+), 6 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
tools/perf/tests/Build +1 -0
index 1692529639b0..2c45ac4a9581 100644
@@ -55,6 +55,7 @@ perf-y += mem2node.o
 perf-y += maps.o
 perf-y += time-utils-test.o
 perf-y += genelf.o
+perf-y += demangle-java-test.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
        $(call rule_mkdir)
tools/perf/tests/builtin-test.c +4 -0
index 54d9516c9839..03b362b37f97 100644
@@ -309,6 +309,10 @@ static struct test generic_tests[] = {
                .desc = "maps__merge_in",
                .func = test__maps__merge_in,
        },
+       {
+               .desc = "Demangle Java",
+               .func = test__demangle_java,
+       },
        {
                .func = NULL,
        },
tools/perf/tests/demangle-java-test.c new file +42 -0
new file mode 100644
index 000000000000..8f3b90832fb0
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include "tests.h"
+#include "session.h"
+#include "debug.h"
+#include "demangle-java.h"
+
+int test__demangle_java(struct test *test __maybe_unused, int subtest __maybe_unused)
+{
+       int ret = TEST_OK;
+       char *buf = NULL;
+       size_t i;
+
+       struct {
+               const char *mangled, *demangled;
+       } test_cases[] = {
+               { "Ljava/lang/StringLatin1;equals([B[B)Z",
+                 "boolean java.lang.StringLatin1.equals(byte[], byte[])" },
+               { "Ljava/util/zip/ZipUtils;CENSIZ([BI)J",
+                 "long java.util.zip.ZipUtils.CENSIZ(byte[], int)" },
+               { "Ljava/util/regex/Pattern$BmpCharProperty;match(Ljava/util/regex/Matcher;ILjava/lang/CharSequence;)Z",
+                 "boolean java.util.regex.Pattern$BmpCharProperty.match(java.util.regex.Matcher, int, java.lang.CharSequence)" },
+               { "Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V",
+                 "void java.lang.AbstractStringBuilder.appendChars(java.lang.String, int, int)" },
+               { "Ljava/lang/Object;<init>()V",
+                 "void java.lang.Object<init>()" },
+       };
+
+       for (i = 0; i < sizeof(test_cases) / sizeof(test_cases[0]); i++) {
+               buf = java_demangle_sym(test_cases[i].mangled, 0);
+               if (strcmp(buf, test_cases[i].demangled)) {
+                       pr_debug("FAILED: %s: %s != %s\n", test_cases[i].mangled,
+                                buf, test_cases[i].demangled);
+                       ret = TEST_FAIL;
+               }
+               free(buf);
+       }
+
+       return ret;
+}
tools/perf/tests/tests.h +1 -0
index 9a160fef47c9..49b791d978f6 100644
@@ -111,6 +111,7 @@ int test__mem2node(struct test *t, int subtest);
 int test__maps__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 int test__jit_write_elf(struct test *test, int subtest);
+int test__demangle_java(struct test *test, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__bp_account_is_supported(void);
tools/perf/util/demangle-java.c +7 -7
index 6fb7f34c0814..39c05200ed65 100644
@@ -15,7 +15,7 @@ enum {
        MODE_CLASS  = 1,
        MODE_FUNC   = 2,
        MODE_TYPE   = 3,
-       MODE_CTYPE  = 3, /* class arg */
+       MODE_CTYPE  = 4, /* class arg */
 };
 
 #define BASE_ENT(c, n) [c - 'A']=n
@@ -27,7 +27,7 @@ static const char *base_types['Z' - 'A' + 1] = {
        BASE_ENT('I', "int" ),
        BASE_ENT('J', "long" ),
        BASE_ENT('S', "short" ),
-       BASE_ENT('Z', "bool" ),
+       BASE_ENT('Z', "boolean" ),
 };
 
 /*
@@ -59,15 +59,16 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 
                switch (*q) {
                case 'L':
-                       if (mode == MODE_PREFIX || mode == MODE_CTYPE) {
-                               if (mode == MODE_CTYPE) {
+                       if (mode == MODE_PREFIX || mode == MODE_TYPE) {
+                               if (mode == MODE_TYPE) {
                                        if (narg)
                                                rlen += scnprintf(buf + rlen, maxlen - rlen, ", ");
                                        narg++;
                                }
-                               rlen += scnprintf(buf + rlen, maxlen - rlen, "class ");
                                if (mode == MODE_PREFIX)
                                        mode = MODE_CLASS;
+                               else
+                                       mode = MODE_CTYPE;
                        } else
                                buf[rlen++] = *q;
                        break;
@@ -120,7 +121,7 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
                        if (mode != MODE_CLASS && mode != MODE_CTYPE)
                                goto error;
                        /* safe because at least one other char to process */
-                       if (isalpha(*(q + 1)))
+                       if (isalpha(*(q + 1)) && mode == MODE_CLASS)
                                rlen += scnprintf(buf + rlen, maxlen - rlen, ".");
                        if (mode == MODE_CLASS)
                                mode = MODE_FUNC;
--

2.26.1

maintainer Applied
#9 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

In reply to Nick Gasson

Em Mon, Apr 27, 2020 at 02:15:16PM +0800, Nick Gasson escreveu:

Quoted 7 lines

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

Cool, one more 'perf test' entry, thanks a lot!

Applied,

- Arnaldo

Quoted 164 lines

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Also added a test case.

---
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 5 files changed, 55 insertions(+), 6 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
tools/perf/tests/Build +1 -0
index 1692529639b0..2c45ac4a9581 100644
@@ -55,6 +55,7 @@ perf-y += mem2node.o
 perf-y += maps.o
 perf-y += time-utils-test.o
 perf-y += genelf.o
+perf-y += demangle-java-test.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
tools/perf/tests/builtin-test.c +4 -0
index 54d9516c9839..03b362b37f97 100644
@@ -309,6 +309,10 @@ static struct test generic_tests[] = {
 		.desc = "maps__merge_in",
 		.func = test__maps__merge_in,
 	},
+	{
+		.desc = "Demangle Java",
+		.func = test__demangle_java,
+	},
 	{
 		.func = NULL,
 	},
tools/perf/tests/demangle-java-test.c new file +42 -0
new file mode 100644
index 000000000000..8f3b90832fb0
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include "tests.h"
+#include "session.h"
+#include "debug.h"
+#include "demangle-java.h"
+
+int test__demangle_java(struct test *test __maybe_unused, int subtest __maybe_unused)
+{
+	int ret = TEST_OK;
+	char *buf = NULL;
+	size_t i;
+
+	struct {
+		const char *mangled, *demangled;
+	} test_cases[] = {
+		{ "Ljava/lang/StringLatin1;equals([B[B)Z",
+		  "boolean java.lang.StringLatin1.equals(byte[], byte[])" },
+		{ "Ljava/util/zip/ZipUtils;CENSIZ([BI)J",
+		  "long java.util.zip.ZipUtils.CENSIZ(byte[], int)" },
+		{ "Ljava/util/regex/Pattern$BmpCharProperty;match(Ljava/util/regex/Matcher;ILjava/lang/CharSequence;)Z",
+		  "boolean java.util.regex.Pattern$BmpCharProperty.match(java.util.regex.Matcher, int, java.lang.CharSequence)" },
+		{ "Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V",
+		  "void java.lang.AbstractStringBuilder.appendChars(java.lang.String, int, int)" },
+		{ "Ljava/lang/Object;<init>()V",
+		  "void java.lang.Object<init>()" },
+	};
+
+	for (i = 0; i < sizeof(test_cases) / sizeof(test_cases[0]); i++) {
+		buf = java_demangle_sym(test_cases[i].mangled, 0);
+		if (strcmp(buf, test_cases[i].demangled)) {
+			pr_debug("FAILED: %s: %s != %s\n", test_cases[i].mangled,
+				 buf, test_cases[i].demangled);
+			ret = TEST_FAIL;
+		}
+		free(buf);
+	}
+
+	return ret;
+}
tools/perf/tests/tests.h +1 -0
index 9a160fef47c9..49b791d978f6 100644
@@ -111,6 +111,7 @@ int test__mem2node(struct test *t, int subtest);
 int test__maps__merge_in(struct test *t, int subtest);
 int test__time_utils(struct test *t, int subtest);
 int test__jit_write_elf(struct test *test, int subtest);
+int test__demangle_java(struct test *test, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__bp_account_is_supported(void);
tools/perf/util/demangle-java.c +7 -6
index 6fb7f34c0814..39c05200ed65 100644
@@ -15,7 +15,7 @@ enum {
 	MODE_CLASS  = 1,
 	MODE_FUNC   = 2,
 	MODE_TYPE   = 3,
-	MODE_CTYPE  = 3, /* class arg */
+	MODE_CTYPE  = 4, /* class arg */
 };
 
 #define BASE_ENT(c, n)	[c - 'A']=n
@@ -27,7 +27,7 @@ static const char *base_types['Z' - 'A' + 1] = {
 	BASE_ENT('I', "int" ),
 	BASE_ENT('J', "long" ),
 	BASE_ENT('S', "short" ),
-	BASE_ENT('Z', "bool" ),
+	BASE_ENT('Z', "boolean" ),
 };
 
 /*
@@ -59,15 +59,16 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 
 		switch (*q) {
 		case 'L':
-			if (mode == MODE_PREFIX || mode == MODE_CTYPE) {
-				if (mode == MODE_CTYPE) {
+			if (mode == MODE_PREFIX || mode == MODE_TYPE) {
+				if (mode == MODE_TYPE) {
 					if (narg)
 						rlen += scnprintf(buf + rlen, maxlen - rlen, ", ");
 					narg++;
 				}
-				rlen += scnprintf(buf + rlen, maxlen - rlen, "class ");
 				if (mode == MODE_PREFIX)
 					mode = MODE_CLASS;
+				else
+					mode = MODE_CTYPE;
 			} else
 				buf[rlen++] = *q;
 			break;
@@ -120,7 +121,7 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 			if (mode != MODE_CLASS && mode != MODE_CTYPE)
 				goto error;
 			/* safe because at least one other char to process */
-			if (isalpha(*(q + 1)))
+			if (isalpha(*(q + 1)) && mode == MODE_CLASS)
 				rlen += scnprintf(buf + rlen, maxlen - rlen, ".");
 			if (mode == MODE_CLASS)
 				mode = MODE_FUNC;
Signature · 1 line
2.26.1
Signature · 1 line
- Arnaldo
maintainer Reviewed-by Ian Rogers Tested-by Tested-by Ian Rogers on v1 3/3
#10 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

In reply to Nick Gasson

Em Mon, Apr 27, 2020 at 02:15:16PM +0800, Nick Gasson escreveu:

Quoted 12 lines

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

Quoted 8 lines

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Also added a test case.

So, I took this and split into a patch for the new 'perf test java' and then the fix, so that we can see the problem being detected and then apply the fix and see it fixed, the last patch in this series thus became:

commit 341e11c1d445999932da3f5d626c9fe096949ae3

Author: Nick Gasson <[email protected]>
Date:   Mon Apr 27 14:15:16 2020 +0800

perf jvmti: Fix demangling Java symbols

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Committer notes:

This was split from a larger patch that also added a java demangler 'perf test' entry, that, before this patch shows the error being fixed by it:

$ perf test java
65: Demangle Java : FAILED!
$ perf test -v java
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
65: Demangle Java :
--- start ---
test child forked, pid 307264
FAILED: Ljava/lang/StringLatin1;equals([B[B)Z: bool class java.lang.StringLatin1.equals(byte[], byte[]) != boolean java.lang.StringLatin1.equals(byte[], byte[])
FAILED: Ljava/util/zip/ZipUtils;CENSIZ([BI)J: long class java.util.zip.ZipUtils.CENSIZ(byte[], int) != long java.util.zip.ZipUtils.CENSIZ(byte[], int)
FAILED: Ljava/util/regex/Pattern$BmpCharProperty;match(Ljava/util/regex/Matcher;ILjava/lang/CharSequence;)Z: bool class java.util.regex.Pattern$BmpCharProperty.match(class java.util.regex.Matcher., int, class java.lang., charhar, shortequence) != boolean java.util.regex.Pattern$BmpCharProperty.match(java.util.regex.Matcher, int, java.lang.CharSequence)
FAILED: Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V: void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int) != void java.lang.AbstractStringBuilder.appendChars(java.lang.String, int, int)
FAILED: Ljava/lang/Object;<init>()V: void class java.lang.Object<init>() != void java.lang.Object<init>()
test child finished with -1
---- end ----
Demangle Java: FAILED!
$

After applying this patch:

$ perf test java
65: Demangle Java : Ok
$

tools/perf/util/demangle-java.c +7 -6
index 6fb7f34c0814..39c05200ed65 100644
@@ -15,7 +15,7 @@ enum {
 	MODE_CLASS  = 1,
 	MODE_FUNC   = 2,
 	MODE_TYPE   = 3,
-	MODE_CTYPE  = 3, /* class arg */
+	MODE_CTYPE  = 4, /* class arg */
 };
 
 #define BASE_ENT(c, n)	[c - 'A']=n
@@ -27,7 +27,7 @@ static const char *base_types['Z' - 'A' + 1] = {
 	BASE_ENT('I', "int" ),
 	BASE_ENT('J', "long" ),
 	BASE_ENT('S', "short" ),
-	BASE_ENT('Z', "bool" ),
+	BASE_ENT('Z', "boolean" ),
 };
 
 /*
@@ -59,15 +59,16 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 
 		switch (*q) {
 		case 'L':
-			if (mode == MODE_PREFIX || mode == MODE_CTYPE) {
-				if (mode == MODE_CTYPE) {
+			if (mode == MODE_PREFIX || mode == MODE_TYPE) {
+				if (mode == MODE_TYPE) {
 					if (narg)
 						rlen += scnprintf(buf + rlen, maxlen - rlen, ", ");
 					narg++;
 				}
-				rlen += scnprintf(buf + rlen, maxlen - rlen, "class ");
 				if (mode == MODE_PREFIX)
 					mode = MODE_CLASS;
+				else
+					mode = MODE_CTYPE;
 			} else
 				buf[rlen++] = *q;
 			break;
@@ -120,7 +121,7 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 			if (mode != MODE_CLASS && mode != MODE_CTYPE)
 				goto error;
 			/* safe because at least one other char to process */
-			if (isalpha(*(q + 1)))
+			if (isalpha(*(q + 1)) && mode == MODE_CLASS)
 				rlen += scnprintf(buf + rlen, maxlen - rlen, ".");
 			if (mode == MODE_CLASS)
 				mode = MODE_FUNC;
#11 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

Em Wed, May 27, 2020 at 11:20:57AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 20 lines

Em Mon, Apr 27, 2020 at 02:15:16PM +0800, Nick Gasson escreveu:

Quoted 19 lines

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Quoted 6 lines
Quoted 1 line

Also added a test case.

So, I took this and split into a patch for the new 'perf test java' and then the fix, so that we can see the problem being detected and then apply the fix and see it fixed, the last patch in this series thus became:

This is in my tmp.perf/core branch pending a round of testing, after that it'll move to perf/core on its way to 5.8, thanks.

- Arnaldo

Quoted 118 lines

commit 341e11c1d445999932da3f5d626c9fe096949ae3

Author: Nick Gasson <[email protected]>
Date:   Mon Apr 27 14:15:16 2020 +0800

perf jvmti: Fix demangling Java symbols

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Committer notes:

This was split from a larger patch that also added a java demangler 'perf test' entry, that, before this patch shows the error being fixed by it:

$ perf test java
65: Demangle Java : FAILED!
$ perf test -v java
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
65: Demangle Java :
--- start ---
test child forked, pid 307264
FAILED: Ljava/lang/StringLatin1;equals([B[B)Z: bool class java.lang.StringLatin1.equals(byte[], byte[]) != boolean java.lang.StringLatin1.equals(byte[], byte[])
FAILED: Ljava/util/zip/ZipUtils;CENSIZ([BI)J: long class java.util.zip.ZipUtils.CENSIZ(byte[], int) != long java.util.zip.ZipUtils.CENSIZ(byte[], int)
FAILED: Ljava/util/regex/Pattern$BmpCharProperty;match(Ljava/util/regex/Matcher;ILjava/lang/CharSequence;)Z: bool class java.util.regex.Pattern$BmpCharProperty.match(class java.util.regex.Matcher., int, class java.lang., charhar, shortequence) != boolean java.util.regex.Pattern$BmpCharProperty.match(java.util.regex.Matcher, int, java.lang.CharSequence)
FAILED: Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V: void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int) != void java.lang.AbstractStringBuilder.appendChars(java.lang.String, int, int)
FAILED: Ljava/lang/Object;<init>()V: void class java.lang.Object<init>() != void java.lang.Object<init>()
test child finished with -1
---- end ----
Demangle Java: FAILED!
$

After applying this patch:

$ perf test java
65: Demangle Java : Ok
$

tools/perf/util/demangle-java.c +7 -6
index 6fb7f34c0814..39c05200ed65 100644
@@ -15,7 +15,7 @@ enum {
 	MODE_CLASS  = 1,
 	MODE_FUNC   = 2,
 	MODE_TYPE   = 3,
-	MODE_CTYPE  = 3, /* class arg */
+	MODE_CTYPE  = 4, /* class arg */
 };
 
 #define BASE_ENT(c, n)	[c - 'A']=n
@@ -27,7 +27,7 @@ static const char *base_types['Z' - 'A' + 1] = {
 	BASE_ENT('I', "int" ),
 	BASE_ENT('J', "long" ),
 	BASE_ENT('S', "short" ),
-	BASE_ENT('Z', "bool" ),
+	BASE_ENT('Z', "boolean" ),
 };
 
 /*
@@ -59,15 +59,16 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 
 		switch (*q) {
 		case 'L':
-			if (mode == MODE_PREFIX || mode == MODE_CTYPE) {
-				if (mode == MODE_CTYPE) {
+			if (mode == MODE_PREFIX || mode == MODE_TYPE) {
+				if (mode == MODE_TYPE) {
 					if (narg)
 						rlen += scnprintf(buf + rlen, maxlen - rlen, ", ");
 					narg++;
 				}
-				rlen += scnprintf(buf + rlen, maxlen - rlen, "class ");
 				if (mode == MODE_PREFIX)
 					mode = MODE_CLASS;
+				else
+					mode = MODE_CTYPE;
 			} else
 				buf[rlen++] = *q;
 			break;
@@ -120,7 +121,7 @@ __demangle_java_sym(const char *str, const char *end, char *buf, int maxlen, int
 			if (mode != MODE_CLASS && mode != MODE_CTYPE)
 				goto error;
 			/* safe because at least one other char to process */
-			if (isalpha(*(q + 1)))
+			if (isalpha(*(q + 1)) && mode == MODE_CLASS)
 				rlen += scnprintf(buf + rlen, maxlen - rlen, ".");
 			if (mode == MODE_CLASS)
 				mode = MODE_FUNC;
Signature · 1 line
- Arnaldo
#12 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

Em Wed, May 27, 2020 at 01:23:00PM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 23 lines

Em Wed, May 27, 2020 at 11:20:57AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 22 lines

Em Mon, Apr 27, 2020 at 02:15:16PM +0800, Nick Gasson escreveu:

Quoted 20 lines

For a Java method signature like:

Ljava/lang/AbstractStringBuilder;appendChars(Ljava/lang/String;II)V

The demangler produces:

void class java.lang.AbstractStringBuilder.appendChars(class java.lang., shorttring., int, int)

The arguments should be (java.lang.String, int, int) but the demangler interprets the "S" in String as the type code for "short". Correct this and two other minor things:

- There is no "bool" type in Java, should be "boolean".

- The demangler prepends "class" to every Java class name. This is not standard Java syntax and it wastes a lot of horizontal space if the signature is long. Remove this as there isn't any ambiguity between class names and primitives.

Also added a test case.

Quoted 7 lines
Quoted 4 lines

So, I took this and split into a patch for the new 'perf test java' and then the fix, so that we can see the problem being detected and then apply the fix and see it fixed, the last patch in this series thus became:

This is in my tmp.perf/core branch pending a round of testing, after that it'll move to perf/core on its way to 5.8, thanks.

All tests passed, moved to perf/core.

- Arnaldo

#13 lore raw

Re: [PATCH 3/3] perf jvmti: Fix demangling Java symbols

Quoted 6 lines · Arnaldo Carvalho de Melo
Quoted 3 lines

This is in my tmp.perf/core branch pending a round of testing, after that it'll move to perf/core on its way to 5.8, thanks.

All tests passed, moved to perf/core.

Great, thank you!

--
Nick

#14 lore raw
In reply to Nick Gasson
Quoted 1 line · Nick Gasson

Hi,

adding Stephane to the loop

jirka

Quoted 27 lines

These three patches fix a couple of issues I ran into while using the jitdump JVMTI agent to profile the SPECjbb benchmark.

Tested with OpenJDK 8 and 14.

Thanks,
Nick

Nick Gasson (3):
perf jvmti: Fix jitdump for methods without debug info perf jvmti: Do not report error when missing debug information perf jvmti: Fix demangling Java symbols

 tools/perf/jvmti/libjvmti.c           | 24 +++++++--------
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 6 files changed, 66 insertions(+), 19 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
Signature · 1 line
2.26.1
#15 lore raw
Quoted 9 lines · Jiri Olsa

adding Stephane to the loop

jirka

Quoted 4 lines

These three patches fix a couple of issues I ran into while using the jitdump JVMTI agent to profile the SPECjbb benchmark.

Hi, any feedback on these patches?

Thanks,
Nick

Quoted 19 lines
Quoted 19 lines

Nick Gasson (3):
perf jvmti: Fix jitdump for methods without debug info perf jvmti: Do not report error when missing debug information perf jvmti: Fix demangling Java symbols

 tools/perf/jvmti/libjvmti.c           | 24 +++++++--------
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 6 files changed, 66 insertions(+), 19 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
Signature · 1 line
2.26.1
#16 lore raw

Em Thu, May 14, 2020 at 04:56:07PM +0800, Nick Gasson escreveu:

Quoted 3 lines · Jiri Olsa
Quoted 2 lines

adding Stephane to the loop

Stephane, Ian, can you guys please take a look at this one and provide a Reviewed-by or Acked-by?

Thanks,

- Arnaldo

Quoted 32 lines
Quoted 6 lines

jirka

Quoted 4 lines

These three patches fix a couple of issues I ran into while using the jitdump JVMTI agent to profile the SPECjbb benchmark.

Hi, any feedback on these patches?

Thanks,
Nick

Quoted 19 lines
Quoted 19 lines

Nick Gasson (3):
perf jvmti: Fix jitdump for methods without debug info perf jvmti: Do not report error when missing debug information perf jvmti: Fix demangling Java symbols

 tools/perf/jvmti/libjvmti.c           | 24 +++++++--------
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 6 files changed, 66 insertions(+), 19 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
Signature · 1 line
2.26.1
Signature · 1 line
- Arnaldo
reviewer
#17 lore raw
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Thu, May 14, 2020 at 04:56:07PM +0800, Nick Gasson escreveu:

Quoted 2 lines · Jiri Olsa
Quoted 1 line
Quoted 22 lines
Quoted 1 line
Quoted 1 line

adding Stephane to the loop

Stephane, Ian, can you guys please take a look at this one and provide a Reviewed-by or Acked-by?

Thanks,

- Arnaldo

Quoted 11 lines
Quoted 6 lines

jirka

Quoted 4 lines

These three patches fix a couple of issues I ran into while using the jitdump JVMTI agent to profile the SPECjbb benchmark.

Hi, any feedback on these patches?

Thanks,
Nick

Thanks Nick,

If you are looking at this code I believe there is a bug in that the loop handling jvmtiCompiledMethodLoadInlineRecord is writing out the entire line number table before a pc and not just the line number table at the pc. This loop in do_get_line_numbers:

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 */
tab[lines].methodID = m;
lines++;
} else {

It could possibly make sense if it were iterating over the inline data in the jvmtiCompiledMethodLoadInlineRecord rather than the line number table.
Fixing this is toward the end of a list of things I need to look at.

Thanks,
Ian

Quoted 24 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines

Nick Gasson (3):
perf jvmti: Fix jitdump for methods without debug info perf jvmti: Do not report error when missing debug information perf jvmti: Fix demangling Java symbols

 tools/perf/jvmti/libjvmti.c           | 24 +++++++--------
 tools/perf/tests/Build                |  1 +
 tools/perf/tests/builtin-test.c       |  4 +++
 tools/perf/tests/demangle-java-test.c | 42 +++++++++++++++++++++++++++
 tools/perf/tests/tests.h              |  1 +
 tools/perf/util/demangle-java.c       | 13 +++++----
 6 files changed, 66 insertions(+), 19 deletions(-)
 create mode 100644 tools/perf/tests/demangle-java-test.c
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines

--
2.26.1

--

- Arnaldo

#18 lore raw
Quoted 14 lines · Ian Rogers

If you are looking at this code I believe there is a bug in that the loop handling jvmtiCompiledMethodLoadInlineRecord is writing out the entire line number table before a pc and not just the line number table at the pc. This loop in do_get_line_numbers:

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 */
tab[lines].methodID = m;
lines++;
} else {

Quoted 5 lines

It could possibly make sense if it were iterating over the inline data in the jvmtiCompiledMethodLoadInlineRecord rather than the line number table.
Fixing this is toward the end of a list of things I need to look at.

OK sure, I'll have a look at this. Thanks for the reviews.

--
Thanks,
Nick