perf tools: Display eBPF code in intel_pt trace

PATCH Superseded Jiri Olsa 3 participants 21 messages May 3, 2019 – May 8, 2019 perf events

Share & tools

Jiri Olsa

author reviewer

2019-05-03 08:18 UTC #0 lore raw

v2 cover perf tools: Display eBPF code in intel_pt trace

hi,
this patchset adds dso support to read and display bpf code in intel_pt trace output. I had to change some of the kernel maps processing code, so hopefully I did not break too many things ;-)

It's now possible to see bpf code flow via:

  # perf-with-kcore record pt -e intel_pt//ku -- sleep 1
  # perf-with-kcore script pt --insn-trace --xed
  ...
           sleep 36660 [016] 1057036.806464404:  ffffffff811dfba5 trace_call_bpf+0x65 ([kernel.kallsyms])               nopl  %eax, (%rax,%rax,1)
           sleep 36660 [016] 1057036.806464404:  ffffffff811dfbaa trace_call_bpf+0x6a ([kernel.kallsyms])               movq  0x30(%rbx), %rax
           sleep 36660 [016] 1057036.806464404:  ffffffff811dfbae trace_call_bpf+0x6e ([kernel.kallsyms])               leaq  0x38(%rbx), %rsi
           sleep 36660 [016] 1057036.806464404:  ffffffff811dfbb2 trace_call_bpf+0x72 ([kernel.kallsyms])               mov %r13, %rdi
           sleep 36660 [016] 1057036.806464404:  ffffffff811dfbb5 trace_call_bpf+0x75 ([kernel.kallsyms])               callq  0xffffffff81c00c30
           sleep 36660 [016] 1057036.806464404:  ffffffff81c00c30 __x86_indirect_thunk_rax+0x0 ([kernel.kallsyms])              callq  0xffffffff81c00c3c
           sleep 36660 [016] 1057036.806464404:  ffffffff81c00c3c __x86_indirect_thunk_rax+0xc ([kernel.kallsyms])              movq  %rax, (%rsp)
           sleep 36660 [016] 1057036.806464404:  ffffffff81c00c40 __x86_indirect_thunk_rax+0x10 ([kernel.kallsyms])             retq
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e89ef bpf_prog_da4fe6b3d2c29b25_trace_return+0x0 (bpf_prog_da4fe6b3d2c29b25_trace_return)           pushq  %rbp
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e89f0 bpf_prog_da4fe6b3d2c29b25_trace_return+0x1 (bpf_prog_da4fe6b3d2c29b25_trace_return)           mov %rsp, %rbp
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e89f3 bpf_prog_da4fe6b3d2c29b25_trace_return+0x4 (bpf_prog_da4fe6b3d2c29b25_trace_return)           sub $0x158, %rsp
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e89fa bpf_prog_da4fe6b3d2c29b25_trace_return+0xb (bpf_prog_da4fe6b3d2c29b25_trace_return)           sub $0x28, %rbp
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e89fe bpf_prog_da4fe6b3d2c29b25_trace_return+0xf (bpf_prog_da4fe6b3d2c29b25_trace_return)           movq  %rbx, (%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a02 bpf_prog_da4fe6b3d2c29b25_trace_return+0x13 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r13, 0x8(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a06 bpf_prog_da4fe6b3d2c29b25_trace_return+0x17 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r14, 0x10(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a0a bpf_prog_da4fe6b3d2c29b25_trace_return+0x1b (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r15, 0x18(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a0e bpf_prog_da4fe6b3d2c29b25_trace_return+0x1f (bpf_prog_da4fe6b3d2c29b25_trace_return)          xor %eax, %eax
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a10 bpf_prog_da4fe6b3d2c29b25_trace_return+0x21 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rax, 0x20(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a14 bpf_prog_da4fe6b3d2c29b25_trace_return+0x25 (bpf_prog_da4fe6b3d2c29b25_trace_return)          mov %rdi, %rbx
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a17 bpf_prog_da4fe6b3d2c29b25_trace_return+0x28 (bpf_prog_da4fe6b3d2c29b25_trace_return)          callq  0xffffffff811fed50
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed50 bpf_get_current_pid_tgid+0x0 ([kernel.kallsyms])              nopl  %eax, (%rax,%rax,1)
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed55 bpf_get_current_pid_tgid+0x5 ([kernel.kallsyms])              movq  %gs:0x15c00, %rdx
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed5e bpf_get_current_pid_tgid+0xe ([kernel.kallsyms])              test %rdx, %rdx
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed61 bpf_get_current_pid_tgid+0x11 ([kernel.kallsyms])             jz 0xffffffff811fed79
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed63 bpf_get_current_pid_tgid+0x13 ([kernel.kallsyms])             movsxdl  0x494(%rdx), %rax
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed6a bpf_get_current_pid_tgid+0x1a ([kernel.kallsyms])             movsxdl  0x490(%rdx), %rdx
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed71 bpf_get_current_pid_tgid+0x21 ([kernel.kallsyms])             shl $0x20, %rax
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed75 bpf_get_current_pid_tgid+0x25 ([kernel.kallsyms])             or %rdx, %rax
           sleep 36660 [016] 1057036.806464725:  ffffffff811fed78 bpf_get_current_pid_tgid+0x28 ([kernel.kallsyms])             retq
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a1c bpf_prog_da4fe6b3d2c29b25_trace_return+0x2d (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rax, -0x8(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a20 bpf_prog_da4fe6b3d2c29b25_trace_return+0x31 (bpf_prog_da4fe6b3d2c29b25_trace_return)          xor %edi, %edi
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a22 bpf_prog_da4fe6b3d2c29b25_trace_return+0x33 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rdi, -0x10(%rbp)
           sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a26 bpf_prog_da4fe6b3d2c29b25_trace_return+0x37 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rdi, -0x18(%rbp)

  # perf-core/perf-with-kcore script pt --call-trace
  ...
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                kretprobe_perf_func
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                    trace_call_bpf
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                        __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                            __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_pid_tgid
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_ktime_get_ns
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                    __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                        __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                __htab_map_lookup_elem
           sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms]                       )                                    memcmp
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_uid_gid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kgid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kuid
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                    perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                        perf_prepare_sample
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                            perf_misc_flags
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                    __x86_indirect_thunk_rax

v2 changes:
- fix missing pthread_mutex_unlock [Stanislav Fomichev]
- removed eBPF dso reusing [Adrian Hunter]
- merge eBPF and kcore maps [Adrian Hunter]
- few patches already applied
- added some perf script helper options
- patches are rebased on top of Arnaldo's perf/core with perf/urgent merged in

It's also available in here:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/fixes

thanks,
jirka

Cc Jesper Dangaard Brouer <[email protected]>

Cc Alexei Starovoitov <[email protected]>

Cc Daniel Borkmann <[email protected]>

---
Jiri Olsa (12):
perf tools: Separate generic code in dso__data_file_size perf tools: Separate generic code in dso_cache__read perf tools: Simplify dso_cache__read function
perf tools: Add bpf dso read and size hooks
perf tools: Read also the end of the kernel
perf tools: Keep zero in pgoff bpf map
perf script: Pad dso name for --call-trace
perf tools: Preserve eBPF maps when loading kcore perf tests: Add map_groups__merge_in test
perf script: Add --show-bpf-events to show eBPF related events perf script: Remove superfluous bpf event titles perf script: Add --show-all-events option

 tools/include/linux/kernel.h             |   1 +
 tools/lib/vsprintf.c                     |  19 +++++++++++++++++++
 tools/perf/Documentation/perf-script.txt |   6 ++++++
 tools/perf/builtin-script.c              |  56 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/Build                   |   1 +
 tools/perf/tests/builtin-test.c          |   4 ++++
 tools/perf/tests/map_groups.c            | 120 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h                 |   1 +
 tools/perf/util/dso.c                    | 125 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------------------------------
 tools/perf/util/event.c                  |   4 ++--
 tools/perf/util/machine.c                |  31 ++++++++++++++++++++-----------
 tools/perf/util/map.c                    |   6 ++++++
 tools/perf/util/map_groups.h             |   2 ++
 tools/perf/util/symbol.c                 |  97 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++----
 tools/perf/util/symbol_conf.h            |   1 +
 15 files changed, 421 insertions(+), 53 deletions(-)
 create mode 100644 tools/perf/tests/map_groups.c

Jiri Olsa

author reviewer

2019-05-03 08:18 UTC #1 lore raw

PATCH v2 1/12 perf tools: Separate generic code in dso__data_file_size

Moving file specific code in dso__data_file_size function into separate file_size function. I'll add bpf specific code in following patches.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/dso.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

tools/perf/util/dso.c +12 -7

index e059976d9d93..cb6199c1390a 100644
@@ -898,18 +898,12 @@ static ssize_t cached_read(struct dso *dso, struct machine *machine,
 	return r;
 }
 
-int dso__data_file_size(struct dso *dso, struct machine *machine)
+static int file_size(struct dso *dso, struct machine *machine)
 {
 	int ret = 0;
 	struct stat st;
 	char sbuf[STRERR_BUFSIZE];
 
-	if (dso->data.file_size)
-		return 0;
-
-	if (dso->data.status == DSO_DATA_STATUS_ERROR)
-		return -1;
-
 	pthread_mutex_lock(&dso__data_open_lock);
 
 	/*

@@ -938,6 +932,17 @@ int dso__data_file_size(struct dso *dso, struct machine *machine)
 	return ret;
 }
 
+int dso__data_file_size(struct dso *dso, struct machine *machine)
+{
+	if (dso->data.file_size)
+		return 0;
+
+	if (dso->data.status == DSO_DATA_STATUS_ERROR)
+		return -1;
+
+	return file_size(dso, machine);
+}
+
 /**
  * dso__data_size - Return dso data size
  * @dso: dso object

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:18 UTC #2 lore raw

PATCH v2 2/12 perf tools: Separate generic code in dso_cache__read

In reply to Jiri Olsa

Moving file specific code in dso_cache__read function into separate file_read function. I'll add bpf specific code in following patches.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/dso.c | 48 ++++++++++++++++++++++++-------------------
 1 file changed, 27 insertions(+), 21 deletions(-)

tools/perf/util/dso.c +27 ~~-21~~

index cb6199c1390a..7734f50a6912 100644
@@ -794,6 +794,31 @@ dso_cache__memcpy(struct dso_cache *cache, u64 offset,
 	return cache_size;
 }
 
+static ssize_t file_read(struct dso *dso, struct machine *machine,
+			 u64 offset, char *data)
+{
+	ssize_t ret;
+
+	pthread_mutex_lock(&dso__data_open_lock);
+
+	/*
+	 * dso->data.fd might be closed if other thread opened another
+	 * file (dso) due to open file limit (RLIMIT_NOFILE).
+	 */
+	try_to_open_dso(dso, machine);
+
+	if (dso->data.fd < 0) {
+		dso->data.status = DSO_DATA_STATUS_ERROR;
+		ret = -errno;
+		goto out;
+	}
+
+	ret = pread(dso->data.fd, data, DSO__DATA_CACHE_SIZE, offset);
+out:
+	pthread_mutex_unlock(&dso__data_open_lock);
+	return ret;
+}
+
 static ssize_t
 dso_cache__read(struct dso *dso, struct machine *machine,
 		u64 offset, u8 *data, ssize_t size)

@@ -803,37 +828,18 @@ dso_cache__read(struct dso *dso, struct machine *machine,
 	ssize_t ret;
 
 	do {
-		u64 cache_offset;
+		u64 cache_offset = offset & DSO__DATA_CACHE_MASK;
 
 		cache = zalloc(sizeof(*cache) + DSO__DATA_CACHE_SIZE);
 		if (!cache)
 			return -ENOMEM;
 
-		pthread_mutex_lock(&dso__data_open_lock);
-
-		/*
-		 * dso->data.fd might be closed if other thread opened another
-		 * file (dso) due to open file limit (RLIMIT_NOFILE).
-		 */
-		try_to_open_dso(dso, machine);
-
-		if (dso->data.fd < 0) {
-			ret = -errno;
-			dso->data.status = DSO_DATA_STATUS_ERROR;
-			break;
-		}
-
-		cache_offset = offset & DSO__DATA_CACHE_MASK;
-
-		ret = pread(dso->data.fd, cache->data, DSO__DATA_CACHE_SIZE, cache_offset);
-		if (ret <= 0)
-			break;
+		ret = file_read(dso, machine, cache_offset, cache->data);
 
 		cache->offset = cache_offset;
 		cache->size   = ret;
 	} while (0);
 
-	pthread_mutex_unlock(&dso__data_open_lock);
 
 	if (ret > 0) {
 		old = dso_cache__insert(dso, cache);

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #3 lore raw

PATCH v2 3/12 perf tools: Simplify dso_cache__read function

In reply to Jiri Olsa

There's no need for the while loop now, also we can connect two (ret > 0) condition legs together.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/dso.c | 17 ++++++-----------
 1 file changed, 6 insertions(+), 11 deletions(-)

tools/perf/util/dso.c +6 ~~-11~~

index 7734f50a6912..1e6a045adb8c 100644
@@ -823,25 +823,20 @@ static ssize_t
 dso_cache__read(struct dso *dso, struct machine *machine,
 		u64 offset, u8 *data, ssize_t size)
 {
+	u64 cache_offset = offset & DSO__DATA_CACHE_MASK;
 	struct dso_cache *cache;
 	struct dso_cache *old;
 	ssize_t ret;
 
-	do {
-		u64 cache_offset = offset & DSO__DATA_CACHE_MASK;
-
-		cache = zalloc(sizeof(*cache) + DSO__DATA_CACHE_SIZE);
-		if (!cache)
-			return -ENOMEM;
-
-		ret = file_read(dso, machine, cache_offset, cache->data);
+	cache = zalloc(sizeof(*cache) + DSO__DATA_CACHE_SIZE);
+	if (!cache)
+		return -ENOMEM;
 
+	ret = file_read(dso, machine, cache_offset, cache->data);
+	if (ret > 0) {
 		cache->offset = cache_offset;
 		cache->size   = ret;
-	} while (0);
-
 
-	if (ret > 0) {
 		old = dso_cache__insert(dso, cache);
 		if (old) {
 			/* we lose the race */

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #4 lore raw

PATCH v2 4/12 perf tools: Add bpf dso read and size hooks

In reply to Jiri Olsa

Adding bpf related code into dso reading paths to return size (bpf_size) and read the bpf code (bpf_read).

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/dso.c | 49 ++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 48 insertions(+), 1 deletion(-)

tools/perf/util/dso.c +48 -1

index 1e6a045adb8c..2c9289621efd 100644
@@ -9,6 +9,8 @@
 #include <errno.h>
 #include <fcntl.h>
 #include <libgen.h>
+#include <bpf/libbpf.h>
+#include "bpf-event.h"
 #include "compress.h"
 #include "namespaces.h"
 #include "path.h"

@@ -706,6 +708,44 @@ bool dso__data_status_seen(struct dso *dso, enum dso_data_status_seen by)
 	return false;
 }
 
+static ssize_t bpf_read(struct dso *dso, u64 offset, char *data)
+{
+	struct bpf_prog_info_node *node;
+	ssize_t size = DSO__DATA_CACHE_SIZE;
+	u64 len;
+	u8 *buf;
+
+	node = perf_env__find_bpf_prog_info(dso->bpf_prog.env, dso->bpf_prog.id);
+	if (!node || !node->info_linear) {
+		dso->data.status = DSO_DATA_STATUS_ERROR;
+		return -1;
+	}
+
+	len = node->info_linear->info.jited_prog_len;
+	buf = (u8 *) node->info_linear->info.jited_prog_insns;
+
+	if (offset >= len)
+		return -1;
+
+	size = (ssize_t) min(len - offset, (u64) size);
+	memcpy(data, buf + offset, size);
+	return size;
+}
+
+static int bpf_size(struct dso *dso)
+{
+	struct bpf_prog_info_node *node;
+
+	node = perf_env__find_bpf_prog_info(dso->bpf_prog.env, dso->bpf_prog.id);
+	if (!node || !node->info_linear) {
+		dso->data.status = DSO_DATA_STATUS_ERROR;
+		return -1;
+	}
+
+	dso->data.file_size = node->info_linear->info.jited_prog_len;
+	return 0;
+}
+
 static void
 dso_cache__free(struct dso *dso)
 {

@@ -832,7 +872,11 @@ dso_cache__read(struct dso *dso, struct machine *machine,
 	if (!cache)
 		return -ENOMEM;
 
-	ret = file_read(dso, machine, cache_offset, cache->data);
+	if (dso->binary_type == DSO_BINARY_TYPE__BPF_PROG_INFO)
+		ret = bpf_read(dso, cache_offset, cache->data);
+	else
+		ret = file_read(dso, machine, cache_offset, cache->data);
+
 	if (ret > 0) {
 		cache->offset = cache_offset;
 		cache->size   = ret;

@@ -941,6 +985,9 @@ int dso__data_file_size(struct dso *dso, struct machine *machine)
 	if (dso->data.status == DSO_DATA_STATUS_ERROR)
 		return -1;
 
+	if (dso->binary_type == DSO_BINARY_TYPE__BPF_PROG_INFO)
+		return bpf_size(dso);
+
 	return file_size(dso, machine);
 }

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #5 lore raw

PATCH v2 5/12 perf tools: Read also the end of the kernel

In reply to Jiri Olsa

We mark the end of kernel based on the first module, but that could cover some bpf program maps. Reading _etext symbol if it's present to get precise kernel map end.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/machine.c | 27 ++++++++++++++++++---------
 1 file changed, 18 insertions(+), 9 deletions(-)

tools/perf/util/machine.c +18 -9

index 3c520baa198c..ad0205fbb506 100644
@@ -924,7 +924,8 @@ const char *ref_reloc_sym_names[] = {"_text", "_stext", NULL};
  * symbol_name if it's not that important.
  */
 static int machine__get_running_kernel_start(struct machine *machine,
-					     const char **symbol_name, u64 *start)
+					     const char **symbol_name,
+					     u64 *start, u64 *end)
 {
 	char filename[PATH_MAX];
 	int i, err = -1;

@@ -949,6 +950,11 @@ static int machine__get_running_kernel_start(struct machine *machine,
 		*symbol_name = name;
 
 	*start = addr;
+
+	err = kallsyms__get_function_start(filename, "_etext", &addr);
+	if (!err)
+		*end = addr;
+
 	return 0;
 }

@@ -1440,7 +1446,7 @@ int machine__create_kernel_maps(struct machine *machine)
 	struct dso *kernel = machine__get_kernel(machine);
 	const char *name = NULL;
 	struct map *map;
-	u64 addr = 0;
+	u64 start = 0, end = ~0ULL;
 	int ret;
 
 	if (kernel == NULL)

@@ -1459,9 +1465,9 @@ int machine__create_kernel_maps(struct machine *machine)
 				 "continuing anyway...\n", machine->pid);
 	}
 
-	if (!machine__get_running_kernel_start(machine, &name, &addr)) {
+	if (!machine__get_running_kernel_start(machine, &name, &start, &end)) {
 		if (name &&
-		    map__set_kallsyms_ref_reloc_sym(machine->vmlinux_map, name, addr)) {
+		    map__set_kallsyms_ref_reloc_sym(machine->vmlinux_map, name, start)) {
 			machine__destroy_kernel_maps(machine);
 			ret = -1;
 			goto out_put;

@@ -1471,16 +1477,19 @@ int machine__create_kernel_maps(struct machine *machine)
 		 * we have a real start address now, so re-order the kmaps
 		 * assume it's the last in the kmaps
 		 */
-		machine__update_kernel_mmap(machine, addr, ~0ULL);
+		machine__update_kernel_mmap(machine, start, end);
 	}
 
 	if (machine__create_extra_kernel_maps(machine, kernel))
 		pr_debug("Problems creating extra kernel maps, continuing anyway...\n");
 
-	/* update end address of the kernel map using adjacent module address */
-	map = map__next(machine__kernel_map(machine));
-	if (map)
-		machine__set_kernel_mmap(machine, addr, map->start);
+	if (end == ~0ULL) {
+		/* update end address of the kernel map using adjacent module address */
+		map = map__next(machine__kernel_map(machine));
+		if (map)
+			machine__set_kernel_mmap(machine, start, map->start);
+	}
+
 out_put:
 	dso__put(kernel);
 	return ret;

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #6 lore raw

PATCH v2 6/12 perf tools: Keep zero in pgoff bpf map

In reply to Jiri Olsa

With pgoff set to zero, the map__map_ip function
will return bpf address based from 0, which is
what we need when we read the data from bpf dso.

Adding bpf symbols with mapped ip addresses as well.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/machine.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

tools/perf/util/machine.c +2 -2

index ad0205fbb506..d4aa44489011 100644
@@ -704,12 +704,12 @@ static int machine__process_ksymbol_register(struct machine *machine,
 			return -ENOMEM;
 
 		map->start = event->ksymbol_event.addr;
-		map->pgoff = map->start;
 		map->end = map->start + event->ksymbol_event.len;
 		map_groups__insert(&machine->kmaps, map);
 	}
 
-	sym = symbol__new(event->ksymbol_event.addr, event->ksymbol_event.len,
+	sym = symbol__new(map->map_ip(map, map->start),
+			  event->ksymbol_event.len,
 			  0, 0, event->ksymbol_event.name);
 	if (!sym)
 		return -ENOMEM;

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #7 lore raw

PATCH v2 7/12 perf script: Pad dso name for --call-trace

In reply to Jiri Olsa

Padding dso name for --call-trace so we don't have the indent screwed by different dso name lengths, as now for kernel there's also bpf code displayed.

  # perf-with-kcore record pt -e intel_pt//ku -- sleep 1
  # perf-core/perf-with-kcore script pt --call-trace

Before:

           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                  kretprobe_perf_func
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                      trace_call_bpf
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                          __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                              __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_get_current_pid_tgid
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_ktime_get_ns
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms])                                                      __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms])                                                          __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     __htab_map_lookup_elem
           sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms])                                                      memcmp
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                      probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                              check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                      probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                              check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_get_current_uid_gid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms])                                                      from_kgid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms])                                                      from_kuid
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                      perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                          perf_prepare_sample
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                              perf_misc_flags
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                                  __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                                      __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806466328: ([kvm])                                                                  kvm_is_in_guest
           sleep 36660 [016] 1057036.806466649: ([kernel.kallsyms])                                                              __perf_event_header__init_id.isra.0
           sleep 36660 [016] 1057036.806466649: ([kernel.kallsyms])                                                          perf_output_begin

After:

           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                kretprobe_perf_func
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                    trace_call_bpf
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                        __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                            __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_pid_tgid
           sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_ktime_get_ns
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                    __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                        __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                __htab_map_lookup_elem
           sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms]                       )                                    memcmp
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
           sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_uid_gid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kgid
           sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kuid
           sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                    perf_event_output
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                        perf_prepare_sample
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                            perf_misc_flags
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                __x86_indirect_thunk_rax
           sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                    __x86_indirect_thunk_rax

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/include/linux/kernel.h  |  1 +
 tools/lib/vsprintf.c          | 19 +++++++++++++++++++
 tools/perf/builtin-script.c   |  1 +
 tools/perf/util/map.c         |  6 ++++++
 tools/perf/util/symbol_conf.h |  1 +
 5 files changed, 28 insertions(+)

tools/include/linux/kernel.h +1 -0

index 857d9e22826e..cba226948a0c 100644
@@ -102,6 +102,7 @@
 
 int vscnprintf(char *buf, size_t size, const char *fmt, va_list args);
 int scnprintf(char * buf, size_t size, const char * fmt, ...);
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...);
 
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))

tools/lib/vsprintf.c +19 -0

index e08ee147eab4..149a15013b23 100644
@@ -23,3 +23,22 @@ int scnprintf(char * buf, size_t size, const char * fmt, ...)
 
        return (i >= ssize) ? (ssize - 1) : i;
 }
+
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...)
+{
+	ssize_t ssize = size;
+	va_list args;
+	int i;
+
+	va_start(args, fmt);
+	i = vsnprintf(buf, size, fmt, args);
+	va_end(args);
+
+	if (i < (int) size) {
+		for (; i < (int) size; i++)
+			buf[i] = ' ';
+		buf[i] = 0x0;
+	}
+
+	return (i >= ssize) ? (ssize - 1) : i;
+}

tools/perf/builtin-script.c +1 -0

index 61cfd8f70989..7adaa6c63a0b 100644
@@ -3297,6 +3297,7 @@ static int parse_call_trace(const struct option *opt __maybe_unused,
 	parse_output_fields(NULL, "-ip,-addr,-event,-period,+callindent", 0);
 	itrace_parse_synth_opts(opt, "cewp", 0);
 	symbol_conf.nanosecs = true;
+	symbol_conf.pad_output_len_dso = 50;
 	return 0;
 }

tools/perf/util/map.c +6 -0

index ee71efb9db62..c3fbd6e556b0 100644
@@ -405,6 +405,7 @@ size_t map__fprintf(struct map *map, FILE *fp)
 
 size_t map__fprintf_dsoname(struct map *map, FILE *fp)
 {
+	char buf[PATH_MAX];
 	const char *dsoname = "[unknown]";
 
 	if (map && map->dso) {

@@ -414,6 +415,11 @@ size_t map__fprintf_dsoname(struct map *map, FILE *fp)
 			dsoname = map->dso->name;
 	}
 
+	if (symbol_conf.pad_output_len_dso) {
+		scnprintf_pad(buf, symbol_conf.pad_output_len_dso, "%s", dsoname);
+		dsoname = buf;
+	}
+
 	return fprintf(fp, "%s", dsoname);
 }

tools/perf/util/symbol_conf.h +1 -0

index 6c55fa6fccec..382ba63fc554 100644
@@ -69,6 +69,7 @@ struct symbol_conf {
 			*tid_list;
 	const char	*symfs;
 	int		res_sample;
+	int		pad_output_len_dso;
 };
 
 extern struct symbol_conf symbol_conf;

Signature · 1 line

2.20.1

Song Liu

2019-05-06 21:39 UTC #8 lore raw

Re: [PATCH 07/12] perf script: Pad dso name for --call-trace

Quoted 103 lines · Jiri Olsa

Padding dso name for --call-trace so we don't have the indent screwed by different dso name lengths, as now for kernel there's also bpf code displayed.

 # perf-with-kcore record pt -e intel_pt//ku -- sleep 1
 # perf-core/perf-with-kcore script pt --call-trace

Before:

          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                  kretprobe_perf_func
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                      trace_call_bpf
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                          __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms])                                              __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_get_current_pid_tgid
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_ktime_get_ns
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms])                                                      __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms])                                                          __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     __htab_map_lookup_elem
          sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms])                                                      memcmp
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                      probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                              check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                      probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                              check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms])                                                          copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_get_current_uid_gid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms])                                                      from_kgid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms])                                                      from_kuid
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return)                                                     bpf_perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                      perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                          perf_prepare_sample
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                              perf_misc_flags
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                                  __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms])                                                                      __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806466328: ([kvm])                                                                  kvm_is_in_guest
          sleep 36660 [016] 1057036.806466649: ([kernel.kallsyms])                                                              __perf_event_header__init_id.isra.0
          sleep 36660 [016] 1057036.806466649: ([kernel.kallsyms])                                                          perf_output_begin

After:

          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                kretprobe_perf_func
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                    trace_call_bpf
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                        __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                            __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_pid_tgid
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_ktime_get_ns
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                    __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                        __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                __htab_map_lookup_elem
          sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms]                       )                                    memcmp
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_uid_gid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kgid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kuid
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                    perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                        perf_prepare_sample
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                            perf_misc_flags
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                    __x86_indirect_thunk_rax

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---

tools/include/linux/kernel.h  |  1 +
tools/lib/vsprintf.c          | 19 +++++++++++++++++++
tools/perf/builtin-script.c   |  1 +
tools/perf/util/map.c         |  6 ++++++
tools/perf/util/symbol_conf.h |  1 +
5 files changed, 28 insertions(+)

tools/include/linux/kernel.h +1 -0

index 857d9e22826e..cba226948a0c 100644
@@ -102,6 +102,7 @@
 
 int vscnprintf(char *buf, size_t size, const char *fmt, va_list args);
 int scnprintf(char * buf, size_t size, const char * fmt, ...);
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...);
 
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))

tools/lib/vsprintf.c +0 -0

index e08ee147eab4..149a15013b23 100644

tools/lib/vsprintf.c int scnprintf(char * buf, size_t size, const char * fmt, ...) · line 31

Quoted 9 lines

tools/lib/vsprintf.c +6 -0

@@ -23,3 +23,22 @@ int scnprintf(char * buf, size_t size, const char * fmt, ...)
 
       return (i >= ssize) ? (ssize - 1) : i;
 }
+
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...)
+{
+	ssize_t ssize = size;
+	va_list args;
+	int i;

nit: I guess we can avoid mixing int, ssize_t and size_t here?

Quoted 30 lines

tools/lib/vsprintf.c +13 -0

+
+	va_start(args, fmt);
+	i = vsnprintf(buf, size, fmt, args);
+	va_end(args);
+
+	if (i < (int) size) {
+		for (; i < (int) size; i++)
+			buf[i] = ' ';
+		buf[i] = 0x0;
+	}
+
+	return (i >= ssize) ? (ssize - 1) : i;
+}

tools/perf/builtin-script.c +1 -0

index 61cfd8f70989..7adaa6c63a0b 100644
@@ -3297,6 +3297,7 @@ static int parse_call_trace(const struct option *opt __maybe_unused,
 	parse_output_fields(NULL, "-ip,-addr,-event,-period,+callindent", 0);
 	itrace_parse_synth_opts(opt, "cewp", 0);
 	symbol_conf.nanosecs = true;
+	symbol_conf.pad_output_len_dso = 50;
 	return 0;
 }

tools/perf/util/map.c +0 -0

index ee71efb9db62..c3fbd6e556b0 100644

tools/perf/util/map.c size_t map__fprintf(struct map *map, FILE *fp) · line 408

Quoted 4 lines

tools/perf/util/map.c +1 -0

@@ -405,6 +405,7 @@ size_t map__fprintf(struct map *map, FILE *fp)
 
 size_t map__fprintf_dsoname(struct map *map, FILE *fp)
 {
+	char buf[PATH_MAX];

nit: PATH_MAX vs. 50 is a little weird.

Quoted 30 lines

tools/perf/util/map.c +5 -0

 	const char *dsoname = "[unknown]";
 
 	if (map && map->dso) {

@@ -414,6 +415,11 @@ size_t map__fprintf_dsoname(struct map *map, FILE *fp)
 			dsoname = map->dso->name;
 	}
 
+	if (symbol_conf.pad_output_len_dso) {
+		scnprintf_pad(buf, symbol_conf.pad_output_len_dso, "%s", dsoname);
+		dsoname = buf;
+	}
+
 	return fprintf(fp, "%s", dsoname);
 }

tools/perf/util/symbol_conf.h +1 -0

index 6c55fa6fccec..382ba63fc554 100644
@@ -69,6 +69,7 @@ struct symbol_conf {
 			*tid_list;
 	const char	*symfs;
 	int		res_sample;
+	int		pad_output_len_dso;
 };
 
 extern struct symbol_conf symbol_conf;

Signature · 1 line

2.20.1

Jiri Olsa

2019-05-07 08:13 UTC #9 lore raw

Re: [PATCH 07/12] perf script: Pad dso name for --call-trace

On Mon, May 06, 2019 at 09:38:55PM +0000, Song Liu wrote:

SNIP

Quoted 29 lines

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---

tools/include/linux/kernel.h  |  1 +
tools/lib/vsprintf.c          | 19 +++++++++++++++++++
tools/perf/builtin-script.c   |  1 +
tools/perf/util/map.c         |  6 ++++++
tools/perf/util/symbol_conf.h |  1 +
5 files changed, 28 insertions(+)

tools/include/linux/kernel.h +1 -0

index 857d9e22826e..cba226948a0c 100644
@@ -102,6 +102,7 @@
 
 int vscnprintf(char *buf, size_t size, const char *fmt, va_list args);
 int scnprintf(char * buf, size_t size, const char * fmt, ...);
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...);
 
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))

tools/lib/vsprintf.c +0 -0

index e08ee147eab4..149a15013b23 100644
@@ -23,3 +23,22 @@ int scnprintf(char * buf, size_t size, const char * fmt, ...)

Quoted 10 lines

tools/lib/vsprintf.c int scnprintf(char * buf, size_t size, const char * fmt, ...) · line 31
Quoted 8 lines
tools/lib/vsprintf.c +6 -0
       return (i >= ssize) ? (ssize - 1) : i;
 }
+
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...)
+{
+	ssize_t ssize = size;
+	va_list args;
+	int i;
nit: I guess we can avoid mixing int, ssize_t and size_t here?

I copied that from scnprintf ;-)

the thing is that at the end we call vsnprintf, which takes size_t as size param and returns int, so there will be casting at some point in any case..

I guess the ssize_t was introduced to compare the size_t value with int

Quoted 33 lines

tools/lib/vsprintf.c +13 -0

+
+	va_start(args, fmt);
+	i = vsnprintf(buf, size, fmt, args);
+	va_end(args);
+
+	if (i < (int) size) {
+		for (; i < (int) size; i++)
+			buf[i] = ' ';
+		buf[i] = 0x0;
+	}
+
+	return (i >= ssize) ? (ssize - 1) : i;
+}

tools/perf/builtin-script.c +1 -0

index 61cfd8f70989..7adaa6c63a0b 100644
@@ -3297,6 +3297,7 @@ static int parse_call_trace(const struct option *opt __maybe_unused,
 	parse_output_fields(NULL, "-ip,-addr,-event,-period,+callindent", 0);
 	itrace_parse_synth_opts(opt, "cewp", 0);
 	symbol_conf.nanosecs = true;
+	symbol_conf.pad_output_len_dso = 50;
 	return 0;
 }

tools/perf/util/map.c +0 -0

index ee71efb9db62..c3fbd6e556b0 100644
@@ -405,6 +405,7 @@ size_t map__fprintf(struct map *map, FILE *fp)

Quoted 5 lines

tools/perf/util/map.c size_t map__fprintf(struct map *map, FILE *fp) · line 408
Quoted 3 lines
tools/perf/util/map.c +1 -0
 size_t map__fprintf_dsoname(struct map *map, FILE *fp)
 {
+	char buf[PATH_MAX];
nit: PATH_MAX vs. 50 is a little weird.

right, the "staying on the safe side" is too big here, will change

thanks,
jirka

Song Liu

2019-05-07 18:29 UTC #10 lore raw

Re: [PATCH 07/12] perf script: Pad dso name for --call-trace

Quoted 48 lines · Jiri Olsa

On Mon, May 06, 2019 at 09:38:55PM +0000, Song Liu wrote:
SNIP
Quoted 29 lines
Quoted 29 lines
Link http://lkml.kernel.org/n/[email protected] open here
Signed-off-by Jiri Olsa <[email protected]>
---
tools/include/linux/kernel.h  |  1 +
tools/lib/vsprintf.c          | 19 +++++++++++++++++++
tools/perf/builtin-script.c   |  1 +
tools/perf/util/map.c         |  6 ++++++
tools/perf/util/symbol_conf.h |  1 +
5 files changed, 28 insertions(+)
tools/include/linux/kernel.h +1 -0
index 857d9e22826e..cba226948a0c 100644
@@ -102,6 +102,7 @@
 
 int vscnprintf(char *buf, size_t size, const char *fmt, va_list args);
 int scnprintf(char * buf, size_t size, const char * fmt, ...);
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...);
 
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
 
tools/lib/vsprintf.c +0 -0
index e08ee147eab4..149a15013b23 100644
@@ -23,3 +23,22 @@ int scnprintf(char * buf, size_t size, const char * fmt, ...)
 
Quoted 10 lines
tools/lib/vsprintf.c int scnprintf(char * buf, size_t size, const char * fmt, ...) · line 31
Quoted 8 lines
tools/lib/vsprintf.c +6 -0
      return (i >= ssize) ? (ssize - 1) : i;
 }
+
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...)
+{
+	ssize_t ssize = size;
+	va_list args;
+	int i;
nit: I guess we can avoid mixing int, ssize_t and size_t here?

Quoted 8 lines

I copied that from scnprintf ;-)
the thing is that at the end we call vsnprintf, which takes size_t as size param and returns int, so there will be casting at some point in any case..
I guess the ssize_t was introduced to compare the size_t value with int

Interesting. Given scnprintf works fine, I think we can keep the patch as-is.

Thanks,
Song

Jiri Olsa

2019-05-08 07:40 UTC #11 lore raw

Re: [PATCH 07/12] perf script: Pad dso name for --call-trace

Quoted 54 lines · Song Liu

Quoted 54 lines · Jiri Olsa
On Mon, May 06, 2019 at 09:38:55PM +0000, Song Liu wrote:
SNIP
Quoted 29 lines
Quoted 29 lines
Link http://lkml.kernel.org/n/[email protected] open here
Signed-off-by Jiri Olsa <[email protected]>
---
tools/include/linux/kernel.h  |  1 +
tools/lib/vsprintf.c          | 19 +++++++++++++++++++
tools/perf/builtin-script.c   |  1 +
tools/perf/util/map.c         |  6 ++++++
tools/perf/util/symbol_conf.h |  1 +
5 files changed, 28 insertions(+)
tools/include/linux/kernel.h +1 -0
index 857d9e22826e..cba226948a0c 100644
@@ -102,6 +102,7 @@
 
 int vscnprintf(char *buf, size_t size, const char *fmt, va_list args);
 int scnprintf(char * buf, size_t size, const char * fmt, ...);
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...);
 
 #define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
 
tools/lib/vsprintf.c +0 -0
index e08ee147eab4..149a15013b23 100644
@@ -23,3 +23,22 @@ int scnprintf(char * buf, size_t size, const char * fmt, ...)
 
Quoted 10 lines
tools/lib/vsprintf.c int scnprintf(char * buf, size_t size, const char * fmt, ...) · line 31
Quoted 8 lines
tools/lib/vsprintf.c +6 -0
      return (i >= ssize) ? (ssize - 1) : i;
 }
+
+int scnprintf_pad(char * buf, size_t size, const char * fmt, ...)
+{
+	ssize_t ssize = size;
+	va_list args;
+	int i;
nit: I guess we can avoid mixing int, ssize_t and size_t here?
I copied that from scnprintf ;-)
the thing is that at the end we call vsnprintf, which takes size_t as size param and returns int, so there will be casting at some point in any case..

Quoted 5 lines

Quoted 2 lines
I guess the ssize_t was introduced to compare the size_t value with int
Interesting. Given scnprintf works fine, I think we can keep the patch as-is.

I actualy found off by one issue in here.. I'll send new version

thanks,
jirka

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #12 lore raw

PATCH v2 8/12 perf tools: Preserve eBPF maps when loading kcore

In reply to Jiri Olsa

We need to preserve eBPF maps even if they are covered by kcore, because we need to access eBPF dso for source data.

Adding map_groups__merge_in function to do that. It merges map into map_groups by splitting the new map within the existing map regions.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/symbol.c | 97 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 93 insertions(+), 4 deletions(-)

tools/perf/util/symbol.c +93 -4

index 5cbad55cd99d..29780fcd049c 100644
@@ -1166,6 +1166,85 @@ static int kcore_mapfn(u64 start, u64 len, u64 pgoff, void *data)
 	return 0;
 }
 
+/*
+ * Merges map into map_groups by splitting the new map
+ * within the existing map regions.
+ */
+static int map_groups__merge_in(struct map_groups *kmaps, struct map *new_map)
+{
+	struct map *old_map;
+	LIST_HEAD(merged);
+
+	for (old_map = map_groups__first(kmaps); old_map;
+	     old_map = map_groups__next(old_map)) {
+
+		/* no overload with this one */
+		if (new_map->end < old_map->start ||
+		    new_map->start >= old_map->end)
+			continue;
+
+		if (new_map->start < old_map->start) {
+			/*
+			 * |new......
+			 *       |old....
+			 */
+			if (new_map->end < old_map->end) {
+				/*
+				 * |new......|     -> |new..|
+				 *       |old....| ->       |old....|
+				 */
+				new_map->end = old_map->start;
+			} else {
+				/*
+				 * |new.............| -> |new..|       |new..|
+				 *       |old....|    ->       |old....|
+				 */
+				struct map *m = map__clone(new_map);
+
+				if (!m)
+					return -ENOMEM;
+
+				m->end = old_map->start;
+				list_add_tail(&m->node, &merged);
+				new_map->start = old_map->end;
+			}
+		} else {
+			/*
+			 *      |new......
+			 * |old....
+			 */
+			if (new_map->end < old_map->end) {
+				/*
+				 *      |new..|   -> x
+				 * |old.........| -> |old.........|
+				 */
+				map__put(new_map);
+				new_map = NULL;
+				break;
+			} else {
+				/*
+				 *      |new......| ->         |new...|
+				 * |old....|        -> |old....|
+				 */
+				new_map->start = old_map->end;
+			}
+		}
+	}
+
+	while (!list_empty(&merged)) {
+		old_map = list_entry(merged.next, struct map, node);
+		list_del_init(&old_map->node);
+		map_groups__insert(kmaps, old_map);
+		map__put(old_map);
+	}
+
+	if (new_map) {
+		map_groups__insert(kmaps, new_map);
+		map__put(new_map);
+	}
+	return 0;
+}
+
 static int dso__load_kcore(struct dso *dso, struct map *map,
 			   const char *kallsyms_filename)
 {

@@ -1222,7 +1301,12 @@ static int dso__load_kcore(struct dso *dso, struct map *map,
 	while (old_map) {
 		struct map *next = map_groups__next(old_map);
 
-		if (old_map != map)
+		/*
+		 * We need to preserve eBPF maps even if they are
+		 * covered by kcore, because we need to access
+		 * eBPF dso for source data.
+		 */
+		if (old_map != map && !__map__is_bpf_prog(old_map))
 			map_groups__remove(kmaps, old_map);
 		old_map = next;
 	}

@@ -1256,11 +1340,16 @@ static int dso__load_kcore(struct dso *dso, struct map *map,
 			map_groups__remove(kmaps, map);
 			map_groups__insert(kmaps, map);
 			map__put(map);
+			map__put(new_map);
 		} else {
-			map_groups__insert(kmaps, new_map);
+			/*
+			 * Merge kcore map into existing maps,
+			 * and ensure that current maps (eBPF)
+			 * stay intact.
+			 */
+			if (map_groups__merge_in(kmaps, new_map))
+				goto out_err;
 		}
-
-		map__put(new_map);
 	}
 
 	if (machine__is(machine, "x86_64")) {

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #13 lore raw

PATCH v2 9/12 perf tests: Add map_groups__merge_in test

In reply to Jiri Olsa

Adding map_groups__merge_in to test map_groups__merge_in function usage - merging kcore maps into existing eBPF maps.

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   4 ++
 tools/perf/tests/map_groups.c   | 120 ++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/map_groups.h    |   2 +
 tools/perf/util/symbol.c        |   2 +-
 6 files changed, 129 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/tests/map_groups.c

tools/perf/tests/Build +1 -0

index 0b2b8305c965..4afb6319ed51 100644
@@ -50,6 +50,7 @@ perf-y += perf-hooks.o
 perf-y += clang.o
 perf-y += unit_number__scnprintf.o
 perf-y += mem2node.o
+perf-y += map_groups.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 9852b5d624a5..941c5456d625 100644
@@ -289,6 +289,10 @@ static struct test generic_tests[] = {
 		.desc = "mem2node",
 		.func = test__mem2node,
 	},
+	{
+		.desc = "map_groups__merge_in",
+		.func = test__map_groups__merge_in,
+	},
 	{
 		.func = NULL,
 	},

tools/perf/tests/map_groups.c new file +120 -0

120 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0

index 399f18ca71a3..e5e3a57cd373 100644
@@ -107,6 +107,7 @@ const char *test__clang_subtest_get_desc(int subtest);
 int test__clang_subtest_get_nr(void);
 int test__unit_number__scnprint(struct test *test, int subtest);
 int test__mem2node(struct test *t, int subtest);
+int test__map_groups__merge_in(struct test *t, int subtest);
 
 bool test__bp_signal_is_supported(void);
 bool test__wp_is_supported(void);

tools/perf/util/map_groups.h +2 -0

index 4dcda33e0fdf..5f25efa6d6bc 100644
@@ -88,4 +88,6 @@ int map_groups__fixup_overlappings(struct map_groups *mg, struct map *map, FILE
 
 struct map *map_groups__find_by_name(struct map_groups *mg, const char *name);
 
+int map_groups__merge_in(struct map_groups *kmaps, struct map *new_map);
+
 #endif // __PERF_MAP_GROUPS_H

tools/perf/util/symbol.c +1 -1

index 29780fcd049c..f4540f8bbed1 100644
@@ -1170,7 +1170,7 @@ static int kcore_mapfn(u64 start, u64 len, u64 pgoff, void *data)
  * Merges map into map_groups by splitting the new map
  * within the existing map regions.
  */
-static int map_groups__merge_in(struct map_groups *kmaps, struct map *new_map)
+int map_groups__merge_in(struct map_groups *kmaps, struct map *new_map)
 {
 	struct map *old_map;
 	LIST_HEAD(merged);

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #14 lore raw

PATCH v2 10/12 perf script: Add --show-bpf-events to show eBPF related events

In reply to Jiri Olsa

Adding --show-bpf-events to show eBPF related events:
PERF_RECORD_KSYMBOL
PERF_RECORD_BPF_EVENT

Usage:

  # perf record -a
  ...
  # perf script --show-bpf-events
  ...
  swapper     0 [000]     0.000000: PERF_RECORD_KSYMBOL ksymbol event with addr ffffffffc0ef971d len 229 type 1 flags 0x0 name bpf_prog_2a142ef67aaad174
  swapper     0 [000]     0.000000: PERF_RECORD_BPF_EVENT bpf event with type 1, flags 0, id 36
  ...

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/Documentation/perf-script.txt |  3 ++
 tools/perf/builtin-script.c              | 42 ++++++++++++++++++++++++
 2 files changed, 45 insertions(+)

tools/perf/Documentation/perf-script.txt +3 -0

index 9b0d04dd2a61..af8282782911 100644
@@ -313,6 +313,9 @@ OPTIONS
 --show-round-events
 	Display finished round events i.e. events of type PERF_RECORD_FINISHED_ROUND.
 
+--show-bpf-events
+	Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
+
 --demangle::
 	Demangle symbol names to human readable form. It's enabled by default,
 	disable with --no-demangle.

tools/perf/builtin-script.c +42 -0

66 lines not on this page · raw message

Signature · 1 line

2.20.1

Song Liu

2019-05-06 21:42 UTC #15 lore raw

Re: [PATCH 10/12] perf script: Add --show-bpf-events to show eBPF related events

Quoted 93 lines · Jiri Olsa

Adding --show-bpf-events to show eBPF related events:
PERF_RECORD_KSYMBOL
PERF_RECORD_BPF_EVENT

Usage:

 # perf record -a
 ...
 # perf script --show-bpf-events
 ...
 swapper     0 [000]     0.000000: PERF_RECORD_KSYMBOL ksymbol event with addr ffffffffc0ef971d len 229 type 1 flags 0x0 name bpf_prog_2a142ef67aaad174
 swapper     0 [000]     0.000000: PERF_RECORD_BPF_EVENT bpf event with type 1, flags 0, id 36
 ...

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
tools/perf/Documentation/perf-script.txt | 3 ++
tools/perf/builtin-script.c | 42 ++++++++++++++++++++++++

2 files changed, 45 insertions(+)

tools/perf/Documentation/perf-script.txt +3 -2

index 9b0d04dd2a61..af8282782911 100644
@@ -313,6 +313,9 @@ OPTIONS
--show-round-events
 	Display finished round events i.e. events of type PERF_RECORD_FINISHED_ROUND.
 
+--show-bpf-events
+	Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
+
--demangle::
 	Demangle symbol names to human readable form. It's enabled by default,
 	disable with --no-demangle.

tools/perf/builtin-script.c +36 -0

index 7adaa6c63a0b..3a48a2627670 100644
@@ -1606,6 +1606,7 @@ struct perf_script {
 	bool			show_namespace_events;
 	bool			show_lost_events;
 	bool			show_round_events;
+	bool			show_bpf_events;
 	bool			allocated;
 	bool			per_event_dump;
 	struct cpu_map		*cpus;

@@ -2318,6 +2319,41 @@ process_finished_round_event(struct perf_tool *tool __maybe_unused,
 	return 0;
 }
 
+static int
+process_bpf_events(struct perf_tool *tool __maybe_unused,
+		   union perf_event *event,
+		   struct perf_sample *sample,
+		   struct machine *machine)
+{
+	struct thread *thread;
+	struct perf_script *script = container_of(tool, struct perf_script, tool);
+	struct perf_session *session = script->session;
+	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
+
+	if (machine__process_ksymbol(machine, event, sample) < 0)
+		return -1;
+
+	if (!evsel->attr.sample_id_all) {
+		perf_event__fprintf(event, stdout);
+		return 0;
+	}
+
+	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
+	if (thread == NULL) {
+		pr_debug("problem processing MMAP event, skipping it.\n");
+		return -1;
+	}
+
+	if (!filter_cpu(sample)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   event->header.type, stdout);
+		perf_event__fprintf(event, stdout);
+	}
+
+	thread__put(thread);
+	return 0;
+}
+
 static void sig_handler(int sig __maybe_unused)
 {
 	session_done = 1;

tools/perf/builtin-script.c static int __cmd_script(struct perf_script *script) · line 2461

Quoted 6 lines

tools/perf/builtin-script.c +3 -0

@@ -2420,6 +2456,10 @@ static int __cmd_script(struct perf_script *script)
 		script->tool.ordered_events = false;
 		script->tool.finished_round = process_finished_round_event;
 	}
+	if (script->show_bpf_events) {
+		script->tool.ksymbol   = process_bpf_events;
+		script->tool.bpf_event = process_bpf_events;

Why do we need both set to process_bpf_events?

Thanks,
Song

Quoted 16 lines

tools/perf/builtin-script.c +3 -0

+	}
 
 	if (perf_script__setup_per_event_dump(script)) {
 		pr_err("Couldn't create the per event dump files\n");

@@ -3439,6 +3479,8 @@ int cmd_script(int argc, const char **argv)
 		    "Show lost events (if recorded)"),
 	OPT_BOOLEAN('\0', "show-round-events", &script.show_round_events,
 		    "Show round events (if recorded)"),
+	OPT_BOOLEAN('\0', "show-bpf-events", &script.show_bpf_events,
+		    "Show bpf related events (if recorded)"),
 	OPT_BOOLEAN('\0', "per-event-dump", &script.per_event_dump,
 		    "Dump trace output to files named by the monitored events"),
 	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),

Signature · 1 line

2.20.1

Jiri Olsa

2019-05-07 08:18 UTC #16 lore raw

Re: [PATCH 10/12] perf script: Add --show-bpf-events to show eBPF related events

On Mon, May 06, 2019 at 09:42:44PM +0000, Song Liu wrote:

SNIP

Quoted 39 lines

tools/perf/builtin-script.c +35 -0

+static int
+process_bpf_events(struct perf_tool *tool __maybe_unused,
+		   union perf_event *event,
+		   struct perf_sample *sample,
+		   struct machine *machine)
+{
+	struct thread *thread;
+	struct perf_script *script = container_of(tool, struct perf_script, tool);
+	struct perf_session *session = script->session;
+	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
+
+	if (machine__process_ksymbol(machine, event, sample) < 0)
+		return -1;
+
+	if (!evsel->attr.sample_id_all) {
+		perf_event__fprintf(event, stdout);
+		return 0;
+	}
+
+	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
+	if (thread == NULL) {
+		pr_debug("problem processing MMAP event, skipping it.\n");
+		return -1;
+	}
+
+	if (!filter_cpu(sample)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   event->header.type, stdout);
+		perf_event__fprintf(event, stdout);
+	}
+
+	thread__put(thread);
+	return 0;
+}
+

static void sig_handler(int sig __maybe_unused)

{
	session_done = 1;

(quoted diff) +0 -0

@@ -2420,6 +2456,10 @@ static int __cmd_script(struct perf_script *script)

Quoted 8 lines

tools/perf/builtin-script.c static int __cmd_script(struct perf_script *script) · line 2461
Quoted 6 lines
tools/perf/builtin-script.c +3 -0
 		script->tool.ordered_events = false;
 		script->tool.finished_round = process_finished_round_event;
 	}
+	if (script->show_bpf_events) {
+		script->tool.ksymbol   = process_bpf_events;
+		script->tool.bpf_event = process_bpf_events;
Why do we need both set to process_bpf_events?

--show-*-events option is there to display all the related events for given '*'

we want to display both ksymbol and bpf_event in here, process_bpf_events takes care of it for both of them

thanks,
jirka

Song Liu

2019-05-07 18:27 UTC #17 lore raw

Re: [PATCH 10/12] perf script: Add --show-bpf-events to show eBPF related events

Quoted 56 lines · Jiri Olsa

On Mon, May 06, 2019 at 09:42:44PM +0000, Song Liu wrote:

SNIP

Quoted 39 lines

tools/perf/builtin-script.c +35 -0

+static int
+process_bpf_events(struct perf_tool *tool __maybe_unused,
+		   union perf_event *event,
+		   struct perf_sample *sample,
+		   struct machine *machine)
+{
+	struct thread *thread;
+	struct perf_script *script = container_of(tool, struct perf_script, tool);
+	struct perf_session *session = script->session;
+	struct perf_evsel *evsel = perf_evlist__id2evsel(session->evlist, sample->id);
+
+	if (machine__process_ksymbol(machine, event, sample) < 0)
+		return -1;
+
+	if (!evsel->attr.sample_id_all) {
+		perf_event__fprintf(event, stdout);
+		return 0;
+	}
+
+	thread = machine__findnew_thread(machine, sample->pid, sample->tid);
+	if (thread == NULL) {
+		pr_debug("problem processing MMAP event, skipping it.\n");
+		return -1;
+	}
+
+	if (!filter_cpu(sample)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   event->header.type, stdout);
+		perf_event__fprintf(event, stdout);
+	}
+
+	thread__put(thread);
+	return 0;
+}
+

static void sig_handler(int sig __maybe_unused)

{
	session_done = 1;

(quoted diff) +0 -0

@@ -2420,6 +2456,10 @@ static int __cmd_script(struct perf_script *script)

Quoted 8 lines

tools/perf/builtin-script.c static int __cmd_script(struct perf_script *script) · line 2461
Quoted 6 lines
tools/perf/builtin-script.c +3 -0
 		script->tool.ordered_events = false;
 		script->tool.finished_round = process_finished_round_event;
 	}
+	if (script->show_bpf_events) {
+		script->tool.ksymbol   = process_bpf_events;
+		script->tool.bpf_event = process_bpf_events;
Why do we need both set to process_bpf_events?

Quoted 4 lines

--show-*-events option is there to display all the related events for given '*'
we want to display both ksymbol and bpf_event in here, process_bpf_events takes care of it for both of them

I see. Thanks for the explanation!

Song

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #18 lore raw

PATCH v2 11/12 perf script: Remove superfluous bpf event titles

In reply to Jiri Olsa

There's no need to display "ksymbol event with" text for PERF_RECORD_KSYMBOL event and "bpf event with" test for PERF_RECORD_BPF_EVENT event.

Removing it so it also goes along with other side-band events display.

Before:

  # perf script --show-bpf-events
  ...
  swapper     0 [000]     0.000000: PERF_RECORD_KSYMBOL ksymbol event with addr ffffffffc0ef971d len 229 type 1 flags 0x0 name bpf_prog_2a142ef67aaad174
  swapper     0 [000]     0.000000: PERF_RECORD_BPF_EVENT bpf event with type 1, flags 0, id 36

After:

  # perf script --show-bpf-events
  ...
  swapper     0 [000]     0.000000: PERF_RECORD_KSYMBOL addr ffffffffc0ef971d len 229 type 1 flags 0x0 name bpf_prog_2a142ef67aaad174
  swapper     0 [000]     0.000000: PERF_RECORD_BPF_EVENT type 1, flags 0, id 36

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/util/event.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

tools/perf/util/event.c +2 -2

index ba7be74fad6e..198920b7c56a 100644
@@ -1485,7 +1485,7 @@ static size_t perf_event__fprintf_lost(union perf_event *event, FILE *fp)
 
 size_t perf_event__fprintf_ksymbol(union perf_event *event, FILE *fp)
 {
-	return fprintf(fp, " ksymbol event with addr %" PRIx64 " len %u type %u flags 0x%x name %s\n",
+	return fprintf(fp, " addr %" PRIx64 " len %u type %u flags 0x%x name %s\n",
 		       event->ksymbol_event.addr, event->ksymbol_event.len,
 		       event->ksymbol_event.ksym_type,
 		       event->ksymbol_event.flags, event->ksymbol_event.name);

@@ -1493,7 +1493,7 @@ size_t perf_event__fprintf_ksymbol(union perf_event *event, FILE *fp)
 
 size_t perf_event__fprintf_bpf_event(union perf_event *event, FILE *fp)
 {
-	return fprintf(fp, " bpf event with type %u, flags %u, id %u\n",
+	return fprintf(fp, " type %u, flags %u, id %u\n",
 		       event->bpf_event.type, event->bpf_event.flags,
 		       event->bpf_event.id);
 }

Signature · 1 line

2.20.1

Jiri Olsa

author reviewer

2019-05-03 08:19 UTC #19 lore raw

PATCH v2 12/12 perf script: Add --show-all-events option

In reply to Jiri Olsa

Adding --show-all-events option to show all
side-bad events with single option, like:

  $ perf script --show-all-events
  swapper     0 [000]     0.000000: PERF_RECORD_MMAP -1/0: [0xffffffffa6000000(0xc00e41) @ 0xffffffffa6000000]: x [kernel.kallsyms]_text
  ...
  swapper     0 [000]     0.000000: PERF_RECORD_KSYMBOL addr ffffffffc01bc362 len 229 type 1 flags 0x0 name bpf_prog_7be49e3934a125ba
  swapper     0 [000]     0.000000: PERF_RECORD_BPF_EVENT type 1, flags 0, id 29
  ...
  swapper     0 [000]     0.000000: PERF_RECORD_FORK(1:1):(0:0)
  systemd     0 [000]     0.000000: PERF_RECORD_COMM: systemd:1/1
  ...
  swapper     0 [000] 63587.039518:          1 cycles:  ffffffffa60698b4 [unknown] ([kernel.kallsyms])
  swapper     0 [000] 63587.039522:          1 cycles:  ffffffffa60698b4 [unknown] ([kernel.kallsyms])

Link http://lkml.kernel.org/n/[email protected] open here

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/Documentation/perf-script.txt |  3 +++
 tools/perf/builtin-script.c              | 13 +++++++++++++
 2 files changed, 16 insertions(+)

tools/perf/Documentation/perf-script.txt +3 -0

index af8282782911..ddcd08bf0172 100644
@@ -316,6 +316,9 @@ OPTIONS
 --show-bpf-events
 	Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
 
+--show-all-events
+	Display all side-band events i.e. enable all --show-*-events options.
+
 --demangle::
 	Demangle symbol names to human readable form. It's enabled by default,
 	disable with --no-demangle.

tools/perf/builtin-script.c +13 -0

index 3a48a2627670..e7462dcab2c6 100644
@@ -1607,6 +1607,7 @@ struct perf_script {
 	bool			show_lost_events;
 	bool			show_round_events;
 	bool			show_bpf_events;
+	bool			show_all_events;
 	bool			allocated;
 	bool			per_event_dump;
 	struct cpu_map		*cpus;

@@ -2436,6 +2437,16 @@ static int __cmd_script(struct perf_script *script)
 
 	perf_stat__init_shadow_stats();
 
+	if (script->show_all_events) {
+		script->show_task_events	= true;
+		script->show_mmap_events	= true;
+		script->show_switch_events	= true;
+		script->show_namespace_events	= true;
+		script->show_lost_events	= true;
+		script->show_round_events	= true;
+		script->show_bpf_events		= true;
+	}
+
 	/* override event processing functions */
 	if (script->show_task_events) {
 		script->tool.comm = process_comm_event;

@@ -3481,6 +3492,8 @@ int cmd_script(int argc, const char **argv)
 		    "Show round events (if recorded)"),
 	OPT_BOOLEAN('\0', "show-bpf-events", &script.show_bpf_events,
 		    "Show bpf related events (if recorded)"),
+	OPT_BOOLEAN('\0', "show-all-events", &script.show_all_events,
+		    "Show all side-band events (if recorded)"),
 	OPT_BOOLEAN('\0', "per-event-dump", &script.per_event_dump,
 		    "Dump trace output to files named by the monitored events"),
 	OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),

Signature · 1 line

2.20.1

Song Liu

Acked-by on the v2 series

2019-05-06 21:46 UTC #20 lore raw

In reply to Jiri Olsa

Quoted 90 lines · Jiri Olsa

hi,
this patchset adds dso support to read and display bpf code in intel_pt trace output. I had to change some of the kernel maps processing code, so hopefully I did not break too many things ;-)

It's now possible to see bpf code flow via:

 # perf-with-kcore record pt -e intel_pt//ku -- sleep 1
 # perf-with-kcore script pt --insn-trace --xed
 ...
          sleep 36660 [016] 1057036.806464404:  ffffffff811dfba5 trace_call_bpf+0x65 ([kernel.kallsyms])               nopl  %eax, (%rax,%rax,1)
          sleep 36660 [016] 1057036.806464404:  ffffffff811dfbaa trace_call_bpf+0x6a ([kernel.kallsyms])               movq  0x30(%rbx), %rax
          sleep 36660 [016] 1057036.806464404:  ffffffff811dfbae trace_call_bpf+0x6e ([kernel.kallsyms])               leaq  0x38(%rbx), %rsi
          sleep 36660 [016] 1057036.806464404:  ffffffff811dfbb2 trace_call_bpf+0x72 ([kernel.kallsyms])               mov %r13, %rdi
          sleep 36660 [016] 1057036.806464404:  ffffffff811dfbb5 trace_call_bpf+0x75 ([kernel.kallsyms])               callq  0xffffffff81c00c30
          sleep 36660 [016] 1057036.806464404:  ffffffff81c00c30 __x86_indirect_thunk_rax+0x0 ([kernel.kallsyms])              callq  0xffffffff81c00c3c
          sleep 36660 [016] 1057036.806464404:  ffffffff81c00c3c __x86_indirect_thunk_rax+0xc ([kernel.kallsyms])              movq  %rax, (%rsp)
          sleep 36660 [016] 1057036.806464404:  ffffffff81c00c40 __x86_indirect_thunk_rax+0x10 ([kernel.kallsyms])             retq
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e89ef bpf_prog_da4fe6b3d2c29b25_trace_return+0x0 (bpf_prog_da4fe6b3d2c29b25_trace_return)           pushq  %rbp
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e89f0 bpf_prog_da4fe6b3d2c29b25_trace_return+0x1 (bpf_prog_da4fe6b3d2c29b25_trace_return)           mov %rsp, %rbp
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e89f3 bpf_prog_da4fe6b3d2c29b25_trace_return+0x4 (bpf_prog_da4fe6b3d2c29b25_trace_return)           sub $0x158, %rsp
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e89fa bpf_prog_da4fe6b3d2c29b25_trace_return+0xb (bpf_prog_da4fe6b3d2c29b25_trace_return)           sub $0x28, %rbp
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e89fe bpf_prog_da4fe6b3d2c29b25_trace_return+0xf (bpf_prog_da4fe6b3d2c29b25_trace_return)           movq  %rbx, (%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a02 bpf_prog_da4fe6b3d2c29b25_trace_return+0x13 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r13, 0x8(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a06 bpf_prog_da4fe6b3d2c29b25_trace_return+0x17 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r14, 0x10(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a0a bpf_prog_da4fe6b3d2c29b25_trace_return+0x1b (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %r15, 0x18(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a0e bpf_prog_da4fe6b3d2c29b25_trace_return+0x1f (bpf_prog_da4fe6b3d2c29b25_trace_return)          xor %eax, %eax
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a10 bpf_prog_da4fe6b3d2c29b25_trace_return+0x21 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rax, 0x20(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a14 bpf_prog_da4fe6b3d2c29b25_trace_return+0x25 (bpf_prog_da4fe6b3d2c29b25_trace_return)          mov %rdi, %rbx
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a17 bpf_prog_da4fe6b3d2c29b25_trace_return+0x28 (bpf_prog_da4fe6b3d2c29b25_trace_return)          callq  0xffffffff811fed50
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed50 bpf_get_current_pid_tgid+0x0 ([kernel.kallsyms])              nopl  %eax, (%rax,%rax,1)
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed55 bpf_get_current_pid_tgid+0x5 ([kernel.kallsyms])              movq  %gs:0x15c00, %rdx
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed5e bpf_get_current_pid_tgid+0xe ([kernel.kallsyms])              test %rdx, %rdx
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed61 bpf_get_current_pid_tgid+0x11 ([kernel.kallsyms])             jz 0xffffffff811fed79
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed63 bpf_get_current_pid_tgid+0x13 ([kernel.kallsyms])             movsxdl  0x494(%rdx), %rax
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed6a bpf_get_current_pid_tgid+0x1a ([kernel.kallsyms])             movsxdl  0x490(%rdx), %rdx
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed71 bpf_get_current_pid_tgid+0x21 ([kernel.kallsyms])             shl $0x20, %rax
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed75 bpf_get_current_pid_tgid+0x25 ([kernel.kallsyms])             or %rdx, %rax
          sleep 36660 [016] 1057036.806464725:  ffffffff811fed78 bpf_get_current_pid_tgid+0x28 ([kernel.kallsyms])             retq
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a1c bpf_prog_da4fe6b3d2c29b25_trace_return+0x2d (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rax, -0x8(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a20 bpf_prog_da4fe6b3d2c29b25_trace_return+0x31 (bpf_prog_da4fe6b3d2c29b25_trace_return)          xor %edi, %edi
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a22 bpf_prog_da4fe6b3d2c29b25_trace_return+0x33 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rdi, -0x10(%rbp)
          sleep 36660 [016] 1057036.806464725:  ffffffffa05e8a26 bpf_prog_da4fe6b3d2c29b25_trace_return+0x37 (bpf_prog_da4fe6b3d2c29b25_trace_return)          movq  %rdi, -0x18(%rbp)

 # perf-core/perf-with-kcore script pt --call-trace
 ...
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                kretprobe_perf_func
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                    trace_call_bpf
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                        __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464404: ([kernel.kallsyms]                       )                            __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_pid_tgid
          sleep 36660 [016] 1057036.806464725: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_ktime_get_ns
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                    __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806464725: ([kernel.kallsyms]                       )                                        __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806465045: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                __htab_map_lookup_elem
          sleep 36660 [016] 1057036.806465366: ([kernel.kallsyms]                       )                                    memcmp
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806465687: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_probe_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                    probe_kernel_read
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        __check_object_size
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                            check_stack_object
          sleep 36660 [016] 1057036.806465687: ([kernel.kallsyms]                       )                                        copy_user_enhanced_fast_string
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_get_current_uid_gid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kgid
          sleep 36660 [016] 1057036.806466008: ([kernel.kallsyms]                       )                                    from_kuid
          sleep 36660 [016] 1057036.806466008: (bpf_prog_da4fe6b3d2c29b25_trace_return  )                                bpf_perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                    perf_event_output
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                        perf_prepare_sample
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                            perf_misc_flags
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                __x86_indirect_thunk_rax
          sleep 36660 [016] 1057036.806466328: ([kernel.kallsyms]                       )                                                    __x86_indirect_thunk_rax

Quoted 6 lines

It's also available in here:
git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
perf/fixes
thanks,
jirka

Besides minor comments in a few patches, for the series:

Acked-by Song Liu <[email protected]>

Quoted 36 lines

Cc Jesper Dangaard Brouer <[email protected]>
Cc Alexei Starovoitov <[email protected]>
Cc Daniel Borkmann <[email protected]>
---
Jiri Olsa (12):
perf tools: Separate generic code in dso__data_file_size perf tools: Separate generic code in dso_cache__read perf tools: Simplify dso_cache__read function
perf tools: Add bpf dso read and size hooks
perf tools: Read also the end of the kernel
perf tools: Keep zero in pgoff bpf map
perf script: Pad dso name for --call-trace
perf tools: Preserve eBPF maps when loading kcore perf tests: Add map_groups__merge_in test
perf script: Add --show-bpf-events to show eBPF related events perf script: Remove superfluous bpf event titles perf script: Add --show-all-events option
tools/include/linux/kernel.h             |   1 +
tools/lib/vsprintf.c                     |  19 +++++++++++++++++++
tools/perf/Documentation/perf-script.txt |   6 ++++++
tools/perf/builtin-script.c              |  56 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
tools/perf/tests/Build                   |   1 +
tools/perf/tests/builtin-test.c          |   4 ++++
tools/perf/tests/map_groups.c            | 120 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
tools/perf/tests/tests.h                 |   1 +
tools/perf/util/dso.c                    | 125 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------------------------------
tools/perf/util/event.c                  |   4 ++--
tools/perf/util/machine.c                |  31 ++++++++++++++++++++-----------
tools/perf/util/map.c                    |   6 ++++++
tools/perf/util/map_groups.h             |   2 ++
tools/perf/util/symbol.c                 |  97 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++----
tools/perf/util/symbol_conf.h            |   1 +
15 files changed, 421 insertions(+), 53 deletions(-)
create mode 100644 tools/perf/tests/map_groups.c