Skip to contentSkip to conversation

csky: perf callchain dwarf support

PATCH Superseded Mao Han 3 participants 10 messages Apr 10, 2019 – Apr 16, 2019

Subsystems (2)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Mao Han

author
#0 lore raw

v3 cover csky: perf callchain dwarf support

This patch set add perf DWARF unwinding support for C-SKY. Including user registers/stack dump API, and libdw support.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
CC Guo Ren <[email protected]>

Changes since v2:
- use same registers name as struct pt_regs
- code convention

Changes since v1:
- seperate the callchain support using frame pointer

Mao Han (3):
perf: use hweight64 instead of hweight_long
csky: Add support for perf registers sampling
csky: add support for libdw

 arch/csky/Kconfig                            |   2 +
 arch/csky/include/uapi/asm/perf_regs.h       |  51 ++++++++++++++
 arch/csky/kernel/Makefile                    |   1 +
 arch/csky/kernel/perf_regs.c                 |  40 +++++++++++
 tools/arch/csky/include/uapi/asm/perf_regs.h |  51 ++++++++++++++
 tools/perf/Makefile.config                   |   6 +-
 tools/perf/arch/csky/Build                   |   1 +
 tools/perf/arch/csky/Makefile                |   3 +
 tools/perf/arch/csky/include/perf_regs.h     | 100 +++++++++++++++++++++++++++
 tools/perf/arch/csky/util/Build              |   2 +
 tools/perf/arch/csky/util/dwarf-regs.c       |  49 +++++++++++++
 tools/perf/arch/csky/util/unwind-libdw.c     |  78 +++++++++++++++++++++
 tools/perf/util/evsel.c                      |   2 +-
 13 files changed, 384 insertions(+), 2 deletions(-)
 create mode 100644 arch/csky/include/uapi/asm/perf_regs.h
 create mode 100644 arch/csky/kernel/perf_regs.c
 create mode 100644 tools/arch/csky/include/uapi/asm/perf_regs.h
 create mode 100644 tools/perf/arch/csky/Build
 create mode 100644 tools/perf/arch/csky/Makefile
 create mode 100644 tools/perf/arch/csky/include/perf_regs.h
 create mode 100644 tools/perf/arch/csky/util/Build
 create mode 100644 tools/perf/arch/csky/util/dwarf-regs.c
 create mode 100644 tools/perf/arch/csky/util/unwind-libdw.c
Signature · 1 line
2.7.4
author
#1 lore raw

PATCH v3 1/3 perf: use hweight64 instead of hweight_long

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
---
 tools/perf/util/evsel.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evsel.c +1 -1
index 7835e05..73c78be 100644
@@ -2322,7 +2322,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;
Signature · 1 line
2.7.4
#2 lore raw

Re: [PATCH v3 1/3] perf: use hweight64 instead of hweight_long

Em Wed, Apr 10, 2019 at 04:16:43PM +0800, Mao Han escreveu:

Quoted 23 lines

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
---
 tools/perf/util/evsel.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evsel.c +0 -0
index 7835e05..73c78be 100644

tools/perf/util/evsel.c int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event, · line 2328

Quoted 8 lines
tools/perf/util/evsel.c +1 -1
@@ -2322,7 +2322,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;

Later on, in the same function, perf_evsel__parse_sample() we have:

        data->intr_regs.abi = PERF_SAMPLE_REGS_ABI_NONE;
        if (type & PERF_SAMPLE_REGS_INTR) {
                OVERFLOW_CHECK_u64(array);
                data->intr_regs.abi = *array;
                array++;

                if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
                        u64 mask = evsel->attr.sample_regs_intr;

                        sz = hweight_long(mask) * sizeof(u64);
                        OVERFLOW_CHECK(array, sz, max_size);
                        data->intr_regs.mask = mask;
                        data->intr_regs.regs = (u64 *)array;
                        array = (void *)array + sz;
                }
        }

You forgot to convert that one, doing it for you,

Thanks,

- Arnaldo

#3 lore raw

Re: [PATCH v3 1/3] perf: use hweight64 instead of hweight_long

Em Wed, Apr 10, 2019 at 10:08:41AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 52 lines

Em Wed, Apr 10, 2019 at 04:16:43PM +0800, Mao Han escreveu:

Quoted 23 lines

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
---
 tools/perf/util/evsel.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evsel.c +0 -0
index 7835e05..73c78be 100644

tools/perf/util/evsel.c int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event, · line 2328

Quoted 8 lines
tools/perf/util/evsel.c +1 -1
@@ -2322,7 +2322,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;

Later on, in the same function, perf_evsel__parse_sample() we have:

        data->intr_regs.abi = PERF_SAMPLE_REGS_ABI_NONE;
        if (type & PERF_SAMPLE_REGS_INTR) {
                OVERFLOW_CHECK_u64(array);
                data->intr_regs.abi = *array;
                array++;

                if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
                        u64 mask = evsel->attr.sample_regs_intr;

                        sz = hweight_long(mask) * sizeof(u64);
                        OVERFLOW_CHECK(array, sz, max_size);
                        data->intr_regs.mask = mask;
                        data->intr_regs.regs = (u64 *)array;
                        array = (void *)array + sz;
                }
        }
Quoted 1 line

You forgot to convert that one, doing it for you,

Also in perf_event__sample_event_size() we need to do the same thing, right?

- Arnaldo

#4 lore raw

Re: [PATCH v3 1/3] perf: use hweight64 instead of hweight_long

Em Wed, Apr 10, 2019 at 10:10:42AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 53 lines

Em Wed, Apr 10, 2019 at 10:08:41AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 52 lines

Em Wed, Apr 10, 2019 at 04:16:43PM +0800, Mao Han escreveu:

Quoted 23 lines

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
---
 tools/perf/util/evsel.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evsel.c +0 -0
index 7835e05..73c78be 100644

tools/perf/util/evsel.c int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event, · line 2328

Quoted 8 lines
tools/perf/util/evsel.c +1 -1
@@ -2322,7 +2322,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;

Later on, in the same function, perf_evsel__parse_sample() we have:

        data->intr_regs.abi = PERF_SAMPLE_REGS_ABI_NONE;
        if (type & PERF_SAMPLE_REGS_INTR) {
                OVERFLOW_CHECK_u64(array);
                data->intr_regs.abi = *array;
                array++;

                if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
                        u64 mask = evsel->attr.sample_regs_intr;

                        sz = hweight_long(mask) * sizeof(u64);
                        OVERFLOW_CHECK(array, sz, max_size);
                        data->intr_regs.mask = mask;
                        data->intr_regs.regs = (u64 *)array;
                        array = (void *)array + sz;
                }
        }
Quoted 4 lines
Quoted 1 line

You forgot to convert that one, doing it for you,

Also in perf_event__sample_event_size() we need to do the same thing, right?

and perf_event__synthesize_sample()

Done, resulting patch is at the end of this messages, and matches the kernel, that uses only hweight64().

I've also added Fixes tags to the patches that used hweight_long() in various places, to help with the stable trees backporting process, please consider doing it next time.

- Arnaldo

commit 21e6dfe04861c2c1b529f2759850bc62a80ca050

Author: Mao Han <[email protected]>
Date:   Wed Apr 10 16:16:43 2019 +0800

perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Fixes 6a21c0b5c2ab ("perf tools: Add core support for sampling intr machine state regs")
Fixes d03f2170546d ("perf tools: Expand perf_event__synthesize_sample()")
Fixes 0f6a30150ca2 ("perf tools: Support user regs and stack in sample parsing")
tools/perf/util/evsel.c +6 -6
index 66d066f18b5b..966360844fff 100644
@@ -2368,7 +2368,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;
@@ -2424,7 +2424,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
 			u64 mask = evsel->attr.sample_regs_intr;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->intr_regs.mask = mask;
 			data->intr_regs.regs = (u64 *)array;
@@ -2552,7 +2552,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2580,7 +2580,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2710,7 +2710,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			*array++ = sample->user_regs.abi;
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			memcpy(array, sample->user_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
@@ -2746,7 +2746,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			*array++ = sample->intr_regs.abi;
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			memcpy(array, sample->intr_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
author
#5 lore raw

Re: [PATCH v3 1/3] perf: use hweight64 instead of hweight_long

Quoted 7 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Quoted 1 line

You forgot to convert that one, doing it for you,

Also in perf_event__sample_event_size() we need to do the same thing, right?

and perf_event__synthesize_sample()

Quoted 9 lines

Done, resulting patch is at the end of this messages, and matches the kernel, that uses only hweight64().

I've also added Fixes tags to the patches that used hweight_long() in various places, to help with the stable trees backporting process, please consider doing it next time.

- Arnaldo

Thanks for help improving the patch and the suggestion. Tested with the new patch on C-SKY, seems work fine.

Mao Han

#6 lore raw

[tip:perf/urgent] perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)

In reply to Mao Han
perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)
Commit-ID:  21e6dfe04861c2c1b529f2759850bc62a80ca050
Gitweb:     https://git.kernel.org/tip/21e6dfe04861c2c1b529f2759850bc62a80ca050
Author:     Mao Han <[email protected]>
AuthorDate: Wed, 10 Apr 2019 16:16:43 +0800
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Wed, 10 Apr 2019 10:25:28 -0300

perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Fixes 6a21c0b5c2ab ("perf tools: Add core support for sampling intr machine state regs")
Fixes d03f2170546d ("perf tools: Expand perf_event__synthesize_sample()")
Fixes 0f6a30150ca2 ("perf tools: Support user regs and stack in sample parsing")
---
 tools/perf/util/evsel.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
tools/perf/util/evsel.c +6 -6
index 66d066f18b5b..966360844fff 100644
@@ -2368,7 +2368,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;
@@ -2424,7 +2424,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
 			u64 mask = evsel->attr.sample_regs_intr;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->intr_regs.mask = mask;
 			data->intr_regs.regs = (u64 *)array;
@@ -2552,7 +2552,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2580,7 +2580,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2710,7 +2710,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			*array++ = sample->user_regs.abi;
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			memcpy(array, sample->user_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
@@ -2746,7 +2746,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			*array++ = sample->intr_regs.abi;
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			memcpy(array, sample->intr_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
#7 lore raw

[tip:perf/urgent] perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)

In reply to Mao Han
perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)
Commit-ID:  3a5b64f05d7fe36dea0dde26423e3044fbacd482
Gitweb:     https://git.kernel.org/tip/3a5b64f05d7fe36dea0dde26423e3044fbacd482
Author:     Mao Han <[email protected]>
AuthorDate: Wed, 10 Apr 2019 16:16:43 +0800
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 16 Apr 2019 11:27:53 -0300

perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)

On 32-bits platform with more than 32 registers, the 64 bits mask is truncate to the lower 32 bits and the return value of hweight_long will always smaller than 32. When kernel outputs more than 32 registers, but the user perf program only counts 32, there will be a data mismatch result to overflow check fail.

Cc Adrian Hunter <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
Fixes 6a21c0b5c2ab ("perf tools: Add core support for sampling intr machine state regs")
Fixes d03f2170546d ("perf tools: Expand perf_event__synthesize_sample()")
Fixes 0f6a30150ca2 ("perf tools: Support user regs and stack in sample parsing")
---
 tools/perf/util/evsel.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
tools/perf/util/evsel.c +6 -6
index 66d066f18b5b..966360844fff 100644
@@ -2368,7 +2368,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->user_regs.abi) {
 			u64 mask = evsel->attr.sample_regs_user;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->user_regs.mask = mask;
 			data->user_regs.regs = (u64 *)array;
@@ -2424,7 +2424,7 @@ int perf_evsel__parse_sample(struct perf_evsel *evsel, union perf_event *event,
 		if (data->intr_regs.abi != PERF_SAMPLE_REGS_ABI_NONE) {
 			u64 mask = evsel->attr.sample_regs_intr;
 
-			sz = hweight_long(mask) * sizeof(u64);
+			sz = hweight64(mask) * sizeof(u64);
 			OVERFLOW_CHECK(array, sz, max_size);
 			data->intr_regs.mask = mask;
 			data->intr_regs.regs = (u64 *)array;
@@ -2552,7 +2552,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2580,7 +2580,7 @@ size_t perf_event__sample_event_size(const struct perf_sample *sample, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			result += sizeof(u64);
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			result += sz;
 		} else {
 			result += sizeof(u64);
@@ -2710,7 +2710,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_USER) {
 		if (sample->user_regs.abi) {
 			*array++ = sample->user_regs.abi;
-			sz = hweight_long(sample->user_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->user_regs.mask) * sizeof(u64);
 			memcpy(array, sample->user_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
@@ -2746,7 +2746,7 @@ int perf_event__synthesize_sample(union perf_event *event, u64 type,
 	if (type & PERF_SAMPLE_REGS_INTR) {
 		if (sample->intr_regs.abi) {
 			*array++ = sample->intr_regs.abi;
-			sz = hweight_long(sample->intr_regs.mask) * sizeof(u64);
+			sz = hweight64(sample->intr_regs.mask) * sizeof(u64);
 			memcpy(array, sample->intr_regs.regs, sz);
 			array = (void *)array + sz;
 		} else {
author
#8 lore raw

PATCH v3 2/3 csky: Add support for perf registers sampling

In reply to Mao Han

This patch implements the perf registers sampling and validation API for csky arch. The valid registers and their register ID are defined in perf_regs.h. Perf tool can backtrace in userspace with unwind library and the registers/user stack dump support.

CC Guo Ren <[email protected]>
---
 arch/csky/Kconfig                      |  2 ++
 arch/csky/include/uapi/asm/perf_regs.h | 51 ++++++++++++++++++++++++++++++++++
 arch/csky/kernel/Makefile              |  1 +
 arch/csky/kernel/perf_regs.c           | 40 ++++++++++++++++++++++++++
 4 files changed, 94 insertions(+)
 create mode 100644 arch/csky/include/uapi/asm/perf_regs.h
 create mode 100644 arch/csky/kernel/perf_regs.c
arch/csky/Kconfig +2 -0
index c4974cf..8e45c7a 100644
@@ -38,6 +38,8 @@ config CSKY
 	select HAVE_KERNEL_LZO
 	select HAVE_KERNEL_LZMA
 	select HAVE_PERF_EVENTS
+	select HAVE_PERF_REGS
+	select HAVE_PERF_USER_STACK_DUMP
 	select HAVE_DMA_API_DEBUG
 	select HAVE_DMA_CONTIGUOUS
 	select HAVE_SYSCALL_TRACEPOINTS
arch/csky/include/uapi/asm/perf_regs.h new file +51 -0

51 lines not on this page · raw message

arch/csky/kernel/Makefile +1 -0
index 4c462f5..1624b04 100644
@@ -10,6 +10,7 @@ obj-$(CONFIG_FUNCTION_TRACER)		+= ftrace.o
 obj-$(CONFIG_STACKTRACE)		+= stacktrace.o
 obj-$(CONFIG_CSKY_PMU_V1)		+= perf_event.o
 obj-$(CONFIG_PERF_EVENTS)		+= perf_callchain.o
+obj-$(CONFIG_HAVE_PERF_REGS)            += perf_regs.o
 
 ifdef CONFIG_FUNCTION_TRACER
 CFLAGS_REMOVE_ftrace.o = $(CC_FLAGS_FTRACE)
arch/csky/kernel/perf_regs.c new file +40 -0
new file mode 100644
index 0000000..88f1875
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (C) 2019 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/perf_event.h>
+#include <linux/bug.h>
+#include <asm/perf_regs.h>
+#include <asm/ptrace.h>
+
+u64 perf_reg_value(struct pt_regs *regs, int idx)
+{
+	if (WARN_ON_ONCE((u32)idx >= PERF_REG_CSKY_MAX))
+		return 0;
+
+	return ((long *)regs)[idx];
+}
+
+#define REG_RESERVED (~((1ULL << PERF_REG_CSKY_MAX) - 1))
+
+int perf_reg_validate(u64 mask)
+{
+	if (!mask || mask & REG_RESERVED)
+		return -EINVAL;
+
+	return 0;
+}
+
+u64 perf_reg_abi(struct task_struct *task)
+{
+	return PERF_SAMPLE_REGS_ABI_32;
+}
+
+void perf_get_regs_user(struct perf_regs *regs_user,
+			struct pt_regs *regs,
+			struct pt_regs *regs_user_copy)
+{
+	regs_user->regs = task_pt_regs(current);
+	regs_user->abi = perf_reg_abi(current);
+}
Signature · 1 line
2.7.4
author
#9 lore raw

PATCH v3 3/3 csky: add support for libdw

In reply to Mao Han

This patch add support for DWARF register mappings and libdw registers initialization, which is used by perf callchain analyzing when --call-graph=dwarf is given.

CC Peter Zijlstra <[email protected]>
CC Ingo Molnar <[email protected]>
CC Arnaldo Carvalho de Melo <[email protected]>
CC Alexander Shishkin <[email protected]>
CC Jiri Olsa <[email protected]>
CC Namhyung Kim <[email protected]>
---
 tools/arch/csky/include/uapi/asm/perf_regs.h |  51 ++++++++++++++
 tools/perf/Makefile.config                   |   6 +-
 tools/perf/arch/csky/Build                   |   1 +
 tools/perf/arch/csky/Makefile                |   3 +
 tools/perf/arch/csky/include/perf_regs.h     | 100 +++++++++++++++++++++++++++
 tools/perf/arch/csky/util/Build              |   2 +
 tools/perf/arch/csky/util/dwarf-regs.c       |  49 +++++++++++++
 tools/perf/arch/csky/util/unwind-libdw.c     |  78 +++++++++++++++++++++
 8 files changed, 289 insertions(+), 1 deletion(-)
 create mode 100644 tools/arch/csky/include/uapi/asm/perf_regs.h
 create mode 100644 tools/perf/arch/csky/Build
 create mode 100644 tools/perf/arch/csky/Makefile
 create mode 100644 tools/perf/arch/csky/include/perf_regs.h
 create mode 100644 tools/perf/arch/csky/util/Build
 create mode 100644 tools/perf/arch/csky/util/dwarf-regs.c
 create mode 100644 tools/perf/arch/csky/util/unwind-libdw.c
tools/arch/csky/include/uapi/asm/perf_regs.h new file +51 -0
new file mode 100644
index 0000000..ee323d8
@@ -0,0 +1,51 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2019 Hangzhou C-SKY Microsystems co.,ltd.
+
+#ifndef _ASM_CSKY_PERF_REGS_H
+#define _ASM_CSKY_PERF_REGS_H
+
+/* Index of struct pt_regs */
+enum perf_event_csky_regs {
+	PERF_REG_CSKY_TLS,
+	PERF_REG_CSKY_LR,
+	PERF_REG_CSKY_PC,
+	PERF_REG_CSKY_SR,
+	PERF_REG_CSKY_SP,
+	PERF_REG_CSKY_ORIG_A0,
+	PERF_REG_CSKY_A0,
+	PERF_REG_CSKY_A1,
+	PERF_REG_CSKY_A2,
+	PERF_REG_CSKY_A3,
+	PERF_REG_CSKY_REGS0,
+	PERF_REG_CSKY_REGS1,
+	PERF_REG_CSKY_REGS2,
+	PERF_REG_CSKY_REGS3,
+	PERF_REG_CSKY_REGS4,
+	PERF_REG_CSKY_REGS5,
+	PERF_REG_CSKY_REGS6,
+	PERF_REG_CSKY_REGS7,
+	PERF_REG_CSKY_REGS8,
+	PERF_REG_CSKY_REGS9,
+#if defined(__CSKYABIV2__)
+	PERF_REG_CSKY_EXREGS0,
+	PERF_REG_CSKY_EXREGS1,
+	PERF_REG_CSKY_EXREGS2,
+	PERF_REG_CSKY_EXREGS3,
+	PERF_REG_CSKY_EXREGS4,
+	PERF_REG_CSKY_EXREGS5,
+	PERF_REG_CSKY_EXREGS6,
+	PERF_REG_CSKY_EXREGS7,
+	PERF_REG_CSKY_EXREGS8,
+	PERF_REG_CSKY_EXREGS9,
+	PERF_REG_CSKY_EXREGS10,
+	PERF_REG_CSKY_EXREGS11,
+	PERF_REG_CSKY_EXREGS12,
+	PERF_REG_CSKY_EXREGS13,
+	PERF_REG_CSKY_EXREGS14,
+	PERF_REG_CSKY_HI,
+	PERF_REG_CSKY_LO,
+	PERF_REG_CSKY_DCSR,
+#endif
+	PERF_REG_CSKY_MAX,
+};
+#endif /* _ASM_CSKY_PERF_REGS_H */
tools/perf/Makefile.config +5 -1
index fe3f97e..42985ae 100644
@@ -59,6 +59,10 @@ ifeq ($(SRCARCH),arm64)
   LIBUNWIND_LIBS = -lunwind -lunwind-aarch64
 endif
 
+ifeq ($(SRCARCH),csky)
+  NO_PERF_REGS := 0
+endif
+
 ifeq ($(ARCH),s390)
   NO_PERF_REGS := 0
   NO_SYSCALL_TABLE := 0
@@ -77,7 +81,7 @@ endif
 # Disable it on all other architectures in case libdw unwind
 # support is detected in system. Add supported architectures
 # to the check.
-ifneq ($(SRCARCH),$(filter $(SRCARCH),x86 arm arm64 powerpc s390))
+ifneq ($(SRCARCH),$(filter $(SRCARCH),x86 arm arm64 powerpc s390 csky))
   NO_LIBDW_DWARF_UNWIND := 1
 endif
 
tools/perf/arch/csky/Build new file +1 -0
new file mode 100644
index 0000000..e4e5f33
@@ -0,0 +1 @@
+perf-y += util/
tools/perf/arch/csky/Makefile new file +3 -0
new file mode 100644
index 0000000..7fbca17
@@ -0,0 +1,3 @@
+ifndef NO_DWARF
+PERF_HAVE_DWARF_REGS := 1
+endif
tools/perf/arch/csky/include/perf_regs.h new file +100 -0

100 lines not on this page · raw message

tools/perf/arch/csky/util/Build new file +2 -0
new file mode 100644
index 0000000..1160bb2
@@ -0,0 +1,2 @@
+perf-$(CONFIG_DWARF) += dwarf-regs.o
+perf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
tools/perf/arch/csky/util/dwarf-regs.c new file +49 -0

49 lines not on this page · raw message

tools/perf/arch/csky/util/unwind-libdw.c new file +78 -0

78 lines not on this page · raw message

Signature · 1 line
2.7.4