Skip to contentSkip to conversation

perf tools: Allow build-id with trailing zeros

PATCH Superseded Namhyung Kim 2 participants 3 messages Sep 10, 2021 perf events

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

PATCH 1/1 perf tools: Allow build-id with trailing zeros

Current perf saves a build-id with size but old versions assumes the size of 20. In case the build-id is less than 20 (like for MD5), it'd fill the rest with 0s.

I saw a problem when old version of perf record saved binary in the build-id cache and new version of perf reads the data. The symbols should be read from the build-id cache (as the path no longer has the same binary) but it failed due to mismatch in the build-id.

symsrc__init: build id mismatch for /home/namhyung/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf.

The build-id event in the data has 20 byte build-ids, but it saw a different size (16) when it reads the build-id of the elf file in the build-id cache.

  $ readelf -n ~/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf

Displaying notes found in: .note.gnu.build-id

    Owner                Data size 	Description
    GNU                  0x00000010	NT_GNU_BUILD_ID (unique build ID bitstring)
      Build ID: 53e4c2f42a4c61a2d632d92a72afa08f

Let's fix this by allowing trailing zeros if the size is different.

Fixes 39be8d0115b3 ("perf tools: Pass build_id object to dso__build_id_equal()")
---
 tools/perf/util/dso.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
tools/perf/util/dso.c +12 -0
index ee15db2be2f4..0c0dd877d4e9 100644
@@ -1349,6 +1349,18 @@ void dso__set_build_id(struct dso *dso, struct build_id *bid)
 
 bool dso__build_id_equal(const struct dso *dso, struct build_id *bid)
 {
+	static const char zeros[BUILD_ID_SIZE];
+
+	if (dso->bid.size > bid->size && dso->bid.size == BUILD_ID_SIZE) {
+		/*
+		 * For the backward compatibility, it allows a build-id has
+		 * trailing zeros.
+		 */
+		return !memcmp(dso->bid.data, bid->data, bid->size) &&
+			!memcmp(&dso->bid.data[bid->size], zeros,
+				dso->bid.size - bid->size);
+	}
+
 	return dso->bid.size == bid->size &&
 	       memcmp(dso->bid.data, bid->data, dso->bid.size) == 0;
 }
Signature · 1 line
2.33.0.309.g3052b89438-goog
Acked-by on v1 1/1
#1 lore raw
Quoted 39 lines · Namhyung Kim

Current perf saves a build-id with size but old versions assumes the size of 20. In case the build-id is less than 20 (like for MD5), it'd fill the rest with 0s.

I saw a problem when old version of perf record saved binary in the build-id cache and new version of perf reads the data. The symbols should be read from the build-id cache (as the path no longer has the same binary) but it failed due to mismatch in the build-id.

symsrc__init: build id mismatch for /home/namhyung/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf.

The build-id event in the data has 20 byte build-ids, but it saw a different size (16) when it reads the build-id of the elf file in the build-id cache.

  $ readelf -n ~/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf

Displaying notes found in: .note.gnu.build-id

    Owner                Data size 	Description
    GNU                  0x00000010	NT_GNU_BUILD_ID (unique build ID bitstring)
      Build ID: 53e4c2f42a4c61a2d632d92a72afa08f

Let's fix this by allowing trailing zeros if the size is different.

Fixes 39be8d0115b3 ("perf tools: Pass build_id object to dso__build_id_equal()")
---
 tools/perf/util/dso.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
tools/perf/util/dso.c +0 -0
index ee15db2be2f4..0c0dd877d4e9 100644
@@ -1349,6 +1349,18 @@ void dso__set_build_id(struct dso *dso, struct build_id *bid)
 
 bool dso__build_id_equal(const struct dso *dso, struct build_id *bid)
 {

tools/perf/util/dso.c void dso__set_build_id(struct dso *dso, struct build_id *bid) · line 1361

Quoted 10 lines
tools/perf/util/dso.c +10 -0
+	static const char zeros[BUILD_ID_SIZE];
+
+	if (dso->bid.size > bid->size && dso->bid.size == BUILD_ID_SIZE) {
+		/*
+		 * For the backward compatibility, it allows a build-id has
+		 * trailing zeros.
+		 */
+		return !memcmp(dso->bid.data, bid->data, bid->size) &&
+			!memcmp(&dso->bid.data[bid->size], zeros,
+				dso->bid.size - bid->size);

we now have memchr_inv in tools, so you could use:
memchr_inv(&dso->bid.data[bid->size], 0, dso->bid.size - bid->size);

and save 20 bytes in bss ;-) other than that, nice catch

thanks,
jirka

Quoted 8 lines
tools/perf/util/dso.c +2 -0
+	}
+
 	return dso->bid.size == bid->size &&
 	       memcmp(dso->bid.data, bid->data, dso->bid.size) == 0;
 }
Signature · 1 line
2.33.0.309.g3052b89438-goog
author maintainer
#2 lore raw

Hi Jiri,

Quoted 55 lines · Jiri Olsa
Quoted 39 lines · Namhyung Kim

Current perf saves a build-id with size but old versions assumes the size of 20. In case the build-id is less than 20 (like for MD5), it'd fill the rest with 0s.

I saw a problem when old version of perf record saved binary in the build-id cache and new version of perf reads the data. The symbols should be read from the build-id cache (as the path no longer has the same binary) but it failed due to mismatch in the build-id.

symsrc__init: build id mismatch for /home/namhyung/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf.

The build-id event in the data has 20 byte build-ids, but it saw a different size (16) when it reads the build-id of the elf file in the build-id cache.

  $ readelf -n ~/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf

Displaying notes found in: .note.gnu.build-id

    Owner                Data size    Description
    GNU                  0x00000010   NT_GNU_BUILD_ID (unique build ID bitstring)
      Build ID: 53e4c2f42a4c61a2d632d92a72afa08f

Let's fix this by allowing trailing zeros if the size is different.

Fixes 39be8d0115b3 ("perf tools: Pass build_id object to dso__build_id_equal()")
---
 tools/perf/util/dso.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
tools/perf/util/dso.c +0 -0
index ee15db2be2f4..0c0dd877d4e9 100644
@@ -1349,6 +1349,18 @@ void dso__set_build_id(struct dso *dso, struct build_id *bid)
 
 bool dso__build_id_equal(const struct dso *dso, struct build_id *bid)
 {

tools/perf/util/dso.c void dso__set_build_id(struct dso *dso, struct build_id *bid) · line 1361

Quoted 10 lines
tools/perf/util/dso.c +10 -0
+     static const char zeros[BUILD_ID_SIZE];
+
+     if (dso->bid.size > bid->size && dso->bid.size == BUILD_ID_SIZE) {
+             /*
+              * For the backward compatibility, it allows a build-id has
+              * trailing zeros.
+              */
+             return !memcmp(dso->bid.data, bid->data, bid->size) &&
+                     !memcmp(&dso->bid.data[bid->size], zeros,
+                             dso->bid.size - bid->size);

we now have memchr_inv in tools, so you could use:
memchr_inv(&dso->bid.data[bid->size], 0, dso->bid.size - bid->size);

Quoted 3 lines

and save 20 bytes in bss ;-) other than that, nice catch

right, will update!

Thanks,
Namhyung