Skip to contentSkip to conversation

perf tools: Fix not finding kcore in buildid cache

PATCH In review Adrian Hunter 4 participants 8 messages Nov 26, 2013 – Nov 30, 2013 perf events

lore
Share & tools
lore

Adrian Hunter

author reviewer
#0 lore raw

PATCH 1/1 perf tools: Fix not finding kcore in buildid cache

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

---
 tools/perf/util/symbol.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
tools/perf/util/symbol.c +7 -2
index ce9ce10..360eefe 100644
@@ -1526,14 +1526,15 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 
 	build_id__sprintf(dso->build_id, sizeof(dso->build_id), sbuild_id);
 
+	scnprintf(path, sizeof(path), "%s/[kernel.kcore]/%s", buildid_dir,
+		  sbuild_id);
+
 	/* Use /proc/kallsyms if possible */
 	if (is_host) {
 		DIR *d;
 		int fd;
 
 		/* If no cached kcore go with /proc/kallsyms */
-		scnprintf(path, sizeof(path), "%s/[kernel.kcore]/%s",
-			  buildid_dir, sbuild_id);
 		d = opendir(path);
 		if (!d)
 			goto proc_kallsyms;
@@ -1558,6 +1559,10 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 		goto proc_kallsyms;
 	}
 
+	/* Find kallsyms in build-id cache with kcore */
+	if (!find_matching_kcore(map, path, sizeof(path)))
+		return strdup(path);
+
 	scnprintf(path, sizeof(path), "%s/[kernel.kallsyms]/%s",
 		  buildid_dir, sbuild_id);
 
Signature · 1 line
1.7.11.7
#1 lore raw
Quoted 3 lines · Adrian Hunter

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

Thanks,

Ingo

author reviewer
#2 lore raw
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Adrian Hunter

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

The preconditions for the problem are that kcore is in the buildid cache and that the running kernel has a different buildid e.g.

$ perf buildid-list | grep kernel.kallsyms
b18fbe769dea12513028a4287dae02efc1582c29 [kernel.kallsyms]
$ perf record -o /tmp/junk true >/dev/null 2>&1 ; perf buildid-list -i /tmp/junk | grep kernel.kallsyms
504f9fd9328c40714995ce76fe1354c6e90136ce [kernel.kallsyms]

Before the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kallsyms]/b18fbe769dea12513028a4287dae02efc1582c29 for symbols

perf script stopped

After the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kcore for kernel object code
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kallsyms for symbols

perf script stopped

#3 lore raw
Quoted 38 lines · Adrian Hunter
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Adrian Hunter

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

The preconditions for the problem are that kcore is in the buildid cache and that the running kernel has a different buildid e.g.

$ perf buildid-list | grep kernel.kallsyms
b18fbe769dea12513028a4287dae02efc1582c29 [kernel.kallsyms]
$ perf record -o /tmp/junk true >/dev/null 2>&1 ; perf buildid-list -i /tmp/junk | grep kernel.kallsyms
504f9fd9328c40714995ce76fe1354c6e90136ce [kernel.kallsyms]

Before the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kallsyms]/b18fbe769dea12513028a4287dae02efc1582c29 for symbols

perf script stopped

After the patch:

Quoted 7 lines
$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kcore for kernel object code
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kallsyms for symbols

perf script stopped

So the result of the patch is that kernel symbols get resolved/annotated correctly, as expected from the live kcore feature, right?

Thanks,

Ingo

author reviewer
#4 lore raw
Quoted 47 lines · Ingo Molnar
Quoted 44 lines · Adrian Hunter
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Adrian Hunter

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

The preconditions for the problem are that kcore is in the buildid cache and that the running kernel has a different buildid e.g.

$ perf buildid-list | grep kernel.kallsyms
b18fbe769dea12513028a4287dae02efc1582c29 [kernel.kallsyms]
$ perf record -o /tmp/junk true >/dev/null 2>&1 ; perf buildid-list -i /tmp/junk | grep kernel.kallsyms
504f9fd9328c40714995ce76fe1354c6e90136ce [kernel.kallsyms]

Before the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kallsyms]/b18fbe769dea12513028a4287dae02efc1582c29 for symbols

perf script stopped

After the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kcore for kernel object code
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kallsyms for symbols
Quoted 5 lines
Quoted 1 line

perf script stopped

So the result of the patch is that kernel symbols get resolved/annotated correctly, as expected from the live kcore feature, right?

Yes but this is the non-live case i.e. the kernel has changed and we are looking at a special copy of kcore that we placed in the buildid cache (using "perf buildid-cache -v -k /proc/kcore") when the data was recorded.

#5 lore raw
Quoted 50 lines · Adrian Hunter
Quoted 49 lines · Ingo Molnar
Quoted 45 lines · Adrian Hunter
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Adrian Hunter

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

The preconditions for the problem are that kcore is in the buildid cache and that the running kernel has a different buildid e.g.

$ perf buildid-list | grep kernel.kallsyms
b18fbe769dea12513028a4287dae02efc1582c29 [kernel.kallsyms]
$ perf record -o /tmp/junk true >/dev/null 2>&1 ; perf buildid-list -i /tmp/junk | grep kernel.kallsyms
504f9fd9328c40714995ce76fe1354c6e90136ce [kernel.kallsyms]

Before the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kallsyms]/b18fbe769dea12513028a4287dae02efc1582c29 for symbols

perf script stopped

After the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kcore for kernel object code
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kallsyms for symbols

perf script stopped

Quoted 7 lines
Quoted 3 lines

So the result of the patch is that kernel symbols get resolved/annotated correctly, as expected from the live kcore feature, right?

Yes but this is the non-live case i.e. the kernel has changed and we are looking at a special copy of kcore that we placed in the buildid cache (using "perf buildid-cache -v -k /proc/kcore") when the data was recorded.

Okay.

Thanks,

Ingo

#6 lore raw

Em Wed, Nov 27, 2013 at 12:52:47PM +0100, Ingo Molnar escreveu:

Quoted 57 lines · Adrian Hunter
Quoted 55 lines · Ingo Molnar
Quoted 52 lines · Adrian Hunter
Quoted 39 lines · Ingo Molnar
Quoted 11 lines · Adrian Hunter
Quoted 4 lines

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

It would be nice to also include the specific failure mode in the
changelog: exactly how can users run into this. That will be more useful to most people reading these changelogs than a more abstract bug description.

The preconditions for the problem are that kcore is in the buildid cache and that the running kernel has a different buildid e.g.

$ perf buildid-list | grep kernel.kallsyms
b18fbe769dea12513028a4287dae02efc1582c29 [kernel.kallsyms]
$ perf record -o /tmp/junk true >/dev/null 2>&1 ; perf buildid-list -i /tmp/junk | grep kernel.kallsyms
504f9fd9328c40714995ce76fe1354c6e90136ce [kernel.kallsyms]

Before the patch:

$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kallsyms]/b18fbe769dea12513028a4287dae02efc1582c29 for symbols

perf script stopped

After the patch:

Quoted 7 lines
$ perf script -v >out.txt
build id event received for [kernel.kallsyms]: b18fbe769dea12513028a4287dae02efc1582c29
<SNIP>
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kcore for kernel object code
Using /home/ahunter/.debug/[kernel.kcore]/b18fbe769dea12513028a4287dae02efc1582c29/2013112617084061/kallsyms for symbols

perf script stopped

So the result of the patch is that kernel symbols get resolved/annotated correctly, as expected from the live kcore feature, right?

Quoted 5 lines
Quoted 3 lines

Yes but this is the non-live case i.e. the kernel has changed and we are looking at a special copy of kcore that we placed in the buildid cache (using "perf buildid-cache -v -k /proc/kcore") when the data was recorded.

Okay.

Thanks for the questions and answers, I'll edit it and add to the patch.

- Arnaldo

#7 lore raw

[tip:perf/core] perf symbols: Fix not finding kcore in buildid cache

In reply to Adrian Hunter
perf symbols: Fix not finding kcore in buildid cache
Commit-ID:  449867e346bfd52c5df6bba5b706a795c35e78d4
Gitweb:     http://git.kernel.org/tip/449867e346bfd52c5df6bba5b706a795c35e78d4
Author:     Adrian Hunter <[email protected]>
AuthorDate: Tue, 26 Nov 2013 15:19:24 +0200
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Wed, 27 Nov 2013 14:58:38 -0300

perf symbols: Fix not finding kcore in buildid cache

The logic was not looking in the buildid cache for kcore if the host kernel buildid did not match the recorded kernel buildid.

This affects the non-live case i.e. the kernel has changed and we are looking at a special copy of kcore that we placed in the buildid cache (using "perf buildid-cache -v -k /proc/kcore") when the data was recorded.

After this fix kernel symbols get resolved/annotated correctly.

Cc David Ahern <[email protected]>
Cc Frederic Weisbecker <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mike Galbraith <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
[ Added further explanation extracted from conversation between Ingo & Adrian on lkml ]
---
 tools/perf/util/symbol.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
tools/perf/util/symbol.c +7 -2
index ce9ce10..360eefe 100644
@@ -1526,14 +1526,15 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 
 	build_id__sprintf(dso->build_id, sizeof(dso->build_id), sbuild_id);
 
+	scnprintf(path, sizeof(path), "%s/[kernel.kcore]/%s", buildid_dir,
+		  sbuild_id);
+
 	/* Use /proc/kallsyms if possible */
 	if (is_host) {
 		DIR *d;
 		int fd;
 
 		/* If no cached kcore go with /proc/kallsyms */
-		scnprintf(path, sizeof(path), "%s/[kernel.kcore]/%s",
-			  buildid_dir, sbuild_id);
 		d = opendir(path);
 		if (!d)
 			goto proc_kallsyms;
@@ -1558,6 +1559,10 @@ static char *dso__find_kallsyms(struct dso *dso, struct map *map)
 		goto proc_kallsyms;
 	}
 
+	/* Find kallsyms in build-id cache with kcore */
+	if (!find_matching_kcore(map, path, sizeof(path)))
+		return strdup(path);
+
 	scnprintf(path, sizeof(path), "%s/[kernel.kallsyms]/%s",
 		  buildid_dir, sbuild_id);