perf: Set build-id using build-id header on new mmap records
cover perf: Set build-id using build-id header on new mmap records
Hi,
We are seeing an issue with doing Coresight decode off target where initially the correct dso from ~/.debug is used, but after a new thread in the perf.data file is passed with its mmap record, the local version of the dso is picked up instead. This happens if the binary exists in the same path on both devices, for example /bin/ls.
Initially when parsing the build-ids in the header, the dso for /bin/ls will be created, and the file will correctly point to ~/.debug/bin/ls/2f15ad836be3339dec0e2e6a3c637e08e48aacbd/elf, but for any new threads or mmaps that are also for /bin/ls, they will not have a build-id set so they point to /bin/ls on the local machine rather than the debug folder.
To fix this I made it possible to look up which existing dsos have build id's set that originate from the header and then copy that build-id onto the new dso if the name matches. Another way to do it would be to stop comparing the mmap id so it matches on filename only, but I think we do want to differentiate between different mmaps, even if they have the same name, which is how it works in this version.
Applies to perf/core 859f7e4554
Thanks
James
James Clark (1):
perf: Set build-id using build-id header on new mmap records
tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)
Signature · 1 line
2.28.0
PATCH 1/1 perf: Set build-id using build-id header on new mmap records
MMAP records that occur after the build-id header is parsed do not have their build-id set even if the filename matches an entry from the header. Set the build-id on these dsos as long as the MMAP record doesn't have its own build-id set.
This fixes an issue with off target analysis where the local version of a dso is loaded rather than one from ~/.debug via a build-id.
--- tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)
tools/perf/util/dso.h +1 -0
@@ -167,6 +167,7 @@ struct dso {enum dso_load_errno load_errno; u8 adjust_symbols:1; u8 has_build_id:1; + u8 header_build_id:1; u8 has_srcline:1; u8 hit:1; u8 annotate_warned:1;
tools/perf/util/header.c +1 -0
@@ -2200,6 +2200,7 @@ static int __event_process_build_id(struct perf_record_header_build_id *bev,build_id__init (&bid, bev->data, size);dso__set_build_id (dso, &bid); + dso->header_build_id =1 ;if (dso_space !=DSO_SPACE__USER ) {struct kmod_path m = { .name =NULL , };
tools/perf/util/map.c +14 -2
@@ -127,7 +127,7 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (map !=NULL ) {char newfilename[PATH_MAX ]; -struct dso *dso; +struct dso *dso, *header_bid_dso;int anon, no_dso, vdso, android; android =is_android_lib (filename);
@@ -185,7 +185,19 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (build_id__is_defined (bid))dso__set_build_id (dso, bid); - +else { +/* +* If the mmap event had no build ID, search for an existing dso from the +* build ID header by name. Otherwise only the dso loaded at the time of +* reading the header will have the build ID set and all future mmaps will +* have it missing. +*/ + header_bid_dso =__dsos__find (&machine->dsos, filename, false); +if (header_bid_dso && header_bid_dso->header_build_id) { +dso__set_build_id (dso, &header_bid_dso->bid); + dso->header_build_id =1 ; + } + }dso__put (dso); }return map;
Signature · 1 line
2.28.0
Quoted 7 lines · James Clark
MMAP records that occur after the build-id header is parsed do not have their build-id set even if the filename matches an entry from the header. Set the build-id on these dsos as long as the MMAP record doesn't have its own build-id set.
This fixes an issue with off target analysis where the local version of a dso is loaded rather than one from ~/.debug via a build-id.
nice catch :)
Quoted 47 lines
Reported-by Denis Nikitin <[email protected]>Signed-off-by James Clark <[email protected]>--- tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)tools/perf/util/dso.h +1
-0@@ -167,6 +167,7 @@ struct dso {enum dso_load_errno load_errno; u8 adjust_symbols:1; u8 has_build_id:1; + u8 header_build_id:1; u8 has_srcline:1; u8 hit:1; u8 annotate_warned:1;tools/perf/util/header.c +1
-0@@ -2200,6 +2200,7 @@ static int __event_process_build_id(struct perf_record_header_build_id *bev,build_id__init (&bid, bev->data, size);dso__set_build_id (dso, &bid); + dso->header_build_id =1 ;if (dso_space !=DSO_SPACE__USER ) {struct kmod_path m = { .name =NULL , };tools/perf/util/map.c +1
-1@@ -127,7 +127,7 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (map !=NULL ) {char newfilename[PATH_MAX ]; -struct dso *dso; +struct dso *dso, *header_bid_dso;int anon, no_dso, vdso, android; android =is_android_lib (filename);
tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 188
Quoted 5 lines
tools/perf/util/map.c +1
-1@@ -185,7 +185,19 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (build_id__is_defined (bid))dso__set_build_id (dso, bid); - +else {
nit please add { } to the if clause as well
tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 195
Quoted 7 lines
tools/perf/util/map.c +7
-0+/* +* If the mmap event had no build ID, search for an existing dso from the +* build ID header by name. Otherwise only the dso loaded at the time of +* reading the header will have the build ID set and all future mmaps will +* have it missing. +*/ + header_bid_dso =__dsos__find (&machine->dsos, filename, false);
is this 'perf top' safe? I think dso should be added in the same thread, but please check and add comment why we don't need locking in here
thanks,
jirka
Quoted 11 lines
tools/perf/util/map.c +5
-0+if (header_bid_dso && header_bid_dso->header_build_id) { +dso__set_build_id (dso, &header_bid_dso->bid); + dso->header_build_id =1 ; + } + }dso__put (dso); }return map;Signature · 1 line
2.28.0
Quoted 66 lines · Jiri Olsa
Quoted 7 lines · James Clark
MMAP records that occur after the build-id header is parsed do not have their build-id set even if the filename matches an entry from the header. Set the build-id on these dsos as long as the MMAP record doesn't have its own build-id set.
This fixes an issue with off target analysis where the local version of a dso is loaded rather than one from ~/.debug via a build-id.
nice catch :)
Quoted 47 lines
Reported-by Denis Nikitin <[email protected]>Signed-off-by James Clark <[email protected]>--- tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)tools/perf/util/dso.h +1
-0@@ -167,6 +167,7 @@ struct dso {enum dso_load_errno load_errno; u8 adjust_symbols:1; u8 has_build_id:1; + u8 header_build_id:1; u8 has_srcline:1; u8 hit:1; u8 annotate_warned:1;tools/perf/util/header.c +1
-0@@ -2200,6 +2200,7 @@ static int __event_process_build_id(struct perf_record_header_build_id *bev,build_id__init (&bid, bev->data, size);dso__set_build_id (dso, &bid); + dso->header_build_id =1 ;if (dso_space !=DSO_SPACE__USER ) {struct kmod_path m = { .name =NULL , };tools/perf/util/map.c +1
-1@@ -127,7 +127,7 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (map !=NULL ) {char newfilename[PATH_MAX ]; -struct dso *dso; +struct dso *dso, *header_bid_dso;int anon, no_dso, vdso, android; android =is_android_lib (filename);tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 188
Quoted 5 lines
tools/perf/util/map.c +1
-1@@ -185,7 +185,19 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (build_id__is_defined (bid))dso__set_build_id (dso, bid); - +else {nit please add { } to the if clause as well
Quoted 11 lines
tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 195
Quoted 7 lines
tools/perf/util/map.c +7
-0+/* +* If the mmap event had no build ID, search for an existing dso from the +* build ID header by name. Otherwise only the dso loaded at the time of +* reading the header will have the build ID set and all future mmaps will +* have it missing. +*/ + header_bid_dso =__dsos__find (&machine->dsos, filename, false);is this 'perf top' safe? I think dso should be added in the same thread, but please check and add comment why we don't need locking in here
Seems like there are multiple synthesize_threads_workers using the same machine->dsos object so I think locking is needed.
At first I thought of doing this:
diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
index 4ae91e491e23..b87b81e3d41c 100644
--- a/tools/perf/util/map.c
+++ b/tools/perf/util/map.c
@@ -192,7 +192,9 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,
* reading the header will have the build ID set and all future mmaps will
* have it missing.
*/
+ down_read(&machine->dsos.lock);
header_bid_dso = __dsos__find(&machine->dsos, filename, false);
+ up_read(&machine->dsos.lock);
if (header_bid_dso && header_bid_dso->header_build_id) {
dso__set_build_id(dso, &header_bid_dso->bid);
dso->header_build_id = 1;But then I was wondering why it doesn't need a write lock all the way from machine__findnew_dso_id() to dso__put()? At the moment there are writes to the dso like dso__set_loaded(), dso->nsinfo = nsi and dso__set_build_id(), so another thread could find the dso in a partially constructed state.
Not sure if this is an issue currently without my patch, but at least with it they would have to be found with header_build_id already set to 1 otherwise it will mess things up.
Extending the write lock outside of machine__findnew_dso_id() is difficult because it already releases it before it returns. Does it need to be changed so that machine__findnew_dso_id() takes all the arguments needed to construct it inside the lock?
James
Quoted 15 lines
thanks,
jirkaQuoted 11 lines
tools/perf/util/map.c +5
-0+if (header_bid_dso && header_bid_dso->header_build_id) { +dso__set_build_id (dso, &header_bid_dso->bid); + dso->header_build_id =1 ; + } + }dso__put (dso); }return map;Signature · 1 line
2.28.0
Quoted 101 lines · James Clark
Quoted 66 lines · Jiri Olsa
Quoted 7 lines · James Clark
MMAP records that occur after the build-id header is parsed do not have their build-id set even if the filename matches an entry from the header. Set the build-id on these dsos as long as the MMAP record doesn't have its own build-id set.
This fixes an issue with off target analysis where the local version of a dso is loaded rather than one from ~/.debug via a build-id.
nice catch :)
Quoted 47 lines
Reported-by Denis Nikitin <[email protected]>Signed-off-by James Clark <[email protected]>--- tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)tools/perf/util/dso.h +1
-0@@ -167,6 +167,7 @@ struct dso {enum dso_load_errno load_errno; u8 adjust_symbols:1; u8 has_build_id:1; + u8 header_build_id:1; u8 has_srcline:1; u8 hit:1; u8 annotate_warned:1;tools/perf/util/header.c +1
-0@@ -2200,6 +2200,7 @@ static int __event_process_build_id(struct perf_record_header_build_id *bev,build_id__init (&bid, bev->data, size);dso__set_build_id (dso, &bid); + dso->header_build_id =1 ;if (dso_space !=DSO_SPACE__USER ) {struct kmod_path m = { .name =NULL , };tools/perf/util/map.c +1
-1@@ -127,7 +127,7 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (map !=NULL ) {char newfilename[PATH_MAX ]; -struct dso *dso; +struct dso *dso, *header_bid_dso;int anon, no_dso, vdso, android; android =is_android_lib (filename);tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 188
Quoted 5 lines
tools/perf/util/map.c +1
-1@@ -185,7 +185,19 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (build_id__is_defined (bid))dso__set_build_id (dso, bid); - +else {nit please add { } to the if clause as well
Quoted 11 lines
tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 195
Quoted 7 lines
tools/perf/util/map.c +7
-0+/* +* If the mmap event had no build ID, search for an existing dso from the +* build ID header by name. Otherwise only the dso loaded at the time of +* reading the header will have the build ID set and all future mmaps will +* have it missing. +*/ + header_bid_dso =__dsos__find (&machine->dsos, filename, false);is this 'perf top' safe? I think dso should be added in the same thread, but please check and add comment why we don't need locking in here
Seems like there are multiple
synthesize_threads_workersusing the samemachine->dsosobject so I think locking is needed.At first I thought of doing this:
diff --git
a/tools/perf/util/map.cb/tools/perf/util/map.c
index4ae91e491e23..b87b81e3d41c100644
---a/tools/perf/util/map.c
+++b/tools/perf/util/map.c
@@ -192,7 +192,9 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,
* reading the header will have the build ID set and all future mmaps will
* have it missing.
*/+ down_read(&machine->dsos.lock); header_bid_dso = __dsos__find(&machine->dsos, filename, false); + up_read(&machine->dsos.lock); if (header_bid_dso && header_bid_dso->header_build_id) { dso__set_build_id(dso, &header_bid_dso->bid); dso->header_build_id = 1;
Quoted 3 lines
But then I was wondering why it doesn't need a write lock all the way from
machine__findnew_dso_id()todso__put()? At the moment there are writes to the dso likedso__set_loaded(),dso->nsinfo= nsi anddso__set_build_id(), so another thread could find the dso in a partially constructed state.
I think that's fine, machine__findnew_dso_id takes the dso ref so the real 'release' is when the machine object goes down at dsos__purge
Quoted 3 lines
Not sure if this is an issue currently without my patch, but at least with it they would have to be found with
header_build_idalready set to 1 otherwise it will mess things up.
as for the partial changes I think it's also fine, because it happens at separate places.. we'd need to investigate specific example to see if there's a problem
Quoted 4 lines
Extending the write lock outside of
machine__findnew_dso_id()is difficult because it already releases it before it returns. Does it need to be changed so thatmachine__findnew_dso_id()takes all the arguments needed to construct it inside the lock?
the change above should do it.. as for top threads I don't think it actually matters because there's events processing and hists processing, and I did not find them clashing in this.. but there's some setup session code that touches this, so it's better to be safe
thanks,
jirka
Quoted 118 lines · Jiri Olsa
Quoted 101 lines · James Clark
Quoted 66 lines · Jiri Olsa
Quoted 7 lines · James Clark
MMAP records that occur after the build-id header is parsed do not have their build-id set even if the filename matches an entry from the header. Set the build-id on these dsos as long as the MMAP record doesn't have its own build-id set.
This fixes an issue with off target analysis where the local version of a dso is loaded rather than one from ~/.debug via a build-id.
nice catch :)
Quoted 47 lines
Reported-by Denis Nikitin <[email protected]>Signed-off-by James Clark <[email protected]>--- tools/perf/util/dso.h | 1 + tools/perf/util/header.c | 1 + tools/perf/util/map.c | 16 ++++++++++++++-- 3 files changed, 16 insertions(+), 2 deletions(-)tools/perf/util/dso.h +1
-0@@ -167,6 +167,7 @@ struct dso {enum dso_load_errno load_errno; u8 adjust_symbols:1; u8 has_build_id:1; + u8 header_build_id:1; u8 has_srcline:1; u8 hit:1; u8 annotate_warned:1;tools/perf/util/header.c +1
-0@@ -2200,6 +2200,7 @@ static int __event_process_build_id(struct perf_record_header_build_id *bev,build_id__init (&bid, bev->data, size);dso__set_build_id (dso, &bid); + dso->header_build_id =1 ;if (dso_space !=DSO_SPACE__USER ) {struct kmod_path m = { .name =NULL , };tools/perf/util/map.c +1
-1@@ -127,7 +127,7 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (map !=NULL ) {char newfilename[PATH_MAX ]; -struct dso *dso; +struct dso *dso, *header_bid_dso;int anon, no_dso, vdso, android; android =is_android_lib (filename);tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 188
Quoted 5 lines
tools/perf/util/map.c +1
-1@@ -185,7 +185,19 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,if (build_id__is_defined (bid))dso__set_build_id (dso, bid); - +else {nit please add { } to the if clause as well
Quoted 11 lines
tools/perf/util/map.c struct map *map__new(struct machine *machine, u64 start, u64 len, · line 195
Quoted 7 lines
tools/perf/util/map.c +7
-0+/* +* If the mmap event had no build ID, search for an existing dso from the +* build ID header by name. Otherwise only the dso loaded at the time of +* reading the header will have the build ID set and all future mmaps will +* have it missing. +*/ + header_bid_dso =__dsos__find (&machine->dsos, filename, false);is this 'perf top' safe? I think dso should be added in the same thread, but please check and add comment why we don't need locking in here
Seems like there are multiple
synthesize_threads_workersusing the samemachine->dsosobject so I think locking is needed.At first I thought of doing this:
diff --git
a/tools/perf/util/map.cb/tools/perf/util/map.c
index4ae91e491e23..b87b81e3d41c100644
---a/tools/perf/util/map.c
+++b/tools/perf/util/map.c
@@ -192,7 +192,9 @@ struct map *map__new(struct machine *machine, u64 start, u64 len,
* reading the header will have the build ID set and all future mmaps will
* have it missing.
*/+ down_read(&machine->dsos.lock); header_bid_dso = __dsos__find(&machine->dsos, filename, false); + up_read(&machine->dsos.lock); if (header_bid_dso && header_bid_dso->header_build_id) { dso__set_build_id(dso, &header_bid_dso->bid); dso->header_build_id = 1;Quoted 3 lines
But then I was wondering why it doesn't need a write lock all the way from
machine__findnew_dso_id()todso__put()? At the moment there are writes to the dso likedso__set_loaded(),dso->nsinfo= nsi anddso__set_build_id(), so another thread could find the dso in a partially constructed state.I think that's fine,
machine__findnew_dso_idtakes the dso ref so the real 'release' is when the machine object goes down atdsos__purgeQuoted 3 lines
Not sure if this is an issue currently without my patch, but at least with it they would have to be found with
header_build_idalready set to 1 otherwise it will mess things up.as for the partial changes I think it's also fine, because it happens at separate places.. we'd need to investigate specific example to see if there's a problem
Quoted 1 line
Quoted 8 lines
Quoted 3 lines
Extending the write lock outside of
machine__findnew_dso_id()is difficult because it already releases it before it returns. Does it need to be changed so thatmachine__findnew_dso_id()takes all the arguments needed to construct it inside the lock?the change above should do it.. as for top threads I don't think it actually matters because there's events processing and hists processing, and I did not find them clashing in this.. but there's some setup session code that touches this, so it's better to be safe
Ok yes I think I agree. The header flag isn't important for perf top so having it all constructed in place isn't necessary.
I've submitted v2 with this change.
Thanks for the review.
Quoted 3 lines
thanks,
jirka