[profiler] Rework GC roots reporting. #6144
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
Author
DavidKarlas
reviewed
Dec 2, 2017
Contributor
There was a problem hiding this comment.
Maybe rename this also to RootPointer to match new HeapRootRegisterEvent naming
Contributor
|
LGTM |
kumpera
approved these changes
Dec 3, 2017
Contributor
Author
|
Rebased again as there were some conflicts from the merging of #5972. |
This change re-implement GC roots reporting in a way that allows users to correctly track the source of a GC root. The new design is based on emitting GC root information upfront and only report addresses during heap dumps. When a GC root is registered we emit an event that is an address range, the root kind, a key and a text description. A decoder can use this information to match a reported root address with all registration addresses to figure out what they mean. A GC Root key is used to further its meaning. For example, if kind is thread, key is a tid, if kind is static variables, key is the vtable pointer and so on. The build of the change is introducing this key argument across all root registration code and rework our root reporting code to encode their addresses. Some roots use pseudo-addresses when they are not registrable. This is the case of thread stacks/registers and the finalizer queue. Finally, root reporting was changed to happen only once per collection and at the end, leading to correct and useful data being produced.
Contributor
|
the weak fields test was disabled for the interp by a recent comimt by @lewurm |
Contributor
|
<3 |
picenka21
pushed a commit
to picenka21/runtime
that referenced
this pull request
Feb 18, 2022
This change re-implement GC roots reporting in a way that allows users to correctly track the source of a GC root. The new design is based on emitting GC root information upfront and only report addresses during heap dumps. When a GC root is registered we emit an event that is an address range, the root kind, a key and a text description. A decoder can use this information to match a reported root address with all registration addresses to figure out what they mean. A GC Root key is used to further its meaning. For example, if kind is thread, key is a tid, if kind is static variables, key is the vtable pointer and so on. The build of the change is introducing this key argument across all root registration code and rework our root reporting code to encode their addresses. Some roots use pseudo-addresses when they are not registrable. This is the case of thread stacks/registers and the finalizer queue. Finally, root reporting was changed to happen only once per collection and at the end, leading to correct and useful data being produced. Commit migrated from mono/mono@d7ac79c
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is rebased on master and with a few changes compared to #5710:
[Obsolete]has been applied to the old roots-related stuff inMono.Profiler.Log.MONO_PROFILER_EVENT_5definition toprofiler-events.hto avoid breaking existing users.TYPE_HEAP_ROOTevent is now written asaddressfollowed byobject, as opposed to the other way around.TYPE_HEAP_ROOTinmprof-report(it handled v14 incorrectly).staticvariable roots is now theMonoVTablepointer.MonoDomainpointer).TYPE_HEAP_OBJECTandTYPE_ALLOC.emit_ptrin the profiler, which could write nonsense to a buffer if the first pointer written in a buffer wasNULL.gc_resizeprofiler event in SGen.