Linux: Implement perf live event reordering. #62
Merged
Conversation
This file contains 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
ishitatsuyuki
force-pushed
the
reordering
branch
from
af81d43 to
14ec8ba
Compare
mstange
approved these changes
Aug 24, 2023
There was a problem hiding this comment.
Looks great, thank you so much!!
Since it's not used by the perf file importer at the moment, I would move it to samply/src/linux rather than linux_shared - this will fix the macOS build warnings. Were you planning to use it for the perf file importer in the future?
This transposes the perf event loop to an internal iteration-like manner. The `consume_events` function calls the callback as long as there are events to consume, then returns, at which the caller should call into epoll to wait for new events. This simplifies logic especially with the forthcoming live event reordering.
The perf ring buffers are divided by CPUs and processes and naively concatenating read events can yield wrong ordering. In particular, this is problematic for fork and exec events since mmaped modules will go into the wrong process and cause unwind to fail. Introduce a streaming event sorter to tackle this. Note that timestamps should not be considered 100% reliable and cases where timestamps go backwards should be still gracefully handled. In particular for perf, a NMI could interrupt event writes, thus even timestamp inversion within the same ring buffer can happen.
ishitatsuyuki
force-pushed
the
reordering
branch
from
f53f4c0 to
0bb348b
Compare
It makes sense, but I have no problem with the current file importer either. I have moved it to the |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
The perf ring buffers are divided by CPUs and processes and naively concatenating read events can yield wrong ordering. In particular, this is problematic for fork and exec events since mmaped modules will go into the wrong process and cause unwind to fail.
Introduce a streaming event sorter to tackle this. Note that timestamps should not be considered 100% reliable and cases where timestamps go backwards should be still gracefully handled. In particular for perf, a NMI could interrupt event writes, thus even timestamp inversion within the same ring buffer can happen.
This is based on top of #61 for convenience, but still split into commits. I can rebase on top of main if needed.
Link: #26