BEGIN/END: Add more reliable implementation #2264
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
danobi
force-pushed
the
special_raw_tp
branch
3 times, most recently
from
72eb896 to
f86d940
Compare
viktormalik
reviewed
Jun 13, 2022
Which parts are you referring to? I left the old codepath in b/c BPF_PROG_RUN for raw_tp was only added in late 2020: https://lore.kernel.org/bpf/[email protected]/ . So I'm concerned we'll leave some of our users hanging if we only use this new approach. |
I found 2 occurrences that could be removed: |
|
Ah yeah good finds. I'll fix that up and debug the test failures some time this week. |
|
Done |
ajor
approved these changes
Jul 11, 2022
viktormalik
approved these changes
Jul 13, 2022
Better to remove as many default cases as possible cuz when someone adds a new enum type they'll get a compiler warning instead of a runtime failure some time later.
Create a separate probe type for BEGIN/END probes. Currently a noop in this commit. We will use this new probe type in the next commit to implement a more reliable special probe.
We will be using BPF_PROG_RUN on this prog type to implement BEGIN/END probes next.
Before, we implemented BEGIN/END probes as uprobes, attached the probes to trigger functions in /proc/self/exe, and triggered the prog by calling the trigger functions. This works well most of the time except for when distros decide to strip symbols from bpftrace. This commit makes BEGIN/END probes more reliable by implementing special probes as BPF_PROG_TYPE_RAW_TRACEPOINT and calling BPF_PROG_RUN on them. This is ok b/c we don't allow users to access any associated `ctx` in speical probes anyways. And we choose raw_tp prog type b/c it has BPF_PROG_RUN implemented. I did try to implement BPF_PROG_RUN on BPF_PROG_TYPE_KPROBE in the kernel but upstream didn't like it too much: https://lore.kernel.org/bpf/[email protected]/ . So we have raw_tp instead.
Even though this change should be somewhat transparent to the user, most distros have stopped stripping bpftrace symbols to work around bpftrace limitations. So add entry in changelog in case anyone is checking.
kmcdonell
added a commit
to kmcdonell/pcp
that referenced
this pull request
Aug 17, 2022
Tests: 1701 1722 1723
Issue:
nm: /usr/bin/bpftrace: no symbols
fixed: bpftrace/bpftrace#2264 but yet arrive
upstream (or is that downstream?)
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.
Before, we implemented BEGIN/END probes as uprobes, attached the probes
to trigger functions in /proc/self/exe, and triggered the prog by calling
the trigger functions. This works well most of the time except for when
distros decide to strip symbols from bpftrace.
This commit makes BEGIN/END probes more reliable by implementing special
probes as BPF_PROG_TYPE_RAW_TRACEPOINT and calling BPF_PROG_RUN on them.
This is ok b/c we don't allow users to access any associated
ctxinspeical probes anyways. And we choose raw_tp prog type b/c it has
BPF_PROG_RUN implemented.
I did try to implement BPF_PROG_RUN on BPF_PROG_TYPE_KPROBE in the
kernel but upstream didn't like it too much:
https://lore.kernel.org/bpf/[email protected]/ .
So we have raw_tp instead.
Checklist
man/adoc/bpftrace.adocand if needed indocs/reference_guide.mdCHANGELOG.md