libbpf-tools/klockstat: Search for correct stack offset when printing #5203
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
tohojo
requested review from
drzaeus77,
goldshtn,
yonghong-song,
4ast,
brendangregg and
davemarchevsky
as code owners
|
|
||
| for (j = 0; j < ARRAY_SIZE(internal_ksym_names); j++) | ||
| if (!strcmp(ksym->name, internal_ksym_names[j])) | ||
| ret = i + 1; |
There was a problem hiding this comment.
Should we break out if 'ret = i + 1' is not triggered?
There was a problem hiding this comment.
Hmm, no, I don't think so; the loop is trying to find the top-most instance of one of the names in the list, so to do that we'll keep matching while walking the whole stack. I.e., I'm not assuming that it's definitely an off-by-one thing here...
There was a problem hiding this comment.
@tohojo Could you show an stack example before and after the change for the above case? It will give a visible example so we can decide what is the proper way for the stack trace.
The klockstat BPF code contains a hard-coded offset (of 4) for the
captured stack entries, which is supposed to get rid of the topmost
entries that are just locking functions. However, when running the
klockstat tool on RHEL, this offset is not correct, leading to output
like the following:
Caller Avg Wait Count Max Wait Total Wait
mutex_lock+0x5 1.1 us 3 1.8 us 3.2 us
mutex_lock+0x5 825 ns 5 1.6 us 4.1 us
mutex_lock+0x5 1.1 us 4 1.5 us 4.3 us
Caller Avg Hold Count Max Hold Total Hold
mutex_lock+0x5 18.0 ms 3 23.0 ms 54.0 ms
mutex_lock+0x5 1.7 ms 15 2.2 ms 25.1 ms
mutex_lock+0x5 1.7 ms 15 2.2 ms 25.1 ms
To fix this, add logic in the userspace component of the utility to
walk the stack frames, skipping any entries that correspond to any of
the kprobe hooks that the tool attaches to. This fixes the output
on RHEL, without affecting other systems where the offset works
correctly.
Signed-off-by: Toke Høiland-Jørgensen <[email protected]>
tohojo
force-pushed
the
klockstat-stack-fix
branch
from
447cd67 to
de3cde1
Compare
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 klockstat BPF code contains a hard-coded offset (of 4) for the
captured stack entries, which is supposed to get rid of the topmost
entries that are just locking functions. However, when running the
klockstat tool on RHEL, this offset is not correct, leading to output
like the following:
To fix this, add logic in the userspace component of the utility to
walk the stack frames, skipping any entries that correspond to any of
the kprobe hooks that the tool attaches to. This fixes the output
on RHEL, without affecting other systems where the offset works
correctly.
Signed-off-by: Toke Høiland-Jørgensen [email protected]