Merged
Conversation
dylandreimerink
force-pushed
the
feature/weak-kfuncs
branch
from
07aa2bb to
41fcb79
Compare
dylandreimerink
force-pushed
the
feature/weak-kfuncs
branch
3 times, most recently
from
e95384b to
fb4c071
Compare
This commit allows users to mark their kfunc defintions as `__weak`.
Weak kfuncs do not cause an error during loading if the kfunc in
question can't be found. Instead, a poison value is written which will
cause the verifier to bail out if the instruction is evaluated.
In addition, this commit relocates LDIMM64 instructions with kfunc
relocation entries. When relocated, the BTF id of the kfunc is written
to the instruction's immediate field.
Together these two changes allow users to write eBPF programs which can
gracefully handle the absence of a kfunc. To do so the `bpf_ksym_exists`
macro from `bpf_helpers.h` can be used to check if a kfunc is present.
```
void invalid_kfunc(void) __ksym __weak;
__section("tp_btf/task_newtask") int weak_kfunc_missing(void *ctx) {
if (bpf_ksym_exists(invalid_kfunc)) {
invalid_kfunc();
return 0;
}
return 1;
}
```
In this example the kfunc is always invalid, yet the program can still
load since the branch calling the kfunc is unreachable.
So this effectively provides CO-RE capabilities for kfuncs. Especially
important since kfuncs are less stable than helpers, their availability
depending on kernel version, kernel compilation flags, or kernel modules
being loaded or not.
Signed-off-by: Dylan Reimerink <[email protected]>
lmb
force-pushed
the
feature/weak-kfuncs
branch
from
fb4c071 to
034e302
Compare
lmb
approved these changes
Mar 15, 2024
dylandreimerink
added a commit
to dylandreimerink/ebpf
that referenced
this pull request
Mar 21, 2024
The "test_log_fixup" selftest is broken on purpose, with the goal of testing error messages. Due to changes in cilium#1364 we now emit a poison instruction in `bad_relo_subprog` on a instruction which also has a reference to a BPF-to-BPF function. This causes the symbol/reference resolution step to fail. This error is flaky, it only happens when `bad_relo_subprog` is loaded first. If any other program in the collection is loaded first due to random hash map ordering we skip the test instead due to the test checking for verifier errors. Since this test is broken on purpose, we previously just always hit this skip condition. This commit disabled this selftest explicitly which resolves the CI flakiness. Signed-off-by: Dylan Reimerink <[email protected]>
lmb
pushed a commit
that referenced
this pull request
Mar 21, 2024
The "test_log_fixup" selftest is broken on purpose, with the goal of testing error messages. Due to changes in #1364 we now emit a poison instruction in `bad_relo_subprog` on a instruction which also has a reference to a BPF-to-BPF function. This causes the symbol/reference resolution step to fail. This error is flaky, it only happens when `bad_relo_subprog` is loaded first. If any other program in the collection is loaded first due to random hash map ordering we skip the test instead due to the test checking for verifier errors. Since this test is broken on purpose, we previously just always hit this skip condition. This commit disabled this selftest explicitly which resolves the CI flakiness. Signed-off-by: Dylan Reimerink <[email protected]>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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 PR allows users to mark their kfunc defintions as
__weak. Weak kfuncs do not cause an error during loading if the kfunc in question can't be found. Instead, a poison value is written which will cause the verifier to bail out if the instruction is evaluated.In addition, this PR relocates LDIMM64 instructions with kfunc relocation entries. When relocated, the BTF id of the kfunc is written to the instruction's immediate field.
Together these two changes allow users to write eBPF programs which can gracefully handle the absence of a kfunc. To do so the
bpf_ksym_existsmacro frombpf_helpers.hcan be used to check if a kfunc is present.In this example the kfunc is always invalid, yet the program can still load since the branch calling the kfunc is unreachable.
So this effectively provides CO-RE capabilities for kfuncs. Especially important since kfuncs are less stable than helpers, their availability depending on kernel version, kernel compilation flags, or kernel modules being loaded or not.
Fixes: #1355