Fix rewrite_extern_constant_load crash on values exceeding int32 range#1054
Open
Fix rewrite_extern_constant_load crash on values exceeding int32 range#1054
Conversation
BPF MOV imm has a 32-bit immediate field that is sign-extended to 64 bits. When rewrite_extern_constant_load() rewrites a LDDW+LDX pair into a MOV imm, it called gsl::narrow<int32_t>(narrowed_value) which throws narrowing_error if the resolved 64-bit constant exceeds INT32_MAX (e.g., 0x80000000 or larger). Fix: check whether the value fits in a sign-extended int32 before rewriting. If it doesn't fit, return false so the caller falls back to the original LDDW+LDX instruction sequence. Add regression test covering boundary values. Signed-off-by: Michael Agun <danielagun@microsoft.com>
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository UI Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (1)
WalkthroughAdds a sign-extension safety check when rewriting 64-bit extern constant loads into a 32-bit BPF MOV immediate, and a regression test that verifies successful rewrites for in-range values and graceful bail-out (return false, no mutation) for out-of-range values. Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~10 minutes Suggested reviewers
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
📝 Coding Plan
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment Tip Migrating from UI to YAML configuration.Use the |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/io/elf_extern_resolve.cpp`:
- Around line 91-93: Replace the current bounds check that compares signed_value
to int32_t limits with an explicit sign-extension round-trip check: compute
int32_t truncated = static_cast<int32_t>(narrowed_value) and then compare
static_cast<int64_t>(truncated) to signed_value (or the int64_t view of
narrowed_value); return false if they differ. Update the comment to state we
perform an explicit sign-extension round-trip for clarity (rather than claiming
implementation-defined behavior), and reference the variables narrowed_value and
signed_value in the code change.
In `@src/test/test_elf_loader.cpp`:
- Around line 547-574: The tests that call rewrite_extern_constant_load (in
sections for 0x80000000, 0x100000000, 0xFFFFFFFF, and 0xDEADBEEFCAFEBABE) must
also assert the instruction list is unchanged when the function bails out; for
each SECTION, make a deep copy of the instructions produced by make_instructions
(e.g., saved_insts = insts or a cloned copy) before calling
rewrite_extern_constant_load, then after the CHECK_NOTHROW and CHECK_FALSE calls
assert equality between the post-call inst list and the saved copy (use the
existing instruction equality comparator used elsewhere in tests) to ensure no
mutation occurred by rewrite_extern_constant_load.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 0da84078-18fb-4f7a-b065-eefb0b28ff45
📒 Files selected for processing (2)
src/io/elf_extern_resolve.cppsrc/test/test_elf_loader.cpp
Signed-off-by: Michael Agun <danielagun@microsoft.com>
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/test/test_elf_loader.cpp`:
- Around line 556-570: Add a positive success case for the sign-extension path:
create a SECTION that uses make_instructions() and calls
rewrite_extern_constant_load(insts, 0, 0xFFFFFFFFFFFFFFFFULL) and assert it
returns true (e.g., via REQUIRE) to verify that -1 (all-ones uint64) survives
the sign-extension round-trip and is rewritable; place this alongside the
existing SECTIONs and reference the helper check functions (make_instructions,
rewrite_extern_constant_load) used in the surrounding tests.
- Around line 538-544: The bail-out test lambda check_bailout_preserves_program
redundantly calls rewrite_extern_constant_load twice on the same insts; change
it to call rewrite_extern_constant_load(insts, 0, value) once, store the
returned bool, assert the call did not throw and that the returned value is
false (e.g., CHECK_NOTHROW and then CHECK(result == false) or
CHECK_FALSE(result)), and finally CHECK(insts == original) to verify no
mutation; update the lambda around rewrite_extern_constant_load, insts, original
to use the single-call pattern.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: c45fd28a-9fb3-4f12-8c6f-6a8391914be5
📒 Files selected for processing (2)
src/io/elf_extern_resolve.cppsrc/test/test_elf_loader.cpp
…uccess case Signed-off-by: Michael Agun <danielagun@microsoft.com>
elazarg
reviewed
Mar 14, 2026
| // sign-extension round-trip; the caller will fall back to the original | ||
| // LDDW+LDX instruction sequence. | ||
| const auto truncated = static_cast<int32_t>(narrowed_value); | ||
| if (static_cast<uint64_t>(static_cast<int64_t>(truncated)) != narrowed_value) { |
Collaborator
There was a problem hiding this comment.
can you add an appropriate abstraction to num_safety.hpp?
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.
Fixes #1053
rewrite_extern_constant_load() rewrites LDDW + LDX instruction pairs into MOV imm when resolving .kconfig extern constants.
It called gsl::narrow<int32_t>(narrowed_value) to store the resolved value into the BPF instruction's 32-bit imm field.
BPF MOV imm has a 32-bit immediate that is sign-extended to 64 bits at runtime.
When the resolved constant doesn't fit in that range (e.g., 0x80000000 or larger as uint64_t), gsl::narrow throws narrowing_error.
This crashes the verifier on any ELF containing a LDDW + LDX DW extern-constant pattern with a large value.
Fix: Check whether the value fits in a sign-extended int32 before rewriting. If it doesn't fit, return false so the caller falls back to the original LDDW + LDX instruction sequence. Replace gsl::narrow with static_cast (now safe — guarded by the range check).
Test: Add regression test with 6 sections covering boundary values (INT32_MAX, 0x80000000, 0x100000000, 0xFFFFFFFF, 0xDEADBEEFCAFEBABE).
Confirms no exception is thrown and the function correctly returns false for out-of-range values.
Summary by CodeRabbit
Bug Fixes
Tests