Avoid writing jitdump code for block data

[tommcdon]

Fixes #120419. The implementation of JIT dumps in the linux perf tool creates dynamic shared objects representing JIT'd methods. In .NET 10, #113943 changed the default granularity for stubs, which now reports the block of memory that stores stubs instead of the individual methods. If the event is fired before the block memory is committed (only is reserved), PerfJitDumpState::LogMethod fails to read the memory from the block and then all subsequent JIT dump events will fail. The linux perf tool will favor jit dumps over perfmaps if jit dumps are available, which breaks DOTNET_PerfMapEnabled=1 or 2 settings. This fix addresses the issue by no longer reporting the code byte for block allocations, which is only enabled when DOTNET_PerfMapStubGranularity is set to 0 or 1.

[Copilot]

Pull request overview

This PR fixes an issue (#120419) where JIT dumps fail when reporting stub blocks whose memory is reserved but not yet committed. The fix prevents writing code bytes for block-level allocations while maintaining backward compatibility for individual method reporting.

• Adds a reportCodeBlock boolean parameter to PAL_PerfJitDump_LogMethod (defaults to true)
• Updates the stub logging logic to skip code bytes for block-type allocations
• Removes dead code check that would never execute

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/coreclr/pal/inc/pal.h	Adds reportCodeBlock parameter with default value true to maintain backward compatibility
src/coreclr/pal/src/misc/perfjitdump.cpp	Implements conditional code size reporting based on reportCodeBlock parameter; removes dead code check
src/coreclr/vm/perfmap.cpp	Updates call site to pass false for block-type stub allocations

[dotnet-policy-service]

Tagging subscribers to this area: @steveisok, @dotnet/dotnet-diag
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

[noahfalk]

👍

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

src/coreclr/pal/src/misc/perfjitdump.cpp:268

• reportedCodeSize controls how many bytes are appended to the record, but record.code_size remains set to codeSize even when reportCodeBlock is false. Since header.total_size will then not equal sizeof(record)+name+code_size, this looks inconsistent and can be confusing for future maintainers and for jitdump consumers. Consider adding an explicit comment here clarifying that code_size describes the code region size for symbolization while the native code payload is intentionally omitted (and total_size reflects the actual emitted payload size).

            size_t reportedCodeSize = reportCodeBlock ? codeSize : 0;

            size_t bytesRemaining = sizeof(JitCodeLoadRecord) + symbolLen + 1 + reportedCodeSize;

            record.header.timestamp = GetTimeStampNS();
            record.vma = (uint64_t) pCode;
            record.code_addr = (uint64_t) pCode;
            record.code_size = codeSize;
            record.header.total_size = bytesRemaining;

---

You can also share your feedback on Copilot code review. Take the survey.

[tommcdon]

/ba-g failures are known