Add ISOSDacInterface17 for StressLog enumeration via cDAC by max-charlamb · Pull Request #129272 · dotnet/runtime

max-charlamb · 2026-06-11T02:29:19Z

Note

This PR was created with assistance from GitHub Copilot.

Summary

Add a new ISOSDacInterface17 COM interface that exposes the cDAC's IStressLog contract to SOS and clrmd consumers. This enables reading stress logs from both CoreCLR and NativeAOT processes without hardcoded struct offsets.

Motivation

SOS (!DumpLog) and clrmd currently read stress logs via raw memory reads with hardcoded struct offsets matching the CoreCLR layout. NativeAOT's StressLog struct has a different layout (different lock type, no module table, no padding sentinel), so these tools cannot read NativeAOT stress logs.

The cDAC already has a working IStressLog contract (StressLog_1/StressLog_2) that reads stress logs from both runtimes using datadescriptor.inc field offsets. This PR bridges that contract to the COM interface layer so SOS and clrmd can consume it.

Changes

cDAC contract updates

Add StartTime (wall-clock FILETIME) to data descriptors (CoreCLR + NativeAOT), Data/StressLog.cs, StressLogData record, and contract implementation
Add Address field to ThreadStressLogData for thread identification across the COM boundary
Add StressLog to ContractRegistry

ISOSDacInterface17 definition

Data structs: SOSStressLogData, SOSThreadStressLogData, SOSStressMsgData (defined inline in IDL following modern convention)
Enumerator interfaces: ISOSStressLogThreadEnum, ISOSStressLogMsgEnum
ISOSDacInterface17: GetStressLogData, GetStressLogThreadEnumerator, GetStressLogMessageEnumerator
All APIs return S_FALSE when stress log is not enabled (no separate availability check needed)

SOSDacImpl bridge

Thread enumerator: materialized array with try/catch COM boundary protection
Message enumerator: eagerly materialized with full Reset/GetCount support and GetArguments for per-batch arg retrieval
The native DAC does not implement this interface -- QueryInterface for the IID will fail on the legacy DAC, and only the cDAC handles it

IDL + testing

Full IDL definitions in sospriv.idl
StressLog debuggee + dump-based integration tests validating the contract and the COM interface layer

Test Results

All dump tests pass:

StressLogIsAvailable -- PASSED
StressLogDataIsValid -- PASSED
CanEnumerateThreadsAndMessages -- PASSED

Related PRs

Diagnostics: Use ISOSDacInterface17 for DumpLog when available diagnostics#5873 -- SOS !DumpLog updated to use ISOSDacInterface17 with fallback
CI run: runtime-diagnostics build 1458939

Add a new ISOSDacInterface17 COM interface that exposes the cDAC's IStressLog contract to SOS and clrmd consumers. This enables reading stress logs from both CoreCLR and NativeAOT without hardcoded struct offsets. Changes: - Add StartTime (wall-clock FILETIME) field to cDAC data descriptors, Data/StressLog.cs, StressLogData record, and contract implementation - Add StressLog to ContractRegistry - Add Address field to ThreadStressLogData for thread identification - Define ISOSDacInterface17 with IsStressLogAvailable, GetStressLogData, GetStressLogThreadEnumerator, GetStressLogMessageEnumerator, and GetStressLogMemoryRanges - Define ISOSStressLogThreadEnum, ISOSStressLogMsgEnum, and ISOSStressLogMemoryEnum enumerator interfaces - Implement SOSDacImpl bridge delegating to _target.Contracts.StressLog - Add IDL definitions in sospriv.idl (native DAC does not implement this interface; QueryInterface for the IID will fail on the legacy DAC) - Add StressLog debuggee and StressLogDumpTests for dump-based validation - Add design document explaining the architecture and migration path Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

This PR extends the cDAC stress log support by (1) enriching the StressLog contract data model (notably adding StartTime and thread-log address identity) and (2) introducing a new ISOSDacInterface17 COM surface (IDL + managed declarations + SOSDacImpl bridge) intended to let SOS/clrmd enumerate stress logs without hardcoded layout offsets. It also adds a new dump-test debuggee plus contract-level dump tests and a detailed design document.

Changes:

Add StartTime to StressLog data descriptors (CoreCLR + NativeAOT) and flow it through the managed contract types.
Add ISOSDacInterface17 + related StressLog enumerator interfaces/structs and implement the bridge in SOSDacImpl.
Add StressLog dump debuggee and dump-based integration tests for the StressLog contract, plus a design doc.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
src/native/managed/cdac/tests/DumpTests/StressLogDumpTests.cs	Adds dump-based integration tests for the StressLog contract (availability, header data, basic enumeration).
src/native/managed/cdac/tests/DumpTests/Debuggees/StressLog/StressLog.csproj	New debuggee project enabling stress logging via environment variables.
src/native/managed/cdac/tests/DumpTests/Debuggees/StressLog/Program.cs	New debuggee that generates allocations/GC activity then `FailFast`s to produce a dump.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Legacy/SOSDacImpl.cs	Implements `ISOSDacInterface17` and new StressLog enumerator implementations bridging to the cDAC `IStressLog` contract.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Legacy/ISOSDacInterface.cs	Adds public COM interface/struct declarations for `ISOSDacInterface17` and StressLog enumerators.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Contracts/Data/StressLog.cs	Adds `[Field] StartTime` to the StressLog data model.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Contracts/Contracts/StressLog.cs	Plumbs `StartTime` into `StressLogData` and includes thread-log address in `ThreadStressLogData`.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Abstractions/Contracts/IStressLog.cs	Extends `StressLogData` with `StartTime` and `ThreadStressLogData` with `Address`.
src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Abstractions/ContractRegistry.cs	Exposes `StressLog` via `Target.Contracts.StressLog`.
src/coreclr/vm/datadescriptor/datadescriptor.inc	Exports `StressLog.StartTime` for CoreCLR via data descriptors.
src/coreclr/nativeaot/Runtime/datadescriptor/datadescriptor.inc	Exports `StressLog.StartTime` for NativeAOT via data descriptors.
src/coreclr/inc/sospriv.idl	Adds IDL definitions for `ISOSDacInterface17` and associated StressLog structs/enumerator interfaces.
docs/design/datacontracts/stresslog-isosinterface-design.md	Design rationale and planned consumer migration strategy for exposing StressLog via SOS DAC interface.

jkotas · 2026-06-11T03:26:18Z

Is there a specific reason that regular CoreCLR and NativeAOT stress logs must be different?

- Add [GeneratedComClass] to all three enumerator classes - Use ToTargetPointer for ClrDataAddress comparison - Shrink _lastBatchRaw to actual fetched size - Return ex.HResult instead of E_FAIL in IsStressLogAvailable Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 5 comments.

jkotas · 2026-06-11T22:09:43Z

We should be able to delete https://github.com/max-charlamb/runtime/blob/bd3de83c6195bdbda612e67672fb1248a1014d28/src/coreclr/inc/stresslog.h#L225 once this is in place. (It does not have to be done in this PR.)

- Remove DacpStressLogMemoryRange, ISOSStressLogMemoryEnum, and GetStressLogMemoryRanges from IDL, managed interface, and impl since no SOS consumer uses them. - Refactor SOSStressLogMsgEnum to eagerly materialize messages into an array, enabling Reset and GetCount support consistent with other ISOSEnum implementations. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Rename DacpStressLogData -> SOSStressLogData, DacpThreadStressLogData -> SOSThreadStressLogData, DacpStressMsgData -> SOSStressMsgData - Define structs inline in sospriv.idl with typedef struct and cpp_quote guards, following the SOSMethodData/SOSMemoryRegion convention used by newer interfaces - Remove Dacp forward declarations and dacprivate.h definitions - Put ISOSDacInterface16, ISOSDacInterface17 on same line in class decl Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Use 'written < count' pattern for S_FALSE (standard COM IEnumXxx convention matching SOSMethodEnum/SOSMemoryEnum) - Return E_POINTER directly for null output in GetStressLogData Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 4 comments.

StressMsgHeaderSize, ChunkSize, MaxMessageSize, and PointerSize are implementation details for raw memory walking that the message enumerator now abstracts away. Consumers no longer need them. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Callers can just call GetStressLogData/GetStressLogThreadEnumerator directly -- they return S_FALSE when stress log is not enabled. No need for a separate availability check method. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Messages with Timestamp==0 can appear at chunk boundaries on x86 with small stress log sizes. These represent unwritten/partial message slots and are valid contract output per the spec. Update tests to find the first message with a non-zero timestamp rather than assuming the first message is valid. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 2 comments.

noahfalk

👍

Matches the convention used by other SOSDAC sized-buffer APIs where S_FALSE indicates fewer items returned than requested. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

On x86 with small stress logs, partially-written messages can have non-zero timestamps but null format strings due to chunk-boundary wrapping. Look for messages with both valid timestamp and format string. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 3 comments.

Temporary diagnostics to understand why x86 can't find valid messages: prints thread count, pointer size, per-thread metadata, and per-message timestamp/format/facility/argcount. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 4 comments.

StressLog.modules is an inline array (ModuleDesc modules[MAX_MODULES]), not a pointer. Using [Field] reads the first bytes of the array as a pointer value, which on x86 happens to be the coreclr base address (from modules[0].baseAddress), causing GetFormatPointer's module table lookup to iterate from the PE header instead of the module descriptors. Using [FieldAddress] returns the address of the inline array itself, which is the correct base for iterating the ModuleDesc entries. Verified fix locally against CI x86 dump that previously failed. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

max-charlamb · 2026-06-19T01:30:21Z

/ba-g known test infra issue

> [!NOTE] > This PR was created with assistance from GitHub Copilot. ## Summary Update `!DumpLog` to prefer the cDAC-based `ISOSDacInterface17` for stress log reading when available, falling back to the existing raw-memory path for older runtimes. This enables reading stress logs from both CoreCLR and NativeAOT processes. Companion to dotnet/runtime#129272 which defines `ISOSDacInterface17` and implements the cDAC bridge. ## Changes ### SOS `!DumpLog` update - Added `StressLog::DumpViaInterface17()` in `stressLogDump.cpp` -- a new code path that: - `QueryInterface` for `ISOSDacInterface17` - If available: uses `GetStressLogData`, `GetStressLogThreadEnumerator`, and `GetStressLogMessageEnumerator` to read stress log data through the cDAC contract - Produces identical output format to the legacy path (header block, merged timestamp-ordered messages, `formatOutput` for special format specifiers like `%pM`/`%pT`) - Returns `E_NOINTERFACE` if the interface is not available - Updated `!DumpLog` in `strike.cpp` to try the new path first, falling back to `StressLog::Dump()` if `E_NOINTERFACE` - Added `-legacy` flag to force the legacy raw-read path ### Interface definitions - Added `ISOSDacInterface17`, `ISOSStressLogThreadEnum`, `ISOSStressLogMsgEnum` interfaces and SOS* data structs to `sospriv.h` - Added IID GUIDs to `sospriv_i.cpp` ### Bug fix in legacy `StressLog::Dump` Fixed a pre-existing bug where the legacy `StressLog::Dump` could return a failure HRESULT even after successfully processing all stress log entries. The issue was that `hr` was being overwritten by `ReadVirtual` calls for format strings inside the message loop. If the last format string read failed (e.g., unmapped memory page in the dump on macOS), the error propagated as the function's return value despite all messages being processed correctly. The fix avoids clobbering `hr` in the loop and explicitly sets it based on whether messages were processed. ### Test coverage - Added `!DumpLog` to the `OtherCommands.script` test with `VERIFY:SUCCESS` - Gated stress log enablement (`DOTNET_StressLog=1`) to only the OtherCommands test via `EnableStressLog` property on `TestInformation` ## Backward Compatibility - When running against a runtime **with** `ISOSDacInterface17` (dotnet/runtime#129272): uses the new cDAC-based path, which supports both CoreCLR and NativeAOT - When running against a runtime **without** `ISOSDacInterface17`: `QueryInterface` returns `E_NOINTERFACE`, falls back to the existing `StressLog::Dump()` raw-memory path -- zero behavior change ## Related PRs - **Runtime**: dotnet/runtime#129272 -- Defines `ISOSDacInterface17` and implements the cDAC bridge Co-authored-by: Max Charlamb <maxcharlamb@microsoft.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings June 11, 2026 02:29

max-charlamb requested a review from MichalStrehovsky as a code owner June 11, 2026 02:29

Copilot started reviewing on behalf of max-charlamb June 11, 2026 02:29 View session

github-actions Bot added the area-Diagnostics-coreclr label Jun 11, 2026

max-charlamb added area-Diagnostics-cdac and removed area-Diagnostics-coreclr labels Jun 11, 2026

max-charlamb requested a review from noahfalk June 11, 2026 02:30

max-charlamb force-pushed the maxcharlamb/stresslog-cdac-bridge branch from 1c1c89a to 714265b Compare June 11, 2026 02:33

max-charlamb force-pushed the maxcharlamb/stresslog-cdac-bridge branch from 714265b to b31ec46 Compare June 11, 2026 02:34

Copilot AI reviewed Jun 11, 2026

View reviewed changes

max-charlamb mentioned this pull request Jun 11, 2026

Use ISOSDacInterface17 for DumpLog when available dotnet/diagnostics#5873

Merged

This was referenced Jun 11, 2026

The Operation will be canceled. The next steps may not contain expected logs. dotnet/dnceng#3008

Open

[browser][coreCLR] Wasm.Console.Node.Sample - undefined symbol: SystemInteropJS_GetManagedStackTrace #129229

Closed

Copilot AI review requested due to automatic review settings June 11, 2026 15:21

Copilot started reviewing on behalf of max-charlamb June 11, 2026 15:21 View session

Copilot AI reviewed Jun 11, 2026

View reviewed changes

max-charlamb mentioned this pull request Jun 11, 2026

Make NativeAOT StressLog binary-compatible with CoreCLR #129300

Closed

Max Charlamb and others added 3 commits June 16, 2026 11:50

Copilot AI review requested due to automatic review settings June 17, 2026 13:50

Copilot started reviewing on behalf of max-charlamb June 17, 2026 13:51 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

Max Charlamb and others added 2 commits June 17, 2026 10:13

Copilot AI review requested due to automatic review settings June 17, 2026 14:18

Copilot AI review requested due to automatic review settings June 18, 2026 00:35

Copilot started reviewing on behalf of max-charlamb June 18, 2026 00:35 View session

Copilot AI reviewed Jun 18, 2026

View reviewed changes

Comment thread src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Legacy/SOSDacImpl.cs

Comment thread src/native/managed/cdac/Microsoft.Diagnostics.DataContractReader.Legacy/SOSDacImpl.cs

noahfalk approved these changes Jun 18, 2026

View reviewed changes

This was referenced Jun 18, 2026

[ci-scan] Test failure: cDAC stress test on Windows.11.Arm64.Open #129556

Open

slow macOS - "##[error]The job running on agent Azure Pipelines 9 ran longer than the maximum time of 60 minutes." dotnet/dnceng#1883

Open

Return S_FALSE from GetArguments when fewer args than requested

c8087f8

Matches the convention used by other SOSDAC sized-buffer APIs where S_FALSE indicates fewer items returned than requested. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

steveisok self-requested a review June 18, 2026 15:02

steveisok approved these changes Jun 18, 2026

View reviewed changes

Copilot AI review requested due to automatic review settings June 18, 2026 15:41

Copilot started reviewing on behalf of max-charlamb June 18, 2026 15:42 View session

Copilot AI reviewed Jun 18, 2026

View reviewed changes

Comment thread src/native/managed/cdac/tests/DumpTests/StressLogDumpTests.cs

Comment thread src/native/managed/cdac/tests/DumpTests/StressLogDumpTests.cs

Comment thread src/native/managed/cdac/tests/DumpTests/StressLogDumpTests.cs Outdated

Copilot AI review requested due to automatic review settings June 18, 2026 18:50

Copilot started reviewing on behalf of max-charlamb June 18, 2026 18:50 View session

noahfalk approved these changes Jun 18, 2026

View reviewed changes

Copilot AI reviewed Jun 18, 2026

View reviewed changes

max-charlamb force-pushed the maxcharlamb/stresslog-cdac-bridge branch from bded221 to cb02ef6 Compare June 18, 2026 18:59

rcj1 approved these changes Jun 18, 2026

View reviewed changes

build-analysis Bot mentioned this pull request Jun 19, 2026

Assertion failed: new_time >= loop->time in Component Governance Detection dotnet/dnceng#6510

Open

3 tasks

max-charlamb merged commit ecbe39a into dotnet:main Jun 19, 2026
112 of 129 checks passed

max-charlamb deleted the maxcharlamb/stresslog-cdac-bridge branch June 19, 2026 01:30

dotnet-maestro Bot mentioned this pull request Jun 19, 2026

[main] Source code updates from dotnet/runtime dotnet/dotnet#7293

Merged

dotnet-milestone-bot Bot added this to the 11.0-preview6 milestone Jun 19, 2026

max-charlamb mentioned this pull request Jun 21, 2026

Slim ThreadStressLogData and simplify GetStressMessages API #129673

Open

Conversation

max-charlamb commented Jun 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Motivation

Changes

cDAC contract updates

ISOSDacInterface17 definition

SOSDacImpl bridge

IDL + testing

Test Results

Related PRs

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jkotas commented Jun 11, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jkotas commented Jun 11, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

noahfalk left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

max-charlamb commented Jun 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

max-charlamb commented Jun 11, 2026 •

edited

Loading