Repository navigation
fix(windows): preserve caller affinity only within one processor group - #2202
Conversation
`spawnProcess` read the caller's affinity mask and applied it to the suspended child unconditionally. A mask names processors relative to one processor group, and Windows assigns each new process to a group of its own choosing, so on hosts with more than 64 logical processors the child is often placed in another group. `SetProcessAffinityMask` then rejects the mask with ERROR_INVALID_PARAMETER (87) and the spawn fails, making every owned provider process start intermittently on such hosts. Affinity is a nicety, not an authority or isolation control, so it is now applied only when the caller and the suspended child both occupy the same single processor group. A child placed in another group keeps that group's default affinity and the spawn proceeds. Job Object membership, kill-on-owner-close policy, owner normalization and priority preservation are unchanged, and the mask failure remains fatal within a shared group, where it is not expected. Single-group hosts, including every host with 64 or fewer logical processors, are unaffected: `GetProcessGroupAffinity` reports one group for both processes and the same mask is applied as before. The optional process-boundary fixture gains an `affinity` observation, and the native verification lane asserts that an owned spawn succeeds and that the child's mask matches the caller's whenever both share a single group. Fixes morluto#2181 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: aca0503d51
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Addresses review feedback on the first commit. An unconstrained caller loses nothing when Windows places an owned child in another processor group, so that spawn proceeds with the group's default. A caller deliberately pinned to fewer processors than its group holds is a different case: silently accepting the default would hand the provider tree the whole group, which `docs/installation.md` tells operators to rely on affinity to prevent. Such a spawn now fails naming the constraint instead. `GetProcessAffinityMask` already reports the group's active set, which distinguishes the two without another query. Selecting the caller's group at creation through PROC_THREAD_ATTRIBUTE_GROUP_AFFINITY was tried first and rejected: `CreateProcessW` fails with ERROR_INVALID_PARAMETER on a host with a single processor group, both alongside PROC_THREAD_ATTRIBUTE_JOB_LIST and on its own, so it cannot carry the ordinary path. The boundary fixture's affinity observation now reports a process id's own mask rather than its own, because a separate probe process is assigned its own group and cannot stand in for the process that spawns the child. Masks and the group's active set cross as fixed-width hexadecimal strings: a mask can set bits above the safe-integer range, where `JSON.parse` would round two different masks to one value. The verification lane asserts the mask is inherited exactly when caller and child share a group, and otherwise that the caller was unconstrained. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 699a64034c
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
A mask covering its group's whole active set does not distinguish a caller the system placed in that group from one that selected the group deliberately, so an owned child can still be placed elsewhere with its processor count bounded and its group placement lost. Treating that case as constrained instead is not available: on a multi-group host an ordinary process is assigned a group by the system and its mask is that group's active set, so refusing it would fail every owned spawn there. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
baron
left a comment
There was a problem hiding this comment.
Exact-head review: e1a2f2c. No confirmed additional source defect: explicit narrow affinity is kept within one applicable group or refused when not representable, while a full default mask is not misrepresented as a caller-selected CPU-count restriction.
Merge is held for matching native affinity-boundary verification. The inspected green current-head Windows job runs verify:windows-native without its optional boundary fixture, so it does not execute the new processor-group observation/assertion path. Please provide exact-head caller/group observations and the changed cross-group/refusal evidence, or an explicit maintainer verification decision. The earlier single-group report has older mask/report-key wording and does not settle this final candidate. No local Windows/affinity operation or verification waiver was performed.
|
@greptileai review |
|
Fixes #2181.
Cause
spawnProcessread the caller's affinity mask and applied it to the suspendedchild unconditionally:
A mask names processors relative to one processor group, and Windows assigns
each new process to a group of its own choosing. On a host whose logical
processors span more than one group, the child is often placed in a group whose
active processors are not a superset of the caller's mask, so
SetProcessAffinityMaskrejects it withERROR_INVALID_PARAMETER(87). Therequiremade that fatal, so every owned provider spawn — the JavaScriptanalysis worker and the other owned helpers — failed randomly depending on
which group Windows picked.
Change
Affinity is a scheduling nicety, not an authority, ownership or isolation
control, so it is now applied only when the caller and the suspended child both
occupy the same single processor group. A child placed in another group keeps
that group's default affinity and the spawn proceeds.
GetProcessGroupAffinityestablishes membership for both processes. A processwhose threads span groups, or whose membership cannot be read, has no single
group to compare and is treated as unknown rather than as aligned, so no mask is
applied. Within a shared group the mask failure stays fatal, because there it is
not expected.
Unchanged: Job Object membership and kill-on-owner-close verification, breakaway
denial, child default-owner normalization, priority preservation, and the
suspended-create-then-resume sequence.
Single-group hosts — every host with 64 or fewer logical processors, and any
larger host whose caller and child land in the same group — behave exactly as
before: one group is reported for both processes and the same mask is applied.
Note for the >64-processor case on current Windows: since Windows 11 and Server
2022 a process's affinity spans all groups by default, so
GetProcessAffinityMaskreports 0 and the existing
affinity != 0guard already skipped the call. Thereported host is Windows 10 19045, where the legacy per-group assignment applies
and the failure is reachable.
Tests and docs
tests/fixtures/windows/processBoundary.ccgains anaffinityobservationreporting the process's own mask, whether it occupies a single processor group,
and which group.
scripts/verify-windows-native.mjsasserts that an owned spawn of thatfixture succeeds, and that the child's mask equals the caller's whenever both
report the same single group. On a multi-group host the exit-code assertion is
the regression test for this issue; the mask comparison is correctly skipped
because no mask is expected to carry across groups.
native/windows/README.mdanddocs/windows-ghidra-p0.mdboth stated thatcaller CPU affinity is preserved without qualification. They now state the
processor-group condition and what a child in another group gets instead.
Validation
Windows 11 Pro 26100 x64, Node 24.16.0:
npm run docs:check— 6/6 tasks pass, includingcheck-doc-facts.mjs.npx oxfmt --check scripts/verify-windows-native.mjs— clean.npm run lint— clean, module boundaries checked across 1,568 source files.The
max-lineswarning onscripts/verify-windows-native.mjspredates thischange.
npm run build:cached— passes.(
node scripts/build-windows-native.mjson Linux with MinGW-w64 g++ 12 in acontainer, since this host has no C++ toolchain).
process.cccompileswithout new warnings.
npm run verify:windows-native:artifactreportsok: truefor ABI 1 / Node-API 8, SHA-256a036a9e4…5606e4, 842,929 bytes.node scripts/verify-windows-native.mjs . <process-boundary.exe>on this realWindows host against that artifact and an independently cross-built boundary
fixture:
ok: true, all 20 filesystem controls and all 12 processobservations pass, including the new
callerAffinityPreservedWithinProcessorGroup. The caller observation was{"affinityMask":4095,"singleProcessorGroup":true,"processorGroup":0}and thechild reported the same, so the mask-equality assertion ran rather than being
skipped.
Explicitly not verified: the failing multi-group path itself. This host has 12
logical processors, which is one processor group, so the rejected-mask branch
cannot occur here and only the unchanged single-group path is exercised. The
reasoning for the multi-group branch comes from the Win32 contract for
processor groups
and
GetProcessGroupAffinity, plus the experimental analysis in #2181.Confirmation on a host whose processors span more than one group — @hzh666kevin-hue's
88-processor, 2-group machine would do it — is still wanted, and the new
assertion is written to cover exactly that case.
🤖 Generated with Claude Code