aarch64: blocking-syscall resume uses entry-captured PC/SP/LR — engine passes ICU init#10
Merged
Merged
Conversation
…ICU ret-to-0) save_return_context / _reexec set a yielding thread's resume rip/rsp from the caller-passed values, which came from the LIVE per-CPU scratch (user_rip()/ user_rsp()/user_lr()). That scratch is wrong in two cases: - on the wake path (save_return_context(wpid,...)) it holds the WAKER's regs, not wpid's; - on a self-yield, a timer-preempted sibling's SVC can clobber it before the yield decision. Either way the thread resumed on the wrong SP and its `ldp x29,x30,[sp,#..]; ret` read a stale/zero frame record → `ret` to 0 (observed crashing the engine inside fml::icu::ICUContext::ICUContext). Fix: capture the user PC/SP eagerly at the SVC boundary (new per-proc entry_user_rip/entry_user_rsp, set in the IRQ-masked entry capture alongside user_lr), and have save_return_context_inner restore from the TARGET thread's own entry snapshot (the GPRS_CAPTURED gate guarantees each pid's entry_* are written only by its own syscall entry, never clobbered by a cross-pid save). x30 is no longer re-read lazily — the eager capture's value stands. x86 path unchanged (cfg-gated; still uses the passed rip/rsp). Verified: engine now passes ICU init and reaches FlutterEngineInitialize → EmbedderThreadHost worker-pool creation (next blocker is kernel-side in sys_thread_create, separate). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
squirelboy360
added a commit
that referenced
this pull request
Jun 9, 2026
…ICU ret-to-0) (#10) save_return_context / _reexec set a yielding thread's resume rip/rsp from the caller-passed values, which came from the LIVE per-CPU scratch (user_rip()/ user_rsp()/user_lr()). That scratch is wrong in two cases: - on the wake path (save_return_context(wpid,...)) it holds the WAKER's regs, not wpid's; - on a self-yield, a timer-preempted sibling's SVC can clobber it before the yield decision. Either way the thread resumed on the wrong SP and its `ldp x29,x30,[sp,#..]; ret` read a stale/zero frame record → `ret` to 0 (observed crashing the engine inside fml::icu::ICUContext::ICUContext). Fix: capture the user PC/SP eagerly at the SVC boundary (new per-proc entry_user_rip/entry_user_rsp, set in the IRQ-masked entry capture alongside user_lr), and have save_return_context_inner restore from the TARGET thread's own entry snapshot (the GPRS_CAPTURED gate guarantees each pid's entry_* are written only by its own syscall entry, never clobbered by a cross-pid save). x30 is no longer re-read lazily — the eager capture's value stands. x86 path unchanged (cfg-gated; still uses the passed rip/rsp). Verified: engine now passes ICU init and reaches FlutterEngineInitialize → EmbedderThreadHost worker-pool creation (next blocker is kernel-side in sys_thread_create, separate).
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
1 participant
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.
Builds on the merged ARM shell bring-up (#9).
save_return_context/_reexecset a yielding thread's resumerip/rspfrom caller-passed values read from the live per-CPU scratch, which is wrong on the wake path (holds the waker's regs, not the target's) and on a self-yield (a timer-preempted sibling's SVC clobbers it). The thread then resumed on the wrong SP and itsldp x29,x30,[sp,#..]; retread a stale/zero frame record →retto 0, crashing the engine insidefml::icu::ICUContext::ICUContext.Fix: capture the user PC/SP eagerly at the SVC boundary (new per-proc
entry_user_rip/entry_user_rsp, set in the IRQ-masked entry capture alongsideuser_lr);save_return_context_innerrestores from the target thread's own entry snapshot. TheGPRS_CAPTUREDgate guarantees each pid'sentry_*are written only by its own syscall entry. x30 no longer re-read lazily. x86 cfg-gated/unchanged.Verified: engine now passes ICU init and reaches
FlutterEngineInitialize → EmbedderThreadHostworker-pool creation. Next blocker is kernel-side insys_thread_create(a write to addr -1 during worker spawn) — separate, tracked.x86_64 + aarch64 kernels both compile.