Skip to content

Let the kernel run restart_syscall for interrupted syscalls - #4163

Open
Keno wants to merge 1 commit into
rr-debugger:masterfrom
ChronitonAI:restart-syscall-orig-syscallno
Open

Keno wants to merge 1 commit into
rr-debugger:masterfrom
ChronitonAI:restart-syscall-orig-syscallno

Conversation

@Keno

@Keno Keno commented Nov 18, 2026

Copy link
Copy Markdown
Member

[Encountered, debugged, patch, test by AI. Part of a series of 5 patches around restart handling. Note that this follows up on #4142 which was the same issue for syscalls interrupted by a signal handler. This is the case where the signal is ignored]

At a restart_syscall entry, rr set the task's syscall number back to the interrupted syscall's, so the kernel ran that syscall again from the start: relative sleeps and timeouts started over after every signal without a handler, and a restart that a signal handler had cancelled ran the syscall again instead of failing with EINTR. Put that number only in the recorded registers, so emulated ptracers see restart_syscall at the entry, as natively.

Test: restart_syscall_sleep, ptrace_restart_syscall.

At a restart_syscall entry, rr set the task's syscall number back to
the interrupted syscall's, so the kernel ran that syscall again from
the start: relative sleeps and timeouts started over after every
signal without a handler, and a restart that a signal handler had
cancelled ran the syscall again instead of failing with EINTR. Put
that number only in the recorded registers, so emulated ptracers see
restart_syscall at the entry, as natively.

Test: restart_syscall_sleep, ptrace_restart_syscall.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant