Hacktoberfest 2026：维护者为十月标记出来的 issue，仍然开放、适合新手。 浏览 Hacktoberfest issue

A handler installed with `SA_RESTART` does not restart the interrupted syscall

未关闭
  • 哪里不对？
#418 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 3 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
25/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
c, linux
领域
operating-systems

调研方向

Start with syscall_restart_arm in src/syscall/proc.h, which today handles only execve on the thread group leader, and read the comment explaining why a signal delivered before resume cancels it. The fix depends on #384, since the handler frame carries the wire value in X8 rather than the syscall number. Build the reproducer from the issue, which should match Linux for FUTEX_WAIT and a pipe read (rc=0 and rc=1 respectively) instead of EINTR.

由索引模型根据 Issue 内容生成。

描述

Linux restarts a syscall that a signal interrupted when the handler was installed with SA_RESTART and the syscall is one of the restartable ones in signal(7): read on a pipe or socket, futex FUTEX_WAIT without a timeout, wait4, accept and so on. elfuse returns EINTR to the guest instead. Measured for the two below; the others are untested.

The dispatcher has a restart path (syscall_restart_arm in src/syscall/proc.h), but it serves only an execve handed to the thread group leader, and a signal delivered before the guest resumes cancels it. The comment there gives the reason: the frame the handler returns through carries the wire value in X8 and not the syscall number, so a rewound PC would re-execute a different syscall. That is #384, which this depends on.

Reproducer: a SIGUSR1 handler installed with SA_RESTART, one thread parked in each wait, pthread_kill(SIGUSR1) at it, then the wait released (a FUTEX_WAKE after storing 1, a one-byte write).

#include <errno.h>
#include <linux/futex.h>
#include <pthread.h>
#include <signal.h>
#include <stdint.h>
#include <stdio.h>
#include <string.h>
#include <sys/syscall.h>
#include <unistd.h>

static uint32_t word;
static int pfd[2];
static volatile int handled;

static void on(int s)
{
    (void) s;
    handled++;
}

static void *w_futex(void *a)
{
    (void) a;
    long r = syscall(SYS_futex, &word, FUTEX_WAIT, 0, NULL, NULL, 0);
    printf("futex FUTEX_WAIT: rc=%ld errno=%d\n", r, r < 0 ? errno : 0);
    return NULL;
}

static void *w_read(void *a)
{
    (void) a;
    char c;
    long r = read(pfd[0], &c, 1);
    printf("read on a pipe:   rc=%ld errno=%d\n", r, r < 0 ? errno : 0);
    return NULL;
}

static void rel_futex(void)
{
    word = 1;
    syscall(SYS_futex, &word, FUTEX_WAKE, 1, NULL, NULL, 0);
}

static void rel_read(void)
{
    write(pfd[1], "x", 1);
}

static void run(void *(*fn)(void *), void (*release)(void))
{
    pthread_t t;
    pthread_create(&t, NULL, fn, NULL);
    usleep(200000);
    pthread_kill(t, SIGUSR1);
    usleep(300000);
    release();
    pthread_join(t, NULL);
}

int main(void)
{
    struct sigaction sa;
    memset(&sa, 0, sizeof sa);
    sa.sa_handler = on;
    sa.sa_flags = SA_RESTART;
    sigaction(SIGUSR1, &sa, NULL);
    pipe(pfd);
    run(w_futex, rel_futex);
    run(w_read, rel_read);
    printf("handler ran %d times\n", handled);
    return 0;
}
wait elfuse Linux 6.12 under qemu
futex(FUTEX_WAIT), no timeout rc=-1 errno=4 rc=0
read on an empty pipe rc=-1 errno=4 rc=1

The handler runs both times on both.

The waits signal(7) lists as never restarted (nanosleep, select, poll, epoll_wait among them) are out of scope: EINTR is right for those.

主要语言
C
星标
273
派生
29
平均合并
2 天 8 小时
30 天内合并 PR
17

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

sysprog21/elfuse 的其他 Issue

查看 sysprog21/elfuse 的全部 Issue

相似的 Issue

更多 C Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。