Date: Tue, 15 Jun 2021 23:55:19 GMT From: Konstantin Belousov <kib@FreeBSD.org> To: src-committers@FreeBSD.org, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: git: a12e901a5a65 - main - Add a knob to disable dequeueing SIGCHLD on waiting for live process Message-ID: <202106152355.15FNtJQj086720@gitrepo.freebsd.org>
next in thread | raw e-mail | index | archive | help
The branch main has been updated by kib: URL: https://cgit.FreeBSD.org/src/commit/?id=a12e901a5a65417849c1ccf1e37b8d092fa438da commit a12e901a5a65417849c1ccf1e37b8d092fa438da Author: Konstantin Belousov <kib@FreeBSD.org> AuthorDate: 2021-06-05 18:24:35 +0000 Commit: Konstantin Belousov <kib@FreeBSD.org> CommitDate: 2021-06-15 23:00:19 +0000 Add a knob to disable dequeueing SIGCHLD on waiting for live process It seems that Linux does not dequeue siginfo for SIGCHLD when wait*(2) reports status of the running process. In particular, sigwaitinfo(2) and other signal querying syscalls can observe the siginfo after wait. FreeBSD dequeued siginfo from the beginning, so we cannot change the default ABI to be more compatible. Still, add a knob to enable to change to the other behavior for debugging purposes. Reported by: dchagin Reviewed by: dchagin, markj Sponsored by: The FreeBSD Foundation MFC after: 1 week Differential revision: https://reviews.freebsd.org/D30675 --- sys/kern/kern_exit.c | 14 +++++++++++--- sys/sys/sysent.h | 1 + 2 files changed, 12 insertions(+), 3 deletions(-) diff --git a/sys/kern/kern_exit.c b/sys/kern/kern_exit.c index cb5996982a3a..fcd8b39a8ee2 100644 --- a/sys/kern/kern_exit.c +++ b/sys/kern/kern_exit.c @@ -105,6 +105,11 @@ SYSCTL_INT(_kern, OID_AUTO, kill_on_debugger_exit, CTLFLAG_RWTUN, &kern_kill_on_dbg_exit, 0, "Kill ptraced processes when debugger exits"); +static bool kern_wait_dequeue_sigchld = 1; +SYSCTL_BOOL(_kern, OID_AUTO, wait_dequeue_sigchld, CTLFLAG_RWTUN, + &kern_wait_dequeue_sigchld, 0, + "Dequeue SIGCHLD on wait(2) for live process"); + struct proc * proc_realparent(struct proc *child) { @@ -1207,9 +1212,12 @@ report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo, p->p_flag &= ~P_CONTINUED; else p->p_flag |= P_WAITED; - PROC_LOCK(td->td_proc); - sigqueue_take(p->p_ksi); - PROC_UNLOCK(td->td_proc); + if (kern_wait_dequeue_sigchld && + (td->td_proc->p_sysent->sv_flags & SV_SIG_WAITNDQ) == 0) { + PROC_LOCK(td->td_proc); + sigqueue_take(p->p_ksi); + PROC_UNLOCK(td->td_proc); + } } sx_xunlock(&proctree_lock); if (siginfo != NULL) { diff --git a/sys/sys/sysent.h b/sys/sys/sysent.h index 95e9dcb1a335..8b0903f7dcc0 100644 --- a/sys/sys/sysent.h +++ b/sys/sys/sysent.h @@ -162,6 +162,7 @@ struct sysentvec { #define SV_ASLR 0x080000 /* ASLR allowed. */ #define SV_RNG_SEED_VER 0x100000 /* random(4) reseed generation. */ #define SV_SIG_DISCIGN 0x200000 /* Do not discard ignored signals */ +#define SV_SIG_WAITNDQ 0x400000 /* Wait does not dequeue SIGCHLD */ #define SV_ABI_MASK 0xff #define SV_PROC_FLAG(p, x) ((p)->p_sysent->sv_flags & (x))
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202106152355.15FNtJQj086720>