From: "Theo de Raadt" <deraadt@openbsd.org>
Subject: Re: Unlock getpgrp(2)
To: Vitaliy Makkoveev <mvs@openbsd.org>
Cc: tech@openbsd.org
Date: Mon, 08 Jun 2026 09:57:07 -0600

I don't see a problem with this.

Vitaliy Makkoveev <mvs@openbsd.org> wrote:

> `ps_pgrp' is protected by both kernel lock and `ps_mtx' mutex, so only
> one is enough for read only access. `pg_id' has the type of pid_t.
> 
> The corresponding enternewpgrp() sets `ps_pgrp' to the initialized new
> one with both locks held.
> 
> Index: sys/kern/kern_prot.c
> ===================================================================
> RCS file: /cvs/src/sys/kern/kern_prot.c,v
> diff -u -p -r1.84 kern_prot.c
> --- sys/kern/kern_prot.c	3 Dec 2025 17:05:53 -0000	1.84
> +++ sys/kern/kern_prot.c	8 Jun 2026 13:13:21 -0000
> @@ -94,7 +94,9 @@ int
>  sys_getpgrp(struct proc *p, void *v, register_t *retval)
>  {
>  
> +	mtx_enter(&p->p_p->ps_mtx);
>  	*retval = p->p_p->ps_pgrp->pg_id;
> +	mtx_leave(&p->p_p->ps_mtx);
>  	return (0);
>  }
>  
> Index: sys/kern/syscalls.master
> ===================================================================
> RCS file: /cvs/src/sys/kern/syscalls.master,v
> diff -u -p -r1.272 syscalls.master
> --- sys/kern/syscalls.master	2 Jun 2026 21:46:21 -0000	1.272
> +++ sys/kern/syscalls.master	8 Jun 2026 13:13:21 -0000
> @@ -183,7 +183,7 @@
>  			    gid_t *gidset); }
>  80	STD		{ int sys_setgroups(int gidsetsize, \
>  			    const gid_t *gidset); }
> -81	STD		{ int sys_getpgrp(void); }
> +81	STD NOLOCK	{ int sys_getpgrp(void); }
>  82	STD		{ int sys_setpgid(pid_t pid, pid_t pgid); }
>  83	STD NOLOCK	{ int sys_futex(uint32_t *f, int op, int val, \
>  			    const struct timespec *timeout, uint32_t *g); }
>