Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 11 Jul 2015 22:47:32 +0300
From:      Konstantin Belousov <kostikbel@gmail.com>
To:        Ed Schouten <ed@FreeBSD.org>
Cc:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   Re: svn commit: r285404 - head/sys/compat/cloudabi
Message-ID:  <20150711194732.GB2404@kib.kiev.ua>
In-Reply-To: <201507111941.t6BJfV1a072539@repo.freebsd.org>
References:  <201507111941.t6BJfV1a072539@repo.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jul 11, 2015 at 07:41:31PM +0000, Ed Schouten wrote:
> Author: ed
> Date: Sat Jul 11 19:41:31 2015
> New Revision: 285404
> URL: https://svnweb.freebsd.org/changeset/base/285404
> 
> Log:
>   Implement normal and abnormal process termination.
>   
>   CloudABI does not provide an explicit kill() system call, for the reason
>   that there is no access to the global process namespace. Instead, it
>   offers a raise() system call that can at least be used to terminate the
>   process abnormally.
>   
>   CloudABI does not support installing signal handlers. CloudABI's raise()
>   system call should behave as if the default policy is set up. Call into
>   kern_sigaction(SIG_DFL) before calling sys_kill() to force this.
>   
>   Obtained from:	https://github.com/NuxiNL/freebsd
> 
> Modified:
>   head/sys/compat/cloudabi/cloudabi_proc.c
> 
> Modified: head/sys/compat/cloudabi/cloudabi_proc.c
> ==============================================================================
> --- head/sys/compat/cloudabi/cloudabi_proc.c	Sat Jul 11 19:14:09 2015	(r285403)
> +++ head/sys/compat/cloudabi/cloudabi_proc.c	Sat Jul 11 19:41:31 2015	(r285404)
> @@ -26,6 +26,11 @@
>  #include <sys/cdefs.h>
>  __FBSDID("$FreeBSD$");
>  
> +#include <sys/param.h>
> +#include <sys/proc.h>
> +#include <sys/syscallsubr.h>
> +#include <sys/sysproto.h>
> +
>  #include <compat/cloudabi/cloudabi_proto.h>
>  
>  int
> @@ -42,8 +47,8 @@ cloudabi_sys_proc_exit(struct thread *td
>      struct cloudabi_sys_proc_exit_args *uap)
>  {
>  
> -	/* Not implemented. */
> -	return (ENOSYS);
> +	exit1(td, W_EXITCODE(uap->rval, 0));
> +	/* NOTREACHED */
>  }
>  
>  int
> @@ -59,7 +64,49 @@ int
>  cloudabi_sys_proc_raise(struct thread *td,
>      struct cloudabi_sys_proc_raise_args *uap)
>  {
> -
> -	/* Not implemented. */
> -	return (ENOSYS);
> +	static const int signals[] = {
> +		[CLOUDABI_SIGABRT] = SIGABRT,
> +		[CLOUDABI_SIGALRM] = SIGALRM,
> +		[CLOUDABI_SIGBUS] = SIGBUS,
> +		[CLOUDABI_SIGCHLD] = SIGCHLD,
> +		[CLOUDABI_SIGCONT] = SIGCONT,
> +		[CLOUDABI_SIGFPE] = SIGFPE,
> +		[CLOUDABI_SIGHUP] = SIGHUP,
> +		[CLOUDABI_SIGILL] = SIGILL,
> +		[CLOUDABI_SIGINT] = SIGINT,
> +		[CLOUDABI_SIGKILL] = SIGKILL,
> +		[CLOUDABI_SIGPIPE] = SIGPIPE,
> +		[CLOUDABI_SIGQUIT] = SIGQUIT,
> +		[CLOUDABI_SIGSEGV] = SIGSEGV,
> +		[CLOUDABI_SIGSTOP] = SIGSTOP,
> +		[CLOUDABI_SIGSYS] = SIGSYS,
> +		[CLOUDABI_SIGTERM] = SIGTERM,
> +		[CLOUDABI_SIGTRAP] = SIGTRAP,
> +		[CLOUDABI_SIGTSTP] = SIGTSTP,
> +		[CLOUDABI_SIGTTIN] = SIGTTIN,
> +		[CLOUDABI_SIGTTOU] = SIGTTOU,
> +		[CLOUDABI_SIGURG] = SIGURG,
> +		[CLOUDABI_SIGUSR1] = SIGUSR1,
> +		[CLOUDABI_SIGUSR2] = SIGUSR2,
> +		[CLOUDABI_SIGVTALRM] = SIGVTALRM,
> +		[CLOUDABI_SIGXCPU] = SIGXCPU,
> +		[CLOUDABI_SIGXFSZ] = SIGXFSZ,
> +	};
> +	static const struct sigaction sigdfl = {
> +		.sa_handler = SIG_DFL,
> +	};
> +	struct kill_args kill_args = {
> +		.pid = td->td_proc->p_pid,
> +	};
> +
> +	if (uap->sig >= nitems(signals) ||
> +	    (uap->sig != 0 && signals[uap->sig] == 0)) {
> +		/* Invalid signal. */
> +		return (EINVAL);
> +	}
> +	kill_args.signum = signals[uap->sig];
> +
> +	/* Restore to default signal action and send signal. */
> +	kern_sigaction(td, kill_args.signum, &sigdfl, NULL, 0);
> +	return (sys_kill(td, &kill_args));
>  }

This is very strange and unmaintanable approach. Also, you cannot handle
traps this way.

If your ABI does not provide any way to install a signal handler, then
you probably should be already fully set up, due to execsigs() performed
on each exec(2).




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150711194732.GB2404>