PERFORCE change 122167 for review

[Alfred Perlstein]

er, can't you just do a signal(SIGSYS, SIG_IGN)?

* Peter Wemm <peter at FreeBSD.org> [070622 14:38] wrote:
> http://perforce.freebsd.org/chv.cgi?CH=122167
> 
> Change 122167 by peter at peter_daintree on 2007/06/22 21:38:41
> 
> 	Add kern.do_sigsys to control whether an enosys() also causes SIGSYS.  This would
> 	have been very useful to probe for presence of syscalls.  Avoids the need to install
> 	a sigsys handler.   If only this was present for the last 10 years....
> 
> Affected files ...
> 
> .. //depot/projects/hammer/sys/kern/kern_sig.c#80 edit
> 
> Differences ...
> 
> ==== //depot/projects/hammer/sys/kern/kern_sig.c#80 (text+ko) ====
> 
> @@ -111,6 +111,10 @@
>  SYSCTL_INT(_kern, OID_AUTO, forcesigexit, CTLFLAG_RW,
>      &kern_forcesigexit, 0, "Force trap signal to be handled");
>  
> +static int	kern_do_sigsys = 1;
> +SYSCTL_INT(_kern, OID_AUTO, do_sigsys, CTLFLAG_RW,
> +    &kern_do_sigsys, 0, "Have the kernel send SIGSYS on a bad syscall");
> +
>  SYSCTL_NODE(_kern, OID_AUTO, sigqueue, CTLFLAG_RW, 0, "POSIX real time signal");
>  
>  static int	max_pending_per_proc = 128;
> @@ -3168,9 +3172,11 @@
>  {
>  	struct proc *p = td->td_proc;
>  
> -	PROC_LOCK(p);
> -	psignal(p, SIGSYS);
> -	PROC_UNLOCK(p);
> +	if (kern_do_sigsys) {
> +		PROC_LOCK(p);
> +		psignal(p, SIGSYS);
> +		PROC_UNLOCK(p);
> +	}
>  	return (ENOSYS);
>  }
>  

-- 
- Alfred Perlstein