Re: arch depending sys/io.h

[Geoff Keating <geoffk at cygnus dot com>]

> Mailing-List: contact libc-alpha-help@sourceware.cygnus.com; run by ezmlm
> List-Unsubscribe: <mailto:libc-alpha-unsubscribe-geoffk=cygnus.com@sourceware.cygnus.com>
> List-Subscribe: <mailto:libc-alpha-subscribe@sourceware.cygnus.com>
> List-Archive: <http://sourceware.cygnus.com/ml/libc-alpha/>
> List-Post: <mailto:libc-alpha@sourceware.cygnus.com>
> List-Help: <mailto:libc-alpha-help@sourceware.cygnus.com>, <http://sourceware.cygnus.com/ml/#faqs>
> Date: Fri, 21 Jan 2000 21:27:09 +0100
> From: Thorsten Kukuk <kukuk@suse.de>
> Cc: libc-alpha@sourceware.cygnus.com
> 
> 
> Hi,
> 
> On Fri, Jan 21, Geoff Keating wrote:
> 
> > > Date: Fri, 21 Jan 2000 11:55:55 +0100
> > > From: Thorsten Kukuk <kukuk@suse.de>
> > > 
> > > 
> > > Hi,
> > > 
> > > In the moment, we install on every plattform sys/io.h.
> > > 
> > > But for example SPARC and PPC doesn't have ioperm nor iopl.
> > > Should we remove sys/io.h for this platforms complete ? Or should
> > > we create an empty dummy header ?
> > 
> > Actually, PPC does have a iopl and ioperm syscalls.  I have no idea
> > what if anything they do.
> 
> Why are they not implemented in glibc ? I cannot find them. 

So far, no-one had noticed...

I just now looked in the kernel and I don't feel it's worth
implementing stubs to call them:

asmlinkage int sys_ioperm(unsigned long from, unsigned long num, int on)
{
        printk(KERN_ERR "sys_ioperm()\n");
        return -EIO;
}

int sys_iopl(int a1, int a2, int a3, int a4)
{
        lock_kernel();
        printk(KERN_ERR "sys_iopl(%x, %x, %x, %x)!\n", a1, a2, a3, a4);
        unlock_kernel();
        return (-ENOSYS);
}

int sys_vm86(int a1, int a2, int a3, int a4)
{
        lock_kernel();
        printk(KERN_ERR "sys_vm86(%x, %x, %x, %x)!\n", a1, a2, a3, a4);
        unlock_kernel();
        return (-ENOSYS);
}

int sys_modify_ldt(int a1, int a2, int a3, int a4)
{
        lock_kernel();
        printk(KERN_ERR "sys_modify_ldt(%x, %x, %x, %x)!\n", a1, a2, a3, a4);
        unlock_kernel();
        return (-ENOSYS);
}

:-)

Yes, take the header out.

-- 
- Geoffrey Keating <geoffk@cygnus.com>