Re: [PATCH] rio: switch to ANSI prototypes

[Al Viro <viro@ftp.linux.org.uk>]

On Thu, Sep 29, 2005 at 08:52:36PM +0400, Alexey Dobriyan wrote:
> On Thu, Sep 29, 2005 at 04:25:56PM +0100, Al Viro wrote:
> > Uh-oh...  Well, if you want to play with it...  FWIW, I'm disabling rio as
> > hopeless FPOS; if you feel masochistic, go ahead but keep in mind that its
> > handling of tty glue is severely b0rken.
> 
> Well, duh... It clutters _my_ logs.

diff -urN RC13-git12-nfs-endian/drivers/char/Kconfig RC13-git12-rio/drivers/char/Kconfig
--- RC13-git12-nfs-endian/drivers/char/Kconfig	2005-09-10 15:41:34.000000000 -0400
+++ RC13-git12-rio/drivers/char/Kconfig	2005-09-12 14:50:05.000000000 -0400
@@ -282,12 +282,13 @@
 
 config RIO
 	tristate "Specialix RIO system support"
-	depends on SERIAL_NONSTANDARD && BROKEN_ON_SMP
+	depends on SERIAL_NONSTANDARD && BROKEN
 	help
 	  This is a driver for the Specialix RIO, a smart serial card which
 	  drives an outboard box that can support up to 128 ports.  Product
 	  information is at <http://www.perle.com/support/documentation.html#multiport>.
 	  There are both ISA and PCI versions.
+	  Note that while card might be smart, driver most certainly isn't.
 
 config RIO_OLDPCI
 	bool "Support really old RIO/PCI cards"

had solved that one nicely for me (and that's the only driver that got
such treatment - this FPOS is really something special).

> > Use of caddr_t is *always* broken.
> 
> "unsigned long arg" or do you keep in mind something more fundamental? 

In this case - unsigned long, in other...  Some are <something> __user *,
some are <something> __iomem *, some are simply void *...