Re: x86 BIOS interface for partitioning and system serial number on SGI UV

[Russ Anderson <rja@sgi.com>]

On Fri, Aug 01, 2008 at 10:34:30AM -0700, H. Peter Anvin wrote:
> Russ Anderson wrote:
> >x86 BIOS interface for partitioning and system serial number on SGI UV
> 
> >-	if (!proc_mkdir("sgi_uv", NULL))
> >+	if (!sgi_uv_dir)
> >+		sgi_uv_dir = proc_mkdir("sgi_uv", NULL);
> >+	if (!sgi_uv_dir)
> > 		return -EINVAL;
> 
> This really seems more that it should belong in sysfs -- 
> /sys/class/firmware presumably.

/proc/sgi_uv already exists, similar to /proc/sgi_sn on Itanium systems.

Would it be /sys/class/firmware?  Reading Documentation/sysfs-rules.txt
seems to indicate somewhere under /sys/devices, such as /sys/devices/system/,
but I may be wrong.

> >+# define BIOS_CALL(result, a0, a1, a2, a3, a4, a5, a6, a7)		\
> >+	do {								\
> >+		/* XXX - the real call goes here */			\
> >+		result.status = BIOS_STATUS_UNIMPLEMENTED;		\
> >+		result.v0 = 0;						\
> >+		result.v1 = 0;						\
> >+	} while (0)
> 
> I have more than a little problem with submitting patches like this.  We 
> have no way to judge the suitability of the coding or the interface with 
> the "meat" of the driver stubbed out!

This is also code for hardware that does not exist.  In order for the
code to be in distro releases in time for the hardware to ship, we
must push it _before_ we have hardware.  The main concern is getting in
interfaces now, because the interfaces cannot change in a minor release.
Providing more "meat" is on the ToDo list.  

-- 
Russ Anderson, OS RAS/Partitioning Project Lead  
SGI - Silicon Graphics Inc          rja@sgi.com