linux-mips
[Top] [All Lists]

Re: [PATCH] I/O helpers rework

To: Herbert Valerio Riedel <hvr@inso.tuwien.ac.at>
Subject: Re: [PATCH] I/O helpers rework
From: "Maciej W. Rozycki" <macro@mips.com>
Date: Fri, 14 Jan 2005 20:07:56 +0000 (GMT)
Cc: Ralf Baechle <ralf@linux-mips.org>, linux-mips@linux-mips.org, "Maciej W. Rozycki" <macro@linux-mips.org>
In-reply-to: <1105368598.21670.7.camel@shswe.inso.tuwien.ac.at>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <Pine.LNX.4.61.0412151936460.14855@perivale.mips.com> <1105029224.4361.21.camel@xterm.intra> <Pine.LNX.4.61.0501101249280.18023@perivale.mips.com> <1105368380.21670.4.camel@shswe.inso.tuwien.ac.at> <1105368598.21670.7.camel@shswe.inso.tuwien.ac.at>
Sender: linux-mips-bounce@linux-mips.org
On Mon, 10 Jan 2005, Herbert Valerio Riedel wrote:

> On Mon, 2005-01-10 at 15:46 +0100, Herbert Valerio Riedel wrote:
> > so this means, the mtd subsystem should use them and that the patch
> > below is the way to fix it? (hoping it won't brake on other systems?)
> 
> sorry, wrong patch... the patch below should have s/__raw_/bus_/g
> instead of s/__raw_//g;

 Probably, although <linux/mtd/map.h> seems device-independent and that 
may mean you might want to take the endianness of the bus an MTD is 
connected to, to preserve data consistency when moving the MTD between 
systems of different endiannesses (it may be non-volatile memory of some 
sort).

> anyway, are the bus_# memory accessors defined for all archs at all?

 Not at all, I'm afraid, and fixing that might be a good idea.
Unfortunately most platforms use a single endianness only and therefore
have a biased view on these accessory functions.  Especially little-endian
folks may have troubles understanding what's this all about.

  Maciej

<Prev in Thread] Current Thread [Next in Thread>