[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <AANLkTiki4wk9dkd2J93FKJTOdcWUrXSLNv0KPR1gMEYk@mail.gmail.com>
Date: Tue, 18 Jan 2011 18:49:29 +0900
From: Akinobu Mita <akinobu.mita@...il.com>
To: Linus Torvalds <torvalds@...ux-foundation.org>,
Akinobu Mita <akinobu.mita@...il.com>,
linux-kernel@...r.kernel.org, linux-arch@...r.kernel.org,
akpm@...ux-foundation.org, Arnd Bergmann <arnd@...db.de>
Subject: Re: [PATCH v4 00/24] Introduce little endian bitops
2011/1/17 Russell King <rmk@....linux.org.uk>:
> The ARM bitops (set_bit/clear_bit/change_bit) have always taken an
> unsigned long argument and we have casts in our preprocessor macros
> for them. Only a couple of the find_bit functions have taken a
> void pointer.
>
> I really don't want to have to change the function prototypes on ARM,
> and I really don't want to hide this fact from non-fs users that ARM
> bitops require such pointers, with the exception of what's required
> for ext2/minix. If we do hide it, at some point we'll have someone
> believing that ARM's wrong to be requiring stricter alignment.
I have updated in my local branch and attached the diff between v4 and
updated one. It only changes the function prototypes for le bitops.
The function prototypes for the other bitops are not changed.
Assuming that ext2_*_bit() bitops is renamed to *_bit_le() bitops
by this patch series, it doesn't make the things worse.
(i.e. both bitops hide the fact that require long aligned pointers)
What do you think?
View attachment "v4-fix.patch" of type "text/x-patch" (23958 bytes)
Powered by blists - more mailing lists