On Mon, Apr 12, 2021 at 10:55 AM Christoph Hellwig <[email protected]> wrote:
>
> Hi all,
>
> currently we deal with the slight differents in the various architecture
> variants of the flock and flock64 stuctures in a very cruft way. This
> series switches to just use small arch hooks and define the rest in
> asm-generic and linux/compat.h instead.
Nice cleanup. I can merge it through the asm-generic tree if you like,
though it's a little late just ahead of the merge window.
I would not want to change the compat_loff_t definition to compat_s64
to avoid the padding at this time, though that might be a useful cleanup
for a future cycle.
Arnd
From: Arnd Bergmann
> Sent: 12 April 2021 11:04
>
> On Mon, Apr 12, 2021 at 10:55 AM Christoph Hellwig <[email protected]> wrote:
> >
> > Hi all,
> >
> > currently we deal with the slight differents in the various architecture
> > variants of the flock and flock64 stuctures in a very cruft way. This
> > series switches to just use small arch hooks and define the rest in
> > asm-generic and linux/compat.h instead.
>
> Nice cleanup. I can merge it through the asm-generic tree if you like,
> though it's a little late just ahead of the merge window.
>
> I would not want to change the compat_loff_t definition to compat_s64
> to avoid the padding at this time, though that might be a useful cleanup
> for a future cycle.
Is x86 the only architecture that has 32bit and 64bit variants where
the 32bit variant aligns 64bit items on 32bit boundaries?
I've just checked MIPS and ARM, and I'm fairly sure sparc 64bit
aligns them.
Are there any others?
Might also be interesting to check whether compat_loff_t gets
used anywhere else - where the x64-64 compat code will get it
wrong.
ISTM that fixing compat_loff_t shouldn't have any fallout.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)