Re: [PATCH 29/35] Btrfs ioctl code

[Chris Mason <chris.mason@oracle.com>]

On Thu, 2009-01-15 at 12:17 -0800, Linus Torvalds wrote:
> 
> On Fri, 16 Jan 2009, Ryusuke Konishi wrote:
> 
> > Hi Chris,
> > On Wed,  7 Jan 2009 22:57:19 -0500, Chris Mason wrote:
> > > +
> > > +struct btrfs_ioctl_vol_args {
> > > +	__s64 fd;
> > > +	char name[BTRFS_PATH_NAME_MAX + 1];
> > > +};
> > 
> > The size of btrfs_ioctl_vol_args seems to depend on architectures.
> > 
> > For example, the size was 3084 bytes on a x86-32 machine, but was 3088
> > bytes on a x86-64 machine.
> 
> Grr. Different alignment for 64-bit values, coupled with "name" not 
> having a size that is a multiple-of-eight.
> 
> Chris, I think we want to fix this.  Even if it breaks binary 
> compatibility with previous tools. I don't think we want to have compat 
> ioctls for something that is so close to already working.

I've fixed both the kernel and the progs to use a 4k struct, and tested
all the btrfs_ioctl_var_arg ioctls on x86-64 with a program compiled on
x86-32.

Please pull from:

git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable.git

And I'll roll out v0.18 of btrfs-progs with a notice about the ioctl
interface change.

Ryusuke, thank you very much for finding this now.

-chris