Re: [PATCH] make last_inode counter in new_inode 32-bit on kernels that offer x86 compatability

[Jeff Layton <jlayton@redhat.com>]

On Tue, 2006-11-07 at 14:20 -0700, Matthew Wilcox wrote:
> On Tue, Nov 07, 2006 at 04:13:00PM -0500, Jeff Layton wrote:
> > +	/* ino must not collide with any ino assigned in the loop below. Set
> > +	   it to the highest possible inode number */
> > +	inode->i_ino = (1 << (sizeof(s->s_lastino) * 8)) - 1;
> 
> This really isn't a good idiom to be using; GCC now takes this to mean
> "I can reformat your hard drive because you did something outside the
> spec".
> 
> Try instead:
> +	inode->i_ino = -1;
> 

The problem there is that on platforms with a 64-bit ino_t, this will be
too large to fit in a 32-bit field and we'll end up with the same
EOVERFLOW problem. Is there a more correct way to make it size
appropriately given the different possible sizes of s_lastino?

I suppose we could just set it to 0xffffffff and hope that that is "big
enough" for most cases.

-- Jeff