Re: [PATCH] Bind Mount Extensions 0.04.1 3/5

[Herbert Poetzl <herbert@13thfloor.at>]

On Mon, Mar 15, 2004 at 02:10:04PM -0800, Andrew Morton wrote:
> Herbert Poetzl <herbert@13thfloor.at> wrote:
> >
> > --- linux-2.6.4-20040314_2308/fs/ext2/xattr.c	2004-03-11 03:55:34.000000000 +0100
> > +++ linux-2.6.4-20040314_2308-bme0.04.1/fs/ext2/xattr.c	2004-03-15 06:27:13.000000000 +0100
> > @@ -496,7 +496,7 @@ ext2_xattr_set(struct inode *inode, int 
> >  	ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
> >  		  name_index, name, value, (long)value_len);
> >  
> > -	if (IS_RDONLY(inode))
> > +	if (IS_RDONLY_INODE(inode))
> 
> I do think that if we're going to do this thing it should have 100%
> coverage, and not have little exceptions because the volume of changes got
> too high.
> 
> The number of places where you need IS_RDONLY_INODE() are encouragingly
> small.  It appears that all we need to do to get rid of it is to propagate
> the file* down through the ext2 and ext3 xattr code.  A NULL value will
> need to be permitted because ext2_new_inode doesn't have a file*, and we've
> already performed the check.

I'm not sure we actually want to do this, but I'll
try ... seems to go up all the way to the syscall entry.

> Sure, it's a largeish patch but it is very safe: if it compiles, it works.

agreed.

> Could you please work that with Andreas?

sure, Andreas who?

> IS_RDONLY_INODE() is also used in intermezzo, but that doesn't compile any
> more anyway.

so I do not bother with that, but what about the nfs(d)
stuff? 

TIA,
Herbert