Re: [CFT] [JANITORIAL] Unbork fs.h

[Daniel Phillips <phillips@bonn-fries.net>]

On January 3, 2002 05:47 pm, Alexander Viro wrote:
> On Thu, 3 Jan 2002, Daniel Phillips wrote:
> > On January 3, 2002 04:45 pm, Christoph Hellwig wrote:
> > > In article <E16M7Gz-00015E-00@starship.berlin> you wrote:
> > > > -	inode = get_empty_inode();
> > > > +	inode = get_empty_inode(sb);
> > > 
> > > How about killing get_empty_inode completly and using new_inode() instead?
> > > There should be no regularly allocated inode without a superblock.
> > 
> > There are: sock_alloc rd_load_image.  However that's a nit because the new, 
> > improved get_empty_inode understands the concept of null sb.  (Another thing 
> > we could do is require every inode to have a superblock - that's probably 
> > where it will go in time.)
> 
> It's _already_ there.  RTFS, please - sock_alloc() creates inodes with
> sockfs superblock in ->i_sb and rd_load_image() just does normal open()
> for device nodes on rootfs.

Sockfs - yes, you'll see my patch already does it correctly, I was getting
a little tired when writing the previous reply... rd_load_image in 2.4.17 is
still using kdev_t there, however, no super_block anywhere to be seen.  I'll
track that change if it happens before I bring the patches forward to 2.5.

--
Daniel