Re: [PATCH] r/o bind mounts: add vfsmount and superblock writer counts

[Christoph Hellwig]

On Mon, Jan 08, 2007 at 10:34:45AM -0800, Dave Hansen wrote:
> Removing it had a few side-effects.  First of all, it made me move
> all of the operations on the counts of writers to underneath the
> spinlock that was already there.  I guess this could be a cause
> for potential contention because there _are_ locks in the common
> code paths now.  But, I do agree with Christoph that it would be
> awfully hard to get it contended.
> 
> The other side-effect is that we can't have the bit in mnt_flags
> to be a shortcut to the superblock's writeable state since we
> don't have a way to go find the mounts and that bit when a fs
> changes writeable state.  This causes a potential cache miss
> when we have to check the superblock directly during the
> relatively common __mnt_is_readonly() function.

Why? We _only_ need to check the vfsmount flag.  vfsmount can
become r/w if the superblock is marked r/o which means the
underlying (block/network/etc) device is fundamentally not writeable.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/