On Sat, Jul 23, 2011 at 7:46 AM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> Checking DCACHE_OP_PERMISSION on _what_? inode_permission() gets inode,
> not dentry... Now, we could mirror those into struct inode itself (and
> that might make more sense, actually), but I'd rather not go through
> the equivalent of d_set_d_op() patchset for i_op,
I thought about that, and indeed could not see a good place to hook
into. However, we do have "only" 317 places that do
inode->i_op = &..myops..
and an alternative would be to simply do an automated replacement with
set_inode_ops(inode, &..myops..);
because the scripting wouldn't be all that bad.
Then, if we just make sure that all the inode fields we touch are in
the same cacheline, we should be good. I think it's mainly i_mode,
i_uid and the acl cache entries. And there is space for some bits in
there, umode_t is just 16 bits, so we could easily put another 16-bit
"inode op flags" in next to it.
So that would be fairly straightforward.
Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
