Re: [POC/RFC PATCH] overlayfs: constant inode numbers

[Miklos Szeredi]

On Tue, Nov 29, 2016 at 1:03 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
> On Tue, Nov 29, 2016 at 1:34 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>> On Tue, Nov 29, 2016 at 12:16 PM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
>>> On Mon, Nov 28, 2016 at 12:56 PM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:

>> Not sure that I understand what you are suggesting, but I would be happy
>> to make the needed adjustments to redirect_fh per your request if you clarify
>> what you mean. From what I understand:
>>
>> 1. If redirect_dir=fh (and supported by layers), store lower handle
>>     on dir copy up in new xattr OVL_XATTR_FH
>> 2. In ovl_rename(), set OVL_XATTR_REDIRECT regardless of OVL_XATTR_FH
>> 3. In ovl_lookup_single(), carry both d.redirct and d.redirect_fh to next layer
>> 4. In ovl_lookup_layer(), lookup by handle first then by path
>>
>> Mind you that unlike OVL_XATTR_INO, OVL_XATTR_FH points to *next* inode
>> rather then *lowest* inode, so it does not really improve anything wrt
>> getting the
>> stable inode. Stable inode of merged dir is available after lookup in
>> oe->lowerstack[oe->numlower - 1].dentry->d_inode->i_ino
>> regardless of whether lookup was by handle or by path or no redirect at all, so
>> not sure what you meant by "... and provide a good way to get the stable ino."
>> Either I managed to confuse you, or I am missing something?

I meant that we can unify OVL_XATTR_INO  with "redirect/fh"
functionality and get something good out of it.

> Perhaps you meant for non-dir:
>
> 5. If redirect_dir=fh, *propagate* lowest-handle on non-dir copy up
> 6. In ovl_lookup() of non-dir, decode lowest-handle to set oe->ino

Yes.

OVL_XATTR_FH would be safe to ignore, so this is back and forward
compatible..  And the cost is probably not prohitive, since copy ups
should be relatively rare.

After a backup + restore it is not expected that we get back the old
inode numbers so it's fine to ignore the stale file handles.

The following issues are left:

 - performance of readdir;
 - what to do if not all layers are on the same fs;
 - hard link copy ups.

Thanks,
Miklos