On Thu, Apr 27, 2017 at 10:48 AM, Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
> On Thu, Apr 27, 2017 at 8:27 AM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>
>> Let me try to explain the problem with a worse case, but not
>> improbable example:
>>
>> Suppose I have an overlay with deep file at /a/b/c/.../z
>> Suppose the layers are at /old/{lower,upper} I copy them
>> over to /new/{lower,upper} and mount the overlay at new path.
>>
>> Suppose that dcache is fully populated under /new and fully
>> evicted under /old.
>>
>> When trying to decode the file handle for z, exportfs_decode_fh()
>> will call the file system to actually read all directories a..z from disk
>> in order to reconnect the dentry of old z all the way up to /old
>> and it will do that *before* calling the acceptable() callback.
>>
>> Alternatively, if we first try to decode the file handle for /old/lower,
>> decoding will be very fast (most likely already in cache) and we will
>> not have to continue to decoding z and reading all directories a..z
>> from disk.
>
> To answer my own question in the prev mail: we need to decode the fh
> and not just blindly use the inum to prevent issues with
> copied/mutilited/etc lower layers.
>
I was going to refer you to this example when reading you question
in prev email. That's what we get for no read/write barriers in emails ;-)
> And yes, in the copied case decoding origin.root first would be a good
> optimization that couldn't be done without it.
>
Good, so we seem to have an agreement w.r.t. the lookup fh patch.
I've already applied a change to disable redirect_fh if lower s_uuid is
zeros and I verified that it works as expected by running the hard-link
constant inode test that relies on redirect_fh over xfs mounted with
-o nouuid.
I will be posting the enhanced xfstest for constant inodes later today.
Let me know when are are done reviewing the series, so I can rework it
with the binary blob change you requested.
Thanks,
Amir.
