Frank Filz

See this documentation for the Linux client: https://www.kernel.org/doc/html/latest/admin-guide/nfs/nfs-client.html#the-nfs4-unique-id-parameter However that doesn't solve the issue for containers (see the bottom of the section)... This Red Hat Bugzilla sheds some light on...

Adding wontfix to make it clear there will be no Ganesha code changes for this.

Look at this upstream kernel patch and discussion: https://lore.kernel.org/all/6e05bf321ff50785360e6c339d111db368e7dfda.1644349990.git.bcodding@redhat.com/ I don't know if that is heading towards a resolution for this issue.

This looks like an enhancement worth putting in the backlog.

Yes, READDIR would be a good candidate. GETATTR might also be a good candidate. Hmm, OPEN also (since it could block while a delegation is recalled). But as Daniel mentions,...

Yea, in general READDIR should be quick, but I could see scenarios where reading the first chunk is somewhat slow, especially if the readdir is done to some different node....

Note also that readdir is already halfway to async since all the content is returned via a callback. Probable changes to make fully async: * return eof in callback instead...

I'm going to flag this as an enhancement request at this point.

I've assigned this to you. Feel free to ask questions here or on our IRC channel (#ganesha on Libera.chat)

Simon, could you take a look at this one?