Re: [PATCH v7 1/1] NFSD: Enforce timeout on layout recall and integrate lease manager fencing

[Chuck Lever <cel@kernel.org>]

On 2/9/26 2:24 PM, Dai Ngo wrote:
>>> diff --git a/fs/nfsd/nfs4layouts.c b/fs/nfsd/nfs4layouts.c
>>> index ad7af8cfcf1f..c02b3219ebeb 100644
>>> --- a/fs/nfsd/nfs4layouts.c
>>> +++ b/fs/nfsd/nfs4layouts.c
>>> @@ -27,6 +27,25 @@ static struct kmem_cache *nfs4_layout_stateid_cache;
>>>   static const struct nfsd4_callback_ops nfsd4_cb_layout_ops;
>>>   static const struct lease_manager_operations nfsd4_layouts_lm_ops;
>>>
>>> +/*
>>> + * By default, if the server fails to fence a client, it retries the
>>> fencing
>>> + * operation indefinitely to prevent data corruption. The admin
>>> needs to take
>>> + * the following actions to restore access to the file for other
>>> clients:
>>> + *
>>> + *    . shutdown or power off the client being fenced.
>>> + *    . manually expire the client to release all its state on the
>>> server;
>>> + *      echo 'expire' > proc/fs/nfsd/clients/clientid/ctl'.
>> Has there been any testing that shows expiring that client actually
>> breaks the fence retry loop below?
> 
> nfsd4_revoke_states calls nfsd4_close_layout to remove all file leases.
> I manually tested it.

Excellent!

-- 
Chuck Lever