linux-kernel - Re: [PATCH 3/8] nfsd: when CB_SEQUENCE gets NFS4ERR

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <19d4f8bd-b139-43bf-9ca6-716e59c9ee0d@talpey.com>
Date: Thu, 23 Jan 2025 20:30:16 -0500
From: Tom Talpey <tom@...pey.com>
To: Jeff Layton <jlayton@...nel.org>, Chuck Lever <chuck.lever@...cle.com>,
 Neil Brown <neilb@...e.de>, Olga Kornievskaia <okorniev@...hat.com>,
 Dai Ngo <Dai.Ngo@...cle.com>, "J. Bruce Fields" <bfields@...ldses.org>,
 Kinglong Mee <kinglongmee@...il.com>, Trond Myklebust <trondmy@...nel.org>,
 Anna Schumaker <anna@...nel.org>, "David S. Miller" <davem@...emloft.net>,
 Eric Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>,
 Paolo Abeni <pabeni@...hat.com>, Simon Horman <horms@...nel.org>
Cc: linux-nfs@...r.kernel.org, linux-kernel@...r.kernel.org,
 netdev@...r.kernel.org
Subject: Re: [PATCH 3/8] nfsd: when CB_SEQUENCE gets NFS4ERR_DELAY, release
 the slot

On 1/23/2025 6:20 PM, Jeff Layton wrote:
> On Thu, 2025-01-23 at 17:18 -0500, Chuck Lever wrote:
>> On 1/23/25 3:25 PM, Jeff Layton wrote:
>>> RFC8881, 15.1.1.3 says this about NFS4ERR_DELAY:
>>>
>>> "For any of a number of reasons, the replier could not process this
>>>    operation in what was deemed a reasonable time. The client should wait
>>>    and then try the request with a new slot and sequence value."
>>
>> A little farther down, Section 15.1.1.3 says this:
>>
>> "If NFS4ERR_DELAY is returned on a SEQUENCE operation, the request is
>>    retried in full with the SEQUENCE operation containing the same slot
>>    and sequence values."
>>
>> And:
>>
>> "If NFS4ERR_DELAY is returned on an operation other than the first in
>>    the request, the request when retried MUST contain a SEQUENCE operation
>>    that is different than the original one, with either the slot ID or the
>>    sequence value different from that in the original request."
>>
>> My impression is that the slot needs to be held and used again only if
>> the server responded with NFS4ERR_DELAY on the SEQUENCE operation. If
>> the NFS4ERR_DELAY was the status of the 2nd or later operation in the
>> COMPOUND, then yes, a different slot, or the same slot with a bumped
>> sequence number, must be used.
>>
>> The current code in nfsd4_cb_sequence_done() appears to be correct in
>> this regard.
>>
> 
> Ok! I stand corrected. We should be able to just drop this patch, but
> some of the later patches may need some trivial merge conflicts fixed
> up.
> 
> Any idea why SEQUENCE is different in this regard? This rule seems a
> bit arbitrary. If the response is NFS4ERR_DELAY, then why would it
> matter which slot you use when retransmitting? The responder is just
> saying "go away and come back later".

SEQUENCE is special because of the minor versioning rules. It is
prepended to COMPOUNDs in a way that allows it to be optional in
order to comply with the "no new required operations" rule.

As such, it's not handled quite the same as other compounded ops,
and has some exceptional handling. It's not beautiful, honestly.
But it allowed the session to be introduced in v4.1.

> What if the responder repeatedly returns NFS4ERR_DELAY (maybe because
> it's under resource pressure), and also shrinks the slot table in the
> meantime? It seems like that might put the requestor in an untenable
> position.
> 
> Maybe we should lobby to get this changed in the spec?

If there's a defect, absolutely. There's an IETF errata process.

Tom.

>>
>>> This is CB_SEQUENCE, but I believe the same rule applies. Release the
>>> slot before submitting the delayed RPC.
>>>
>>> Fixes: 7ba6cad6c88f ("nfsd: New helper nfsd4_cb_sequence_done() for processing more cb errors")
>>> Signed-off-by: Jeff Layton <jlayton@...nel.org>
>>> ---
>>>    fs/nfsd/nfs4callback.c | 1 +
>>>    1 file changed, 1 insertion(+)
>>>
>>> diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
>>> index bfc9de1fcb67b4f05ed2f7a28038cd8290809c17..c26ccb9485b95499fc908833a384d741e966a8db 100644
>>> --- a/fs/nfsd/nfs4callback.c
>>> +++ b/fs/nfsd/nfs4callback.c
>>> @@ -1392,6 +1392,7 @@ static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback
>>>    		goto need_restart;
>>>    	case -NFS4ERR_DELAY:
>>>    		cb->cb_seq_status = 1;
>>> +		nfsd41_cb_release_slot(cb);
>>>    		if (!rpc_restart_call(task))
>>>    			goto out;
>>>    
>>>
>>
>>
>