Received-SPF: pass (google.com: domain of linux-nfs+bounces-3344-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-ID: <81fa32fb-c5a3-4af8-a8ca-08905a8b62ef@grimberg.me>
Date: Thu, 23 May 2024 00:19:39 +0300
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH rfc] nfs: propagate readlink errors in nfs_symlink_filler
To: Trond Myklebust <trondmy@hammerspace.com>,
 "linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>,
 "jlayton@kernel.org" <jlayton@kernel.org>
Cc: "hch@lst.de" <hch@lst.de>, "dan.aloni@vastdata.com"
 <dan.aloni@vastdata.com>, "chuck.lever@oracle.com" <chuck.lever@oracle.com>
References: <20240521125840.186618-1-sagi@grimberg.me>
 <fa1a77fee6403454444fffce839924157778df95.camel@kernel.org>
 <ac2bfa20-d952-4917-8a70-1e821f9b57ca@grimberg.me>
 <d5409ff9ce51e439f442abb1cded7c7ab732b726.camel@hammerspace.com>
 <4d2bc7f1-b5c2-469c-9351-772626c707d7@grimberg.me>
 <c1d98e76-1b5c-4d91-a7fe-9412df7c2fab@grimberg.me>
 <e4e181e4a7b2db4b27b6ce3e6bb26b23e514cdb1.camel@hammerspace.com>
Content-Language: en-US
From: Sagi Grimberg <sagi@grimberg.me>
In-Reply-To: <e4e181e4a7b2db4b27b6ce3e6bb26b23e514cdb1.camel@hammerspace.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit


>> So what do you suggest we do here? IMO at a minimum NFS should retry
>> once similar
>> to nfs4_file_open (it would probably address 99.9% of the use-cases
>> where symlinks are
>> not overwritten in a high enough frequency for the client to see 2
>> consecutive stale readlink
>> rpc rplies).
>>
>> I can send a patch paired with a vfs ESTALE conversion patch?
>> alternatively retry locally in NFS...
>> I would like to understand your position here.
>>
> Looking more closely at nfs_get_link(), it is obvious that it can
> already return ESTALE (thanks to the call to nfs_revalidate_mapping())
> and looking at do_readlinkat(), it has already been plumbed through
> with a call to retry_estale().
>
> So I think we can take your patch as is, since it doesn't add any error
> cases that callers of readlink() don't have to handle already.

Sounds good.

>
> We might still want to think about cleaning up the output of the VFS in
> all these cases, so that we don't return ESTALE when it isn't allowed
> by POSIX, but that would be a separate task.
>

Yes, I can follow up on that later...