Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Sat, 27 Aug 2022 04:48:27 +0100
From:   Al Viro <viro@zeniv.linux.org.uk>
To:     NeilBrown <neilb@suse.de>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        Daire Byrne <daire@dneg.com>,
        Trond Myklebust <trond.myklebust@hammerspace.com>,
        Chuck Lever <chuck.lever@oracle.com>,
        Linux NFS Mailing List <linux-nfs@vger.kernel.org>,
        linux-fsdevel@vger.kernel.org, LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 03/10] VFS: move want_write checks into
 lookup_hash_update()
Message-ID: <YwmUC7UNvNFdLTlc@ZenIV>
References: <166147828344.25420.13834885828450967910.stgit@noble.brown>
 <166147984374.25420.3477094952897986387.stgit@noble.brown>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <166147984374.25420.3477094952897986387.stgit@noble.brown>
Sender: Al Viro <viro@ftp.linux.org.uk>
Precedence: bulk

On Fri, Aug 26, 2022 at 12:10:43PM +1000, NeilBrown wrote:
> mnt_want_write() is always called before lookup_hash_update(), so
> we can simplify the code by moving the call into that function.
> 
> If lookup_hash_update() succeeds, it now will have claimed the
> want_write lock.  If it fails, the want_write lock isn't held.
> 
> To allow this, lookup_hash_update() now receives a 'struct path *'
> instead of 'struct dentry *'
> 
> Note that when creating a name, any error from mnt_want_write() does not
> get reported unless there is no other error.  For unlink/rmdir though,
> an error from mnt_want_write() is immediately fatal - overriding ENOENT
> for example.  This behaviour seems strange, but this patch is careful to
> preserve it.

Would be a nice idea, if not for the mess with LOOKUP_SILLY_RENAME later
in the series...