From mboxrd@z Thu Jan  1 00:00:00 1970
From: Christoph Hellwig <hch@lst.de>
Date: Fri, 15 May 2020 17:24:59 +0200
Subject: [Ocfs2-devel] [PATCH 27/33] sctp: export sctp_setsockopt_bindx
In-Reply-To: <129070.1589556002@warthog.procyon.org.uk>
References: <20200514062820.GC8564@lst.de>
	<20200513062649.2100053-1-hch@lst.de>
	<20200513062649.2100053-28-hch@lst.de>
	<20200513180058.GB2491@localhost.localdomain>
	<129070.1589556002@warthog.procyon.org.uk>
Message-ID: <20200515152459.GA28995@lst.de>
List-Id: <ocfs2-devel.oss.oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: David Howells <dhowells@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>, Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>, Eric Dumazet <edumazet@google.com>, linux-nvme@lists.infradead.org, linux-sctp@vger.kernel.org, target-devel@vger.kernel.org, linux-afs@lists.infradead.org, drbd-dev@lists.linbit.com, linux-cifs@vger.kernel.org, rds-devel@oss.oracle.com, linux-rdma@vger.kernel.org, cluster-devel@redhat.com, Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>, linux-block@vger.kernel.org, Jakub Kicinski <kuba@kernel.org>, ceph-devel@vger.kernel.org, linux-nfs@vger.kernel.org, Neil Horman <nhorman@tuxdriver.com>, Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>, netdev@vger.kernel.org, Vlad Yasevich <vyasevich@gmail.com>, linux-kernel@vger.kernel.org, Jon Maloy <jmaloy@redhat.com>, Ying Xue <ying.xue@windriver.com>

On Fri, May 15, 2020 at 04:20:02PM +0100, David Howells wrote:
> Christoph Hellwig <hch@lst.de> wrote:
> 
> > > The advantage on using kernel_setsockopt here is that sctp module will
> > > only be loaded if dlm actually creates a SCTP socket.  With this
> > > change, sctp will be loaded on setups that may not be actually using
> > > it. It's a quite big module and might expose the system.
> > 
> > True.  Not that the intent is to kill kernel space callers of setsockopt,
> > as I plan to remove the set_fs address space override used for it.
> 
> For getsockopt, does it make sense to have the core kernel load optval/optlen
> into a buffer before calling the protocol driver?  Then the driver need not
> see the userspace pointer at all.
> 
> Similar could be done for setsockopt - allocate a buffer of the size requested
> by the user inside the kernel and pass it into the driver, then copy the data
> back afterwards.

I did look into that initially.  The problem is that tons of sockopts
entirely ignore optlen and just use a fixed size.  So I fear that there
could be tons of breakage if we suddently respect it.  Otherwise that
would be a pretty nice way to handle the situation.