Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Sat, 14 Jan 2023 01:58:31 +0000
From:   Al Viro <viro@zeniv.linux.org.uk>
To:     "Martin K. Petersen" <martin.petersen@oracle.com>
Cc:     Bart Van Assche <bart.vanassche@gmail.com>,
        Christoph Hellwig <hch@infradead.org>,
        David Howells <dhowells@redhat.com>,
        Matthew Wilcox <willy@infradead.org>,
        Jens Axboe <axboe@kernel.dk>, Jan Kara <jack@suse.cz>,
        Jeff Layton <jlayton@kernel.org>,
        Logan Gunthorpe <logang@deltatee.com>,
        linux-fsdevel@vger.kernel.org, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        Douglas Gilbert <dgilbert@interlog.com>
Subject: Re: [PATCH v5 3/9] iov_iter: Use IOCB/IOMAP_WRITE if available
 rather than iterator direction
Message-ID: <Y8IMR8BK3bLnr9ps@ZenIV>
References: <Y7+8r1IYQS3sbbVz@infradead.org>
 <167344725490.2425628.13771289553670112965.stgit@warthog.procyon.org.uk>
 <167344727810.2425628.4715663653893036683.stgit@warthog.procyon.org.uk>
 <15330.1673519461@warthog.procyon.org.uk>
 <Y8AUTlRibL+pGDJN@infradead.org>
 <Y8BFVgdGYNQqK3sB@ZenIV>
 <c6f4014e-d199-d5e8-515c-5ffcd9946c80@gmail.com>
 <yq1ilh9ucg3.fsf@ca-mkp.ca.oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <yq1ilh9ucg3.fsf@ca-mkp.ca.oracle.com>
Sender: Al Viro <viro@ftp.linux.org.uk>
Precedence: bulk

On Fri, Jan 13, 2023 at 08:34:50PM -0500, Martin K. Petersen wrote:
> 
> Bart,
> 
> > I'm not sure that we still need the double copy in the sg driver. It
> > seems obscure to me that there is user space software that relies on
> > finding "0xec" in bytes not originating from a SCSI
> > device. Additionally, SCSI drivers that do not support residuals
> > should be something from the past.
> 
> Yeah. I'm not aware of anything that relies on this still. But obviously
> Doug has more experience in the app dependency department.

Are we guaranteed to know the accurate amount of data that got transferred
for all surviving drivers?  If we do, we can do accurate copy-out and all
apps will keep seeing what they currently do.  If we don't, the best we
can do is replacing copy-in + IO + copy-out with memset + IO + copy-out.