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:   Fri, 23 Sep 2022 01:39:23 -0700
From:   Christoph Hellwig <hch@infradead.org>
To:     John Hubbard <jhubbard@nvidia.com>
Cc:     Al Viro <viro@zeniv.linux.org.uk>, Jan Kara <jack@suse.cz>,
        Christoph Hellwig <hch@infradead.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Jens Axboe <axboe@kernel.dk>,
        Miklos Szeredi <miklos@szeredi.hu>,
        "Darrick J . Wong" <djwong@kernel.org>,
        Trond Myklebust <trond.myklebust@hammerspace.com>,
        Anna Schumaker <anna@kernel.org>,
        David Hildenbrand <david@redhat.com>,
        Logan Gunthorpe <logang@deltatee.com>,
        linux-block@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-xfs@vger.kernel.org, linux-nfs@vger.kernel.org,
        linux-mm@kvack.org, LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 4/7] iov_iter: new iov_iter_pin_pages*() routines
Message-ID: <Yy1wu9tKo/sbsi1N@infradead.org>
References: <YxhaJktqtHw3QTSG@infradead.org>
 <YyFPtTtxYozCuXvu@ZenIV>
 <20220914145233.cyeljaku4egeu4x2@quack3>
 <YyIEgD8ksSZTsUdJ@ZenIV>
 <20220915081625.6a72nza6yq4l5etp@quack3>
 <YyvG+Oih2A37Grcf@ZenIV>
 <a6f95605-c2d5-6ec5-b85c-d1f3f8664646@nvidia.com>
 <20220922112935.pep45vfqfw5766gq@quack3>
 <Yy0lztxfwfGXFme4@ZenIV>
 <7e652ba4-8b03-59e0-a9ef-1118c4bbd492@nvidia.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <7e652ba4-8b03-59e0-a9ef-1118c4bbd492@nvidia.com>
Precedence: bulk

On Thu, Sep 22, 2022 at 09:05:16PM -0700, John Hubbard wrote:
> I certainly hope not. And in fact, we should really just say that that's
> a rule: the whole time the page is pinned, it simply must remain dirty
> and writable, at least with the way things are right now.

Yes, if we can stick to that rule and make sure shared pagecache is
never dirtied through get_user_pags anywhere that will allow us to
fix a lot of mess

> To fix those cases, IIUC, the answer is: you must make the page dirty
> properly, with page_mkwrite(), not just with set_page_dirty_lock(). And
> that has to be done probably a lot earlier, for reasons that I'm still
> vague on. But perhaps right after pinning the page. (Assuming that we
> hold off writeback while the page is pinned.)

I think we need to hold off the writeback for it to work properly.
The big question is, is if there are callers that do expect data
to be written back on mappings that are longterm pinned.  RDMA or
vfio would come to mind.