Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 13 Jan 2021 19:39:36 -0400
From:   Jason Gunthorpe <jgg@ziepe.ca>
To:     Jerome Glisse <jglisse@redhat.com>
Cc:     Andrea Arcangeli <aarcange@redhat.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, Yu Zhao <yuzhao@google.com>,
        Andy Lutomirski <luto@kernel.org>,
        Peter Xu <peterx@redhat.com>,
        Pavel Emelyanov <xemul@openvz.org>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Mike Rapoport <rppt@linux.vnet.ibm.com>,
        Minchan Kim <minchan@kernel.org>,
        Will Deacon <will@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Hugh Dickins <hughd@google.com>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Matthew Wilcox <willy@infradead.org>,
        Oleg Nesterov <oleg@redhat.com>, Jann Horn <jannh@google.com>,
        Kees Cook <keescook@chromium.org>,
        John Hubbard <jhubbard@nvidia.com>,
        Leon Romanovsky <leonro@nvidia.com>, Jan Kara <jack@suse.cz>,
        Kirill Tkhai <ktkhai@virtuozzo.com>
Subject: Re: [PATCH 0/2] page_count can't be used to decide when wp_page_copy
Message-ID: <20210113233936.GE4605@ziepe.ca>
References: <B1B85771-B211-4FCC-AEEF-BDFD37332C25@vmware.com>
 <20210107200402.31095-1-aarcange@redhat.com>
 <20210107202525.GD504133@ziepe.ca>
 <X/eA/f1r5GXvcRWH@redhat.com>
 <20210108133649.GE504133@ziepe.ca>
 <X/iPtCktcQHwuK5T@redhat.com>
 <20210108181945.GF504133@ziepe.ca>
 <X/jgLGPgPb+Xms1t@redhat.com>
 <20210109004255.GG504133@ziepe.ca>
 <20210113215638.GA528828@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210113215638.GA528828@redhat.com>
Precedence: bulk

On Wed, Jan 13, 2021 at 04:56:38PM -0500, Jerome Glisse wrote:

> is a broken model and the way GPU use GUP is less broken then RDMA. In
> GPU driver GUP contract with userspace is that the data the GPU can
> access is a snapshot of what the process memory was at the time you
> asked for the GUP. Process can start using different pages right after.
> There is no constant coherency contract (ie CPU and GPU can be working
> on different pages).

Look at the habana labs "totally not a GPU" driver, it doesn't work
that way, GPU compute operations do want coherency.

The mmu notifier hackery some of the other GPU drivers use to get
coherency requires putting the kernel between every single work
submission, and has all kinds of wonky issues and limitations - I
think it is net worse approach than GUP, honestly.

Jason