[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20220216172853.GI4160@nvidia.com>
Date: Wed, 16 Feb 2022 13:28:53 -0400
From: Jason Gunthorpe <jgg@...dia.com>
To: Felix Kuehling <felix.kuehling@....com>
Cc: Christoph Hellwig <hch@....de>,
David Hildenbrand <david@...hat.com>,
Alex Sierra <alex.sierra@....com>, akpm@...ux-foundation.org,
linux-mm@...ck.org, rcampbell@...dia.com,
linux-ext4@...r.kernel.org, linux-xfs@...r.kernel.org,
amd-gfx@...ts.freedesktop.org, dri-devel@...ts.freedesktop.org,
jglisse@...hat.com, apopple@...dia.com, willy@...radead.org
Subject: Re: [PATCH v6 01/10] mm: add zone device coherent type memory support
On Wed, Feb 16, 2022 at 11:56:51AM -0500, Felix Kuehling wrote:
> In the case of DEVICE_COHERENT memory, the pfns correspond to real physical
> memory addresses. I don't think they have those PFN_DEV|PFN_MAP bits set.
So do DAX pages. The PTE flag does several things. As this would be
the first time ZONE_DEVICE pages do not set devmap it needs a full
audit.
eg the gup_fast bug Alistair pointed at needs fixing at least.
Jason
Powered by blists - more mailing lists