From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dmitry Osipenko Subject: Re: [PATCH v1 1/3] gpu: host1x: Remove implicit IOMMU backing on client's registration Date: Mon, 24 Jun 2019 15:55:04 +0300 Message-ID: References: <20190623173743.24088-1-digetx@gmail.com> <20190624070413.GA23846@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: <20190624070413.GA23846@infradead.org> Content-Language: en-US Sender: linux-kernel-owner@vger.kernel.org To: Christoph Hellwig Cc: Thierry Reding , dri-devel@lists.freedesktop.org, linux-tegra@vger.kernel.org, linux-kernel@vger.kernel.org List-Id: linux-tegra@vger.kernel.org 24.06.2019 10:04, Christoph Hellwig пишет: > Don't we have a device tree problem here if there is a domain covering > them? I though we should only pick up an IOMMU for a given device > if DT explicitly asked for that? > There is no specific domain that "covering them". The IOMMU domain is allocated dynamically during of the Tegra DRM's driver initialization (see tegra_drm_load) and then all of DRM devices are attached to that domain once all of the DRM sub-drivers are probed successfully. On Tegra SoCs it's up to software (driver) to decide how to separate hardware devices from each other, in a case of DRM we're putting all the relevant graphics devices into a single domain. Is it even possible to express IOMMU domain (not group!) assignments in a device-tree?