From patchwork Wed May 4 09:00:03 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Marek Szyprowski X-Patchwork-Id: 67110 Delivered-To: patch@linaro.org Received: by 10.140.92.199 with SMTP id b65csp63428qge; Wed, 4 May 2016 02:00:21 -0700 (PDT) X-Received: by 10.67.6.3 with SMTP id cq3mr10480308pad.115.1462352421818; Wed, 04 May 2016 02:00:21 -0700 (PDT) Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id k70si3750254pfk.75.2016.05.04.02.00.21; Wed, 04 May 2016 02:00:21 -0700 (PDT) Received-SPF: pass (google.com: best guess record for domain of linux-media-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; spf=pass (google.com: best guess record for domain of linux-media-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-media-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757881AbcEDJAQ (ORCPT + 4 others); Wed, 4 May 2016 05:00:16 -0400 Received: from mailout2.w1.samsung.com ([210.118.77.12]:21529 "EHLO mailout2.w1.samsung.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757878AbcEDJAM (ORCPT ); Wed, 4 May 2016 05:00:12 -0400 Received: from eucpsbgm1.samsung.com (unknown [203.254.199.244]) by mailout2.w1.samsung.com (Oracle Communications Messaging Server 7.0.5.31.0 64bit (built May 5 2014)) with ESMTP id <0O6N00JYKAC95V40@mailout2.w1.samsung.com>; Wed, 04 May 2016 10:00:09 +0100 (BST) X-AuditID: cbfec7f4-f796c6d000001486-14-5729ba1812a1 Received: from eusync3.samsung.com ( [203.254.199.213]) by eucpsbgm1.samsung.com (EUCPMTA) with SMTP id 12.CD.05254.81AB9275; Wed, 4 May 2016 10:00:08 +0100 (BST) Received: from amdc1339.digital.local ([106.116.147.30]) by eusync3.samsung.com (Oracle Communications Messaging Server 7.0.5.31.0 64bit (built May 5 2014)) with ESMTPA id <0O6N0078LAC5HZ80@eusync3.samsung.com>; Wed, 04 May 2016 10:00:08 +0100 (BST) From: Marek Szyprowski To: linux-media@vger.kernel.org, linux-samsung-soc@vger.kernel.org Cc: Marek Szyprowski , Sylwester Nawrocki , Laurent Pinchart , Krzysztof Kozlowski , Bartlomiej Zolnierkiewicz , Hans Verkuil , Sakari Ailus Subject: [PATCH v4] media: vb2-dma-contig: configure DMA max segment size properly Date: Wed, 04 May 2016 11:00:03 +0200 Message-id: <1462352403-27418-1-git-send-email-m.szyprowski@samsung.com> X-Mailer: git-send-email 1.9.2 In-reply-to: <5729B396.1020706@xs4all.nl> References: <5729B396.1020706@xs4all.nl> X-Brightmail-Tracker: H4sIAAAAAAAAA+NgFprNLMWRmVeSWpSXmKPExsVy+t/xq7oSuzTDDeZ81rHYOGM9q8WSn7uY LF6/MLTonLiE3aJnw1ZWixnn9zFZrD1yl93i8Jt2VotPW74xOXB6TPm9kdVjdsdMVo95JwM9 +rasYvT4vEkugDWKyyYlNSezLLVI3y6BK6Nt1hTmgoWaFdOfvmNuYHyi2MXIySEhYCKxYFIT M4QtJnHh3nq2LkYuDiGBpYwSfdu2MEM4TUwSZ+6/ZwSpYhMwlOh628UGYosIOEksnPWXHcRm FrjOJHG5M7aLkYNDWCBEon9mCkiYRUBVYs2Fl6wgNq+Ah8TXf9dYIZbJSfx/uYIJxOYU0JR4 1PGbBcQWEtCQOL/qCuMERt4FjAyrGEVTS5MLipPScw31ihNzi0vz0vWS83M3MUKC7MsOxsXH rA4xCnAwKvHwvvDWDBdiTSwrrsw9xCjBwawkwiu8DSjEm5JYWZValB9fVJqTWnyIUZqDRUmc d+6u9yFCAumJJanZqakFqUUwWSYOTqkGxtZJXAuVms4/Sc69EL5CQ4JnxsLt8toOxzp2FXyZ djdGa4X1bqb5HscZfF/wJtS4cp7pCZh5fe+Kz0cjrePd2TsE8o+kiB4+Jit56QPr/OALJ+e8 090uxPREYeG31FNTgd644heofMn039d7jp9+HfCe+OWtsGo5j56sotW6/7pmJ4XeMcrMUWIp zkg01GIuKk4EAC404MUuAgAA Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org This patch lets vb2-dma-contig memory allocator to configure DMA max segment size properly for the client device. Setting it is needed to let DMA-mapping subsystem to create a single, contiguous mapping in DMA address space. This is essential for all devices, which use dma-contig videobuf2 memory allocator and shared buffers (in USERPTR or DMAbuf modes of operations). Signed-off-by: Marek Szyprowski Acked-by: Sakari Ailus --- Hello, This patch is a follow-up of my previous attempts to let Exynos multimedia devices to work properly with shared buffers when IOMMU is enabled: 1. https://www.mail-archive.com/linux-media@vger.kernel.org/msg96946.html 2. http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/97316 3. https://patchwork.linuxtv.org/patch/30870/ As sugested by Hans, configuring DMA max segment size should be done by videobuf2-dma-contig module instead of requiring all device drivers to do it on their own. Here is some backgroud why this is done in videobuf2-dc not in the respective generic bus code: http://lists.infradead.org/pipermail/linux-arm-kernel/2014-November/305913.html Best regards, Marek Szyprowski changelog: v4: - rebased onto media master tree - call vb2_dc_set_max_seg_size after allocating vb2 buf object v3: - added FIXME note about possible memory leak v2: - fixes typos and other language issues in the comments v1: http://article.gmane.org/gmane.linux.kernel.samsung-soc/53690 --- drivers/media/v4l2-core/videobuf2-dma-contig.c | 53 +++++++++++++++++++++++++- 1 file changed, 51 insertions(+), 2 deletions(-) -- 1.9.2 -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html diff --git a/drivers/media/v4l2-core/videobuf2-dma-contig.c b/drivers/media/v4l2-core/videobuf2-dma-contig.c index 5361197f3e57..6291842a889f 100644 --- a/drivers/media/v4l2-core/videobuf2-dma-contig.c +++ b/drivers/media/v4l2-core/videobuf2-dma-contig.c @@ -448,6 +448,42 @@ static void vb2_dc_put_userptr(void *buf_priv) } /* + * To allow mapping the scatter-list into a single chunk in the DMA + * address space, the device is required to have the DMA max segment + * size parameter set to a value larger than the buffer size. Otherwise, + * the DMA-mapping subsystem will split the mapping into max segment + * size chunks. This function increases the DMA max segment size + * parameter to let DMA-mapping map a buffer as a single chunk in DMA + * address space. + * This code assumes that the DMA-mapping subsystem will merge all + * scatterlist segments if this is really possible (for example when + * an IOMMU is available and enabled). + * Ideally, this parameter should be set by the generic bus code, but it + * is left with the default 64KiB value due to historical litmiations in + * other subsystems (like limited USB host drivers) and there no good + * place to set it to the proper value. It is done here to avoid fixing + * all the vb2-dc client drivers. + * + * FIXME: the allocated dma_params structure is leaked because there + * is completely no way to determine when to free it (dma_params might have + * been also already allocated by the bus code). However in typical + * use cases this function will be called for platform devices, which are + * not hot-plugged and exist all the time in the target system. + */ +static int vb2_dc_set_max_seg_size(struct device *dev, unsigned int size) +{ + if (!dev->dma_parms) { + dev->dma_parms = kzalloc(sizeof(dev->dma_parms), GFP_KERNEL); + if (!dev->dma_parms) + return -ENOMEM; + } + if (dma_get_max_seg_size(dev) < size) + return dma_set_max_seg_size(dev, size); + + return 0; +} + +/* * For some kind of reserved memory there might be no struct page available, * so all that can be done to support such 'pages' is to try to convert * pfn to dma address or at the last resort just assume that @@ -509,6 +545,10 @@ static void *vb2_dc_get_userptr(void *alloc_ctx, unsigned long vaddr, if (!buf) return ERR_PTR(-ENOMEM); + ret = vb2_dc_set_max_seg_size(conf->dev, PAGE_ALIGN(size + PAGE_SIZE)); + if (!ret) + goto fail_buf; + buf->dev = conf->dev; buf->dma_dir = dma_dir; @@ -682,6 +722,7 @@ static void *vb2_dc_attach_dmabuf(void *alloc_ctx, struct dma_buf *dbuf, struct vb2_dc_conf *conf = alloc_ctx; struct vb2_dc_buf *buf; struct dma_buf_attachment *dba; + int ret; if (dbuf->size < size) return ERR_PTR(-EFAULT); @@ -690,13 +731,17 @@ static void *vb2_dc_attach_dmabuf(void *alloc_ctx, struct dma_buf *dbuf, if (!buf) return ERR_PTR(-ENOMEM); + ret = vb2_dc_set_max_seg_size(conf->dev, PAGE_ALIGN(size)); + if (!ret) + goto fail_buf; + buf->dev = conf->dev; /* create attachment for the dmabuf with the user device */ dba = dma_buf_attach(dbuf, buf->dev); if (IS_ERR(dba)) { pr_err("failed to attach dmabuf\n"); - kfree(buf); - return dba; + ret = PTR_ERR(dba); + goto fail_buf; } buf->dma_dir = dma_dir; @@ -704,6 +749,10 @@ static void *vb2_dc_attach_dmabuf(void *alloc_ctx, struct dma_buf *dbuf, buf->db_attach = dba; return buf; + +fail_buf: + kfree(buf); + return ERR_PTR(ret); } /*********************************************/