From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756089AbaKSPqv (ORCPT ); Wed, 19 Nov 2014 10:46:51 -0500 Received: from foss-mx-na.foss.arm.com ([217.140.108.86]:33634 "EHLO foss-mx-na.foss.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756065AbaKSPqu (ORCPT ); Wed, 19 Nov 2014 10:46:50 -0500 Date: Wed, 19 Nov 2014 15:46:35 +0000 From: Catalin Marinas To: Arnd Bergmann Cc: Ding Tianhong , "linux-arm-kernel@lists.infradead.org" , Will Deacon , "linux-kernel@vger.kernel.org" Subject: Re: For the problem when using swiotlb Message-ID: <20141119154634.GE7156@e104818-lin.cambridge.arm.com> References: <5469E26B.2010905@huawei.com> <1535751.CcvIi3DN4F@wuerfel> <20141119112910.GD7156@e104818-lin.cambridge.arm.com> <1497289.TPKf3qDiKk@wuerfel> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1497289.TPKf3qDiKk@wuerfel> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Nov 19, 2014 at 12:48:58PM +0000, Arnd Bergmann wrote: > On Wednesday 19 November 2014 11:29:10 Catalin Marinas wrote: > > > The driver should call 'dma_set_mask_and_coherent()' with the appropriate > > > dma mask, and check whether that succeeded. However, the code implementing > > > dma_set_mask_and_coherent on arm64 also needs to be changed to look up > > > the dma-ranges property (see of_dma_configure()), and check if the mask > > > is possible. > > > > dma_set_mask_and_coherent() is a generic function. I think the > > of_dma_configure() should start with a coherent_dma_mask based on > > dma-ranges if given rather than defaulting to DMA_BIT_MASK(32). The > > comment in of_dma_configure() says that devices should set up the > > supported mask but it's not always up to them but the bus they are > > connected to. > > > > Something like below, untested: > > > > diff --git a/drivers/of/platform.c b/drivers/of/platform.c > > index 3b64d0bf5bba..dff34883db45 100644 > > --- a/drivers/of/platform.c > > +++ b/drivers/of/platform.c > > @@ -200,6 +200,10 @@ static void of_dma_configure(struct device *dev) > > /* DMA ranges found. Calculate and set dma_pfn_offset */ > > dev->dma_pfn_offset = PFN_DOWN(paddr - dma_addr); > > dev_dbg(dev, "dma_pfn_offset(%#08lx)\n", dev->dma_pfn_offset); > > + > > + /* Set the coherent_dma_mask based on the dma-ranges property */ > > + if (size) > > + dev->coherent_dma_mask = DMA_BIT_MASK(ilog2(size)); > > } > > We have discussed this in the past, and the problem with this is > that the actual mask depends both on the capabilities of the > device and the bus. In particular, if the device can only do > 32-bit DMA, we must not set the mask to something higher. So is the dma-ranges property allowed to specify a size bigger than what the device supports? > The normal rule here is that a driver that wants to do 64-bit > DMA must call dma_set_mask_and_coherent() with the higher mask, > while a device that can not access all of the 32-bit address space > must call dma_set_mask_and_coherent() with the smaller mask > before doing calling any of the other DMA interfaces. OK, looking at the DMA API docs, it looks like the default mask is 32-bit and any other value should be explicitly set by the driver. What we don't have on arm64 yet is taking dma_pfn_offset into account when generating the dma address (but so far I haven't seen any request for this from hardware vendors; it can easily be fixed). So if that's not the case for Ding, I'm not sure dma-ranges property would help. > However, if the bus is not capable of addressing the entire > 32-bit range (as on some modern shmobile machines, or some of the > really old machines), we need to limit the mask here already. Would the limiting be based on the dma-ranges size property? Such information is not easily available after of_dma_configure(), maybe we could store it somewhere in struct device. Going back to original topic, the dma_supported() function on arm64 calls swiotlb_dma_supported() which actually checks whether the swiotlb bounce buffer is within the dma mask. This transparent bouncing (unlike arm32 where it needs to be explicit) is not always optimal, though required for 32-bit only devices on a 64-bit system. The problem is when the driver is 64-bit capable but forgets to call dma_set_mask_and_coherent() (that's not the only question I got about running out of swiotlb buffers). -- Catalin