[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200930165212.GA8833@lst.de>
Date: Wed, 30 Sep 2020 18:52:12 +0200
From: Christoph Hellwig <hch@....de>
To: Paul Cercueil <paul@...pouillou.net>
Cc: Christoph Hellwig <hch@....de>,
Stephen Rothwell <sfr@...b.auug.org.au>,
Dave Airlie <airlied@...ux.ie>,
DRI <dri-devel@...ts.freedesktop.org>,
Linux Next Mailing List <linux-next@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: linux-next: build failure after merge of the drm tree
On Wed, Sep 30, 2020 at 06:45:02PM +0200, Paul Cercueil wrote:
>> We don't have such a thing in the Linux API at all.
>
> dma_pgprot(dev, vma->vm_page_prot, DMA_ATTR_NON_CONSISTENT);
>
> That was giving me non-coherent cached memory, and now I don't have an
> alternative.
Looking at Linux 5.9-rc dma_pgprot is defined as:
pgprot_t dma_pgprot(struct device *dev, pgprot_t prot, unsigned long attrs)
{
if (force_dma_unencrypted(dev))
prot = pgprot_decrypted(prot);
if (dev_is_dma_coherent(dev) ||
(IS_ENABLED(CONFIG_DMA_NONCOHERENT_CACHE_SYNC) &&
(attrs & DMA_ATTR_NON_CONSISTENT)))
return prot;
#ifdef CONFIG_ARCH_HAS_DMA_WRITE_COMBINE
if (attrs & DMA_ATTR_WRITE_COMBINE)
return pgprot_writecombine(prot);
#endif
return pgprot_dmacoherent(prot);
}
so it doesn't change vma->vm_page_prot at all.
The only place that uses _CACHE_CACHABLE_NONCOHERENT is the MIPS specific
kmap_noncoherent which ha sa single caller that doesn't leak anywhere
into driver code.
Powered by blists - more mailing lists