linux-kernel - Re: [PATCH] dma-direct: Set SG_DMA

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Thu, 9 May 2024 11:26:22 -0700
From: "T.J. Mercier" <tjmercier@...gle.com>
To: Robin Murphy <robin.murphy@....com>
Cc: Christoph Hellwig <hch@....de>, Marek Szyprowski <m.szyprowski@...sung.com>, isaacmanjarres@...gle.com, 
	Catalin Marinas <catalin.marinas@....com>, iommu@...ts.linux.dev, 
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] dma-direct: Set SG_DMA_SWIOTLB flag for dma-direct

On Thu, May 9, 2024 at 5:55 AM Robin Murphy <robin.murphy@....com> wrote:
>
> On 08/05/2024 12:33 pm, Christoph Hellwig wrote:
> > On Tue, May 07, 2024 at 01:07:25PM -0700, T.J. Mercier wrote:
> >> On Mon, May 6, 2024 at 10:43???PM Christoph Hellwig <hch@....de> wrote:
> >>>
> >>> On Mon, May 06, 2024 at 09:39:53AM -0700, T.J. Mercier wrote:
> >>>>> You should not check, you simply must handle it by doing the proper
> >>>>> DMA API based ownership management.
> >>>>
> >>>> That doesn't really work for uncached buffers.
> >>>
> >>> What uncached buffers?
> >>
> >> For example these ones:
> >> https://android.googlesource.com/kernel/common/+/refs/heads/android-mainline/drivers/dma-buf/heaps/system_heap.c#141
> >
> > Whatever that code is doing is probably not upstream because it's too
> > broken to live.
>
> Indeed, at a glance it appears to be trying to reinvent
> dma_alloc_noncontiguous(). What's not immediately obvious is whether
> it's particular about allocations being DMA-contiguous; if not then I
> think it comes down to the same thing as vb2-dma-sg and the ideas we
> were tossing around for that[1].

This heap isn't too picky about the memory being contiguous, but there
is an attempt to allocate high order pages if possible to reduce the
number of translation entries. These page orders are one thing vendors
sometimes change for the hardware they have.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/dma-buf/heaps/system_heap.c#n48

I think one problem there'd be attempting to use
dma_alloc_noncontiguous for dmabuf exporters is that memory is
typically (always?) allocated when the buffer is exported, and there
won't be any device attachments at that time.


> Thanks,
> Robin.
>
> [1]
> https://lore.kernel.org/linux-media/20231228074645.765yytb2a7hvz7ti@chromium.org/