Re: [PATCH] dma-pool: Do not allocate pool memory from CMA

From: Amit Pundir <amit.pundir@linaro.org>
To: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Cc: lkml <linux-kernel@vger.kernel.org>,
	jeremy.linton@arm.com, Sumit Semwal <sumit.semwal@linaro.org>,
	iommu@lists.linux-foundation.org,
	John Stultz <john.stultz@linaro.org>,
	linux-rpi-kernel@lists.infradead.org,
	David Rientjes <rientjes@google.com>,
	Robin Murphy <robin.murphy@arm.com>,
	Christoph Hellwig <hch@lst.de>
Subject: Re: [PATCH] dma-pool: Do not allocate pool memory from CMA
Date: Thu, 23 Jul 2020 10:44:58 +0530	[thread overview]
Message-ID: <CAMi1Hd0Xz6kOJFpA5PEpi6RDDGOcz0RmQ7tTOkuXq4QneOO_vQ@mail.gmail.com> (raw)
In-Reply-To: <6db722947546221ed99d3f473f78e1a6de65d7d6.camel@suse.de>

Hi Nicolas,

Sorry I got stuck on other things yesterday.

On Tue, 21 Jul 2020 at 21:57, Nicolas Saenz Julienne
<nsaenzjulienne@suse.de> wrote:
>
> On Tue, 2020-07-21 at 20:52 +0530, Amit Pundir wrote:
>
> [...]
>
> > > > > Can you try booting *without* my patch and this in the kernel
> > > > > command
> > > > > line: "cma=16M@0x100000000-0x200000000".
> > > >
> > > > It doesn't boot with this added kernel command line.
> > >
> > > For the record, this placed the CMA in the [4GB, 8GB] address space
> > > instead of you setup's default: [3GB, 4GB]. All atomic pools fall
> > > in
> > > that memory area without my patch, which makes me think some of the
> > > devices on your board might not like higher addresses.
> > >
> >
> > Thank you Nicolas for the details. Though we don't set the CMA
> > alloc-ranges explicitly in upstream sdm845 dts, but I dug around and
> > found that CMA alloc-ranges in the downstream kernel are indeed in
> > lower address space.
> > https://github.com/MiCode/Xiaomi_Kernel_OpenSource/blob/dipper-q-oss/arch/arm64/boot/dts/qcom/sdm845.dtsi#L662
> >
> > /* global autoconfigured region for contiguous allocations */
> > linux,cma {
> >         compatible = "shared-dma-pool";
> >         alloc-ranges = <0 0x00000000 0 0xffffffff>;
> >         reusable;
> >         alignment = <0 0x400000>;
> >         size = <0 0x2000000>;
> >         linux,cma-default;
> > };
>
> Pretty standard, and similar to what it's being used upstream by
> default.
>
> >
> > > What happens if you boot with my troublesome patch with this in
> > > your
> > > device tree? (insert it at the bottom of sdm845-beryllium.dts)
> > >
> > > &soc {
> > >         dma-ranges = <0 0 0 0 0x1 0>;
> > > };
> > >
> >
> > Device still doesn't boot up to adb shell.
>
> Let's get a bigger hammer, I'm just looking for clues here. Can you
> apply this and provide the dmesg output.
>
> diff --git a/kernel/dma/pool.c b/kernel/dma/pool.c
> index 6bc74a2d5127..2160676bf488 100644
> --- a/kernel/dma/pool.c
> +++ b/kernel/dma/pool.c
> @@ -268,6 +268,8 @@ void *dma_alloc_from_pool(struct device *dev, size_t size,
>                         schedule_work(&atomic_pool_work);
>         }
>
> +       dev_info(dev, "%s: size %lx, phys addr %llx, flags 0x%x\n", __func__, size, phys, flags);
> +
>         return ptr;
>  }

I never made it to dma_alloc_from_pool() call from
dma_direct_alloc_pages(), dma_should_alloc_from_pool() returns False
from gfpflags_allow_blocking() block.

Regards,
Amit Pundir

>
>
> Regards,
> Nicolas
>
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu