Re: Regression in 543cea9a - was: Re: Kernel problem on rx2800 i2

[Frank Scheiner <frank.scheiner@web.de>]

On 6/25/19 12:32, Christoph Hellwig wrote:
> Thanks Frank.
>
> It seems like there is something odd going on with the zonelist
> on your system.  Maybe there is not actual ZONE_DMA32, or something
> is messed with the numa node setup.

Do you suspect a firmware issue? Because the firmware of that machine is
actually quite old (the model was retired in 2015):

```
***********************************************************
* ROM Version : 01.93
* ROM Date    : Wed Sep 12 22:10:03 PDT 2012
***********************************************************
```

...but since HP has a different opinion about what is considered part of
the hardware and hence should be "corrected" by the manufacturer in any
case without asking for large amounts of money, I have no means to
upgrade it.

>  Below is a band aid patch to
> try theory number two above:
>
>
> diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c
> index fcdb23e8d2fc..8e3f7b8bdb33 100644
> --- a/kernel/dma/direct.c
> +++ b/kernel/dma/direct.c
> @@ -119,7 +119,7 @@ struct page *__dma_direct_alloc_pages(struct device *dev, size_t size,
>   		}
>   	}
>   	if (!page)
> -		page = alloc_pages_node(dev_to_node(dev), gfp, page_order);
> +		page = alloc_pages(gfp, page_order);
>
>   	if (page && !dma_coherent_ok(dev, page_to_phys(page), size)) {
>   		__free_pages(page, page_order);
>

Ok, will try that patch - actually (1) in addition or (2) without the
first one?