[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20201001171922.GO21544@gaia>
Date: Thu, 1 Oct 2020 18:19:22 +0100
From: Catalin Marinas <catalin.marinas@....com>
To: Nicolas Saenz Julienne <nsaenzjulienne@...e.de>
Cc: robh+dt@...nel.org, hch@....de, robin.murphy@....com,
devicetree@...r.kernel.org, iommu@...ts.linux-foundation.org,
linux-mm@...ck.org, linux-kernel@...r.kernel.org, will@...nel.org,
linux-arm-kernel@...ts.infradead.org,
linux-rpi-kernel@...ts.infradead.org
Subject: Re: [PATCH 3/4] arm64: Default to 32-bit ZONE_DMA
On Thu, Oct 01, 2020 at 06:17:39PM +0200, Nicolas Saenz Julienne wrote:
> diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
> index e1a69a618832..3c3f462466eb 100644
> --- a/arch/arm64/mm/init.c
> +++ b/arch/arm64/mm/init.c
> @@ -43,8 +43,6 @@
> #include <asm/tlb.h>
> #include <asm/alternative.h>
>
> -#define ARM64_ZONE_DMA_BITS 30
> -
> /*
> * We need to be able to catch inadvertent references to memstart_addr
> * that occur (potentially in generic code) before arm64_memblock_init()
> @@ -388,8 +386,14 @@ void __init arm64_memblock_init(void)
> early_init_fdt_scan_reserved_mem();
>
> if (IS_ENABLED(CONFIG_ZONE_DMA)) {
> - zone_dma_bits = ARM64_ZONE_DMA_BITS;
> - arm64_dma_phys_limit = max_zone_phys(ARM64_ZONE_DMA_BITS);
> + /*
> + * early_init_dt_scan() might alter zone_dma_bits based on the
> + * device's DT. Otherwise, have it cover the 32-bit address
> + * space.
> + */
> + if (zone_dma_bits == ZONE_DMA_BITS_DEFAULT)
> + zone_dma_bits = 32;
> + arm64_dma_phys_limit = max_zone_phys(zone_dma_bits);
So here we assume that if zone_dma_bits is 24, it wasn't initialised. I
think it may be simpler if we just set it in setup_machine_fdt() to 32
or 30 if RPi4. This way we don't have to depend on what the core kernel
sets.
--
Catalin
Powered by blists - more mailing lists