Drew found "CONFIG_DMA_GLOBAL_POOL=y causes ADMA buffer alloc to fail"
the log looks like:
[ 3.741083] mmc0: Unable to allocate ADMA buffers - falling back to standard DMA
The logic is: generic riscv defconfig selects ARCH_RENESAS then
ARCH_R9A07G043 which selects DMA_GLOBAL_POOL, which assumes all
non-dma-coherent riscv platforms have a dma global pool, this assumption
seems not correct. And I believe DMA_GLOBAL_POOL should not be
selected by ARCH_SOCFAMILIY, instead, only ARCH under some specific
conditions can select it globaly, for example NOMMU ARM and so on,
because it's designed for special cases such as "nommu cases where
non-cacheable memory lives in a fixed place in the physical address
map" as pointed out by Robin.
Fix the issue by making ARCH_R9A07G043 (riscv version) depend on
NONPORTABLE, thus generic defconfig won't select ARCH_R9A07G043 by
default. And even for random config case, there will be less debug
effort once we see NONPORTABLE is enabled.
Reported-by: Drew Fustini <[email protected]>
Closes: https://lore.kernel.org/linux-riscv/ZRuamJuShOnvP1pr@x1/
Fixes: 484861e09f3e ("soc: renesas: Kconfig: Select the required configs for RZ/Five SoC")
Signed-off-by: Jisheng Zhang <[email protected]>
---
drivers/soc/renesas/Kconfig | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/soc/renesas/Kconfig b/drivers/soc/renesas/Kconfig
index 5a75ab64d1ed..12040ce116a5 100644
--- a/drivers/soc/renesas/Kconfig
+++ b/drivers/soc/renesas/Kconfig
@@ -333,6 +333,7 @@ if RISCV
config ARCH_R9A07G043
bool "RISC-V Platform support for RZ/Five"
+ depends on NONPORTABLE
select ARCH_RZG2L
select AX45MP_L2_CACHE if RISCV_DMA_NONCOHERENT
select DMA_GLOBAL_POOL
--
2.40.1
On 04/10/2023 4:08 pm, Jisheng Zhang wrote:
> Drew found "CONFIG_DMA_GLOBAL_POOL=y causes ADMA buffer alloc to fail"
> the log looks like:
> [ 3.741083] mmc0: Unable to allocate ADMA buffers - falling back to standard DMA
>
> The logic is: generic riscv defconfig selects ARCH_RENESAS then
> ARCH_R9A07G043 which selects DMA_GLOBAL_POOL, which assumes all
> non-dma-coherent riscv platforms have a dma global pool, this assumption
> seems not correct. And I believe DMA_GLOBAL_POOL should not be
> selected by ARCH_SOCFAMILIY, instead, only ARCH under some specific
> conditions can select it globaly, for example NOMMU ARM and so on,
> because it's designed for special cases such as "nommu cases where
> non-cacheable memory lives in a fixed place in the physical address
> map" as pointed out by Robin.
>
> Fix the issue by making ARCH_R9A07G043 (riscv version) depend on
> NONPORTABLE, thus generic defconfig won't select ARCH_R9A07G043 by
> default. And even for random config case, there will be less debug
> effort once we see NONPORTABLE is enabled.
FWIW,
Reviewed-by: Robin Murphy <[email protected]>
> Reported-by: Drew Fustini <[email protected]>
> Closes: https://lore.kernel.org/linux-riscv/ZRuamJuShOnvP1pr@x1/
> Fixes: 484861e09f3e ("soc: renesas: Kconfig: Select the required configs for RZ/Five SoC")
> Signed-off-by: Jisheng Zhang <[email protected]>
> ---
> drivers/soc/renesas/Kconfig | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/soc/renesas/Kconfig b/drivers/soc/renesas/Kconfig
> index 5a75ab64d1ed..12040ce116a5 100644
> --- a/drivers/soc/renesas/Kconfig
> +++ b/drivers/soc/renesas/Kconfig
> @@ -333,6 +333,7 @@ if RISCV
>
> config ARCH_R9A07G043
> bool "RISC-V Platform support for RZ/Five"
> + depends on NONPORTABLE
> select ARCH_RZG2L
> select AX45MP_L2_CACHE if RISCV_DMA_NONCOHERENT
> select DMA_GLOBAL_POOL
On Wed, Oct 04, 2023 at 11:08:56PM +0800, Jisheng Zhang wrote:
> Drew found "CONFIG_DMA_GLOBAL_POOL=y causes ADMA buffer alloc to fail"
> the log looks like:
> [ 3.741083] mmc0: Unable to allocate ADMA buffers - falling back to standard DMA
>
> The logic is: generic riscv defconfig selects ARCH_RENESAS then
> ARCH_R9A07G043 which selects DMA_GLOBAL_POOL, which assumes all
> non-dma-coherent riscv platforms have a dma global pool, this assumption
> seems not correct. And I believe DMA_GLOBAL_POOL should not be
> selected by ARCH_SOCFAMILIY, instead, only ARCH under some specific
> conditions can select it globaly, for example NOMMU ARM and so on,
> because it's designed for special cases such as "nommu cases where
> non-cacheable memory lives in a fixed place in the physical address
> map" as pointed out by Robin.
>
> Fix the issue by making ARCH_R9A07G043 (riscv version) depend on
> NONPORTABLE, thus generic defconfig won't select ARCH_R9A07G043 by
> default. And even for random config case, there will be less debug
> effort once we see NONPORTABLE is enabled.
>
> Reported-by: Drew Fustini <[email protected]>
> Closes: https://lore.kernel.org/linux-riscv/ZRuamJuShOnvP1pr@x1/
> Fixes: 484861e09f3e ("soc: renesas: Kconfig: Select the required configs for RZ/Five SoC")
> Signed-off-by: Jisheng Zhang <[email protected]>
> ---
> drivers/soc/renesas/Kconfig | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/soc/renesas/Kconfig b/drivers/soc/renesas/Kconfig
> index 5a75ab64d1ed..12040ce116a5 100644
> --- a/drivers/soc/renesas/Kconfig
> +++ b/drivers/soc/renesas/Kconfig
> @@ -333,6 +333,7 @@ if RISCV
>
> config ARCH_R9A07G043
> bool "RISC-V Platform support for RZ/Five"
> + depends on NONPORTABLE
> select ARCH_RZG2L
> select AX45MP_L2_CACHE if RISCV_DMA_NONCOHERENT
> select DMA_GLOBAL_POOL
> --
> 2.40.1
>
Tested-by: Drew Fustini <[email protected]>
With this patch, I can use the defconfig and sdhci-of-dwcmshc driver is
able to use ADMA mode without needing to add shared-dma-pool to dts:
# zcat /proc/config.gz |grep DMA_GLOBAL_POOL
# dtc -I fs /sys/firmware/devicetree/base/ 2>/dev/null |grep pool
# dmesg |grep ADMA
[ 3.910269] mmc0: SDHCI controller on ffe7080000.mmc [ffe7080000.mmc] using ADMA 64-bit
Thanks,
Drew
On Wed, Oct 4, 2023 at 5:21 PM Jisheng Zhang <[email protected]> wrote:
> Drew found "CONFIG_DMA_GLOBAL_POOL=y causes ADMA buffer alloc to fail"
> the log looks like:
> [ 3.741083] mmc0: Unable to allocate ADMA buffers - falling back to standard DMA
>
> The logic is: generic riscv defconfig selects ARCH_RENESAS then
> ARCH_R9A07G043 which selects DMA_GLOBAL_POOL, which assumes all
> non-dma-coherent riscv platforms have a dma global pool, this assumption
> seems not correct. And I believe DMA_GLOBAL_POOL should not be
> selected by ARCH_SOCFAMILIY, instead, only ARCH under some specific
> conditions can select it globaly, for example NOMMU ARM and so on,
> because it's designed for special cases such as "nommu cases where
> non-cacheable memory lives in a fixed place in the physical address
> map" as pointed out by Robin.
>
> Fix the issue by making ARCH_R9A07G043 (riscv version) depend on
> NONPORTABLE, thus generic defconfig won't select ARCH_R9A07G043 by
> default. And even for random config case, there will be less debug
> effort once we see NONPORTABLE is enabled.
>
> Reported-by: Drew Fustini <[email protected]>
> Closes: https://lore.kernel.org/linux-riscv/ZRuamJuShOnvP1pr@x1/
> Fixes: 484861e09f3e ("soc: renesas: Kconfig: Select the required configs for RZ/Five SoC")
> Signed-off-by: Jisheng Zhang <[email protected]>
Reviewed-by: Geert Uytterhoeven <[email protected]>
I guess this is the right thing to do until a better solution is found.
I.e. will queue in renesas-fixes for v6.6.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Wed, Oct 4, 2023 at 4:21 PM Jisheng Zhang <[email protected]> wrote:
>
> Drew found "CONFIG_DMA_GLOBAL_POOL=y causes ADMA buffer alloc to fail"
> the log looks like:
> [ 3.741083] mmc0: Unable to allocate ADMA buffers - falling back to standard DMA
>
> The logic is: generic riscv defconfig selects ARCH_RENESAS then
> ARCH_R9A07G043 which selects DMA_GLOBAL_POOL, which assumes all
> non-dma-coherent riscv platforms have a dma global pool, this assumption
> seems not correct. And I believe DMA_GLOBAL_POOL should not be
> selected by ARCH_SOCFAMILIY, instead, only ARCH under some specific
> conditions can select it globaly, for example NOMMU ARM and so on,
> because it's designed for special cases such as "nommu cases where
> non-cacheable memory lives in a fixed place in the physical address
> map" as pointed out by Robin.
>
> Fix the issue by making ARCH_R9A07G043 (riscv version) depend on
> NONPORTABLE, thus generic defconfig won't select ARCH_R9A07G043 by
> default. And even for random config case, there will be less debug
> effort once we see NONPORTABLE is enabled.
>
> Reported-by: Drew Fustini <[email protected]>
> Closes: https://lore.kernel.org/linux-riscv/ZRuamJuShOnvP1pr@x1/
> Fixes: 484861e09f3e ("soc: renesas: Kconfig: Select the required configs for RZ/Five SoC")
> Signed-off-by: Jisheng Zhang <[email protected]>
> ---
> drivers/soc/renesas/Kconfig | 1 +
> 1 file changed, 1 insertion(+)
>
Reviewed-by: Lad Prabhakar <[email protected]>
Cheers,
Prabhakar
> diff --git a/drivers/soc/renesas/Kconfig b/drivers/soc/renesas/Kconfig
> index 5a75ab64d1ed..12040ce116a5 100644
> --- a/drivers/soc/renesas/Kconfig
> +++ b/drivers/soc/renesas/Kconfig
> @@ -333,6 +333,7 @@ if RISCV
>
> config ARCH_R9A07G043
> bool "RISC-V Platform support for RZ/Five"
> + depends on NONPORTABLE
> select ARCH_RZG2L
> select AX45MP_L2_CACHE if RISCV_DMA_NONCOHERENT
> select DMA_GLOBAL_POOL
> --
> 2.40.1
>