2020-01-10 17:21:16

by Nicolas Saenz Julienne

Subject: [PATCH] dma-contiguous: CMA: give precedence to cmdline

Although the device tree might contain a reserved-memory DT node
dedicated as the default CMA pool, users might want to change CMA's
parameters using the kernel command line for debugging purposes and
whatnot. Honor this by bypassing the reserved memory CMA setup, which
will ultimately end up freeing the memblock and allow the command line
CMA configuration routine to run.

Signed-off-by: Nicolas Saenz Julienne <[email protected]>
---

NOTE: Tested this on arm and arm64 with the Raspberry Pi 4.

kernel/dma/contiguous.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/dma/contiguous.c b/kernel/dma/contiguous.c
index daa4e6eefdde..8bc6f2d670f9 100644
--- a/kernel/dma/contiguous.c
+++ b/kernel/dma/contiguous.c
@@ -302,9 +302,16 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
phys_addr_t align = PAGE_SIZE << max(MAX_ORDER - 1, pageblock_order);
phys_addr_t mask = align - 1;
unsigned long node = rmem->fdt_node;
+ bool default_cma = of_get_flat_dt_prop(node, "linux,cma-default", NULL);
struct cma *cma;
int err;

+ if (size_cmdline != -1 && default_cma) {
+ pr_info("Reserved memory: bypass %s node, using cmdline CMA params instead\n",
+ rmem->name);
+ return -EBUSY;
+ }
+
if (!of_get_flat_dt_prop(node, "reusable", NULL) ||
of_get_flat_dt_prop(node, "no-map", NULL))
return -EINVAL;
@@ -322,7 +329,7 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
/* Architecture specific contiguous memory fixup. */
dma_contiguous_early_fixup(rmem->base, rmem->size);

- if (of_get_flat_dt_prop(node, "linux,cma-default", NULL))
+ if (default_cma)
dma_contiguous_set_default(cma);

rmem->ops = &rmem_cma_ops;
--
2.24.1

2020-01-10 17:52:25

by Phil Elwell

Subject: Re: [PATCH] dma-contiguous: CMA: give precedence to cmdline

Hi Nicolas,

On 10/01/2020 17:19, Nicolas Saenz Julienne wrote:
> Although the device tree might contain a reserved-memory DT node
> dedicated as the default CMA pool, users might want to change CMA's
> parameters using the kernel command line for debugging purposes and
> whatnot. Honor this by bypassing the reserved memory CMA setup, which
> will ultimately end up freeing the memblock and allow the command line
> CMA configuration routine to run.
>
> Signed-off-by: Nicolas Saenz Julienne <[email protected]>
> ---
>
> NOTE: Tested this on arm and arm64 with the Raspberry Pi 4.
>
> kernel/dma/contiguous.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/dma/contiguous.c b/kernel/dma/contiguous.c
> index daa4e6eefdde..8bc6f2d670f9 100644
> --- a/kernel/dma/contiguous.c
> +++ b/kernel/dma/contiguous.c
> @@ -302,9 +302,16 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
> phys_addr_t align = PAGE_SIZE << max(MAX_ORDER - 1, pageblock_order);
> phys_addr_t mask = align - 1;
> unsigned long node = rmem->fdt_node;
> + bool default_cma = of_get_flat_dt_prop(node, "linux,cma-default", NULL);
> struct cma *cma;
> int err;
>
> + if (size_cmdline != -1 && default_cma) {
> + pr_info("Reserved memory: bypass %s node, using cmdline CMA params instead\n",
> + rmem->name);
> + return -EBUSY;
> + }
> +
> if (!of_get_flat_dt_prop(node, "reusable", NULL) ||
> of_get_flat_dt_prop(node, "no-map", NULL))
> return -EINVAL;
> @@ -322,7 +329,7 @@ static int __init rmem_cma_setup(struct reserved_mem *rmem)
> /* Architecture specific contiguous memory fixup. */
> dma_contiguous_early_fixup(rmem->base, rmem->size);
>
> - if (of_get_flat_dt_prop(node, "linux,cma-default", NULL))
> + if (default_cma)
> dma_contiguous_set_default(cma);
>
> rmem->ops = &rmem_cma_ops;
>

For what it's worth,

Reviewed-by: Phil Elwell <[email protected]>

Phil

2020-01-30 13:43:38

by Christoph Hellwig

Subject: Re: [PATCH] dma-contiguous: CMA: give precedence to cmdline

I've picked this up for Linux 5.6, sorry for the delay.

2020-01-30 16:12:54

by Nicolas Saenz Julienne

Subject: Re: [PATCH] dma-contiguous: CMA: give precedence to cmdline

On Thu Jan 30, 2020 at 2:42 PM, Christoph Hellwig wrote:
> I've picked this up for Linux 5.6, sorry for the delay.

Thanks!