Re: [PATCH] of: of_reserved_mem: match memblock_free with memblock_reserve

Re: [PATCH] of: of_reserved_mem: match memblock_free with memblock_reserve

[Rob Herring]

On Thu, Jul 1, 2021 at 8:10 AM 권오훈 <ohoono.kwon@samsung.com> wrote:
>
> When __reserved_mem_init_node called from fdt_init_reserved_mem fails,
>
> we try to undo __reserved_mem_alloc_size to prevent memory leak.
>
> 'commit d0b8ed47e83a ("of: reserved_mem: fix reserve memory leak")'

Your patch is corrupted and not plain text.

In any case, I believe this issue has already been fixed. Check the
latest kernel tree(s).

>
> Meanwhile, __reserved_mem_alloc_size calls
>
> early_init_dt_alloc_reserved_memory_arch to allocate memory,
>
> which calls
>
> 1) memblock_remove when rmem is declared nomap,
>
> 2) memblock_reserve, otherwise.
>
>
>
> static int __init early_init_dt_alloc_reserved_memory_arch(
>
> ...
>
>         if (nomap)
>
>                 return memblock_remove(base, size);
>
>
>
>         return memblock_reserve(base, size);
>
> }
>
>
>
> Therefore the proper undo-logic should be as follows:
>
> 1) memblock_add when rmem is declared nomap,
>
> 2) memblock_free, otherwise.
>
>
>
> Match the undo functions for readability.
>
>
>
> Signed-off-by: Ohhoon Kwon <ohoono.kwon@samsung.com>
>
> ---
>
>  drivers/of/of_reserved_mem.c | 3 ++-
>
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
>
>
> diff --git a/drivers/of/of_reserved_mem.c b/drivers/of/of_reserved_mem.c
>
> index 15e2417974d6..2279e1b55d1d 100644
>
> --- a/drivers/of/of_reserved_mem.c
>
> +++ b/drivers/of/of_reserved_mem.c
>
> @@ -273,9 +273,10 @@ void __init fdt_init_reserved_mem(void)
>
>                          if (err != 0 && err != -ENOENT) {
>
>                                  pr_info("node %s compatible matching fail\n",
>
>                                          rmem->name);
>
> -                                memblock_free(rmem->base, rmem->size);
>
>                                  if (nomap)
>
>                                          memblock_add(rmem->base, rmem->size);
>
> +                                else
>
> +                                        memblock_free(rmem->base, rmem->size);
>
>                          }
>
>                  }
>
>          }
>
> --
>
> 2.17.1
>
>
>
>
>
>

[PATCH] of: of_reserved_mem: match memblock_free with memblock_reserve

[권오훈]

When __reserved_mem_init_node called from fdt_init_reserved_mem fails,
we try to undo __reserved_mem_alloc_size to prevent memory leak.
'commit d0b8ed47e83a ("of: reserved_mem: fix reserve memory leak")'

Meanwhile, __reserved_mem_alloc_size calls
early_init_dt_alloc_reserved_memory_arch to allocate memory,
which calls
1) memblock_remove when rmem is declared nomap,
2) memblock_reserve, otherwise.

static int __init early_init_dt_alloc_reserved_memory_arch(
...
	if (nomap)
		return memblock_remove(base, size);

	return memblock_reserve(base, size);
}

Therefore the proper undo-logic should be as follows:
1) memblock_add when rmem is declared nomap,
2) memblock_free, otherwise.

Match the undo functions for readability.

Signed-off-by: Ohhoon Kwon <ohoono.kwon@samsung.com>
---
 drivers/of/of_reserved_mem.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/of/of_reserved_mem.c b/drivers/of/of_reserved_mem.c
index 15e2417974d6..2279e1b55d1d 100644
--- a/drivers/of/of_reserved_mem.c
+++ b/drivers/of/of_reserved_mem.c
@@ -273,9 +273,10 @@ void __init fdt_init_reserved_mem(void)
 			if (err != 0 && err != -ENOENT) {
 				pr_info("node %s compatible matching fail\n",
 					rmem->name);
-				memblock_free(rmem->base, rmem->size);
 				if (nomap)
 					memblock_add(rmem->base, rmem->size);
+				else
+					memblock_free(rmem->base, rmem->size);
 			}
 		}
 	}
-- 
2.17.1