[PATCH 3/8] ARM: mmu: use reserve mem entries to modify maps

Ahmad Fatoum a.fatoum at pengutronix.de
Tue Apr 20 10:09:35 BST 2021


On 20.04.21 09:56, Rouven Czerwinski wrote:
> Use the information from the reserved memory entries to modify the
> mapping of memory regions to mark them as uncachable and not-executable.
> This also prevents the processor from speculating into these regions,
> preventing hard to debug scenarios where boots fail for unknown reasons.
> 
> Signed-off-by: Rouven Czerwinski <r.czerwinski at pengutronix.de>
> ---
>  arch/arm/cpu/mmu.c | 60 ++++++++++++++++++++++++++++++++++++++++++++--
>  1 file changed, 58 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
> index 6af228505d..71d6cad1ef 100644
> --- a/arch/arm/cpu/mmu.c
> +++ b/arch/arm/cpu/mmu.c
> @@ -17,6 +17,7 @@
>  #include <memory.h>
>  #include <asm/system_info.h>
>  #include <asm/sections.h>
> +#include <of.h>
>  
>  #include "mmu.h"
>  
> @@ -387,6 +388,57 @@ static void vectors_init(void)
>  	create_vector_table(ARM_LOW_VECTORS);
>  }
>  
> +static void create_sections_with_intersect(struct memory_bank *bank)
> +{
> +	int i;
> +	unsigned flag;

u32 pmd_flags fits the intent better.

> +	struct of_reserve_map *res_map = of_get_reserve_map();

Drop this call. It's overwritten later.

> +	unsigned long j_end;
> +	unsigned long end;
> +	unsigned long j;
> +
> +	res_map = of_get_reserve_map();
> +	if (!res_map)

Check unneeded, you never enter here if !of_get_reserve_map() anyway.

> +		return;
> +
> +	end = bank->start + bank->size - 1;
> +	j = bank->start;
> +
> +	while(j < end) {

Missing space after while. Same for if() a few lines down.
IMO a for loop would be clearer here.

> +		flag = PMD_SECT_DEF_CACHED;
> +		j_end = j + PGDIR_SIZE - 1;
> +
> +		for (i = 0; i < res_map->num_entries; i++) {
> +			if(BIT(i) & res_map->xn && (j_end >= res_map->start[i] &&

Always wrap bitwise comparisons in parenthesis. Doesn't matter for code
correctness here however. The parens for the && stuff can be dropped.

> +						    j_end <= res_map->end[i]))

What about a section that starts in a reserved region, but doesn't end in one?
This wouldn't be considered here, would it?

> +				flag = PMD_SECT_DEF_UNCACHED | PMD_SECT_XN;
> +		}
> +
> +		create_sections(ttb, j, j_end, flag);
> +		j += PGDIR_SIZE;
> +	}
> +}
> +
> +static bool intersects_reserved_map_xn(struct memory_bank *bank)
> +{
> +	struct of_reserve_map *res_map;
> +	unsigned long bank_end;
> +	int i;
> +
> +	res_map = of_get_reserve_map();
> +
> +	if (!res_map)

Also early-exit if !res_map->xn.

> +		return false;
> +
> +	bank_end = bank->start + bank->size;

band_end is one past the end here...

> +	for (i = 0; i < res_map->num_entries; i++) {
> +		if(res_map->start[i] >= bank->start || res_map->end[i] <= bank_end)

So this should read < bank_end.

> +			return true;
> +	}
> +	return false;
> +}
> +
> +
>  /*
>   * Prepare MMU for usage enable it.
>   */
> @@ -448,8 +500,12 @@ void __mmu_init(bool mmu_on)
>  	vectors_init();
>  
>  	for_each_memory_bank(bank) {
> -		create_sections(ttb, bank->start, bank->start + bank->size - 1,
> -				PMD_SECT_DEF_CACHED);
> +		if (intersects_reserved_map_xn(bank)) {

I think it would be cleaner to remove this check here and just have one section
creation function. You can check whether there are any XN sections and if any
XN sections are intersecting at the start and then just have a flag you check
in the loop to skip the section search


> +			create_sections_with_intersect(bank);
> +		} else {
> +			create_sections(ttb, bank->start, bank->start + bank->size - 1,
> +					PMD_SECT_DEF_CACHED);
> +		}
>  		__mmu_cache_flush();
>  	}
>  
> 

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |



More information about the barebox mailing list