[PATCH] mm/memblock:Do not retry a range that has already been checked

[PATCH] mm/memblock:Do not retry a range that has already been checked

[daeroro]

The range that has already been checked
don't have to be checked in a second attempt.

Signed-off-by: daeroro <skseofh@naver.com>
---
 mm/memblock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/memblock.c b/mm/memblock.c
index 39aceafc57f6..6f72fae415ee 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -1489,7 +1489,7 @@ static void * __init memblock_alloc_internal(
 
 	/* retry allocation without lower limit */
 	if (!alloc && min_addr)
-		alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid,
+		alloc = memblock_alloc_range_nid(size, align, 0, min_addr, nid,
 						exact_nid);
 
 	if (!alloc)
-- 
2.17.1

Re: [PATCH] mm/memblock:Do not retry a range that has already been checked

[Mike Rapoport]

On Sun, May 24, 2020 at 11:16:40PM +0900, daeroro wrote:
> The range that has already been checked
> don't have to be checked in a second attempt.

The first attempts tries to find free memory in the interval [min_addr,
max_addr) and the second attempt does not care about min_addr and looks
for free memory in the interval [0, max_addr).

Is there a problem you see with this algorthim?

> Signed-off-by: daeroro <skseofh@naver.com>
> ---
>  mm/memblock.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 39aceafc57f6..6f72fae415ee 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -1489,7 +1489,7 @@ static void * __init memblock_alloc_internal(
>  
>  	/* retry allocation without lower limit */
>  	if (!alloc && min_addr)
> -		alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid,
> +		alloc = memblock_alloc_range_nid(size, align, 0, min_addr, nid,
>  						exact_nid);
>  
>  	if (!alloc)
> -- 
> 2.17.1
> 

-- 
Sincerely yours,
Mike.

Re: [PATCH] mm/memblock:Do not retry a range that has already been checked

[Mike Rapoport]

On Tue, May 26, 2020 at 12:48:38AM +0900, DaeRoLee wrote:
> Sorry for the invalid mail format..
> 
> Actually, I didn't experience a real issue.
> 
> On second thought, I totally agree with you.
> 
> Could you explain "size" in "min_addr + size -1"?
> Does this mean memblock size?

"size" is the allocation size. It could be that there is no free slot
above min_addr, but there is free memory slightly below it. If you ceil
the second search by min_addr, the allocation must start below min_addr
- size, rather than below min_addr.

-- 
Sincerely yours,
Mike.