Re: [PATCH v2] mm/mm_init: Don't iterate pages below ARCH_PFN_OFFSET

[Mike Rapoport <rppt@kernel.org>]

On Sat, Apr 19, 2025 at 08:28:01PM +0800, Ruihan Li wrote:
> Currently, memmap_init initializes pfn_hole with 0 instead of
> ARCH_PFN_OFFSET. Then init_unavailable_range will start iterating each
> page from the page at address zero to the first available page, but it
> won't do anything for pages below ARCH_PFN_OFFSET because pfn_valid
> won't pass.
> 
> If ARCH_PFN_OFFSET is very large (e.g., something like 2^64-2GiB if the
> kernel is used as a library and loaded at a very high address), the
> pointless iteration for pages below ARCH_PFN_OFFSET will take a very
> long time, and the kernel will look stuck at boot time.
> 
> This commit sets the initial value of pfn_hole to ARCH_PFN_OFFSET, which
> avoids the problematic and useless iteration mentioned above.
> 
> This problem has existed since commit 907ec5fca3dc ("mm: zero remaining
> unavailable struct pages").
> 
> Signed-off-by: Ruihan Li <lrh2000@pku.edu.cn>
> ---
> Link to v1:
>  - https://lore.kernel.org/linux-mm/20250418162727.1535335-1-lrh2000@pku.edu.cn/
> Changes since v1:
>  - Removed the unnecessary Fixes tag.
>  - Fixed the build issue for CONFIG_SPARSEMEM.
> 
>  mm/mm_init.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/mm/mm_init.c b/mm/mm_init.c
> index 84f14fa12..a697a83ff 100644
> --- a/mm/mm_init.c
> +++ b/mm/mm_init.c
> @@ -969,6 +969,15 @@ static void __init memmap_init(void)
>  	unsigned long hole_pfn = 0;
>  	int i, j, zone_id = 0, nid;
>  
> +#ifdef CONFIG_FLATMEM
> +	/*
> +	 * Pages below ARCH_PFN_OFFSET are invalid as far as pfn_valid is
> +	 * concerned, so don't waste time iterating on them when looking
> +	 * for holes.
> +	 */
> +	hole_pfn = ARCH_PFN_OFFSET;
> +#endif
> +

I'd prefer a solution for both FLATMEM and SPARSMEM. 

David Woodhouse proposed a for_each_valid_pfn() a while ago:

https://lore.kernel.org/all/20250404155959.3442111-1-dwmw2@infradead.org

It can be used in init_unavailable_range() and will essentially skip the
unpopulated memory map.

>  	for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
>  		struct pglist_data *node = NODE_DATA(nid);
>  
> -- 
> 2.49.0
> 

-- 
Sincerely yours,
Mike.