Re: [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock_skip on reserved pages

[Mike Rapoport <rppt@linux.ibm.com>]

Hello Andrea,

On Thu, Nov 26, 2020 at 07:46:05PM +0200, Mike Rapoport wrote:
> On Thu, Nov 26, 2020 at 11:05:14AM +0100, David Hildenbrand wrote:
> 
> Let's try to merge init_unavailable_memory() into memmap_init().
> Than it'll be able to set zone/nid for those nasty pfns that BIOS
> decided to keep to itself, like in Andrea's case and will also take care
> of struct pages that do not really have a frame in DRAM, but are there
> because of arbitrary section size.

Did you have a chance to check if this solves your issue?
If yes, I'll resend this as a formal patch.

> Something like this:
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index eaa227a479e4..072e94042a11 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6185,24 +6185,84 @@ static void __meminit zone_init_free_lists(struct zone *zone)
>  	}
>  }
>  
> -void __meminit __weak memmap_init(unsigned long size, int nid,
> -				  unsigned long zone,
> -				  unsigned long range_start_pfn)
> +#if !defined(CONFIG_FLAT_NODE_MEM_MAP)
> +/*
> + * Only struct pages that are backed by physical memory available to the
> + * kernel are zeroed and initialized by memmap_init_zone().
> + * But, there are some struct pages that are either reserved by firmware or
> + * do not correspond to physical page frames becuase actual memory bank is
> + * not a multiple of SECTION_SIZE. Fields of those struct pages may be
> + * accessed (for example page_to_pfn() on some configuration accesses
> + * flags) so we must explicitly initialize those struct pages.
> + */
> +static u64 __init init_unavailable_range(unsigned long spfn, unsigned long epfn,
> +					 int zone, int node)
>  {
> -	unsigned long start_pfn, end_pfn;
> +	unsigned long pfn;
> +	u64 pgcnt = 0;
> +
> +	for (pfn = spfn; pfn < epfn; pfn++) {
> +		if (!pfn_valid(ALIGN_DOWN(pfn, pageblock_nr_pages))) {
> +			pfn = ALIGN_DOWN(pfn, pageblock_nr_pages)
> +				+ pageblock_nr_pages - 1;
> +			continue;
> +		}
> +		__init_single_page(pfn_to_page(pfn), pfn, zone, node);
> +		__SetPageReserved(pfn_to_page(pfn));
> +		pgcnt++;
> +	}
> +
> +	return pgcnt;
> +}
> +#else
> +static inline u64 init_unavailable_range(unsigned long spfn, unsigned long epfn,
> +					 int zone, int node)
> +{
> +	return 0;
> +}
> +#endif
> +
> +void __init __weak memmap_init(unsigned long size, int nid,
> +			       unsigned long zone,
> +			       unsigned long range_start_pfn)
> +{
> +	unsigned long start_pfn, end_pfn, next_pfn = 0;
>  	unsigned long range_end_pfn = range_start_pfn + size;
> +	u64 pgcnt = 0;
>  	int i;
>  
>  	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
>  		start_pfn = clamp(start_pfn, range_start_pfn, range_end_pfn);
>  		end_pfn = clamp(end_pfn, range_start_pfn, range_end_pfn);
> +		next_pfn = clamp(next_pfn, range_start_pfn, range_end_pfn);
>  
>  		if (end_pfn > start_pfn) {
>  			size = end_pfn - start_pfn;
>  			memmap_init_zone(size, nid, zone, start_pfn,
>  					 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
>  		}
> +
> +		if (next_pfn < start_pfn)
> +			pgcnt += init_unavailable_range(next_pfn, start_pfn,
> +							zone, nid);
> +		next_pfn = end_pfn;
>  	}
> +
> +	/*
> +	 * Early sections always have a fully populated memmap for the whole
> +	 * section - see pfn_valid(). If the last section has holes at the
> +	 * end and that section is marked "online", the memmap will be
> +	 * considered initialized. Make sure that memmap has a well defined
> +	 * state.
> +	 */
> +	if (next_pfn < range_end_pfn)
> +		pgcnt += init_unavailable_range(next_pfn, range_end_pfn,
> +						zone, nid);
> +
> +	if (pgcnt)
> +		pr_info("%s: Zeroed struct page in unavailable ranges: %lld\n",
> +			zone_names[zone], pgcnt);
> +
>  }
>  
>  static int zone_batchsize(struct zone *zone)
> @@ -6995,88 +7055,6 @@ void __init free_area_init_memoryless_node(int nid)
>  	free_area_init_node(nid);
>  }
>  
> -#if !defined(CONFIG_FLAT_NODE_MEM_MAP)
> -/*
> - * Initialize all valid struct pages in the range [spfn, epfn) and mark them
> - * PageReserved(). Return the number of struct pages that were initialized.
> - */
> -static u64 __init init_unavailable_range(unsigned long spfn, unsigned long epfn)
> -{
> -	unsigned long pfn;
> -	u64 pgcnt = 0;
> -
> -	for (pfn = spfn; pfn < epfn; pfn++) {
> -		if (!pfn_valid(ALIGN_DOWN(pfn, pageblock_nr_pages))) {
> -			pfn = ALIGN_DOWN(pfn, pageblock_nr_pages)
> -				+ pageblock_nr_pages - 1;
> -			continue;
> -		}
> -		/*
> -		 * Use a fake node/zone (0) for now. Some of these pages
> -		 * (in memblock.reserved but not in memblock.memory) will
> -		 * get re-initialized via reserve_bootmem_region() later.
> -		 */
> -		__init_single_page(pfn_to_page(pfn), pfn, 0, 0);
> -		__SetPageReserved(pfn_to_page(pfn));
> -		pgcnt++;
> -	}
> -
> -	return pgcnt;
> -}
> -
> -/*
> - * Only struct pages that are backed by physical memory are zeroed and
> - * initialized by going through __init_single_page(). But, there are some
> - * struct pages which are reserved in memblock allocator and their fields
> - * may be accessed (for example page_to_pfn() on some configuration accesses
> - * flags). We must explicitly initialize those struct pages.
> - *
> - * This function also addresses a similar issue where struct pages are left
> - * uninitialized because the physical address range is not covered by
> - * memblock.memory or memblock.reserved. That could happen when memblock
> - * layout is manually configured via memmap=, or when the highest physical
> - * address (max_pfn) does not end on a section boundary.
> - */
> -static void __init init_unavailable_mem(void)
> -{
> -	phys_addr_t start, end;
> -	u64 i, pgcnt;
> -	phys_addr_t next = 0;
> -
> -	/*
> -	 * Loop through unavailable ranges not covered by memblock.memory.
> -	 */
> -	pgcnt = 0;
> -	for_each_mem_range(i, &start, &end) {
> -		if (next < start)
> -			pgcnt += init_unavailable_range(PFN_DOWN(next),
> -							PFN_UP(start));
> -		next = end;
> -	}
> -
> -	/*
> -	 * Early sections always have a fully populated memmap for the whole
> -	 * section - see pfn_valid(). If the last section has holes at the
> -	 * end and that section is marked "online", the memmap will be
> -	 * considered initialized. Make sure that memmap has a well defined
> -	 * state.
> -	 */
> -	pgcnt += init_unavailable_range(PFN_DOWN(next),
> -					round_up(max_pfn, PAGES_PER_SECTION));
> -
> -	/*
> -	 * Struct pages that do not have backing memory. This could be because
> -	 * firmware is using some of this memory, or for some other reasons.
> -	 */
> -	if (pgcnt)
> -		pr_info("Zeroed struct page in unavailable ranges: %lld pages", pgcnt);
> -}
> -#else
> -static inline void __init init_unavailable_mem(void)
> -{
> -}
> -#endif /* !CONFIG_FLAT_NODE_MEM_MAP */
> -
>  #if MAX_NUMNODES > 1
>  /*
>   * Figure out the number of possible node ids.
> @@ -7500,7 +7478,6 @@ void __init free_area_init(unsigned long *max_zone_pfn)
>  	/* Initialise every node */
>  	mminit_verify_pageflags_layout();
>  	setup_nr_node_ids();
> -	init_unavailable_mem();
>  	for_each_online_node(nid) {
>  		pg_data_t *pgdat = NODE_DATA(nid);
>  		free_area_init_node(nid);
> 
> 
> > I agree that this is sub-optimal, as such pages are impossible to detect
> > (PageReserved is just not clear as discussed with Andrea). The basic
> > question is how we want to proceed:
> > 
> > a) Make sure any online struct page has a valid nid/zid, and is spanned
> > by the nid/zid.
> > b) Use a fake nid that will bail out when used for page_zone() and
> > page_pgdat(), and make pfn walkers detect that.
> > 
> > AFAIU, Andrea seems to prefer a). I thing b) might be easier in corner
> > cases. Thoughts?
> 
> I'd also prefer (a). 
> 
> The hardware defines what physical addresses correspond to which node,
> so for any populated DIMM (or soldered DRAM for that matter) we can
> detect page <-> node relationship.
> 
> As for the struct pages that just "hang" in the end of a section (your
> example with 4000M), the addresses of these pages still obey the same
> rules, so again, we have page <-> node correspondence.
> 
> The zones are software construct but they also correspond to some
> hardware defined restrictions - each zone has a maximal PFN that HW
> allows. Here again we can always know which zone spans that or another
> page.
> 
> We'd have to fix a couple of things to get there, though :)
> 
> > ---
> > 
> > The tricky thing for b) is to fix all call places that do a
> > page_zone()/page_pgdat() without checking if they are allowed to do so.
> > We would have to audit all callers of page_zone() / page_zone_id() /
> > page_pgdat() ...
> > 
> > 
> > E.g., using a pattern like
> > mm/memory_hotplug.c:find_biggest_section_pfn() is fine
> > 
> > 
> > if (unlikely(!pfn_to_online_page(pfn)))
> > 	continue;
> > if (unlikely(pfn_to_nid(pfn) != nid))
> > 	continue;
> > 
> > 
> > e.g., mm/page_alloc.c:__pageblock_pfn_to_page() would have to be adjusted
> > 
> > start_page = pfn_to_online_page(start_pfn);
> > if (!start_page)
> > 	return NULL;
> > 
> > // necessary to avoid ...
> > if (page_to_nid(page) == NODE_INVALID)
> > 	return NULL;
> > 
> > // page_zone() from bailing out
> > if (page_zone(start_page) != zone)
> > 	return NULL;
> > 
> > -- 
> > Thanks,
> > 
> > David / dhildenb
> > 
> 
> -- 
> Sincerely yours,
> Mike.

-- 
Sincerely yours,
Mike.