Re: [PATCH 1/4] mm/mm_init: use deferred_init_memmap_chunk() in deferred_grow_zone()

[Wei Yang]

On Wed, Aug 20, 2025 at 12:20:10PM +0300, Mike Rapoport wrote:
>On Tue, Aug 19, 2025 at 11:51:58PM +0000, Wei Yang wrote:
>> On Tue, Aug 19, 2025 at 01:54:46PM +0300, Mike Rapoport wrote:
>> >On Tue, Aug 19, 2025 at 09:52:23AM +0000, Wei Yang wrote:
>> >> Hi, Mike
>> >> 
>> >> After going through the code again, I have some trivial thoughts to discuss
>> >> with you. If not right, please let me know.
>> >> 
>> >> On Mon, Aug 18, 2025 at 09:46:12AM +0300, Mike Rapoport wrote:
>> >> 
>> >> In the file above this line, there is a compare between first_deferred_pfn and
>> >> its original value after grab pgdat_resize_lock.
>> >
>> >Do you mean this one:
>> >
>> >	if (first_deferred_pfn != pgdat->first_deferred_pfn) {
>> >		pgdat_resize_unlock(pgdat, &flags);
>> >		return true;
>> >	}
>> > 
>> 
>> Yes.
>> 
>> I am thinking something like this:
>> 
>>  	if (first_deferred_pfn != pgdat->first_deferred_pfn || 
>> 	    first_deferred_pfn == ULONG_MAX)
>> 
>> This means
>> 
>>   * someone else has grow zone before we grab the lock
>>   * or the whole zone has already been initialized
>
>deferred_grow_zone() can be called only before deferred_init_memmap(), so
>it's very unlikely that a zone will be completely initialized here. We
>start with at least one section with each deferred zone and every call to
>deferred_grow_zone() adds a section.
>
>And even if that was a case and first_deferred_pfn is ULONG_MAX, the loop
>below will end immediately, so I don't think additional condition here
>would be helpful.
> 

I think you are right.

>> >> I am thinking to compare first_deferred_pfn with ULONG_MAX, as it compared in
>> >> deferred_init_memmap(). This indicate this zone has already been initialized
>> >> totally.
>> >
>> >It may be another CPU ran deferred_grow_zone() and won the race for resize
>> >lock. Then pgdat->first_deferred_pfn will be larger than
>> >first_deferred_pfn, but still not entire zone would be initialized.
>> > 
>> >> Current code guard this by spfn < zone_end_pfn(zone). Maybe a check ahead
>> >> would be more clear?
>> >
>> >Not sure I follow you here. The check that we don't pass zone_end_pfn is
>> >inside the loop for every section we initialize.
>> > 
>> 
>> In case the zone has been initialized totally, first_deferred_pfn = ULONG_MAX.
>> 
>> Then we come to the loop with initial state:
>> 
>>     spfn = ULONG_MAX
>>     epfn = 0 (which is wrap around)
>> 
>> And loop condition check (spfn < zone_end_pfn(zone)) is false, so the loop is
>> skipped. This is how we handle a fully initialized zone now.
>> 
>> Would this be a little un-common?
>
>Why? The important thing is (spfn < zone_end_pfn(zone)) is false, and I
>think that's good enough.
> 

Well, no more else.

>> >> > 
>> >> >-	/* If the zone is empty somebody else may have cleared out the zone */
>> >> >-	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
>> >> >-						 first_deferred_pfn)) {
>> >> >-		pgdat->first_deferred_pfn = ULONG_MAX;
>> >> >-		pgdat_resize_unlock(pgdat, &flags);
>> >> >-		/* Retry only once. */
>> >> >-		return first_deferred_pfn != ULONG_MAX;
>> >> >+	/*
>> >> >+	 * Initialize at least nr_pages_needed in section chunks.
>> >> >+	 * If a section has less free memory than nr_pages_needed, the next
>> >> >+	 * section will be also initalized.
>> 
>> Nit, one typo here. s/initalized/initialized/
>
>Thanks, will fix.
> 
>-- 
>Sincerely yours,
>Mike.

-- 
Wei Yang
Help you, Help me