Re: [PATCH v3] mm/page_alloc: simplify lowmem_reserve max calculation

[Zi Yan <ziy@nvidia.com>]

On 16 Aug 2025, at 3:27, Andrew Morton wrote:

> On Fri, 15 Aug 2025 10:45:09 +0800 Ye Liu <ye.liu@linux.dev> wrote:
>
>> From: Ye Liu <liuye@kylinos.cn>
>>
>> Use max() to find the maximum lowmem_reserve value and min_t() to
>> cap it to managed_pages in calculate_totalreserve_pages(), instead
>> of open-coding the comparisons. No functional change.
>>
>> ...
>>
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -6235,16 +6235,13 @@ static void calculate_totalreserve_pages(void)
>>  			unsigned long managed_pages = zone_managed_pages(zone);
>>
>>  			/* Find valid and maximum lowmem_reserve in the zone */
>> -			for (j = i; j < MAX_NR_ZONES; j++) {
>> -				if (zone->lowmem_reserve[j] > max)
>> -					max = zone->lowmem_reserve[j];
>> -			}
>> +			for (j = i; j < MAX_NR_ZONES; j++)
>> +				max = max(max, zone->lowmem_reserve[j]);
>>
>>  			/* we treat the high watermark as reserved pages. */
>>  			max += high_wmark_pages(zone);
>>
>> -			if (max > managed_pages)
>> -				max = managed_pages;
>> +			max = min_t(unsigned long, max, managed_pages);
>>
>
> Use of max_t/min_t is usually a sign that we messed up the type choices
> somewhere.
>
> In this case, I'd say that zone.lowmem_reserve[] should have been
> ulong.  Oh well.

It was, but then changed into long because in __zone_watermark_ok() it is compared
to can-be-negative free_pages (see commit 3484b2de9499). Maybe it is better to revert
that change and convert ulong lowmem_reserve to long in __zone_watermark_ok()
for the comparison.

+Mel, for his opinion on it, since he made the change in 3484b2de9499.

--
Best Regards,
Yan, Zi