Re: [PATCH RFC 2/4] mm/page_alloc: place pages to tail in __putback_isolated_page()

[David Hildenbrand <david@redhat.com>]

On 24.09.20 12:37, Vlastimil Babka wrote:
> On 9/16/20 8:34 PM, David Hildenbrand wrote:
>> __putback_isolated_page() already documents that pages will be placed to
>> the tail of the freelist - this is, however, not the case for
>> "order >= MAX_ORDER - 2" (see buddy_merge_likely()) - which should be
>> the case for all existing users.
> 
> I think here should be a sentence saying something along "Thus this patch
> introduces a FOP_TO_TAIL flag to really ensure moving pages to tail."

Agreed, thanks!

> 
>> This change affects two users:
>> - free page reporting
>> - page isolation, when undoing the isolation.
>>
>> This behavior is desireable for pages that haven't really been touched
>> lately, so exactly the two users that don't actually read/write page
>> content, but rather move untouched pages.
>>
>> The new behavior is especially desirable for memory onlining, where we
>> allow allocation of newly onlined pages via undo_isolate_page_range()
>> in online_pages(). Right now, we always place them to the head of the
>> free list, resulting in undesireable behavior: Assume we add
>> individual memory chunks via add_memory() and online them right away to
>> the NORMAL zone. We create a dependency chain of unmovable allocations
>> e.g., via the memmap. The memmap of the next chunk will be placed onto
>> previous chunks - if the last block cannot get offlined+removed, all
>> dependent ones cannot get offlined+removed. While this can already be
>> observed with individual DIMMs, it's more of an issue for virtio-mem
>> (and I suspect also ppc DLPAR).
>>
>> Note: If we observe a degradation due to the changed page isolation
>> behavior (which I doubt), we can always make this configurable by the
>> instance triggering undo of isolation (e.g., alloc_contig_range(),
>> memory onlining, memory offlining).
>>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
>> Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
>> Cc: Mel Gorman <mgorman@techsingularity.net>
>> Cc: Michal Hocko <mhocko@kernel.org>
>> Cc: Dave Hansen <dave.hansen@intel.com>
>> Cc: Vlastimil Babka <vbabka@suse.cz>
>> Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
>> Cc: Oscar Salvador <osalvador@suse.de>
>> Cc: Mike Rapoport <rppt@kernel.org>
>> Cc: Scott Cheloha <cheloha@linux.ibm.com>
>> Cc: Michael Ellerman <mpe@ellerman.id.au>
>> Signed-off-by: David Hildenbrand <david@redhat.com>
>> ---
>>  mm/page_alloc.c | 10 +++++++++-
>>  1 file changed, 9 insertions(+), 1 deletion(-)
>>
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index 91cefb8157dd..bba9a0f60c70 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -89,6 +89,12 @@ typedef int __bitwise fop_t;
>>   */
>>  #define FOP_SKIP_REPORT_NOTIFY	((__force fop_t)BIT(0))
>>  
>> +/*
>> + * Place the freed page to the tail of the freelist after buddy merging. Will
>> + * get ignored with page shuffling enabled.
>> + */
>> +#define FOP_TO_TAIL		((__force fop_t)BIT(1))
>> +
>>  /* prevent >1 _updater_ of zone percpu pageset ->high and ->batch fields */
>>  static DEFINE_MUTEX(pcp_batch_high_lock);
>>  #define MIN_PERCPU_PAGELIST_FRACTION	(8)
>> @@ -1040,6 +1046,8 @@ static inline void __free_one_page(struct page *page, unsigned long pfn,
>>  
>>  	if (is_shuffle_order(order))
>>  		to_tail = shuffle_pick_tail();
>> +	else if (fop_flags & FOP_TO_TAIL)
>> +		to_tail = true;
> 
> Should we really let random shuffling decision have a larger priority than
> explicit FOP_TO_TAIL request? Wei Yang mentioned that there's a call to
> shuffle_zone() anyway to process a freshly added memory, so we don't need to do
> that also during the process of addition itself? Might help with your goal of
> reducing dependencies even on systems that do have shuffling enabled?

So, we do have cases where generic_online_page() -> __free_pages_core()
isn't called (see patch #4):

generic_online_page() is used in two cases:

1. Direct memory onlining in online_pages(). Here, we call
   shuffle_zone().
2. Deferred memory onlining in memory-ballooning-like mechanisms (HyperV
   balloon and virtio-mem), when parts of a section are kept
   fake-offline to be fake-onlined later on.

While we shuffle in the fist instance the whole zone, we wouldn't
shuffle in the second case.

But maybe this should be tackled (just like when alloc_contig_free() a
large contiguous range, memory offlining failing, alloc_contig_range()
failing) by manually shuffling the zone again. That would be cleaner,
and the right thing to do when exposing large, contiguous ranges again
to the buddy.

Thanks!


-- 
Thanks,

David / dhildenb