Re: [PATCH v2 03/10] mm/memory_hotplug: simplify page offlining

[David Hildenbrand <david@redhat.com>]

> Am 03.09.2020 um 23:58 schrieb Andrew Morton <akpm@linux-foundation.org>:
> 
> On Wed, 19 Aug 2020 19:59:50 +0200 David Hildenbrand <david@redhat.com> wrote:
> 
>> We make sure that we cannot have any memory holes right at the beginning
>> of offline_pages(). We no longer need walk_system_ram_range() and can
>> call test_pages_isolated() and __offline_isolated_pages() directly.
>> 
>> offlined_pages always corresponds to nr_pages, so we can simplify that.
> 
> This patch ran afoul of Pavel's "mm/memory_hotplug: drain per-cpu pages
> again during memory offline", here:
> 
>> @@ -1481,7 +1459,7 @@ static int count_system_ram_pages_cb(unsigned long start_pfn,
>> int __ref offline_pages(unsigned long start_pfn, unsigned long nr_pages)
>> {
>>    const unsigned long end_pfn = start_pfn + nr_pages;
>> -    unsigned long pfn, system_ram_pages = 0, offlined_pages = 0;
>> +    unsigned long pfn, system_ram_pages = 0;
>>    int ret, node, nr_isolate_pageblock;
>>    unsigned long flags;
>>    struct zone *zone;
>> @@ -1579,16 +1557,12 @@ int __ref offline_pages(unsigned long start_pfn, unsigned long nr_pages)
>>            reason = "failure to dissolve huge pages";
>>            goto failed_removal_isolated;
>>        }
>> -        /* check again */
>> -        ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn,
>> -                        NULL, check_pages_isolated_cb);
>> -    } while (ret);
>> -
>> -    /* Ok, all of our target is isolated.
>> -       We cannot do rollback at this point. */
>> -    walk_system_ram_range(start_pfn, end_pfn - start_pfn,
>> -                  &offlined_pages, offline_isolated_pages_cb);
>> -    pr_info("Offlined Pages %ld\n", offlined_pages);
>> +    } while (test_pages_isolated(start_pfn, end_pfn, MEMORY_OFFLINE));
>> +
>> +    /* Mark all sections offline and remove free pages from the buddy. */
>> +    __offline_isolated_pages(start_pfn, end_pfn);
>> +    pr_info("Offlined Pages %ld\n", nr_pages);
>> +
>>    /*
>>     * Onlining will reset pagetype flags and makes migrate type
> 
> I did this.  Looks OK?
> 

Reading on my smartphone, it looks like you squashed both patches?

> From: David Hildenbrand <david@redhat.com>
> Subject: mm/memory_hotplug: simplify page offlining
> 
> We make sure that we cannot have any memory holes right at the beginning
> of offline_pages().  We no longer need walk_system_ram_range() and can
> call test_pages_isolated() and __offline_isolated_pages() directly.
> 
> offlined_pages always corresponds to nr_pages, so we can simplify that.
> 
> Link: https://lkml.kernel.org/r/20200819175957.28465-4-david@redhat.com
> Signed-off-by: David Hildenbrand <david@redhat.com>
> Acked-by: Michal Hocko <mhocko@suse.com>
> Reviewed-by: Oscar Salvador <osalvador@suse.de>
> Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
> Cc: Baoquan He <bhe@redhat.com>
> Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
> Cc: Charan Teja Reddy <charante@codeaurora.org>
> Cc: Dan Williams <dan.j.williams@intel.com>
> Cc: Fenghua Yu <fenghua.yu@intel.com>
> Cc: Logan Gunthorpe <logang@deltatee.com>
> Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Mel Gorman <mgorman@techsingularity.net>
> Cc: Michel Lespinasse <walken@google.com>
> Cc: Mike Rapoport <rppt@kernel.org>
> Cc: Tony Luck <tony.luck@intel.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
> mm/memory_hotplug.c |   61 +++++++++++++++++-------------------------
> 1 file changed, 25 insertions(+), 36 deletions(-)
> 
> --- a/mm/memory_hotplug.c~mm-memory_hotplug-simplify-page-offlining
> +++ a/mm/memory_hotplug.c
> @@ -1383,28 +1383,6 @@ do_migrate_range(unsigned long start_pfn
>    return ret;
> }
> 
> -/* Mark all sections offline and remove all free pages from the buddy. */
> -static int
> -offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages,
> -            void *data)
> -{
> -    unsigned long *offlined_pages = (unsigned long *)data;
> -
> -    *offlined_pages += __offline_isolated_pages(start, start + nr_pages);
> -    return 0;
> -}
> -
> -/*
> - * Check all pages in range, recorded as memory resource, are isolated.
> - */
> -static int
> -check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages,
> -            void *data)
> -{
> -    return test_pages_isolated(start_pfn, start_pfn + nr_pages,
> -                   MEMORY_OFFLINE);
> -}
> -
> static int __init cmdline_parse_movable_node(char *p)
> {
>    movable_node_enabled = true;
> @@ -1491,7 +1469,7 @@ static int count_system_ram_pages_cb(uns
> int __ref offline_pages(unsigned long start_pfn, unsigned long nr_pages)
> {
>    const unsigned long end_pfn = start_pfn + nr_pages;
> -    unsigned long pfn, system_ram_pages = 0, offlined_pages = 0;
> +    unsigned long pfn, system_ram_pages = 0;
>    int ret, node, nr_isolate_pageblock;
>    unsigned long flags;
>    struct zone *zone;
> @@ -1589,16 +1567,27 @@ int __ref offline_pages(unsigned long st
>            reason = "failure to dissolve huge pages";
>            goto failed_removal_isolated;
>        }
> -        /* check again */
> -        ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn,
> -                        NULL, check_pages_isolated_cb);
> -    } while (ret);
> -
> -    /* Ok, all of our target is isolated.
> -       We cannot do rollback at this point. */
> -    walk_system_ram_range(start_pfn, end_pfn - start_pfn,
> -                  &offlined_pages, offline_isolated_pages_cb);
> -    pr_info("Offlined Pages %ld\n", offlined_pages);
> +
> +        /*
> +         * per-cpu pages are drained in start_isolate_page_range, but if
> +         * there are still pages that are not free, make sure that we
> +         * drain again, because when we isolated range we might
> +         * have raced with another thread that was adding pages to pcp
> +         * list.
> +         *
> +         * Forward progress should be still guaranteed because
> +         * pages on the pcp list can only belong to MOVABLE_ZONE
> +         * because has_unmovable_pages explicitly checks for
> +         * PageBuddy on freed pages on other zones.
> +         */
> +        if (ret)
> +            drain_all_pages(zone);
> +    } while (test_pages_isolated(start_pfn, end_pfn, MEMORY_OFFLINE));

I think we have to do

ret = test_pages_isolated()
if (ret)
...
} while (ret);

So keeping the old code flow. I cannot resend before next Tuesday.

> +
> +    /* Mark all sections offline and remove free pages from the buddy. */
> +    __offline_isolated_pages(start_pfn, end_pfn);
> +    pr_info("Offlined Pages %ld\n", nr_pages);
> +
>    /*
>     * Onlining will reset pagetype flags and makes migrate type
>     * MOVABLE, so just need to decrease the number of isolated
> @@ -1609,11 +1598,11 @@ int __ref offline_pages(unsigned long st
>    spin_unlock_irqrestore(&zone->lock, flags);
> 
>    /* removal success */
> -    adjust_managed_page_count(pfn_to_page(start_pfn), -offlined_pages);
> -    zone->present_pages -= offlined_pages;
> +    adjust_managed_page_count(pfn_to_page(start_pfn), -nr_pages);
> +    zone->present_pages -= nr_pages;
> 
>    pgdat_resize_lock(zone->zone_pgdat, &flags);
> -    zone->zone_pgdat->node_present_pages -= offlined_pages;
> +    zone->zone_pgdat->node_present_pages -= nr_pages;
>    pgdat_resize_unlock(zone->zone_pgdat, &flags);
> 
>    init_per_zone_wmark_min();
> _
>