Re: [PATCH v4 4/4] mm: introduce MADV_PAGEOUT

From: Johannes Weiner <hannes@cmpxchg.org>
To: Minchan Kim <minchan@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	linux-mm <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>,
	linux-api@vger.kernel.org, Michal Hocko <mhocko@suse.com>,
	Tim Murray <timmurray@google.com>,
	Joel Fernandes <joel@joelfernandes.org>,
	Suren Baghdasaryan <surenb@google.com>,
	Daniel Colascione <dancol@google.com>,
	Shakeel Butt <shakeelb@google.com>,
	Sonny Rao <sonnyrao@google.com>,
	oleksandr@redhat.com, hdanton@sina.com, lizeb@google.com,
	Dave Hansen <dave.hansen@intel.com>,
	"Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Subject: Re: [PATCH v4 4/4] mm: introduce MADV_PAGEOUT
Date: Thu, 11 Jul 2019 14:42:23 -0400	[thread overview]
Message-ID: <20190711184223.GD20341@cmpxchg.org> (raw)
In-Reply-To: <20190711012528.176050-5-minchan@kernel.org>

On Thu, Jul 11, 2019 at 10:25:28AM +0900, Minchan Kim wrote:
> @@ -480,6 +482,198 @@ static long madvise_cold(struct vm_area_struct *vma,
>  	return 0;
>  }
>  
> +static int madvise_pageout_pte_range(pmd_t *pmd, unsigned long addr,
> +				unsigned long end, struct mm_walk *walk)
> +{
> +	struct mmu_gather *tlb = walk->private;
> +	struct mm_struct *mm = tlb->mm;
> +	struct vm_area_struct *vma = walk->vma;
> +	pte_t *orig_pte, *pte, ptent;
> +	spinlock_t *ptl;
> +	LIST_HEAD(page_list);
> +	struct page *page;
> +	unsigned long next;
> +
> +	if (fatal_signal_pending(current))
> +		return -EINTR;
> +
> +	next = pmd_addr_end(addr, end);
> +	if (pmd_trans_huge(*pmd)) {
> +		pmd_t orig_pmd;
> +
> +		tlb_change_page_size(tlb, HPAGE_PMD_SIZE);
> +		ptl = pmd_trans_huge_lock(pmd, vma);
> +		if (!ptl)
> +			return 0;
> +
> +		orig_pmd = *pmd;
> +		if (is_huge_zero_pmd(orig_pmd))
> +			goto huge_unlock;
> +
> +		if (unlikely(!pmd_present(orig_pmd))) {
> +			VM_BUG_ON(thp_migration_supported() &&
> +					!is_pmd_migration_entry(orig_pmd));
> +			goto huge_unlock;
> +		}
> +
> +		page = pmd_page(orig_pmd);
> +		if (next - addr != HPAGE_PMD_SIZE) {
> +			int err;
> +
> +			if (page_mapcount(page) != 1)
> +				goto huge_unlock;
> +			get_page(page);
> +			spin_unlock(ptl);
> +			lock_page(page);
> +			err = split_huge_page(page);
> +			unlock_page(page);
> +			put_page(page);
> +			if (!err)
> +				goto regular_page;
> +			return 0;
> +		}
> +
> +		if (isolate_lru_page(page))
> +			goto huge_unlock;
> +
> +		if (pmd_young(orig_pmd)) {
> +			pmdp_invalidate(vma, addr, pmd);
> +			orig_pmd = pmd_mkold(orig_pmd);
> +
> +			set_pmd_at(mm, addr, pmd, orig_pmd);
> +			tlb_remove_tlb_entry(tlb, pmd, addr);
> +		}
> +
> +		ClearPageReferenced(page);
> +		test_and_clear_page_young(page);
> +		list_add(&page->lru, &page_list);
> +huge_unlock:
> +		spin_unlock(ptl);
> +		reclaim_pages(&page_list);
> +		return 0;
> +	}
> +
> +	if (pmd_trans_unstable(pmd))
> +		return 0;
> +regular_page:
> +	tlb_change_page_size(tlb, PAGE_SIZE);
> +	orig_pte = pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
> +	flush_tlb_batched_pending(mm);
> +	arch_enter_lazy_mmu_mode();
> +	for (; addr < end; pte++, addr += PAGE_SIZE) {
> +		ptent = *pte;
> +		if (!pte_present(ptent))
> +			continue;
> +
> +		page = vm_normal_page(vma, addr, ptent);
> +		if (!page)
> +			continue;
> +
> +		/*
> +		 * creating a THP page is expensive so split it only if we
> +		 * are sure it's worth. Split it if we are only owner.
> +		 */
> +		if (PageTransCompound(page)) {
> +			if (page_mapcount(page) != 1)
> +				break;
> +			get_page(page);
> +			if (!trylock_page(page)) {
> +				put_page(page);
> +				break;
> +			}
> +			pte_unmap_unlock(orig_pte, ptl);
> +			if (split_huge_page(page)) {
> +				unlock_page(page);
> +				put_page(page);
> +				pte_offset_map_lock(mm, pmd, addr, &ptl);
> +				break;
> +			}
> +			unlock_page(page);
> +			put_page(page);
> +			pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
> +			pte--;
> +			addr -= PAGE_SIZE;
> +			continue;
> +		}
> +
> +		VM_BUG_ON_PAGE(PageTransCompound(page), page);
> +
> +		if (isolate_lru_page(page))
> +			continue;
> +
> +		if (pte_young(ptent)) {
> +			ptent = ptep_get_and_clear_full(mm, addr, pte,
> +							tlb->fullmm);
> +			ptent = pte_mkold(ptent);
> +			set_pte_at(mm, addr, pte, ptent);
> +			tlb_remove_tlb_entry(tlb, pte, addr);
> +		}
> +		ClearPageReferenced(page);
> +		test_and_clear_page_young(page);
> +		list_add(&page->lru, &page_list);
> +	}
> +
> +	arch_leave_lazy_mmu_mode();
> +	pte_unmap_unlock(orig_pte, ptl);
> +	reclaim_pages(&page_list);
> +	cond_resched();
> +
> +	return 0;
> +}

I know you have briefly talked about code sharing already.

While I agree that sharing with MADV_FREE is maybe a stretch, I
applied these patches and compared the pageout and the cold page table
functions, and they are line for line the same EXCEPT for 2-3 lines at
the very end, where one reclaims and the other deactivates. It would
be good to share here, it shouldn't be hard or result in fragile code.

Something like int madvise_cold_or_pageout_range(..., bool pageout)?