| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20190714231144.GB128252@google.com>
Date: Mon, 15 Jul 2019 08:11:44 +0900
From: Minchan Kim <minchan@...nel.org>
To: Johannes Weiner <hannes@...xchg.org>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
linux-mm <linux-mm@...ck.org>,
LKML <linux-kernel@...r.kernel.org>, linux-api@...r.kernel.org,
Michal Hocko <mhocko@...e.com>,
Tim Murray <timmurray@...gle.com>,
Joel Fernandes <joel@...lfernandes.org>,
Suren Baghdasaryan <surenb@...gle.com>,
Daniel Colascione <dancol@...gle.com>,
Shakeel Butt <shakeelb@...gle.com>,
Sonny Rao <sonnyrao@...gle.com>, oleksandr@...hat.com,
hdanton@...a.com, lizeb@...gle.com,
Dave Hansen <dave.hansen@...el.com>,
"Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>
Subject: Re: [PATCH v4 4/4] mm: introduce MADV_PAGEOUT
On Fri, Jul 12, 2019 at 09:58:09AM -0400, Johannes Weiner wrote:
> On Fri, Jul 12, 2019 at 02:18:28PM +0900, Minchan Kim wrote:
> > Hi Johannes,
> >
> > On Thu, Jul 11, 2019 at 02:42:23PM -0400, Johannes Weiner wrote:
> > > 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.
> >
> > Fair enough if we leave MADV_FREE.
> >
> > >
> > > Something like int madvise_cold_or_pageout_range(..., bool pageout)?
> >
> > How about this?
> >
> > From 41592f23e876ec21e49dc3c76dc89538e2bb16be Mon Sep 17 00:00:00 2001
> > From: Minchan Kim <minchan@...nel.org>
> > Date: Fri, 12 Jul 2019 14:05:36 +0900
> > Subject: [PATCH] mm: factor out common parts between MADV_COLD and
> > MADV_PAGEOUT
> >
> > There are many common parts between MADV_COLD and MADV_PAGEOUT.
> > This patch factor them out to save code duplication.
> >
> > Signed-off-by: Minchan Kim <minchan@...nel.org>
>
> This looks much better, thanks!
>
> > @@ -423,6 +445,12 @@ static int madvise_cold_pte_range(pmd_t *pmd, unsigned long addr,
> >
> > VM_BUG_ON_PAGE(PageTransCompound(page), page);
> >
> > + if (pageout) {
> > + if (isolate_lru_page(page))
> > + continue;
> > + list_add(&page->lru, &page_list);
> > + }
> > +
> > if (pte_young(ptent)) {
> > ptent = ptep_get_and_clear_full(mm, addr, pte,
> > tlb->fullmm);
>
> One thought on the ordering here.
>
> When LRU isolation fails, it would still make sense to clear the young
> bit: we cannot reclaim the page as we wanted to, but the user still
> provided a clear hint that the page is cold and she won't be touching
> it for a while. MADV_PAGEOUT is basically MADV_COLD + try_to_reclaim.
> So IMO isolation should go to the end next to deactivate_page().
Sure, I will modify MADV_PAGEOUT patch instead of refactoring one.
Thanks for the review, Johannes!
Powered by blists - more mailing lists