On Fri, Apr 29, 2022 at 3:15 PM Yang Shi <shy828301@xxxxxxxxx> wrote:
>
> On Thu, Apr 28, 2022 at 5:09 PM Jiaqi Yan <jiaqiyan@xxxxxxxxxx> wrote:
> >
> > Make __collapse_huge_page_copy return whether
> > collapsing/copying anonymous pages succeeded,
> > and make collapse_huge_page handle the return status.
> >
> > Break existing PTE scan loop into two for-loops.
> > The first loop copies source pages into target huge page,
> > and can fail gracefully when running into memory errors in
> > source pages. Roll back the page table and page states
> > in the 2nd loop when copying failed:
> > 1) re-establish the PTEs-to-PMD connection.
> > 2) release pages back to their LRU list.
>
> Could you please include a changelog next time? It is really helpful
> for reviewers.
Of course, I will include change logs for both v2 + v3 changes.
>
>
> >
> > Signed-off-by: Jiaqi Yan <jiaqiyan@xxxxxxxxxx>
> > ---
> > include/linux/highmem.h | 19 ++++++
> > mm/khugepaged.c | 138 ++++++++++++++++++++++++++++++----------
> > 2 files changed, 124 insertions(+), 33 deletions(-)
> >
> > diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> > index 39bb9b47fa9cd..0ccb1e92c4b06 100644
> > --- a/include/linux/highmem.h
> > +++ b/include/linux/highmem.h
> > @@ -298,6 +298,25 @@ static inline void copy_highpage(struct page *to, struct page *from)
> >
> > #endif
> >
> > +/*
> > + * Machine check exception handled version of copy_highpage.
> > + * Return true if copying page content failed; otherwise false.
> > + * Note handling #MC requires arch opt-in.
> > + */
> > +static inline bool copy_highpage_mc(struct page *to, struct page *from)
> > +{
> > + char *vfrom, *vto;
> > + unsigned long ret;
> > +
> > + vfrom = kmap_local_page(from);
> > + vto = kmap_local_page(to);
> > + ret = copy_mc_to_kernel(vto, vfrom, PAGE_SIZE);
> > + kunmap_local(vto);
> > + kunmap_local(vfrom);
> > +
> > + return ret > 0;
> > +}
> > +
> > static inline void memcpy_page(struct page *dst_page, size_t dst_off,
> > struct page *src_page, size_t src_off,
> > size_t len)
> > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > index 131492fd1148b..8e69a0640e551 100644
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -52,6 +52,7 @@ enum scan_result {
> > SCAN_CGROUP_CHARGE_FAIL,
> > SCAN_TRUNCATED,
> > SCAN_PAGE_HAS_PRIVATE,
> > + SCAN_COPY_MC,
>
> You need to update the tracepoint in
> include/trace/events/huge_memory.h to include the new result.
Will add tracepoint in v3.
>
> > };
> >
> > #define CREATE_TRACE_POINTS
> > @@ -739,44 +740,98 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
> > return 0;
> > }
> >
> > -static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> > - struct vm_area_struct *vma,
> > - unsigned long address,
> > - spinlock_t *ptl,
> > - struct list_head *compound_pagelist)
> > +/*
>
> Better to use "/**" so that it could be converted to kernel doc.
Will update in v3.
>
> > + * __collapse_huge_page_copy - attempts to copy memory contents from normal
> > + * pages to a hugepage. Cleanup the normal pages if copying succeeds;
> > + * otherwise restore the original pmd page table. Returns true if copying
> > + * succeeds, otherwise returns false.
> > + *
> > + * @pte: starting of the PTEs to copy from
> > + * @page: the new hugepage to copy contents to
> > + * @pmd: pointer to the new hugepage's PMD
> > + * @rollback: the original normal PTEs' PMD
>
> You may not need pmd and rollback, please see the below comments for the reason.
>
> > + * @address: starting address to copy
> > + * @pte_ptl: lock on normal pages' PTEs
> > + * @compound_pagelist: list that stores compound pages
> > + */
> > +static bool __collapse_huge_page_copy(pte_t *pte,
> > + struct page *page,
> > + pmd_t *pmd,
> > + pmd_t rollback,
> > + struct vm_area_struct *vma,
> > + unsigned long address,
> > + spinlock_t *pte_ptl,
> > + struct list_head *compound_pagelist)
> > {
> > struct page *src_page, *tmp;
> > pte_t *_pte;
> > - for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
> > - _pte++, page++, address += PAGE_SIZE) {
> > - pte_t pteval = *_pte;
> > + pte_t pteval;
> > + unsigned long _address;
> > + spinlock_t *pmd_ptl;
> > + bool copy_succeeded = true;
> >
> > - if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
> > + /*
> > + * Copying pages' contents is subject to memory poison at any iteration.
> > + */
> > + for (_pte = pte, _address = address;
> > + _pte < pte + HPAGE_PMD_NR;
> > + _pte++, page++, _address += PAGE_SIZE) {
> > + pteval = *_pte;
> > +
> > + if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval)))
> > clear_user_highpage(page, address);
> > - add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> > - if (is_zero_pfn(pte_pfn(pteval))) {
> > - /*
> > - * ptl mostly unnecessary.
> > - */
> > - spin_lock(ptl);
> > - ptep_clear(vma->vm_mm, address, _pte);
> > - spin_unlock(ptl);
> > + else {
> > + src_page = pte_page(pteval);
> > + if (copy_highpage_mc(page, src_page)) {
> > + copy_succeeded = false;
> > + break;
> > + }
> > + }
> > + }
> > +
> > + if (!copy_succeeded) {
> > + /*
> > + * Copying failed, re-establish the regular PMD that
> > + * points to regular page table. Since PTEs are still
> > + * isolated and locked, acquiring anon_vma_lock is unnecessary.
> > + */
> > + pmd_ptl = pmd_lock(vma->vm_mm, pmd);
> > + pmd_populate(vma->vm_mm, pmd, pmd_pgtable(rollback));
> > + spin_unlock(pmd_ptl);
> > + }
>
> I think the above section could be moved to out of
> __collapse_huge_page_copy(), just like what is done after
> __collapse_huge_page_isolate() is failed.
>
> You don't have to restore the pmd here since khugepaged holds write
> mmap_lock, there can't be page fault running in parallel. Hence you
> don't have to add pmd and rollback parameters to
> __collapse_huge_page_copy().
>
The reason I re-establish the pmd table here is to avoid anon_vma_lock_write
since here PTEs are still isolated and locked. Are you implying that, we should
do the PMD re-establishment after PTEs released? and more importantly,
with or without holding the anon_vma_lock_write?
> > +
> > + for (_pte = pte, _address = address; _pte < pte + HPAGE_PMD_NR;
> > + _pte++, _address += PAGE_SIZE) {
> > + pteval = *_pte;
> > + if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
> > + if (copy_succeeded) {
> > + add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> > + if (is_zero_pfn(pte_pfn(pteval))) {
> > + /*
> > + * ptl mostly unnecessary.
> > + */
> > + spin_lock(pte_ptl);
> > + pte_clear(vma->vm_mm, _address, _pte);
> > + spin_unlock(pte_ptl);
> > + }
> > }
> > } else {
> > src_page = pte_page(pteval);
> > - copy_user_highpage(page, src_page, address, vma);
> > if (!PageCompound(src_page))
> > release_pte_page(src_page);
> > - /*
> > - * ptl mostly unnecessary, but preempt has to
> > - * be disabled to update the per-cpu stats
> > - * inside page_remove_rmap().
> > - */
> > - spin_lock(ptl);
> > - ptep_clear(vma->vm_mm, address, _pte);
> > - page_remove_rmap(src_page, false);
> > - spin_unlock(ptl);
> > - free_page_and_swap_cache(src_page);
> > +
> > + if (copy_succeeded) {
> > + /*
> > + * ptl mostly unnecessary, but preempt has to
> > + * be disabled to update the per-cpu stats
> > + * inside page_remove_rmap().
> > + */
> > + spin_lock(pte_ptl);
> > + pte_clear(vma->vm_mm, _address, _pte);
> > + page_remove_rmap(src_page, false);
> > + spin_unlock(pte_ptl);
> > + free_page_and_swap_cache(src_page);
> > + }
> > }
> > }
> >
> > @@ -784,6 +839,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> > list_del(&src_page->lru);
> > release_pte_page(src_page);
> > }
> > +
> > + return copy_succeeded;
> > }
> >
> > static void khugepaged_alloc_sleep(void)
> > @@ -1066,6 +1123,7 @@ static void collapse_huge_page(struct mm_struct *mm,
> > struct vm_area_struct *vma;
> > struct mmu_notifier_range range;
> > gfp_t gfp;
> > + bool copied = false;
> >
> > VM_BUG_ON(address & ~HPAGE_PMD_MASK);
> >
> > @@ -1177,9 +1235,13 @@ static void collapse_huge_page(struct mm_struct *mm,
> > */
> > anon_vma_unlock_write(vma->anon_vma);
> >
> > - __collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
> > - &compound_pagelist);
> > + copied = __collapse_huge_page_copy(pte, new_page, pmd, _pmd,
> > + vma, address, pte_ptl, &compound_pagelist);
> > pte_unmap(pte);
> > + if (!copied) {
> > + result = SCAN_COPY_MC;
> > + goto out_up_write;
> > + }
> > /*
> > * spin_lock() below is not the equivalent of smp_wmb(), but
> > * the smp_wmb() inside __SetPageUptodate() can be reused to
> > @@ -1364,9 +1426,14 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
> > pte_unmap_unlock(pte, ptl);
> > if (ret) {
> > node = khugepaged_find_target_node();
> > - /* collapse_huge_page will return with the mmap_lock released */
> > - collapse_huge_page(mm, address, hpage, node,
> > - referenced, unmapped);
> > + /*
> > + * collapse_huge_page will return with the mmap_r+w_lock released.
> > + * It is uncertain if *hpage is NULL or not when collapse_huge_page
> > + * returns, so keep ret=1 to jump to breakouterloop_mmap_lock
> > + * in khugepaged_scan_mm_slot, then *hpage will be freed
> > + * if collapse failed.
> > + */
> > + collapse_huge_page(mm, address, hpage, node, referenced, unmapped);
> > }
> > out:
> > trace_mm_khugepaged_scan_pmd(mm, page, writable, referenced,
> > @@ -2168,6 +2235,11 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
> > khugepaged_scan_file(mm, file, pgoff, hpage);
> > fput(file);
> > } else {
> > + /*
> > + * mmap_read_lock is
> > + * 1) released if both scan and collapse succeeded;
> > + * 2) still held if either scan or collapse failed.
>
> The #2 doesn't look correct. Even though collapse is failed, the
> mmap_lock is released as long as scan is succeeded.
>
> IIUC the collapse does:
> read unlock (passed in by scan)
> read lock
> read unlock
> write lock
> write unlock
>
You are right! I will update this comment block in v3.
> > + */
> > ret = khugepaged_scan_pmd(mm, vma,
> > khugepaged_scan.address,
> > hpage);
> > --
> > 2.35.1.1178.g4f1659d476-goog
> >
