Zi Yan <zi.yan@xxxxxxxx> writes:
> From: Zi Yan <ziy@xxxxxxxxxx>
>
> do_numa_page() and do_huge_pmd_numa_page() share a lot of common code. To
> reduce redundancy, move common code to numa_migrate_prep() and rename
> the function to numa_migrate_check() to reflect its functionality.
>
> There is some code difference between do_numa_page() and
> do_huge_pmd_numa_page() before the code move:
>
> 1. do_huge_pmd_numa_page() did not check shared folios to set TNF_SHARED.
> 2. do_huge_pmd_numa_page() did not check and skip zone device folios.
>
> Signed-off-by: Zi Yan <ziy@xxxxxxxxxx>
> ---
> mm/huge_memory.c | 28 ++++++-----------
> mm/internal.h | 5 +--
> mm/memory.c | 81 +++++++++++++++++++++++-------------------------
> 3 files changed, 52 insertions(+), 62 deletions(-)
>
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 8c11d6da4b36..66d67d13e0dc 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -1670,10 +1670,10 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
> pmd_t pmd;
> struct folio *folio;
> unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
> - int nid = NUMA_NO_NODE;
> - int target_nid, last_cpupid = (-1 & LAST_CPUPID_MASK);
> + int target_nid = NUMA_NO_NODE;
> + int last_cpupid = (-1 & LAST_CPUPID_MASK);
> bool writable = false;
> - int flags = 0;
> + int flags = 0, nr_pages;
>
> vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
> if (unlikely(!pmd_same(oldpmd, *vmf->pmd))) {
> @@ -1693,21 +1693,13 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
> writable = true;
>
> folio = vm_normal_folio_pmd(vma, haddr, pmd);
> - if (!folio)
> + if (!folio || folio_is_zone_device(folio))
This change appears unrelated. Can we put it in a separate patch?
IIUC, this isn't necessary even in do_numa_page()? Because in
change_pte_range(), folio_is_zone_device() has been checked already.
But It doesn't hurt too.
> goto out_map;
>
> - /* See similar comment in do_numa_page for explanation */
> - if (!writable)
> - flags |= TNF_NO_GROUP;
> + nr_pages = folio_nr_pages(folio);
>
> - nid = folio_nid(folio);
> - /*
> - * For memory tiering mode, cpupid of slow memory page is used
> - * to record page access time. So use default value.
> - */
> - if (folio_has_cpupid(folio))
> - last_cpupid = folio_last_cpupid(folio);
> - target_nid = numa_migrate_prep(folio, vmf, haddr, nid, &flags);
> + target_nid = numa_migrate_check(folio, vmf, haddr, writable,
> + &flags, &last_cpupid);
> if (target_nid == NUMA_NO_NODE)
> goto out_map;
> if (migrate_misplaced_folio_prepare(folio, vma, target_nid)) {
> @@ -1720,8 +1712,8 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
>
> if (!migrate_misplaced_folio(folio, vma, target_nid)) {
> flags |= TNF_MIGRATED;
> - nid = target_nid;
> } else {
> + target_nid = NUMA_NO_NODE;
> flags |= TNF_MIGRATE_FAIL;
> vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
> if (unlikely(!pmd_same(oldpmd, *vmf->pmd))) {
> @@ -1732,8 +1724,8 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
> }
>
> out:
> - if (nid != NUMA_NO_NODE)
> - task_numa_fault(last_cpupid, nid, HPAGE_PMD_NR, flags);
> + if (target_nid != NUMA_NO_NODE)
> + task_numa_fault(last_cpupid, target_nid, nr_pages, flags);
This appears a behavior change. IIUC, there are 2 possible issues.
1) if migrate_misplaced_folio() fails, folio_nid() should be used as
nid. "target_nid" as variable name here is confusing, because
folio_nid() is needed in fact.
2) if !pmd_same(), task_numa_fault() should be skipped. The original
code is buggy.
Similar issues for do_numa_page().
If my understanding were correct, we should implement a separate patch
to fix 2) above. And that may need to be backported.
>
> return 0;
>
> diff --git a/mm/internal.h b/mm/internal.h
> index b4d86436565b..7782b7bb3383 100644
> --- a/mm/internal.h
> +++ b/mm/internal.h
> @@ -1217,8 +1217,9 @@ void vunmap_range_noflush(unsigned long start, unsigned long end);
>
> void __vunmap_range_noflush(unsigned long start, unsigned long end);
>
> -int numa_migrate_prep(struct folio *folio, struct vm_fault *vmf,
> - unsigned long addr, int page_nid, int *flags);
> +int numa_migrate_check(struct folio *folio, struct vm_fault *vmf,
> + unsigned long addr, bool writable,
> + int *flags, int *last_cpupid);
>
> void free_zone_device_folio(struct folio *folio);
> int migrate_device_coherent_page(struct page *page);
> diff --git a/mm/memory.c b/mm/memory.c
> index 96c2f5b3d796..a252c0f13755 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -5209,16 +5209,42 @@ static vm_fault_t do_fault(struct vm_fault *vmf)
> return ret;
> }
>
> -int numa_migrate_prep(struct folio *folio, struct vm_fault *vmf,
> - unsigned long addr, int page_nid, int *flags)
> +int numa_migrate_check(struct folio *folio, struct vm_fault *vmf,
> + unsigned long addr, bool writable,
> + int *flags, int *last_cpupid)
> {
> struct vm_area_struct *vma = vmf->vma;
>
> + /*
> + * Avoid grouping on RO pages in general. RO pages shouldn't hurt as
> + * much anyway since they can be in shared cache state. This misses
> + * the case where a mapping is writable but the process never writes
> + * to it but pte_write gets cleared during protection updates and
> + * pte_dirty has unpredictable behaviour between PTE scan updates,
> + * background writeback, dirty balancing and application behaviour.
> + */
> + if (!writable)
> + *flags |= TNF_NO_GROUP;
> +
> + /*
> + * Flag if the folio is shared between multiple address spaces. This
> + * is later used when determining whether to group tasks together
> + */
> + if (folio_likely_mapped_shared(folio) && (vma->vm_flags & VM_SHARED))
> + *flags |= TNF_SHARED;
> +
> + /*
> + * For memory tiering mode, cpupid of slow memory page is used
> + * to record page access time.
> + */
> + if (folio_has_cpupid(folio))
> + *last_cpupid = folio_last_cpupid(folio);
> +
> /* Record the current PID acceesing VMA */
> vma_set_access_pid_bit(vma);
>
> count_vm_numa_event(NUMA_HINT_FAULTS);
> - if (page_nid == numa_node_id()) {
> + if (folio_nid(folio) == numa_node_id()) {
> count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
> *flags |= TNF_FAULT_LOCAL;
> }
> @@ -5284,12 +5310,11 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
> {
> struct vm_area_struct *vma = vmf->vma;
> struct folio *folio = NULL;
> - int nid = NUMA_NO_NODE;
> + int target_nid = NUMA_NO_NODE;
> bool writable = false, ignore_writable = false;
> bool pte_write_upgrade = vma_wants_manual_pte_write_upgrade(vma);
> - int last_cpupid;
> - int target_nid;
> - pte_t pte, old_pte;
> + int last_cpupid = (-1 & LAST_CPUPID_MASK);
> + pte_t pte, old_pte = vmf->orig_pte;
> int flags = 0, nr_pages;
>
> /*
> @@ -5297,10 +5322,7 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
> * table lock, that its contents have not changed during fault handling.
> */
> spin_lock(vmf->ptl);
> - /* Read the live PTE from the page tables: */
> - old_pte = ptep_get(vmf->pte);
> -
> - if (unlikely(!pte_same(old_pte, vmf->orig_pte))) {
> + if (unlikely(!pte_same(old_pte, *vmf->pte))) {
> pte_unmap_unlock(vmf->pte, vmf->ptl);
> goto out;
> }
> @@ -5320,35 +5342,10 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
> if (!folio || folio_is_zone_device(folio))
> goto out_map;
>
> - /*
> - * Avoid grouping on RO pages in general. RO pages shouldn't hurt as
> - * much anyway since they can be in shared cache state. This misses
> - * the case where a mapping is writable but the process never writes
> - * to it but pte_write gets cleared during protection updates and
> - * pte_dirty has unpredictable behaviour between PTE scan updates,
> - * background writeback, dirty balancing and application behaviour.
> - */
> - if (!writable)
> - flags |= TNF_NO_GROUP;
> -
> - /*
> - * Flag if the folio is shared between multiple address spaces. This
> - * is later used when determining whether to group tasks together
> - */
> - if (folio_likely_mapped_shared(folio) && (vma->vm_flags & VM_SHARED))
> - flags |= TNF_SHARED;
> -
> - nid = folio_nid(folio);
> nr_pages = folio_nr_pages(folio);
> - /*
> - * For memory tiering mode, cpupid of slow memory page is used
> - * to record page access time. So use default value.
> - */
> - if (!folio_has_cpupid(folio))
> - last_cpupid = (-1 & LAST_CPUPID_MASK);
> - else
> - last_cpupid = folio_last_cpupid(folio);
> - target_nid = numa_migrate_prep(folio, vmf, vmf->address, nid, &flags);
> +
> + target_nid = numa_migrate_check(folio, vmf, vmf->address, writable,
> + &flags, &last_cpupid);
> if (target_nid == NUMA_NO_NODE)
> goto out_map;
> if (migrate_misplaced_folio_prepare(folio, vma, target_nid)) {
> @@ -5362,9 +5359,9 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
>
> /* Migrate to the requested node */
> if (!migrate_misplaced_folio(folio, vma, target_nid)) {
> - nid = target_nid;
> flags |= TNF_MIGRATED;
> } else {
> + target_nid = NUMA_NO_NODE;
> flags |= TNF_MIGRATE_FAIL;
> vmf->pte = pte_offset_map_lock(vma->vm_mm, vmf->pmd,
> vmf->address, &vmf->ptl);
> @@ -5378,8 +5375,8 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
> }
>
> out:
> - if (nid != NUMA_NO_NODE)
> - task_numa_fault(last_cpupid, nid, nr_pages, flags);
> + if (target_nid != NUMA_NO_NODE)
> + task_numa_fault(last_cpupid, target_nid, nr_pages, flags);
> return 0;
> out_map:
> /*
--
Best Regards,
Huang, Ying
