Re: [PATCH rfc 3/4] mm: filemap: move __lruvec_stat_mod_folio() out of filemap_set_pte_range()

David Hildenbrand <david@xxxxxxxxxx> · Wed, 8 May 2024 11:33:45 +0200

On 07.05.24 15:12, Kefeng Wang wrote:


On 2024/5/7 19:11, David Hildenbrand wrote:
On 29.04.24 09:24, Kefeng Wang wrote:
Adding __folio_add_file_rmap_ptes() which don't update lruvec stat, it
is used in filemap_set_pte_range(), with it, lruvec stat updating is
moved into the caller, no functional changes.

Signed-off-by: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
---
   include/linux/rmap.h |  2 ++
   mm/filemap.c         | 27 ++++++++++++++++++---------
   mm/rmap.c            | 16 ++++++++++++++++
   3 files changed, 36 insertions(+), 9 deletions(-)

diff --git a/include/linux/rmap.h b/include/linux/rmap.h
index 7229b9baf20d..43014ddd06f9 100644
--- a/include/linux/rmap.h
+++ b/include/linux/rmap.h
@@ -242,6 +242,8 @@ void folio_add_anon_rmap_pmd(struct folio *,
struct page *,
           struct vm_area_struct *, unsigned long address, rmap_t flags);
   void folio_add_new_anon_rmap(struct folio *, struct vm_area_struct *,
           unsigned long address);
+int __folio_add_file_rmap_ptes(struct folio *, struct page *, int
nr_pages,
+        struct vm_area_struct *);
   void folio_add_file_rmap_ptes(struct folio *, struct page *, int
nr_pages,
           struct vm_area_struct *);
   #define folio_add_file_rmap_pte(folio, page, vma) \
diff --git a/mm/filemap.c b/mm/filemap.c
index 7019692daddd..3966b6616d02 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -3501,14 +3501,15 @@ static struct folio
*next_uptodate_folio(struct xa_state *xas,
   static void filemap_set_pte_range(struct vm_fault *vmf, struct folio
*folio,
               struct page *page, unsigned int nr, unsigned long addr,
-            unsigned long *rss)
+            unsigned long *rss, int *nr_mapped)
   {
       struct vm_area_struct *vma = vmf->vma;
       pte_t entry;
       entry = prepare_range_pte_entry(vmf, false, folio, page, nr, addr);
-    folio_add_file_rmap_ptes(folio, page, nr, vma);
+    *nr_mapped += __folio_add_file_rmap_ptes(folio, page, nr, vma);
+
       set_ptes(vma->vm_mm, addr, vmf->pte, entry, nr);
       /* no need to invalidate: a not-present page won't be cached */
@@ -3525,7 +3526,8 @@ static void filemap_set_pte_range(struct
vm_fault *vmf, struct folio *folio,
   static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
               struct folio *folio, unsigned long start,
               unsigned long addr, unsigned int nr_pages,
-            unsigned long *rss, unsigned int *mmap_miss)
+            unsigned long *rss, int *nr_mapped,
+            unsigned int *mmap_miss)
   {
       vm_fault_t ret = 0;
       struct page *page = folio_page(folio, start);
@@ -3558,7 +3560,8 @@ static vm_fault_t filemap_map_folio_range(struct
vm_fault *vmf,
           continue;
   skip:
           if (count) {
-            filemap_set_pte_range(vmf, folio, page, count, addr, rss);
+            filemap_set_pte_range(vmf, folio, page, count, addr,
+                          rss, nr_mapped);
               if (in_range(vmf->address, addr, count * PAGE_SIZE))
                   ret = VM_FAULT_NOPAGE;
           }
@@ -3571,7 +3574,8 @@ static vm_fault_t filemap_map_folio_range(struct
vm_fault *vmf,
       } while (--nr_pages > 0);
       if (count) {
-        filemap_set_pte_range(vmf, folio, page, count, addr, rss);
+        filemap_set_pte_range(vmf, folio, page, count, addr, rss,
+                      nr_mapped);
           if (in_range(vmf->address, addr, count * PAGE_SIZE))
               ret = VM_FAULT_NOPAGE;
       }
@@ -3583,7 +3587,7 @@ static vm_fault_t filemap_map_folio_range(struct
vm_fault *vmf,
   static vm_fault_t filemap_map_order0_folio(struct vm_fault *vmf,
           struct folio *folio, unsigned long addr,
-        unsigned long *rss, unsigned int *mmap_miss)
+        unsigned long *rss, int *nr_mapped, unsigned int *mmap_miss)
   {
       vm_fault_t ret = 0;
       struct page *page = &folio->page;
@@ -3606,7 +3610,7 @@ static vm_fault_t
filemap_map_order0_folio(struct vm_fault *vmf,
       if (vmf->address == addr)
           ret = VM_FAULT_NOPAGE;
-    filemap_set_pte_range(vmf, folio, page, 1, addr, rss);
+    filemap_set_pte_range(vmf, folio, page, 1, addr, rss, nr_mapped);
       return ret;
   }
@@ -3646,6 +3650,7 @@ vm_fault_t filemap_map_pages(struct vm_fault *vmf,
       folio_type = mm_counter_file(folio);
       do {
           unsigned long end;
+        int nr_mapped = 0;
           addr += (xas.xa_index - last_pgoff) << PAGE_SHIFT;
           vmf->pte += xas.xa_index - last_pgoff;
@@ -3655,11 +3660,15 @@ vm_fault_t filemap_map_pages(struct vm_fault
*vmf,
           if (!folio_test_large(folio))
               ret |= filemap_map_order0_folio(vmf,
-                    folio, addr, &rss, &mmap_miss);
+                    folio, addr, &rss, &nr_mapped,
+                    &mmap_miss);
           else
               ret |= filemap_map_folio_range(vmf, folio,
                       xas.xa_index - folio->index, addr,
-                    nr_pages, &rss, &mmap_miss);
+                    nr_pages, &rss, &nr_mapped,
+                    &mmap_miss);
+
+        __lruvec_stat_mod_folio(folio, NR_FILE_MAPPED, nr_mapped);
           folio_unlock(folio);
           folio_put(folio);
diff --git a/mm/rmap.c b/mm/rmap.c
index 2608c40dffad..55face4024f2 100644
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1452,6 +1452,22 @@ static __always_inline void
__folio_add_file_rmap(struct folio *folio,
           mlock_vma_folio(folio, vma);
   }
+int __folio_add_file_rmap_ptes(struct folio *folio, struct page *page,
+        int nr_pages, struct vm_area_struct *vma)
+{
+    int nr, nr_pmdmapped = 0;
+
+    VM_WARN_ON_FOLIO(folio_test_anon(folio), folio);
+
+    nr = __folio_add_rmap(folio, page, nr_pages, RMAP_LEVEL_PTE,
+                  &nr_pmdmapped);
+
+    /* See comments in folio_add_anon_rmap_*() */
+    if (!folio_test_large(folio))
+        mlock_vma_folio(folio, vma);
+
+    return nr;
+}

I'm not really a fan :/ It does make the code more complicated, and it
will be harder to extend if we decide to ever account differently (e.g.,
NR_SHMEM_MAPPED, additional tracking for mTHP etc).

If more different accounts, this may lead to bad scalability.

We already do it for PMD mappings.


With large folios we'll be naturally batching already here, and I do

Yes, it is batched with large folios，but our fs is ext4/tmpfs, there
are not support large folio or still upstreaming.

Okay, so that will be sorted out sooner or later.


wonder, if this is really worth for performance, or if we could find
another way of batching (let the caller activate batching and drain
afterwards) without exposing these details to the caller.

It does reduce latency when batch lruvec stat updating without large
folio, but I can't find better way, or let's wait for the large folio
support on ext4/tmpfs, I also Cced memcg maintainers in patch4 to see if
there are any other ideas.

I'm not convinced this benefit here is worth making the code more 
complicated.

Maybe we can find another way to optimize this batching in rmap code 
without having to leak these details to the callers.

For example, we could pass an optional batching structure to all rmap 
add/rel functions that would collect these stat updates. Then we could 
have one function to flush it and update the counters combined.

Such batching could be beneficial also for page unmapping/zapping where 
we might unmap various different folios in one go.

--
Cheers,

David / dhildenb








