mm/smaps: simplify shmem handling of pte holes
authorPeter Xu <peterx@redhat.com>
Fri, 5 Nov 2021 20:36:05 +0000 (13:36 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 6 Nov 2021 20:30:33 +0000 (13:30 -0700)
Firstly, check_shmem_swap variable is actually not necessary, because
it's always set with pte_hole hook; checking each would work.

Meanwhile, the check within smaps_pte_entry is not easy to follow.
E.g., pte_none() check is not needed as "!pte_present && !is_swap_pte"
is the same.  Since at it, use the pte_hole() helper rather than dup the
page cache lookup.

Still keep the CONFIG_SHMEM part so the code can be optimized to nop for
!SHMEM.

There will be a very slight functional change in smaps_pte_entry(), that
for !SHMEM we'll return early for pte_none (before checking page==NULL),
but that's even nicer.

Link: https://lkml.kernel.org/r/20210917164756.8586-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/task_mmu.c

index 2197f66..ad667db 100644 (file)
@@ -397,7 +397,6 @@ struct mem_size_stats {
        u64 pss_shmem;
        u64 pss_locked;
        u64 swap_pss;
-       bool check_shmem_swap;
 };
 
 static void smaps_page_accumulate(struct mem_size_stats *mss,
@@ -490,6 +489,16 @@ static int smaps_pte_hole(unsigned long addr, unsigned long end,
 #define smaps_pte_hole         NULL
 #endif /* CONFIG_SHMEM */
 
+static void smaps_pte_hole_lookup(unsigned long addr, struct mm_walk *walk)
+{
+#ifdef CONFIG_SHMEM
+       if (walk->ops->pte_hole) {
+               /* depth is not used */
+               smaps_pte_hole(addr, addr + PAGE_SIZE, 0, walk);
+       }
+#endif
+}
+
 static void smaps_pte_entry(pte_t *pte, unsigned long addr,
                struct mm_walk *walk)
 {
@@ -518,12 +527,8 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr,
                        }
                } else if (is_pfn_swap_entry(swpent))
                        page = pfn_swap_entry_to_page(swpent);
-       } else if (unlikely(IS_ENABLED(CONFIG_SHMEM) && mss->check_shmem_swap
-                                                       && pte_none(*pte))) {
-               page = xa_load(&vma->vm_file->f_mapping->i_pages,
-                                               linear_page_index(vma, addr));
-               if (xa_is_value(page))
-                       mss->swap += PAGE_SIZE;
+       } else {
+               smaps_pte_hole_lookup(addr, walk);
                return;
        }
 
@@ -737,8 +742,6 @@ static void smap_gather_stats(struct vm_area_struct *vma,
                return;
 
 #ifdef CONFIG_SHMEM
-       /* In case of smaps_rollup, reset the value from previous vma */
-       mss->check_shmem_swap = false;
        if (vma->vm_file && shmem_mapping(vma->vm_file->f_mapping)) {
                /*
                 * For shared or readonly shmem mappings we know that all
@@ -756,7 +759,6 @@ static void smap_gather_stats(struct vm_area_struct *vma,
                                        !(vma->vm_flags & VM_WRITE))) {
                        mss->swap += shmem_swapped;
                } else {
-                       mss->check_shmem_swap = true;
                        ops = &smaps_shmem_walk_ops;
                }
        }