s390/gmap: improve THP splitting

During s390_enable_sie(), we need to take care of splitting all qemu user
process THP mappings. This is currently done with follow_page(FOLL_SPLIT),
by simply iterating over all vma ranges, with PAGE_SIZE increment.

This logic is sub-optimal and can result in a lot of unnecessary overhead,
especially when using qemu and ASAN with large shadow map. Ilya reported
significant system slow-down with one CPU busy for a long time and overall
unresponsiveness.

Fix this by using walk_page_vma() and directly calling split_huge_pmd()
only for present pmds, which greatly reduces overhead.

Cc: <stable@vger.kernel.org> # v5.4+
Reported-by: Ilya Leoshkevich <iii@linux.ibm.com>
Tested-by: Ilya Leoshkevich <iii@linux.ibm.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

diff --git a/arch/s390/mm/gmap.c b/arch/s390/mm/gmap.c
index 190357f..46c1bf2 100644 (file)
--- a/arch/s390/mm/gmap.c
+++ b/arch/s390/mm/gmap.c
@@ -2485,23 +2485,36 @@ void gmap_sync_dirty_log_pmd(struct gmap *gmap, unsigned long bitmap[4],
 }
 EXPORT_SYMBOL_GPL(gmap_sync_dirty_log_pmd);
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+static int thp_split_walk_pmd_entry(pmd_t *pmd, unsigned long addr,
+                                   unsigned long end, struct mm_walk *walk)
+{
+       struct vm_area_struct *vma = walk->vma;
+
+       split_huge_pmd(vma, pmd, addr);
+       return 0;
+}
+
+static const struct mm_walk_ops thp_split_walk_ops = {
+       .pmd_entry      = thp_split_walk_pmd_entry,
+};
+
 static inline void thp_split_mm(struct mm_struct *mm)
 {
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
        struct vm_area_struct *vma;
-       unsigned long addr;
 
        for (vma = mm->mmap; vma != NULL; vma = vma->vm_next) {
-               for (addr = vma->vm_start;
-                    addr < vma->vm_end;
-                    addr += PAGE_SIZE)
-                       follow_page(vma, addr, FOLL_SPLIT);
                vma->vm_flags &= ~VM_HUGEPAGE;
                vma->vm_flags |= VM_NOHUGEPAGE;
+               walk_page_vma(vma, &thp_split_walk_ops, NULL);
        }
        mm->def_flags |= VM_NOHUGEPAGE;
-#endif
 }
+#else
+static inline void thp_split_mm(struct mm_struct *mm)
+{
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 
 /*
  * Remove all empty zero pages from the mapping for lazy refaulting