diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index e35a0398db63..865ad026cab3 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -856,6 +856,39 @@ static void __show_smap(struct seq_file *m, const struct mem_size_stats *mss,
 	seq_puts(m, " kB\n");
 }
 
+static void anon_vma_dump_recursive(struct seq_file *m,
+				    struct vm_area_struct *vma,
+				    struct anon_vma *parent,
+				    int level)
+{
+	struct anon_vma_chain *avc;
+	int idx = -1;
+	int i;
+
+	list_for_each_entry_reverse(avc, &vma->anon_vma_chain, same_vma) {
+		struct anon_vma *av = avc->anon_vma;
+		struct anon_vma_chain *avc2;
+		unsigned long anon_vma_count = 0;
+
+		idx++;
+		if ((parent == NULL) ? (av != vma->anon_vma->root) :
+		    (av->parent != parent || av == parent))
+			continue;
+
+		anon_vma_interval_tree_foreach(avc2, &av->rb_root, 0, -(pgoff_t)1) {
+			anon_vma_count++;
+		}
+
+		seq_puts(m, (av == vma->anon_vma) ? " # " : "   ");
+		for (i=0; i<level; i++)
+			seq_puts(m, "  ");
+		seq_printf(m, "[%3d] %px root=%px refcount=%d anon_vmas=%lu num_children=%lu num_active_vmas=%lu parent=%px\n",
+			   idx, av, av->root, atomic_read(&av->refcount),
+			   anon_vma_count, av->num_children, av->num_active_vmas, av->parent);
+		anon_vma_dump_recursive(m, vma, av, level+1);
+	}
+}
+
 static int show_smap(struct seq_file *m, void *v)
 {
 	struct vm_area_struct *vma = v;
@@ -881,6 +914,14 @@ static int show_smap(struct seq_file *m, void *v)
 		seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
 	show_smap_vma_flags(m, vma);
 
+	/* dump anon_vma information */
+	if (READ_ONCE(vma->anon_vma)) {
+		seq_printf(m, "AnonVMAs:\n");
+		anon_vma_lock_write(vma->anon_vma);
+		anon_vma_dump_recursive(m, vma, NULL, 0);
+		anon_vma_unlock_write(vma->anon_vma);
+	}
+
 	return 0;
 }
 
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 5cb401aa2b9d..f25e409eb1f1 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -19,6 +19,7 @@
 #include <linux/page_table_check.h>
 #include <linux/swapops.h>
 #include <linux/shmem_fs.h>
+#include <linux/delay.h>
 
 #include <asm/tlb.h>
 #include <asm/pgalloc.h>
@@ -1619,6 +1620,9 @@ static int retract_page_tables(struct address_space *mapping, pgoff_t pgoff,
 	struct vm_area_struct *vma;
 	int target_result = SCAN_FAIL;
 
+	pr_warn("%s: retract_page_tables(), cc->is_khugepaged=%d, target_mm=%px\n",
+		__func__, cc->is_khugepaged, target_mm);
+
 	i_mmap_lock_write(mapping);
 	vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
 		int result = SCAN_FAIL;
@@ -1645,20 +1649,37 @@ static int retract_page_tables(struct address_space *mapping, pgoff_t pgoff,
 		 * the anon_vma.
 		 */
 		if (vma->anon_vma) {
+			pr_warn("%s: skipping vma at 0x%lx in mm=%px because anon_vma\n",
+				__func__, vma->vm_start, vma->vm_mm);
 			result = SCAN_PAGE_ANON;
 			goto next;
 		}
 		addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
 		if (addr & ~HPAGE_PMD_MASK ||
 		    vma->vm_end < addr + HPAGE_PMD_SIZE) {
+			pr_warn("%s: skipping vma at 0x%lx in mm=%px because align\n",
+				__func__, vma->vm_start, vma->vm_mm);
 			result = SCAN_VMA_CHECK;
 			goto next;
 		}
 		mm = vma->vm_mm;
 		is_target = mm == target_mm && addr == target_addr;
 		result = find_pmd_or_thp_or_none(mm, addr, &pmd);
-		if (result != SCAN_SUCCEED)
+		if (result != SCAN_SUCCEED) {
+			pr_warn("%s: skipping vma at 0x%lx in mm=%px because no PMD\n",
+				__func__, vma->vm_start, vma->vm_mm);
 			goto next;
+		}
+
+		if (cc->is_khugepaged && mm == target_mm && vma->vm_start == 0x2400000-0x1000) {
+			pr_warn("%s: sleeping before mmap lock...\n", __func__);
+			mdelay(3000);
+			pr_warn("%s: done sleeping\n", __func__);
+		} else {
+			pr_warn("%s: running on vma at 0x%lx in mm=%px\n",
+				__func__, vma->vm_start, vma->vm_mm);
+		}
+
 		/*
 		 * We need exclusive mmap_lock to retract page table.
 		 *
@@ -1688,7 +1709,9 @@ static int retract_page_tables(struct address_space *mapping, pgoff_t pgoff,
 				result = SCAN_PTE_UFFD_WP;
 				goto unlock_next;
 			}
+			pr_warn("%s: about to call collapse_and_free_pmd()\n", __func__);
 			collapse_and_free_pmd(mm, vma, addr, pmd);
+			pr_warn("%s: done with collapse_and_free_pmd()\n", __func__);
 			if (!cc->is_khugepaged && is_target)
 				result = set_huge_pmd(vma, addr, pmd, hpage);
 			else
diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
index 93e13fc17d3c..771a0124329b 100644
--- a/mm/page_vma_mapped.c
+++ b/mm/page_vma_mapped.c
@@ -4,6 +4,7 @@
 #include <linux/hugetlb.h>
 #include <linux/swap.h>
 #include <linux/swapops.h>
+#include <linux/delay.h>
 
 #include "internal.h"
 
@@ -16,6 +17,19 @@ static inline bool not_found(struct page_vma_mapped_walk *pvmw)
 static bool map_pte(struct page_vma_mapped_walk *pvmw)
 {
 	pvmw->pte = pte_offset_map(pvmw->pmd, pvmw->address);
+
+	if (strcmp(current->comm, "TEST1") == 0) {
+		pr_warn("%s: executing under TEST1 pvmw->vma->vm_mm==%px pvmw->vma->vm_start=0x%lx\n",
+			__func__, pvmw->vma->vm_mm, pvmw->vma->vm_start);
+		if (current->mm == pvmw->vma->vm_mm || pvmw->vma->vm_start != 0x2400000-0x1000) {
+			pr_warn("%s:   ... continuing\n", __func__);
+		} else {
+			pr_warn("%s:   ... begin delay...\n", __func__);
+			mdelay(3500);
+			pr_warn("%s:   ... end delay\n", __func__);
+		}
+	}
+
 	if (!(pvmw->flags & PVMW_SYNC)) {
 		if (pvmw->flags & PVMW_MIGRATION) {
 			if (!is_swap_pte(*pvmw->pte))