Useful to see whats going on.
Signed-off-by: Christoph Lameter <cl@xxxxxxxxx>
Index: linux/lib/xarray.c
===================================================================
--- linux.orig/lib/xarray.c
+++ linux/lib/xarray.c
@@ -1583,11 +1583,13 @@ void xa_object_migrate(struct xa_node *n
new_node = kmem_cache_alloc_node(radix_tree_node_cachep, GFP_KERNEL, numa_node);
+ printk(KERN_INFO "xa_object_migrate(%px, %d)\n", node, numa_node);
xa_lock_irq(xa);
/* Check again..... */
if (xa != node->array || !list_empty(&node->private_list)) {
node = new_node;
+ printk(KERN_ERR "Skip temporary object\n");
goto unlock;
}
@@ -1606,6 +1608,7 @@ void xa_object_migrate(struct xa_node *n
else
slot = &xa_parent_locked(xa, new_node)->slots[new_node->offset];
rcu_assign_pointer(*slot, xa_mk_node(new_node));
+ printk(KERN_ERR "Success\n");
unlock:
xa_unlock_irq(xa);
Index: linux/mm/slub.c
===================================================================
--- linux.orig/mm/slub.c
+++ linux/mm/slub.c
@@ -4245,6 +4245,7 @@ static void kmem_cache_move(struct page
unsigned long flags;
unsigned long objects;
+ printk(KERN_ERR "kmem_cache_move in: page=%px inuse=%d\n", page, page->inuse);
local_irq_save(flags);
slab_lock(page);
@@ -4267,6 +4268,7 @@ static void kmem_cache_move(struct page
if (test_bit(slab_index(p, s, addr), map))
vector[count++] = p;
+ printk(KERN_ERR "Vector of %d items\n", count);
if (s->isolate)
private = s->isolate(s, vector, count);
else
@@ -4295,6 +4297,7 @@ static void kmem_cache_move(struct page
* Perform callbacks to move the objects.
*/
s->migrate(s, vector, count, node, private);
+ printk(KERN_ERR "kmem_cache_move out: page=%px inuse=%d\n", page, page->inuse);
}
/*
@@ -4312,6 +4315,7 @@ static unsigned long __move(struct kmem_
LIST_HEAD(move_list);
struct kmem_cache_node *n = get_node(s, node);
+ printk(KERN_ERR "__move(%s, %d, %d, %d) migrate=%px\n", s->name, node, target_node, ratio, s->migrate);
if (node == target_node && n->nr_partial <= 1)
/*
* Trying to reduce fragmentataion on a node but there is
@@ -4322,9 +4326,16 @@ static unsigned long __move(struct kmem_
spin_lock_irqsave(&n->list_lock, flags);
list_for_each_entry_safe(page, page2, &n->partial, lru) {
- if (!slab_trylock(page))
+ printk(KERN_ERR "Slab page %px inuse=%d ", page, page->inuse);
+ if (page->inuse > 1000) {
+ printk("Page->inuse too high....\n");
+ break;
+ }
+ if (!slab_trylock(page)) {
+ printk("Locked\n");
/* Busy slab. Get out of the way */
continue;
+ }
if (page->inuse) {
if (page->inuse > ratio * page->objects / 100) {
@@ -4333,10 +4344,13 @@ static unsigned long __move(struct kmem_
* Skip slab because the object density
* in the slab page is high enough
*/
+ printk("Below ratio. Skipping\n");
continue;
}
list_move(&page->lru, &move_list);
+ printk("Added to list to move\n");
+
if (s->migrate) {
/* Remove page from being considered for allocations */
n->nr_partial--;
@@ -4345,6 +4359,7 @@ static unsigned long __move(struct kmem_
slab_unlock(page);
} else {
/* Empty slab page */
+ printk("Empty\n");
list_del(&page->lru);
n->nr_partial--;
slab_unlock(page);
@@ -4374,11 +4389,17 @@ static unsigned long __move(struct kmem_
struct page *page;
struct page *page2;
+ printk(KERN_ERR "Beginning to migrate pages\n");
if (scratch) {
/* Try to remove / move the objects left */
list_for_each_entry(page, &move_list, lru) {
- if (page->inuse)
+ if (page->inuse) {
kmem_cache_move(page, scratch, target_node);
+ if (page->inuse > 1000) {
+ printk(KERN_ERR "Page corrupted. Abort\n");
+ break;
+ }
+ }
}
kfree(scratch);
}
@@ -4404,9 +4425,11 @@ static unsigned long __move(struct kmem_
} else {
slab_unlock(page);
discard_slab(s, page);
+ printk(KERN_ERR "Freed one page %px\n", page);
}
}
spin_unlock_irqrestore(&n->list_lock, flags);
+ printk(KERN_ERR "Finished migrating slab objects\n");
}
out:
return atomic_long_read(&n->nr_slabs);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
