On Sun 24-03-13 15:33:21, Li Zefan wrote:
> >> Thanks for identifying and fixing this.
> >>
> >> Li is right. The cache name will live long, but this is because the
> >> slab/slub caches will strdup it internally. So the actual memcg
> >> allocation is short lived.
> >
> > OK, I have totally missed that. Sorry about the confusion. Then all the
> > churn around the allocation is pointless, no?
> > What about:
> > ---
> >>From 7ed7f53bb597e8cb40d9ac91ce16142fb60f1e93 Mon Sep 17 00:00:00 2001
> > From: Michal Hocko <mhocko@xxxxxxx>
> > Date: Fri, 22 Mar 2013 10:22:54 +0100
> > Subject: [PATCH] memcg: fix memcg_cache_name() to use cgroup_name()
> >
> > As cgroup supports rename, it's unsafe to dereference dentry->d_name
> > without proper vfs locks. Fix this by using cgroup_name() rather than
> > dentry directly.
> >
> > Also open code memcg_cache_name because it is called only from
> > kmem_cache_dup which frees the returned name right after
> > kmem_cache_create_memcg makes a copy of it. Such a short-lived
> > allocation doesn't make too much sense. So replace it by a static
> > buffer as kmem_cache_dup is called with memcg_cache_mutex.
> >
>
> I doubt it's a win to add 4K to kernel text size instead of adding
> a few extra lines of code... but it's up to you.
I will leave the decision to Glauber. The updated version which uses
kmalloc for the static buffer is bellow.
> > Signed-off-by: Li Zefan <lizefan@xxxxxxxxxx>
> > Signed-off-by: Michal Hocko <mhocko@xxxxxxx>
> > ---
> > mm/memcontrol.c | 33 +++++++++++----------------------
> > 1 file changed, 11 insertions(+), 22 deletions(-)
> ...
> > static struct kmem_cache *kmem_cache_dup(struct mem_cgroup *memcg,
> > struct kmem_cache *s)
> > {
> > char *name;
> > struct kmem_cache *new;
> > + static char tmp_name[PAGE_SIZE];
> >
> > - name = memcg_cache_name(memcg, s);
> > - if (!name)
> > - return NULL;
> > + lockdep_assert_held(&memcg_cache_mutex);
> > +
> > + rcu_read_lock();
> > + tmp_name = snprintf(tmp_name, sizeof(tmp_name), "%s(%d:%s)", s->name,
> > + memcg_cache_id(memcg), cgroup_name(memcg->css.cgroup));
>
> I guess you didn't turn on CONFIG_MEMCG_KMEM?
dohh. Friday effect...
> snprintf() returns a int value.
[...]
---
>From 6f5d4c08cde5c82ac9432608adf517916ab54634 Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@xxxxxxx>
Date: Mon, 25 Mar 2013 10:04:02 +0100
Subject: [PATCH] memcg: fix memcg_cache_name() to use cgroup_name()
As cgroup supports rename, it's unsafe to dereference dentry->d_name
without proper vfs locks. Fix this by using cgroup_name() rather than
dentry directly.
Also open code memcg_cache_name because it is called only from
kmem_cache_dup which frees the returned name right after
kmem_cache_create_memcg makes a copy of it. Such a short-lived
allocation doesn't make too much sense. So replace it by a static
buffer as kmem_cache_dup is called with memcg_cache_mutex.
Signed-off-by: Li Zefan <lizefan@xxxxxxxxxx>
Signed-off-by: Michal Hocko <mhocko@xxxxxxx>
---
mm/memcontrol.c | 64 ++++++++++++++++++++++++++++---------------------------
1 file changed, 33 insertions(+), 31 deletions(-)
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 53b8201..3a75f2c 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -3214,52 +3214,54 @@ void mem_cgroup_destroy_cache(struct kmem_cache *cachep)
schedule_work(&cachep->memcg_params->destroy);
}
-static char *memcg_cache_name(struct mem_cgroup *memcg, struct kmem_cache *s)
-{
- char *name;
- struct dentry *dentry;
-
- rcu_read_lock();
- dentry = rcu_dereference(memcg->css.cgroup->dentry);
- rcu_read_unlock();
-
- BUG_ON(dentry == NULL);
-
- name = kasprintf(GFP_KERNEL, "%s(%d:%s)", s->name,
- memcg_cache_id(memcg), dentry->d_name.name);
-
- return name;
-}
+/*
+ * This lock protects updaters, not readers. We want readers to be as fast as
+ * they can, and they will either see NULL or a valid cache value. Our model
+ * allow them to see NULL, in which case the root memcg will be selected.
+ *
+ * We need this lock because multiple allocations to the same cache from a non
+ * will span more than one worker. Only one of them can create the cache.
+ */
+static DEFINE_MUTEX(memcg_cache_mutex);
+/*
+ * Called with memcg_cache_mutex held
+ */
static struct kmem_cache *kmem_cache_dup(struct mem_cgroup *memcg,
struct kmem_cache *s)
{
- char *name;
struct kmem_cache *new;
+ static char *tmp_name = NULL;
- name = memcg_cache_name(memcg, s);
- if (!name)
- return NULL;
+ lockdep_assert_held(&memcg_cache_mutex);
+
+ /*
+ * kmem_cache_create_memcg duplicates the given name and
+ * cgroup_name for this name requires RCU context.
+ * This static temporary buffer is used to prevent from
+ * pointless shortliving allocation.
+ */
+ if (!tmp_name) {
+ tmp_name = kmalloc(PAGE_SIZE, GFP_KERNEL);
+ WARN_ON_ONCE(!tmp_name);
+ if (!tmp_name)
+ return NULL;
+ }
+
+ rcu_read_lock();
+ snprintf(tmp_name, PAGE_SIZE, "%s(%d:%s)", s->name,
+ memcg_cache_id(memcg), cgroup_name(memcg->css.cgroup));
+ rcu_read_unlock();
- new = kmem_cache_create_memcg(memcg, name, s->object_size, s->align,
+ new = kmem_cache_create_memcg(memcg, tmp_name, s->object_size, s->align,
(s->flags & ~SLAB_PANIC), s->ctor, s);
if (new)
new->allocflags |= __GFP_KMEMCG;
- kfree(name);
return new;
}
-/*
- * This lock protects updaters, not readers. We want readers to be as fast as
- * they can, and they will either see NULL or a valid cache value. Our model
- * allow them to see NULL, in which case the root memcg will be selected.
- *
- * We need this lock because multiple allocations to the same cache from a non
- * will span more than one worker. Only one of them can create the cache.
- */
-static DEFINE_MUTEX(memcg_cache_mutex);
static struct kmem_cache *memcg_create_kmem_cache(struct mem_cgroup *memcg,
struct kmem_cache *cachep)
{
--
1.7.10.4
--
Michal Hocko
SUSE Labs
--
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>