| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <1335138820-26590-8-git-send-email-glommer@parallels.com>
Date: Sun, 22 Apr 2012 20:53:36 -0300
From: Glauber Costa <glommer@...allels.com>
To: <cgroups@...r.kernel.org>
Cc: <linux-mm@...ck.org>, <linux-kernel@...r.kernel.org>,
<devel@...nvz.org>, <kamezawa.hiroyu@...fujitsu.com>,
Michal Hocko <mhocko@...e.cz>,
Johannes Weiner <hannes@...xchg.org>, fweisbec@...il.com,
Greg Thelen <gthelen@...gle.com>,
Suleiman Souhlal <suleiman@...gle.com>,
Glauber Costa <glommer@...allels.com>,
Christoph Lameter <cl@...ux.com>,
Pekka Enberg <penberg@...helsinki.fi>
Subject: [PATCH 19/23] slab: per-memcg accounting of slab caches
This patch charges allocation of a slab object to a particular
memcg.
The cache is selected with mem_cgroup_get_kmem_cache(),
which is the biggest overhead we pay here, because
it happens at all allocations. However, other than forcing
a function call, this function is not very expensive, and
try to return as soon as we realize we are not a memcg cache.
The charge/uncharge functions are heavier, but are only called
for new page allocations.
Code is heavily inspired by Suleiman's, with adaptations to
the patchset and minor simplifications by me.
Signed-off-by: Glauber Costa <glommer@...allels.com>
CC: Christoph Lameter <cl@...ux.com>
CC: Pekka Enberg <penberg@...helsinki.fi>
CC: Michal Hocko <mhocko@...e.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
CC: Johannes Weiner <hannes@...xchg.org>
CC: Suleiman Souhlal <suleiman@...gle.com>
---
include/linux/slab_def.h | 66 ++++++++++++++++++++++++++++-
mm/slab.c | 105 ++++++++++++++++++++++++++++++++++++++++++----
2 files changed, 162 insertions(+), 9 deletions(-)
diff --git a/include/linux/slab_def.h b/include/linux/slab_def.h
index 54d25d7..c4f7e45 100644
--- a/include/linux/slab_def.h
+++ b/include/linux/slab_def.h
@@ -51,7 +51,7 @@ struct kmem_cache {
void (*ctor)(void *obj);
/* 4) cache creation/removal */
- const char *name;
+ char *name;
struct list_head next;
/* 5) statistics */
@@ -219,4 +219,68 @@ found:
#endif /* CONFIG_NUMA */
+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+
+void kmem_cache_drop_ref(struct kmem_cache *cachep);
+
+static inline void
+kmem_cache_get_ref(struct kmem_cache *cachep)
+{
+ if (cachep->memcg_params.id == -1 &&
+ unlikely(!atomic_add_unless(&cachep->memcg_params.refcnt, 1, 0)))
+ BUG();
+}
+
+static inline void
+mem_cgroup_put_kmem_cache(struct kmem_cache *cachep)
+{
+ rcu_read_unlock();
+}
+
+static inline void
+mem_cgroup_kmem_cache_prepare_sleep(struct kmem_cache *cachep)
+{
+ /*
+ * Make sure the cache doesn't get freed while we have interrupts
+ * enabled.
+ */
+ kmem_cache_get_ref(cachep);
+ rcu_read_unlock();
+}
+
+static inline void
+mem_cgroup_kmem_cache_finish_sleep(struct kmem_cache *cachep)
+{
+ rcu_read_lock();
+ kmem_cache_drop_ref(cachep);
+}
+
+#else /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */
+
+static inline void
+kmem_cache_get_ref(struct kmem_cache *cachep)
+{
+}
+
+static inline void
+kmem_cache_drop_ref(struct kmem_cache *cachep)
+{
+}
+
+static inline void
+mem_cgroup_put_kmem_cache(struct kmem_cache *cachep)
+{
+}
+
+static inline void
+mem_cgroup_kmem_cache_prepare_sleep(struct kmem_cache *cachep)
+{
+}
+
+static inline void
+mem_cgroup_kmem_cache_finish_sleep(struct kmem_cache *cachep)
+{
+}
+#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */
+
#endif /* _LINUX_SLAB_DEF_H */
diff --git a/mm/slab.c b/mm/slab.c
index 13948c3..ac0916b 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -1818,20 +1818,28 @@ static void *kmem_getpages(struct kmem_cache *cachep, gfp_t flags, int nodeid)
if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
flags |= __GFP_RECLAIMABLE;
+ nr_pages = (1 << cachep->gfporder);
+ if (!mem_cgroup_charge_slab(cachep, flags, nr_pages * PAGE_SIZE))
+ return NULL;
+
page = alloc_pages_exact_node(nodeid, flags | __GFP_NOTRACK, cachep->gfporder);
if (!page) {
if (!(flags & __GFP_NOWARN) && printk_ratelimit())
slab_out_of_memory(cachep, flags, nodeid);
+
+ mem_cgroup_uncharge_slab(cachep, nr_pages * PAGE_SIZE);
return NULL;
}
- nr_pages = (1 << cachep->gfporder);
if (cachep->flags & SLAB_RECLAIM_ACCOUNT)
add_zone_page_state(page_zone(page),
NR_SLAB_RECLAIMABLE, nr_pages);
else
add_zone_page_state(page_zone(page),
NR_SLAB_UNRECLAIMABLE, nr_pages);
+
+ kmem_cache_get_ref(cachep);
+
for (i = 0; i < nr_pages; i++)
__SetPageSlab(page + i);
@@ -1864,6 +1872,8 @@ static void kmem_freepages(struct kmem_cache *cachep, void *addr)
else
sub_zone_page_state(page_zone(page),
NR_SLAB_UNRECLAIMABLE, nr_freed);
+ mem_cgroup_uncharge_slab(cachep, i * PAGE_SIZE);
+ kmem_cache_drop_ref(cachep);
while (i--) {
BUG_ON(!PageSlab(page));
__ClearPageSlab(page);
@@ -2823,12 +2833,28 @@ void kmem_cache_destroy(struct kmem_cache *cachep)
if (unlikely(cachep->flags & SLAB_DESTROY_BY_RCU))
rcu_barrier();
+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+ /* Not a memcg cache */
+ if (cachep->memcg_params.id != -1) {
+ mem_cgroup_release_cache(cachep);
+ mem_cgroup_flush_cache_create_queue();
+ }
+#endif
__kmem_cache_destroy(cachep);
mutex_unlock(&cache_chain_mutex);
put_online_cpus();
}
EXPORT_SYMBOL(kmem_cache_destroy);
+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+void kmem_cache_drop_ref(struct kmem_cache *cachep)
+{
+ if (cachep->memcg_params.id == -1 &&
+ unlikely(atomic_dec_and_test(&cachep->memcg_params.refcnt)))
+ mem_cgroup_destroy_cache(cachep);
+}
+#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */
+
/*
* Get the memory for a slab management obj.
* For a slab cache when the slab descriptor is off-slab, slab descriptors
@@ -3028,8 +3054,10 @@ static int cache_grow(struct kmem_cache *cachep,
offset *= cachep->colour_off;
- if (local_flags & __GFP_WAIT)
+ if (local_flags & __GFP_WAIT) {
local_irq_enable();
+ mem_cgroup_kmem_cache_prepare_sleep(cachep);
+ }
/*
* The test for missing atomic flag is performed here, rather than
@@ -3058,8 +3086,10 @@ static int cache_grow(struct kmem_cache *cachep,
cache_init_objs(cachep, slabp);
- if (local_flags & __GFP_WAIT)
+ if (local_flags & __GFP_WAIT) {
local_irq_disable();
+ mem_cgroup_kmem_cache_finish_sleep(cachep);
+ }
check_irq_off();
spin_lock(&l3->list_lock);
@@ -3072,8 +3102,10 @@ static int cache_grow(struct kmem_cache *cachep,
opps1:
kmem_freepages(cachep, objp);
failed:
- if (local_flags & __GFP_WAIT)
+ if (local_flags & __GFP_WAIT) {
local_irq_disable();
+ mem_cgroup_kmem_cache_finish_sleep(cachep);
+ }
return 0;
}
@@ -3834,11 +3866,15 @@ static inline void __cache_free(struct kmem_cache *cachep, void *objp,
*/
void *kmem_cache_alloc(struct kmem_cache *cachep, gfp_t flags)
{
- void *ret = __cache_alloc(cachep, flags, __builtin_return_address(0));
+ void *ret;
+
+ rcu_read_lock();
+ cachep = mem_cgroup_get_kmem_cache(cachep, flags);
+ rcu_read_unlock();
+ ret = __cache_alloc(cachep, flags, __builtin_return_address(0));
trace_kmem_cache_alloc(_RET_IP_, ret,
obj_size(cachep), cachep->buffer_size, flags);
-
return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc);
@@ -3849,6 +3885,10 @@ kmem_cache_alloc_trace(size_t size, struct kmem_cache *cachep, gfp_t flags)
{
void *ret;
+ rcu_read_lock();
+ cachep = mem_cgroup_get_kmem_cache(cachep, flags);
+ rcu_read_unlock();
+
ret = __cache_alloc(cachep, flags, __builtin_return_address(0));
trace_kmalloc(_RET_IP_, ret,
@@ -3861,13 +3901,17 @@ EXPORT_SYMBOL(kmem_cache_alloc_trace);
#ifdef CONFIG_NUMA
void *kmem_cache_alloc_node(struct kmem_cache *cachep, gfp_t flags, int nodeid)
{
- void *ret = __cache_alloc_node(cachep, flags, nodeid,
+ void *ret;
+
+ rcu_read_lock();
+ cachep = mem_cgroup_get_kmem_cache(cachep, flags);
+ rcu_read_unlock();
+ ret = __cache_alloc_node(cachep, flags, nodeid,
__builtin_return_address(0));
trace_kmem_cache_alloc_node(_RET_IP_, ret,
obj_size(cachep), cachep->buffer_size,
flags, nodeid);
-
return ret;
}
EXPORT_SYMBOL(kmem_cache_alloc_node);
@@ -3880,6 +3924,9 @@ void *kmem_cache_alloc_node_trace(size_t size,
{
void *ret;
+ rcu_read_lock();
+ cachep = mem_cgroup_get_kmem_cache(cachep, flags);
+ rcu_read_unlock();
ret = __cache_alloc_node(cachep, flags, nodeid,
__builtin_return_address(0));
trace_kmalloc_node(_RET_IP_, ret,
@@ -4011,9 +4058,33 @@ void kmem_cache_free(struct kmem_cache *cachep, void *objp)
local_irq_save(flags);
debug_check_no_locks_freed(objp, obj_size(cachep));
+
+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+ {
+ struct kmem_cache *actual_cachep;
+
+ actual_cachep = virt_to_cache(objp);
+ if (actual_cachep != cachep) {
+ VM_BUG_ON(actual_cachep->memcg_params.id != -1);
+ cachep = actual_cachep;
+ }
+ /*
+ * Grab a reference so that the cache is guaranteed to stay
+ * around.
+ * If we are freeing the last object of a dead memcg cache,
+ * the kmem_cache_drop_ref() at the end of this function
+ * will end up freeing the cache.
+ */
+ kmem_cache_get_ref(cachep);
+ }
+#endif
+
if (!(cachep->flags & SLAB_DEBUG_OBJECTS))
debug_check_no_obj_freed(objp, obj_size(cachep));
__cache_free(cachep, objp, __builtin_return_address(0));
+
+ kmem_cache_drop_ref(cachep);
+
local_irq_restore(flags);
trace_kmem_cache_free(_RET_IP_, objp);
@@ -4041,9 +4112,19 @@ void kfree(const void *objp)
local_irq_save(flags);
kfree_debugcheck(objp);
c = virt_to_cache(objp);
+
+ /*
+ * Grab a reference so that the cache is guaranteed to stay around.
+ * If we are freeing the last object of a dead memcg cache, the
+ * kmem_cache_drop_ref() at the end of this function will end up
+ * freeing the cache.
+ */
+ kmem_cache_get_ref(c);
+
debug_check_no_locks_freed(objp, obj_size(c));
debug_check_no_obj_freed(objp, obj_size(c));
__cache_free(c, (void *)objp, __builtin_return_address(0));
+ kmem_cache_drop_ref(c);
local_irq_restore(flags);
}
EXPORT_SYMBOL(kfree);
@@ -4312,6 +4393,13 @@ static void cache_reap(struct work_struct *w)
list_for_each_entry(searchp, &cache_chain, next) {
check_irq_on();
+#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
+ /* For memcg caches, make sure we only reap the active ones. */
+ if (searchp->memcg_params.id == -1 &&
+ !atomic_add_unless(&searchp->memcg_params.refcnt, 1, 0))
+ continue;
+#endif
+
/*
* We only take the l3 lock if absolutely necessary and we
* have established with reasonable certainty that
@@ -4344,6 +4432,7 @@ static void cache_reap(struct work_struct *w)
STATS_ADD_REAPED(searchp, freed);
}
next:
+ kmem_cache_drop_ref(searchp);
cond_resched();
}
check_irq_on();
--
1.7.7.6
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists