memcg: separate local_trylock for memcg and obj

The per-cpu stock_lock protects cached memcg and cached objcg and their
respective fields.  However there is no dependency between these fields
and it is better to have fine grained separate locks for cached memcg and
cached objcg.  This decoupling of locks allows us to make the memcg charge
cache and objcg charge cache to be nmi safe independently.

At the moment, memcg charge cache is already nmi safe and this decoupling
will allow to make memcg charge cache work without disabling irqs.

Link: https://lkml.kernel.org/r/20250506225533.2580386-3-shakeel.butt@linux.dev
Signed-off-by: Shakeel Butt <shakeel.butt@linux.dev>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Eric Dumaze <edumazet@google.com>
Cc: Jakub Kacinski <kuba@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index d5b1df2326f74f74bbd04fa49de1652f0c3cad1c..352aaaeaa536d8cb58099308fc178b9157f89df8 100644 (file)
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1779,13 +1779,14 @@ void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
  */
 #define NR_MEMCG_STOCK 7
 struct memcg_stock_pcp {
-       local_trylock_t stock_lock;
+       local_trylock_t memcg_lock;
        uint8_t nr_pages[NR_MEMCG_STOCK];
        struct mem_cgroup *cached[NR_MEMCG_STOCK];
 
+       local_trylock_t obj_lock;
+       unsigned int nr_bytes;
        struct obj_cgroup *cached_objcg;
        struct pglist_data *cached_pgdat;
-       unsigned int nr_bytes;
        int nr_slab_reclaimable_b;
        int nr_slab_unreclaimable_b;
 
@@ -1794,7 +1795,8 @@ struct memcg_stock_pcp {
 #define FLUSHING_CACHED_CHARGE 0
 };
 static DEFINE_PER_CPU_ALIGNED(struct memcg_stock_pcp, memcg_stock) = {
-       .stock_lock = INIT_LOCAL_TRYLOCK(stock_lock),
+       .memcg_lock = INIT_LOCAL_TRYLOCK(memcg_lock),
+       .obj_lock = INIT_LOCAL_TRYLOCK(obj_lock),
 };
 static DEFINE_MUTEX(percpu_charge_mutex);
 
@@ -1822,7 +1824,7 @@ static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
        int i;
 
        if (nr_pages > MEMCG_CHARGE_BATCH ||
-           !local_trylock_irqsave(&memcg_stock.stock_lock, flags))
+           !local_trylock_irqsave(&memcg_stock.memcg_lock, flags))
                return ret;
 
        stock = this_cpu_ptr(&memcg_stock);
@@ -1839,7 +1841,7 @@ static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
                break;
        }
 
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.memcg_lock, flags);
 
        return ret;
 }
@@ -1885,19 +1887,19 @@ static void drain_local_stock(struct work_struct *dummy)
        struct memcg_stock_pcp *stock;
        unsigned long flags;
 
-       /*
-        * The only protection from cpu hotplug (memcg_hotplug_cpu_dead) vs.
-        * drain_stock races is that we always operate on local CPU stock
-        * here with IRQ disabled
-        */
-       local_lock_irqsave(&memcg_stock.stock_lock, flags);
+       if (WARN_ONCE(!in_task(), "drain in non-task context"))
+               return;
 
+       local_lock_irqsave(&memcg_stock.obj_lock, flags);
        stock = this_cpu_ptr(&memcg_stock);
        drain_obj_stock(stock);
+       local_unlock_irqrestore(&memcg_stock.obj_lock, flags);
+
+       local_lock_irqsave(&memcg_stock.memcg_lock, flags);
+       stock = this_cpu_ptr(&memcg_stock);
        drain_stock_fully(stock);
        clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
-
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.memcg_lock, flags);
 }
 
 static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
@@ -1920,10 +1922,10 @@ static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
        VM_WARN_ON_ONCE(mem_cgroup_is_root(memcg));
 
        if (nr_pages > MEMCG_CHARGE_BATCH ||
-           !local_trylock_irqsave(&memcg_stock.stock_lock, flags)) {
+           !local_trylock_irqsave(&memcg_stock.memcg_lock, flags)) {
                /*
                 * In case of larger than batch refill or unlikely failure to
-                * lock the percpu stock_lock, uncharge memcg directly.
+                * lock the percpu memcg_lock, uncharge memcg directly.
                 */
                memcg_uncharge(memcg, nr_pages);
                return;
@@ -1955,7 +1957,7 @@ static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
                WRITE_ONCE(stock->nr_pages[i], nr_pages);
        }
 
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.memcg_lock, flags);
 }
 
 static bool is_drain_needed(struct memcg_stock_pcp *stock,
@@ -2030,11 +2032,12 @@ static int memcg_hotplug_cpu_dead(unsigned int cpu)
 
        stock = &per_cpu(memcg_stock, cpu);
 
-       /* drain_obj_stock requires stock_lock */
-       local_lock_irqsave(&memcg_stock.stock_lock, flags);
+       /* drain_obj_stock requires obj_lock */
+       local_lock_irqsave(&memcg_stock.obj_lock, flags);
        drain_obj_stock(stock);
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.obj_lock, flags);
 
+       /* no need for the local lock */
        drain_stock_fully(stock);
 
        return 0;
@@ -2887,7 +2890,7 @@ static bool consume_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes,
        unsigned long flags;
        bool ret = false;
 
-       local_lock_irqsave(&memcg_stock.stock_lock, flags);
+       local_lock_irqsave(&memcg_stock.obj_lock, flags);
 
        stock = this_cpu_ptr(&memcg_stock);
        if (objcg == READ_ONCE(stock->cached_objcg) && stock->nr_bytes >= nr_bytes) {
@@ -2898,7 +2901,7 @@ static bool consume_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes,
                        __account_obj_stock(objcg, stock, nr_bytes, pgdat, idx);
        }
 
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.obj_lock, flags);
 
        return ret;
 }
@@ -2987,7 +2990,7 @@ static void refill_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes,
        unsigned long flags;
        unsigned int nr_pages = 0;
 
-       local_lock_irqsave(&memcg_stock.stock_lock, flags);
+       local_lock_irqsave(&memcg_stock.obj_lock, flags);
 
        stock = this_cpu_ptr(&memcg_stock);
        if (READ_ONCE(stock->cached_objcg) != objcg) { /* reset if necessary */
@@ -3009,7 +3012,7 @@ static void refill_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes,
                stock->nr_bytes &= (PAGE_SIZE - 1);
        }
 
-       local_unlock_irqrestore(&memcg_stock.stock_lock, flags);
+       local_unlock_irqrestore(&memcg_stock.obj_lock, flags);
 
        if (nr_pages)
                obj_cgroup_uncharge_pages(objcg, nr_pages);