mm/vmpressure: fix data-race with memcg->socket_pressure

When reading memcg->socket_pressure in mem_cgroup_under_socket_pressure()
and writing memcg->socket_pressure in vmpressure() at the same time, the
following data-race occurs:

  BUG: KCSAN: data-race in __sk_mem_reduce_allocated / vmpressure

  write to 0xffff8881286f4938 of 8 bytes by task 24550 on cpu 3:
   vmpressure+0x218/0x230 mm/vmpressure.c:307
   shrink_node_memcgs+0x2b9/0x410 mm/vmscan.c:2658
   shrink_node+0x9d2/0x11d0 mm/vmscan.c:2769
   shrink_zones+0x29f/0x470 mm/vmscan.c:2972
   do_try_to_free_pages+0x193/0x6e0 mm/vmscan.c:3027
   try_to_free_mem_cgroup_pages+0x1c0/0x3f0 mm/vmscan.c:3345
   reclaim_high mm/memcontrol.c:2440 [inline]
   mem_cgroup_handle_over_high+0x18b/0x4d0 mm/memcontrol.c:2624
   tracehook_notify_resume include/linux/tracehook.h:197 [inline]
   exit_to_user_mode_loop kernel/entry/common.c:164 [inline]
   exit_to_user_mode_prepare+0x110/0x170 kernel/entry/common.c:191
   syscall_exit_to_user_mode+0x16/0x30 kernel/entry/common.c:266
   ret_from_fork+0x15/0x30 arch/x86/entry/entry_64.S:289

  read to 0xffff8881286f4938 of 8 bytes by interrupt on cpu 1:
   mem_cgroup_under_socket_pressure include/linux/memcontrol.h:1483 [inline]
   sk_under_memory_pressure include/net/sock.h:1314 [inline]
   __sk_mem_reduce_allocated+0x1d2/0x270 net/core/sock.c:2696
   __sk_mem_reclaim+0x44/0x50 net/core/sock.c:2711
   sk_mem_reclaim include/net/sock.h:1490 [inline]
   ......
   net_rx_action+0x17a/0x480 net/core/dev.c:6864
   __do_softirq+0x12c/0x2af kernel/softirq.c:298
   run_ksoftirqd+0x13/0x20 kernel/softirq.c:653
   smpboot_thread_fn+0x33f/0x510 kernel/smpboot.c:165
   kthread+0x1fc/0x220 kernel/kthread.c:292
   ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:296

Fix it by using READ_ONCE() and WRITE_ONCE() to read and write
memcg->socket_pressure.

Link: https://lkml.kernel.org/r/20211025082843.671690-1-songyuanzheng@huawei.com
Signed-off-by: Yuanzheng Song <songyuanzheng@huawei.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Alex Shi <alexs@kernel.org>
Cc: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index f207f98bdb76b67b30a32fce1f35916d417a8d8d..9d96238d9c210722fe727ec0447390c20c9833f4 100644 (file)
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -1606,7 +1606,7 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
        if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_pressure)
                return true;
        do {
-               if (time_before(jiffies, memcg->socket_pressure))
+               if (time_before(jiffies, READ_ONCE(memcg->socket_pressure)))
                        return true;
        } while ((memcg = parent_mem_cgroup(memcg)));
        return false;

diff --git a/mm/vmpressure.c b/mm/vmpressure.c
index 76518e4166dc931acf0b4ae59f80d97609738b74..b52644771cc438f80f032390eaec7c87423c9ae7 100644 (file)
--- a/mm/vmpressure.c
+++ b/mm/vmpressure.c
@@ -308,7 +308,7 @@ void vmpressure(gfp_t gfp, struct mem_cgroup *memcg, bool tree,
                         * asserted for a second in which subsequent
                         * pressure events can occur.
                         */
-                       memcg->socket_pressure = jiffies + HZ;
+                       WRITE_ONCE(memcg->socket_pressure, jiffies + HZ);
                }
        }
 }