On Mon, Apr 29, 2019 at 12:59:39PM +0200, Jiri Slaby wrote:
> We have a single node system with node 0 disabled:
> Scanning NUMA topology in Northbridge 24
> Number of physical nodes 2
> Skipping disabled node 0
> Node 1 MemBase 0000000000000000 Limit 00000000fbff0000
> NODE_DATA(1) allocated [mem 0xfbfda000-0xfbfeffff]
>
> This causes crashes in memcg when system boots:
> BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
> #PF error: [normal kernel read fault]
> ...
> RIP: 0010:list_lru_add+0x94/0x170
> ...
> Call Trace:
> d_lru_add+0x44/0x50
> dput.part.34+0xfc/0x110
> __fput+0x108/0x230
> task_work_run+0x9f/0xc0
> exit_to_usermode_loop+0xf5/0x100
>
> It is reproducible as far as 4.12. I did not try older kernels. You have
> to have a new enough systemd, e.g. 241 (the reason is unknown -- was not
> investigated). Cannot be reproduced with systemd 234.
>
> The system crashes because the size of lru array is never updated in
> memcg_update_all_list_lrus and the reads are past the zero-sized array,
> causing dereferences of random memory.
>
> The root cause are list_lru_memcg_aware checks in the list_lru code.
> The test in list_lru_memcg_aware is broken: it assumes node 0 is always
> present, but it is not true on some systems as can be seen above.
>
> So fix this by checking the first online node instead of node 0.
>
> Signed-off-by: Jiri Slaby <jslaby@xxxxxxx>
> Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> Cc: Michal Hocko <mhocko@xxxxxxxxxx>
> Cc: Vladimir Davydov <vdavydov.dev@xxxxxxxxx>
> Cc: <cgroups@xxxxxxxxxxxxxxx>
> Cc: <linux-mm@xxxxxxxxx>
> Cc: Raghavendra K T <raghavendra.kt@xxxxxxxxxxxxxxxxxx>
> ---
> mm/list_lru.c | 6 +-----
> 1 file changed, 1 insertion(+), 5 deletions(-)
>
> diff --git a/mm/list_lru.c b/mm/list_lru.c
> index 0730bf8ff39f..7689910f1a91 100644
> --- a/mm/list_lru.c
> +++ b/mm/list_lru.c
> @@ -37,11 +37,7 @@ static int lru_shrinker_id(struct list_lru *lru)
>
> static inline bool list_lru_memcg_aware(struct list_lru *lru)
> {
> - /*
> - * This needs node 0 to be always present, even
> - * in the systems supporting sparse numa ids.
> - */
> - return !!lru->node[0].memcg_lrus;
> + return !!lru->node[first_online_node].memcg_lrus;
> }
>
> static inline struct list_lru_one *
Yep, I didn't expect node 0 could ever be unavailable, my bad.
The patch looks fine to me:
Acked-by: Vladimir Davydov <vdavydov.dev@xxxxxxxxx>
However, I tend to agree with Michal that (ab)using node[0].memcg_lrus
to check if a list_lru is memcg aware looks confusing. I guess we could
simply add a bool flag to list_lru instead. Something like this, may be:
diff --git a/include/linux/list_lru.h b/include/linux/list_lru.h
index aa5efd9351eb..d5ceb2839a2d 100644
--- a/include/linux/list_lru.h
+++ b/include/linux/list_lru.h
@@ -54,6 +54,7 @@ struct list_lru {
#ifdef CONFIG_MEMCG_KMEM
struct list_head list;
int shrinker_id;
+ bool memcg_aware;
#endif
};
diff --git a/mm/list_lru.c b/mm/list_lru.c
index 0730bf8ff39f..8e605e40a4c6 100644
--- a/mm/list_lru.c
+++ b/mm/list_lru.c
@@ -37,11 +37,7 @@ static int lru_shrinker_id(struct list_lru *lru)
static inline bool list_lru_memcg_aware(struct list_lru *lru)
{
- /*
- * This needs node 0 to be always present, even
- * in the systems supporting sparse numa ids.
- */
- return !!lru->node[0].memcg_lrus;
+ return lru->memcg_aware;
}
static inline struct list_lru_one *
@@ -451,6 +447,7 @@ static int memcg_init_list_lru(struct list_lru *lru, bool memcg_aware)
{
int i;
+ lru->memcg_aware = memcg_aware;
if (!memcg_aware)
return 0;
