Message ID | 20190708124120.3400683-1-arnd@arndb.de |
---|---|
State | New |
Headers | show |
Series | vmscan: fix memcg_kmem build failure | expand |
On Mon, Jul 8, 2019 at 2:42 PM Arnd Bergmann <arnd@arndb.de> wrote: > > When CONFIG_MEMCG_KMEM is disabled, we get a build failure > for calling a nonexisting memcg_expand_shrinker_maps(): > > mm/vmscan.c:220:7: error: implicit declaration of function 'memcg_expand_shrinker_maps' [-Werror,-Wimplicit-function-declaration] > if (memcg_expand_shrinker_maps(id)) { > ^ I see now that a fix for this is already in today's linux-next, please ignore my patch. Arnd
diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 5901a90f58eb..6b15e2066fc7 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -1407,6 +1407,11 @@ static inline void memcg_put_cache_ids(void) { } +static inline int memcg_expand_shrinker_maps(int new_id) +{ + return -ENOMEM; +} + static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id) { } #endif /* CONFIG_MEMCG_KMEM */ diff --git a/mm/vmscan.c b/mm/vmscan.c index a0301edd8d03..323a9c50c0fe 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -591,7 +591,7 @@ static unsigned long do_shrink_slab(struct shrink_control *shrinkctl, return freed; } -#ifdef CONFIG_MEMCG +#ifdef CONFIG_MEMCG_KMEM static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, int priority) {
When CONFIG_MEMCG_KMEM is disabled, we get a build failure for calling a nonexisting memcg_expand_shrinker_maps(): mm/vmscan.c:220:7: error: implicit declaration of function 'memcg_expand_shrinker_maps' [-Werror,-Wimplicit-function-declaration] if (memcg_expand_shrinker_maps(id)) { ^ mm/vmscan.c:220:7: error: this function declaration is not a prototype [-Werror,-Wstrict-prototypes] mm/vmscan.c:608:56: error: no member named 'shrinker_map' in 'struct mem_cgroup_per_node' map = rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_map, ~~~~~~~~~~~~~~~~~~~~ ^ include/linux/rcupdate.h:498:31: note: expanded from macro 'rcu_dereference_protected' __rcu_dereference_protected((p), (c), __rcu) ^ include/linux/rcupdate.h:321:12: note: expanded from macro '__rcu_dereference_protected' ((typeof(*p) __force __kernel *)(p)); \ ^ mm/vmscan.c:608:6: error: assigning to 'struct memcg_shrinker_map *' from incompatible type 'void' map = rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_map, and another issue trying to access invalid struct fields: mm/vmscan.c:608:56: error: no member named 'shrinker_map' in 'struct mem_cgroup_per_node' map = rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_map, ~~~~~~~~~~~~~~~~~~~~ ^ include/linux/rcupdate.h:498:31: note: expanded from macro 'rcu_dereference_protected' __rcu_dereference_protected((p), (c), __rcu) ^ include/linux/rcupdate.h:321:12: note: expanded from macro '__rcu_dereference_protected' ((typeof(*p) __force __kernel *)(p)); \ ^ mm/vmscan.c:608:6: error: assigning to 'struct memcg_shrinker_map *' from incompatible type 'void' map = rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_map, Add a dummy definition for memcg_expand_shrinker_maps() that always fails, and hide the obviously nonworking shrink_slab_memcg() function. Fixes: 8236f517d69e ("mm: shrinker: make shrinker not depend on memcg kmem") Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- No idea what the intended behavior is supposed to be for this case. Rather than failing, should we actually provide that function? Or maybe a more elaborate change is needed? --- include/linux/memcontrol.h | 5 +++++ mm/vmscan.c | 2 +- 2 files changed, 6 insertions(+), 1 deletion(-) -- 2.20.0