On 2023/12/11 17:12, Yuan Can wrote:
From: Jian Zhang zhangjian210@huawei.com
hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8NC0E
In Ascend, we use tmp hugepage and disable OOM-killer, when we cause a OOM, and after some time, the memory is enough for process, the process will not return to run normal. In this case, we must use oom recover to let the process run.
Signed-off-by: Jian Zhang zhangjian210@huawei.com
mm/memcontrol.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
1. 这个代码没有做到隔离；
memcg_oom_recover都会被调用到
2. 需要单独写一个函数 #ifdef CONFIG_ASCEND_OOM static fucn{} {
这里需要一种隔离，让当前代码不执行
} #else static func{} #endif
diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 08af3c8df6f3..e1e79339dbd2 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -3154,12 +3154,22 @@ void __memcg_kmem_uncharge_page(struct page *page, int order) struct folio *folio = page_folio(page); struct obj_cgroup *objcg; unsigned int nr_pages = 1 << order; +#ifdef CONFIG_ASCEND_OOM
- struct mem_cgroup *memcg;
+#endif
if (!folio_memcg_kmem(folio)) return;
objcg = __folio_objcg(folio); obj_cgroup_uncharge_pages(objcg, nr_pages); +#ifdef CONFIG_ASCEND_OOM
- memcg = get_mem_cgroup_from_objcg(objcg);
- if (!mem_cgroup_is_root(memcg))
memcg_oom_recover(memcg);- css_put(&memcg->css);
+#endif
- folio->memcg_data = 0; obj_cgroup_put(objcg); }