On Wed, Jan 29, 2020 at 1:28 PM David Rientjes rientjes@google.com wrote:
On Tue, 14 Jan 2020, Mina Almasry wrote:
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h index dea6143aa0685..5491932ea5758 100644 --- a/include/linux/hugetlb.h +++ b/include/linux/hugetlb.h @@ -46,6 +46,16 @@ struct resv_map { long adds_in_progress; struct list_head region_cache; long region_cache_count; +#ifdef CONFIG_CGROUP_HUGETLB
/** On private mappings, the counter to uncharge reservations is stored* here. If these fields are 0, then either the mapping is shared, or* cgroup accounting is disabled for this resv_map.*/struct page_counter *reservation_counter;unsigned long pages_per_hpage;struct cgroup_subsys_state *css;+#endif }; extern struct resv_map *resv_map_alloc(void); void resv_map_release(struct kref *ref); diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h index eab8a70d5bcb5..8c320accefe87 100644 --- a/include/linux/hugetlb_cgroup.h +++ b/include/linux/hugetlb_cgroup.h @@ -25,6 +25,33 @@ struct hugetlb_cgroup; #define HUGETLB_CGROUP_MIN_ORDER 2
#ifdef CONFIG_CGROUP_HUGETLB +enum hugetlb_memory_event {
HUGETLB_MAX,HUGETLB_NR_MEMORY_EVENTS,+};
+struct hugetlb_cgroup {
struct cgroup_subsys_state css;/** the counter to account for hugepages from hugetlb.*/struct page_counter hugepage[HUGE_MAX_HSTATE];/** the counter to account for hugepage reservations from hugetlb.*/struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];/* Handle for "hugetlb.events" */struct cgroup_file events_file[HUGE_MAX_HSTATE];/* Handle for "hugetlb.events.local" */struct cgroup_file events_local_file[HUGE_MAX_HSTATE];+};
static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page, bool reserved) diff --git a/mm/hugetlb.c b/mm/hugetlb.c index 62a4cf3db4090..f1b63946ee95c 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c @@ -666,6 +666,17 @@ struct resv_map *resv_map_alloc(void) INIT_LIST_HEAD(&resv_map->regions);
resv_map->adds_in_progress = 0;+#ifdef CONFIG_CGROUP_HUGETLB
/** Initialize these to 0. On shared mappings, 0's here indicate these* fields don't do cgroup accounting. On private mappings, these will be* re-initialized to the proper values, to indicate that hugetlb cgroup* reservations are to be un-charged from here.*/resv_map->reservation_counter = NULL;resv_map->pages_per_hpage = 0;resv_map->css = NULL;+#endif
Might be better to extract out a resv_map_init() that does the initialization when CONFIG_CGROUP_HUGETLB is enabled? Could be used here as well as hugetlb_reserve_pages().
INIT_LIST_HEAD(&resv_map->region_cache); list_add(&rg->link, &resv_map->region_cache);@@ -3194,7 +3205,11 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
reserve = (end - start) - region_count(resv, start, end);
kref_put(&resv->refs, resv_map_release);+#ifdef CONFIG_CGROUP_HUGETLB
hugetlb_cgroup_uncharge_counter(resv->reservation_counter,(end - start) * resv->pages_per_hpage,resv->css);+#endif
if (reserve) { /*Mike has given is Reviewed-by so likely not a big concern for the generic hugetlb code, but I was wondering if we can reduce the number of #ifdef's if we defined a CONFIG_CGROUP_HUGETLB helper to take the resv, end, and start? If CONFIG_CGROUP_HUGETLB is defined, it converts into the above, otherwise it's a no-op and we don't run into any compile errors because we are accessing fields that don't exist without the option.
Yes wherever possible I refactored the code a bit to remove #ifdefs in the middle of hugetlb logic.
Otherwise looks good!
Acked-by: David Rientjes rientjes@google.com