From c4843a7593a9df3ff5b1806084cefdfa81dd7c79 Mon Sep 17 00:00:00 2001 From: Greg Thelen Date: Fri, 22 May 2015 17:13:16 -0400 Subject: memcg: add per cgroup dirty page accounting When modifying PG_Dirty on cached file pages, update the new MEM_CGROUP_STAT_DIRTY counter. This is done in the same places where global NR_FILE_DIRTY is managed. The new memcg stat is visible in the per memcg memory.stat cgroupfs file. The most recent past attempt at this was http://thread.gmane.org/gmane.linux.kernel.cgroups/8632 The new accounting supports future efforts to add per cgroup dirty page throttling and writeback. It also helps an administrator break down a container's memory usage and provides evidence to understand memcg oom kills (the new dirty count is included in memcg oom kill messages). The ability to move page accounting between memcg (memory.move_charge_at_immigrate) makes this accounting more complicated than the global counter. The existing mem_cgroup_{begin,end}_page_stat() lock is used to serialize move accounting with stat updates. Typical update operation: memcg = mem_cgroup_begin_page_stat(page) if (TestSetPageDirty()) { [...] mem_cgroup_update_page_stat(memcg) } mem_cgroup_end_page_stat(memcg) Summary of mem_cgroup_end_page_stat() overhead: - Without CONFIG_MEMCG it's a no-op - With CONFIG_MEMCG and no inter memcg task movement, it's just rcu_read_lock() - With CONFIG_MEMCG and inter memcg task movement, it's rcu_read_lock() + spin_lock_irqsave() A memcg parameter is added to several routines because their callers now grab mem_cgroup_begin_page_stat() which returns the memcg later needed by for mem_cgroup_update_page_stat(). Because mem_cgroup_begin_page_stat() may disable interrupts, some adjustments are needed: - move __mark_inode_dirty() from __set_page_dirty() to its caller. __mark_inode_dirty() locking does not want interrupts disabled. - use spin_lock_irqsave(tree_lock) rather than spin_lock_irq() in __delete_from_page_cache(), replace_page_cache_page(), invalidate_complete_page2(), and __remove_mapping(). text data bss dec hex filename 8925147 1774832 1785856 12485835 be84cb vmlinux-!CONFIG_MEMCG-before 8925339 1774832 1785856 12486027 be858b vmlinux-!CONFIG_MEMCG-after +192 text bytes 8965977 1784992 1785856 12536825 bf4bf9 vmlinux-CONFIG_MEMCG-before 8966750 1784992 1785856 12537598 bf4efe vmlinux-CONFIG_MEMCG-after +773 text bytes Performance tests run on v4.0-rc1-36-g4f671fe2f952. Lower is better for all metrics, they're all wall clock or cycle counts. The read and write fault benchmarks just measure fault time, they do not include I/O time. * CONFIG_MEMCG not set: baseline patched kbuild 1m25.030000(+-0.088% 3 samples) 1m25.426667(+-0.120% 3 samples) dd write 100 MiB 0.859211561 +-15.10% 0.874162885 +-15.03% dd write 200 MiB 1.670653105 +-17.87% 1.669384764 +-11.99% dd write 1000 MiB 8.434691190 +-14.15% 8.474733215 +-14.77% read fault cycles 254.0(+-0.000% 10 samples) 253.0(+-0.000% 10 samples) write fault cycles 2021.2(+-3.070% 10 samples) 1984.5(+-1.036% 10 samples) * CONFIG_MEMCG=y root_memcg: baseline patched kbuild 1m25.716667(+-0.105% 3 samples) 1m25.686667(+-0.153% 3 samples) dd write 100 MiB 0.855650830 +-14.90% 0.887557919 +-14.90% dd write 200 MiB 1.688322953 +-12.72% 1.667682724 +-13.33% dd write 1000 MiB 8.418601605 +-14.30% 8.673532299 +-15.00% read fault cycles 266.0(+-0.000% 10 samples) 266.0(+-0.000% 10 samples) write fault cycles 2051.7(+-1.349% 10 samples) 2049.6(+-1.686% 10 samples) * CONFIG_MEMCG=y non-root_memcg: baseline patched kbuild 1m26.120000(+-0.273% 3 samples) 1m25.763333(+-0.127% 3 samples) dd write 100 MiB 0.861723964 +-15.25% 0.818129350 +-14.82% dd write 200 MiB 1.669887569 +-13.30% 1.698645885 +-13.27% dd write 1000 MiB 8.383191730 +-14.65% 8.351742280 +-14.52% read fault cycles 265.7(+-0.172% 10 samples) 267.0(+-0.000% 10 samples) write fault cycles 2070.6(+-1.512% 10 samples) 2084.4(+-2.148% 10 samples) As expected anon page faults are not affected by this patch. tj: Updated to apply on top of the recent cancel_dirty_page() changes. Signed-off-by: Sha Zhengju Signed-off-by: Greg Thelen Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- mm/memcontrol.c | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) (limited to 'mm/memcontrol.c') diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 14c2f2017e37..c23c1a3e8e16 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -90,6 +90,7 @@ static const char * const mem_cgroup_stat_names[] = { "rss", "rss_huge", "mapped_file", + "dirty", "writeback", "swap", }; @@ -2011,6 +2012,7 @@ again: return memcg; } +EXPORT_SYMBOL(mem_cgroup_begin_page_stat); /** * mem_cgroup_end_page_stat - finish a page state statistics transaction @@ -2029,6 +2031,7 @@ void mem_cgroup_end_page_stat(struct mem_cgroup *memcg) rcu_read_unlock(); } +EXPORT_SYMBOL(mem_cgroup_end_page_stat); /** * mem_cgroup_update_page_stat - update page state statistics @@ -4746,6 +4749,7 @@ static int mem_cgroup_move_account(struct page *page, { unsigned long flags; int ret; + bool anon; VM_BUG_ON(from == to); VM_BUG_ON_PAGE(PageLRU(page), page); @@ -4771,15 +4775,33 @@ static int mem_cgroup_move_account(struct page *page, if (page->mem_cgroup != from) goto out_unlock; + anon = PageAnon(page); + spin_lock_irqsave(&from->move_lock, flags); - if (!PageAnon(page) && page_mapped(page)) { + if (!anon && page_mapped(page)) { __this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_FILE_MAPPED], nr_pages); __this_cpu_add(to->stat->count[MEM_CGROUP_STAT_FILE_MAPPED], nr_pages); } + /* + * move_lock grabbed above and caller set from->moving_account, so + * mem_cgroup_update_page_stat() will serialize updates to PageDirty. + * So mapping should be stable for dirty pages. + */ + if (!anon && PageDirty(page)) { + struct address_space *mapping = page_mapping(page); + + if (mapping_cap_account_dirty(mapping)) { + __this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_DIRTY], + nr_pages); + __this_cpu_add(to->stat->count[MEM_CGROUP_STAT_DIRTY], + nr_pages); + } + } + if (PageWriteback(page)) { __this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_WRITEBACK], nr_pages); -- cgit v1.2.3-70-g09d2 From 56161634e4824380a67243a4cf3fa52eb1e5d836 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 17:13:20 -0400 Subject: memcg: add mem_cgroup_root_css Add global mem_cgroup_root_css which points to the root memcg css. This will be used by cgroup writeback support. If memcg is disabled, it's defined as ERR_PTR(-EINVAL). Signed-off-by: Tejun Heo Cc: Johannes Weiner aCc: Michal Hocko Signed-off-by: Jens Axboe --- include/linux/memcontrol.h | 4 ++++ mm/memcontrol.c | 2 ++ 2 files changed, 6 insertions(+) (limited to 'mm/memcontrol.c') diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 5fe6411b5e54..294498f4f6fc 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -68,6 +68,8 @@ enum mem_cgroup_events_index { }; #ifdef CONFIG_MEMCG +extern struct cgroup_subsys_state *mem_cgroup_root_css; + void mem_cgroup_events(struct mem_cgroup *memcg, enum mem_cgroup_events_index idx, unsigned int nr); @@ -196,6 +198,8 @@ void mem_cgroup_split_huge_fixup(struct page *head); #else /* CONFIG_MEMCG */ struct mem_cgroup; +#define mem_cgroup_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) + static inline void mem_cgroup_events(struct mem_cgroup *memcg, enum mem_cgroup_events_index idx, unsigned int nr) diff --git a/mm/memcontrol.c b/mm/memcontrol.c index c23c1a3e8e16..b22a92b37480 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -77,6 +77,7 @@ EXPORT_SYMBOL(memory_cgrp_subsys); #define MEM_CGROUP_RECLAIM_RETRIES 5 static struct mem_cgroup *root_mem_cgroup __read_mostly; +struct cgroup_subsys_state *mem_cgroup_root_css __read_mostly; /* Whether the swap controller is active */ #ifdef CONFIG_MEMCG_SWAP @@ -4441,6 +4442,7 @@ mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css) /* root ? */ if (parent_css == NULL) { root_mem_cgroup = memcg; + mem_cgroup_root_css = &memcg->css; page_counter_init(&memcg->memory, NULL); memcg->high = PAGE_COUNTER_MAX; memcg->soft_limit = PAGE_COUNTER_MAX; -- cgit v1.2.3-70-g09d2 From ad7fa852d3d2816d68a138ebc5bc8967aeb7fd86 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 27 May 2015 20:00:02 -0400 Subject: memcg: implement mem_cgroup_css_from_page() Implement mem_cgroup_css_from_page() which returns the cgroup_subsys_state of the memcg associated with a given page on the default hierarchy. This will be used by cgroup writeback support. This function assumes that page->mem_cgroup association doesn't change until the page is released, which is true on the default hierarchy as long as replace_page_cache_page() is not used. As the only user of replace_page_cache_page() is FUSE which won't support cgroup writeback for the time being, this works for now, and replace_page_cache_page() will soon be updated so that the invariant actually holds. Note that the RCU protected page->mem_cgroup access is consistent with other usages across memcg but ultimately incorrect. These unlocked accesses are missing required barriers. page->mem_cgroup should be made an RCU pointer and updated and accessed using RCU operations. v4: Instead of triggering WARN, return the root css on the traditional hierarchies. This makes the function a lot easier to deal with especially as there's no light way to synchronize against hierarchy rebinding. v3: s/mem_cgroup_migrate()/mem_cgroup_css_from_page()/ v2: Trigger WARN if the function is used on the traditional hierarchies and add comment about the assumed invariant. Signed-off-by: Tejun Heo Cc: Johannes Weiner Cc: Michal Hocko Signed-off-by: Jens Axboe --- include/linux/memcontrol.h | 1 + mm/memcontrol.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+) (limited to 'mm/memcontrol.c') diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 294498f4f6fc..637ef626008e 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -115,6 +115,7 @@ static inline bool mm_match_cgroup(struct mm_struct *mm, } extern struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg); +extern struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page); struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *, struct mem_cgroup *, diff --git a/mm/memcontrol.c b/mm/memcontrol.c index b22a92b37480..5c270a03729c 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -598,6 +598,39 @@ struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg) return &memcg->css; } +/** + * mem_cgroup_css_from_page - css of the memcg associated with a page + * @page: page of interest + * + * If memcg is bound to the default hierarchy, css of the memcg associated + * with @page is returned. The returned css remains associated with @page + * until it is released. + * + * If memcg is bound to a traditional hierarchy, the css of root_mem_cgroup + * is returned. + * + * XXX: The above description of behavior on the default hierarchy isn't + * strictly true yet as replace_page_cache_page() can modify the + * association before @page is released even on the default hierarchy; + * however, the current and planned usages don't mix the the two functions + * and replace_page_cache_page() will soon be updated to make the invariant + * actually true. + */ +struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page) +{ + struct mem_cgroup *memcg; + + rcu_read_lock(); + + memcg = page->mem_cgroup; + + if (!memcg || !cgroup_on_dfl(memcg->css.cgroup)) + memcg = root_mem_cgroup; + + rcu_read_unlock(); + return &memcg->css; +} + static struct mem_cgroup_per_zone * mem_cgroup_page_zoneinfo(struct mem_cgroup *memcg, struct page *page) { -- cgit v1.2.3-70-g09d2 From 52ebea749aaed195245701a8f90a23d672c7a933 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 17:13:37 -0400 Subject: writeback: make backing_dev_info host cgroup-specific bdi_writebacks For the planned cgroup writeback support, on each bdi (backing_dev_info), each memcg will be served by a separate wb (bdi_writeback). This patch updates bdi so that a bdi can host multiple wbs (bdi_writebacks). On the default hierarchy, blkcg implicitly enables memcg. This allows using memcg's page ownership for attributing writeback IOs, and every memcg - blkcg combination can be served by its own wb by assigning a dedicated wb to each memcg. This means that there may be multiple wb's of a bdi mapped to the same blkcg. As congested state is per blkcg - bdi combination, those wb's should share the same congested state. This is achieved by tracking congested state via bdi_writeback_congested structs which are keyed by blkcg. bdi->wb remains unchanged and will keep serving the root cgroup. cgwb's (cgroup wb's) for non-root cgroups are created on-demand or looked up while dirtying an inode according to the memcg of the page being dirtied or current task. Each cgwb is indexed on bdi->cgwb_tree by its memcg id. Once an inode is associated with its wb, it can be retrieved using inode_to_wb(). Currently, none of the filesystems has FS_CGROUP_WRITEBACK and all pages will keep being associated with bdi->wb. v3: inode_attach_wb() in account_page_dirtied() moved inside mapping_cap_account_dirty() block where it's known to be !NULL. Also, an unnecessary NULL check before kfree() removed. Both detected by the kbuild bot. v2: Updated so that wb association is per inode and wb is per memcg rather than blkcg. Signed-off-by: Tejun Heo Cc: kbuild test robot Cc: Dan Carpenter Cc: Jens Axboe Cc: Jan Kara Signed-off-by: Jens Axboe --- block/blk-cgroup.c | 7 +- fs/fs-writeback.c | 8 +- fs/inode.c | 1 + include/linux/backing-dev-defs.h | 59 +++++- include/linux/backing-dev.h | 195 +++++++++++++++++++ include/linux/blk-cgroup.h | 4 + include/linux/fs.h | 4 + include/linux/memcontrol.h | 4 + mm/backing-dev.c | 397 +++++++++++++++++++++++++++++++++++++++ mm/memcontrol.c | 19 +- mm/page-writeback.c | 11 +- 11 files changed, 698 insertions(+), 11 deletions(-) (limited to 'mm/memcontrol.c') diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c index 54ec1721cd3e..979cfdbb94e0 100644 --- a/block/blk-cgroup.c +++ b/block/blk-cgroup.c @@ -15,6 +15,7 @@ #include #include #include +#include #include #include #include @@ -797,6 +798,8 @@ static void blkcg_css_offline(struct cgroup_subsys_state *css) } spin_unlock_irq(&blkcg->lock); + + wb_blkcg_offline(blkcg); } static void blkcg_css_free(struct cgroup_subsys_state *css) @@ -827,7 +830,9 @@ done: spin_lock_init(&blkcg->lock); INIT_RADIX_TREE(&blkcg->blkg_tree, GFP_ATOMIC); INIT_HLIST_HEAD(&blkcg->blkg_list); - +#ifdef CONFIG_CGROUP_WRITEBACK + INIT_LIST_HEAD(&blkcg->cgwb_list); +#endif return &blkcg->css; } diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c index 34d1cb854fc1..99a2440c5588 100644 --- a/fs/fs-writeback.c +++ b/fs/fs-writeback.c @@ -185,11 +185,11 @@ void bdi_start_background_writeback(struct backing_dev_info *bdi) */ void inode_wb_list_del(struct inode *inode) { - struct backing_dev_info *bdi = inode_to_bdi(inode); + struct bdi_writeback *wb = inode_to_wb(inode); - spin_lock(&bdi->wb.list_lock); + spin_lock(&wb->list_lock); list_del_init(&inode->i_wb_list); - spin_unlock(&bdi->wb.list_lock); + spin_unlock(&wb->list_lock); } /* @@ -1268,6 +1268,8 @@ void __mark_inode_dirty(struct inode *inode, int flags) if ((inode->i_state & flags) != flags) { const int was_dirty = inode->i_state & I_DIRTY; + inode_attach_wb(inode, NULL); + if (flags & I_DIRTY_INODE) inode->i_state &= ~I_DIRTY_TIME; inode->i_state |= flags; diff --git a/fs/inode.c b/fs/inode.c index ea37cd17b53f..efc9edacfb9b 100644 --- a/fs/inode.c +++ b/fs/inode.c @@ -223,6 +223,7 @@ EXPORT_SYMBOL(free_inode_nonrcu); void __destroy_inode(struct inode *inode) { BUG_ON(inode_has_buffers(inode)); + inode_detach_wb(inode); security_inode_free(inode); fsnotify_inode_delete(inode); locks_free_lock_context(inode->i_flctx); diff --git a/include/linux/backing-dev-defs.h b/include/linux/backing-dev-defs.h index 9e9eafa5f5aa..a1e9c407a59a 100644 --- a/include/linux/backing-dev-defs.h +++ b/include/linux/backing-dev-defs.h @@ -2,8 +2,11 @@ #define __LINUX_BACKING_DEV_DEFS_H #include +#include +#include #include #include +#include #include #include #include @@ -37,10 +40,43 @@ enum wb_stat_item { #define WB_STAT_BATCH (8*(1+ilog2(nr_cpu_ids))) +/* + * For cgroup writeback, multiple wb's may map to the same blkcg. Those + * wb's can operate mostly independently but should share the congested + * state. To facilitate such sharing, the congested state is tracked using + * the following struct which is created on demand, indexed by blkcg ID on + * its bdi, and refcounted. + */ struct bdi_writeback_congested { unsigned long state; /* WB_[a]sync_congested flags */ + +#ifdef CONFIG_CGROUP_WRITEBACK + struct backing_dev_info *bdi; /* the associated bdi */ + atomic_t refcnt; /* nr of attached wb's and blkg */ + int blkcg_id; /* ID of the associated blkcg */ + struct rb_node rb_node; /* on bdi->cgwb_congestion_tree */ +#endif }; +/* + * Each wb (bdi_writeback) can perform writeback operations, is measured + * and throttled, independently. Without cgroup writeback, each bdi + * (bdi_writeback) is served by its embedded bdi->wb. + * + * On the default hierarchy, blkcg implicitly enables memcg. This allows + * using memcg's page ownership for attributing writeback IOs, and every + * memcg - blkcg combination can be served by its own wb by assigning a + * dedicated wb to each memcg, which enables isolation across different + * cgroups and propagation of IO back pressure down from the IO layer upto + * the tasks which are generating the dirty pages to be written back. + * + * A cgroup wb is indexed on its bdi by the ID of the associated memcg, + * refcounted with the number of inodes attached to it, and pins the memcg + * and the corresponding blkcg. As the corresponding blkcg for a memcg may + * change as blkcg is disabled and enabled higher up in the hierarchy, a wb + * is tested for blkcg after lookup and removed from index on mismatch so + * that a new wb for the combination can be created. + */ struct bdi_writeback { struct backing_dev_info *bdi; /* our parent bdi */ @@ -78,6 +114,19 @@ struct bdi_writeback { spinlock_t work_lock; /* protects work_list & dwork scheduling */ struct list_head work_list; struct delayed_work dwork; /* work item used for writeback */ + +#ifdef CONFIG_CGROUP_WRITEBACK + struct percpu_ref refcnt; /* used only for !root wb's */ + struct cgroup_subsys_state *memcg_css; /* the associated memcg */ + struct cgroup_subsys_state *blkcg_css; /* and blkcg */ + struct list_head memcg_node; /* anchored at memcg->cgwb_list */ + struct list_head blkcg_node; /* anchored at blkcg->cgwb_list */ + + union { + struct work_struct release_work; + struct rcu_head rcu; + }; +#endif }; struct backing_dev_info { @@ -92,9 +141,13 @@ struct backing_dev_info { unsigned int min_ratio; unsigned int max_ratio, max_prop_frac; - struct bdi_writeback wb; /* default writeback info for this bdi */ - struct bdi_writeback_congested wb_congested; - + struct bdi_writeback wb; /* the root writeback info for this bdi */ + struct bdi_writeback_congested wb_congested; /* its congested state */ +#ifdef CONFIG_CGROUP_WRITEBACK + struct radix_tree_root cgwb_tree; /* radix tree of active cgroup wbs */ + struct rb_root cgwb_congested_tree; /* their congested states */ + atomic_t usage_cnt; /* counts both cgwbs and cgwb_contested's */ +#endif struct device *dev; struct timer_list laptop_mode_wb_timer; diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h index 6bb31234e6a9..8ae59df2e3d1 100644 --- a/include/linux/backing-dev.h +++ b/include/linux/backing-dev.h @@ -13,6 +13,7 @@ #include #include #include +#include #include int __must_check bdi_init(struct backing_dev_info *bdi); @@ -234,6 +235,16 @@ static inline int bdi_sched_wait(void *word) #ifdef CONFIG_CGROUP_WRITEBACK +struct bdi_writeback_congested * +wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp); +void wb_congested_put(struct bdi_writeback_congested *congested); +struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi, + struct cgroup_subsys_state *memcg_css, + gfp_t gfp); +void __inode_attach_wb(struct inode *inode, struct page *page); +void wb_memcg_offline(struct mem_cgroup *memcg); +void wb_blkcg_offline(struct blkcg *blkcg); + /** * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode * @inode: inode of interest @@ -250,6 +261,135 @@ static inline bool inode_cgwb_enabled(struct inode *inode) (inode->i_sb->s_type->fs_flags & FS_CGROUP_WRITEBACK); } +/** + * wb_tryget - try to increment a wb's refcount + * @wb: bdi_writeback to get + */ +static inline bool wb_tryget(struct bdi_writeback *wb) +{ + if (wb != &wb->bdi->wb) + return percpu_ref_tryget(&wb->refcnt); + return true; +} + +/** + * wb_get - increment a wb's refcount + * @wb: bdi_writeback to get + */ +static inline void wb_get(struct bdi_writeback *wb) +{ + if (wb != &wb->bdi->wb) + percpu_ref_get(&wb->refcnt); +} + +/** + * wb_put - decrement a wb's refcount + * @wb: bdi_writeback to put + */ +static inline void wb_put(struct bdi_writeback *wb) +{ + if (wb != &wb->bdi->wb) + percpu_ref_put(&wb->refcnt); +} + +/** + * wb_find_current - find wb for %current on a bdi + * @bdi: bdi of interest + * + * Find the wb of @bdi which matches both the memcg and blkcg of %current. + * Must be called under rcu_read_lock() which protects the returend wb. + * NULL if not found. + */ +static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi) +{ + struct cgroup_subsys_state *memcg_css; + struct bdi_writeback *wb; + + memcg_css = task_css(current, memory_cgrp_id); + if (!memcg_css->parent) + return &bdi->wb; + + wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id); + + /* + * %current's blkcg equals the effective blkcg of its memcg. No + * need to use the relatively expensive cgroup_get_e_css(). + */ + if (likely(wb && wb->blkcg_css == task_css(current, blkio_cgrp_id))) + return wb; + return NULL; +} + +/** + * wb_get_create_current - get or create wb for %current on a bdi + * @bdi: bdi of interest + * @gfp: allocation mask + * + * Equivalent to wb_get_create() on %current's memcg. This function is + * called from a relatively hot path and optimizes the common cases using + * wb_find_current(). + */ +static inline struct bdi_writeback * +wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp) +{ + struct bdi_writeback *wb; + + rcu_read_lock(); + wb = wb_find_current(bdi); + if (wb && unlikely(!wb_tryget(wb))) + wb = NULL; + rcu_read_unlock(); + + if (unlikely(!wb)) { + struct cgroup_subsys_state *memcg_css; + + memcg_css = task_get_css(current, memory_cgrp_id); + wb = wb_get_create(bdi, memcg_css, gfp); + css_put(memcg_css); + } + return wb; +} + +/** + * inode_attach_wb - associate an inode with its wb + * @inode: inode of interest + * @page: page being dirtied (may be NULL) + * + * If @inode doesn't have its wb, associate it with the wb matching the + * memcg of @page or, if @page is NULL, %current. May be called w/ or w/o + * @inode->i_lock. + */ +static inline void inode_attach_wb(struct inode *inode, struct page *page) +{ + if (!inode->i_wb) + __inode_attach_wb(inode, page); +} + +/** + * inode_detach_wb - disassociate an inode from its wb + * @inode: inode of interest + * + * @inode is being freed. Detach from its wb. + */ +static inline void inode_detach_wb(struct inode *inode) +{ + if (inode->i_wb) { + wb_put(inode->i_wb); + inode->i_wb = NULL; + } +} + +/** + * inode_to_wb - determine the wb of an inode + * @inode: inode of interest + * + * Returns the wb @inode is currently associated with. + */ +static inline struct bdi_writeback *inode_to_wb(struct inode *inode) +{ + return inode->i_wb; +} + #else /* CONFIG_CGROUP_WRITEBACK */ static inline bool inode_cgwb_enabled(struct inode *inode) @@ -257,6 +397,61 @@ static inline bool inode_cgwb_enabled(struct inode *inode) return false; } +static inline struct bdi_writeback_congested * +wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp) +{ + return bdi->wb.congested; +} + +static inline void wb_congested_put(struct bdi_writeback_congested *congested) +{ +} + +static inline bool wb_tryget(struct bdi_writeback *wb) +{ + return true; +} + +static inline void wb_get(struct bdi_writeback *wb) +{ +} + +static inline void wb_put(struct bdi_writeback *wb) +{ +} + +static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi) +{ + return &bdi->wb; +} + +static inline struct bdi_writeback * +wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp) +{ + return &bdi->wb; +} + +static inline void inode_attach_wb(struct inode *inode, struct page *page) +{ +} + +static inline void inode_detach_wb(struct inode *inode) +{ +} + +static inline struct bdi_writeback *inode_to_wb(struct inode *inode) +{ + return &inode_to_bdi(inode)->wb; +} + +static inline void wb_memcg_offline(struct mem_cgroup *memcg) +{ +} + +static inline void wb_blkcg_offline(struct blkcg *blkcg) +{ +} + #endif /* CONFIG_CGROUP_WRITEBACK */ #endif /* _LINUX_BACKING_DEV_H */ diff --git a/include/linux/blk-cgroup.h b/include/linux/blk-cgroup.h index 4dc643f2046e..3033eb173eb4 100644 --- a/include/linux/blk-cgroup.h +++ b/include/linux/blk-cgroup.h @@ -53,6 +53,10 @@ struct blkcg { /* TODO: per-policy storage in blkcg */ unsigned int cfq_weight; /* belongs to cfq */ unsigned int cfq_leaf_weight; + +#ifdef CONFIG_CGROUP_WRITEBACK + struct list_head cgwb_list; +#endif }; struct blkg_stat { diff --git a/include/linux/fs.h b/include/linux/fs.h index 74e0ae0626a8..67a42ec95065 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -35,6 +35,7 @@ #include struct backing_dev_info; +struct bdi_writeback; struct export_operations; struct hd_geometry; struct iovec; @@ -635,6 +636,9 @@ struct inode { struct hlist_node i_hash; struct list_head i_wb_list; /* backing dev IO list */ +#ifdef CONFIG_CGROUP_WRITEBACK + struct bdi_writeback *i_wb; /* the associated cgroup wb */ +#endif struct list_head i_lru; /* inode LRU list */ struct list_head i_sb_list; union { diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 637ef626008e..662a953ea8ad 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -388,6 +388,10 @@ enum { OVER_LIMIT, }; +#ifdef CONFIG_CGROUP_WRITEBACK +struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); +#endif + struct sock; #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM) void sock_update_memcg(struct sock *sk); diff --git a/mm/backing-dev.c b/mm/backing-dev.c index 5ec7658cb984..4c9386c98ec1 100644 --- a/mm/backing-dev.c +++ b/mm/backing-dev.c @@ -368,6 +368,401 @@ static void wb_exit(struct bdi_writeback *wb) fprop_local_destroy_percpu(&wb->completions); } +#ifdef CONFIG_CGROUP_WRITEBACK + +#include + +/* + * cgwb_lock protects bdi->cgwb_tree, bdi->cgwb_congested_tree, + * blkcg->cgwb_list, and memcg->cgwb_list. bdi->cgwb_tree is also RCU + * protected. cgwb_release_wait is used to wait for the completion of cgwb + * releases from bdi destruction path. + */ +static DEFINE_SPINLOCK(cgwb_lock); +static DECLARE_WAIT_QUEUE_HEAD(cgwb_release_wait); + +/** + * wb_congested_get_create - get or create a wb_congested + * @bdi: associated bdi + * @blkcg_id: ID of the associated blkcg + * @gfp: allocation mask + * + * Look up the wb_congested for @blkcg_id on @bdi. If missing, create one. + * The returned wb_congested has its reference count incremented. Returns + * NULL on failure. + */ +struct bdi_writeback_congested * +wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp) +{ + struct bdi_writeback_congested *new_congested = NULL, *congested; + struct rb_node **node, *parent; + unsigned long flags; + + if (blkcg_id == 1) + return &bdi->wb_congested; +retry: + spin_lock_irqsave(&cgwb_lock, flags); + + node = &bdi->cgwb_congested_tree.rb_node; + parent = NULL; + + while (*node != NULL) { + parent = *node; + congested = container_of(parent, struct bdi_writeback_congested, + rb_node); + if (congested->blkcg_id < blkcg_id) + node = &parent->rb_left; + else if (congested->blkcg_id > blkcg_id) + node = &parent->rb_right; + else + goto found; + } + + if (new_congested) { + /* !found and storage for new one already allocated, insert */ + congested = new_congested; + new_congested = NULL; + rb_link_node(&congested->rb_node, parent, node); + rb_insert_color(&congested->rb_node, &bdi->cgwb_congested_tree); + atomic_inc(&bdi->usage_cnt); + goto found; + } + + spin_unlock_irqrestore(&cgwb_lock, flags); + + /* allocate storage for new one and retry */ + new_congested = kzalloc(sizeof(*new_congested), gfp); + if (!new_congested) + return NULL; + + atomic_set(&new_congested->refcnt, 0); + new_congested->bdi = bdi; + new_congested->blkcg_id = blkcg_id; + goto retry; + +found: + atomic_inc(&congested->refcnt); + spin_unlock_irqrestore(&cgwb_lock, flags); + kfree(new_congested); + return congested; +} + +/** + * wb_congested_put - put a wb_congested + * @congested: wb_congested to put + * + * Put @congested and destroy it if the refcnt reaches zero. + */ +void wb_congested_put(struct bdi_writeback_congested *congested) +{ + struct backing_dev_info *bdi = congested->bdi; + unsigned long flags; + + if (congested->blkcg_id == 1) + return; + + local_irq_save(flags); + if (!atomic_dec_and_lock(&congested->refcnt, &cgwb_lock)) { + local_irq_restore(flags); + return; + } + + rb_erase(&congested->rb_node, &congested->bdi->cgwb_congested_tree); + spin_unlock_irqrestore(&cgwb_lock, flags); + kfree(congested); + + if (atomic_dec_and_test(&bdi->usage_cnt)) + wake_up_all(&cgwb_release_wait); +} + +static void cgwb_release_workfn(struct work_struct *work) +{ + struct bdi_writeback *wb = container_of(work, struct bdi_writeback, + release_work); + struct backing_dev_info *bdi = wb->bdi; + + wb_shutdown(wb); + + css_put(wb->memcg_css); + css_put(wb->blkcg_css); + wb_congested_put(wb->congested); + + percpu_ref_exit(&wb->refcnt); + wb_exit(wb); + kfree_rcu(wb, rcu); + + if (atomic_dec_and_test(&bdi->usage_cnt)) + wake_up_all(&cgwb_release_wait); +} + +static void cgwb_release(struct percpu_ref *refcnt) +{ + struct bdi_writeback *wb = container_of(refcnt, struct bdi_writeback, + refcnt); + schedule_work(&wb->release_work); +} + +static void cgwb_kill(struct bdi_writeback *wb) +{ + lockdep_assert_held(&cgwb_lock); + + WARN_ON(!radix_tree_delete(&wb->bdi->cgwb_tree, wb->memcg_css->id)); + list_del(&wb->memcg_node); + list_del(&wb->blkcg_node); + percpu_ref_kill(&wb->refcnt); +} + +static int cgwb_create(struct backing_dev_info *bdi, + struct cgroup_subsys_state *memcg_css, gfp_t gfp) +{ + struct mem_cgroup *memcg; + struct cgroup_subsys_state *blkcg_css; + struct blkcg *blkcg; + struct list_head *memcg_cgwb_list, *blkcg_cgwb_list; + struct bdi_writeback *wb; + unsigned long flags; + int ret = 0; + + memcg = mem_cgroup_from_css(memcg_css); + blkcg_css = cgroup_get_e_css(memcg_css->cgroup, &blkio_cgrp_subsys); + blkcg = css_to_blkcg(blkcg_css); + memcg_cgwb_list = mem_cgroup_cgwb_list(memcg); + blkcg_cgwb_list = &blkcg->cgwb_list; + + /* look up again under lock and discard on blkcg mismatch */ + spin_lock_irqsave(&cgwb_lock, flags); + wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id); + if (wb && wb->blkcg_css != blkcg_css) { + cgwb_kill(wb); + wb = NULL; + } + spin_unlock_irqrestore(&cgwb_lock, flags); + if (wb) + goto out_put; + + /* need to create a new one */ + wb = kmalloc(sizeof(*wb), gfp); + if (!wb) + return -ENOMEM; + + ret = wb_init(wb, bdi, gfp); + if (ret) + goto err_free; + + ret = percpu_ref_init(&wb->refcnt, cgwb_release, 0, gfp); + if (ret) + goto err_wb_exit; + + wb->congested = wb_congested_get_create(bdi, blkcg_css->id, gfp); + if (!wb->congested) + goto err_ref_exit; + + wb->memcg_css = memcg_css; + wb->blkcg_css = blkcg_css; + INIT_WORK(&wb->release_work, cgwb_release_workfn); + set_bit(WB_registered, &wb->state); + + /* + * The root wb determines the registered state of the whole bdi and + * memcg_cgwb_list and blkcg_cgwb_list's next pointers indicate + * whether they're still online. Don't link @wb if any is dead. + * See wb_memcg_offline() and wb_blkcg_offline(). + */ + ret = -ENODEV; + spin_lock_irqsave(&cgwb_lock, flags); + if (test_bit(WB_registered, &bdi->wb.state) && + blkcg_cgwb_list->next && memcg_cgwb_list->next) { + /* we might have raced another instance of this function */ + ret = radix_tree_insert(&bdi->cgwb_tree, memcg_css->id, wb); + if (!ret) { + atomic_inc(&bdi->usage_cnt); + list_add(&wb->memcg_node, memcg_cgwb_list); + list_add(&wb->blkcg_node, blkcg_cgwb_list); + css_get(memcg_css); + css_get(blkcg_css); + } + } + spin_unlock_irqrestore(&cgwb_lock, flags); + if (ret) { + if (ret == -EEXIST) + ret = 0; + goto err_put_congested; + } + goto out_put; + +err_put_congested: + wb_congested_put(wb->congested); +err_ref_exit: + percpu_ref_exit(&wb->refcnt); +err_wb_exit: + wb_exit(wb); +err_free: + kfree(wb); +out_put: + css_put(blkcg_css); + return ret; +} + +/** + * wb_get_create - get wb for a given memcg, create if necessary + * @bdi: target bdi + * @memcg_css: cgroup_subsys_state of the target memcg (must have positive ref) + * @gfp: allocation mask to use + * + * Try to get the wb for @memcg_css on @bdi. If it doesn't exist, try to + * create one. The returned wb has its refcount incremented. + * + * This function uses css_get() on @memcg_css and thus expects its refcnt + * to be positive on invocation. IOW, rcu_read_lock() protection on + * @memcg_css isn't enough. try_get it before calling this function. + * + * A wb is keyed by its associated memcg. As blkcg implicitly enables + * memcg on the default hierarchy, memcg association is guaranteed to be + * more specific (equal or descendant to the associated blkcg) and thus can + * identify both the memcg and blkcg associations. + * + * Because the blkcg associated with a memcg may change as blkcg is enabled + * and disabled closer to root in the hierarchy, each wb keeps track of + * both the memcg and blkcg associated with it and verifies the blkcg on + * each lookup. On mismatch, the existing wb is discarded and a new one is + * created. + */ +struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi, + struct cgroup_subsys_state *memcg_css, + gfp_t gfp) +{ + struct bdi_writeback *wb; + + might_sleep_if(gfp & __GFP_WAIT); + + if (!memcg_css->parent) + return &bdi->wb; + + do { + rcu_read_lock(); + wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id); + if (wb) { + struct cgroup_subsys_state *blkcg_css; + + /* see whether the blkcg association has changed */ + blkcg_css = cgroup_get_e_css(memcg_css->cgroup, + &blkio_cgrp_subsys); + if (unlikely(wb->blkcg_css != blkcg_css || + !wb_tryget(wb))) + wb = NULL; + css_put(blkcg_css); + } + rcu_read_unlock(); + } while (!wb && !cgwb_create(bdi, memcg_css, gfp)); + + return wb; +} + +void __inode_attach_wb(struct inode *inode, struct page *page) +{ + struct backing_dev_info *bdi = inode_to_bdi(inode); + struct bdi_writeback *wb = NULL; + + if (inode_cgwb_enabled(inode)) { + struct cgroup_subsys_state *memcg_css; + + if (page) { + memcg_css = mem_cgroup_css_from_page(page); + wb = wb_get_create(bdi, memcg_css, GFP_ATOMIC); + } else { + /* must pin memcg_css, see wb_get_create() */ + memcg_css = task_get_css(current, memory_cgrp_id); + wb = wb_get_create(bdi, memcg_css, GFP_ATOMIC); + css_put(memcg_css); + } + } + + if (!wb) + wb = &bdi->wb; + + /* + * There may be multiple instances of this function racing to + * update the same inode. Use cmpxchg() to tell the winner. + */ + if (unlikely(cmpxchg(&inode->i_wb, NULL, wb))) + wb_put(wb); +} + +static void cgwb_bdi_init(struct backing_dev_info *bdi) +{ + bdi->wb.memcg_css = mem_cgroup_root_css; + bdi->wb.blkcg_css = blkcg_root_css; + bdi->wb_congested.blkcg_id = 1; + INIT_RADIX_TREE(&bdi->cgwb_tree, GFP_ATOMIC); + bdi->cgwb_congested_tree = RB_ROOT; + atomic_set(&bdi->usage_cnt, 1); +} + +static void cgwb_bdi_destroy(struct backing_dev_info *bdi) +{ + struct radix_tree_iter iter; + void **slot; + + WARN_ON(test_bit(WB_registered, &bdi->wb.state)); + + spin_lock_irq(&cgwb_lock); + radix_tree_for_each_slot(slot, &bdi->cgwb_tree, &iter, 0) + cgwb_kill(*slot); + spin_unlock_irq(&cgwb_lock); + + /* + * All cgwb's and their congested states must be shutdown and + * released before returning. Drain the usage counter to wait for + * all cgwb's and cgwb_congested's ever created on @bdi. + */ + atomic_dec(&bdi->usage_cnt); + wait_event(cgwb_release_wait, !atomic_read(&bdi->usage_cnt)); +} + +/** + * wb_memcg_offline - kill all wb's associated with a memcg being offlined + * @memcg: memcg being offlined + * + * Also prevents creation of any new wb's associated with @memcg. + */ +void wb_memcg_offline(struct mem_cgroup *memcg) +{ + LIST_HEAD(to_destroy); + struct list_head *memcg_cgwb_list = mem_cgroup_cgwb_list(memcg); + struct bdi_writeback *wb, *next; + + spin_lock_irq(&cgwb_lock); + list_for_each_entry_safe(wb, next, memcg_cgwb_list, memcg_node) + cgwb_kill(wb); + memcg_cgwb_list->next = NULL; /* prevent new wb's */ + spin_unlock_irq(&cgwb_lock); +} + +/** + * wb_blkcg_offline - kill all wb's associated with a blkcg being offlined + * @blkcg: blkcg being offlined + * + * Also prevents creation of any new wb's associated with @blkcg. + */ +void wb_blkcg_offline(struct blkcg *blkcg) +{ + LIST_HEAD(to_destroy); + struct bdi_writeback *wb, *next; + + spin_lock_irq(&cgwb_lock); + list_for_each_entry_safe(wb, next, &blkcg->cgwb_list, blkcg_node) + cgwb_kill(wb); + blkcg->cgwb_list.next = NULL; /* prevent new wb's */ + spin_unlock_irq(&cgwb_lock); +} + +#else /* CONFIG_CGROUP_WRITEBACK */ + +static void cgwb_bdi_init(struct backing_dev_info *bdi) { } +static void cgwb_bdi_destroy(struct backing_dev_info *bdi) { } + +#endif /* CONFIG_CGROUP_WRITEBACK */ + int bdi_init(struct backing_dev_info *bdi) { int err; @@ -386,6 +781,7 @@ int bdi_init(struct backing_dev_info *bdi) bdi->wb_congested.state = 0; bdi->wb.congested = &bdi->wb_congested; + cgwb_bdi_init(bdi); return 0; } EXPORT_SYMBOL(bdi_init); @@ -459,6 +855,7 @@ void bdi_destroy(struct backing_dev_info *bdi) /* make sure nobody finds us on the bdi_list anymore */ bdi_remove_from_list(bdi); wb_shutdown(&bdi->wb); + cgwb_bdi_destroy(bdi); if (bdi->dev) { bdi_debug_unregister(bdi); diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 5c270a03729c..49e5aa6abca6 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -348,6 +348,10 @@ struct mem_cgroup { atomic_t numainfo_updating; #endif +#ifdef CONFIG_CGROUP_WRITEBACK + struct list_head cgwb_list; +#endif + /* List of events which userspace want to receive */ struct list_head event_list; spinlock_t event_list_lock; @@ -4030,6 +4034,15 @@ static void memcg_destroy_kmem(struct mem_cgroup *memcg) } #endif +#ifdef CONFIG_CGROUP_WRITEBACK + +struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg) +{ + return &memcg->cgwb_list; +} + +#endif /* CONFIG_CGROUP_WRITEBACK */ + /* * DO NOT USE IN NEW FILES. * @@ -4494,7 +4507,9 @@ mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css) #ifdef CONFIG_MEMCG_KMEM memcg->kmemcg_id = -1; #endif - +#ifdef CONFIG_CGROUP_WRITEBACK + INIT_LIST_HEAD(&memcg->cgwb_list); +#endif return &memcg->css; free_out: @@ -4582,6 +4597,8 @@ static void mem_cgroup_css_offline(struct cgroup_subsys_state *css) vmpressure_cleanup(&memcg->vmpressure); memcg_deactivate_kmem(memcg); + + wb_memcg_offline(memcg); } static void mem_cgroup_css_free(struct cgroup_subsys_state *css) diff --git a/mm/page-writeback.c b/mm/page-writeback.c index 78ef5511a198..9b95cf80b407 100644 --- a/mm/page-writeback.c +++ b/mm/page-writeback.c @@ -2097,16 +2097,21 @@ int __set_page_dirty_no_writeback(struct page *page) void account_page_dirtied(struct page *page, struct address_space *mapping, struct mem_cgroup *memcg) { + struct inode *inode = mapping->host; + trace_writeback_dirty_page(page, mapping); if (mapping_cap_account_dirty(mapping)) { - struct backing_dev_info *bdi = inode_to_bdi(mapping->host); + struct bdi_writeback *wb; + + inode_attach_wb(inode, page); + wb = inode_to_wb(inode); mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_DIRTY); __inc_zone_page_state(page, NR_FILE_DIRTY); __inc_zone_page_state(page, NR_DIRTIED); - __inc_wb_stat(&bdi->wb, WB_RECLAIMABLE); - __inc_wb_stat(&bdi->wb, WB_DIRTIED); + __inc_wb_stat(wb, WB_RECLAIMABLE); + __inc_wb_stat(wb, WB_DIRTIED); task_io_account_write(PAGE_CACHE_SIZE); current->nr_dirtied++; this_cpu_inc(bdp_ratelimits); -- cgit v1.2.3-70-g09d2 From 733a572e66d2a23c852fdce34dba5bbd40667817 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 18:23:18 -0400 Subject: memcg: make mem_cgroup_read_{stat|event}() iterate possible cpus instead of online cpu_possible_mask represents the CPUs which are actually possible during that boot instance. For systems which don't support CPU hotplug, this will match cpu_online_mask exactly in most cases. Even for systems which support CPU hotplug, the number of possible CPU slots is highly unlikely to diverge greatly from the number of online CPUs. The only cases where the difference between possible and online caused problems were when the boot code failed to initialize the possible mask and left it fully set at NR_CPUS - 1. As such, most per-cpu constructs allocate for all possible CPUs and often iterate over the possibles, which also has the benefit of avoiding the blocking CPU hotplug synchronization. memcg open codes per-cpu stat counting for mem_cgroup_read_stat() and mem_cgroup_read_events(), which iterates over online CPUs and handles CPU hotplug operations explicitly. This complexity doesn't actually buy anything. Switch to iterating over the possibles and drop the explicit CPU hotplug handling. Eventually, we want to convert memcg to use percpu_counter instead of its own custom implementation which also benefits from quick access w/o summing for cases where larger error margin is acceptable. This will allow mem_cgroup_read_stat() to be called from non-sleepable contexts which will be used by cgroup writeback. Signed-off-by: Tejun Heo Cc: Michal Hocko Acked-by: Johannes Weiner Signed-off-by: Jens Axboe --- mm/memcontrol.c | 51 ++------------------------------------------------- 1 file changed, 2 insertions(+), 49 deletions(-) (limited to 'mm/memcontrol.c') diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 49e5aa6abca6..701cbee9acba 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -324,11 +324,6 @@ struct mem_cgroup { * percpu counter. */ struct mem_cgroup_stat_cpu __percpu *stat; - /* - * used when a cpu is offlined or other synchronizations - * See mem_cgroup_read_stat(). - */ - struct mem_cgroup_stat_cpu nocpu_base; spinlock_t pcp_counter_lock; #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET) @@ -834,15 +829,8 @@ static long mem_cgroup_read_stat(struct mem_cgroup *memcg, long val = 0; int cpu; - get_online_cpus(); - for_each_online_cpu(cpu) + for_each_possible_cpu(cpu) val += per_cpu(memcg->stat->count[idx], cpu); -#ifdef CONFIG_HOTPLUG_CPU - spin_lock(&memcg->pcp_counter_lock); - val += memcg->nocpu_base.count[idx]; - spin_unlock(&memcg->pcp_counter_lock); -#endif - put_online_cpus(); return val; } @@ -852,15 +840,8 @@ static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg, unsigned long val = 0; int cpu; - get_online_cpus(); - for_each_online_cpu(cpu) + for_each_possible_cpu(cpu) val += per_cpu(memcg->stat->events[idx], cpu); -#ifdef CONFIG_HOTPLUG_CPU - spin_lock(&memcg->pcp_counter_lock); - val += memcg->nocpu_base.events[idx]; - spin_unlock(&memcg->pcp_counter_lock); -#endif - put_online_cpus(); return val; } @@ -2210,37 +2191,12 @@ static void drain_all_stock(struct mem_cgroup *root_memcg) mutex_unlock(&percpu_charge_mutex); } -/* - * This function drains percpu counter value from DEAD cpu and - * move it to local cpu. Note that this function can be preempted. - */ -static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu) -{ - int i; - - spin_lock(&memcg->pcp_counter_lock); - for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) { - long x = per_cpu(memcg->stat->count[i], cpu); - - per_cpu(memcg->stat->count[i], cpu) = 0; - memcg->nocpu_base.count[i] += x; - } - for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) { - unsigned long x = per_cpu(memcg->stat->events[i], cpu); - - per_cpu(memcg->stat->events[i], cpu) = 0; - memcg->nocpu_base.events[i] += x; - } - spin_unlock(&memcg->pcp_counter_lock); -} - static int memcg_cpu_hotplug_callback(struct notifier_block *nb, unsigned long action, void *hcpu) { int cpu = (unsigned long)hcpu; struct memcg_stock_pcp *stock; - struct mem_cgroup *iter; if (action == CPU_ONLINE) return NOTIFY_OK; @@ -2248,9 +2204,6 @@ static int memcg_cpu_hotplug_callback(struct notifier_block *nb, if (action != CPU_DEAD && action != CPU_DEAD_FROZEN) return NOTIFY_OK; - for_each_mem_cgroup(iter) - mem_cgroup_drain_pcp_counter(iter, cpu); - stock = &per_cpu(memcg_stock, cpu); drain_stock(stock); return NOTIFY_OK; -- cgit v1.2.3-70-g09d2 From 841710aa6e4acd066ab9fe8c8cb6f4e4e6709d83 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 18:23:33 -0400 Subject: writeback: implement memcg wb_domain Dirtyable memory is distributed to a wb (bdi_writeback) according to the relative bandwidth the wb is writing out in the whole system. This distribution is global - each wb is measured against all other wb's and gets the proportinately sized portion of the memory in the whole system. For cgroup writeback, the amount of dirtyable memory is scoped by memcg and thus each wb would need to be measured and controlled in its memcg. IOW, a wb will belong to two writeback domains - the global and memcg domains. The previous patches laid the groundwork to support the two wb_domains and this patch implements memcg wb_domain. memcg->cgwb_domain is initialized on css online and destroyed on css release, wb->memcg_completions is added, and __wb_writeout_inc() is updated to increment completions against both global and memcg wb_domains. The following patches will update balance_dirty_pages() and its subroutines to actually consider memcg wb_domain for throttling. Signed-off-by: Tejun Heo Cc: Jens Axboe Cc: Jan Kara Cc: Wu Fengguang Cc: Greg Thelen Signed-off-by: Jens Axboe --- include/linux/backing-dev-defs.h | 1 + include/linux/memcontrol.h | 12 +++++++++++- include/linux/writeback.h | 3 +++ mm/backing-dev.c | 9 ++++++++- mm/memcontrol.c | 39 +++++++++++++++++++++++++++++++++++++++ mm/page-writeback.c | 25 +++++++++++++++++++++++++ 6 files changed, 87 insertions(+), 2 deletions(-) (limited to 'mm/memcontrol.c') diff --git a/include/linux/backing-dev-defs.h b/include/linux/backing-dev-defs.h index 97a92fa0cdb5..8d470b73824f 100644 --- a/include/linux/backing-dev-defs.h +++ b/include/linux/backing-dev-defs.h @@ -118,6 +118,7 @@ struct bdi_writeback { #ifdef CONFIG_CGROUP_WRITEBACK struct percpu_ref refcnt; /* used only for !root wb's */ + struct fprop_local_percpu memcg_completions; struct cgroup_subsys_state *memcg_css; /* the associated memcg */ struct cgroup_subsys_state *blkcg_css; /* and blkcg */ struct list_head memcg_node; /* anchored at memcg->cgwb_list */ diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 662a953ea8ad..e3177bed23ea 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -389,8 +389,18 @@ enum { }; #ifdef CONFIG_CGROUP_WRITEBACK + struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); -#endif +struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); + +#else /* CONFIG_CGROUP_WRITEBACK */ + +static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) +{ + return NULL; +} + +#endif /* CONFIG_CGROUP_WRITEBACK */ struct sock; #if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM) diff --git a/include/linux/writeback.h b/include/linux/writeback.h index b57c2786b5aa..04a3786c456f 100644 --- a/include/linux/writeback.h +++ b/include/linux/writeback.h @@ -167,6 +167,9 @@ static inline void laptop_sync_completion(void) { } void throttle_vm_writeout(gfp_t gfp_mask); bool zone_dirty_ok(struct zone *zone); int wb_domain_init(struct wb_domain *dom, gfp_t gfp); +#ifdef CONFIG_CGROUP_WRITEBACK +void wb_domain_exit(struct wb_domain *dom); +#endif extern struct wb_domain global_wb_domain; diff --git a/mm/backing-dev.c b/mm/backing-dev.c index 9c8b7b5a0eee..84ebf7c8d006 100644 --- a/mm/backing-dev.c +++ b/mm/backing-dev.c @@ -482,6 +482,7 @@ static void cgwb_release_workfn(struct work_struct *work) css_put(wb->blkcg_css); wb_congested_put(wb->congested); + fprop_local_destroy_percpu(&wb->memcg_completions); percpu_ref_exit(&wb->refcnt); wb_exit(wb); kfree_rcu(wb, rcu); @@ -548,9 +549,13 @@ static int cgwb_create(struct backing_dev_info *bdi, if (ret) goto err_wb_exit; + ret = fprop_local_init_percpu(&wb->memcg_completions, gfp); + if (ret) + goto err_ref_exit; + wb->congested = wb_congested_get_create(bdi, blkcg_css->id, gfp); if (!wb->congested) - goto err_ref_exit; + goto err_fprop_exit; wb->memcg_css = memcg_css; wb->blkcg_css = blkcg_css; @@ -587,6 +592,8 @@ static int cgwb_create(struct backing_dev_info *bdi, err_put_congested: wb_congested_put(wb->congested); +err_fprop_exit: + fprop_local_destroy_percpu(&wb->memcg_completions); err_ref_exit: percpu_ref_exit(&wb->refcnt); err_wb_exit: diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 701cbee9acba..ce113ddf2fb5 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -345,6 +345,7 @@ struct mem_cgroup { #ifdef CONFIG_CGROUP_WRITEBACK struct list_head cgwb_list; + struct wb_domain cgwb_domain; #endif /* List of events which userspace want to receive */ @@ -3994,6 +3995,37 @@ struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg) return &memcg->cgwb_list; } +static int memcg_wb_domain_init(struct mem_cgroup *memcg, gfp_t gfp) +{ + return wb_domain_init(&memcg->cgwb_domain, gfp); +} + +static void memcg_wb_domain_exit(struct mem_cgroup *memcg) +{ + wb_domain_exit(&memcg->cgwb_domain); +} + +struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) +{ + struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css); + + if (!memcg->css.parent) + return NULL; + + return &memcg->cgwb_domain; +} + +#else /* CONFIG_CGROUP_WRITEBACK */ + +static int memcg_wb_domain_init(struct mem_cgroup *memcg, gfp_t gfp) +{ + return 0; +} + +static void memcg_wb_domain_exit(struct mem_cgroup *memcg) +{ +} + #endif /* CONFIG_CGROUP_WRITEBACK */ /* @@ -4380,9 +4412,15 @@ static struct mem_cgroup *mem_cgroup_alloc(void) memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu); if (!memcg->stat) goto out_free; + + if (memcg_wb_domain_init(memcg, GFP_KERNEL)) + goto out_free_stat; + spin_lock_init(&memcg->pcp_counter_lock); return memcg; +out_free_stat: + free_percpu(memcg->stat); out_free: kfree(memcg); return NULL; @@ -4409,6 +4447,7 @@ static void __mem_cgroup_free(struct mem_cgroup *memcg) free_mem_cgroup_per_zone_info(memcg, node); free_percpu(memcg->stat); + memcg_wb_domain_exit(memcg); kfree(memcg); } diff --git a/mm/page-writeback.c b/mm/page-writeback.c index a7ba5cee950b..a146e3389e78 100644 --- a/mm/page-writeback.c +++ b/mm/page-writeback.c @@ -171,6 +171,11 @@ static struct dirty_throttle_control *mdtc_gdtc(struct dirty_throttle_control *m return mdtc->gdtc; } +static struct fprop_local_percpu *wb_memcg_completions(struct bdi_writeback *wb) +{ + return &wb->memcg_completions; +} + static void wb_min_max_ratio(struct bdi_writeback *wb, unsigned long *minp, unsigned long *maxp) { @@ -213,6 +218,11 @@ static struct dirty_throttle_control *mdtc_gdtc(struct dirty_throttle_control *m return NULL; } +static struct fprop_local_percpu *wb_memcg_completions(struct bdi_writeback *wb) +{ + return NULL; +} + static void wb_min_max_ratio(struct bdi_writeback *wb, unsigned long *minp, unsigned long *maxp) { @@ -530,9 +540,16 @@ static void wb_domain_writeout_inc(struct wb_domain *dom, */ static inline void __wb_writeout_inc(struct bdi_writeback *wb) { + struct wb_domain *cgdom; + __inc_wb_stat(wb, WB_WRITTEN); wb_domain_writeout_inc(&global_wb_domain, &wb->completions, wb->bdi->max_prop_frac); + + cgdom = mem_cgroup_wb_domain(wb); + if (cgdom) + wb_domain_writeout_inc(cgdom, wb_memcg_completions(wb), + wb->bdi->max_prop_frac); } void wb_writeout_inc(struct bdi_writeback *wb) @@ -583,6 +600,14 @@ int wb_domain_init(struct wb_domain *dom, gfp_t gfp) return fprop_global_init(&dom->completions, gfp); } +#ifdef CONFIG_CGROUP_WRITEBACK +void wb_domain_exit(struct wb_domain *dom) +{ + del_timer_sync(&dom->period_timer); + fprop_global_destroy(&dom->completions); +} +#endif + /* * bdi_min_ratio keeps the sum of the minimum dirty shares of all * registered backing devices, which, for obvious reasons, can not -- cgit v1.2.3-70-g09d2 From 2529bb3aadc40a93e642f5f3650f63379a964467 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 18:23:34 -0400 Subject: writeback: reset wb_domain->dirty_limit[_tstmp] when memcg domain size changes The amount of available memory to a memcg wb_domain can change as memcg configuration changes. A domain's ->dirty_limit exists to smooth out sudden drops in dirty threshold; however, when a domain's size actually drops significantly, it hinders the dirty throttling from adjusting to the new configuration leading to unexpected behaviors including unnecessary OOM kills. This patch resolves the issue by adding wb_domain_size_changed() which resets ->dirty_limit[_tstmp] and making memcg call it on configuration changes. Signed-off-by: Tejun Heo Cc: Jens Axboe Cc: Jan Kara Cc: Wu Fengguang Cc: Greg Thelen Signed-off-by: Jens Axboe --- include/linux/writeback.h | 20 ++++++++++++++++++++ mm/memcontrol.c | 12 ++++++++++++ 2 files changed, 32 insertions(+) (limited to 'mm/memcontrol.c') diff --git a/include/linux/writeback.h b/include/linux/writeback.h index 04a3786c456f..3b73e97ecfc7 100644 --- a/include/linux/writeback.h +++ b/include/linux/writeback.h @@ -132,6 +132,26 @@ struct wb_domain { unsigned long dirty_limit; }; +/** + * wb_domain_size_changed - memory available to a wb_domain has changed + * @dom: wb_domain of interest + * + * This function should be called when the amount of memory available to + * @dom has changed. It resets @dom's dirty limit parameters to prevent + * the past values which don't match the current configuration from skewing + * dirty throttling. Without this, when memory size of a wb_domain is + * greatly reduced, the dirty throttling logic may allow too many pages to + * be dirtied leading to consecutive unnecessary OOMs and may get stuck in + * that situation. + */ +static inline void wb_domain_size_changed(struct wb_domain *dom) +{ + spin_lock(&dom->lock); + dom->dirty_limit_tstamp = jiffies; + dom->dirty_limit = 0; + spin_unlock(&dom->lock); +} + /* * fs/fs-writeback.c */ diff --git a/mm/memcontrol.c b/mm/memcontrol.c index ce113ddf2fb5..c0b0406ae5ca 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -4005,6 +4005,11 @@ static void memcg_wb_domain_exit(struct mem_cgroup *memcg) wb_domain_exit(&memcg->cgwb_domain); } +static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg) +{ + wb_domain_size_changed(&memcg->cgwb_domain); +} + struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) { struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css); @@ -4026,6 +4031,10 @@ static void memcg_wb_domain_exit(struct mem_cgroup *memcg) { } +static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg) +{ +} + #endif /* CONFIG_CGROUP_WRITEBACK */ /* @@ -4624,6 +4633,7 @@ static void mem_cgroup_css_reset(struct cgroup_subsys_state *css) memcg->low = 0; memcg->high = PAGE_COUNTER_MAX; memcg->soft_limit = PAGE_COUNTER_MAX; + memcg_wb_domain_size_changed(memcg); } #ifdef CONFIG_MMU @@ -5361,6 +5371,7 @@ static ssize_t memory_high_write(struct kernfs_open_file *of, memcg->high = high; + memcg_wb_domain_size_changed(memcg); return nbytes; } @@ -5393,6 +5404,7 @@ static ssize_t memory_max_write(struct kernfs_open_file *of, if (err) return err; + memcg_wb_domain_size_changed(memcg); return nbytes; } -- cgit v1.2.3-70-g09d2 From c2aa723a6093633ae4ec15b08a4db276643cab3e Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 22 May 2015 18:23:35 -0400 Subject: writeback: implement memcg writeback domain based throttling While cgroup writeback support now connects memcg and blkcg so that writeback IOs are properly attributed and controlled, the IO back pressure propagation mechanism implemented in balance_dirty_pages() and its subroutines wasn't aware of cgroup writeback. Processes belonging to a memcg may have access to only subset of total memory available in the system and not factoring this into dirty throttling rendered it completely ineffective for processes under memcg limits and memcg ended up building a separate ad-hoc degenerate mechanism directly into vmscan code to limit page dirtying. The previous patches updated balance_dirty_pages() and its subroutines so that they can deal with multiple wb_domain's (writeback domains) and defined per-memcg wb_domain. Processes belonging to a non-root memcg are bound to two wb_domains, global wb_domain and memcg wb_domain, and should be throttled according to IO pressures from both domains. This patch updates dirty throttling code so that it repeats similar calculations for the two domains - the differences between the two are few and minor - and applies the lower of the two sets of resulting constraints. wb_over_bg_thresh(), which controls when background writeback terminates, is also updated to consider both global and memcg wb_domains. It returns true if dirty is over bg_thresh for either domain. This makes the dirty throttling mechanism operational for memcg domains including writeback-bandwidth-proportional dirty page distribution inside them but the ad-hoc memcg throttling mechanism in vmscan is still in place. The next patch will rip it out. Signed-off-by: Tejun Heo Cc: Jens Axboe Cc: Jan Kara Cc: Wu Fengguang Cc: Greg Thelen Signed-off-by: Jens Axboe --- include/linux/memcontrol.h | 9 +++ mm/memcontrol.c | 43 ++++++++++++ mm/page-writeback.c | 158 ++++++++++++++++++++++++++++++++++++++------- 3 files changed, 188 insertions(+), 22 deletions(-) (limited to 'mm/memcontrol.c') diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index e3177bed23ea..c3eb19e2bc1c 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -392,6 +392,8 @@ enum { struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg); struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb); +void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pavail, + unsigned long *pdirty, unsigned long *pwriteback); #else /* CONFIG_CGROUP_WRITEBACK */ @@ -400,6 +402,13 @@ static inline struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) return NULL; } +static inline void mem_cgroup_wb_stats(struct bdi_writeback *wb, + unsigned long *pavail, + unsigned long *pdirty, + unsigned long *pwriteback) +{ +} + #endif /* CONFIG_CGROUP_WRITEBACK */ struct sock; diff --git a/mm/memcontrol.c b/mm/memcontrol.c index c0b0406ae5ca..f816d91c643b 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -4020,6 +4020,49 @@ struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb) return &memcg->cgwb_domain; } +/** + * mem_cgroup_wb_stats - retrieve writeback related stats from its memcg + * @wb: bdi_writeback in question + * @pavail: out parameter for number of available pages + * @pdirty: out parameter for number of dirty pages + * @pwriteback: out parameter for number of pages under writeback + * + * Determine the numbers of available, dirty, and writeback pages in @wb's + * memcg. Dirty and writeback are self-explanatory. Available is a bit + * more involved. + * + * A memcg's headroom is "min(max, high) - used". The available memory is + * calculated as the lowest headroom of itself and the ancestors plus the + * number of pages already being used for file pages. Note that this + * doesn't consider the actual amount of available memory in the system. + * The caller should further cap *@pavail accordingly. + */ +void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pavail, + unsigned long *pdirty, unsigned long *pwriteback) +{ + struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css); + struct mem_cgroup *parent; + unsigned long head_room = PAGE_COUNTER_MAX; + unsigned long file_pages; + + *pdirty = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_DIRTY); + + /* this should eventually include NR_UNSTABLE_NFS */ + *pwriteback = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_WRITEBACK); + + file_pages = mem_cgroup_nr_lru_pages(memcg, (1 << LRU_INACTIVE_FILE) | + (1 << LRU_ACTIVE_FILE)); + while ((parent = parent_mem_cgroup(memcg))) { + unsigned long ceiling = min(memcg->memory.limit, memcg->high); + unsigned long used = page_counter_read(&memcg->memory); + + head_room = min(head_room, ceiling - min(ceiling, used)); + memcg = parent; + } + + *pavail = file_pages + head_room; +} + #else /* CONFIG_CGROUP_WRITEBACK */ static int memcg_wb_domain_init(struct mem_cgroup *memcg, gfp_t gfp) diff --git a/mm/page-writeback.c b/mm/page-writeback.c index a146e3389e78..e8903356f423 100644 --- a/mm/page-writeback.c +++ b/mm/page-writeback.c @@ -160,6 +160,14 @@ struct dirty_throttle_control { #define GDTC_INIT(__wb) .dom = &global_wb_domain, \ DTC_INIT_COMMON(__wb) #define GDTC_INIT_NO_WB .dom = &global_wb_domain +#define MDTC_INIT(__wb, __gdtc) .dom = mem_cgroup_wb_domain(__wb), \ + .gdtc = __gdtc, \ + DTC_INIT_COMMON(__wb) + +static bool mdtc_valid(struct dirty_throttle_control *dtc) +{ + return dtc->dom; +} static struct wb_domain *dtc_dom(struct dirty_throttle_control *dtc) { @@ -207,6 +215,12 @@ static void wb_min_max_ratio(struct bdi_writeback *wb, #define GDTC_INIT(__wb) DTC_INIT_COMMON(__wb) #define GDTC_INIT_NO_WB +#define MDTC_INIT(__wb, __gdtc) + +static bool mdtc_valid(struct dirty_throttle_control *dtc) +{ + return false; +} static struct wb_domain *dtc_dom(struct dirty_throttle_control *dtc) { @@ -668,6 +682,15 @@ static unsigned long hard_dirty_limit(struct wb_domain *dom, return max(thresh, dom->dirty_limit); } +/* memory available to a memcg domain is capped by system-wide clean memory */ +static void mdtc_cap_avail(struct dirty_throttle_control *mdtc) +{ + struct dirty_throttle_control *gdtc = mdtc_gdtc(mdtc); + unsigned long clean = gdtc->avail - min(gdtc->avail, gdtc->dirty); + + mdtc->avail = min(mdtc->avail, clean); +} + /** * __wb_calc_thresh - @wb's share of dirty throttling threshold * @dtc: dirty_throttle_context of interest @@ -1269,11 +1292,12 @@ static void wb_update_dirty_ratelimit(struct dirty_throttle_control *dtc, trace_bdi_dirty_ratelimit(wb->bdi, dirty_rate, task_ratelimit); } -static void __wb_update_bandwidth(struct dirty_throttle_control *dtc, +static void __wb_update_bandwidth(struct dirty_throttle_control *gdtc, + struct dirty_throttle_control *mdtc, unsigned long start_time, bool update_ratelimit) { - struct bdi_writeback *wb = dtc->wb; + struct bdi_writeback *wb = gdtc->wb; unsigned long now = jiffies; unsigned long elapsed = now - wb->bw_time_stamp; unsigned long dirtied; @@ -1298,8 +1322,17 @@ static void __wb_update_bandwidth(struct dirty_throttle_control *dtc, goto snapshot; if (update_ratelimit) { - domain_update_bandwidth(dtc, now); - wb_update_dirty_ratelimit(dtc, dirtied, elapsed); + domain_update_bandwidth(gdtc, now); + wb_update_dirty_ratelimit(gdtc, dirtied, elapsed); + + /* + * @mdtc is always NULL if !CGROUP_WRITEBACK but the + * compiler has no way to figure that out. Help it. + */ + if (IS_ENABLED(CONFIG_CGROUP_WRITEBACK) && mdtc) { + domain_update_bandwidth(mdtc, now); + wb_update_dirty_ratelimit(mdtc, dirtied, elapsed); + } } wb_update_write_bandwidth(wb, elapsed, written); @@ -1313,7 +1346,7 @@ void wb_update_bandwidth(struct bdi_writeback *wb, unsigned long start_time) { struct dirty_throttle_control gdtc = { GDTC_INIT(wb) }; - __wb_update_bandwidth(&gdtc, start_time, false); + __wb_update_bandwidth(&gdtc, NULL, start_time, false); } /* @@ -1480,7 +1513,11 @@ static void balance_dirty_pages(struct address_space *mapping, unsigned long pages_dirtied) { struct dirty_throttle_control gdtc_stor = { GDTC_INIT(wb) }; + struct dirty_throttle_control mdtc_stor = { MDTC_INIT(wb, &gdtc_stor) }; struct dirty_throttle_control * const gdtc = &gdtc_stor; + struct dirty_throttle_control * const mdtc = mdtc_valid(&mdtc_stor) ? + &mdtc_stor : NULL; + struct dirty_throttle_control *sdtc; unsigned long nr_reclaimable; /* = file_dirty + unstable_nfs */ long period; long pause; @@ -1497,6 +1534,7 @@ static void balance_dirty_pages(struct address_space *mapping, for (;;) { unsigned long now = jiffies; unsigned long dirty, thresh, bg_thresh; + unsigned long m_dirty, m_thresh, m_bg_thresh; /* * Unstable writes are a feature of certain networked @@ -1523,6 +1561,32 @@ static void balance_dirty_pages(struct address_space *mapping, bg_thresh = gdtc->bg_thresh; } + if (mdtc) { + unsigned long writeback; + + /* + * If @wb belongs to !root memcg, repeat the same + * basic calculations for the memcg domain. + */ + mem_cgroup_wb_stats(wb, &mdtc->avail, &mdtc->dirty, + &writeback); + mdtc_cap_avail(mdtc); + mdtc->dirty += writeback; + + domain_dirty_limits(mdtc); + + if (unlikely(strictlimit)) { + wb_dirty_limits(mdtc); + m_dirty = mdtc->wb_dirty; + m_thresh = mdtc->wb_thresh; + m_bg_thresh = mdtc->wb_bg_thresh; + } else { + m_dirty = mdtc->dirty; + m_thresh = mdtc->thresh; + m_bg_thresh = mdtc->bg_thresh; + } + } + /* * Throttle it only when the background writeback cannot * catch-up. This avoids (excessively) small writeouts @@ -1531,18 +1595,31 @@ static void balance_dirty_pages(struct address_space *mapping, * In strictlimit case make decision based on the wb counters * and limits. Small writeouts when the wb limits are ramping * up are the price we consciously pay for strictlimit-ing. + * + * If memcg domain is in effect, @dirty should be under + * both global and memcg freerun ceilings. */ - if (dirty <= dirty_freerun_ceiling(thresh, bg_thresh)) { + if (dirty <= dirty_freerun_ceiling(thresh, bg_thresh) && + (!mdtc || + m_dirty <= dirty_freerun_ceiling(m_thresh, m_bg_thresh))) { + unsigned long intv = dirty_poll_interval(dirty, thresh); + unsigned long m_intv = ULONG_MAX; + current->dirty_paused_when = now; current->nr_dirtied = 0; - current->nr_dirtied_pause = - dirty_poll_interval(dirty, thresh); + if (mdtc) + m_intv = dirty_poll_interval(m_dirty, m_thresh); + current->nr_dirtied_pause = min(intv, m_intv); break; } if (unlikely(!writeback_in_progress(wb))) wb_start_background_writeback(wb); + /* + * Calculate global domain's pos_ratio and select the + * global dtc by default. + */ if (!strictlimit) wb_dirty_limits(gdtc); @@ -1550,6 +1627,25 @@ static void balance_dirty_pages(struct address_space *mapping, ((gdtc->dirty > gdtc->thresh) || strictlimit); wb_position_ratio(gdtc); + sdtc = gdtc; + + if (mdtc) { + /* + * If memcg domain is in effect, calculate its + * pos_ratio. @wb should satisfy constraints from + * both global and memcg domains. Choose the one + * w/ lower pos_ratio. + */ + if (!strictlimit) + wb_dirty_limits(mdtc); + + dirty_exceeded |= (mdtc->wb_dirty > mdtc->wb_thresh) && + ((mdtc->dirty > mdtc->thresh) || strictlimit); + + wb_position_ratio(mdtc); + if (mdtc->pos_ratio < gdtc->pos_ratio) + sdtc = mdtc; + } if (dirty_exceeded && !wb->dirty_exceeded) wb->dirty_exceeded = 1; @@ -1557,14 +1653,15 @@ static void balance_dirty_pages(struct address_space *mapping, if (time_is_before_jiffies(wb->bw_time_stamp + BANDWIDTH_INTERVAL)) { spin_lock(&wb->list_lock); - __wb_update_bandwidth(gdtc, start_time, true); + __wb_update_bandwidth(gdtc, mdtc, start_time, true); spin_unlock(&wb->list_lock); } + /* throttle according to the chosen dtc */ dirty_ratelimit = wb->dirty_ratelimit; - task_ratelimit = ((u64)dirty_ratelimit * gdtc->pos_ratio) >> + task_ratelimit = ((u64)dirty_ratelimit * sdtc->pos_ratio) >> RATELIMIT_CALC_SHIFT; - max_pause = wb_max_pause(wb, gdtc->wb_dirty); + max_pause = wb_max_pause(wb, sdtc->wb_dirty); min_pause = wb_min_pause(wb, max_pause, task_ratelimit, dirty_ratelimit, &nr_dirtied_pause); @@ -1587,11 +1684,11 @@ static void balance_dirty_pages(struct address_space *mapping, */ if (pause < min_pause) { trace_balance_dirty_pages(bdi, - gdtc->thresh, - gdtc->bg_thresh, - gdtc->dirty, - gdtc->wb_thresh, - gdtc->wb_dirty, + sdtc->thresh, + sdtc->bg_thresh, + sdtc->dirty, + sdtc->wb_thresh, + sdtc->wb_dirty, dirty_ratelimit, task_ratelimit, pages_dirtied, @@ -1616,11 +1713,11 @@ static void balance_dirty_pages(struct address_space *mapping, pause: trace_balance_dirty_pages(bdi, - gdtc->thresh, - gdtc->bg_thresh, - gdtc->dirty, - gdtc->wb_thresh, - gdtc->wb_dirty, + sdtc->thresh, + sdtc->bg_thresh, + sdtc->dirty, + sdtc->wb_thresh, + sdtc->wb_dirty, dirty_ratelimit, task_ratelimit, pages_dirtied, @@ -1651,7 +1748,7 @@ pause: * more page. However wb_dirty has accounting errors. So use * the larger and more IO friendly wb_stat_error. */ - if (gdtc->wb_dirty <= wb_stat_error(wb)) + if (sdtc->wb_dirty <= wb_stat_error(wb)) break; if (fatal_signal_pending(current)) @@ -1775,7 +1872,10 @@ EXPORT_SYMBOL(balance_dirty_pages_ratelimited); bool wb_over_bg_thresh(struct bdi_writeback *wb) { struct dirty_throttle_control gdtc_stor = { GDTC_INIT(wb) }; + struct dirty_throttle_control mdtc_stor = { MDTC_INIT(wb, &gdtc_stor) }; struct dirty_throttle_control * const gdtc = &gdtc_stor; + struct dirty_throttle_control * const mdtc = mdtc_valid(&mdtc_stor) ? + &mdtc_stor : NULL; /* * Similar to balance_dirty_pages() but ignores pages being written @@ -1792,6 +1892,20 @@ bool wb_over_bg_thresh(struct bdi_writeback *wb) if (wb_stat(wb, WB_RECLAIMABLE) > __wb_calc_thresh(gdtc)) return true; + if (mdtc) { + unsigned long writeback; + + mem_cgroup_wb_stats(wb, &mdtc->avail, &mdtc->dirty, &writeback); + mdtc_cap_avail(mdtc); + domain_dirty_limits(mdtc); /* ditto, ignore writeback */ + + if (mdtc->dirty > mdtc->bg_thresh) + return true; + + if (wb_stat(wb, WB_RECLAIMABLE) > __wb_calc_thresh(mdtc)) + return true; + } + return false; } -- cgit v1.2.3-70-g09d2