On Mon, 8 Mar 2010 11:37:11 +0900, KAMEZAWA Hiroyuki <[email protected]> wrote: > On Mon, 8 Mar 2010 11:17:24 +0900 > Daisuke Nishimura <[email protected]> wrote: > > > > But IIRC, clear_writeback is done under treelock.... No ? > > > > > The place where NR_WRITEBACK is updated is out of tree_lock. > > > > 1311 int test_clear_page_writeback(struct page *page) > > 1312 { > > 1313 struct address_space *mapping = page_mapping(page); > > 1314 int ret; > > 1315 > > 1316 if (mapping) { > > 1317 struct backing_dev_info *bdi = > > mapping->backing_dev_info; > > 1318 unsigned long flags; > > 1319 > > 1320 spin_lock_irqsave(&mapping->tree_lock, flags); > > 1321 ret = TestClearPageWriteback(page); > > 1322 if (ret) { > > 1323 radix_tree_tag_clear(&mapping->page_tree, > > 1324 page_index(page), > > 1325 > > PAGECACHE_TAG_WRITEBACK); > > 1326 if (bdi_cap_account_writeback(bdi)) { > > 1327 __dec_bdi_stat(bdi, BDI_WRITEBACK); > > 1328 __bdi_writeout_inc(bdi); > > 1329 } > > 1330 } > > 1331 spin_unlock_irqrestore(&mapping->tree_lock, flags); > > 1332 } else { > > 1333 ret = TestClearPageWriteback(page); > > 1334 } > > 1335 if (ret) > > 1336 dec_zone_page_state(page, NR_WRITEBACK); > > 1337 return ret; > > 1338 } > > We can move this up to under tree_lock. Considering memcg, all our target has > "mapping". > > If we newly account bounce-buffers (for NILFS, FUSE, etc..), which has no > ->mapping, > we need much more complex new charge/uncharge theory. > > But yes, adding new lock scheme seems complicated. (Sorry Andrea.) > My concerns is performance. We may need somehing new re-implementation of > locks/migrate/charge/uncharge. > I agree. Performance is my concern too.
I made a patch below and measured the time(average of 10 times) of kernel build on tmpfs(make -j8 on 8 CPU machine with 2.6.33 defconfig). <before> - root cgroup: 190.47 sec - child cgroup: 192.81 sec <after> - root cgroup: 191.06 sec - child cgroup: 193.06 sec Hmm... about 0.3% slower for root, 0.1% slower for child. === From: Daisuke Nishimura <[email protected]> In current implementation, we don't have to disable irq at lock_page_cgroup() because the lock is never acquired in interrupt context. But we are going to do it in later patch, so this patch encloses all of lock_page_cgroup()/unlock_page_cgroup() with irq_disabled()/irq_enabled(). Signed-off-by: Daisuke Nishimura <[email protected]> --- mm/memcontrol.c | 17 +++++++++++++++++ 1 files changed, 17 insertions(+), 0 deletions(-) diff --git a/mm/memcontrol.c b/mm/memcontrol.c index 02ea959..e5ae1a1 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -1359,6 +1359,7 @@ void mem_cgroup_update_file_mapped(struct page *page, int val) if (unlikely(!pc)) return; + local_irq_disable(); lock_page_cgroup(pc); mem = pc->mem_cgroup; if (!mem) @@ -1374,6 +1375,7 @@ void mem_cgroup_update_file_mapped(struct page *page, int val) done: unlock_page_cgroup(pc); + local_irq_enable(); } /* @@ -1711,6 +1713,7 @@ struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page) VM_BUG_ON(!PageLocked(page)); pc = lookup_page_cgroup(page); + local_irq_disable(); lock_page_cgroup(pc); if (PageCgroupUsed(pc)) { mem = pc->mem_cgroup; @@ -1726,6 +1729,7 @@ struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page) rcu_read_unlock(); } unlock_page_cgroup(pc); + local_irq_enable(); return mem; } @@ -1742,9 +1746,11 @@ static void __mem_cgroup_commit_charge(struct mem_cgroup *mem, if (!mem) return; + local_irq_disable(); lock_page_cgroup(pc); if (unlikely(PageCgroupUsed(pc))) { unlock_page_cgroup(pc); + local_irq_enable(); mem_cgroup_cancel_charge(mem); return; } @@ -1775,6 +1781,7 @@ static void __mem_cgroup_commit_charge(struct mem_cgroup *mem, mem_cgroup_charge_statistics(mem, pc, true); unlock_page_cgroup(pc); + local_irq_enable(); /* * "charge_statistics" updated event counter. Then, check it. * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree. @@ -1844,12 +1851,14 @@ static int mem_cgroup_move_account(struct page_cgroup *pc, struct mem_cgroup *from, struct mem_cgroup *to, bool uncharge) { int ret = -EINVAL; + local_irq_disable(); lock_page_cgroup(pc); if (PageCgroupUsed(pc) && pc->mem_cgroup == from) { __mem_cgroup_move_account(pc, from, to, uncharge); ret = 0; } unlock_page_cgroup(pc); + local_irq_enable(); /* * check events */ @@ -1981,12 +1990,15 @@ int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm, pc = lookup_page_cgroup(page); if (!pc) return 0; + local_irq_disable(); lock_page_cgroup(pc); if (PageCgroupUsed(pc)) { unlock_page_cgroup(pc); + local_irq_enable(); return 0; } unlock_page_cgroup(pc); + local_irq_enable(); } if (unlikely(!mm && !mem)) @@ -2182,6 +2194,7 @@ __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype) if (unlikely(!pc || !PageCgroupUsed(pc))) return NULL; + local_irq_disable(); lock_page_cgroup(pc); mem = pc->mem_cgroup; @@ -2222,6 +2235,7 @@ __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype) mz = page_cgroup_zoneinfo(pc); unlock_page_cgroup(pc); + local_irq_enable(); memcg_check_events(mem, page); /* at swapout, this memcg will be accessed to record to swap */ @@ -2232,6 +2246,7 @@ __mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype) unlock_out: unlock_page_cgroup(pc); + local_irq_enable(); return NULL; } @@ -2424,12 +2439,14 @@ int mem_cgroup_prepare_migration(struct page *page, struct mem_cgroup **ptr) return 0; pc = lookup_page_cgroup(page); + local_irq_disable(); lock_page_cgroup(pc); if (PageCgroupUsed(pc)) { mem = pc->mem_cgroup; css_get(&mem->css); } unlock_page_cgroup(pc); + local_irq_enable(); if (mem) { ret = __mem_cgroup_try_charge(NULL, GFP_KERNEL, &mem, false); -- 1.6.4 _______________________________________________ Containers mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/containers _______________________________________________ Devel mailing list [email protected] https://openvz.org/mailman/listinfo/devel
