Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S941539AbcKQRRx (ORCPT ); Thu, 17 Nov 2016 12:17:53 -0500 Received: from szxga02-in.huawei.com ([119.145.14.65]:65242 "EHLO szxga02-in.huawei.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S941523AbcKQRRt (ORCPT ); Thu, 17 Nov 2016 12:17:49 -0500 Subject: Re: [f2fs-dev] [PATCH] f2fs: fix to account total free nid correctly To: heyunlei , Jaegeuk Kim References: <20161114112456.113074-1-yuchao0@huawei.com> <20161114204557.GA28366@jaegeuk> <1baaeb63-4557-a209-3f6b-9090f44345e1@huawei.com> CC: , , From: Chao Yu Message-ID: Date: Thu, 17 Nov 2016 20:09:43 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.0 MIME-Version: 1.0 In-Reply-To: <1baaeb63-4557-a209-3f6b-9090f44345e1@huawei.com> Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: 7bit X-Originating-IP: [10.134.22.195] X-CFilter-Loop: Reflected Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5292 Lines: 163 Hi Yunlei, On 2016/11/17 17:42, heyunlei wrote: > > > On 2016/11/15 4:45, Jaegeuk Kim wrote: >> On Mon, Nov 14, 2016 at 07:24:56PM +0800, Chao Yu wrote: >>> Thread A Thread B Thread C >>> - f2fs_create >>> - f2fs_new_inode >>> - f2fs_lock_op >>> - alloc_nid >>> alloc last nid >>> - f2fs_unlock_op >>> - f2fs_create >>> - f2fs_new_inode >>> - f2fs_lock_op >>> - alloc_nid >>> as node count still not >>> be increased, we will >>> loop in alloc_nid >>> - f2fs_write_node_pages >>> - f2fs_balance_fs_bg >>> - f2fs_sync_fs >>> - write_checkpoint >>> - block_operations >>> - f2fs_lock_all >>> - f2fs_lock_op >>> >>> While creating new inode, we do not allocate and account nid atomically, >>> so that when there is almost no free nids left, we may encounter deadloop >>> like above stack. >>> >>> In order to avoid that, add nm_i::free_nid_cnt for accounting free nids >>> and do nid allocation atomically during node creation. >> >> How about using nm_i::avaiable_nids for this? >> It seems that we don't need both of variables at the same time. >> >> Thanks, >> >>> >>> Signed-off-by: Chao Yu >>> --- >>> fs/f2fs/f2fs.h | 1 + >>> fs/f2fs/node.c | 19 +++++++++++++++---- >>> 2 files changed, 16 insertions(+), 4 deletions(-) >>> >>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h >>> index 6de1fbf..9de6f20 100644 >>> --- a/fs/f2fs/f2fs.h >>> +++ b/fs/f2fs/f2fs.h >>> @@ -551,6 +551,7 @@ struct f2fs_nm_info { >>> struct radix_tree_root free_nid_root;/* root of the free_nid cache */ >>> struct list_head nid_list[MAX_NID_LIST];/* lists for free nids */ >>> unsigned int nid_cnt[MAX_NID_LIST]; /* the number of free node id */ >>> + unsigned int free_nid_cnt; /* the number of total free nid */ >>> spinlock_t nid_list_lock; /* protect nid lists ops */ >>> struct mutex build_lock; /* lock for build free nids */ >>> >>> diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c >>> index d58438f..e412d0e 100644 >>> --- a/fs/f2fs/node.c >>> +++ b/fs/f2fs/node.c >>> @@ -1885,11 +1885,13 @@ bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) >>> return false; >>> } >>> #endif >>> - if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) >>> - return false; >>> - >>> spin_lock(&nm_i->nid_list_lock); >>> >>> + if (unlikely(nm_i->free_nid_cnt == 0)) { >>> + spin_unlock(&nm_i->nid_list_lock); >>> + return false; >>> + } >>> + >>> /* We should not use stale free nids created by build_free_nids */ >>> if (nm_i->nid_cnt[FREE_NID_LIST] && !on_build_free_nids(nm_i)) { >>> f2fs_bug_on(sbi, list_empty(&nm_i->nid_list[FREE_NID_LIST])); >>> @@ -1900,6 +1902,7 @@ bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) >>> __remove_nid_from_list(sbi, i, FREE_NID_LIST, true); >>> i->state = NID_ALLOC; >>> __insert_nid_to_list(sbi, i, ALLOC_NID_LIST, false); >>> + nm_i->free_nid_cnt--; >>> spin_unlock(&nm_i->nid_list_lock); >>> return true; >>> } >>> @@ -1951,6 +1954,9 @@ void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) >>> i->state = NID_NEW; >>> __insert_nid_to_list(sbi, i, FREE_NID_LIST, false); >>> } >>> + >>> + nm_i->free_nid_cnt++; >>> + >>> spin_unlock(&nm_i->nid_list_lock); >>> >>> if (need_free) >>> @@ -2222,8 +2228,12 @@ static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, >>> raw_nat_from_node_info(raw_ne, &ne->ni); >>> nat_reset_flag(ne); >>> __clear_nat_cache_dirty(NM_I(sbi), ne); >>> - if (nat_get_blkaddr(ne) == NULL_ADDR) >>> + if (nat_get_blkaddr(ne) == NULL_ADDR) { >>> add_free_nid(sbi, nid, false); >>> + spin_lock(&NM_I(sbi)->nid_list_lock); >>> + NM_I(sbi)->free_nid_cnt++; > Hi all, > Here, we should consider clean NULL_ADDR nat entry in journal. > > diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c > index dcfab29..b22ecb0 100644 > --- a/fs/f2fs/node.c > +++ b/fs/f2fs/node.c > @@ -158,6 +158,13 @@ static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, > if (get_nat_flag(ne, IS_DIRTY)) > return; > > + if (ne->ni.blk_addr == NULL_ADDR) { > + spin_lock(&nm_i->free_nid_list_lock); > + nm_i->available_nids--; > + spin_unlock(&nm_i->free_nid_list_lock); > + } > + > + Thanks for pointing this out, as we discussed, it should be moved to remove_nats_in_journal, anyway, I will send v2. Thanks, > > Thanks. > >>> + spin_unlock(&NM_I(sbi)->nid_list_lock); >>> + } >>> } >>> >>> if (to_journal) >>> @@ -2302,6 +2312,7 @@ static int init_node_manager(struct f2fs_sb_info *sbi) >>> nm_i->nid_cnt[FREE_NID_LIST] = 0; >>> nm_i->nid_cnt[ALLOC_NID_LIST] = 0; >>> nm_i->nat_cnt = 0; >>> + nm_i->free_nid_cnt = nm_i->available_nids - sbi->total_valid_node_count; >>> nm_i->ram_thresh = DEF_RAM_THRESHOLD; >>> nm_i->ra_nid_pages = DEF_RA_NID_PAGES; >>> nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; >>> -- >>> 2.8.2.311.gee88674 >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> Linux-f2fs-devel mailing list >> Linux-f2fs-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel >> >> . >> > > > . >