From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chao Yu Subject: Re: [PATCH] Introduce lifetime write IO statistics Date: Thu, 28 Jan 2016 09:20:39 +0800 Message-ID: <00e801d1596a$314a8db0$93dfa910$@samsung.com> References: <1453794038-189520-1-git-send-email-liushuoran@huawei.com> <00a101d1581b$6a7ad090$3f7071b0$@samsung.com> <56A7468A.5050600@huawei.com> <00c801d158d3$4507ae10$cf170a30$@samsung.com> <56A87462.7010502@huawei.com> <00da01d158e9$f170a290$d451e7b0$@samsung.com> <20160127222652.GB62463@jaegeuk.gateway> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from sog-mx-4.v43.ch3.sourceforge.com ([172.29.43.194] helo=mx.sourceforge.net) by sfs-ml-1.v29.ch3.sourceforge.com with esmtp (Exim 4.76) (envelope-from ) id 1aObHA-0007ps-Lx for linux-f2fs-devel@lists.sourceforge.net; Thu, 28 Jan 2016 01:21:32 +0000 Received: from mailout1.samsung.com ([203.254.224.24]) by sog-mx-4.v43.ch3.sourceforge.com with esmtps (TLSv1:AES128-SHA:128) (Exim 4.76) id 1aObH7-0008Ja-S9 for linux-f2fs-devel@lists.sourceforge.net; Thu, 28 Jan 2016 01:21:32 +0000 Received: from epcpsbgm2new.samsung.com (epcpsbgm2 [203.254.230.27]) by mailout1.samsung.com (Oracle Communications Messaging Server 7.0.5.31.0 64bit (built May 5 2014)) with ESMTP id <0O1N022DJ2FL2240@mailout1.samsung.com> for linux-f2fs-devel@lists.sourceforge.net; Thu, 28 Jan 2016 10:21:22 +0900 (KST) In-reply-to: <20160127222652.GB62463@jaegeuk.gateway> Content-language: zh-cn List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: linux-f2fs-devel-bounces@lists.sourceforge.net To: 'Jaegeuk Kim' Cc: linux-f2fs-devel@lists.sourceforge.net > -----Original Message----- > From: Jaegeuk Kim [mailto:jaegeuk@kernel.org] > Sent: Thursday, January 28, 2016 6:27 AM > To: Chao Yu > Cc: 'Shuoran'; linux-f2fs-devel@lists.sourceforge.net > Subject: Re: [f2fs-dev] [PATCH] Introduce lifetime write IO statistics > > Hi Chao, > > On Wed, Jan 27, 2016 at 06:02:37PM +0800, Chao Yu wrote: > > > -----Original Message----- > > > From: Shuoran [mailto:liushuoran@huawei.com] > > > Sent: Wednesday, January 27, 2016 3:40 PM > > > To: Chao Yu > > > Cc: jaegeuk@kernel.org; linux-f2fs-devel@lists.sourceforge.net > > > Subject: Re: [f2fs-dev] [PATCH] Introduce lifetime write IO statistics > > > > > > On 2016/1/27 15:20, Chao Yu wrote: > > > >> -----Original Message----- > > > >> From: Shuoran [mailto:liushuoran@huawei.com] > > > >> Sent: Tuesday, January 26, 2016 6:12 PM > > > >> To: Chao Yu > > > >> Cc: jaegeuk@kernel.org; linux-f2fs-devel@lists.sourceforge.net > > > >> Subject: Re: [f2fs-dev] [PATCH] Introduce lifetime write IO statistics > > > >> > > > >> On 2016/1/26 17:24, Chao Yu wrote: > > > >>> Hi, > > > >>> > > > >>>> -----Original Message----- > > > >>>> From: Liu Shuoran [mailto:liushuoran@huawei.com] > > > >>>> Sent: Tuesday, January 26, 2016 3:41 PM > > > >>>> To: jaegeuk@kernel.org > > > >>>> Cc: linux-f2fs-devel@lists.sourceforge.net > > > >>>> Subject: [f2fs-dev] [PATCH] Introduce lifetime write IO statistics > > > >>>> > > > >>>> This patch introduces lifetime IO write statistics exposed to the sysfs interface. > > > >>>> The write IO amount is obtained from block layer, accumulated in the file system and > > > >>>> stored in the hot node summary of checkpoint. > > > >>>> > > > >>>> Signed-off-by: Shuoran Liu > > > >>>> Signed-off-by: Pengyang Hou > > > >>>> Signed-off-by: Jaegeuk Kim > > > >>>> --- > > > >>>> fs/f2fs/checkpoint.c | 12 ++++++++++++ > > > >>>> fs/f2fs/f2fs.h | 11 +++++++++++ > > > >>>> fs/f2fs/super.c | 35 +++++++++++++++++++++++++++++++++++ > > > >>>> include/linux/f2fs_fs.h | 14 +++++++++++++- > > > >>>> 4 files changed, 71 insertions(+), 1 deletion(-) > > > >>>> > > > >>>> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > > > >>>> index 3842af9..998de47 100644 > > > >>>> --- a/fs/f2fs/checkpoint.c > > > >>>> +++ b/fs/f2fs/checkpoint.c > > > >>>> @@ -921,6 +921,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct > cp_control > > > >> *cpc) > > > >>>> int cp_payload_blks = __cp_payload(sbi); > > > >>>> block_t discard_blk = NEXT_FREE_BLKADDR(sbi, curseg); > > > >>>> bool invalidate = false; > > > >>>> + struct super_block *sb = sbi->sb; > > > >>>> + struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); > > > >>>> + u64 kbytes_written; > > > >>>> + > > > >>>> > > > >>>> /* > > > >>>> * This avoids to conduct wrong roll-forward operations and uses > > > >>>> @@ -1034,6 +1038,14 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct > cp_control > > > >>>> *cpc) > > > >>>> > > > >>>> write_data_summaries(sbi, start_blk); > > > >>>> start_blk += data_sum_blocks; > > > >>>> + > > > >>>> + /* Record write statistics in the hot node summary */ > > > >>>> + kbytes_written = sbi->kbytes_written; > > > >>>> + if (sb->s_bdev->bd_part) > > > >>>> + kbytes_written += BD_PART_WRITTEN(sbi); > > > >>>> + > > > >>>> + seg_i->sum_blk->info.kbytes_written = cpu_to_le64(kbytes_written); > > > >>>> + > > > >>>> if (__remain_node_summaries(cpc->reason)) { > > > >>>> write_node_summaries(sbi, start_blk); > > > >>>> start_blk += NR_CURSEG_NODE_TYPE; > > > >>>> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h > > > >>>> index ff79054..cee5fab0 100644 > > > >>>> --- a/fs/f2fs/f2fs.h > > > >>>> +++ b/fs/f2fs/f2fs.h > > > >>>> @@ -844,8 +844,19 @@ struct f2fs_sb_info { > > > >>>> struct list_head s_list; > > > >>>> struct mutex umount_mutex; > > > >>>> unsigned int shrinker_run_no; > > > >>>> + > > > >>>> + /* For write statistics */ > > > >>>> + u64 sectors_written_start; > > > >>>> + u64 kbytes_written; > > > >>>> }; > > > >>>> > > > >>>> +/* For write statistics. Suppose sector size is 512 bytes, > > > >>>> + * and the return value is in kbytes. s is of struct f2fs_sb_info. > > > >>>> + */ > > > >>>> +#define BD_PART_WRITTEN(s) \ > > > >>>> +(((u64)part_stat_read(s->sb->s_bdev->bd_part, sectors[1]) - \ > > > >>>> + s->sectors_written_start) >> 1) > > > >>>> + > > > >>>> static inline void f2fs_update_time(struct f2fs_sb_info *sbi, int type) > > > >>>> { > > > >>>> sbi->last_time[type] = jiffies; > > > >>>> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c > > > >>>> index 3bf990b..7ed10f1 100644 > > > >>>> --- a/fs/f2fs/super.c > > > >>>> +++ b/fs/f2fs/super.c > > > >>>> @@ -126,6 +126,19 @@ static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int > > > >>>> struct_type) > > > >>>> return NULL; > > > >>>> } > > > >>>> > > > >>>> +static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a, > > > >>>> + struct f2fs_sb_info *sbi, char *buf) > > > >>>> +{ > > > >>>> + struct super_block *sb = sbi->sb; > > > >>>> + > > > >>>> + if (!sb->s_bdev->bd_part) > > > >>>> + return snprintf(buf, PAGE_SIZE, "0\n"); > > > >>>> + > > > >>>> + return snprintf(buf, PAGE_SIZE, "%llu\n", > > > >>>> + (unsigned long long)(sbi->kbytes_written + > > > >>>> + BD_PART_WRITTEN(sbi))); > > > >>>> +} > > > >>>> + > > > >>>> static ssize_t f2fs_sbi_show(struct f2fs_attr *a, > > > >>>> struct f2fs_sb_info *sbi, char *buf) > > > >>>> { > > > >>>> @@ -204,6 +217,9 @@ static struct f2fs_attr f2fs_attr_##_name = { \ > > > >>>> f2fs_sbi_show, f2fs_sbi_store, \ > > > >>>> offsetof(struct struct_name, elname)) > > > >>>> > > > >>>> +#define F2FS_GENERAL_RO_ATTR(name) \ > > > >>>> +static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL) > > > >>>> + > > > >>>> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time); > > > >>>> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time); > > > >>>> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time); > > > >>>> @@ -220,6 +236,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, > > > >>>> max_victim_search); > > > >>>> F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); > > > >>>> F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]); > > > >>>> F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]); > > > >>>> +F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes); > > > >>>> > > > >>>> #define ATTR_LIST(name) (&f2fs_attr_##name.attr) > > > >>>> static struct attribute *f2fs_attrs[] = { > > > >>>> @@ -239,6 +256,7 @@ static struct attribute *f2fs_attrs[] = { > > > >>>> ATTR_LIST(ra_nid_pages), > > > >>>> ATTR_LIST(cp_interval), > > > >>>> ATTR_LIST(idle_interval), > > > >>>> + ATTR_LIST(lifetime_write_kbytes), > > > >>>> NULL, > > > >>>> }; > > > >>>> > > > >>>> @@ -766,6 +784,11 @@ static int f2fs_remount(struct super_block *sb, int *flags, char > *data) > > > >>>> bool need_stop_gc = false; > > > >>>> bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); > > > >>>> > > > >>>> + if (*flags & MS_RDONLY) { > > > >>>> + set_opt(sbi, FASTBOOT); > > > >>> Need to recover to original mode, otherwise it may change option user > > > >>> set. > > > >> Yes, maybe I can bring forward the action of saving old mount options. > > > >>>> + set_sbi_flag(sbi, SBI_IS_DIRTY); > > > >>>> + } > > > >>>> + > > > >>>> sync_filesystem(sb); > > > >>>> > > > >>>> /* > > > >>>> @@ -1242,6 +1265,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, > int > > > >> silent) > > > >>>> bool retry = true, need_fsck = false; > > > >>>> char *options = NULL; > > > >>>> int recovery, i, valid_super_block; > > > >>>> + struct curseg_info *seg_i; > > > >>>> > > > >>>> try_onemore: > > > >>>> err = -EINVAL; > > > >>>> @@ -1372,6 +1396,17 @@ try_onemore: > > > >>>> goto free_nm; > > > >>>> } > > > >>>> > > > >>>> + /* For write statistics */ > > > >>>> + if (sb->s_bdev->bd_part) > > > >>>> + sbi->sectors_written_start = > > > >>>> + (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]); > > > >>>> + > > > >>>> + /* Read accumulated write IO statistics if exists */ > > > >>>> + seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); > > > >>> Why not CURSEG_WARM_DATA? > > > >> Data summary might be compacted. So only hot/warm/cold node summary is > > > >> available, and we just picked hot node summary. > > > > Oh, you're right, warm data journal would not be persisted if data > > > > summaries are compacted, my mistake. > > > > > > > > Wouldn't be better to persist this stat number into super block like > > > > ext4? > > > > > > > > Thanks, > > > > > > I think super block in F2FS is not updated as frequent as that in Ext4 > > > (there are several places invoking ext4_commit_super, while few invoking > > > f2fs_commit_super), although in this design the statistics is not > > > updated very frequently either. Another concern is that, to some extent, > > > the super blocks of f2fs is not supposed to update except for a few > > > cases. So the stat is better stored in the checkpoint, and we just tried > > > very hard finding a spare place. > > > > Got it, so the key point is preventing generating additional IO. > > > > If update points of stat in your approach is enough, so for > > 1) umount case: commit sb approach causes two more blocks updating. > > 2) remount case: commit sb approach causes one less blocks updating when > > FASTBOOT is disabled. > > > > Not a big degression. :) > > > > IMO, maybe it's better to use journal log in data segments to store > > checkpoint related data, maybe more nats/sits entries. > > > > And another thing is I'm in doubt that whether it is enough to use cp + > > data journal to update stat, since abnormal pow-cut after a normal cp would > > make us losing all stats saved before. > > Pros and Cons. > IMO, we don't need to guarantee up-to-date information to users. If users wants > to keep that in high accuracy, they simply can set fastboot. Yes, at least it needs to document this sysfs entry to let user know about this. Thanks, > > Thanks, > > > > > Thanks, > > > > > > > > >>> Thanks, > > > >>> > > > >>>> + if(__exist_node_summaries(sbi)) > > > >>>> + sbi->kbytes_written = > > > >>>> + le64_to_cpu(seg_i->sum_blk->info.kbytes_written); > > > >>>> + > > > >>>> build_gc_manager(sbi); > > > >>>> > > > >>>> /* get an inode for node space */ > > > >>>> diff --git a/include/linux/f2fs_fs.h b/include/linux/f2fs_fs.h > > > >>>> index e59c3be..67aa01d 100644 > > > >>>> --- a/include/linux/f2fs_fs.h > > > >>>> +++ b/include/linux/f2fs_fs.h > > > >>>> @@ -358,6 +358,12 @@ struct summary_footer { > > > >>>> sizeof(struct sit_journal_entry)) > > > >>>> #define SIT_JOURNAL_RESERVED ((SUM_JOURNAL_SIZE - 2) %\ > > > >>>> sizeof(struct sit_journal_entry)) > > > >>>> + > > > >>>> +/* Reserved area should make size of f2fs_extra_info equals to > > > >>>> + * that of nat_journal and sit_journal. > > > >>>> + */ > > > >>>> +#define EXTRA_INFO_RESERVED (SUM_JOURNAL_SIZE - 2 - 8) > > > >>>> + > > > >>>> /* > > > >>>> * frequently updated NAT/SIT entries can be stored in the spare area in > > > >>>> * summary blocks > > > >>>> @@ -387,6 +393,11 @@ struct sit_journal { > > > >>>> __u8 reserved[SIT_JOURNAL_RESERVED]; > > > >>>> } __packed; > > > >>>> > > > >>>> +struct f2fs_extra_info { > > > >>>> + __le64 kbytes_written; > > > >>>> + __u8 reserved[EXTRA_INFO_RESERVED]; > > > >>>> +} __packed; > > > >>>> + > > > >>>> /* 4KB-sized summary block structure */ > > > >>>> struct f2fs_summary_block { > > > >>>> struct f2fs_summary entries[ENTRIES_IN_SUM]; > > > >>>> @@ -394,10 +405,11 @@ struct f2fs_summary_block { > > > >>>> __le16 n_nats; > > > >>>> __le16 n_sits; > > > >>>> }; > > > >>>> - /* spare area is used by NAT or SIT journals */ > > > >>>> + /* spare area is used by NAT or SIT journals or extra info */ > > > >>>> union { > > > >>>> struct nat_journal nat_j; > > > >>>> struct sit_journal sit_j; > > > >>>> + struct f2fs_extra_info info; > > > >>>> }; > > > >>>> struct summary_footer footer; > > > >>>> } __packed; > > > >>>> -- > > > >>>> 1.9.1 > > > >>>> > > > >>>> > > > >>>> ------------------------------------------------------------------------------ > > > >>>> Site24x7 APM Insight: Get Deep Visibility into Application Performance > > > >>>> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month > > > >>>> Monitor end-to-end web transactions and take corrective actions now > > > >>>> Troubleshoot faster and improve end-user experience. Signup Now! > > > >>>> http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140 > > > >>>> _______________________________________________ > > > >>>> Linux-f2fs-devel mailing list > > > >>>> Linux-f2fs-devel@lists.sourceforge.net > > > >>>> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel > > > >>> . > > > >>> > > > > > > > > > > > > . > > > > > > > > > ------------------------------------------------------------------------------ Site24x7 APM Insight: Get Deep Visibility into Application Performance APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month Monitor end-to-end web transactions and take corrective actions now Troubleshoot faster and improve end-user experience. Signup Now! http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140