linux-f2fs-devel.lists.sourceforge.net archive mirror
 help / color / mirror / Atom feed
From: Lihong Kou <koulihong@huawei.com>
To: <yuchao0@huawei.com>, <jaegeuk@kernel.org>
Cc: zengguangyue@hisilicon.com, linux-f2fs-devel@lists.sourceforge.net
Subject: [f2fs-dev] [PATCH v2 2/2] f2fs: introduce a new fio cache for IPU.
Date: Wed, 16 Oct 2019 18:11:29 +0800	[thread overview]
Message-ID: <1571220689-95274-3-git-send-email-koulihong@huawei.com> (raw)
In-Reply-To: <1571220689-95274-1-git-send-email-koulihong@huawei.com>

From: koulihong <koulihong@huawei.com>

Description: introduce a new fio cache for IPU sequential
writes. After commit 8648de2c581(f2fs: add bio cache for IPU)
in the mainline, we still have the problem in SQLite. After
discuss the issue with chao, we know there are still some bugs
in temporary bio cache for IPU. This patch introdce a new fio cache
to fix this bug.

Test case:
fio -filename=/data/1.txt -direct=0 -ioengine=psync -fdatasync=1 -rw=write -bs=256K -size=256M -numjobs=1 -group_reporting -name=test1

Before:
             fio-2635  [015] ....  2017.735514: block_rq_issue: 7,0 WS 4096 () 548888 + 8 [fio]
             fio-2635  [015] ....  2017.735515: block_rq_issue: 7,0 WS 4096 () 548896 + 8 [fio]
             fio-2635  [015] ....  2017.735516: block_rq_issue: 7,0 WS 4096 () 548904 + 8 [fio]
             fio-2635  [015] ....  2017.735516: block_rq_issue: 7,0 WS 4096 () 548912 + 8 [fio]
             fio-2635  [015] ....  2017.735517: block_rq_issue: 7,0 WS 4096 () 548920 + 8 [fio]
             fio-2635  [015] ....  2017.735517: block_rq_issue: 7,0 WS 4096 () 548928 + 8 [fio]
             fio-2635  [015] ....  2017.735518: block_rq_issue: 7,0 WS 4096 () 548936 + 8 [fio]
             fio-2635  [015] ....  2017.735518: block_rq_issue: 7,0 WS 4096 () 548944 + 8 [fio]
             fio-2635  [015] ....  2017.735519: block_rq_issue: 7,0 WS 4096 () 548952 + 8 [fio]
             fio-2635  [015] ....  2017.735519: block_rq_issue: 7,0 WS 4096 () 548960 + 8 [fio]
             fio-2635  [015] ....  2017.735520: block_rq_issue: 7,0 WS 4096 () 548968 + 8 [fio]
             fio-2635  [015] ....  2017.735520: block_rq_issue: 7,0 WS 4096 () 548976 + 8 [fio]
             fio-2635  [015] ....  2017.735521: block_rq_issue: 7,0 WS 4096 () 548984 + 8 [fio]
After:
             fio-14692 [003] ....  1473.946721: block_rq_issue: 7,0 WS 262144 () 195072 + 512 [fio]
             fio-14692 [003] ....  1473.947722: block_rq_issue: 7,0 WS 262144 () 195584 + 512 [fio]
             fio-14692 [003] ....  1473.948716: block_rq_issue: 7,0 WS 262144 () 196096 + 512 [fio]
             fio-14692 [003] ....  1473.949711: block_rq_issue: 7,0 WS 262144 () 196608 + 512 [fio]
             fio-14692 [003] ....  1473.950704: block_rq_issue: 7,0 WS 262144 () 197120 + 512 [fio]
             fio-14692 [003] ....  1473.951700: block_rq_issue: 7,0 WS 262144 () 197632 + 512 [fio]
             fio-14692 [003] ....  1473.952687: block_rq_issue: 7,0 WS 262144 () 198144 + 512 [fio]
             fio-14692 [003] ....  1473.953684: block_rq_issue: 7,0 WS 262144 () 198656 + 512 [fio]
             fio-14692 [003] ....  1473.954683: block_rq_issue: 7,0 WS 262144 () 199168 + 512 [fio]
             fio-14692 [003] ....  1473.955686: block_rq_issue: 7,0 WS 262144 () 199680 + 512 [fio]
             fio-14692 [003] ....  1473.956732: block_rq_issue: 7,0 WS 262144 () 200192 + 512 [fio]
             fio-14692 [003] ....  1473.957749: block_rq_issue: 7,0 WS 262144 () 200704 + 512 [fio]
             fio-14692 [003] ....  1473.958741: block_rq_issue: 7,0 WS 262144 () 201216 + 512 [fio]

Signed-off-by: koulihong <koulihong@huawei.com>
---
 fs/f2fs/checkpoint.c |  1 +
 fs/f2fs/data.c       |  5 +++++
 fs/f2fs/f2fs.h       |  1 +
 fs/f2fs/segment.c    | 19 +++++++++++--------
 fs/f2fs/super.c      |  2 ++
 5 files changed, 20 insertions(+), 8 deletions(-)

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index a0eef95..78d5db0 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1081,6 +1081,7 @@ int f2fs_sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type)
 		 * wribacking dentry pages in the freeing inode.
 		 */
 		f2fs_submit_merged_write(sbi, DATA);
+		f2fs_submit_merged_write(sbi, DATA_IPU);
 		cond_resched();
 	}
 	goto retry;
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 5af9810..16fdab2 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -440,6 +440,7 @@ void f2fs_flush_merged_writes(struct f2fs_sb_info *sbi)
 	f2fs_submit_merged_write(sbi, DATA);
 	f2fs_submit_merged_write(sbi, NODE);
 	f2fs_submit_merged_write(sbi, META);
+	f2fs_submit_merged_write(sbi, DATA_IPU);
 }
 
 /*
@@ -2065,6 +2066,7 @@ static int __write_data_page(struct page *page, bool *submitted,
 
 	if (wbc->for_reclaim) {
 		f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA);
+		f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA_IPU);
 		clear_inode_flag(inode, FI_HOT_DATA);
 		f2fs_remove_dirty_inode(inode);
 		submitted = NULL;
@@ -2077,6 +2079,7 @@ static int __write_data_page(struct page *page, bool *submitted,
 
 	if (unlikely(f2fs_cp_error(sbi))) {
 		f2fs_submit_merged_write(sbi, DATA);
+		f2fs_submit_merged_write(sbi, DATA_IPU);
 		submitted = NULL;
 	}
 
@@ -2253,6 +2256,8 @@ static int f2fs_write_cache_pages(struct address_space *mapping,
 	if (nwritten)
 		f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
 								NULL, 0, DATA);
+		f2fs_submit_merged_write_cond(F2FS_M_SB(mapping), mapping->host,
+								NULL, 0, DATA_IPU);
 
 	return ret;
 }
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index f1cf2843..79bf996 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -983,6 +983,7 @@ enum count_type {
 enum page_type {
 	DATA,
 	NODE,
+	DATA_IPU,
 	META,
 	NR_PAGE_TYPE,
 	META_FLUSH,
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 380e0ab..9c67b55 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -430,8 +430,10 @@ static int __f2fs_commit_inmem_pages(struct inode *inode)
 		list_move_tail(&cur->list, &revoke_list);
 	}
 
-	if (submit_bio)
+	if (submit_bio) {
 		f2fs_submit_merged_write_cond(sbi, inode, NULL, 0, DATA);
+		f2fs_submit_merged_write_cond(sbi, inode, NULL, 0, DATA_IPU);
+	}
 
 	if (err) {
 		/*
@@ -3234,7 +3236,6 @@ void f2fs_outplace_write_data(struct dnode_of_data *dn,
 
 int f2fs_inplace_write_data(struct f2fs_io_info *fio)
 {
-	int err;
 	struct f2fs_sb_info *sbi = fio->sbi;
 	unsigned int segno;
 
@@ -3253,13 +3254,12 @@ int f2fs_inplace_write_data(struct f2fs_io_info *fio)
 
 	stat_inc_inplace_blocks(fio->sbi);
 
-	err = f2fs_submit_page_bio(fio);
-	if (!err) {
-		update_device_state(fio);
-		f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
-	}
+	fio->type = DATA_IPU;
+	f2fs_submit_page_write(fio);
+	update_device_state(fio);
+	f2fs_update_iostat(fio->sbi, fio->io_type, F2FS_BLKSIZE);
 
-	return err;
+	return 0;
 }
 
 static inline int __f2fs_get_curseg(struct f2fs_sb_info *sbi,
@@ -3375,6 +3375,9 @@ void f2fs_wait_on_page_writeback(struct page *page,
 		struct f2fs_sb_info *sbi = F2FS_P_SB(page);
 
 		f2fs_submit_merged_write_cond(sbi, NULL, page, 0, type);
+		if (type == DATA)
+			f2fs_submit_merged_write_cond(sbi, NULL, page, 0, DATA_IPU);
+
 		if (ordered) {
 			wait_on_page_writeback(page);
 			f2fs_bug_on(sbi, locked && PageWriteback(page));
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index d95a681..62e1799 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -899,6 +899,8 @@ static int f2fs_drop_inode(struct inode *inode)
 
 			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
 					inode, NULL, 0, DATA);
+			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
+					inode, NULL, 0, DATA_IPU);
 			truncate_inode_pages_final(inode->i_mapping);
 
 			if (F2FS_HAS_BLOCKS(inode))
-- 
2.7.4



_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

  parent reply	other threads:[~2019-10-16 10:04 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-16 10:11 [f2fs-dev] [PATCH v2 0/2] add a new fio cache for IPU Lihong Kou
2019-10-16 10:11 ` [f2fs-dev] [PATCH v2 1/2] Revert "f2fs: add bio cache for IPU" Lihong Kou
2019-10-16 10:11 ` Lihong Kou [this message]
2019-10-22  8:33 ` [f2fs-dev] [PATCH v2 0/2] add a new fio cache for IPU Chao Yu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1571220689-95274-3-git-send-email-koulihong@huawei.com \
    --to=koulihong@huawei.com \
    --cc=jaegeuk@kernel.org \
    --cc=linux-f2fs-devel@lists.sourceforge.net \
    --cc=yuchao0@huawei.com \
    --cc=zengguangyue@hisilicon.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).