Skip to contentSkip to conversation

f2fs: stop checkpoint only from fault injection

PATCH In review Jaegeuk Kim 2 participants 6 messages Dec 29, 2017 – Jan 2, 2018 F2FS file system

lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/2 f2fs: stop checkpoint only from fault injection

If we got EIO by block layer, we still can proceed since EIO can be recovered. But, if we injected stop_checkpoint, we must stop everything.

This should fix generic/441 failure in xfstests.

Signed-off-by Jaegeuk Kim <[email protected]>
---
 fs/f2fs/checkpoint.c | 16 ++++------------
 fs/f2fs/data.c       |  8 ++------
 fs/f2fs/f2fs.h       |  2 +-
 fs/f2fs/file.c       |  8 ++++----
 fs/f2fs/gc.c         |  2 +-
 fs/f2fs/inode.c      |  2 +-
 fs/f2fs/segment.c    |  2 +-
 7 files changed, 14 insertions(+), 26 deletions(-)
fs/f2fs/checkpoint.c +4 -12
index 14d2fedaf3ca..ce92243549dd 100644
@@ -26,11 +26,10 @@
 static struct kmem_cache *ino_entry_slab;
 struct kmem_cache *inode_entry_slab;
 
-void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io)
+void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi)
 {
 	set_ckpt_flags(sbi, CP_ERROR_FLAG);
-	if (!end_io)
-		f2fs_flush_merged_writes(sbi);
+	f2fs_flush_merged_writes(sbi);
 }
 
 /*

@@ -89,18 +88,11 @@ static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index,
 	}
 
 	lock_page(page);
-	if (unlikely(page->mapping != mapping)) {
+	/* If there is any IO error when accessing device, let's try again. */
+	if (unlikely(page->mapping != mapping || !PageUptodate(page))) {
 		f2fs_put_page(page, 1);
 		goto repeat;
 	}
-
-	/*
-	 * if there is any IO error when accessing device, make our filesystem
-	 * readonly and make sure do not write checkpoint with non-uptodate
-	 * meta page.
-	 */
-	if (unlikely(!PageUptodate(page)))
-		f2fs_stop_checkpoint(sbi, false);
 out:
 	return page;
 }
fs/f2fs/data.c +2 -6
index b9fab6186f28..29523013c5a3 100644
@@ -101,18 +101,14 @@ static void f2fs_write_end_io(struct bio *bio)
 			ClearPagePrivate(page);
 			unlock_page(page);
 			mempool_free(page, sbi->write_io_dummy);
-
-			if (unlikely(bio->bi_status))
-				f2fs_stop_checkpoint(sbi, true);
 			continue;
 		}
 
 		fscrypt_pullback_bio_page(&page, true);
 
-		if (unlikely(bio->bi_status)) {
+		if (unlikely(bio->bi_status))
 			mapping_set_error(page->mapping, -EIO);
-			f2fs_stop_checkpoint(sbi, true);
-		}
+
 		dec_page_count(sbi, type);
 		clear_cold_data(page);
 		end_page_writeback(page);
fs/f2fs/f2fs.h +1 -1
index b32eacf11270..f06fb7b332c7 100644
@@ -2741,7 +2741,7 @@ int rw_hint_to_seg_type(enum rw_hint hint);
 /*
  * checkpoint.c
  */
-void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io);
+void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi);
 struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index);
 struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index);
 struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index);
fs/f2fs/file.c +4 -4
index 66a51e9fbe83..7ccf3e7ff498 100644
@@ -1840,21 +1840,21 @@ static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
 	case F2FS_GOING_DOWN_FULLSYNC:
 		sb = freeze_bdev(sb->s_bdev);
 		if (sb && !IS_ERR(sb)) {
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 			thaw_bdev(sb->s_bdev, sb);
 		}
 		break;
 	case F2FS_GOING_DOWN_METASYNC:
 		/* do checkpoint only */
 		f2fs_sync_fs(sb, 1);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	case F2FS_GOING_DOWN_NOSYNC:
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	case F2FS_GOING_DOWN_METAFLUSH:
 		sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	default:
 		ret = -EINVAL;
fs/f2fs/gc.c +1 -1
index 9bffef153a12..538396ea4d29 100644
@@ -56,7 +56,7 @@ static int gc_thread_func(void *data)
 #ifdef CONFIG_F2FS_FAULT_INJECTION
 		if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
 			f2fs_show_injection_info(FAULT_CHECKPOINT);
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 		}
 #endif
fs/f2fs/inode.c +1 -1
index 234322889e65..daac3d142c2c 100644
@@ -441,7 +441,7 @@ void update_inode_page(struct inode *inode)
 			cond_resched();
 			goto retry;
 		} else if (err != -ENOENT) {
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 		}
 		return;
 	}
fs/f2fs/segment.c +1 -1
index b03f65e5dfdc..890d483ad21e 100644
@@ -454,7 +454,7 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
 #ifdef CONFIG_F2FS_FAULT_INJECTION
 	if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
 		f2fs_show_injection_info(FAULT_CHECKPOINT);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 	}
 #endif
Signature · 1 line
2.15.0.531.g2ccb3012c9-goog
author maintainer
#1 lore raw

PATCH 2/2 f2fs: fix missing error number for xattr operation

This fixes generic/449 hang problem caused by no ENOSPC forever which should be returned by setxattr under disk full scenario.

Signed-off-by Jaegeuk Kim <[email protected]>
---
 fs/f2fs/xattr.c | 2 ++
 1 file changed, 2 insertions(+)
fs/f2fs/xattr.c +2 -0
index 5e2b905cc0c4..600162f4ddbf 100644
@@ -433,6 +433,7 @@ static inline int write_all_xattrs(struct inode *inode, __u32 hsize,
 	if (F2FS_I(inode)->i_xattr_nid) {
 		xpage = get_node_page(sbi, F2FS_I(inode)->i_xattr_nid);
 		if (IS_ERR(xpage)) {
+			err = PTR_ERR(xpage);
 			alloc_nid_failed(sbi, new_nid);
 			goto in_page_out;
 		}

@@ -443,6 +444,7 @@ static inline int write_all_xattrs(struct inode *inode, __u32 hsize,
 		set_new_dnode(&dn, inode, NULL, NULL, new_nid);
 		xpage = new_node_page(&dn, XATTR_NODE_OFFSET);
 		if (IS_ERR(xpage)) {
+			err = PTR_ERR(xpage);
 			alloc_nid_failed(sbi, new_nid);
 			goto in_page_out;
 		}
Signature · 1 line
2.15.0.531.g2ccb3012c9-goog
Reviewed-by on v1 2/2
#2 lore raw

Re: [f2fs-dev] [PATCH 2/2] f2fs: fix missing error number for xattr operation

Quoted 4 lines · Jaegeuk Kim

This fixes generic/449 hang problem caused by no ENOSPC forever which should be returned by setxattr under disk full scenario.

Signed-off-by Jaegeuk Kim <[email protected]>
Reviewed-by Chao Yu <[email protected]>

Thanks,

#3 lore raw
In reply to Jaegeuk Kim
Quoted 35 lines · Jaegeuk Kim

If we got EIO by block layer, we still can proceed since EIO can be recovered. But, if we injected stop_checkpoint, we must stop everything.

This should fix generic/441 failure in xfstests.

Signed-off-by Jaegeuk Kim <[email protected]>
---
 fs/f2fs/checkpoint.c | 16 ++++------------
 fs/f2fs/data.c       |  8 ++------
 fs/f2fs/f2fs.h       |  2 +-
 fs/f2fs/file.c       |  8 ++++----
 fs/f2fs/gc.c         |  2 +-
 fs/f2fs/inode.c      |  2 +-
 fs/f2fs/segment.c    |  2 +-
 7 files changed, 14 insertions(+), 26 deletions(-)
fs/f2fs/checkpoint.c +2 -3
index 14d2fedaf3ca..ce92243549dd 100644
@@ -26,11 +26,10 @@
 static struct kmem_cache *ino_entry_slab;
 struct kmem_cache *inode_entry_slab;
 
-void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io)
+void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi)
 {
 	set_ckpt_flags(sbi, CP_ERROR_FLAG);
-	if (!end_io)
-		f2fs_flush_merged_writes(sbi);
+	f2fs_flush_merged_writes(sbi);
 }
 
 /*

fs/f2fs/checkpoint.c static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index, · line 92

Quoted 6 lines
fs/f2fs/checkpoint.c +2 -1
@@ -89,18 +88,11 @@ static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index,
 	}
 
 	lock_page(page);
-	if (unlikely(page->mapping != mapping)) {
+	/* If there is any IO error when accessing device, let's try again. */
+	if (unlikely(page->mapping != mapping || !PageUptodate(page))) {

If IO error is not recovered, we will run into endless loop?

Quoted 24 lines
fs/f2fs/checkpoint.c +0 -8
 		f2fs_put_page(page, 1);
 		goto repeat;
 	}
-
-	/*
-	 * if there is any IO error when accessing device, make our filesystem
-	 * readonly and make sure do not write checkpoint with non-uptodate
-	 * meta page.
-	 */
-	if (unlikely(!PageUptodate(page)))
-		f2fs_stop_checkpoint(sbi, false);
 out:
 	return page;
 }
fs/f2fs/data.c +0 -2
index b9fab6186f28..29523013c5a3 100644
@@ -101,18 +101,14 @@ static void f2fs_write_end_io(struct bio *bio)
 			ClearPagePrivate(page);
 			unlock_page(page);
 			mempool_free(page, sbi->write_io_dummy);
-
-			if (unlikely(bio->bi_status))

fs/f2fs/data.c static void f2fs_write_end_io(struct bio *bio) · line 114

Quoted 10 lines
fs/f2fs/data.c +1 -3
-				f2fs_stop_checkpoint(sbi, true);
 			continue;
 		}
 
 		fscrypt_pullback_bio_page(&page, true);
 
-		if (unlikely(bio->bi_status)) {
+		if (unlikely(bio->bi_status))
 			mapping_set_error(page->mapping, -EIO);
-			f2fs_stop_checkpoint(sbi, true);

If we do not stop checkpoint here, fs meta/node data and user data can be more and more corrupted and inconsistent... at least, we should stop checkpoint if there is EIO when we writeback fs meta/node data.

Thanks,

Quoted 88 lines
fs/f2fs/data.c +1 -1
-		}
+
 		dec_page_count(sbi, type);
 		clear_cold_data(page);
 		end_page_writeback(page);
fs/f2fs/f2fs.h +1 -1
index b32eacf11270..f06fb7b332c7 100644
@@ -2741,7 +2741,7 @@ int rw_hint_to_seg_type(enum rw_hint hint);
 /*
  * checkpoint.c
  */
-void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io);
+void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi);
 struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index);
 struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index);
 struct page *get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index);
fs/f2fs/file.c +4 -4
index 66a51e9fbe83..7ccf3e7ff498 100644
@@ -1840,21 +1840,21 @@ static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
 	case F2FS_GOING_DOWN_FULLSYNC:
 		sb = freeze_bdev(sb->s_bdev);
 		if (sb && !IS_ERR(sb)) {
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 			thaw_bdev(sb->s_bdev, sb);
 		}
 		break;
 	case F2FS_GOING_DOWN_METASYNC:
 		/* do checkpoint only */
 		f2fs_sync_fs(sb, 1);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	case F2FS_GOING_DOWN_NOSYNC:
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	case F2FS_GOING_DOWN_METAFLUSH:
 		sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 		break;
 	default:
 		ret = -EINVAL;
fs/f2fs/gc.c +1 -1
index 9bffef153a12..538396ea4d29 100644
@@ -56,7 +56,7 @@ static int gc_thread_func(void *data)
 #ifdef CONFIG_F2FS_FAULT_INJECTION
 		if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
 			f2fs_show_injection_info(FAULT_CHECKPOINT);
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 		}
 #endif
fs/f2fs/inode.c +1 -1
index 234322889e65..daac3d142c2c 100644
@@ -441,7 +441,7 @@ void update_inode_page(struct inode *inode)
 			cond_resched();
 			goto retry;
 		} else if (err != -ENOENT) {
-			f2fs_stop_checkpoint(sbi, false);
+			f2fs_stop_checkpoint(sbi);
 		}
 		return;
 	}
fs/f2fs/segment.c +1 -1
index b03f65e5dfdc..890d483ad21e 100644
@@ -454,7 +454,7 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
 #ifdef CONFIG_F2FS_FAULT_INJECTION
 	if (time_to_inject(sbi, FAULT_CHECKPOINT)) {
 		f2fs_show_injection_info(FAULT_CHECKPOINT);
-		f2fs_stop_checkpoint(sbi, false);
+		f2fs_stop_checkpoint(sbi);
 	}
 #endif
author maintainer
#4 lore raw

We can give another chance to write user data, which can resolve generic/441.

Signed-off-by Jaegeuk Kim <[email protected]>

---

Change log from v1:
- skip stop_checkpoint for user data writes

 fs/f2fs/data.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
fs/f2fs/data.c +2 -1
index b9fab6186f28..45f65a5b9871 100644
@@ -111,7 +111,8 @@ static void f2fs_write_end_io(struct bio *bio)
 
 		if (unlikely(bio->bi_status)) {
 			mapping_set_error(page->mapping, -EIO);
-			f2fs_stop_checkpoint(sbi, true);
+			if (type == F2FS_WB_CP_DATA)
+				f2fs_stop_checkpoint(sbi, true);
 		}
 		dec_page_count(sbi, type);
 		clear_cold_data(page);
Signature · 1 line
2.15.0.531.g2ccb3012c9-goog