[PATCH 13/13] f2fs: handle EIO not to break fs consistency
From: Jaegeuk Kim
Date: Tue Aug 12 2014 - 15:49:50 EST
There are two rules when EIO is occurred.
1. don't write any checkpoint data to preserve the previous checkpoint
2. don't lose the cached data/node/meta pages
So, at first, this patch adds set_page_dirty in f2fs_write_end_io's failure.
Then, writing checkpoint/dentry/node blocks is not allowed.
Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
---
fs/f2fs/checkpoint.c | 47 +++++++++++++++++++++++++++++++++--------------
fs/f2fs/data.c | 10 +++++++++-
fs/f2fs/f2fs.h | 1 +
fs/f2fs/node.c | 2 ++
fs/f2fs/super.c | 3 +++
5 files changed, 48 insertions(+), 15 deletions(-)
diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index d074e27..b365c64 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -160,14 +160,11 @@ static int f2fs_write_meta_page(struct page *page,
goto redirty_out;
if (wbc->for_reclaim)
goto redirty_out;
-
- /* Should not write any meta pages, if any IO error was occurred */
if (unlikely(f2fs_cp_error(sbi)))
- goto no_write;
+ goto redirty_out;
f2fs_wait_on_page_writeback(page, META);
write_meta_page(sbi, page);
-no_write:
dec_page_count(sbi, F2FS_DIRTY_META);
unlock_page(page);
return 0;
@@ -348,7 +345,7 @@ bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
return e ? true : false;
}
-static void release_dirty_inode(struct f2fs_sb_info *sbi)
+void release_dirty_inode(struct f2fs_sb_info *sbi)
{
struct ino_entry *e, *tmp;
int i;
@@ -737,7 +734,7 @@ retry:
/*
* Freeze all the FS-operations for checkpoint.
*/
-static void block_operations(struct f2fs_sb_info *sbi)
+static int block_operations(struct f2fs_sb_info *sbi)
{
struct writeback_control wbc = {
.sync_mode = WB_SYNC_ALL,
@@ -745,6 +742,7 @@ static void block_operations(struct f2fs_sb_info *sbi)
.for_reclaim = 0,
};
struct blk_plug plug;
+ int err = 0;
blk_start_plug(&plug);
@@ -754,6 +752,10 @@ retry_flush_dents:
if (get_pages(sbi, F2FS_DIRTY_DENTS)) {
f2fs_unlock_all(sbi);
sync_dirty_dir_inodes(sbi);
+ if (unlikely(f2fs_cp_error(sbi))) {
+ err = -EIO;
+ goto out;
+ }
goto retry_flush_dents;
}
@@ -767,9 +769,16 @@ retry_flush_nodes:
if (get_pages(sbi, F2FS_DIRTY_NODES)) {
up_write(&sbi->node_write);
sync_node_pages(sbi, 0, &wbc);
+ if (unlikely(f2fs_cp_error(sbi))) {
+ f2fs_unlock_all(sbi);
+ err = -EIO;
+ goto out;
+ }
goto retry_flush_nodes;
}
+out:
blk_finish_plug(&plug);
+ return err;
}
static void unblock_operations(struct f2fs_sb_info *sbi)
@@ -813,8 +822,11 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg));
/* Flush all the NAT/SIT pages */
- while (get_pages(sbi, F2FS_DIRTY_META))
+ while (get_pages(sbi, F2FS_DIRTY_META)) {
sync_meta_pages(sbi, META, LONG_MAX);
+ if (unlikely(f2fs_cp_error(sbi)))
+ return;
+ }
next_free_nid(sbi, &last_nid);
@@ -931,14 +943,19 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
sbi->last_valid_block_count = sbi->total_valid_block_count;
sbi->alloc_valid_block_count = 0;
+ if (unlikely(f2fs_cp_error(sbi)))
+ return;
+
/* Here, we only have one bio having CP pack */
sync_meta_pages(sbi, META_FLUSH, LONG_MAX);
- if (!f2fs_cp_error(sbi)) {
- clear_prefree_segments(sbi);
- release_dirty_inode(sbi);
- F2FS_RESET_SB_DIRT(sbi);
- }
+ release_dirty_inode(sbi);
+
+ if (unlikely(f2fs_cp_error(sbi)))
+ return;
+
+ clear_prefree_segments(sbi);
+ F2FS_RESET_SB_DIRT(sbi);
}
/*
@@ -955,8 +972,10 @@ void write_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
if (!sbi->s_dirty)
goto out;
-
- block_operations(sbi);
+ if (unlikely(f2fs_cp_error(sbi)))
+ goto out;
+ if (block_operations(sbi))
+ goto out;
trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish block_ops");
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index ac3ccc2..68834e2 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -53,7 +53,7 @@ static void f2fs_write_end_io(struct bio *bio, int err)
struct page *page = bvec->bv_page;
if (unlikely(err)) {
- SetPageError(page);
+ set_page_dirty(page);
set_bit(AS_EIO, &page->mapping->flags);
f2fs_stop_checkpoint(sbi);
}
@@ -836,10 +836,18 @@ write:
/* Dentry blocks are controlled by checkpoint */
if (S_ISDIR(inode->i_mode)) {
+ if (unlikely(f2fs_cp_error(sbi)))
+ goto redirty_out;
err = do_write_data_page(page, &fio);
goto done;
}
+ if (unlikely(f2fs_cp_error(sbi))) {
+ SetPageError(page);
+ unlock_page(page);
+ return 0;
+ }
+
if (!wbc->for_reclaim)
need_balance_fs = true;
else if (has_not_enough_free_secs(sbi, 0))
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 46ceb88..2d009ae 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1267,6 +1267,7 @@ int ra_meta_pages(struct f2fs_sb_info *, int, int, int);
long sync_meta_pages(struct f2fs_sb_info *, enum page_type, long);
void add_dirty_inode(struct f2fs_sb_info *, nid_t, int type);
void remove_dirty_inode(struct f2fs_sb_info *, nid_t, int type);
+void release_dirty_inode(struct f2fs_sb_info *);
bool exist_written_data(struct f2fs_sb_info *, nid_t, int);
int acquire_orphan_inode(struct f2fs_sb_info *);
void release_orphan_inode(struct f2fs_sb_info *);
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 9f126f8..d2f7842 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -1215,6 +1215,8 @@ static int f2fs_write_node_page(struct page *page,
if (unlikely(sbi->por_doing))
goto redirty_out;
+ if (unlikely(f2fs_cp_error(sbi)))
+ goto redirty_out;
f2fs_wait_on_page_writeback(page, NODE);
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index c4f90cd..c999d67 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -435,6 +435,9 @@ static void f2fs_put_super(struct super_block *sb)
if (sbi->s_dirty)
write_checkpoint(sbi, true);
+ /* EIO will skip do checkpoint, so here we need to release this */
+ release_dirty_inode(sbi);
+
iput(sbi->node_inode);
iput(sbi->meta_inode);
--
1.8.5.2 (Apple Git-48)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/