f2fs: support revoking atomic written pages

f2fs support atomic write with following semantics:
1. open db file
2. ioctl start atomic write
3. (write db file) * n
4. ioctl commit atomic write
5. close db file

With this flow we can avoid file becoming corrupted when abnormal power
cut, because we hold data of transaction in referenced pages linked in
inmem_pages list of inode, but without setting them dirty, so these data
won't be persisted unless we commit them in step 4.

But we should still hold journal db file in memory by using volatile
write, because our semantics of 'atomic write support' is incomplete, in
step 4, we could fail to submit all dirty data of transaction, once
partial dirty data was committed in storage, then after a checkpoint &
abnormal power-cut, db file will be corrupted forever.

So this patch tries to improve atomic write flow by adding a revoking flow,
once inner error occurs in committing, this gives another chance to try to
revoke these partial submitted data of current transaction, it makes
committing operation more like aotmical one.

If we're not lucky, once revoking operation was failed, EAGAIN will be
reported to user for suggesting doing the recovery with held journal file,
or retrying current transaction again.

Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

6 files changed, 92 insertions, 36 deletions

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index ca99a2aca107..8b46e5d9bcdd 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -1072,6 +1072,7 @@ int do_write_data_page(struct f2fs_io_info *fio)
 		return err;
 
 	fio->blk_addr = dn.data_blkaddr;
+	fio->old_blkaddr = dn.data_blkaddr;
 
 	/* This page is already truncated */
 	if (fio->blk_addr == NULL_ADDR) {
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index d8caf0c86593..0d2b1ba9660e 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -686,6 +686,7 @@ enum page_type {
 	META_FLUSH,
 	INMEM,		/* the below types are used by tracepoints only. */
 	INMEM_DROP,
+	INMEM_REVOKE,
 	IPU,
 	OPU,
 };
@@ -695,6 +696,7 @@ struct f2fs_io_info {
 	enum page_type type;	/* contains DATA/NODE/META/META_FLUSH */
 	int rw;			/* contains R/RS/W/WS with REQ_META/REQ_PRIO */
 	block_t blk_addr;	/* block address to be written */
+	block_t old_blkaddr;	/* old block address before Cow */
 	struct page *page;	/* page to be written */
 	struct page *encrypted_page;	/* encrypted page */
 };
@@ -1853,7 +1855,7 @@ void write_node_page(unsigned int, struct f2fs_io_info *);
 void write_data_page(struct dnode_of_data *, struct f2fs_io_info *);
 void rewrite_data_page(struct f2fs_io_info *);
 void f2fs_replace_block(struct f2fs_sb_info *, struct dnode_of_data *,
-				block_t, block_t, unsigned char, bool);
+				block_t, block_t, unsigned char, bool, bool);
 void allocate_data_block(struct f2fs_sb_info *, struct page *,
 		block_t, block_t *, struct f2fs_summary *, int);
 void f2fs_wait_on_page_writeback(struct page *, enum page_type, bool);
diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 92273b6e9f47..05f5f2f8f8fd 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -883,7 +883,7 @@ static int __exchange_data_block(struct inode *inode, pgoff_t src,
 
 		get_node_info(sbi, dn.nid, &ni);
 		f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
-				ni.version, true);
+				ni.version, true, false);
 		f2fs_put_dnode(&dn);
 	} else {
 		struct page *psrc, *pdst;
diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
index 5045dd6a27e9..0b30cd2aeebd 100644
--- a/fs/f2fs/recovery.c
+++ b/fs/f2fs/recovery.c
@@ -465,7 +465,7 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 
 			/* write dummy data page */
 			f2fs_replace_block(sbi, &dn, src, dest,
-							ni.version, false);
+						ni.version, false, false);
 			recovered++;
 		}
 	}
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 5ae66baf6989..ad5da895260a 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -191,24 +191,48 @@ void register_inmem_page(struct inode *inode, struct page *page)
 	trace_f2fs_register_inmem_page(page, INMEM);
 }
 
-static void __revoke_inmem_pages(struct inode *inode,
-							struct list_head *head)
+static int __revoke_inmem_pages(struct inode *inode,
+				struct list_head *head, bool drop, bool recover)
 {
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct inmem_pages *cur, *tmp;
+	int err = 0;
 
 	list_for_each_entry_safe(cur, tmp, head, list) {
-		trace_f2fs_commit_inmem_page(cur->page, INMEM_DROP);
+		struct page *page = cur->page;
+
+		if (drop)
+			trace_f2fs_commit_inmem_page(page, INMEM_DROP);
+
+		lock_page(page);
 
-		lock_page(cur->page);
-		ClearPageUptodate(cur->page);
-		set_page_private(cur->page, 0);
-		ClearPagePrivate(cur->page);
-		f2fs_put_page(cur->page, 1);
+		if (recover) {
+			struct dnode_of_data dn;
+			struct node_info ni;
+
+			trace_f2fs_commit_inmem_page(page, INMEM_REVOKE);
+
+			set_new_dnode(&dn, inode, NULL, NULL, 0);
+			if (get_dnode_of_data(&dn, page->index, LOOKUP_NODE)) {
+				err = -EAGAIN;
+				goto next;
+			}
+			get_node_info(sbi, dn.nid, &ni);
+			f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
+					cur->old_addr, ni.version, true, true);
+			f2fs_put_dnode(&dn);
+		}
+next:
+		ClearPageUptodate(page);
+		set_page_private(page, 0);
+		ClearPageUptodate(page);
+		f2fs_put_page(page, 1);
 
 		list_del(&cur->list);
 		kmem_cache_free(inmem_entry_slab, cur);
 		dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
 	}
+	return err;
 }
 
 void drop_inmem_pages(struct inode *inode)
@@ -216,11 +240,12 @@ void drop_inmem_pages(struct inode *inode)
 	struct f2fs_inode_info *fi = F2FS_I(inode);
 
 	mutex_lock(&fi->inmem_lock);
-	__revoke_inmem_pages(inode, &fi->inmem_pages);
+	__revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
 	mutex_unlock(&fi->inmem_lock);
 }
 
-static int __commit_inmem_pages(struct inode *inode)
+static int __commit_inmem_pages(struct inode *inode,
+					struct list_head *revoke_list)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct f2fs_inode_info *fi = F2FS_I(inode);
@@ -235,34 +260,40 @@ static int __commit_inmem_pages(struct inode *inode)
 	int err = 0;
 
 	list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
-		lock_page(cur->page);
-		if (cur->page->mapping == inode->i_mapping) {
-			set_page_dirty(cur->page);
-			f2fs_wait_on_page_writeback(cur->page, DATA, true);
-			if (clear_page_dirty_for_io(cur->page))
+		struct page *page = cur->page;
+
+		lock_page(page);
+		if (page->mapping == inode->i_mapping) {
+			trace_f2fs_commit_inmem_page(page, INMEM);
+
+			set_page_dirty(page);
+			f2fs_wait_on_page_writeback(page, DATA, true);
+			if (clear_page_dirty_for_io(page))
 				inode_dec_dirty_pages(inode);
-			trace_f2fs_commit_inmem_page(cur->page, INMEM);
-			fio.page = cur->page;
+
+			fio.page = page;
 			err = do_write_data_page(&fio);
 			if (err) {
-				unlock_page(cur->page);
+				unlock_page(page);
 				break;
 			}
-			clear_cold_data(cur->page);
-			submit_bio = true;
-		}
 
-		set_page_private(cur->page, 0);
-		ClearPagePrivate(cur->page);
-		f2fs_put_page(cur->page, 1);
+			/* record old blkaddr for revoking */
+			cur->old_addr = fio.old_blkaddr;
 
-		list_del(&cur->list);
-		kmem_cache_free(inmem_entry_slab, cur);
-		dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
+			clear_cold_data(page);
+			submit_bio = true;
+		}
+		unlock_page(page);
+		list_move_tail(&cur->list, revoke_list);
 	}
 
 	if (submit_bio)
 		f2fs_submit_merged_bio_cond(sbi, inode, NULL, 0, DATA, WRITE);
+
+	if (!err)
+		__revoke_inmem_pages(inode, revoke_list, false, false);
+
 	return err;
 }
 
@@ -270,13 +301,32 @@ int commit_inmem_pages(struct inode *inode)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 	struct f2fs_inode_info *fi = F2FS_I(inode);
-	int err = 0;
+	struct list_head revoke_list;
+	int err;
 
+	INIT_LIST_HEAD(&revoke_list);
 	f2fs_balance_fs(sbi, true);
 	f2fs_lock_op(sbi);
 
 	mutex_lock(&fi->inmem_lock);
-	err = __commit_inmem_pages(inode);
+	err = __commit_inmem_pages(inode, &revoke_list);
+	if (err) {
+		int ret;
+		/*
+		 * try to revoke all committed pages, but still we could fail
+		 * due to no memory or other reason, if that happened, EAGAIN
+		 * will be returned, which means in such case, transaction is
+		 * already not integrity, caller should use journal to do the
+		 * recovery or rewrite & commit last transaction. For other
+		 * error number, revoking was done by filesystem itself.
+		 */
+		ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
+		if (ret)
+			err = ret;
+
+		/* drop all uncommitted pages */
+		__revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
+	}
 	mutex_unlock(&fi->inmem_lock);
 
 	f2fs_unlock_op(sbi);
@@ -1360,7 +1410,7 @@ void rewrite_data_page(struct f2fs_io_info *fio)
 static void __f2fs_replace_block(struct f2fs_sb_info *sbi,
 				struct f2fs_summary *sum,
 				block_t old_blkaddr, block_t new_blkaddr,
-				bool recover_curseg)
+				bool recover_curseg, bool recover_newaddr)
 {
 	struct sit_info *sit_i = SIT_I(sbi);
 	struct curseg_info *curseg;
@@ -1403,7 +1453,7 @@ static void __f2fs_replace_block(struct f2fs_sb_info *sbi,
 	curseg->next_blkoff = GET_BLKOFF_FROM_SEG0(sbi, new_blkaddr);
 	__add_sum_entry(sbi, type, sum);
 
-	if (!recover_curseg)
+	if (!recover_curseg || recover_newaddr)
 		update_sit_entry(sbi, new_blkaddr, 1);
 	if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
 		update_sit_entry(sbi, old_blkaddr, -1);
@@ -1427,13 +1477,15 @@ static void __f2fs_replace_block(struct f2fs_sb_info *sbi,
 
 void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
 				block_t old_addr, block_t new_addr,
-				unsigned char version, bool recover_curseg)
+				unsigned char version, bool recover_curseg,
+				bool recover_newaddr)
 {
 	struct f2fs_summary sum;
 
 	set_summary(&sum, dn->nid, dn->ofs_in_node, version);
 
-	__f2fs_replace_block(sbi, &sum, old_addr, new_addr, recover_curseg);
+	__f2fs_replace_block(sbi, &sum, old_addr, new_addr,
+					recover_curseg, recover_newaddr);
 
 	dn->data_blkaddr = new_addr;
 	set_data_blkaddr(dn);
diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
index cd7111b9a664..2f1a4220f14b 100644
--- a/fs/f2fs/segment.h
+++ b/fs/f2fs/segment.h
@@ -191,6 +191,7 @@ struct segment_allocation {
 struct inmem_pages {
 	struct list_head list;
 	struct page *page;
+	block_t old_addr;		/* for revoking when fail to commit */
 };
 
 struct sit_info {