From: amir73il@users.sourceforge.net
To: linux-ext4@vger.kernel.org
Cc: tytso@mit.edu, Amir Goldstein <amir73il@users.sf.net>,
Yongqiang Yang <xiaoqiangnk@gmail.com>
Subject: [PATCH RFC 08/30] ext4: snapshot hooks - move extent file data blocks
Date: Mon, 9 May 2011 19:41:26 +0300 [thread overview]
Message-ID: <1304959308-11122-9-git-send-email-amir73il@users.sourceforge.net> (raw)
In-Reply-To: <1304959308-11122-1-git-send-email-amir73il@users.sourceforge.net>
From: Amir Goldstein <amir73il@users.sf.net>
Extent mapped file data is moved into snapshot in ext4_ext_map_blocks().
If a part of a extent is to be moved, the extent is splitted. Fragmentation
is light because of delayed-move-on-write.
Signed-off-by: Amir Goldstein <amir73il@users.sf.net>
Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com>
---
fs/ext4/ext4_jbd2.h | 2 -
fs/ext4/extents.c | 143 +++++++++++++++++++++++++++++++++++++++++++++------
fs/ext4/inode.c | 3 +-
3 files changed, 128 insertions(+), 20 deletions(-)
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index 1c119cc..ea3a0a0 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -369,8 +369,6 @@ static inline int ext4_snapshot_should_move_data(struct inode *inode)
return 0;
if (EXT4_JOURNAL(inode) == NULL)
return 0;
- if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
- return 0;
/* when a data block is journaled, it is already COWed as metadata */
if (ext4_should_journal_data(inode))
return 0;
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index c8cab3d..11fe058 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -1256,11 +1256,10 @@ static int ext4_ext_search_left(struct inode *inode,
return 0;
}
- if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
- EXT4_ERROR_INODE(inode,
- "logical %d < ee_block %d + ee_len %d!",
- *logical, le32_to_cpu(ex->ee_block), ee_len);
- return -EIO;
+ if (*logical < (le32_to_cpu(ex->ee_block) + ee_len)) {
+ *logical -= 1;
+ *phys = ext4_ext_pblock(ex) + *logical;
+ return 0;
}
*logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
@@ -1324,11 +1323,10 @@ static int ext4_ext_search_right(struct inode *inode,
return 0;
}
- if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
- EXT4_ERROR_INODE(inode,
- "logical %d < ee_block %d + ee_len %d!",
- *logical, le32_to_cpu(ex->ee_block), ee_len);
- return -EIO;
+ if (*logical < (le32_to_cpu(ex->ee_block) + ee_len)) {
+ *logical += 1;
+ *phys = ext4_ext_pblock(ex) + *logical;
+ return 0;
}
if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) {
@@ -3155,7 +3153,8 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
struct ext4_map_blocks *map, int flags)
{
struct ext4_ext_path *path = NULL;
- struct ext4_extent newex, *ex;
+ struct ext4_extent newex, *ex = NULL;
+ ext4_fsblk_t oldblock = 0;
ext4_fsblk_t newblock = 0;
int err = 0, depth, ret;
unsigned int allocated = 0;
@@ -3184,7 +3183,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
/* number of remaining blocks in the extent */
allocated = ext4_ext_get_actual_len(&newex) -
(map->m_lblk - le32_to_cpu(newex.ee_block));
- goto out;
+ goto found;
}
}
@@ -3235,7 +3234,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
if (!ext4_ext_is_uninitialized(ex)) {
ext4_ext_put_in_cache(inode, ee_block,
ee_len, ee_start);
- goto out;
+ goto found;
}
ret = ext4_ext_handle_uninitialized_extents(handle,
inode, map, path, flags, allocated,
@@ -3256,6 +3255,59 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
goto out2;
}
+
+ /*
+ * two cases:
+ * 1. the request block is found.
+ * a. If EXT4_GET_BLOCKS_CREATE is not set, we will test
+ * if MOW is needed.
+ * b. If EXT4_GET_BLOCKS_CREATE is set. MOW will be done
+ * if MOW is needed.
+ *
+ * 2. the request block is not found, EXT4_GET_BLOCKS_CREATE
+ * must be set and MOW must be not needed.
+ */
+found:
+ if (newblock && (flags & EXT4_GET_BLOCKS_MOVE_ON_WRITE)) {
+ BUG_ON(!ext4_snapshot_should_move_data(inode));
+ /*
+ * Should move 1 block to snapshot?
+ *
+ * XXX With delayed-move-write support,
+ * multi-blocks should be moved each time.
+ */
+ allocated = allocated < map->m_len ? allocated : map->m_len;
+ err = ext4_snapshot_get_move_access(handle, inode, newblock,
+ &allocated, 0);
+ map->m_len = allocated;
+ if (err > 0) {
+ if (!(flags & EXT4_GET_BLOCKS_CREATE)) {
+ /* Do not map found block. */
+ map->m_flags |= EXT4_MAP_REMAP;
+ err = 0;
+ goto out;
+ } else {
+ oldblock = newblock;
+ }
+ } else if (err < 0)
+ goto out2;
+
+ if ((path == NULL) && (flags & EXT4_GET_BLOCKS_CREATE)) {
+ /* find extent for this block */
+ path = ext4_ext_find_extent(inode, map->m_lblk, NULL);
+ if (IS_ERR(path)) {
+ err = PTR_ERR(path);
+ path = NULL;
+ goto out2;
+ }
+ depth = ext_depth(inode);
+ ex = path[depth].p_ext;
+ }
+ }
+
+ if (!(flags & EXT4_GET_BLOCKS_CREATE))
+ goto out;
+
/*
* Okay, we need to do block allocation.
*/
@@ -3265,7 +3317,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
if (err)
goto out2;
- ar.lright = map->m_lblk;
+ ar.lright = map->m_lblk + allocated;
err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright);
if (err)
goto out2;
@@ -3286,7 +3338,11 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
/* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
newex.ee_block = cpu_to_le32(map->m_lblk);
newex.ee_len = cpu_to_le16(map->m_len);
- err = ext4_ext_check_overlap(inode, &newex, path);
+ if (oldblock) {
+ /* Overlap checking is not needed for MOW case. */
+ err = 0;
+ } else
+ err = ext4_ext_check_overlap(inode, &newex, path);
if (err)
allocated = ext4_ext_get_actual_len(&newex);
else
@@ -3337,7 +3393,55 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
if (err)
goto out2;
- err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
+ if (oldblock) {
+ /*
+ * Move oldblocks to snapshot.
+ */
+ map->m_len = ar.len;
+ err = ext4_snapshot_get_move_access(handle, inode,
+ oldblock, &map->m_len, 1);
+ if (err <= 0 || map->m_len < ar.len) {
+ /* failed to move to snapshot - abort! */
+ err = err ? : -EIO;
+ ext4_journal_abort_handle(__func__, __LINE__,
+ "ext4_snapshot_get_move_access", NULL,
+ handle, err);
+ } else {
+ /*
+ * Move to snapshot successfully.
+ * TODO merge extent after finishing MOW
+ */
+ err = ext4_split_extent(handle, inode, path, map, 0,
+ flags | EXT4_GET_BLOCKS_PRE_IO);
+ if (err < 0)
+ goto out;
+
+ /* extent tree may be changed. */
+ depth = ext_depth(inode);
+ ext4_ext_drop_refs(path);
+ path = ext4_ext_find_extent(inode, map->m_lblk, path);
+ if (IS_ERR(path)) {
+ err = PTR_ERR(path);
+ goto out;
+ }
+
+ /* just verify splitting. */
+ ex = path[depth].p_ext;
+ BUG_ON(le32_to_cpu(ex->ee_block) != map->m_lblk ||
+ ext4_ext_get_actual_len(ex) != map->m_len);
+
+ err = ext4_ext_get_access(handle, inode, path + depth);
+ if (!err) {
+ /* splice new blocks to the inode*/
+ ext4_ext_store_pblock(ex, newblock);
+ err = ext4_ext_dirty(handle, inode,
+ path + depth);
+ }
+ }
+
+ } else
+ err = ext4_ext_insert_extent(handle, inode,
+ path, &newex, flags);
if (err) {
/* free data blocks we just allocated */
/* not a good idea to call discard here directly,
@@ -3366,7 +3470,12 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
* Cache the extent and update transaction to commit on fdatasync only
* when it is _not_ an uninitialized extent.
*/
- if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
+ if (IS_COWING(handle)) {
+ /*
+ * snapshot does not supprt fdatasync and fsync
+ * and there is no need to cache extent
+ */
+ } else if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
ext4_ext_put_in_cache(inode, map->m_lblk, allocated, newblock);
ext4_update_inode_fsync_trans(handle, inode, 1);
} else
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 476606b..866ac36 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1357,7 +1357,8 @@ int ext4_map_blocks(handle_t *handle, struct inode *inode,
*/
down_read((&EXT4_I(inode)->i_data_sem));
if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
- retval = ext4_ext_map_blocks(handle, inode, map, 0);
+ retval = ext4_ext_map_blocks(handle, inode, map,
+ flags & EXT4_GET_BLOCKS_MOVE_ON_WRITE);
} else {
retval = ext4_ind_map_blocks(handle, inode, map,
flags & EXT4_GET_BLOCKS_MOVE_ON_WRITE);
--
1.7.0.4
next prev parent reply other threads:[~2011-05-09 16:43 UTC|newest]
Thread overview: 74+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-05-09 16:41 [PATCH RFC 00/30] Ext4 snapshots - core patches amir73il
2011-05-09 16:41 ` [PATCH RFC 01/30] ext4: EXT4 snapshots (Experimental) amir73il
2011-06-06 14:50 ` Lukas Czerner
2011-06-07 9:28 ` Amir G.
2011-06-07 10:42 ` Lukas Czerner
2011-06-07 13:20 ` Amir G.
2011-05-09 16:41 ` [PATCH RFC 02/30] ext4: snapshot debugging support amir73il
2011-06-06 15:08 ` Lukas Czerner
2011-06-07 9:59 ` Amir G.
2011-06-07 10:49 ` Lukas Czerner
2011-05-09 16:41 ` [PATCH RFC 03/30] ext4: snapshot hooks - inside JBD hooks amir73il
2011-06-06 15:53 ` Lukas Czerner
2011-06-06 16:08 ` Amir G.
2011-06-06 19:01 ` Amir G.
2011-05-09 16:41 ` [PATCH RFC 04/30] ext4: snapshot hooks - block bitmap access amir73il
2011-05-09 16:41 ` [PATCH RFC 05/30] ext4: snapshot hooks - delete blocks amir73il
2011-06-07 11:24 ` Lukas Czerner
2011-06-07 13:24 ` Amir G.
2011-06-07 13:32 ` Lukas Czerner
2011-05-09 16:41 ` [PATCH RFC 06/30] ext4: snapshot hooks - move data blocks amir73il
2011-05-09 16:41 ` [PATCH RFC 07/30] ext4: snapshot hooks - direct I/O amir73il
2011-05-09 16:41 ` amir73il [this message]
2011-05-09 16:41 ` [PATCH RFC 09/30] ext4: snapshot file amir73il
2011-06-02 11:52 ` Amir G.
2011-05-09 16:41 ` [PATCH RFC 10/30] ext4: snapshot file - read through to block device amir73il
2011-05-09 16:41 ` [PATCH RFC 11/30] ext4: snapshot file - permissions amir73il
2011-05-09 16:41 ` [PATCH RFC 12/30] ext4: snapshot file - store on disk amir73il
2011-05-09 16:41 ` [PATCH RFC 13/30] ext4: snapshot file - increase maximum file size limit to 16TB amir73il
2011-06-02 11:47 ` Amir G.
2011-06-03 0:48 ` Ted Ts'o
2011-06-03 4:45 ` Amir G.
2011-05-09 16:41 ` [PATCH RFC 14/30] ext4: snapshot block operations amir73il
2011-05-09 16:41 ` [PATCH RFC 15/30] ext4: snapshot block operation - copy blocks to snapshot amir73il
2011-05-09 16:41 ` [PATCH RFC 16/30] ext4: snapshot block operation - move " amir73il
2011-05-09 16:41 ` [PATCH RFC 17/30] ext4: snapshot control amir73il
2011-05-09 16:41 ` [PATCH RFC 18/30] ext4: snapshot control - fix new snapshot amir73il
2011-05-09 16:41 ` [PATCH RFC 19/30] ext4: snapshot control - reserve disk space for snapshot amir73il
2011-05-09 16:41 ` [PATCH RFC 20/30] ext4: snapshot journaled - increase transaction credits amir73il
2011-05-09 16:41 ` [PATCH RFC 21/30] ext4: snapshot journaled - implement journal_release_buffer() amir73il
2011-05-09 16:41 ` [PATCH RFC 22/30] ext4: snapshot journaled - bypass to save credits amir73il
2011-05-09 16:41 ` [PATCH RFC 23/30] ext4: snapshot journaled - trace COW/buffer credits amir73il
2011-05-09 16:41 ` [PATCH RFC 24/30] ext4: snapshot list support amir73il
2011-05-09 16:41 ` [PATCH RFC 25/30] ext4: snapshot race conditions - concurrent COW operations amir73il
2011-05-09 16:41 ` [PATCH RFC 26/30] ext4: snapshot race conditions - tracked reads amir73il
2011-05-09 16:41 ` [PATCH RFC 27/30] ext4: snapshot exclude - the exclude bitmap amir73il
2011-05-09 16:41 ` [PATCH RFC 28/30] ext4: snapshot cleanup amir73il
2011-05-09 16:41 ` [PATCH RFC 29/30] ext4: snapshot cleanup - shrink deleted snapshots amir73il
2011-05-09 16:41 ` [PATCH RFC 30/30] ext4: snapshot rocompat - enable rw mount amir73il
2011-06-06 13:08 ` [PATCH RFC 00/30] Ext4 snapshots - core patches Lukas Czerner
2011-06-06 14:32 ` Amir G.
2011-06-06 15:31 ` Eric Sandeen
2011-06-06 16:05 ` Lukas Czerner
2011-06-06 20:40 ` Ted Ts'o
2011-06-07 13:59 ` Ric Wheeler
2011-06-07 15:37 ` Ted Ts'o
2011-06-06 16:33 ` Andreas Dilger
2011-06-06 16:42 ` Eric Sandeen
2011-06-06 19:58 ` Lukáš Czerner
2011-06-06 18:25 ` Amir G.
2011-06-06 20:55 ` Ted Ts'o
2011-06-07 5:17 ` Andreas Dilger
2011-06-07 5:58 ` Amir G.
2011-06-07 10:09 ` Lukas Czerner
2011-06-07 13:01 ` Amir G.
2011-06-07 13:50 ` Ric Wheeler
2011-06-07 14:39 ` Amir G.
2011-06-07 6:40 ` Amir G.
2011-06-07 15:26 ` Josef Bacik
2011-06-07 16:46 ` Amir G.
2011-06-07 16:54 ` Josef Bacik
2011-06-07 18:22 ` Amir G.
2011-06-07 17:14 ` Sunil Mushran
2011-06-07 17:30 ` Ted Ts'o
2011-06-07 17:54 ` Amir G.
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=1304959308-11122-9-git-send-email-amir73il@users.sourceforge.net \
--to=amir73il@users.sourceforge.net \
--cc=amir73il@users.sf.net \
--cc=linux-ext4@vger.kernel.org \
--cc=tytso@mit.edu \
--cc=xiaoqiangnk@gmail.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).