From: "Theodore Ts'o" <tytso@mit.edu>
To: Ext4 Developers List <linux-ext4@vger.kernel.org>,
Linux Kernel Developers List <linux-kernel@vger.kernel.org>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>,
Mingming Cao <cmm@us.ibm.com>, "Theodore Ts'o" <tytso@mit.edu>
Subject: [PATCH 24/52] ext4: Use inode preallocation with -o noextents
Date: Sat, 5 Jul 2008 13:35:50 -0400 [thread overview]
Message-ID: <1215279378-30504-25-git-send-email-tytso@mit.edu> (raw)
In-Reply-To: <1215279378-30504-24-git-send-email-tytso@mit.edu>
From: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
When mballoc is enabled, block allocation for old block-based
files are allocated using mballoc allocator instead of old
block-based allocator. The old ext3 block reservation is turned
off when mballoc is turned on.
However, the in-core preallocation is not enabled for block-based/
non-extent based file block allocation. This result in performance
regression, as now we don't have "reservation" ore in-core preallocation
to prevent interleaved fragmentation in multiple writes workload.
This patch fix this by enable per inode in-core preallocation
for non extent files when mballoc is used.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
---
fs/ext4/balloc.c | 36 +++++++++++++++++++++++++-
fs/ext4/ext4.h | 7 +++-
fs/ext4/extents.c | 2 +-
fs/ext4/inode.c | 72 +++++++++++++++++++++++++++++++++++++++-------------
fs/ext4/xattr.c | 2 +-
5 files changed, 95 insertions(+), 24 deletions(-)
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 1327ac3..816f1db 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -1923,7 +1923,7 @@ out:
return 0;
}
-ext4_fsblk_t ext4_new_block(handle_t *handle, struct inode *inode,
+ext4_fsblk_t ext4_new_meta_block(handle_t *handle, struct inode *inode,
ext4_fsblk_t goal, int *errp)
{
struct ext4_allocation_request ar;
@@ -1942,9 +1942,29 @@ ext4_fsblk_t ext4_new_block(handle_t *handle, struct inode *inode,
ret = ext4_mb_new_blocks(handle, &ar, errp);
return ret;
}
+ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
+ ext4_fsblk_t goal, unsigned long *count, int *errp)
+{
+ struct ext4_allocation_request ar;
+ ext4_fsblk_t ret;
+
+ if (!test_opt(inode->i_sb, MBALLOC)) {
+ ret = ext4_new_blocks_old(handle, inode, goal, count, errp);
+ return ret;
+ }
+
+ memset(&ar, 0, sizeof(ar));
+ ar.inode = inode;
+ ar.goal = goal;
+ ar.len = *count;
+ ret = ext4_mb_new_blocks(handle, &ar, errp);
+ *count = ar.len;
+ return ret;
+}
ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
- ext4_fsblk_t goal, unsigned long *count, int *errp)
+ ext4_lblk_t iblock, ext4_fsblk_t goal,
+ unsigned long *count, int *errp)
{
struct ext4_allocation_request ar;
ext4_fsblk_t ret;
@@ -1955,9 +1975,21 @@ ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
}
memset(&ar, 0, sizeof(ar));
+ /* Fill with neighbour allocated blocks */
+ ar.lleft = 0;
+ ar.pleft = 0;
+ ar.lright = 0;
+ ar.pright = 0;
+
ar.inode = inode;
ar.goal = goal;
ar.len = *count;
+ ar.logical = iblock;
+ if (S_ISREG(inode->i_mode))
+ ar.flags = EXT4_MB_HINT_DATA;
+ else
+ /* disable in-core preallocation for non-regular files */
+ ar.flags = 0;
ret = ext4_mb_new_blocks(handle, &ar, errp);
*count = ar.len;
return ret;
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index c4aa1ce..b3e62b7 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -970,10 +970,13 @@ extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
ext4_group_t group);
-extern ext4_fsblk_t ext4_new_block (handle_t *handle, struct inode *inode,
+extern ext4_fsblk_t ext4_new_meta_block(handle_t *handle, struct inode *inode,
ext4_fsblk_t goal, int *errp);
-extern ext4_fsblk_t ext4_new_blocks (handle_t *handle, struct inode *inode,
+extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
ext4_fsblk_t goal, unsigned long *count, int *errp);
+extern ext4_fsblk_t ext4_new_blocks(handle_t *handle, struct inode *inode,
+ ext4_lblk_t iblock, ext4_fsblk_t goal,
+ unsigned long *count, int *errp);
extern ext4_fsblk_t ext4_new_blocks_old(handle_t *handle, struct inode *inode,
ext4_fsblk_t goal, unsigned long *count, int *errp);
extern void ext4_free_blocks (handle_t *handle, struct inode *inode,
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 17ea8bb..c729b35 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -187,7 +187,7 @@ ext4_ext_new_block(handle_t *handle, struct inode *inode,
ext4_fsblk_t goal, newblock;
goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
- newblock = ext4_new_block(handle, inode, goal, err);
+ newblock = ext4_new_meta_block(handle, inode, goal, err);
return newblock;
}
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 8d97077..04059ba 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -508,11 +508,12 @@ static int ext4_blks_to_allocate(Indirect *branch, int k, unsigned long blks,
* direct blocks
*/
static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
- ext4_fsblk_t goal, int indirect_blks, int blks,
- ext4_fsblk_t new_blocks[4], int *err)
+ ext4_lblk_t iblock, ext4_fsblk_t goal,
+ int indirect_blks, int blks,
+ ext4_fsblk_t new_blocks[4], int *err)
{
int target, i;
- unsigned long count = 0;
+ unsigned long count = 0, blk_allocated = 0;
int index = 0;
ext4_fsblk_t current_block = 0;
int ret = 0;
@@ -525,12 +526,13 @@ static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
* the first direct block of this branch. That's the
* minimum number of blocks need to allocate(required)
*/
- target = blks + indirect_blks;
-
- while (1) {
+ /* first we try to allocate the indirect blocks */
+ target = indirect_blks;
+ while (target > 0) {
count = target;
/* allocating blocks for indirect blocks and direct blocks */
- current_block = ext4_new_blocks(handle,inode,goal,&count,err);
+ current_block = ext4_new_meta_blocks(handle, inode,
+ goal, &count, err);
if (*err)
goto failed_out;
@@ -540,16 +542,48 @@ static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
new_blocks[index++] = current_block++;
count--;
}
-
- if (count > 0)
+ if (count > 0) {
+ /*
+ * save the new block number
+ * for the first direct block
+ */
+ new_blocks[index] = current_block;
+ printk(KERN_INFO "%s returned more blocks than "
+ "requested\n", __func__);
+ WARN_ON(1);
break;
+ }
}
- /* save the new block number for the first direct block */
- new_blocks[index] = current_block;
-
+ target = blks - count ;
+ blk_allocated = count;
+ if (!target)
+ goto allocated;
+ /* Now allocate data blocks */
+ count = target;
+ /* allocating blocks for indirect blocks and direct blocks */
+ current_block = ext4_new_blocks(handle, inode, iblock,
+ goal, &count, err);
+ if (*err && (target == blks)) {
+ /*
+ * if the allocation failed and we didn't allocate
+ * any blocks before
+ */
+ goto failed_out;
+ }
+ if (!*err) {
+ if (target == blks) {
+ /*
+ * save the new block number
+ * for the first direct block
+ */
+ new_blocks[index] = current_block;
+ }
+ blk_allocated += count;
+ }
+allocated:
/* total number of blocks allocated for direct blocks */
- ret = count;
+ ret = blk_allocated;
*err = 0;
return ret;
failed_out:
@@ -584,8 +618,9 @@ failed_out:
* as described above and return 0.
*/
static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
- int indirect_blks, int *blks, ext4_fsblk_t goal,
- ext4_lblk_t *offsets, Indirect *branch)
+ ext4_lblk_t iblock, int indirect_blks,
+ int *blks, ext4_fsblk_t goal,
+ ext4_lblk_t *offsets, Indirect *branch)
{
int blocksize = inode->i_sb->s_blocksize;
int i, n = 0;
@@ -595,7 +630,7 @@ static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
ext4_fsblk_t new_blocks[4];
ext4_fsblk_t current_block;
- num = ext4_alloc_blocks(handle, inode, goal, indirect_blks,
+ num = ext4_alloc_blocks(handle, inode, iblock, goal, indirect_blks,
*blks, new_blocks, &err);
if (err)
return err;
@@ -855,8 +890,9 @@ int ext4_get_blocks_handle(handle_t *handle, struct inode *inode,
/*
* Block out ext4_truncate while we alter the tree
*/
- err = ext4_alloc_branch(handle, inode, indirect_blks, &count, goal,
- offsets + (partial - chain), partial);
+ err = ext4_alloc_branch(handle, inode, iblock, indirect_blks,
+ &count, goal,
+ offsets + (partial - chain), partial);
/*
* The ext4_splice_branch call will free and forget any buffers
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index ff08633..93c5fdc 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -810,7 +810,7 @@ inserted:
/* We need to allocate a new block */
ext4_fsblk_t goal = ext4_group_first_block_no(sb,
EXT4_I(inode)->i_block_group);
- ext4_fsblk_t block = ext4_new_block(handle, inode,
+ ext4_fsblk_t block = ext4_new_meta_block(handle, inode,
goal, &error);
if (error)
goto cleanup;
--
1.5.6.rc3.1.g36b7.dirty
next prev parent reply other threads:[~2008-07-05 17:46 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-07-05 17:35 Ext4 patches for the next merge window Theodore Ts'o
2008-07-05 17:35 ` [PATCH 01/52] ext4: fix comments to say "ext4" Theodore Ts'o
2008-07-05 17:35 ` [PATCH 02/52] ext4: start searching for the right extent from the goal group Theodore Ts'o
2008-07-05 17:35 ` [PATCH 03/52] ext4: Use BUG_ON() instead of BUG() Theodore Ts'o
2008-07-05 17:35 ` [PATCH 04/52] ext4: switch to seq_files Theodore Ts'o
2008-07-05 17:35 ` [PATCH 05/52] ext4: improve some code in rb tree part of dir.c Theodore Ts'o
2008-07-05 17:35 ` [PATCH 06/52] ext4: Fix ext4_mb_init_cache return error Theodore Ts'o
2008-07-05 17:35 ` [PATCH 07/52] ext4: add error processing when calling ext4_mb_init_cache in mballoc Theodore Ts'o
2008-07-05 17:35 ` [PATCH 08/52] ext4: miscellaneous error checks and coding cleanups for mballoc Theodore Ts'o
2008-07-05 17:35 ` [PATCH 09/52] ext4: remove double definitions of xattr macros Theodore Ts'o
2008-07-05 17:35 ` [PATCH 10/52] ext4: Rename read_block_bitmap() to ext4_read_block_bitmap() Theodore Ts'o
2008-07-05 17:35 ` [PATCH 11/52] ext4: Remove unused variable from ext4_show_options Theodore Ts'o
2008-07-05 17:35 ` [PATCH 12/52] jbd2: Add commit time into the commit block Theodore Ts'o
2008-07-05 17:35 ` [PATCH 13/52] ext4: New inode allocation for FLEX_BG meta-data groups Theodore Ts'o
2008-07-05 17:35 ` [PATCH 14/52] jbd2: fix race between jbd2_journal_try_to_free_buffers() and jbd2 commit transaction Theodore Ts'o
2008-07-05 17:35 ` [PATCH 15/52] ext4: remove redundant code in ext4_fill_super() Theodore Ts'o
2008-07-05 17:35 ` [PATCH 16/52] ext4: remove quota allocation when ext4_mb_new_blocks fails Theodore Ts'o
2008-07-05 17:35 ` [PATCH 17/52] ext4: Update i_disksize properly when allocating from fallocate area Theodore Ts'o
2008-07-05 17:35 ` [PATCH 18/52] ext4: return error when calling ext4_ext_split failed Theodore Ts'o
2008-07-05 17:35 ` [PATCH 19/52] ext4: Make ext4_ext_find_extent fills ext_path completely Theodore Ts'o
2008-07-05 17:35 ` [PATCH 20/52] ext4: Fix ext4_ext_journal_restart() to reflect errors up to the caller Theodore Ts'o
2008-07-05 17:35 ` [PATCH 21/52] ext4: cleanup never-used magic numbers from htree code Theodore Ts'o
2008-07-05 17:35 ` [PATCH 22/52] ext4: Fix sparse warning Theodore Ts'o
2008-07-05 17:35 ` [PATCH 23/52] ext4: fix ext4_init_block_bitmap() for metablock block group Theodore Ts'o
2008-07-05 17:35 ` Theodore Ts'o [this message]
2008-07-05 17:35 ` [PATCH 25/52] ext4: cleanup block allocator Theodore Ts'o
2008-07-05 17:35 ` [PATCH 26/52] ext4: call blkdev_issue_flush on fsync Theodore Ts'o
2008-07-05 17:35 ` [PATCH 27/52] ext4: mballoc avoid use root reserved blocks for non root allocation Theodore Ts'o
2008-07-05 17:35 ` [PATCH 28/52] ext4: Set journal pointer to NULL when journal is released Theodore Ts'o
2008-07-05 17:35 ` [PATCH 29/52] ext4: use atomic functions to set bh_state Theodore Ts'o
2008-07-05 17:35 ` [PATCH 30/52] ext4: Add missing unlock to an error path in ext4_quota_write() Theodore Ts'o
2008-07-05 17:35 ` [PATCH 31/52] ext4: fix online resize with mballoc Theodore Ts'o
2008-07-05 17:35 ` [PATCH 32/52] ext4: Documentation updates Theodore Ts'o
2008-07-05 17:35 ` [PATCH 33/52] ext4: Use page_mkwrite vma_operations to get mmap write notification Theodore Ts'o
2008-07-05 17:36 ` [PATCH 34/52] vfs: Move mark_inode_dirty() from under page lock in generic_write_end() Theodore Ts'o
2008-07-05 17:36 ` [PATCH 35/52] ext4: Invert the locking order of page_lock and transaction start Theodore Ts'o
2008-07-05 17:36 ` [PATCH 36/52] ext4: Fix lock inversion in ext4_ext_truncate() Theodore Ts'o
2008-07-05 17:36 ` [PATCH 37/52] vfs: export filemap_fdatawrite_range() Theodore Ts'o
2008-07-05 17:36 ` [PATCH 38/52] jbd2: Implement data=ordered mode handling via inodes Theodore Ts'o
2008-07-05 17:36 ` [PATCH 39/52] ext4: Use new framework for data=ordered mode in JBD2 Theodore Ts'o
2008-07-05 17:36 ` [PATCH 40/52] jbd2: Remove data=ordered mode support using jbd buffer heads Theodore Ts'o
2008-07-05 17:36 ` [PATCH 41/52] vfs: add basic delayed allocation support Theodore Ts'o
2008-07-05 17:36 ` [PATCH 42/52] ext4: Add " Theodore Ts'o
2008-07-05 17:36 ` [PATCH 43/52] percpu_counter: new function percpu_counter_sum_and_set Theodore Ts'o
2008-07-05 17:36 ` [PATCH 44/52] ext4: delayed allocation ENOSPC handling Theodore Ts'o
2008-07-05 17:36 ` [PATCH 45/52] mm: Add range_cont mode for writeback Theodore Ts'o
2008-07-05 17:36 ` [PATCH 46/52] ext4: Invert lock ordering of page_lock and transaction start in delalloc Theodore Ts'o
2008-07-05 17:36 ` [PATCH 47/52] ext4: Add ordered mode support for delalloc Theodore Ts'o
2008-07-05 17:36 ` [PATCH 48/52] ext4: Handle page without buffers in ext4_*_writepage() Theodore Ts'o
2008-07-05 17:36 ` [PATCH 49/52] ext4: fix delalloc i_disksize early update issue Theodore Ts'o
2008-07-05 17:36 ` [PATCH 50/52] ext4: Enable delalloc by default Theodore Ts'o
2008-07-05 17:36 ` [PATCH 51/52] ext4: Don't allow nonextenst mount option for large filesystem Theodore Ts'o
2008-07-05 17:36 ` [PATCH 52/52] ext4: Documention update for new ordered mode and delayed allocation Theodore Ts'o
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=1215279378-30504-25-git-send-email-tytso@mit.edu \
--to=tytso@mit.edu \
--cc=aneesh.kumar@linux.vnet.ibm.com \
--cc=cmm@us.ibm.com \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
/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
all inboxes | Powered by JetHome®