2019-05-24 03:04:05 -07:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
2021-05-06 18:06:44 -07:00
|
|
|
/*
|
2007-12-18 00:47:03 -07:00
|
|
|
* resize.c
|
|
|
|
|
*
|
|
|
|
|
* volume resize.
|
|
|
|
|
* Inspired by ext3/resize.c.
|
|
|
|
|
*
|
|
|
|
|
* Copyright (C) 2007 Oracle. All rights reserved.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <linux/fs.h>
|
|
|
|
|
#include <linux/types.h>
|
|
|
|
|
|
|
|
|
|
#include <cluster/masklog.h>
|
|
|
|
|
|
|
|
|
|
#include "ocfs2.h"
|
|
|
|
|
|
|
|
|
|
#include "alloc.h"
|
|
|
|
|
#include "dlmglue.h"
|
|
|
|
|
#include "inode.h"
|
|
|
|
|
#include "journal.h"
|
|
|
|
|
#include "super.h"
|
|
|
|
|
#include "sysfile.h"
|
|
|
|
|
#include "uptodate.h"
|
2011-02-21 17:24:01 -07:00
|
|
|
#include "ocfs2_trace.h"
|
2007-12-18 00:47:03 -07:00
|
|
|
|
|
|
|
|
#include "buffer_head_io.h"
|
|
|
|
|
#include "suballoc.h"
|
|
|
|
|
#include "resize.h"
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Check whether there are new backup superblocks exist
|
|
|
|
|
* in the last group. If there are some, mark them or clear
|
|
|
|
|
* them in the bitmap.
|
|
|
|
|
*
|
|
|
|
|
* Return how many backups we find in the last group.
|
|
|
|
|
*/
|
|
|
|
|
static u16 ocfs2_calc_new_backup_super(struct inode *inode,
|
|
|
|
|
struct ocfs2_group_desc *gd,
|
|
|
|
|
u16 cl_cpg,
|
2015-12-29 15:54:06 -07:00
|
|
|
u16 old_bg_clusters,
|
2007-12-18 00:47:03 -07:00
|
|
|
int set)
|
|
|
|
|
{
|
|
|
|
|
int i;
|
|
|
|
|
u16 backups = 0;
|
2015-12-29 15:54:06 -07:00
|
|
|
u32 cluster, lgd_cluster;
|
2007-12-18 00:47:03 -07:00
|
|
|
u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno);
|
|
|
|
|
|
|
|
|
|
for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
|
|
|
|
|
blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
|
|
|
|
|
cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
|
|
|
|
|
|
|
|
|
|
gd_blkno = ocfs2_which_cluster_group(inode, cluster);
|
|
|
|
|
if (gd_blkno < lgd_blkno)
|
|
|
|
|
continue;
|
|
|
|
|
else if (gd_blkno > lgd_blkno)
|
|
|
|
|
break;
|
|
|
|
|
|
2015-12-29 15:54:06 -07:00
|
|
|
/* check if already done backup super */
|
|
|
|
|
lgd_cluster = ocfs2_blocks_to_clusters(inode->i_sb, lgd_blkno);
|
|
|
|
|
lgd_cluster += old_bg_clusters;
|
|
|
|
|
if (lgd_cluster >= cluster)
|
|
|
|
|
continue;
|
|
|
|
|
|
2007-12-18 00:47:03 -07:00
|
|
|
if (set)
|
|
|
|
|
ocfs2_set_bit(cluster % cl_cpg,
|
|
|
|
|
(unsigned long *)gd->bg_bitmap);
|
|
|
|
|
else
|
|
|
|
|
ocfs2_clear_bit(cluster % cl_cpg,
|
|
|
|
|
(unsigned long *)gd->bg_bitmap);
|
|
|
|
|
backups++;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return backups;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int ocfs2_update_last_group_and_inode(handle_t *handle,
|
|
|
|
|
struct inode *bm_inode,
|
|
|
|
|
struct buffer_head *bm_bh,
|
|
|
|
|
struct buffer_head *group_bh,
|
|
|
|
|
u32 first_new_cluster,
|
|
|
|
|
int new_clusters)
|
|
|
|
|
{
|
|
|
|
|
int ret = 0;
|
|
|
|
|
struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb);
|
|
|
|
|
struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data;
|
|
|
|
|
struct ocfs2_chain_list *cl = &fe->id2.i_chain;
|
|
|
|
|
struct ocfs2_chain_rec *cr;
|
|
|
|
|
struct ocfs2_group_desc *group;
|
|
|
|
|
u16 chain, num_bits, backups = 0;
|
|
|
|
|
u16 cl_bpc = le16_to_cpu(cl->cl_bpc);
|
|
|
|
|
u16 cl_cpg = le16_to_cpu(cl->cl_cpg);
|
2015-12-29 15:54:06 -07:00
|
|
|
u16 old_bg_clusters;
|
ocfs2: improve write IO performance when fragmentation is high
The group_search function ocfs2_cluster_group_search() should
bypass groups with insufficient space to avoid unnecessary
searches.
This patch is particularly useful when ocfs2 is handling huge
number small files, and volume fragmentation is very high.
In this case, ocfs2 is busy with looking up available la window
from //global_bitmap.
This patch introduces a new member in the Group Description (gd)
struct called 'bg_contig_free_bits', representing the max
contigous free bits in this gd. When ocfs2 allocates a new
la window from //global_bitmap, 'bg_contig_free_bits' helps
expedite the search process.
Let's image below path.
1. la state (->local_alloc_state) is set THROTTLED or DISABLED.
2. when user delete a large file and trigger
ocfs2_local_alloc_seen_free_bits set osb->local_alloc_state
unconditionally.
3. a write IOs thread run and trigger the worst performance path
```
ocfs2_reserve_clusters_with_limit
ocfs2_reserve_local_alloc_bits
ocfs2_local_alloc_slide_window //[1]
+ ocfs2_local_alloc_reserve_for_window //[2]
+ ocfs2_local_alloc_new_window //[3]
ocfs2_recalc_la_window
```
[1]:
will be called when la window bits used up.
[2]:
under la state is ENABLED, and this func only check global_bitmap
free bits, it will succeed in general.
[3]:
will use the default la window size to search clusters then fail.
ocfs2_recalc_la_window attempts other la window sizes.
the timing complexity is O(n^4), resulting in a significant time
cost for scanning global bitmap. This leads to a dramatic slowdown
in write I/Os (e.g., user space 'dd').
i.e.
an ocfs2 partition size: 1.45TB, cluster size: 4KB,
la window default size: 106MB.
The partition is fragmentation by creating & deleting huge mount of
small files.
before this patch, the timing of [3] should be
(the number got from real world):
- la window size change order (size: MB):
106, 53, 26.5, 13, 6.5, 3.25, 1.6, 0.8
only 0.8MB succeed, 0.8MB also triggers la window to disable.
ocfs2_local_alloc_new_window retries 8 times, first 7 times totally
runs in worst case.
- group chain number: 242
ocfs2_claim_suballoc_bits calls for-loop 242 times
- each chain has 49 block group
ocfs2_search_chain calls while-loop 49 times
- each bg has 32256 blocks
ocfs2_block_group_find_clear_bits calls while-loop for 32256 bits.
for ocfs2_find_next_zero_bit uses ffz() to find zero bit, let's use
(32256/64) (this is not worst value) for timing calucation.
the loop times: 7*242*49*(32256/64) = 41835024 (~42 million times)
In the worst case, user space writes 1MB data will trigger 42M scanning
times.
under this patch, the timing is '7*242*49 = 83006', reduced by three
orders of magnitude.
Link: https://lkml.kernel.org/r/20240328125203.20892-2-heming.zhao@suse.com
Signed-off-by: Heming Zhao <heming.zhao@suse.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Jun Piao <piaojun@huawei.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Mark Fasheh <mark@fasheh.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2024-03-28 05:52:00 -07:00
|
|
|
u16 contig_bits;
|
|
|
|
|
__le16 old_bg_contig_free_bits;
|
2007-12-18 00:47:03 -07:00
|
|
|
|
2011-02-21 17:24:01 -07:00
|
|
|
trace_ocfs2_update_last_group_and_inode(new_clusters,
|
|
|
|
|
first_new_cluster);
|
2007-12-18 00:47:03 -07:00
|
|
|
|
2009-02-12 17:41:25 -07:00
|
|
|
ret = ocfs2_journal_access_gd(handle, INODE_CACHE(bm_inode),
|
|
|
|
|
group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
group = (struct ocfs2_group_desc *)group_bh->b_data;
|
|
|
|
|
|
2015-12-29 15:54:06 -07:00
|
|
|
old_bg_clusters = le16_to_cpu(group->bg_bits) / cl_bpc;
|
2007-12-18 00:47:03 -07:00
|
|
|
/* update the group first. */
|
|
|
|
|
num_bits = new_clusters * cl_bpc;
|
|
|
|
|
le16_add_cpu(&group->bg_bits, num_bits);
|
|
|
|
|
le16_add_cpu(&group->bg_free_bits_count, num_bits);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* check whether there are some new backup superblocks exist in
|
|
|
|
|
* this group and update the group bitmap accordingly.
|
|
|
|
|
*/
|
|
|
|
|
if (OCFS2_HAS_COMPAT_FEATURE(osb->sb,
|
|
|
|
|
OCFS2_FEATURE_COMPAT_BACKUP_SB)) {
|
|
|
|
|
backups = ocfs2_calc_new_backup_super(bm_inode,
|
|
|
|
|
group,
|
2015-12-29 15:54:06 -07:00
|
|
|
cl_cpg, old_bg_clusters, 1);
|
2007-12-18 00:47:03 -07:00
|
|
|
le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
|
|
|
|
|
}
|
|
|
|
|
|
ocfs2: improve write IO performance when fragmentation is high
The group_search function ocfs2_cluster_group_search() should
bypass groups with insufficient space to avoid unnecessary
searches.
This patch is particularly useful when ocfs2 is handling huge
number small files, and volume fragmentation is very high.
In this case, ocfs2 is busy with looking up available la window
from //global_bitmap.
This patch introduces a new member in the Group Description (gd)
struct called 'bg_contig_free_bits', representing the max
contigous free bits in this gd. When ocfs2 allocates a new
la window from //global_bitmap, 'bg_contig_free_bits' helps
expedite the search process.
Let's image below path.
1. la state (->local_alloc_state) is set THROTTLED or DISABLED.
2. when user delete a large file and trigger
ocfs2_local_alloc_seen_free_bits set osb->local_alloc_state
unconditionally.
3. a write IOs thread run and trigger the worst performance path
```
ocfs2_reserve_clusters_with_limit
ocfs2_reserve_local_alloc_bits
ocfs2_local_alloc_slide_window //[1]
+ ocfs2_local_alloc_reserve_for_window //[2]
+ ocfs2_local_alloc_new_window //[3]
ocfs2_recalc_la_window
```
[1]:
will be called when la window bits used up.
[2]:
under la state is ENABLED, and this func only check global_bitmap
free bits, it will succeed in general.
[3]:
will use the default la window size to search clusters then fail.
ocfs2_recalc_la_window attempts other la window sizes.
the timing complexity is O(n^4), resulting in a significant time
cost for scanning global bitmap. This leads to a dramatic slowdown
in write I/Os (e.g., user space 'dd').
i.e.
an ocfs2 partition size: 1.45TB, cluster size: 4KB,
la window default size: 106MB.
The partition is fragmentation by creating & deleting huge mount of
small files.
before this patch, the timing of [3] should be
(the number got from real world):
- la window size change order (size: MB):
106, 53, 26.5, 13, 6.5, 3.25, 1.6, 0.8
only 0.8MB succeed, 0.8MB also triggers la window to disable.
ocfs2_local_alloc_new_window retries 8 times, first 7 times totally
runs in worst case.
- group chain number: 242
ocfs2_claim_suballoc_bits calls for-loop 242 times
- each chain has 49 block group
ocfs2_search_chain calls while-loop 49 times
- each bg has 32256 blocks
ocfs2_block_group_find_clear_bits calls while-loop for 32256 bits.
for ocfs2_find_next_zero_bit uses ffz() to find zero bit, let's use
(32256/64) (this is not worst value) for timing calucation.
the loop times: 7*242*49*(32256/64) = 41835024 (~42 million times)
In the worst case, user space writes 1MB data will trigger 42M scanning
times.
under this patch, the timing is '7*242*49 = 83006', reduced by three
orders of magnitude.
Link: https://lkml.kernel.org/r/20240328125203.20892-2-heming.zhao@suse.com
Signed-off-by: Heming Zhao <heming.zhao@suse.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Jun Piao <piaojun@huawei.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Mark Fasheh <mark@fasheh.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2024-03-28 05:52:00 -07:00
|
|
|
contig_bits = ocfs2_find_max_contig_free_bits(group->bg_bitmap,
|
|
|
|
|
le16_to_cpu(group->bg_bits), 0);
|
|
|
|
|
old_bg_contig_free_bits = group->bg_contig_free_bits;
|
|
|
|
|
group->bg_contig_free_bits = cpu_to_le16(contig_bits);
|
|
|
|
|
|
2010-03-19 14:13:52 -07:00
|
|
|
ocfs2_journal_dirty(handle, group_bh);
|
2007-12-18 00:47:03 -07:00
|
|
|
|
|
|
|
|
/* update the inode accordingly. */
|
2009-02-12 17:41:25 -07:00
|
|
|
ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
|
2008-10-17 19:25:01 -07:00
|
|
|
OCFS2_JOURNAL_ACCESS_WRITE);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_rollback;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
chain = le16_to_cpu(group->bg_chain);
|
|
|
|
|
cr = (&cl->cl_recs[chain]);
|
|
|
|
|
le32_add_cpu(&cr->c_total, num_bits);
|
|
|
|
|
le32_add_cpu(&cr->c_free, num_bits);
|
|
|
|
|
le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
|
|
|
|
|
le32_add_cpu(&fe->i_clusters, new_clusters);
|
|
|
|
|
|
|
|
|
|
if (backups) {
|
|
|
|
|
le32_add_cpu(&cr->c_free, -1 * backups);
|
|
|
|
|
le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
spin_lock(&OCFS2_I(bm_inode)->ip_lock);
|
|
|
|
|
OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
|
2014-06-04 16:06:17 -07:00
|
|
|
le64_add_cpu(&fe->i_size, (u64)new_clusters << osb->s_clustersize_bits);
|
2007-12-18 00:47:03 -07:00
|
|
|
spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
|
|
|
|
|
i_size_write(bm_inode, le64_to_cpu(fe->i_size));
|
|
|
|
|
|
|
|
|
|
ocfs2_journal_dirty(handle, bm_bh);
|
|
|
|
|
|
|
|
|
|
out_rollback:
|
|
|
|
|
if (ret < 0) {
|
|
|
|
|
ocfs2_calc_new_backup_super(bm_inode,
|
|
|
|
|
group,
|
2015-12-29 15:54:06 -07:00
|
|
|
cl_cpg, old_bg_clusters, 0);
|
2007-12-18 00:47:03 -07:00
|
|
|
le16_add_cpu(&group->bg_free_bits_count, backups);
|
|
|
|
|
le16_add_cpu(&group->bg_bits, -1 * num_bits);
|
|
|
|
|
le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
|
ocfs2: improve write IO performance when fragmentation is high
The group_search function ocfs2_cluster_group_search() should
bypass groups with insufficient space to avoid unnecessary
searches.
This patch is particularly useful when ocfs2 is handling huge
number small files, and volume fragmentation is very high.
In this case, ocfs2 is busy with looking up available la window
from //global_bitmap.
This patch introduces a new member in the Group Description (gd)
struct called 'bg_contig_free_bits', representing the max
contigous free bits in this gd. When ocfs2 allocates a new
la window from //global_bitmap, 'bg_contig_free_bits' helps
expedite the search process.
Let's image below path.
1. la state (->local_alloc_state) is set THROTTLED or DISABLED.
2. when user delete a large file and trigger
ocfs2_local_alloc_seen_free_bits set osb->local_alloc_state
unconditionally.
3. a write IOs thread run and trigger the worst performance path
```
ocfs2_reserve_clusters_with_limit
ocfs2_reserve_local_alloc_bits
ocfs2_local_alloc_slide_window //[1]
+ ocfs2_local_alloc_reserve_for_window //[2]
+ ocfs2_local_alloc_new_window //[3]
ocfs2_recalc_la_window
```
[1]:
will be called when la window bits used up.
[2]:
under la state is ENABLED, and this func only check global_bitmap
free bits, it will succeed in general.
[3]:
will use the default la window size to search clusters then fail.
ocfs2_recalc_la_window attempts other la window sizes.
the timing complexity is O(n^4), resulting in a significant time
cost for scanning global bitmap. This leads to a dramatic slowdown
in write I/Os (e.g., user space 'dd').
i.e.
an ocfs2 partition size: 1.45TB, cluster size: 4KB,
la window default size: 106MB.
The partition is fragmentation by creating & deleting huge mount of
small files.
before this patch, the timing of [3] should be
(the number got from real world):
- la window size change order (size: MB):
106, 53, 26.5, 13, 6.5, 3.25, 1.6, 0.8
only 0.8MB succeed, 0.8MB also triggers la window to disable.
ocfs2_local_alloc_new_window retries 8 times, first 7 times totally
runs in worst case.
- group chain number: 242
ocfs2_claim_suballoc_bits calls for-loop 242 times
- each chain has 49 block group
ocfs2_search_chain calls while-loop 49 times
- each bg has 32256 blocks
ocfs2_block_group_find_clear_bits calls while-loop for 32256 bits.
for ocfs2_find_next_zero_bit uses ffz() to find zero bit, let's use
(32256/64) (this is not worst value) for timing calucation.
the loop times: 7*242*49*(32256/64) = 41835024 (~42 million times)
In the worst case, user space writes 1MB data will trigger 42M scanning
times.
under this patch, the timing is '7*242*49 = 83006', reduced by three
orders of magnitude.
Link: https://lkml.kernel.org/r/20240328125203.20892-2-heming.zhao@suse.com
Signed-off-by: Heming Zhao <heming.zhao@suse.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Jun Piao <piaojun@huawei.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Mark Fasheh <mark@fasheh.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2024-03-28 05:52:00 -07:00
|
|
|
group->bg_contig_free_bits = old_bg_contig_free_bits;
|
2007-12-18 00:47:03 -07:00
|
|
|
}
|
|
|
|
|
out:
|
2011-03-07 01:43:21 -07:00
|
|
|
if (ret)
|
|
|
|
|
mlog_errno(ret);
|
2007-12-18 00:47:03 -07:00
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int update_backups(struct inode * inode, u32 clusters, char *data)
|
|
|
|
|
{
|
|
|
|
|
int i, ret = 0;
|
|
|
|
|
u32 cluster;
|
|
|
|
|
u64 blkno;
|
|
|
|
|
struct buffer_head *backup = NULL;
|
|
|
|
|
struct ocfs2_dinode *backup_di = NULL;
|
|
|
|
|
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
|
|
|
|
|
|
|
|
|
|
/* calculate the real backups we need to update. */
|
|
|
|
|
for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
|
|
|
|
|
blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
|
|
|
|
|
cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
|
2016-03-25 14:21:35 -07:00
|
|
|
if (cluster >= clusters)
|
2007-12-18 00:47:03 -07:00
|
|
|
break;
|
|
|
|
|
|
2008-10-09 17:20:29 -07:00
|
|
|
ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
|
|
|
|
|
|
|
|
|
|
backup_di = (struct ocfs2_dinode *)backup->b_data;
|
|
|
|
|
backup_di->i_blkno = cpu_to_le64(blkno);
|
|
|
|
|
|
|
|
|
|
ret = ocfs2_write_super_or_backup(osb, backup);
|
|
|
|
|
brelse(backup);
|
|
|
|
|
backup = NULL;
|
|
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void ocfs2_update_super_and_backups(struct inode *inode,
|
|
|
|
|
int new_clusters)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
|
|
|
|
u32 clusters = 0;
|
|
|
|
|
struct buffer_head *super_bh = NULL;
|
|
|
|
|
struct ocfs2_dinode *super_di = NULL;
|
|
|
|
|
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* update the superblock last.
|
|
|
|
|
* It doesn't matter if the write failed.
|
|
|
|
|
*/
|
2008-10-09 17:20:29 -07:00
|
|
|
ret = ocfs2_read_blocks_sync(osb, OCFS2_SUPER_BLOCK_BLKNO, 1,
|
|
|
|
|
&super_bh);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
super_di = (struct ocfs2_dinode *)super_bh->b_data;
|
|
|
|
|
le32_add_cpu(&super_di->i_clusters, new_clusters);
|
|
|
|
|
clusters = le32_to_cpu(super_di->i_clusters);
|
|
|
|
|
|
|
|
|
|
ret = ocfs2_write_super_or_backup(osb, super_bh);
|
|
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
|
|
|
|
|
ret = update_backups(inode, clusters, super_bh->b_data);
|
|
|
|
|
|
|
|
|
|
out:
|
2008-01-23 19:35:31 -07:00
|
|
|
brelse(super_bh);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret)
|
|
|
|
|
printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
|
|
|
|
|
" during fs resize. This condition is not fatal,"
|
|
|
|
|
" but fsck.ocfs2 should be run to fix it\n",
|
|
|
|
|
osb->dev_str);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Extend the filesystem to the new number of clusters specified. This entry
|
|
|
|
|
* point is only used to extend the current filesystem to the end of the last
|
|
|
|
|
* existing group.
|
|
|
|
|
*/
|
|
|
|
|
int ocfs2_group_extend(struct inode * inode, int new_clusters)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
|
|
|
|
handle_t *handle;
|
|
|
|
|
struct buffer_head *main_bm_bh = NULL;
|
|
|
|
|
struct buffer_head *group_bh = NULL;
|
|
|
|
|
struct inode *main_bm_inode = NULL;
|
|
|
|
|
struct ocfs2_dinode *fe = NULL;
|
|
|
|
|
struct ocfs2_group_desc *group = NULL;
|
|
|
|
|
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
|
|
|
|
|
u16 cl_bpc;
|
|
|
|
|
u32 first_new_cluster;
|
|
|
|
|
u64 lgd_blkno;
|
|
|
|
|
|
|
|
|
|
if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
|
|
|
|
|
return -EROFS;
|
|
|
|
|
|
|
|
|
|
if (new_clusters < 0)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
else if (new_clusters == 0)
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
main_bm_inode = ocfs2_get_system_file_inode(osb,
|
|
|
|
|
GLOBAL_BITMAP_SYSTEM_INODE,
|
|
|
|
|
OCFS2_INVALID_SLOT);
|
|
|
|
|
if (!main_bm_inode) {
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
2016-01-22 13:40:57 -07:00
|
|
|
inode_lock(main_bm_inode);
|
2007-12-18 00:47:03 -07:00
|
|
|
|
|
|
|
|
ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
|
|
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_mutex;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
|
|
|
|
|
|
2008-11-13 15:49:12 -07:00
|
|
|
/* main_bm_bh is validated by inode read inside ocfs2_inode_lock(),
|
|
|
|
|
* so any corruption is a code bug. */
|
|
|
|
|
BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
|
|
|
|
|
|
2007-12-18 00:47:03 -07:00
|
|
|
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
|
2010-04-12 23:38:06 -07:00
|
|
|
ocfs2_group_bitmap_size(osb->sb, 0,
|
|
|
|
|
osb->s_feature_incompat) * 8) {
|
2007-12-18 00:47:03 -07:00
|
|
|
mlog(ML_ERROR, "The disk is too old and small. "
|
|
|
|
|
"Force to do offline resize.");
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
first_new_cluster = le32_to_cpu(fe->i_clusters);
|
|
|
|
|
lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
|
|
|
|
|
first_new_cluster - 1);
|
|
|
|
|
|
2008-11-13 15:49:14 -07:00
|
|
|
ret = ocfs2_read_group_descriptor(main_bm_inode, fe, lgd_blkno,
|
|
|
|
|
&group_bh);
|
2007-12-18 00:47:03 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
group = (struct ocfs2_group_desc *)group_bh->b_data;
|
|
|
|
|
|
|
|
|
|
cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
|
|
|
|
|
if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
|
|
|
|
|
le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
|
2011-02-21 17:24:01 -07:00
|
|
|
|
|
|
|
|
trace_ocfs2_group_extend(
|
2007-12-18 00:47:25 -07:00
|
|
|
(unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
|
2007-12-18 00:47:03 -07:00
|
|
|
|
|
|
|
|
handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
|
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
|
mlog_errno(PTR_ERR(handle));
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* update the last group descriptor and inode. */
|
|
|
|
|
ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
|
|
|
|
|
main_bm_bh, group_bh,
|
|
|
|
|
first_new_cluster,
|
|
|
|
|
new_clusters);
|
|
|
|
|
if (ret) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_commit;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
|
|
|
|
|
|
|
|
|
|
out_commit:
|
|
|
|
|
ocfs2_commit_trans(osb, handle);
|
|
|
|
|
out_unlock:
|
2008-01-23 19:35:31 -07:00
|
|
|
brelse(group_bh);
|
|
|
|
|
brelse(main_bm_bh);
|
2007-12-18 00:47:03 -07:00
|
|
|
|
|
|
|
|
ocfs2_inode_unlock(main_bm_inode, 1);
|
|
|
|
|
|
|
|
|
|
out_mutex:
|
2016-01-22 13:40:57 -07:00
|
|
|
inode_unlock(main_bm_inode);
|
2007-12-18 00:47:03 -07:00
|
|
|
iput(main_bm_inode);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
static int ocfs2_check_new_group(struct inode *inode,
|
|
|
|
|
struct ocfs2_dinode *di,
|
|
|
|
|
struct ocfs2_new_group_input *input,
|
|
|
|
|
struct buffer_head *group_bh)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
2008-11-13 15:49:13 -07:00
|
|
|
struct ocfs2_group_desc *gd =
|
|
|
|
|
(struct ocfs2_group_desc *)group_bh->b_data;
|
2007-12-18 00:47:25 -07:00
|
|
|
u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
|
|
|
|
|
|
2008-11-13 15:49:19 -07:00
|
|
|
ret = ocfs2_check_group_descriptor(inode->i_sb, di, group_bh);
|
2008-11-13 15:49:13 -07:00
|
|
|
if (ret)
|
|
|
|
|
goto out;
|
2007-12-18 00:47:25 -07:00
|
|
|
|
2008-11-13 15:49:13 -07:00
|
|
|
ret = -EINVAL;
|
|
|
|
|
if (le16_to_cpu(gd->bg_chain) != input->chain)
|
2007-12-18 00:47:25 -07:00
|
|
|
mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
|
|
|
|
|
"while input has %u set.\n",
|
|
|
|
|
(unsigned long long)le64_to_cpu(gd->bg_blkno),
|
|
|
|
|
le16_to_cpu(gd->bg_chain), input->chain);
|
|
|
|
|
else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
|
|
|
|
|
mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
|
|
|
|
|
"input has %u clusters set\n",
|
|
|
|
|
(unsigned long long)le64_to_cpu(gd->bg_blkno),
|
|
|
|
|
le16_to_cpu(gd->bg_bits), input->clusters);
|
|
|
|
|
else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
|
|
|
|
|
mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
|
|
|
|
|
"but it should have %u set\n",
|
|
|
|
|
(unsigned long long)le64_to_cpu(gd->bg_blkno),
|
|
|
|
|
le16_to_cpu(gd->bg_bits),
|
|
|
|
|
input->frees * cl_bpc);
|
|
|
|
|
else
|
|
|
|
|
ret = 0;
|
|
|
|
|
|
2008-11-13 15:49:13 -07:00
|
|
|
out:
|
2007-12-18 00:47:25 -07:00
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int ocfs2_verify_group_and_input(struct inode *inode,
|
|
|
|
|
struct ocfs2_dinode *di,
|
|
|
|
|
struct ocfs2_new_group_input *input,
|
|
|
|
|
struct buffer_head *group_bh)
|
|
|
|
|
{
|
|
|
|
|
u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
|
|
|
|
|
u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
|
|
|
|
|
u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
|
|
|
|
|
u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
|
|
|
|
|
u32 total_clusters = le32_to_cpu(di->i_clusters);
|
|
|
|
|
int ret = -EINVAL;
|
|
|
|
|
|
|
|
|
|
if (cluster < total_clusters)
|
|
|
|
|
mlog(ML_ERROR, "add a group which is in the current volume.\n");
|
|
|
|
|
else if (input->chain >= cl_count)
|
|
|
|
|
mlog(ML_ERROR, "input chain exceeds the limit.\n");
|
|
|
|
|
else if (next_free != cl_count && next_free != input->chain)
|
|
|
|
|
mlog(ML_ERROR,
|
|
|
|
|
"the add group should be in chain %u\n", next_free);
|
|
|
|
|
else if (total_clusters + input->clusters < total_clusters)
|
|
|
|
|
mlog(ML_ERROR, "add group's clusters overflow.\n");
|
|
|
|
|
else if (input->clusters > cl_cpg)
|
|
|
|
|
mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
|
|
|
|
|
else if (input->frees > input->clusters)
|
|
|
|
|
mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
|
|
|
|
|
else if (total_clusters % cl_cpg != 0)
|
|
|
|
|
mlog(ML_ERROR,
|
|
|
|
|
"the last group isn't full. Use group extend first.\n");
|
|
|
|
|
else if (input->group != ocfs2_which_cluster_group(inode, cluster))
|
|
|
|
|
mlog(ML_ERROR, "group blkno is invalid\n");
|
|
|
|
|
else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
|
|
|
|
|
mlog(ML_ERROR, "group descriptor check failed.\n");
|
|
|
|
|
else
|
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Add a new group descriptor to global_bitmap. */
|
|
|
|
|
int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
|
|
|
|
|
{
|
|
|
|
|
int ret;
|
|
|
|
|
handle_t *handle;
|
|
|
|
|
struct buffer_head *main_bm_bh = NULL;
|
|
|
|
|
struct inode *main_bm_inode = NULL;
|
|
|
|
|
struct ocfs2_dinode *fe = NULL;
|
|
|
|
|
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
|
|
|
|
|
struct buffer_head *group_bh = NULL;
|
|
|
|
|
struct ocfs2_group_desc *group = NULL;
|
|
|
|
|
struct ocfs2_chain_list *cl;
|
|
|
|
|
struct ocfs2_chain_rec *cr;
|
|
|
|
|
u16 cl_bpc;
|
2013-11-12 16:07:03 -07:00
|
|
|
u64 bg_ptr;
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
|
|
|
|
|
return -EROFS;
|
|
|
|
|
|
|
|
|
|
main_bm_inode = ocfs2_get_system_file_inode(osb,
|
|
|
|
|
GLOBAL_BITMAP_SYSTEM_INODE,
|
|
|
|
|
OCFS2_INVALID_SLOT);
|
|
|
|
|
if (!main_bm_inode) {
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out;
|
|
|
|
|
}
|
|
|
|
|
|
2016-01-22 13:40:57 -07:00
|
|
|
inode_lock(main_bm_inode);
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
|
|
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_mutex;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
|
|
|
|
|
|
|
|
|
|
if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
|
2010-04-12 23:38:06 -07:00
|
|
|
ocfs2_group_bitmap_size(osb->sb, 0,
|
|
|
|
|
osb->s_feature_incompat) * 8) {
|
2007-12-18 00:47:25 -07:00
|
|
|
mlog(ML_ERROR, "The disk is too old and small."
|
|
|
|
|
" Force to do offline resize.");
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
|
2008-10-09 17:20:29 -07:00
|
|
|
ret = ocfs2_read_blocks_sync(osb, input->group, 1, &group_bh);
|
2007-12-18 00:47:25 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog(ML_ERROR, "Can't read the group descriptor # %llu "
|
|
|
|
|
"from the device.", (unsigned long long)input->group);
|
|
|
|
|
goto out_unlock;
|
|
|
|
|
}
|
|
|
|
|
|
2009-02-10 21:00:41 -07:00
|
|
|
ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh);
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
|
|
|
|
|
if (ret) {
|
|
|
|
|
mlog_errno(ret);
|
2013-11-12 16:07:04 -07:00
|
|
|
goto out_free_group_bh;
|
2007-12-18 00:47:25 -07:00
|
|
|
}
|
|
|
|
|
|
2011-02-21 17:24:01 -07:00
|
|
|
trace_ocfs2_group_add((unsigned long long)input->group,
|
|
|
|
|
input->chain, input->clusters, input->frees);
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
|
|
|
|
|
if (IS_ERR(handle)) {
|
|
|
|
|
mlog_errno(PTR_ERR(handle));
|
|
|
|
|
ret = -EINVAL;
|
2013-11-12 16:07:04 -07:00
|
|
|
goto out_free_group_bh;
|
2007-12-18 00:47:25 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
|
|
|
|
|
cl = &fe->id2.i_chain;
|
|
|
|
|
cr = &cl->cl_recs[input->chain];
|
|
|
|
|
|
2009-02-12 17:41:25 -07:00
|
|
|
ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode),
|
|
|
|
|
group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
|
2007-12-18 00:47:25 -07:00
|
|
|
if (ret < 0) {
|
|
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_commit;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
group = (struct ocfs2_group_desc *)group_bh->b_data;
|
2013-11-12 16:07:03 -07:00
|
|
|
bg_ptr = le64_to_cpu(group->bg_next_group);
|
2007-12-18 00:47:25 -07:00
|
|
|
group->bg_next_group = cr->c_blkno;
|
2010-03-19 14:13:52 -07:00
|
|
|
ocfs2_journal_dirty(handle, group_bh);
|
2007-12-18 00:47:25 -07:00
|
|
|
|
2009-02-12 17:41:25 -07:00
|
|
|
ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
|
|
|
|
|
main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);
|
2007-12-18 00:47:25 -07:00
|
|
|
if (ret < 0) {
|
2013-11-12 16:07:03 -07:00
|
|
|
group->bg_next_group = cpu_to_le64(bg_ptr);
|
2007-12-18 00:47:25 -07:00
|
|
|
mlog_errno(ret);
|
|
|
|
|
goto out_commit;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
|
|
|
|
|
le16_add_cpu(&cl->cl_next_free_rec, 1);
|
|
|
|
|
memset(cr, 0, sizeof(struct ocfs2_chain_rec));
|
|
|
|
|
}
|
|
|
|
|
|
2008-03-02 19:53:02 -07:00
|
|
|
cr->c_blkno = cpu_to_le64(input->group);
|
2007-12-18 00:47:25 -07:00
|
|
|
le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
|
|
|
|
|
le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
|
|
|
|
|
|
|
|
|
|
le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
|
|
|
|
|
le32_add_cpu(&fe->id1.bitmap1.i_used,
|
|
|
|
|
(input->clusters - input->frees) * cl_bpc);
|
|
|
|
|
le32_add_cpu(&fe->i_clusters, input->clusters);
|
|
|
|
|
|
|
|
|
|
ocfs2_journal_dirty(handle, main_bm_bh);
|
|
|
|
|
|
|
|
|
|
spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
|
|
|
|
|
OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
|
2014-06-04 16:06:17 -07:00
|
|
|
le64_add_cpu(&fe->i_size, (u64)input->clusters << osb->s_clustersize_bits);
|
2007-12-18 00:47:25 -07:00
|
|
|
spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
|
|
|
|
|
i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
|
|
|
|
|
|
|
|
|
|
ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
|
|
|
|
|
|
|
|
|
|
out_commit:
|
|
|
|
|
ocfs2_commit_trans(osb, handle);
|
2013-11-12 16:07:04 -07:00
|
|
|
|
|
|
|
|
out_free_group_bh:
|
2008-01-23 19:35:31 -07:00
|
|
|
brelse(group_bh);
|
2013-11-12 16:07:04 -07:00
|
|
|
|
|
|
|
|
out_unlock:
|
2008-01-23 19:35:31 -07:00
|
|
|
brelse(main_bm_bh);
|
2007-12-18 00:47:25 -07:00
|
|
|
|
|
|
|
|
ocfs2_inode_unlock(main_bm_inode, 1);
|
|
|
|
|
|
|
|
|
|
out_mutex:
|
2016-01-22 13:40:57 -07:00
|
|
|
inode_unlock(main_bm_inode);
|
2007-12-18 00:47:25 -07:00
|
|
|
iput(main_bm_inode);
|
|
|
|
|
|
|
|
|
|
out:
|
|
|
|
|
return ret;
|
|
|
|
|
}
|