From: Dmitry Monakhov <dmonakhov@xxxxxxxxx>
Currently dq_data_lock is global, which is bad for scalability.
In fact different super_blocks have no shared quota data.
So we may simply convert global the lock to per-sb locks.
Signed-off-by: Dmitry Monakhov <dmonakhov@xxxxxxxxxx>
---
fs/ocfs2/quota_global.c | 20 ++++++++--------
fs/ocfs2/quota_local.c | 13 ++++++-----
fs/quota/dquot.c | 54 ++++++++++++++++++++++++----------------------
fs/quota/quota_tree.c | 8 +++---
fs/quota/quota_v2.c | 4 +-
include/linux/quota.h | 3 +-
6 files changed, 52 insertions(+), 50 deletions(-)
diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c
index b464947..3e7fda8 100644
--- a/fs/ocfs2/quota_global.c
+++ b/fs/ocfs2/quota_global.c
@@ -300,12 +300,12 @@ int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
status = ocfs2_inode_lock(oinfo->dqi_gqinode, &bh, ex);
if (status < 0)
return status;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(oinfo->dqi_gqinode->i_sb)->dq_data_lock);
if (!oinfo->dqi_gqi_count++)
oinfo->dqi_gqi_bh = bh;
else
WARN_ON(bh != oinfo->dqi_gqi_bh);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(oinfo->dqi_gqinode->i_sb)->dq_data_lock);
if (ex) {
mutex_lock(&oinfo->dqi_gqinode->i_mutex);
down_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
@@ -325,10 +325,10 @@ void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
}
ocfs2_inode_unlock(oinfo->dqi_gqinode, ex);
brelse(oinfo->dqi_gqi_bh);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(oinfo->dqi_gqinode->i_sb)->dq_data_lock);
if (!--oinfo->dqi_gqi_count)
oinfo->dqi_gqi_bh = NULL;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(oinfo->dqi_gqinode->i_sb)->dq_data_lock);
}
/* Read information header from global quota file */
@@ -421,11 +421,11 @@ static int __ocfs2_global_write_info(struct super_block *sb, int type)
struct ocfs2_global_disk_dqinfo dinfo;
ssize_t size;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
info->dqi_flags &= ~DQF_INFO_DIRTY;
dinfo.dqi_bgrace = cpu_to_le32(info->dqi_bgrace);
dinfo.dqi_igrace = cpu_to_le32(info->dqi_igrace);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
dinfo.dqi_syncms = cpu_to_le32(oinfo->dqi_syncms);
dinfo.dqi_blocks = cpu_to_le32(oinfo->dqi_gi.dqi_blocks);
dinfo.dqi_free_blk = cpu_to_le32(oinfo->dqi_gi.dqi_free_blk);
@@ -502,7 +502,7 @@ int __ocfs2_sync_dquot(struct dquot *dquot, int freeing)
/* Update space and inode usage. Get also other information from
* global quota file so that we don't overwrite any changes there.
* We are */
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
spacechange = dquot->dq_dqb.dqb_curspace -
OCFS2_DQUOT(dquot)->dq_origspace;
inodechange = dquot->dq_dqb.dqb_curinodes -
@@ -556,7 +556,7 @@ int __ocfs2_sync_dquot(struct dquot *dquot, int freeing)
__clear_bit(DQ_LASTSET_B + QIF_ITIME_B, &dquot->dq_flags);
OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace;
OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
err = ocfs2_qinfo_lock(info, freeing);
if (err < 0) {
mlog(ML_ERROR, "Failed to lock quota info, loosing quota write"
@@ -835,10 +835,10 @@ static int ocfs2_mark_dquot_dirty(struct dquot *dquot)
/* In case user set some limits, sync dquot immediately to global
* quota file so that information propagates quicker */
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
if (dquot->dq_flags & mask)
sync = 1;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
/* This is a slight hack but we can't afford getting global quota
* lock if we already have a transaction started. */
if (!sync || journal_current_handle()) {
diff --git a/fs/ocfs2/quota_local.c b/fs/ocfs2/quota_local.c
index 7c30ba3..2d2e981 100644
--- a/fs/ocfs2/quota_local.c
+++ b/fs/ocfs2/quota_local.c
@@ -288,14 +288,15 @@ static void olq_update_info(struct buffer_head *bh, void *private)
struct mem_dqinfo *info = private;
struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv;
struct ocfs2_local_disk_dqinfo *ldinfo;
+ struct quota_info *dqopt = dqopts(oinfo->dqi_gqinode->i_sb);
ldinfo = (struct ocfs2_local_disk_dqinfo *)(bh->b_data +
OCFS2_LOCAL_INFO_OFF);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopt->dq_data_lock);
ldinfo->dqi_flags = cpu_to_le32(info->dqi_flags & DQF_MASK);
ldinfo->dqi_chunks = cpu_to_le32(oinfo->dqi_chunks);
ldinfo->dqi_blocks = cpu_to_le32(oinfo->dqi_blocks);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopt->dq_data_lock);
}
static int ocfs2_add_recovery_chunk(struct super_block *sb,
@@ -523,7 +524,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode,
goto out_drop_lock;
}
mutex_lock(&dqopts(sb)->dqio_mutex);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
/* Add usage from quota entry into quota changes
* of our node. Auxiliary variables are important
* due to signedness */
@@ -531,7 +532,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode,
inodechange = le64_to_cpu(dqblk->dqb_inodemod);
dquot->dq_dqb.dqb_curspace += spacechange;
dquot->dq_dqb.dqb_curinodes += inodechange;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
/* We want to drop reference held by the crashed
* node. Since we have our own reference we know
* global structure actually won't be freed. */
@@ -876,12 +877,12 @@ static void olq_set_dquot(struct buffer_head *bh, void *private)
+ ol_dqblk_block_offset(sb, od->dq_local_off));
dqblk->dqb_id = cpu_to_le64(od->dq_dquot.dq_id);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
dqblk->dqb_spacemod = cpu_to_le64(od->dq_dquot.dq_dqb.dqb_curspace -
od->dq_origspace);
dqblk->dqb_inodemod = cpu_to_le64(od->dq_dquot.dq_dqb.dqb_curinodes -
od->dq_originodes);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
mlog(0, "Writing local dquot %u space %lld inodes %lld\n",
od->dq_dquot.dq_id, (long long)le64_to_cpu(dqblk->dqb_spacemod),
(long long)le64_to_cpu(dqblk->dqb_inodemod));
diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
index fdaa386..6c68172 100644
--- a/fs/quota/dquot.c
+++ b/fs/quota/dquot.c
@@ -128,8 +128,6 @@
*/
static __cacheline_aligned_in_smp DEFINE_SPINLOCK(dq_fmt_lock);
-__cacheline_aligned_in_smp DEFINE_SPINLOCK(dq_data_lock);
-EXPORT_SYMBOL(dq_data_lock);
void __quota_error(struct super_block *sb, const char *func,
const char *fmt, ...)
@@ -1417,8 +1415,11 @@ static void __dquot_initialize(struct inode *inode, int type)
* did a write before quota was turned on
*/
rsv = inode_get_rsv_space(inode);
- if (unlikely(rsv))
- dquot_resv_space(inode->i_dquot[cnt], rsv);
+ if (unlikely(rsv)) {
+ spin_lock(&got[cnt]->dq_lock);
+ dquot_resv_space(got[cnt], rsv);
+ spin_unlock(&got[cnt]->dq_lock);
+ }
}
}
out_err:
@@ -1574,14 +1575,14 @@ int __dquot_alloc_space(struct inode *inode, qsize_t number, int flags)
for (cnt = 0; cnt < MAXQUOTAS; cnt++)
warntype[cnt] = QUOTA_NL_NOWARN;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
if (!inode->i_dquot[cnt])
continue;
ret = check_bdq(inode->i_dquot[cnt], number, !warn,
warntype+cnt);
if (ret && !nofail) {
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
goto out_flush_warn;
}
}
@@ -1594,7 +1595,7 @@ int __dquot_alloc_space(struct inode *inode, qsize_t number, int flags)
dquot_incr_space(inode->i_dquot[cnt], number);
}
inode_incr_space(inode, number, reserve);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
if (reserve)
goto out_flush_warn;
@@ -1622,7 +1623,7 @@ int dquot_alloc_inode(const struct inode *inode)
for (cnt = 0; cnt < MAXQUOTAS; cnt++)
warntype[cnt] = QUOTA_NL_NOWARN;
down_read(&dqctl(inode->i_sb)->dqptr_sem);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
if (!inode->i_dquot[cnt])
continue;
@@ -1638,7 +1639,7 @@ int dquot_alloc_inode(const struct inode *inode)
}
warn_put_all:
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
if (ret == 0)
mark_all_dquot_dirty(inode->i_dquot);
flush_warnings(inode->i_dquot, warntype);
@@ -1660,7 +1661,7 @@ int dquot_claim_space_nodirty(struct inode *inode, qsize_t number)
}
down_read(&dqctl(inode->i_sb)->dqptr_sem);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
/* Claim reserved quotas to allocated quotas */
for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
if (inode->i_dquot[cnt])
@@ -1669,7 +1670,7 @@ int dquot_claim_space_nodirty(struct inode *inode, qsize_t number)
}
/* Update inode bytes */
inode_claim_rsv_space(inode, number);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
mark_all_dquot_dirty(inode->i_dquot);
up_read(&dqctl(inode->i_sb)->dqptr_sem);
return 0;
@@ -1693,7 +1694,7 @@ void __dquot_free_space(struct inode *inode, qsize_t number, int flags)
}
down_read(&dqctl(inode->i_sb)->dqptr_sem);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
if (!inode->i_dquot[cnt])
continue;
@@ -1704,7 +1705,7 @@ void __dquot_free_space(struct inode *inode, qsize_t number, int flags)
dquot_decr_space(inode->i_dquot[cnt], number);
}
inode_decr_space(inode, number, reserve);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
if (reserve)
goto out_unlock;
@@ -1729,14 +1730,14 @@ void dquot_free_inode(const struct inode *inode)
return;
down_read(&dqctl(inode->i_sb)->dqptr_sem);
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
if (!inode->i_dquot[cnt])
continue;
warntype[cnt] = info_idq_free(inode->i_dquot[cnt], 1);
dquot_decr_inodes(inode->i_dquot[cnt], 1);
}
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
mark_all_dquot_dirty(inode->i_dquot);
flush_warnings(inode->i_dquot, warntype);
up_read(&dqctl(inode->i_sb)->dqptr_sem);
@@ -1775,7 +1776,7 @@ int __dquot_transfer(struct inode *inode, struct dquot **transfer_to)
up_write(&dqctl(inode->i_sb)->dqptr_sem);
return 0;
}
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(inode->i_sb)->dq_data_lock);
cur_space = inode_get_bytes(inode);
rsv_space = inode_get_rsv_space(inode);
space = cur_space + rsv_space;
@@ -1823,7 +1824,7 @@ int __dquot_transfer(struct inode *inode, struct dquot **transfer_to)
inode->i_dquot[cnt] = transfer_to[cnt];
}
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
up_write(&dqctl(inode->i_sb)->dqptr_sem);
mark_all_dquot_dirty(transfer_from);
@@ -1837,7 +1838,7 @@ int __dquot_transfer(struct inode *inode, struct dquot **transfer_to)
transfer_to[cnt] = transfer_from[cnt];
return 0;
over_quota:
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(inode->i_sb)->dq_data_lock);
up_write(&dqctl(inode->i_sb)->dqptr_sem);
flush_warnings(transfer_to, warntype_to);
return ret;
@@ -1958,6 +1959,7 @@ static int alloc_quota_info(struct quota_ctl_info *dqctl) {
mutex_init(&dqopt->dqio_mutex);
spin_lock_init(&dqopt->dq_state_lock);
+ spin_lock_init(&dqopt->dq_data_lock);
spin_lock_init(&dqopt->dq_list_lock);
INIT_LIST_HEAD(&dqopt->dq_inuse_list);
INIT_LIST_HEAD(&dqopt->dq_free_list);
@@ -2415,7 +2417,7 @@ static void do_get_dqblk(struct dquot *dquot, struct fs_disk_quota *di)
FS_USER_QUOTA : FS_GROUP_QUOTA;
di->d_id = dquot->dq_id;
- spin_lock(&dq_data_lock);
+ spin_lock(&sb_dqopts(dquot)->dq_data_lock);
di->d_blk_hardlimit = stoqb(dm->dqb_bhardlimit);
di->d_blk_softlimit = stoqb(dm->dqb_bsoftlimit);
di->d_ino_hardlimit = dm->dqb_ihardlimit;
@@ -2424,7 +2426,7 @@ static void do_get_dqblk(struct dquot *dquot, struct fs_disk_quota *di)
di->d_icount = dm->dqb_curinodes;
di->d_btimer = dm->dqb_btime;
di->d_itimer = dm->dqb_itime;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&sb_dqopts(dquot)->dq_data_lock);
}
int dquot_get_dqblk(struct super_block *sb, int type, qid_t id,
@@ -2467,7 +2469,7 @@ static int do_set_dqblk(struct dquot *dquot, struct fs_disk_quota *di)
(di->d_ino_hardlimit > dqi->dqi_maxilimit)))
return -ERANGE;
- spin_lock(&dq_data_lock);
+ spin_lock(&sb_dqopts(dquot)->dq_data_lock);
if (di->d_fieldmask & FS_DQ_BCOUNT) {
dm->dqb_curspace = di->d_bcount - dm->dqb_rsvspace;
check_blim = 1;
@@ -2533,7 +2535,7 @@ static int do_set_dqblk(struct dquot *dquot, struct fs_disk_quota *di)
clear_bit(DQ_FAKE_B, &dquot->dq_flags);
else
set_bit(DQ_FAKE_B, &dquot->dq_flags);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&sb_dqopts(dquot)->dq_data_lock);
mark_dquot_dirty(dquot);
return 0;
@@ -2568,12 +2570,12 @@ int dquot_get_dqinfo(struct super_block *sb, int type, struct if_dqinfo *ii)
return -ESRCH;
}
mi = dqopts(sb)->info + type;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
ii->dqi_bgrace = mi->dqi_bgrace;
ii->dqi_igrace = mi->dqi_igrace;
ii->dqi_flags = mi->dqi_flags & DQF_MASK;
ii->dqi_valid = IIF_ALL;
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
mutex_unlock(&dqctl(sb)->dqonoff_mutex);
return 0;
}
@@ -2591,7 +2593,7 @@ int dquot_set_dqinfo(struct super_block *sb, int type, struct if_dqinfo *ii)
goto out;
}
mi = dqopts(sb)->info + type;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
if (ii->dqi_valid & IIF_BGRACE)
mi->dqi_bgrace = ii->dqi_bgrace;
if (ii->dqi_valid & IIF_IGRACE)
@@ -2599,7 +2601,7 @@ int dquot_set_dqinfo(struct super_block *sb, int type, struct if_dqinfo *ii)
if (ii->dqi_valid & IIF_FLAGS)
mi->dqi_flags = (mi->dqi_flags & ~DQF_MASK) |
(ii->dqi_flags & DQF_MASK);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
mark_info_dirty(sb, type);
/* Force write to disk */
dqctl(sb)->dq_op->write_info(sb, type);
diff --git a/fs/quota/quota_tree.c b/fs/quota/quota_tree.c
index 21a4a6a..a089c70 100644
--- a/fs/quota/quota_tree.c
+++ b/fs/quota/quota_tree.c
@@ -375,9 +375,9 @@ int qtree_write_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
return ret;
}
}
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
info->dqi_ops->mem2disk_dqblk(ddquot, dquot);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
ret = sb->s_op->quota_write(sb, type, ddquot, info->dqi_entry_size,
dquot->dq_off);
if (ret != info->dqi_entry_size) {
@@ -631,14 +631,14 @@ int qtree_read_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
kfree(ddquot);
goto out;
}
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
info->dqi_ops->disk2mem_dqblk(dquot, ddquot);
if (!dquot->dq_dqb.dqb_bhardlimit &&
!dquot->dq_dqb.dqb_bsoftlimit &&
!dquot->dq_dqb.dqb_ihardlimit &&
!dquot->dq_dqb.dqb_isoftlimit)
set_bit(DQ_FAKE_B, &dquot->dq_flags);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
kfree(ddquot);
out:
dqstats_inc(DQST_READS);
diff --git a/fs/quota/quota_v2.c b/fs/quota/quota_v2.c
index 65444d2..e4ef8de 100644
--- a/fs/quota/quota_v2.c
+++ b/fs/quota/quota_v2.c
@@ -153,12 +153,12 @@ static int v2_write_file_info(struct super_block *sb, int type)
struct qtree_mem_dqinfo *qinfo = info->dqi_priv;
ssize_t size;
- spin_lock(&dq_data_lock);
+ spin_lock(&dqopts(sb)->dq_data_lock);
info->dqi_flags &= ~DQF_INFO_DIRTY;
dinfo.dqi_bgrace = cpu_to_le32(info->dqi_bgrace);
dinfo.dqi_igrace = cpu_to_le32(info->dqi_igrace);
dinfo.dqi_flags = cpu_to_le32(info->dqi_flags & DQF_MASK);
- spin_unlock(&dq_data_lock);
+ spin_unlock(&dqopts(sb)->dq_data_lock);
dinfo.dqi_blocks = cpu_to_le32(qinfo->dqi_blocks);
dinfo.dqi_free_blk = cpu_to_le32(qinfo->dqi_free_blk);
dinfo.dqi_free_entry = cpu_to_le32(qinfo->dqi_free_entry);
diff --git a/include/linux/quota.h b/include/linux/quota.h
index 754aedb..6b04001 100644
--- a/include/linux/quota.h
+++ b/include/linux/quota.h
@@ -186,8 +186,6 @@ enum {
typedef __kernel_uid32_t qid_t; /* Type in which we store ids in memory */
typedef long long qsize_t; /* Type in which we store sizes */
-extern spinlock_t dq_data_lock;
-
/* Maximal numbers of writes for quota operation (insert/delete/update)
* (over VFS all formats) */
#define DQUOT_INIT_ALLOC max(V1_INIT_ALLOC, V2_INIT_ALLOC)
@@ -413,6 +411,7 @@ struct quota_info {
struct mutex dqio_mutex; /* lock device while I/O in progress */
struct mem_dqinfo info[MAXQUOTAS]; /* Information for each quota type */
spinlock_t dq_state_lock; /* serialize quota state changes*/
+ spinlock_t dq_data_lock; /* protect in memory data */
spinlock_t dq_list_lock; /* protect lists and hash*/
struct list_head dq_inuse_list; /* list of inused dquotas */
struct list_head dq_free_list; /* list of free dquotas */
--
1.6.5.2
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
