2018-06-06 09:42:14 +07:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2005-11-02 10:59:41 +07:00
|
|
|
* Copyright (c) 2000-2003 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
#include "xfs.h"
|
|
|
|
#include "xfs_fs.h"
|
2013-08-12 17:49:26 +07:00
|
|
|
#include "xfs_format.h"
|
2013-10-23 06:50:10 +07:00
|
|
|
#include "xfs_log_format.h"
|
2013-10-23 06:36:05 +07:00
|
|
|
#include "xfs_shared.h"
|
2013-10-23 06:50:10 +07:00
|
|
|
#include "xfs_trans_resv.h"
|
2005-11-02 10:38:42 +07:00
|
|
|
#include "xfs_bit.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
#include "xfs_mount.h"
|
2016-08-03 08:15:38 +07:00
|
|
|
#include "xfs_defer.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
#include "xfs_inode.h"
|
|
|
|
#include "xfs_bmap.h"
|
2013-08-12 17:49:42 +07:00
|
|
|
#include "xfs_bmap_util.h"
|
2013-10-23 06:50:10 +07:00
|
|
|
#include "xfs_alloc.h"
|
|
|
|
#include "xfs_quota.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
#include "xfs_error.h"
|
2013-10-23 06:50:10 +07:00
|
|
|
#include "xfs_trans.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
#include "xfs_buf_item.h"
|
|
|
|
#include "xfs_trans_space.h"
|
|
|
|
#include "xfs_trans_priv.h"
|
|
|
|
#include "xfs_qm.h"
|
2013-04-03 12:11:16 +07:00
|
|
|
#include "xfs_cksum.h"
|
2009-12-15 06:14:59 +07:00
|
|
|
#include "xfs_trace.h"
|
2013-10-23 06:50:10 +07:00
|
|
|
#include "xfs_log.h"
|
2013-10-23 06:51:50 +07:00
|
|
|
#include "xfs_bmap_btree.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
2011-12-07 04:58:19 +07:00
|
|
|
* Lock order:
|
|
|
|
*
|
|
|
|
* ip->i_lock
|
2012-03-13 15:52:35 +07:00
|
|
|
* qi->qi_tree_lock
|
2012-03-14 23:53:34 +07:00
|
|
|
* dquot->q_qlock (xfs_dqlock() and friends)
|
|
|
|
* dquot->q_flush (xfs_dqflock() and friends)
|
|
|
|
* qi->qi_lru_lock
|
2011-12-07 04:58:19 +07:00
|
|
|
*
|
|
|
|
* If two dquots need to be locked the order is user before group/project,
|
|
|
|
* otherwise by the lowest id first, see xfs_dqlock2.
|
|
|
|
*/
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-03-13 15:52:37 +07:00
|
|
|
struct kmem_zone *xfs_qm_dqtrxzone;
|
|
|
|
static struct kmem_zone *xfs_qm_dqzone;
|
|
|
|
|
xfs: lockdep needs to know about 3 dquot-deep nesting
Michael Semon reported that xfs/299 generated this lockdep warning:
=============================================
[ INFO: possible recursive locking detected ]
3.12.0-rc2+ #2 Not tainted
---------------------------------------------
touch/21072 is trying to acquire lock:
(&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
but task is already holding lock:
(&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&xfs_dquot_other_class);
lock(&xfs_dquot_other_class);
*** DEADLOCK ***
May be due to missing lock nesting notation
7 locks held by touch/21072:
#0: (sb_writers#10){++++.+}, at: [<c11185b6>] mnt_want_write+0x1e/0x3e
#1: (&type->i_mutex_dir_key#4){+.+.+.}, at: [<c11078ee>] do_last+0x245/0xe40
#2: (sb_internal#2){++++.+}, at: [<c122c9e0>] xfs_trans_alloc+0x1f/0x35
#3: (&(&ip->i_lock)->mr_lock/1){+.+...}, at: [<c126cd1b>] xfs_ilock+0x100/0x1f1
#4: (&(&ip->i_lock)->mr_lock){++++-.}, at: [<c126cf52>] xfs_ilock_nowait+0x105/0x22f
#5: (&dqp->q_qlock){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
#6: (&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
The lockdep annotation for dquot lock nesting only understands
locking for user and "other" dquots, not user, group and quota
dquots. Fix the annotations to match the locking heirarchy we now
have.
Reported-by: Michael L. Semon <mlsemon35@gmail.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-09-30 06:37:03 +07:00
|
|
|
static struct lock_class_key xfs_dquot_group_class;
|
|
|
|
static struct lock_class_key xfs_dquot_project_class;
|
2009-01-19 08:03:25 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* This is called to free all the memory associated with a dquot
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_qm_dqdestroy(
|
|
|
|
xfs_dquot_t *dqp)
|
|
|
|
{
|
2012-03-13 15:52:34 +07:00
|
|
|
ASSERT(list_empty(&dqp->q_lru));
|
2005-04-17 05:20:36 +07:00
|
|
|
|
xfs: allocate log vector buffers outside CIL context lock
One of the problems we currently have with delayed logging is that
under serious memory pressure we can deadlock memory reclaim. THis
occurs when memory reclaim (such as run by kswapd) is reclaiming XFS
inodes and issues a log force to unpin inodes that are dirty in the
CIL.
The CIL is pushed, but this will only occur once it gets the CIL
context lock to ensure that all committing transactions are complete
and no new transactions start being committed to the CIL while the
push switches to a new context.
The deadlock occurs when the CIL context lock is held by a
committing process that is doing memory allocation for log vector
buffers, and that allocation is then blocked on memory reclaim
making progress. Memory reclaim, however, is blocked waiting for
a log force to make progress, and so we effectively deadlock at this
point.
To solve this problem, we have to move the CIL log vector buffer
allocation outside of the context lock so that memory reclaim can
always make progress when it needs to force the log. The problem
with doing this is that a CIL push can take place while we are
determining if we need to allocate a new log vector buffer for
an item and hence the current log vector may go away without
warning. That means we canot rely on the existing log vector being
present when we finally grab the context lock and so we must have a
replacement buffer ready to go at all times.
To ensure this, introduce a "shadow log vector" buffer that is
always guaranteed to be present when we gain the CIL context lock
and format the item. This shadow buffer may or may not be used
during the formatting, but if the log item does not have an existing
log vector buffer or that buffer is too small for the new
modifications, we swap it for the new shadow buffer and format
the modifications into that new log vector buffer.
The result of this is that for any object we modify more than once
in a given CIL checkpoint, we double the memory required
to track dirty regions in the log. For single modifications then
we consume the shadow log vectorwe allocate on commit, and that gets
consumed by the checkpoint. However, if we make multiple
modifications, then the second transaction commit will allocate a
shadow log vector and hence we will end up with double the memory
usage as only one of the log vectors is consumed by the CIL
checkpoint. The remaining shadow vector will be freed when th elog
item is freed.
This can probably be optimised in future - access to the shadow log
vector is serialised by the object lock (as opposited to the active
log vector, which is controlled by the CIL context lock) and so we
can probably free shadow log vector from some objects when the log
item is marked clean on removal from the AIL.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 06:52:35 +07:00
|
|
|
kmem_free(dqp->q_logitem.qli_item.li_lv_shadow);
|
2005-04-17 05:20:36 +07:00
|
|
|
mutex_destroy(&dqp->q_qlock);
|
2009-12-15 06:14:59 +07:00
|
|
|
|
2015-10-12 14:21:22 +07:00
|
|
|
XFS_STATS_DEC(dqp->q_mount, xs_qm_dquot);
|
|
|
|
kmem_zone_free(xfs_qm_dqzone, dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If default limits are in force, push them into the dquot now.
|
|
|
|
* We overwrite the dquot limits only if they are zero and this
|
|
|
|
* is not the root dquot.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_qm_adjust_dqlimits(
|
2013-03-18 21:51:45 +07:00
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_dquot *dq)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2013-03-18 21:51:45 +07:00
|
|
|
struct xfs_quotainfo *q = mp->m_quotainfo;
|
|
|
|
struct xfs_disk_dquot *d = &dq->q_core;
|
2016-02-08 07:27:55 +07:00
|
|
|
struct xfs_def_quota *defq;
|
2013-03-18 21:51:46 +07:00
|
|
|
int prealloc = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
ASSERT(d->d_id);
|
2016-02-08 07:27:55 +07:00
|
|
|
defq = xfs_get_defquota(dq, q);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2016-02-08 07:27:55 +07:00
|
|
|
if (defq->bsoftlimit && !d->d_blk_softlimit) {
|
|
|
|
d->d_blk_softlimit = cpu_to_be64(defq->bsoftlimit);
|
2013-03-18 21:51:46 +07:00
|
|
|
prealloc = 1;
|
|
|
|
}
|
2016-02-08 07:27:55 +07:00
|
|
|
if (defq->bhardlimit && !d->d_blk_hardlimit) {
|
|
|
|
d->d_blk_hardlimit = cpu_to_be64(defq->bhardlimit);
|
2013-03-18 21:51:46 +07:00
|
|
|
prealloc = 1;
|
|
|
|
}
|
2016-02-08 07:27:55 +07:00
|
|
|
if (defq->isoftlimit && !d->d_ino_softlimit)
|
|
|
|
d->d_ino_softlimit = cpu_to_be64(defq->isoftlimit);
|
|
|
|
if (defq->ihardlimit && !d->d_ino_hardlimit)
|
|
|
|
d->d_ino_hardlimit = cpu_to_be64(defq->ihardlimit);
|
|
|
|
if (defq->rtbsoftlimit && !d->d_rtb_softlimit)
|
|
|
|
d->d_rtb_softlimit = cpu_to_be64(defq->rtbsoftlimit);
|
|
|
|
if (defq->rtbhardlimit && !d->d_rtb_hardlimit)
|
|
|
|
d->d_rtb_hardlimit = cpu_to_be64(defq->rtbhardlimit);
|
2013-03-18 21:51:46 +07:00
|
|
|
|
|
|
|
if (prealloc)
|
|
|
|
xfs_dquot_set_prealloc_limits(dq);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check the limits and timers of a dquot and start or reset timers
|
|
|
|
* if necessary.
|
|
|
|
* This gets called even when quota enforcement is OFF, which makes our
|
|
|
|
* life a little less complicated. (We just don't reject any quota
|
|
|
|
* reservations in that case, when enforcement is off).
|
|
|
|
* We also return 0 as the values of the timers in Q_GETQUOTA calls, when
|
|
|
|
* enforcement's off.
|
|
|
|
* In contrast, warnings are a little different in that they don't
|
2005-06-21 12:49:06 +07:00
|
|
|
* 'automatically' get started when limits get exceeded. They do
|
|
|
|
* get reset to zero, however, when we find the count to be under
|
|
|
|
* the soft limit (they are only ever set non-zero via userspace).
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_qm_adjust_dqtimers(
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
xfs_disk_dquot_t *d)
|
|
|
|
{
|
|
|
|
ASSERT(d->d_id);
|
|
|
|
|
2011-07-13 18:43:50 +07:00
|
|
|
#ifdef DEBUG
|
2005-11-02 11:01:12 +07:00
|
|
|
if (d->d_blk_hardlimit)
|
|
|
|
ASSERT(be64_to_cpu(d->d_blk_softlimit) <=
|
|
|
|
be64_to_cpu(d->d_blk_hardlimit));
|
|
|
|
if (d->d_ino_hardlimit)
|
|
|
|
ASSERT(be64_to_cpu(d->d_ino_softlimit) <=
|
|
|
|
be64_to_cpu(d->d_ino_hardlimit));
|
|
|
|
if (d->d_rtb_hardlimit)
|
|
|
|
ASSERT(be64_to_cpu(d->d_rtb_softlimit) <=
|
|
|
|
be64_to_cpu(d->d_rtb_hardlimit));
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif
|
2011-07-13 18:43:50 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
if (!d->d_btimer) {
|
2005-11-02 11:01:12 +07:00
|
|
|
if ((d->d_blk_softlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_bcount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_blk_softlimit))) ||
|
|
|
|
(d->d_blk_hardlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_bcount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_blk_hardlimit)))) {
|
|
|
|
d->d_btimer = cpu_to_be32(get_seconds() +
|
2010-04-20 14:01:30 +07:00
|
|
|
mp->m_quotainfo->qi_btimelimit);
|
2005-06-21 12:49:06 +07:00
|
|
|
} else {
|
|
|
|
d->d_bwarns = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if ((!d->d_blk_softlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_bcount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_blk_softlimit))) &&
|
2005-04-17 05:20:36 +07:00
|
|
|
(!d->d_blk_hardlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_bcount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_blk_hardlimit)))) {
|
2005-04-17 05:20:36 +07:00
|
|
|
d->d_btimer = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!d->d_itimer) {
|
2005-11-02 11:01:12 +07:00
|
|
|
if ((d->d_ino_softlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_icount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_ino_softlimit))) ||
|
|
|
|
(d->d_ino_hardlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_icount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_ino_hardlimit)))) {
|
|
|
|
d->d_itimer = cpu_to_be32(get_seconds() +
|
2010-04-20 14:01:30 +07:00
|
|
|
mp->m_quotainfo->qi_itimelimit);
|
2005-06-21 12:49:06 +07:00
|
|
|
} else {
|
|
|
|
d->d_iwarns = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if ((!d->d_ino_softlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_icount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_ino_softlimit))) &&
|
2005-04-17 05:20:36 +07:00
|
|
|
(!d->d_ino_hardlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_icount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_ino_hardlimit)))) {
|
2005-04-17 05:20:36 +07:00
|
|
|
d->d_itimer = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!d->d_rtbtimer) {
|
2005-11-02 11:01:12 +07:00
|
|
|
if ((d->d_rtb_softlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_rtbcount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_rtb_softlimit))) ||
|
|
|
|
(d->d_rtb_hardlimit &&
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_rtbcount) >
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_rtb_hardlimit)))) {
|
|
|
|
d->d_rtbtimer = cpu_to_be32(get_seconds() +
|
2010-04-20 14:01:30 +07:00
|
|
|
mp->m_quotainfo->qi_rtbtimelimit);
|
2005-06-21 12:49:06 +07:00
|
|
|
} else {
|
|
|
|
d->d_rtbwarns = 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if ((!d->d_rtb_softlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_rtbcount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_rtb_softlimit))) &&
|
2005-04-17 05:20:36 +07:00
|
|
|
(!d->d_rtb_hardlimit ||
|
2012-02-06 19:50:07 +07:00
|
|
|
(be64_to_cpu(d->d_rtbcount) <=
|
2005-11-02 11:01:12 +07:00
|
|
|
be64_to_cpu(d->d_rtb_hardlimit)))) {
|
2005-04-17 05:20:36 +07:00
|
|
|
d->d_rtbtimer = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* initialize a buffer full of dquots and log the whole thing
|
|
|
|
*/
|
|
|
|
STATIC void
|
|
|
|
xfs_qm_init_dquot_blk(
|
|
|
|
xfs_trans_t *tp,
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
uint type,
|
|
|
|
xfs_buf_t *bp)
|
|
|
|
{
|
2010-04-20 14:01:30 +07:00
|
|
|
struct xfs_quotainfo *q = mp->m_quotainfo;
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_dqblk_t *d;
|
2016-02-08 07:22:58 +07:00
|
|
|
xfs_dqid_t curid;
|
|
|
|
int i;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
ASSERT(tp);
|
2011-07-08 19:36:19 +07:00
|
|
|
ASSERT(xfs_buf_islocked(bp));
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-07-23 06:40:15 +07:00
|
|
|
d = bp->b_addr;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* ID of the first dquot in the block - id's are zero based.
|
|
|
|
*/
|
2010-04-20 14:01:30 +07:00
|
|
|
curid = id - (id % q->qi_dqperchunk);
|
|
|
|
memset(d, 0, BBTOB(q->qi_dqchunklen));
|
2011-12-07 04:58:23 +07:00
|
|
|
for (i = 0; i < q->qi_dqperchunk; i++, d++, curid++) {
|
|
|
|
d->dd_diskdq.d_magic = cpu_to_be16(XFS_DQUOT_MAGIC);
|
|
|
|
d->dd_diskdq.d_version = XFS_DQUOT_VERSION;
|
|
|
|
d->dd_diskdq.d_id = cpu_to_be32(curid);
|
|
|
|
d->dd_diskdq.d_flags = type;
|
xfs: rework dquot CRCs
Calculating dquot CRCs when the backing buffer is written back just
doesn't work reliably. There are several places which manipulate
dquots directly in the buffers, and they don't calculate CRCs
appropriately, nor do they always set the buffer up to calculate
CRCs appropriately.
Firstly, if we log a dquot buffer (e.g. during allocation) it gets
logged without valid CRC, and so on recovery we end up with a dquot
that is not valid.
Secondly, if we recover/repair a dquot, we don't have a verifier
attached to the buffer and hence CRCs are not calculated on the way
down to disk.
Thirdly, calculating the CRC after we've changed the contents means
that if we re-read the dquot from the buffer, we cannot verify the
contents of the dquot are valid, as the CRC is invalid.
So, to avoid all the dquot CRC errors that are being detected by the
read verifier, change to using the same model as for inodes. That
is, dquot CRCs are calculated and written to the backing buffer at
the time the dquot is flushed to the backing buffer. If we modify
the dquot directly in the backing buffer, calculate the CRC
immediately after the modification is complete. Hence the dquot in
the on-disk buffer should always have a valid CRC.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-06-03 12:28:46 +07:00
|
|
|
if (xfs_sb_version_hascrc(&mp->m_sb)) {
|
2015-08-19 07:32:01 +07:00
|
|
|
uuid_copy(&d->dd_uuid, &mp->m_sb.sb_meta_uuid);
|
xfs: rework dquot CRCs
Calculating dquot CRCs when the backing buffer is written back just
doesn't work reliably. There are several places which manipulate
dquots directly in the buffers, and they don't calculate CRCs
appropriately, nor do they always set the buffer up to calculate
CRCs appropriately.
Firstly, if we log a dquot buffer (e.g. during allocation) it gets
logged without valid CRC, and so on recovery we end up with a dquot
that is not valid.
Secondly, if we recover/repair a dquot, we don't have a verifier
attached to the buffer and hence CRCs are not calculated on the way
down to disk.
Thirdly, calculating the CRC after we've changed the contents means
that if we re-read the dquot from the buffer, we cannot verify the
contents of the dquot are valid, as the CRC is invalid.
So, to avoid all the dquot CRC errors that are being detected by the
read verifier, change to using the same model as for inodes. That
is, dquot CRCs are calculated and written to the backing buffer at
the time the dquot is flushed to the backing buffer. If we modify
the dquot directly in the backing buffer, calculate the CRC
immediately after the modification is complete. Hence the dquot in
the on-disk buffer should always have a valid CRC.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-06-03 12:28:46 +07:00
|
|
|
xfs_update_cksum((char *)d, sizeof(struct xfs_dqblk),
|
|
|
|
XFS_DQUOT_CRC_OFF);
|
|
|
|
}
|
2011-12-07 04:58:23 +07:00
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_trans_dquot_buf(tp, bp,
|
2010-05-07 08:05:19 +07:00
|
|
|
(type & XFS_DQ_USER ? XFS_BLF_UDQUOT_BUF :
|
|
|
|
((type & XFS_DQ_PROJ) ? XFS_BLF_PDQUOT_BUF :
|
|
|
|
XFS_BLF_GDQUOT_BUF)));
|
2010-04-20 14:01:30 +07:00
|
|
|
xfs_trans_log_buf(tp, bp, 0, BBTOB(q->qi_dqchunklen) - 1);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2013-03-18 21:51:46 +07:00
|
|
|
/*
|
|
|
|
* Initialize the dynamic speculative preallocation thresholds. The lo/hi
|
|
|
|
* watermarks correspond to the soft and hard limits by default. If a soft limit
|
|
|
|
* is not specified, we use 95% of the hard limit.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_dquot_set_prealloc_limits(struct xfs_dquot *dqp)
|
|
|
|
{
|
2017-06-17 01:00:05 +07:00
|
|
|
uint64_t space;
|
2013-03-18 21:51:46 +07:00
|
|
|
|
|
|
|
dqp->q_prealloc_hi_wmark = be64_to_cpu(dqp->q_core.d_blk_hardlimit);
|
|
|
|
dqp->q_prealloc_lo_wmark = be64_to_cpu(dqp->q_core.d_blk_softlimit);
|
|
|
|
if (!dqp->q_prealloc_lo_wmark) {
|
|
|
|
dqp->q_prealloc_lo_wmark = dqp->q_prealloc_hi_wmark;
|
|
|
|
do_div(dqp->q_prealloc_lo_wmark, 100);
|
|
|
|
dqp->q_prealloc_lo_wmark *= 95;
|
|
|
|
}
|
|
|
|
|
|
|
|
space = dqp->q_prealloc_hi_wmark;
|
|
|
|
|
|
|
|
do_div(space, 100);
|
|
|
|
dqp->q_low_space[XFS_QLOWSP_1_PCNT] = space;
|
|
|
|
dqp->q_low_space[XFS_QLOWSP_3_PCNT] = space * 3;
|
|
|
|
dqp->q_low_space[XFS_QLOWSP_5_PCNT] = space * 5;
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2018-05-05 05:30:23 +07:00
|
|
|
* Ensure that the given in-core dquot has a buffer on disk backing it, and
|
|
|
|
* return the buffer. This is called when the bmapi finds a hole.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
STATIC int
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_dquot_disk_alloc(
|
|
|
|
struct xfs_trans **tpp,
|
|
|
|
struct xfs_dquot *dqp,
|
|
|
|
struct xfs_buf **bpp)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_bmbt_irec map;
|
2018-07-12 12:26:11 +07:00
|
|
|
struct xfs_trans *tp = *tpp;
|
|
|
|
struct xfs_mount *mp = tp->t_mountp;
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_buf *bp;
|
|
|
|
struct xfs_inode *quotip = xfs_quota_inode(mp, dqp->dq_flags);
|
|
|
|
int nmaps = 1;
|
|
|
|
int error;
|
2009-12-15 06:14:59 +07:00
|
|
|
|
|
|
|
trace_xfs_dqalloc(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
xfs_ilock(quotip, XFS_ILOCK_EXCL);
|
2012-01-24 00:31:25 +07:00
|
|
|
if (!xfs_this_quota_on(dqp->q_mount, dqp->dq_flags)) {
|
2018-05-05 05:30:23 +07:00
|
|
|
/*
|
|
|
|
* Return if this type of quotas is turned off while we didn't
|
|
|
|
* have an inode lock
|
|
|
|
*/
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_iunlock(quotip, XFS_ILOCK_EXCL);
|
2014-06-25 11:58:08 +07:00
|
|
|
return -ESRCH;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/* Create the block mapping. */
|
2018-07-12 12:26:11 +07:00
|
|
|
xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL);
|
|
|
|
error = xfs_bmapi_write(tp, quotip, dqp->q_fileoffset,
|
2018-05-05 05:30:23 +07:00
|
|
|
XFS_DQUOT_CLUSTER_SIZE_FSB, XFS_BMAPI_METADATA,
|
2018-07-12 12:26:25 +07:00
|
|
|
XFS_QM_DQALLOC_SPACE_RES(mp), &map, &nmaps);
|
2011-09-19 03:40:52 +07:00
|
|
|
if (error)
|
2018-08-08 00:57:13 +07:00
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB);
|
|
|
|
ASSERT(nmaps == 1);
|
|
|
|
ASSERT((map.br_startblock != DELAYSTARTBLOCK) &&
|
|
|
|
(map.br_startblock != HOLESTARTBLOCK));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Keep track of the blkno to save a lookup later
|
|
|
|
*/
|
|
|
|
dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
|
|
|
|
|
|
|
|
/* now we can just get the buffer (there's nothing to read yet) */
|
2018-07-12 12:26:11 +07:00
|
|
|
bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, dqp->q_blkno,
|
2018-05-05 05:30:23 +07:00
|
|
|
mp->m_quotainfo->qi_dqchunklen, 0);
|
2018-08-08 00:57:13 +07:00
|
|
|
if (!bp)
|
|
|
|
return -ENOMEM;
|
2012-11-14 13:54:40 +07:00
|
|
|
bp->b_ops = &xfs_dquot_buf_ops;
|
2011-09-20 20:56:55 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Make a chunk of dquots out of this buffer and log
|
|
|
|
* the entire thing.
|
|
|
|
*/
|
2018-07-12 12:26:11 +07:00
|
|
|
xfs_qm_init_dquot_blk(tp, mp, be32_to_cpu(dqp->q_core.d_id),
|
2005-06-21 12:38:48 +07:00
|
|
|
dqp->dq_flags & XFS_DQ_ALLTYPES, bp);
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_buf_set_ref(bp, XFS_DQUOT_REF);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-09-05 05:29:01 +07:00
|
|
|
/*
|
xfs: release new dquot buffer on defer_finish error
In commit efa092f3d4c6 "[XFS] Fixes a bug in the quota code when
allocating a new dquot record", we allocate a new dquot block, grab a
buffer to initialize it, and return the locked initialized dquot buffer
to the caller for further in-core dquot initialization. Unfortunately,
if the _bmap_finish errored out, _qm_dqalloc would also error out
without bothering to free the (locked) buffer. Leaking a locked buffer
caused hangs in generic/388 when quotas are enabled.
Furthermore, the _bmap_finish -> _defer_finish conversion in
310a75a3c6c747 ("xfs: change xfs_bmap_{finish,cancel,init,free} ->
xfs_defer_*") failed to observe that the buffer was held going into
_defer_finish and therefore failed to notice that the buffer lock is
/not/ maintained afterwards. Now that we can bjoin a buffer to a
defer_ops, use this mechanism to ensure that the buffer stays locked
across the _defer_finish. Release the holds and locks on the buffer as
appropriate if we have to error out.
There is a subtlety here for the caller in that the buffer emerges
locked and held to the transaction, so if the _trans_commit fails we
have to release the buffer explicitly. This fixes the unmount hang
in generic/388 when quotas are enabled.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-05-05 05:30:19 +07:00
|
|
|
* Hold the buffer and join it to the dfops so that we'll still own
|
|
|
|
* the buffer when we return to the caller. The buffer disposal on
|
|
|
|
* error must be paid attention to very carefully, as it has been
|
|
|
|
* broken since commit efa092f3d4c6 "[XFS] Fixes a bug in the quota
|
|
|
|
* code when allocating a new dquot record" in 2005, and the later
|
|
|
|
* conversion to xfs_defer_ops in commit 310a75a3c6c747 failed to keep
|
|
|
|
* the buffer locked across the _defer_finish call. We can now do
|
|
|
|
* this correctly with xfs_defer_bjoin.
|
2005-09-05 05:29:01 +07:00
|
|
|
*
|
2018-08-08 00:57:13 +07:00
|
|
|
* Above, we allocated a disk block for the dquot information and used
|
|
|
|
* get_buf to initialize the dquot. If the _defer_finish fails, the old
|
xfs: release new dquot buffer on defer_finish error
In commit efa092f3d4c6 "[XFS] Fixes a bug in the quota code when
allocating a new dquot record", we allocate a new dquot block, grab a
buffer to initialize it, and return the locked initialized dquot buffer
to the caller for further in-core dquot initialization. Unfortunately,
if the _bmap_finish errored out, _qm_dqalloc would also error out
without bothering to free the (locked) buffer. Leaking a locked buffer
caused hangs in generic/388 when quotas are enabled.
Furthermore, the _bmap_finish -> _defer_finish conversion in
310a75a3c6c747 ("xfs: change xfs_bmap_{finish,cancel,init,free} ->
xfs_defer_*") failed to observe that the buffer was held going into
_defer_finish and therefore failed to notice that the buffer lock is
/not/ maintained afterwards. Now that we can bjoin a buffer to a
defer_ops, use this mechanism to ensure that the buffer stays locked
across the _defer_finish. Release the holds and locks on the buffer as
appropriate if we have to error out.
There is a subtlety here for the caller in that the buffer emerges
locked and held to the transaction, so if the _trans_commit fails we
have to release the buffer explicitly. This fixes the unmount hang
in generic/388 when quotas are enabled.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-05-05 05:30:19 +07:00
|
|
|
* transaction is gone but the new buffer is not joined or held to any
|
|
|
|
* transaction, so we must _buf_relse it.
|
2005-09-05 05:29:01 +07:00
|
|
|
*
|
xfs: release new dquot buffer on defer_finish error
In commit efa092f3d4c6 "[XFS] Fixes a bug in the quota code when
allocating a new dquot record", we allocate a new dquot block, grab a
buffer to initialize it, and return the locked initialized dquot buffer
to the caller for further in-core dquot initialization. Unfortunately,
if the _bmap_finish errored out, _qm_dqalloc would also error out
without bothering to free the (locked) buffer. Leaking a locked buffer
caused hangs in generic/388 when quotas are enabled.
Furthermore, the _bmap_finish -> _defer_finish conversion in
310a75a3c6c747 ("xfs: change xfs_bmap_{finish,cancel,init,free} ->
xfs_defer_*") failed to observe that the buffer was held going into
_defer_finish and therefore failed to notice that the buffer lock is
/not/ maintained afterwards. Now that we can bjoin a buffer to a
defer_ops, use this mechanism to ensure that the buffer stays locked
across the _defer_finish. Release the holds and locks on the buffer as
appropriate if we have to error out.
There is a subtlety here for the caller in that the buffer emerges
locked and held to the transaction, so if the _trans_commit fails we
have to release the buffer explicitly. This fixes the unmount hang
in generic/388 when quotas are enabled.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-05-05 05:30:19 +07:00
|
|
|
* If everything succeeds, the caller of this function is returned a
|
2018-05-05 05:30:23 +07:00
|
|
|
* buffer that is locked and held to the transaction. The caller
|
xfs: release new dquot buffer on defer_finish error
In commit efa092f3d4c6 "[XFS] Fixes a bug in the quota code when
allocating a new dquot record", we allocate a new dquot block, grab a
buffer to initialize it, and return the locked initialized dquot buffer
to the caller for further in-core dquot initialization. Unfortunately,
if the _bmap_finish errored out, _qm_dqalloc would also error out
without bothering to free the (locked) buffer. Leaking a locked buffer
caused hangs in generic/388 when quotas are enabled.
Furthermore, the _bmap_finish -> _defer_finish conversion in
310a75a3c6c747 ("xfs: change xfs_bmap_{finish,cancel,init,free} ->
xfs_defer_*") failed to observe that the buffer was held going into
_defer_finish and therefore failed to notice that the buffer lock is
/not/ maintained afterwards. Now that we can bjoin a buffer to a
defer_ops, use this mechanism to ensure that the buffer stays locked
across the _defer_finish. Release the holds and locks on the buffer as
appropriate if we have to error out.
There is a subtlety here for the caller in that the buffer emerges
locked and held to the transaction, so if the _trans_commit fails we
have to release the buffer explicitly. This fixes the unmount hang
in generic/388 when quotas are enabled.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-05-05 05:30:19 +07:00
|
|
|
* is responsible for unlocking any buffer passed back, either
|
|
|
|
* manually or by committing the transaction.
|
2005-09-05 05:29:01 +07:00
|
|
|
*/
|
2018-07-12 12:26:11 +07:00
|
|
|
xfs_trans_bhold(tp, bp);
|
2018-07-25 03:43:15 +07:00
|
|
|
error = xfs_defer_finish(tpp);
|
2018-07-12 12:26:11 +07:00
|
|
|
tp = *tpp;
|
xfs: release new dquot buffer on defer_finish error
In commit efa092f3d4c6 "[XFS] Fixes a bug in the quota code when
allocating a new dquot record", we allocate a new dquot block, grab a
buffer to initialize it, and return the locked initialized dquot buffer
to the caller for further in-core dquot initialization. Unfortunately,
if the _bmap_finish errored out, _qm_dqalloc would also error out
without bothering to free the (locked) buffer. Leaking a locked buffer
caused hangs in generic/388 when quotas are enabled.
Furthermore, the _bmap_finish -> _defer_finish conversion in
310a75a3c6c747 ("xfs: change xfs_bmap_{finish,cancel,init,free} ->
xfs_defer_*") failed to observe that the buffer was held going into
_defer_finish and therefore failed to notice that the buffer lock is
/not/ maintained afterwards. Now that we can bjoin a buffer to a
defer_ops, use this mechanism to ensure that the buffer stays locked
across the _defer_finish. Release the holds and locks on the buffer as
appropriate if we have to error out.
There is a subtlety here for the caller in that the buffer emerges
locked and held to the transaction, so if the _trans_commit fails we
have to release the buffer explicitly. This fixes the unmount hang
in generic/388 when quotas are enabled.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-05-05 05:30:19 +07:00
|
|
|
if (error) {
|
|
|
|
xfs_buf_relse(bp);
|
2018-08-08 00:57:13 +07:00
|
|
|
return error;
|
2005-09-05 05:29:01 +07:00
|
|
|
}
|
2018-05-05 05:30:23 +07:00
|
|
|
*bpp = bp;
|
2005-04-17 05:20:36 +07:00
|
|
|
return 0;
|
|
|
|
}
|
2013-10-15 05:17:52 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2018-05-05 05:30:23 +07:00
|
|
|
* Read in the in-core dquot's on-disk metadata and return the buffer.
|
|
|
|
* Returns ENOENT to signal a hole.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
STATIC int
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_dquot_disk_read(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_dquot *dqp,
|
|
|
|
struct xfs_buf **bpp)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2013-06-28 05:25:07 +07:00
|
|
|
struct xfs_bmbt_irec map;
|
|
|
|
struct xfs_buf *bp;
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_inode *quotip = xfs_quota_inode(mp, dqp->dq_flags);
|
2017-07-14 02:14:34 +07:00
|
|
|
uint lock_mode;
|
2018-05-05 05:30:23 +07:00
|
|
|
int nmaps = 1;
|
|
|
|
int error;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2017-07-14 02:14:34 +07:00
|
|
|
lock_mode = xfs_ilock_data_map_shared(quotip);
|
2018-05-05 05:30:23 +07:00
|
|
|
if (!xfs_this_quota_on(mp, dqp->dq_flags)) {
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2010-09-06 08:44:45 +07:00
|
|
|
* Return if this type of quotas is turned off while we
|
|
|
|
* didn't have the quota inode lock.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2017-07-14 02:14:34 +07:00
|
|
|
xfs_iunlock(quotip, lock_mode);
|
2014-06-25 11:58:08 +07:00
|
|
|
return -ESRCH;
|
2010-09-06 08:44:45 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find the block map; no allocations yet
|
|
|
|
*/
|
2011-09-19 03:40:45 +07:00
|
|
|
error = xfs_bmapi_read(quotip, dqp->q_fileoffset,
|
2018-05-05 05:30:23 +07:00
|
|
|
XFS_DQUOT_CLUSTER_SIZE_FSB, &map, &nmaps, 0);
|
2017-07-14 02:14:34 +07:00
|
|
|
xfs_iunlock(quotip, lock_mode);
|
2010-09-06 08:44:45 +07:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
ASSERT(nmaps == 1);
|
2018-05-05 05:30:23 +07:00
|
|
|
ASSERT(map.br_blockcount >= 1);
|
|
|
|
ASSERT(map.br_startblock != DELAYSTARTBLOCK);
|
|
|
|
if (map.br_startblock == HOLESTARTBLOCK)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
trace_xfs_dqtobp_read(dqp);
|
2010-09-06 08:44:45 +07:00
|
|
|
|
|
|
|
/*
|
2018-05-05 05:30:23 +07:00
|
|
|
* store the blkno etc so that we don't have to do the
|
|
|
|
* mapping all the time
|
2010-09-06 08:44:45 +07:00
|
|
|
*/
|
2018-05-05 05:30:23 +07:00
|
|
|
dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
|
|
|
|
mp->m_quotainfo->qi_dqchunklen, 0, &bp,
|
|
|
|
&xfs_dquot_buf_ops);
|
|
|
|
if (error) {
|
|
|
|
ASSERT(bp == NULL);
|
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2012-11-14 13:50:13 +07:00
|
|
|
ASSERT(xfs_buf_islocked(bp));
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_buf_set_ref(bp, XFS_DQUOT_REF);
|
|
|
|
*bpp = bp;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2014-06-22 12:03:54 +07:00
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/* Allocate and initialize everything we need for an incore dquot. */
|
|
|
|
STATIC struct xfs_dquot *
|
|
|
|
xfs_dquot_alloc(
|
2011-12-07 04:58:24 +07:00
|
|
|
struct xfs_mount *mp,
|
|
|
|
xfs_dqid_t id,
|
2018-05-05 05:30:23 +07:00
|
|
|
uint type)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2011-12-07 04:58:24 +07:00
|
|
|
struct xfs_dquot *dqp;
|
2012-02-01 20:57:20 +07:00
|
|
|
|
2012-03-13 15:52:37 +07:00
|
|
|
dqp = kmem_zone_zalloc(xfs_qm_dqzone, KM_SLEEP);
|
2012-02-01 20:57:20 +07:00
|
|
|
|
|
|
|
dqp->dq_flags = type;
|
|
|
|
dqp->q_core.d_id = cpu_to_be32(id);
|
|
|
|
dqp->q_mount = mp;
|
2012-03-13 15:52:34 +07:00
|
|
|
INIT_LIST_HEAD(&dqp->q_lru);
|
2012-02-01 20:57:20 +07:00
|
|
|
mutex_init(&dqp->q_qlock);
|
|
|
|
init_waitqueue_head(&dqp->q_pinwait);
|
2018-05-05 05:30:23 +07:00
|
|
|
dqp->q_fileoffset = (xfs_fileoff_t)id / mp->m_quotainfo->qi_dqperchunk;
|
|
|
|
/*
|
|
|
|
* Offset of dquot in the (fixed sized) dquot chunk.
|
|
|
|
*/
|
|
|
|
dqp->q_bufoffset = (id % mp->m_quotainfo->qi_dqperchunk) *
|
|
|
|
sizeof(xfs_dqblk_t);
|
2012-02-01 20:57:20 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Because we want to use a counting completion, complete
|
|
|
|
* the flush completion once to allow a single access to
|
|
|
|
* the flush completion without blocking.
|
|
|
|
*/
|
|
|
|
init_completion(&dqp->q_flush);
|
|
|
|
complete(&dqp->q_flush);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure group quotas have a different lock class than user
|
|
|
|
* quotas.
|
|
|
|
*/
|
xfs: lockdep needs to know about 3 dquot-deep nesting
Michael Semon reported that xfs/299 generated this lockdep warning:
=============================================
[ INFO: possible recursive locking detected ]
3.12.0-rc2+ #2 Not tainted
---------------------------------------------
touch/21072 is trying to acquire lock:
(&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
but task is already holding lock:
(&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&xfs_dquot_other_class);
lock(&xfs_dquot_other_class);
*** DEADLOCK ***
May be due to missing lock nesting notation
7 locks held by touch/21072:
#0: (sb_writers#10){++++.+}, at: [<c11185b6>] mnt_want_write+0x1e/0x3e
#1: (&type->i_mutex_dir_key#4){+.+.+.}, at: [<c11078ee>] do_last+0x245/0xe40
#2: (sb_internal#2){++++.+}, at: [<c122c9e0>] xfs_trans_alloc+0x1f/0x35
#3: (&(&ip->i_lock)->mr_lock/1){+.+...}, at: [<c126cd1b>] xfs_ilock+0x100/0x1f1
#4: (&(&ip->i_lock)->mr_lock){++++-.}, at: [<c126cf52>] xfs_ilock_nowait+0x105/0x22f
#5: (&dqp->q_qlock){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
#6: (&xfs_dquot_other_class){+.+...}, at: [<c12902fb>] xfs_trans_dqlockedjoin+0x57/0x64
The lockdep annotation for dquot lock nesting only understands
locking for user and "other" dquots, not user, group and quota
dquots. Fix the annotations to match the locking heirarchy we now
have.
Reported-by: Michael L. Semon <mlsemon35@gmail.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-09-30 06:37:03 +07:00
|
|
|
switch (type) {
|
|
|
|
case XFS_DQ_USER:
|
|
|
|
/* uses the default lock class */
|
|
|
|
break;
|
|
|
|
case XFS_DQ_GROUP:
|
|
|
|
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_group_class);
|
|
|
|
break;
|
|
|
|
case XFS_DQ_PROJ:
|
|
|
|
lockdep_set_class(&dqp->q_qlock, &xfs_dquot_project_class);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
ASSERT(0);
|
|
|
|
break;
|
|
|
|
}
|
2012-02-01 20:57:20 +07:00
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_qm_dquot_logitem_init(dqp);
|
|
|
|
|
2015-10-12 14:21:22 +07:00
|
|
|
XFS_STATS_INC(mp, xs_qm_dquot);
|
2018-05-05 05:30:23 +07:00
|
|
|
return dqp;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy the in-core quota fields in from the on-disk buffer. */
|
|
|
|
STATIC void
|
|
|
|
xfs_dquot_from_disk(
|
|
|
|
struct xfs_dquot *dqp,
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_buf *bp)
|
2018-05-05 05:30:23 +07:00
|
|
|
{
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_disk_dquot *ddqp = bp->b_addr + dqp->q_bufoffset;
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/* copy everything from disk dquot to the incore dquot */
|
|
|
|
memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reservation counters are defined as reservation plus current usage
|
|
|
|
* to avoid having to add every time.
|
|
|
|
*/
|
|
|
|
dqp->q_res_bcount = be64_to_cpu(ddqp->d_bcount);
|
|
|
|
dqp->q_res_icount = be64_to_cpu(ddqp->d_icount);
|
|
|
|
dqp->q_res_rtbcount = be64_to_cpu(ddqp->d_rtbcount);
|
|
|
|
|
|
|
|
/* initialize the dquot speculative prealloc thresholds */
|
|
|
|
xfs_dquot_set_prealloc_limits(dqp);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/* Allocate and initialize the dquot buffer for this in-core dquot. */
|
|
|
|
static int
|
|
|
|
xfs_qm_dqread_alloc(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_dquot *dqp,
|
|
|
|
struct xfs_buf **bpp)
|
|
|
|
{
|
|
|
|
struct xfs_trans *tp;
|
|
|
|
struct xfs_buf *bp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_qm_dqalloc,
|
|
|
|
XFS_QM_DQALLOC_SPACE_RES(mp), 0, 0, &tp);
|
|
|
|
if (error)
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
error = xfs_dquot_disk_alloc(&tp, dqp, &bp);
|
|
|
|
if (error)
|
|
|
|
goto err_cancel;
|
|
|
|
|
|
|
|
error = xfs_trans_commit(tp);
|
|
|
|
if (error) {
|
|
|
|
/*
|
|
|
|
* Buffer was held to the transaction, so we have to unlock it
|
|
|
|
* manually here because we're not passing it back.
|
|
|
|
*/
|
|
|
|
xfs_buf_relse(bp);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
*bpp = bp;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err_cancel:
|
|
|
|
xfs_trans_cancel(tp);
|
|
|
|
err:
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/*
|
|
|
|
* Read in the ondisk dquot using dqtobp() then copy it to an incore version,
|
2018-05-05 05:30:24 +07:00
|
|
|
* and release the buffer immediately. If @can_alloc is true, fill any
|
|
|
|
* holes in the on-disk metadata.
|
2018-05-05 05:30:23 +07:00
|
|
|
*/
|
2018-05-05 05:30:23 +07:00
|
|
|
static int
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_qm_dqread(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
uint type,
|
2018-05-05 05:30:24 +07:00
|
|
|
bool can_alloc,
|
2018-05-05 05:30:23 +07:00
|
|
|
struct xfs_dquot **dqpp)
|
2018-05-05 05:30:23 +07:00
|
|
|
{
|
|
|
|
struct xfs_dquot *dqp;
|
|
|
|
struct xfs_buf *bp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
dqp = xfs_dquot_alloc(mp, id, type);
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqread(dqp);
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/* Try to read the buffer, allocating if necessary. */
|
|
|
|
error = xfs_dquot_disk_read(mp, dqp, &bp);
|
2018-05-05 05:30:24 +07:00
|
|
|
if (error == -ENOENT && can_alloc)
|
2018-05-05 05:30:23 +07:00
|
|
|
error = xfs_qm_dqread_alloc(mp, dqp, &bp);
|
|
|
|
if (error)
|
|
|
|
goto err;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
2018-05-05 05:30:23 +07:00
|
|
|
* At this point we should have a clean locked buffer. Copy the data
|
|
|
|
* to the incore dquot and release the buffer since the incore dquot
|
|
|
|
* has its own locking protocol so we needn't tie up the buffer any
|
|
|
|
* further.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2011-07-08 19:36:19 +07:00
|
|
|
ASSERT(xfs_buf_islocked(bp));
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_dquot_from_disk(dqp, bp);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
xfs_buf_relse(bp);
|
|
|
|
*dqpp = dqp;
|
2011-12-07 04:58:24 +07:00
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
err:
|
|
|
|
trace_xfs_dqread_fail(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_qm_dqdestroy(dqp);
|
2018-05-05 05:30:23 +07:00
|
|
|
*dqpp = NULL;
|
2011-12-07 04:58:24 +07:00
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2016-02-08 07:27:38 +07:00
|
|
|
/*
|
|
|
|
* Advance to the next id in the current chunk, or if at the
|
|
|
|
* end of the chunk, skip ahead to first id in next allocated chunk
|
|
|
|
* using the SEEK_DATA interface.
|
|
|
|
*/
|
2016-04-06 06:47:21 +07:00
|
|
|
static int
|
2016-02-08 07:27:38 +07:00
|
|
|
xfs_dq_get_next_id(
|
2017-06-30 02:28:36 +07:00
|
|
|
struct xfs_mount *mp,
|
2016-02-08 07:27:38 +07:00
|
|
|
uint type,
|
2017-06-30 02:28:36 +07:00
|
|
|
xfs_dqid_t *id)
|
2016-02-08 07:27:38 +07:00
|
|
|
{
|
2017-06-30 02:28:36 +07:00
|
|
|
struct xfs_inode *quotip = xfs_quota_inode(mp, type);
|
|
|
|
xfs_dqid_t next_id = *id + 1; /* simple advance */
|
|
|
|
uint lock_flags;
|
|
|
|
struct xfs_bmbt_irec got;
|
2017-11-04 00:34:43 +07:00
|
|
|
struct xfs_iext_cursor cur;
|
2016-02-08 07:27:38 +07:00
|
|
|
xfs_fsblock_t start;
|
|
|
|
int error = 0;
|
|
|
|
|
2017-01-18 02:43:38 +07:00
|
|
|
/* If we'd wrap past the max ID, stop */
|
|
|
|
if (next_id < *id)
|
|
|
|
return -ENOENT;
|
|
|
|
|
2016-02-08 07:27:38 +07:00
|
|
|
/* If new ID is within the current chunk, advancing it sufficed */
|
|
|
|
if (next_id % mp->m_quotainfo->qi_dqperchunk) {
|
|
|
|
*id = next_id;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Nope, next_id is now past the current chunk, so find the next one */
|
|
|
|
start = (xfs_fsblock_t)next_id / mp->m_quotainfo->qi_dqperchunk;
|
|
|
|
|
2017-06-30 02:28:36 +07:00
|
|
|
lock_flags = xfs_ilock_data_map_shared(quotip);
|
|
|
|
if (!(quotip->i_df.if_flags & XFS_IFEXTENTS)) {
|
|
|
|
error = xfs_iread_extents(NULL, quotip, XFS_DATA_FORK);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
2016-02-08 07:27:38 +07:00
|
|
|
|
2017-11-04 00:34:43 +07:00
|
|
|
if (xfs_iext_lookup_extent(quotip, "ip->i_df, start, &cur, &got)) {
|
2017-07-06 02:07:52 +07:00
|
|
|
/* contiguous chunk, bump startoff for the id calculation */
|
|
|
|
if (got.br_startoff < start)
|
|
|
|
got.br_startoff = start;
|
2017-06-30 02:28:36 +07:00
|
|
|
*id = got.br_startoff * mp->m_quotainfo->qi_dqperchunk;
|
2017-07-06 02:07:52 +07:00
|
|
|
} else {
|
2017-06-30 02:28:36 +07:00
|
|
|
error = -ENOENT;
|
2017-07-06 02:07:52 +07:00
|
|
|
}
|
|
|
|
|
2017-06-30 02:28:36 +07:00
|
|
|
xfs_iunlock(quotip, lock_flags);
|
2016-02-08 07:27:38 +07:00
|
|
|
|
2017-06-30 02:28:36 +07:00
|
|
|
return error;
|
2016-02-08 07:27:38 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:20 +07:00
|
|
|
/*
|
|
|
|
* Look up the dquot in the in-core cache. If found, the dquot is returned
|
|
|
|
* locked and ready to go.
|
|
|
|
*/
|
|
|
|
static struct xfs_dquot *
|
|
|
|
xfs_qm_dqget_cache_lookup(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_quotainfo *qi,
|
|
|
|
struct radix_tree_root *tree,
|
|
|
|
xfs_dqid_t id)
|
|
|
|
{
|
|
|
|
struct xfs_dquot *dqp;
|
|
|
|
|
|
|
|
restart:
|
|
|
|
mutex_lock(&qi->qi_tree_lock);
|
|
|
|
dqp = radix_tree_lookup(tree, id);
|
|
|
|
if (!dqp) {
|
|
|
|
mutex_unlock(&qi->qi_tree_lock);
|
|
|
|
XFS_STATS_INC(mp, xs_qm_dqcachemisses);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_dqlock(dqp);
|
|
|
|
if (dqp->dq_flags & XFS_DQ_FREEING) {
|
|
|
|
xfs_dqunlock(dqp);
|
|
|
|
mutex_unlock(&qi->qi_tree_lock);
|
|
|
|
trace_xfs_dqget_freeing(dqp);
|
|
|
|
delay(1);
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
|
|
|
dqp->q_nrefs++;
|
|
|
|
mutex_unlock(&qi->qi_tree_lock);
|
|
|
|
|
|
|
|
trace_xfs_dqget_hit(dqp);
|
|
|
|
XFS_STATS_INC(mp, xs_qm_dqcachehits);
|
|
|
|
return dqp;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Try to insert a new dquot into the in-core cache. If an error occurs the
|
|
|
|
* caller should throw away the dquot and start over. Otherwise, the dquot
|
|
|
|
* is returned locked (and held by the cache) as if there had been a cache
|
|
|
|
* hit.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_qm_dqget_cache_insert(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
struct xfs_quotainfo *qi,
|
|
|
|
struct radix_tree_root *tree,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
struct xfs_dquot *dqp)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
mutex_lock(&qi->qi_tree_lock);
|
|
|
|
error = radix_tree_insert(tree, id, dqp);
|
|
|
|
if (unlikely(error)) {
|
|
|
|
/* Duplicate found! Caller must try again. */
|
|
|
|
WARN_ON(error != -EEXIST);
|
|
|
|
mutex_unlock(&qi->qi_tree_lock);
|
|
|
|
trace_xfs_dqget_dup(dqp);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Return a locked dquot to the caller, with a reference taken. */
|
|
|
|
xfs_dqlock(dqp);
|
|
|
|
dqp->q_nrefs = 1;
|
|
|
|
|
|
|
|
qi->qi_dquots++;
|
|
|
|
mutex_unlock(&qi->qi_tree_lock);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:21 +07:00
|
|
|
/* Check our input parameters. */
|
|
|
|
static int
|
|
|
|
xfs_qm_dqget_checks(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
uint type)
|
|
|
|
{
|
|
|
|
if (WARN_ON_ONCE(!XFS_IS_QUOTA_RUNNING(mp)))
|
|
|
|
return -ESRCH;
|
|
|
|
|
|
|
|
switch (type) {
|
|
|
|
case XFS_DQ_USER:
|
|
|
|
if (!XFS_IS_UQUOTA_ON(mp))
|
|
|
|
return -ESRCH;
|
|
|
|
return 0;
|
|
|
|
case XFS_DQ_GROUP:
|
|
|
|
if (!XFS_IS_GQUOTA_ON(mp))
|
|
|
|
return -ESRCH;
|
|
|
|
return 0;
|
|
|
|
case XFS_DQ_PROJ:
|
|
|
|
if (!XFS_IS_PQUOTA_ON(mp))
|
|
|
|
return -ESRCH;
|
|
|
|
return 0;
|
|
|
|
default:
|
|
|
|
WARN_ON_ONCE(0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2018-05-05 05:30:22 +07:00
|
|
|
* Given the file system, id, and type (UDQUOT/GDQUOT), return a a locked
|
|
|
|
* dquot, doing an allocation (if requested) as needed.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqget(
|
2018-05-05 05:30:22 +07:00
|
|
|
struct xfs_mount *mp,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
uint type,
|
2018-05-05 05:30:24 +07:00
|
|
|
bool can_alloc,
|
2018-05-05 05:30:22 +07:00
|
|
|
struct xfs_dquot **O_dqpp)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2012-03-13 15:52:35 +07:00
|
|
|
struct xfs_quotainfo *qi = mp->m_quotainfo;
|
2018-05-05 05:30:22 +07:00
|
|
|
struct radix_tree_root *tree = xfs_dquot_tree(qi, type);
|
2012-03-13 15:52:35 +07:00
|
|
|
struct xfs_dquot *dqp;
|
|
|
|
int error;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:21 +07:00
|
|
|
error = xfs_qm_dqget_checks(mp, type);
|
|
|
|
if (error)
|
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:22 +07:00
|
|
|
restart:
|
|
|
|
dqp = xfs_qm_dqget_cache_lookup(mp, qi, tree, id);
|
|
|
|
if (dqp) {
|
|
|
|
*O_dqpp = dqp;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:24 +07:00
|
|
|
error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp);
|
2018-05-05 05:30:22 +07:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
error = xfs_qm_dqget_cache_insert(mp, qi, tree, id, dqp);
|
|
|
|
if (error) {
|
|
|
|
/*
|
|
|
|
* Duplicate found. Just throw away the new dquot and start
|
|
|
|
* over.
|
|
|
|
*/
|
|
|
|
xfs_qm_dqdestroy(dqp);
|
|
|
|
XFS_STATS_INC(mp, xs_qm_dquot_dups);
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_xfs_dqget_miss(dqp);
|
|
|
|
*O_dqpp = dqp;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:23 +07:00
|
|
|
/*
|
|
|
|
* Given a dquot id and type, read and initialize a dquot from the on-disk
|
|
|
|
* metadata. This function is only for use during quota initialization so
|
|
|
|
* it ignores the dquot cache assuming that the dquot shrinker isn't set up.
|
|
|
|
* The caller is responsible for _qm_dqdestroy'ing the returned dquot.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqget_uncached(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
uint type,
|
|
|
|
struct xfs_dquot **dqpp)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = xfs_qm_dqget_checks(mp, type);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
return xfs_qm_dqread(mp, id, type, 0, dqpp);
|
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:22 +07:00
|
|
|
/* Return the quota id for a given inode and type. */
|
|
|
|
xfs_dqid_t
|
|
|
|
xfs_qm_id_for_quotatype(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
uint type)
|
|
|
|
{
|
|
|
|
switch (type) {
|
|
|
|
case XFS_DQ_USER:
|
|
|
|
return ip->i_d.di_uid;
|
|
|
|
case XFS_DQ_GROUP:
|
|
|
|
return ip->i_d.di_gid;
|
|
|
|
case XFS_DQ_PROJ:
|
|
|
|
return xfs_get_projid(ip);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2018-05-05 05:30:22 +07:00
|
|
|
ASSERT(0);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Return the dquot for a given inode and type. If @can_alloc is true, then
|
|
|
|
* allocate blocks if needed. The inode's ILOCK must be held and it must not
|
|
|
|
* have already had an inode attached.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqget_inode(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
uint type,
|
|
|
|
bool can_alloc,
|
|
|
|
struct xfs_dquot **O_dqpp)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
struct xfs_quotainfo *qi = mp->m_quotainfo;
|
|
|
|
struct radix_tree_root *tree = xfs_dquot_tree(qi, type);
|
|
|
|
struct xfs_dquot *dqp;
|
|
|
|
xfs_dqid_t id;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
error = xfs_qm_dqget_checks(mp, type);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
|
|
|
|
ASSERT(xfs_inode_dquot(ip, type) == NULL);
|
|
|
|
|
|
|
|
id = xfs_qm_id_for_quotatype(ip, type);
|
2011-12-07 04:58:18 +07:00
|
|
|
|
|
|
|
restart:
|
2018-05-05 05:30:20 +07:00
|
|
|
dqp = xfs_qm_dqget_cache_lookup(mp, qi, tree, id);
|
2012-03-13 15:52:35 +07:00
|
|
|
if (dqp) {
|
|
|
|
*O_dqpp = dqp;
|
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Dquot cache miss. We don't want to keep the inode lock across
|
|
|
|
* a (potential) disk read. Also we don't want to deal with the lock
|
|
|
|
* ordering between quotainode and this inode. OTOH, dropping the inode
|
|
|
|
* lock here means dealing with a chown that can happen before
|
|
|
|
* we re-acquire the lock.
|
|
|
|
*/
|
2018-05-05 05:30:22 +07:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2018-05-05 05:30:24 +07:00
|
|
|
error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp);
|
2018-05-05 05:30:22 +07:00
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2011-12-07 04:58:25 +07:00
|
|
|
if (error)
|
|
|
|
return error;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-05-05 05:30:22 +07:00
|
|
|
/*
|
|
|
|
* A dquot could be attached to this inode by now, since we had
|
|
|
|
* dropped the ilock.
|
|
|
|
*/
|
|
|
|
if (xfs_this_quota_on(mp, type)) {
|
|
|
|
struct xfs_dquot *dqp1;
|
|
|
|
|
|
|
|
dqp1 = xfs_inode_dquot(ip, type);
|
|
|
|
if (dqp1) {
|
2012-01-24 00:31:30 +07:00
|
|
|
xfs_qm_dqdestroy(dqp);
|
2018-05-05 05:30:22 +07:00
|
|
|
dqp = dqp1;
|
|
|
|
xfs_dqlock(dqp);
|
|
|
|
goto dqret;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2018-05-05 05:30:22 +07:00
|
|
|
} else {
|
|
|
|
/* inode stays locked on return */
|
|
|
|
xfs_qm_dqdestroy(dqp);
|
|
|
|
return -ESRCH;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:20 +07:00
|
|
|
error = xfs_qm_dqget_cache_insert(mp, qi, tree, id, dqp);
|
|
|
|
if (error) {
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2012-03-13 15:52:35 +07:00
|
|
|
* Duplicate found. Just throw away the new dquot and start
|
|
|
|
* over.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2012-03-13 15:52:35 +07:00
|
|
|
xfs_qm_dqdestroy(dqp);
|
2015-10-12 14:21:22 +07:00
|
|
|
XFS_STATS_INC(mp, xs_qm_dquot_dups);
|
2012-03-13 15:52:35 +07:00
|
|
|
goto restart;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:22 +07:00
|
|
|
dqret:
|
|
|
|
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqget_miss(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
*O_dqpp = dqp;
|
2014-06-22 12:03:54 +07:00
|
|
|
return 0;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2018-05-05 05:30:20 +07:00
|
|
|
/*
|
|
|
|
* Starting at @id and progressing upwards, look for an initialized incore
|
|
|
|
* dquot, lock it, and return it.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqget_next(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
xfs_dqid_t id,
|
|
|
|
uint type,
|
|
|
|
struct xfs_dquot **dqpp)
|
|
|
|
{
|
|
|
|
struct xfs_dquot *dqp;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
*dqpp = NULL;
|
|
|
|
for (; !error; error = xfs_dq_get_next_id(mp, type, &id)) {
|
2018-05-05 05:30:24 +07:00
|
|
|
error = xfs_qm_dqget(mp, id, type, false, &dqp);
|
2018-05-05 05:30:20 +07:00
|
|
|
if (error == -ENOENT)
|
|
|
|
continue;
|
|
|
|
else if (error != 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (!XFS_IS_DQUOT_UNINITIALIZED(dqp)) {
|
|
|
|
*dqpp = dqp;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_qm_dqput(dqp);
|
|
|
|
}
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2012-03-13 15:52:34 +07:00
|
|
|
/*
|
|
|
|
* Release a reference to the dquot (decrement ref-count) and unlock it.
|
|
|
|
*
|
|
|
|
* If there is a group quota attached to this dquot, carefully release that
|
|
|
|
* too without tripping over deadlocks'n'stuff.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_qm_dqput(
|
|
|
|
struct xfs_dquot *dqp)
|
|
|
|
{
|
|
|
|
ASSERT(dqp->q_nrefs > 0);
|
|
|
|
ASSERT(XFS_DQ_IS_LOCKED(dqp));
|
|
|
|
|
|
|
|
trace_xfs_dqput(dqp);
|
|
|
|
|
xfs: remove dquot hints
group and project quota hints are currently stored on the user
dquot. If we are attaching quotas to the inode, then the group and
project dquots are stored as hints on the user dquot to save having
to look them up again later.
The thing is, the hints are not used for that inode for the rest of
the life of the inode - the dquots are attached directly to the
inode itself - so the only time the hints are used is when an inode
first has dquots attached.
When the hints on the user dquot don't match the dquots being
attache dto the inode, they are then removed and replaced with the
new hints. If a user is concurrently modifying files in different
group and/or project contexts, then this leads to thrashing of the
hints attached to user dquot.
If user quotas are not enabled, then hints are never even used.
So, if the hints are used to avoid the cost of the lookup, is the
cost of the lookup significant enough to justify the hint
infrstructure? Maybe it was once, when there was a global quota
manager shared between all XFS filesystems and was hash table based.
However, lookups are now much simpler, requiring only a single lock and
radix tree lookup local to the filesystem and no hash or LRU
manipulations to be made. Hence the cost of lookup is much lower
than when hints were implemented. Turns out that benchmarks show
that, too, with thir being no differnce in performance when doing
file creation workloads as a single user with user, group and
project quotas enabled - the hints do not make the code go any
faster. In fact, removing the hints shows a 2-3% reduction in the
time it takes to create 50 million inodes....
So, let's just get rid of the hints and the complexity around them.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2014-05-05 14:30:15 +07:00
|
|
|
if (--dqp->q_nrefs == 0) {
|
|
|
|
struct xfs_quotainfo *qi = dqp->q_mount->m_quotainfo;
|
|
|
|
trace_xfs_dqput_free(dqp);
|
|
|
|
|
|
|
|
if (list_lru_add(&qi->qi_lru, &dqp->q_lru))
|
2015-10-12 14:21:22 +07:00
|
|
|
XFS_STATS_INC(dqp->q_mount, xs_qm_dquot_unused);
|
xfs: remove dquot hints
group and project quota hints are currently stored on the user
dquot. If we are attaching quotas to the inode, then the group and
project dquots are stored as hints on the user dquot to save having
to look them up again later.
The thing is, the hints are not used for that inode for the rest of
the life of the inode - the dquots are attached directly to the
inode itself - so the only time the hints are used is when an inode
first has dquots attached.
When the hints on the user dquot don't match the dquots being
attache dto the inode, they are then removed and replaced with the
new hints. If a user is concurrently modifying files in different
group and/or project contexts, then this leads to thrashing of the
hints attached to user dquot.
If user quotas are not enabled, then hints are never even used.
So, if the hints are used to avoid the cost of the lookup, is the
cost of the lookup significant enough to justify the hint
infrstructure? Maybe it was once, when there was a global quota
manager shared between all XFS filesystems and was hash table based.
However, lookups are now much simpler, requiring only a single lock and
radix tree lookup local to the filesystem and no hash or LRU
manipulations to be made. Hence the cost of lookup is much lower
than when hints were implemented. Turns out that benchmarks show
that, too, with thir being no differnce in performance when doing
file creation workloads as a single user with user, group and
project quotas enabled - the hints do not make the code go any
faster. In fact, removing the hints shows a 2-3% reduction in the
time it takes to create 50 million inodes....
So, let's just get rid of the hints and the complexity around them.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2014-05-05 14:30:15 +07:00
|
|
|
}
|
|
|
|
xfs_dqunlock(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Release a dquot. Flush it if dirty, then dqput() it.
|
|
|
|
* dquot must not be locked.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_qm_dqrele(
|
|
|
|
xfs_dquot_t *dqp)
|
|
|
|
{
|
2009-06-08 20:33:32 +07:00
|
|
|
if (!dqp)
|
|
|
|
return;
|
|
|
|
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqrele(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
xfs_dqlock(dqp);
|
|
|
|
/*
|
|
|
|
* We don't care to flush it if the dquot is dirty here.
|
|
|
|
* That will create stutters that we want to avoid.
|
|
|
|
* Instead we do a delayed write when we try to reclaim
|
|
|
|
* a dirty dquot. Also xfs_sync will take part of the burden...
|
|
|
|
*/
|
|
|
|
xfs_qm_dqput(dqp);
|
|
|
|
}
|
|
|
|
|
2010-06-23 15:11:15 +07:00
|
|
|
/*
|
|
|
|
* This is the dquot flushing I/O completion routine. It is called
|
|
|
|
* from interrupt level when the buffer containing the dquot is
|
|
|
|
* flushed to disk. It is responsible for removing the dquot logitem
|
|
|
|
* from the AIL if it has not been re-logged, and unlocking the dquot's
|
|
|
|
* flush lock. This behavior is very similar to that of inodes..
|
|
|
|
*/
|
|
|
|
STATIC void
|
|
|
|
xfs_qm_dqflush_done(
|
|
|
|
struct xfs_buf *bp,
|
|
|
|
struct xfs_log_item *lip)
|
|
|
|
{
|
|
|
|
xfs_dq_logitem_t *qip = (struct xfs_dq_logitem *)lip;
|
|
|
|
xfs_dquot_t *dqp = qip->qli_dquot;
|
|
|
|
struct xfs_ail *ailp = lip->li_ailp;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We only want to pull the item from the AIL if its
|
|
|
|
* location in the log has not changed since we started the flush.
|
|
|
|
* Thus, we only bother if the dquot's lsn has
|
|
|
|
* not changed. First we check the lsn outside the lock
|
|
|
|
* since it's cheaper, and then we recheck while
|
|
|
|
* holding the lock before removing the dquot from the AIL.
|
|
|
|
*/
|
2018-05-09 21:47:34 +07:00
|
|
|
if (test_bit(XFS_LI_IN_AIL, &lip->li_flags) &&
|
2017-11-28 23:54:10 +07:00
|
|
|
((lip->li_lsn == qip->qli_flush_lsn) ||
|
2018-05-09 21:47:34 +07:00
|
|
|
test_bit(XFS_LI_FAILED, &lip->li_flags))) {
|
2010-06-23 15:11:15 +07:00
|
|
|
|
|
|
|
/* xfs_trans_ail_delete() drops the AIL lock. */
|
2018-03-08 05:59:39 +07:00
|
|
|
spin_lock(&ailp->ail_lock);
|
2017-11-28 23:54:10 +07:00
|
|
|
if (lip->li_lsn == qip->qli_flush_lsn) {
|
2012-04-23 12:58:41 +07:00
|
|
|
xfs_trans_ail_delete(ailp, lip, SHUTDOWN_CORRUPT_INCORE);
|
2017-11-28 23:54:10 +07:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Clear the failed state since we are about to drop the
|
|
|
|
* flush lock
|
|
|
|
*/
|
2018-05-09 21:47:34 +07:00
|
|
|
xfs_clear_li_failed(lip);
|
2018-03-08 05:59:39 +07:00
|
|
|
spin_unlock(&ailp->ail_lock);
|
2017-11-28 23:54:10 +07:00
|
|
|
}
|
2010-06-23 15:11:15 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Release the dq's flush lock since we're done with it.
|
|
|
|
*/
|
|
|
|
xfs_dqfunlock(dqp);
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Write a modified dquot to disk.
|
|
|
|
* The dquot must be locked and the flush lock too taken by caller.
|
|
|
|
* The flush lock will not be unlocked until the dquot reaches the disk,
|
|
|
|
* but the dquot is free to be unlocked and modified by the caller
|
|
|
|
* in the interim. Dquot is still locked on return. This behavior is
|
|
|
|
* identical to that of inodes.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqflush(
|
2012-04-23 12:58:37 +07:00
|
|
|
struct xfs_dquot *dqp,
|
|
|
|
struct xfs_buf **bpp)
|
2005-04-17 05:20:36 +07:00
|
|
|
{
|
2010-09-06 08:44:45 +07:00
|
|
|
struct xfs_mount *mp = dqp->q_mount;
|
|
|
|
struct xfs_buf *bp;
|
2018-05-07 23:20:18 +07:00
|
|
|
struct xfs_dqblk *dqb;
|
2010-09-06 08:44:45 +07:00
|
|
|
struct xfs_disk_dquot *ddqp;
|
2018-01-09 01:51:25 +07:00
|
|
|
xfs_failaddr_t fa;
|
2005-04-17 05:20:36 +07:00
|
|
|
int error;
|
|
|
|
|
|
|
|
ASSERT(XFS_DQ_IS_LOCKED(dqp));
|
2008-08-13 13:41:43 +07:00
|
|
|
ASSERT(!completion_done(&dqp->q_flush));
|
2010-09-06 08:44:45 +07:00
|
|
|
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqflush(dqp);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-23 12:58:37 +07:00
|
|
|
*bpp = NULL;
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_qm_dqunpin_wait(dqp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This may have been unpinned because the filesystem is shutting
|
|
|
|
* down forcibly. If that's the case we must not write this dquot
|
2012-04-23 12:58:31 +07:00
|
|
|
* to disk, because the log record didn't make it to disk.
|
|
|
|
*
|
|
|
|
* We also have to remove the log item from the AIL in this case,
|
|
|
|
* as we wait for an emptry AIL as part of the unmount process.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2010-09-06 08:44:45 +07:00
|
|
|
if (XFS_FORCED_SHUTDOWN(mp)) {
|
2012-04-23 12:58:31 +07:00
|
|
|
struct xfs_log_item *lip = &dqp->q_logitem.qli_item;
|
2010-09-06 08:44:45 +07:00
|
|
|
dqp->dq_flags &= ~XFS_DQ_DIRTY;
|
2012-04-23 12:58:31 +07:00
|
|
|
|
2015-08-19 07:01:08 +07:00
|
|
|
xfs_trans_ail_remove(lip, SHUTDOWN_CORRUPT_INCORE);
|
|
|
|
|
2014-06-25 11:58:08 +07:00
|
|
|
error = -EIO;
|
2012-04-23 12:58:37 +07:00
|
|
|
goto out_unlock;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get the buffer containing the on-disk dquot
|
|
|
|
*/
|
2010-09-06 08:44:45 +07:00
|
|
|
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
|
2014-08-04 09:43:26 +07:00
|
|
|
mp->m_quotainfo->qi_dqchunklen, 0, &bp,
|
|
|
|
&xfs_dquot_buf_ops);
|
2012-04-23 12:58:37 +07:00
|
|
|
if (error)
|
|
|
|
goto out_unlock;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2010-09-06 08:44:45 +07:00
|
|
|
/*
|
|
|
|
* Calculate the location of the dquot inside the buffer.
|
|
|
|
*/
|
2018-05-07 23:20:18 +07:00
|
|
|
dqb = bp->b_addr + dqp->q_bufoffset;
|
|
|
|
ddqp = &dqb->dd_diskdq;
|
2010-09-06 08:44:45 +07:00
|
|
|
|
|
|
|
/*
|
2018-05-07 23:20:18 +07:00
|
|
|
* A simple sanity check in case we got a corrupted dquot.
|
2010-09-06 08:44:45 +07:00
|
|
|
*/
|
2018-05-07 23:20:18 +07:00
|
|
|
fa = xfs_dqblk_verify(mp, dqb, be32_to_cpu(ddqp->d_id), 0);
|
2018-01-09 01:51:25 +07:00
|
|
|
if (fa) {
|
|
|
|
xfs_alert(mp, "corrupt dquot ID 0x%x in memory at %pS",
|
|
|
|
be32_to_cpu(ddqp->d_id), fa);
|
2010-09-06 08:44:45 +07:00
|
|
|
xfs_buf_relse(bp);
|
|
|
|
xfs_dqfunlock(dqp);
|
|
|
|
xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
|
2014-06-25 11:58:08 +07:00
|
|
|
return -EIO;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/* This is the only portion of data that needs to persist */
|
2010-09-06 08:44:45 +07:00
|
|
|
memcpy(ddqp, &dqp->q_core, sizeof(xfs_disk_dquot_t));
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Clear the dirty field and remember the flush lsn for later use.
|
|
|
|
*/
|
2010-09-06 08:44:45 +07:00
|
|
|
dqp->dq_flags &= ~XFS_DQ_DIRTY;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-10-30 13:39:12 +07:00
|
|
|
xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
|
|
|
|
&dqp->q_logitem.qli_item.li_lsn);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2013-04-03 12:11:16 +07:00
|
|
|
/*
|
|
|
|
* copy the lsn into the on-disk dquot now while we have the in memory
|
|
|
|
* dquot here. This can't be done later in the write verifier as we
|
|
|
|
* can't get access to the log item at that point in time.
|
xfs: rework dquot CRCs
Calculating dquot CRCs when the backing buffer is written back just
doesn't work reliably. There are several places which manipulate
dquots directly in the buffers, and they don't calculate CRCs
appropriately, nor do they always set the buffer up to calculate
CRCs appropriately.
Firstly, if we log a dquot buffer (e.g. during allocation) it gets
logged without valid CRC, and so on recovery we end up with a dquot
that is not valid.
Secondly, if we recover/repair a dquot, we don't have a verifier
attached to the buffer and hence CRCs are not calculated on the way
down to disk.
Thirdly, calculating the CRC after we've changed the contents means
that if we re-read the dquot from the buffer, we cannot verify the
contents of the dquot are valid, as the CRC is invalid.
So, to avoid all the dquot CRC errors that are being detected by the
read verifier, change to using the same model as for inodes. That
is, dquot CRCs are calculated and written to the backing buffer at
the time the dquot is flushed to the backing buffer. If we modify
the dquot directly in the backing buffer, calculate the CRC
immediately after the modification is complete. Hence the dquot in
the on-disk buffer should always have a valid CRC.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-06-03 12:28:46 +07:00
|
|
|
*
|
|
|
|
* We also calculate the CRC here so that the on-disk dquot in the
|
|
|
|
* buffer always has a valid CRC. This ensures there is no possibility
|
|
|
|
* of a dquot without an up-to-date CRC getting to disk.
|
2013-04-03 12:11:16 +07:00
|
|
|
*/
|
|
|
|
if (xfs_sb_version_hascrc(&mp->m_sb)) {
|
|
|
|
dqb->dd_lsn = cpu_to_be64(dqp->q_logitem.qli_item.li_lsn);
|
xfs: rework dquot CRCs
Calculating dquot CRCs when the backing buffer is written back just
doesn't work reliably. There are several places which manipulate
dquots directly in the buffers, and they don't calculate CRCs
appropriately, nor do they always set the buffer up to calculate
CRCs appropriately.
Firstly, if we log a dquot buffer (e.g. during allocation) it gets
logged without valid CRC, and so on recovery we end up with a dquot
that is not valid.
Secondly, if we recover/repair a dquot, we don't have a verifier
attached to the buffer and hence CRCs are not calculated on the way
down to disk.
Thirdly, calculating the CRC after we've changed the contents means
that if we re-read the dquot from the buffer, we cannot verify the
contents of the dquot are valid, as the CRC is invalid.
So, to avoid all the dquot CRC errors that are being detected by the
read verifier, change to using the same model as for inodes. That
is, dquot CRCs are calculated and written to the backing buffer at
the time the dquot is flushed to the backing buffer. If we modify
the dquot directly in the backing buffer, calculate the CRC
immediately after the modification is complete. Hence the dquot in
the on-disk buffer should always have a valid CRC.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
2013-06-03 12:28:46 +07:00
|
|
|
xfs_update_cksum((char *)dqb, sizeof(struct xfs_dqblk),
|
|
|
|
XFS_DQUOT_CRC_OFF);
|
2013-04-03 12:11:16 +07:00
|
|
|
}
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Attach an iodone routine so that we can remove this dquot from the
|
|
|
|
* AIL and release the flush lock once the dquot is synced to disk.
|
|
|
|
*/
|
2010-06-23 15:11:15 +07:00
|
|
|
xfs_buf_attach_iodone(bp, xfs_qm_dqflush_done,
|
|
|
|
&dqp->q_logitem.qli_item);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* If the buffer is pinned then push on the log so we won't
|
|
|
|
* get stuck waiting in the write for too long.
|
|
|
|
*/
|
2011-07-23 06:40:27 +07:00
|
|
|
if (xfs_buf_ispinned(bp)) {
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqflush_force(dqp);
|
2010-01-19 16:56:46 +07:00
|
|
|
xfs_log_force(mp, 0);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2009-12-15 06:14:59 +07:00
|
|
|
trace_xfs_dqflush_done(dqp);
|
2012-04-23 12:58:37 +07:00
|
|
|
*bpp = bp;
|
|
|
|
return 0;
|
2009-12-15 06:14:59 +07:00
|
|
|
|
2012-04-23 12:58:37 +07:00
|
|
|
out_unlock:
|
|
|
|
xfs_dqfunlock(dqp);
|
2014-06-25 11:58:08 +07:00
|
|
|
return -EIO;
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
|
2009-01-19 08:03:19 +07:00
|
|
|
/*
|
|
|
|
* Lock two xfs_dquot structures.
|
|
|
|
*
|
|
|
|
* To avoid deadlocks we always lock the quota structure with
|
|
|
|
* the lowerd id first.
|
|
|
|
*/
|
2005-04-17 05:20:36 +07:00
|
|
|
void
|
|
|
|
xfs_dqlock2(
|
|
|
|
xfs_dquot_t *d1,
|
|
|
|
xfs_dquot_t *d2)
|
|
|
|
{
|
|
|
|
if (d1 && d2) {
|
|
|
|
ASSERT(d1 != d2);
|
2005-11-02 11:01:12 +07:00
|
|
|
if (be32_to_cpu(d1->q_core.d_id) >
|
|
|
|
be32_to_cpu(d2->q_core.d_id)) {
|
2009-01-19 08:03:19 +07:00
|
|
|
mutex_lock(&d2->q_qlock);
|
|
|
|
mutex_lock_nested(&d1->q_qlock, XFS_QLOCK_NESTED);
|
2005-04-17 05:20:36 +07:00
|
|
|
} else {
|
2009-01-19 08:03:19 +07:00
|
|
|
mutex_lock(&d1->q_qlock);
|
|
|
|
mutex_lock_nested(&d2->q_qlock, XFS_QLOCK_NESTED);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
2009-01-19 08:03:19 +07:00
|
|
|
} else if (d1) {
|
|
|
|
mutex_lock(&d1->q_qlock);
|
|
|
|
} else if (d2) {
|
|
|
|
mutex_lock(&d2->q_qlock);
|
2005-04-17 05:20:36 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-03-13 15:52:37 +07:00
|
|
|
int __init
|
|
|
|
xfs_qm_init(void)
|
|
|
|
{
|
|
|
|
xfs_qm_dqzone =
|
|
|
|
kmem_zone_init(sizeof(struct xfs_dquot), "xfs_dquot");
|
|
|
|
if (!xfs_qm_dqzone)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
xfs_qm_dqtrxzone =
|
|
|
|
kmem_zone_init(sizeof(struct xfs_dquot_acct), "xfs_dqtrx");
|
|
|
|
if (!xfs_qm_dqtrxzone)
|
|
|
|
goto out_free_dqzone;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_free_dqzone:
|
|
|
|
kmem_zone_destroy(xfs_qm_dqzone);
|
|
|
|
out:
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2012-03-17 01:36:18 +07:00
|
|
|
void
|
2012-03-13 15:52:37 +07:00
|
|
|
xfs_qm_exit(void)
|
|
|
|
{
|
|
|
|
kmem_zone_destroy(xfs_qm_dqtrxzone);
|
|
|
|
kmem_zone_destroy(xfs_qm_dqzone);
|
|
|
|
}
|
2018-05-05 05:31:21 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Iterate every dquot of a particular type. The caller must ensure that the
|
|
|
|
* particular quota type is active. iter_fn can return negative error codes,
|
|
|
|
* or XFS_BTREE_QUERY_RANGE_ABORT to indicate that it wants to stop iterating.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_qm_dqiterate(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
uint dqtype,
|
|
|
|
xfs_qm_dqiterate_fn iter_fn,
|
|
|
|
void *priv)
|
|
|
|
{
|
|
|
|
struct xfs_dquot *dq;
|
|
|
|
xfs_dqid_t id = 0;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
do {
|
|
|
|
error = xfs_qm_dqget_next(mp, id, dqtype, &dq);
|
|
|
|
if (error == -ENOENT)
|
|
|
|
return 0;
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
error = iter_fn(dq, dqtype, priv);
|
|
|
|
id = be32_to_cpu(dq->q_core.d_id);
|
|
|
|
xfs_qm_dqput(dq);
|
|
|
|
id++;
|
|
|
|
} while (error == 0 && id != 0);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|