diff options
Diffstat (limited to 'fs/xfs/xfs_dquot.c')
| -rw-r--r-- | fs/xfs/xfs_dquot.c | 1903 |
1 files changed, 1102 insertions, 801 deletions
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c index 0adf27ecf3f1..612ca682a513 100644 --- a/fs/xfs/xfs_dquot.c +++ b/fs/xfs/xfs_dquot.c @@ -1,50 +1,37 @@ +// SPDX-License-Identifier: GPL-2.0 /* * Copyright (c) 2000-2003 Silicon Graphics, Inc. * All Rights Reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License as - * published by the Free Software Foundation. - * - * This program is distributed in the hope that it would be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ #include "xfs.h" #include "xfs_fs.h" +#include "xfs_format.h" +#include "xfs_log_format.h" +#include "xfs_shared.h" +#include "xfs_trans_resv.h" #include "xfs_bit.h" -#include "xfs_log.h" -#include "xfs_trans.h" -#include "xfs_sb.h" -#include "xfs_ag.h" -#include "xfs_alloc.h" -#include "xfs_quota.h" #include "xfs_mount.h" -#include "xfs_bmap_btree.h" +#include "xfs_defer.h" #include "xfs_inode.h" #include "xfs_bmap.h" -#include "xfs_rtalloc.h" -#include "xfs_error.h" -#include "xfs_itable.h" -#include "xfs_attr.h" +#include "xfs_quota.h" +#include "xfs_trans.h" #include "xfs_buf_item.h" #include "xfs_trans_space.h" #include "xfs_trans_priv.h" #include "xfs_qm.h" -#include "xfs_cksum.h" #include "xfs_trace.h" +#include "xfs_log.h" +#include "xfs_bmap_btree.h" +#include "xfs_error.h" +#include "xfs_health.h" /* * Lock order: * * ip->i_lock * qi->qi_tree_lock - * dquot->q_qlock (xfs_dqlock() and friends) + * dquot->q_qlock * dquot->q_flush (xfs_dqflock() and friends) * qi->qi_lru_lock * @@ -52,31 +39,75 @@ * otherwise by the lowest id first, see xfs_dqlock2. */ -#ifdef DEBUG -xfs_buftarg_t *xfs_dqerror_target; -int xfs_do_dqerror; -int xfs_dqreq_num; -int xfs_dqerror_mod = 33; -#endif +struct kmem_cache *xfs_dqtrx_cache; +static struct kmem_cache *xfs_dquot_cache; -struct kmem_zone *xfs_qm_dqtrxzone; -static struct kmem_zone *xfs_qm_dqzone; +static struct lock_class_key xfs_dquot_group_class; +static struct lock_class_key xfs_dquot_project_class; -static struct lock_class_key xfs_dquot_other_class; +/* Record observations of quota corruption with the health tracking system. */ +static void +xfs_dquot_mark_sick( + struct xfs_dquot *dqp) +{ + struct xfs_mount *mp = dqp->q_mount; + + switch (dqp->q_type) { + case XFS_DQTYPE_USER: + xfs_fs_mark_sick(mp, XFS_SICK_FS_UQUOTA); + break; + case XFS_DQTYPE_GROUP: + xfs_fs_mark_sick(mp, XFS_SICK_FS_GQUOTA); + break; + case XFS_DQTYPE_PROJ: + xfs_fs_mark_sick(mp, XFS_SICK_FS_PQUOTA); + break; + default: + ASSERT(0); + break; + } +} + +/* + * Detach the dquot buffer if it's still attached, because we can get called + * through dqpurge after a log shutdown. Caller must hold the dqflock or have + * otherwise isolated the dquot. + */ +void +xfs_dquot_detach_buf( + struct xfs_dquot *dqp) +{ + struct xfs_dq_logitem *qlip = &dqp->q_logitem; + struct xfs_buf *bp = NULL; + + spin_lock(&qlip->qli_lock); + if (qlip->qli_item.li_buf) { + bp = qlip->qli_item.li_buf; + qlip->qli_item.li_buf = NULL; + } + spin_unlock(&qlip->qli_lock); + if (bp) { + xfs_buf_lock(bp); + list_del_init(&qlip->qli_item.li_bio_list); + xfs_buf_relse(bp); + } +} /* * This is called to free all the memory associated with a dquot */ void xfs_qm_dqdestroy( - xfs_dquot_t *dqp) + struct xfs_dquot *dqp) { ASSERT(list_empty(&dqp->q_lru)); + ASSERT(dqp->q_logitem.qli_item.li_buf == NULL); + kvfree(dqp->q_logitem.qli_item.li_lv_shadow); mutex_destroy(&dqp->q_qlock); - kmem_zone_free(xfs_qm_dqzone, dqp); - XFS_STATS_DEC(xs_qm_dquot); + XFS_STATS_DEC(dqp->q_mount, xs_qm_dquot); + kmem_cache_free(xfs_dquot_cache, dqp); } /* @@ -86,36 +117,79 @@ xfs_qm_dqdestroy( */ void xfs_qm_adjust_dqlimits( - struct xfs_mount *mp, struct xfs_dquot *dq) { + struct xfs_mount *mp = dq->q_mount; struct xfs_quotainfo *q = mp->m_quotainfo; - struct xfs_disk_dquot *d = &dq->q_core; + struct xfs_def_quota *defq; int prealloc = 0; - ASSERT(d->d_id); + ASSERT(dq->q_id); + defq = xfs_get_defquota(q, xfs_dquot_type(dq)); - if (q->qi_bsoftlimit && !d->d_blk_softlimit) { - d->d_blk_softlimit = cpu_to_be64(q->qi_bsoftlimit); + if (!dq->q_blk.softlimit) { + dq->q_blk.softlimit = defq->blk.soft; prealloc = 1; } - if (q->qi_bhardlimit && !d->d_blk_hardlimit) { - d->d_blk_hardlimit = cpu_to_be64(q->qi_bhardlimit); + if (!dq->q_blk.hardlimit) { + dq->q_blk.hardlimit = defq->blk.hard; prealloc = 1; } - if (q->qi_isoftlimit && !d->d_ino_softlimit) - d->d_ino_softlimit = cpu_to_be64(q->qi_isoftlimit); - if (q->qi_ihardlimit && !d->d_ino_hardlimit) - d->d_ino_hardlimit = cpu_to_be64(q->qi_ihardlimit); - if (q->qi_rtbsoftlimit && !d->d_rtb_softlimit) - d->d_rtb_softlimit = cpu_to_be64(q->qi_rtbsoftlimit); - if (q->qi_rtbhardlimit && !d->d_rtb_hardlimit) - d->d_rtb_hardlimit = cpu_to_be64(q->qi_rtbhardlimit); + if (!dq->q_ino.softlimit) + dq->q_ino.softlimit = defq->ino.soft; + if (!dq->q_ino.hardlimit) + dq->q_ino.hardlimit = defq->ino.hard; + if (!dq->q_rtb.softlimit) + dq->q_rtb.softlimit = defq->rtb.soft; + if (!dq->q_rtb.hardlimit) + dq->q_rtb.hardlimit = defq->rtb.hard; if (prealloc) xfs_dquot_set_prealloc_limits(dq); } +/* Set the expiration time of a quota's grace period. */ +time64_t +xfs_dquot_set_timeout( + struct xfs_mount *mp, + time64_t timeout) +{ + struct xfs_quotainfo *qi = mp->m_quotainfo; + + return clamp_t(time64_t, timeout, qi->qi_expiry_min, + qi->qi_expiry_max); +} + +/* Set the length of the default grace period. */ +time64_t +xfs_dquot_set_grace_period( + time64_t grace) +{ + return clamp_t(time64_t, grace, XFS_DQ_GRACE_MIN, XFS_DQ_GRACE_MAX); +} + +/* + * Determine if this quota counter is over either limit and set the quota + * timers as appropriate. + */ +static inline void +xfs_qm_adjust_res_timer( + struct xfs_mount *mp, + struct xfs_dquot_res *res, + struct xfs_quota_limits *qlim) +{ + ASSERT(res->hardlimit == 0 || res->softlimit <= res->hardlimit); + + if ((res->softlimit && res->count > res->softlimit) || + (res->hardlimit && res->count > res->hardlimit)) { + if (res->timer == 0) + res->timer = xfs_dquot_set_timeout(mp, + ktime_get_real_seconds() + qlim->time); + } else { + res->timer = 0; + } +} + /* * Check the limits and timers of a dquot and start or reset timers * if necessary. @@ -131,324 +205,186 @@ xfs_qm_adjust_dqlimits( */ void xfs_qm_adjust_dqtimers( - xfs_mount_t *mp, - xfs_disk_dquot_t *d) + struct xfs_dquot *dq) { - ASSERT(d->d_id); - -#ifdef DEBUG - 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)); -#endif - - if (!d->d_btimer) { - if ((d->d_blk_softlimit && - (be64_to_cpu(d->d_bcount) > - be64_to_cpu(d->d_blk_softlimit))) || - (d->d_blk_hardlimit && - (be64_to_cpu(d->d_bcount) > - be64_to_cpu(d->d_blk_hardlimit)))) { - d->d_btimer = cpu_to_be32(get_seconds() + - mp->m_quotainfo->qi_btimelimit); - } else { - d->d_bwarns = 0; - } - } else { - if ((!d->d_blk_softlimit || - (be64_to_cpu(d->d_bcount) <= - be64_to_cpu(d->d_blk_softlimit))) && - (!d->d_blk_hardlimit || - (be64_to_cpu(d->d_bcount) <= - be64_to_cpu(d->d_blk_hardlimit)))) { - d->d_btimer = 0; - } - } + struct xfs_mount *mp = dq->q_mount; + struct xfs_quotainfo *qi = mp->m_quotainfo; + struct xfs_def_quota *defq; - if (!d->d_itimer) { - if ((d->d_ino_softlimit && - (be64_to_cpu(d->d_icount) > - be64_to_cpu(d->d_ino_softlimit))) || - (d->d_ino_hardlimit && - (be64_to_cpu(d->d_icount) > - be64_to_cpu(d->d_ino_hardlimit)))) { - d->d_itimer = cpu_to_be32(get_seconds() + - mp->m_quotainfo->qi_itimelimit); - } else { - d->d_iwarns = 0; - } - } else { - if ((!d->d_ino_softlimit || - (be64_to_cpu(d->d_icount) <= - be64_to_cpu(d->d_ino_softlimit))) && - (!d->d_ino_hardlimit || - (be64_to_cpu(d->d_icount) <= - be64_to_cpu(d->d_ino_hardlimit)))) { - d->d_itimer = 0; - } - } + ASSERT(dq->q_id); + defq = xfs_get_defquota(qi, xfs_dquot_type(dq)); - if (!d->d_rtbtimer) { - if ((d->d_rtb_softlimit && - (be64_to_cpu(d->d_rtbcount) > - be64_to_cpu(d->d_rtb_softlimit))) || - (d->d_rtb_hardlimit && - (be64_to_cpu(d->d_rtbcount) > - be64_to_cpu(d->d_rtb_hardlimit)))) { - d->d_rtbtimer = cpu_to_be32(get_seconds() + - mp->m_quotainfo->qi_rtbtimelimit); - } else { - d->d_rtbwarns = 0; - } - } else { - if ((!d->d_rtb_softlimit || - (be64_to_cpu(d->d_rtbcount) <= - be64_to_cpu(d->d_rtb_softlimit))) && - (!d->d_rtb_hardlimit || - (be64_to_cpu(d->d_rtbcount) <= - be64_to_cpu(d->d_rtb_hardlimit)))) { - d->d_rtbtimer = 0; - } - } + xfs_qm_adjust_res_timer(dq->q_mount, &dq->q_blk, &defq->blk); + xfs_qm_adjust_res_timer(dq->q_mount, &dq->q_ino, &defq->ino); + xfs_qm_adjust_res_timer(dq->q_mount, &dq->q_rtb, &defq->rtb); } /* * initialize a buffer full of dquots and log the whole thing */ -STATIC void +void xfs_qm_init_dquot_blk( - xfs_trans_t *tp, - xfs_mount_t *mp, - xfs_dqid_t id, - uint type, - xfs_buf_t *bp) + struct xfs_trans *tp, + xfs_dqid_t id, + xfs_dqtype_t type, + struct xfs_buf *bp) { + struct xfs_mount *mp = tp->t_mountp; struct xfs_quotainfo *q = mp->m_quotainfo; - xfs_dqblk_t *d; - int curid, i; + struct xfs_dqblk *d; + xfs_dqid_t curid; + unsigned int qflag; + unsigned int blftype; + int i; ASSERT(tp); ASSERT(xfs_buf_islocked(bp)); + switch (type) { + case XFS_DQTYPE_USER: + qflag = XFS_UQUOTA_CHKD; + blftype = XFS_BLF_UDQUOT_BUF; + break; + case XFS_DQTYPE_PROJ: + qflag = XFS_PQUOTA_CHKD; + blftype = XFS_BLF_PDQUOT_BUF; + break; + case XFS_DQTYPE_GROUP: + qflag = XFS_GQUOTA_CHKD; + blftype = XFS_BLF_GDQUOT_BUF; + break; + default: + ASSERT(0); + return; + } + d = bp->b_addr; /* * ID of the first dquot in the block - id's are zero based. */ curid = id - (id % q->qi_dqperchunk); - ASSERT(curid >= 0); memset(d, 0, BBTOB(q->qi_dqchunklen)); 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; - if (xfs_sb_version_hascrc(&mp->m_sb)) { - uuid_copy(&d->dd_uuid, &mp->m_sb.sb_uuid); + d->dd_diskdq.d_type = type; + if (curid > 0 && xfs_has_bigtime(mp)) + d->dd_diskdq.d_type |= XFS_DQTYPE_BIGTIME; + if (xfs_has_crc(mp)) { + uuid_copy(&d->dd_uuid, &mp->m_sb.sb_meta_uuid); xfs_update_cksum((char *)d, sizeof(struct xfs_dqblk), XFS_DQUOT_CRC_OFF); } } - xfs_trans_dquot_buf(tp, bp, - (type & XFS_DQ_USER ? XFS_BLF_UDQUOT_BUF : - ((type & XFS_DQ_PROJ) ? XFS_BLF_PDQUOT_BUF : - XFS_BLF_GDQUOT_BUF))); - xfs_trans_log_buf(tp, bp, 0, BBTOB(q->qi_dqchunklen) - 1); -} - -/* - * 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) -{ - __uint64_t space; - - 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; -} - -STATIC bool -xfs_dquot_buf_verify_crc( - struct xfs_mount *mp, - struct xfs_buf *bp) -{ - struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr; - int ndquots; - int i; - - if (!xfs_sb_version_hascrc(&mp->m_sb)) - return true; + xfs_trans_dquot_buf(tp, bp, blftype); /* - * if we are in log recovery, the quota subsystem has not been - * initialised so we have no quotainfo structure. In that case, we need - * to manually calculate the number of dquots in the buffer. + * quotacheck uses delayed writes to update all the dquots on disk in an + * efficient manner instead of logging the individual dquot changes as + * they are made. However if we log the buffer allocated here and crash + * after quotacheck while the logged initialisation is still in the + * active region of the log, log recovery can replay the dquot buffer + * initialisation over the top of the checked dquots and corrupt quota + * accounting. + * + * To avoid this problem, quotacheck cannot log the initialised buffer. + * We must still dirty the buffer and write it back before the + * allocation transaction clears the log. Therefore, mark the buffer as + * ordered instead of logging it directly. This is safe for quotacheck + * because it detects and repairs allocated but initialized dquot blocks + * in the quota inodes. */ - if (mp->m_quotainfo) - ndquots = mp->m_quotainfo->qi_dqperchunk; + if (!(mp->m_qflags & qflag)) + xfs_trans_ordered_buf(tp, bp); else - ndquots = xfs_qm_calc_dquots_per_chunk(mp, - XFS_BB_TO_FSB(mp, bp->b_length)); - - for (i = 0; i < ndquots; i++, d++) { - if (!xfs_verify_cksum((char *)d, sizeof(struct xfs_dqblk), - XFS_DQUOT_CRC_OFF)) - return false; - if (!uuid_equal(&d->dd_uuid, &mp->m_sb.sb_uuid)) - return false; - } - return true; + xfs_trans_log_buf(tp, bp, 0, BBTOB(q->qi_dqchunklen) - 1); } -STATIC bool -xfs_dquot_buf_verify( - struct xfs_mount *mp, - struct xfs_buf *bp) +static void +xfs_dquot_set_prealloc( + struct xfs_dquot_pre *pre, + const struct xfs_dquot_res *res) { - struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr; - xfs_dqid_t id = 0; - int ndquots; - int i; + xfs_qcnt_t space; - /* - * if we are in log recovery, the quota subsystem has not been - * initialised so we have no quotainfo structure. In that case, we need - * to manually calculate the number of dquots in the buffer. - */ - if (mp->m_quotainfo) - ndquots = mp->m_quotainfo->qi_dqperchunk; - else - ndquots = xfs_qm_calc_dquots_per_chunk(mp, bp->b_length); + pre->q_prealloc_hi_wmark = res->hardlimit; + pre->q_prealloc_lo_wmark = res->softlimit; - /* - * On the first read of the buffer, verify that each dquot is valid. - * We don't know what the id of the dquot is supposed to be, just that - * they should be increasing monotonically within the buffer. If the - * first id is corrupt, then it will fail on the second dquot in the - * buffer so corruptions could point to the wrong dquot in this case. - */ - for (i = 0; i < ndquots; i++) { - struct xfs_disk_dquot *ddq; - int error; - - ddq = &d[i].dd_diskdq; + space = div_u64(pre->q_prealloc_hi_wmark, 100); + if (!pre->q_prealloc_lo_wmark) + pre->q_prealloc_lo_wmark = space * 95; - if (i == 0) - id = be32_to_cpu(ddq->d_id); - - error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN, - "xfs_dquot_buf_verify"); - if (error) - return false; - } - return true; -} - -static void -xfs_dquot_buf_read_verify( - struct xfs_buf *bp) -{ - struct xfs_mount *mp = bp->b_target->bt_mount; - - if (!xfs_dquot_buf_verify_crc(mp, bp) || !xfs_dquot_buf_verify(mp, bp)) { - XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); - xfs_buf_ioerror(bp, EFSCORRUPTED); - } + pre->q_low_space[XFS_QLOWSP_1_PCNT] = space; + pre->q_low_space[XFS_QLOWSP_3_PCNT] = space * 3; + pre->q_low_space[XFS_QLOWSP_5_PCNT] = space * 5; } /* - * we don't calculate the CRC here as that is done when the dquot is flushed to - * the buffer after the update is done. This ensures that the dquot in the - * buffer always has an up-to-date CRC value. + * 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_buf_write_verify( - struct xfs_buf *bp) +xfs_dquot_set_prealloc_limits(struct xfs_dquot *dqp) { - struct xfs_mount *mp = bp->b_target->bt_mount; - - if (!xfs_dquot_buf_verify(mp, bp)) { - XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, bp->b_addr); - xfs_buf_ioerror(bp, EFSCORRUPTED); - return; - } + xfs_dquot_set_prealloc(&dqp->q_blk_prealloc, &dqp->q_blk); + xfs_dquot_set_prealloc(&dqp->q_rtb_prealloc, &dqp->q_rtb); } -const struct xfs_buf_ops xfs_dquot_buf_ops = { - .verify_read = xfs_dquot_buf_read_verify, - .verify_write = xfs_dquot_buf_write_verify, -}; - /* - * Allocate a block and fill it with dquots. - * This is called when the bmapi finds a hole. + * Ensure that the given in-core dquot has a buffer on disk backing it, and + * return the buffer locked and held. This is called when the bmapi finds a + * hole. */ STATIC int -xfs_qm_dqalloc( - xfs_trans_t **tpp, - xfs_mount_t *mp, - xfs_dquot_t *dqp, - xfs_inode_t *quotip, - xfs_fileoff_t offset_fsb, - xfs_buf_t **O_bpp) +xfs_dquot_disk_alloc( + struct xfs_dquot *dqp, + struct xfs_buf **bpp) { - xfs_fsblock_t firstblock; - xfs_bmap_free_t flist; - xfs_bmbt_irec_t map; - int nmaps, error, committed; - xfs_buf_t *bp; - xfs_trans_t *tp = *tpp; - - ASSERT(tp != NULL); + struct xfs_bmbt_irec map; + struct xfs_trans *tp; + struct xfs_mount *mp = dqp->q_mount; + struct xfs_buf *bp; + xfs_dqtype_t qtype = xfs_dquot_type(dqp); + struct xfs_inode *quotip = xfs_quota_inode(mp, qtype); + int nmaps = 1; + int error; trace_xfs_dqalloc(dqp); - /* - * Initialize the bmap freelist prior to calling bmapi code. - */ - xfs_bmap_init(&flist, &firstblock); + error = xfs_trans_alloc(mp, &M_RES(mp)->tr_qm_dqalloc, + XFS_QM_DQALLOC_SPACE_RES(mp), 0, 0, &tp); + if (error) + return error; + xfs_ilock(quotip, XFS_ILOCK_EXCL); - /* - * Return if this type of quotas is turned off while we didn't - * have an inode lock - */ - if (!xfs_this_quota_on(dqp->q_mount, dqp->dq_flags)) { - xfs_iunlock(quotip, XFS_ILOCK_EXCL); - return (ESRCH); + xfs_trans_ijoin(tp, quotip, 0); + + if (!xfs_this_quota_on(dqp->q_mount, qtype)) { + /* + * Return if this type of quotas is turned off while we didn't + * have an inode lock + */ + error = -ESRCH; + goto err_cancel; } - xfs_trans_ijoin(tp, quotip, XFS_ILOCK_EXCL); - nmaps = 1; - error = xfs_bmapi_write(tp, quotip, offset_fsb, - XFS_DQUOT_CLUSTER_SIZE_FSB, XFS_BMAPI_METADATA, - &firstblock, XFS_QM_DQALLOC_SPACE_RES(mp), - &map, &nmaps, &flist); + error = xfs_iext_count_extend(tp, quotip, XFS_DATA_FORK, + XFS_IEXT_ADD_NOSPLIT_CNT); if (error) - goto error0; + goto err_cancel; + + /* Create the block mapping. */ + error = xfs_bmapi_write(tp, quotip, dqp->q_fileoffset, + XFS_DQUOT_CLUSTER_SIZE_FSB, XFS_BMAPI_METADATA, 0, &map, + &nmaps); + if (error) + goto err_cancel; + ASSERT(map.br_blockcount == XFS_DQUOT_CLUSTER_SIZE_FSB); - ASSERT(nmaps == 1); ASSERT((map.br_startblock != DELAYSTARTBLOCK) && (map.br_startblock != HOLESTARTBLOCK)); @@ -458,236 +394,151 @@ xfs_qm_dqalloc( dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock); /* now we can just get the buffer (there's nothing to read yet) */ - bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, - dqp->q_blkno, - mp->m_quotainfo->qi_dqchunklen, - 0); - - error = xfs_buf_geterror(bp); + error = xfs_trans_get_buf(tp, mp->m_ddev_targp, dqp->q_blkno, + mp->m_quotainfo->qi_dqchunklen, 0, &bp); if (error) - goto error1; + goto err_cancel; bp->b_ops = &xfs_dquot_buf_ops; /* * Make a chunk of dquots out of this buffer and log * the entire thing. */ - xfs_qm_init_dquot_blk(tp, mp, be32_to_cpu(dqp->q_core.d_id), - dqp->dq_flags & XFS_DQ_ALLTYPES, bp); + xfs_qm_init_dquot_blk(tp, dqp->q_id, qtype, bp); + xfs_buf_set_ref(bp, XFS_DQUOT_REF); /* - * xfs_bmap_finish() may commit the current transaction and - * start a second transaction if the freelist is not empty. + * 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. + * + * 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 + * transaction is gone but the new buffer is not joined or held to any + * transaction, so we must _buf_relse it. * - * Since we still want to modify this buffer, we need to - * ensure that the buffer is not released on commit of - * the first transaction and ensure the buffer is added to the - * second transaction. + * If everything succeeds, the caller of this function is returned a + * buffer that is locked and held to the transaction. The caller + * is responsible for unlocking any buffer passed back, either + * manually or by committing the transaction. On error, the buffer is + * released and not passed back. * - * If there is only one transaction then don't stop the buffer - * from being released when it commits later on. + * Keep the quota inode ILOCKed until after the transaction commit to + * maintain the atomicity of bmap/rmap updates. */ - xfs_trans_bhold(tp, bp); - - if ((error = xfs_bmap_finish(tpp, &flist, &committed))) { - goto error1; - } - - if (committed) { - tp = *tpp; - xfs_trans_bjoin(tp, bp); - } else { - xfs_trans_bhold_release(tp, bp); - } - - *O_bpp = bp; - return 0; - - error1: - xfs_bmap_cancel(&flist); - error0: + error = xfs_trans_commit(tp); xfs_iunlock(quotip, XFS_ILOCK_EXCL); - - return (error); -} -STATIC int -xfs_qm_dqrepair( - struct xfs_mount *mp, - struct xfs_trans *tp, - struct xfs_dquot *dqp, - xfs_dqid_t firstid, - struct xfs_buf **bpp) -{ - int error; - struct xfs_disk_dquot *ddq; - struct xfs_dqblk *d; - int i; - - /* - * Read the buffer without verification so we get the corrupted - * buffer returned to us. make sure we verify it on write, though. - */ - error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, dqp->q_blkno, - mp->m_quotainfo->qi_dqchunklen, - 0, bpp, NULL); - if (error) { - ASSERT(*bpp == NULL); - return XFS_ERROR(error); - } - (*bpp)->b_ops = &xfs_dquot_buf_ops; - - ASSERT(xfs_buf_islocked(*bpp)); - d = (struct xfs_dqblk *)(*bpp)->b_addr; - - /* Do the actual repair of dquots in this buffer */ - for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) { - ddq = &d[i].dd_diskdq; - error = xfs_qm_dqcheck(mp, ddq, firstid + i, - dqp->dq_flags & XFS_DQ_ALLTYPES, - XFS_QMOPT_DQREPAIR, "xfs_qm_dqrepair"); - if (error) { - /* repair failed, we're screwed */ - xfs_trans_brelse(tp, *bpp); - return XFS_ERROR(EIO); - } + xfs_buf_relse(bp); + return error; } + *bpp = bp; return 0; + +err_cancel: + xfs_trans_cancel(tp); + xfs_iunlock(quotip, XFS_ILOCK_EXCL); + return error; } /* - * Maps a dquot to the buffer containing its on-disk version. - * This returns a ptr to the buffer containing the on-disk dquot - * in the bpp param, and a ptr to the on-disk dquot within that buffer + * Read in the in-core dquot's on-disk metadata and return the buffer. + * Returns ENOENT to signal a hole. */ STATIC int -xfs_qm_dqtobp( - xfs_trans_t **tpp, - xfs_dquot_t *dqp, - xfs_disk_dquot_t **O_ddpp, - xfs_buf_t **O_bpp, - uint flags) +xfs_dquot_disk_read( + struct xfs_mount *mp, + struct xfs_dquot *dqp, + struct xfs_buf **bpp) { struct xfs_bmbt_irec map; - int nmaps = 1, error; struct xfs_buf *bp; - struct xfs_inode *quotip = xfs_dq_to_quota_inode(dqp); - struct xfs_mount *mp = dqp->q_mount; - xfs_dqid_t id = be32_to_cpu(dqp->q_core.d_id); - struct xfs_trans *tp = (tpp ? *tpp : NULL); - - dqp->q_fileoffset = (xfs_fileoff_t)id / mp->m_quotainfo->qi_dqperchunk; + xfs_dqtype_t qtype = xfs_dquot_type(dqp); + struct xfs_inode *quotip = xfs_quota_inode(mp, qtype); + uint lock_mode; + int nmaps = 1; + int error; - xfs_ilock(quotip, XFS_ILOCK_SHARED); - if (!xfs_this_quota_on(dqp->q_mount, dqp->dq_flags)) { + lock_mode = xfs_ilock_data_map_shared(quotip); + if (!xfs_this_quota_on(mp, qtype)) { /* * Return if this type of quotas is turned off while we * didn't have the quota inode lock. */ - xfs_iunlock(quotip, XFS_ILOCK_SHARED); - return ESRCH; + xfs_iunlock(quotip, lock_mode); + return -ESRCH; } /* * Find the block map; no allocations yet */ error = xfs_bmapi_read(quotip, dqp->q_fileoffset, - XFS_DQUOT_CLUSTER_SIZE_FSB, &map, &nmaps, 0); - - xfs_iunlock(quotip, XFS_ILOCK_SHARED); + XFS_DQUOT_CLUSTER_SIZE_FSB, &map, &nmaps, 0); + xfs_iunlock(quotip, lock_mode); if (error) return error; ASSERT(nmaps == 1); - ASSERT(map.br_blockcount == 1); - - /* - * Offset of dquot in the (fixed sized) dquot chunk. - */ - dqp->q_bufoffset = (id % mp->m_quotainfo->qi_dqperchunk) * - sizeof(xfs_dqblk_t); - + ASSERT(map.br_blockcount >= 1); ASSERT(map.br_startblock != DELAYSTARTBLOCK); - if (map.br_startblock == HOLESTARTBLOCK) { - /* - * We don't allocate unless we're asked to - */ - if (!(flags & XFS_QMOPT_DQALLOC)) - return ENOENT; + if (map.br_startblock == HOLESTARTBLOCK) + return -ENOENT; - ASSERT(tp); - error = xfs_qm_dqalloc(tpp, mp, dqp, quotip, - dqp->q_fileoffset, &bp); - if (error) - return error; - tp = *tpp; - } else { - trace_xfs_dqtobp_read(dqp); + trace_xfs_dqtobp_read(dqp); - /* - * store the blkno etc so that we don't have to do the - * mapping all the time - */ - dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock); - - error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, - dqp->q_blkno, - mp->m_quotainfo->qi_dqchunklen, - 0, &bp, &xfs_dquot_buf_ops); - - if (error == EFSCORRUPTED && (flags & XFS_QMOPT_DQREPAIR)) { - xfs_dqid_t firstid = (xfs_dqid_t)map.br_startoff * - mp->m_quotainfo->qi_dqperchunk; - ASSERT(bp == NULL); - error = xfs_qm_dqrepair(mp, tp, dqp, firstid, &bp); - } + /* + * store the blkno etc so that we don't have to do the + * mapping all the time + */ + dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock); - if (error) { - ASSERT(bp == NULL); - return XFS_ERROR(error); - } + 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 (xfs_metadata_is_sick(error)) + xfs_dquot_mark_sick(dqp); + if (error) { + ASSERT(bp == NULL); + return error; } ASSERT(xfs_buf_islocked(bp)); - *O_bpp = bp; - *O_ddpp = bp->b_addr + dqp->q_bufoffset; + xfs_buf_set_ref(bp, XFS_DQUOT_REF); + *bpp = bp; - return (0); + return 0; } - -/* - * Read in the ondisk dquot using dqtobp() then copy it to an incore version, - * and release the buffer immediately. - * - * If XFS_QMOPT_DQALLOC is set, allocate a dquot on disk if it needed. - */ -int -xfs_qm_dqread( +/* Allocate and initialize everything we need for an incore dquot. */ +STATIC struct xfs_dquot * +xfs_dquot_alloc( struct xfs_mount *mp, xfs_dqid_t id, - uint type, - uint flags, - struct xfs_dquot **O_dqpp) + xfs_dqtype_t type) { struct xfs_dquot *dqp; - struct xfs_disk_dquot *ddqp; - struct xfs_buf *bp; - struct xfs_trans *tp = NULL; - int error; - int cancelflags = 0; + dqp = kmem_cache_zalloc(xfs_dquot_cache, GFP_KERNEL | __GFP_NOFAIL); - dqp = kmem_zone_zalloc(xfs_qm_dqzone, KM_SLEEP); - - dqp->dq_flags = type; - dqp->q_core.d_id = cpu_to_be32(id); + dqp->q_type = type; + dqp->q_id = id; dqp->q_mount = mp; INIT_LIST_HEAD(&dqp->q_lru); mutex_init(&dqp->q_qlock); init_waitqueue_head(&dqp->q_pinwait); + 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(struct xfs_dqblk); /* * Because we want to use a counting completion, complete @@ -701,322 +552,584 @@ xfs_qm_dqread( * Make sure group quotas have a different lock class than user * quotas. */ - if (!(type & XFS_DQ_USER)) - lockdep_set_class(&dqp->q_qlock, &xfs_dquot_other_class); + switch (type) { + case XFS_DQTYPE_USER: + /* uses the default lock class */ + break; + case XFS_DQTYPE_GROUP: + lockdep_set_class(&dqp->q_qlock, &xfs_dquot_group_class); + break; + case XFS_DQTYPE_PROJ: + lockdep_set_class(&dqp->q_qlock, &xfs_dquot_project_class); + break; + default: + ASSERT(0); + break; + } - XFS_STATS_INC(xs_qm_dquot); + xfs_qm_dquot_logitem_init(dqp); - trace_xfs_dqread(dqp); + XFS_STATS_INC(mp, xs_qm_dquot); + return dqp; +} - if (flags & XFS_QMOPT_DQALLOC) { - tp = xfs_trans_alloc(mp, XFS_TRANS_QM_DQALLOC); - error = xfs_trans_reserve(tp, XFS_QM_DQALLOC_SPACE_RES(mp), - XFS_QM_DQALLOC_LOG_RES(mp), 0, - XFS_TRANS_PERM_LOG_RES, - XFS_WRITE_LOG_COUNT); - if (error) - goto error1; - cancelflags = XFS_TRANS_RELEASE_LOG_RES; - } +/* Check the ondisk dquot's id and type match what the incore dquot expects. */ +static bool +xfs_dquot_check_type( + struct xfs_dquot *dqp, + struct xfs_disk_dquot *ddqp) +{ + uint8_t ddqp_type; + uint8_t dqp_type; + + ddqp_type = ddqp->d_type & XFS_DQTYPE_REC_MASK; + dqp_type = xfs_dquot_type(dqp); + + if (be32_to_cpu(ddqp->d_id) != dqp->q_id) + return false; /* - * get a pointer to the on-disk dquot and the buffer containing it - * dqp already knows its own type (GROUP/USER). + * V5 filesystems always expect an exact type match. V4 filesystems + * expect an exact match for user dquots and for non-root group and + * project dquots. */ - error = xfs_qm_dqtobp(&tp, dqp, &ddqp, &bp, flags); - if (error) { - /* - * This can happen if quotas got turned off (ESRCH), - * or if the dquot didn't exist on disk and we ask to - * allocate (ENOENT). - */ - trace_xfs_dqread_fail(dqp); - cancelflags |= XFS_TRANS_ABORT; - goto error1; + if (xfs_has_crc(dqp->q_mount) || + dqp_type == XFS_DQTYPE_USER || dqp->q_id != 0) + return ddqp_type == dqp_type; + + /* + * V4 filesystems support either group or project quotas, but not both + * at the same time. The non-user quota file can be switched between + * group and project quota uses depending on the mount options, which + * means that we can encounter the other type when we try to load quota + * defaults. Quotacheck will soon reset the entire quota file + * (including the root dquot) anyway, but don't log scary corruption + * reports to dmesg. + */ + return ddqp_type == XFS_DQTYPE_GROUP || ddqp_type == XFS_DQTYPE_PROJ; +} + +/* Copy the in-core quota fields in from the on-disk buffer. */ +STATIC int +xfs_dquot_from_disk( + struct xfs_dquot *dqp, + struct xfs_buf *bp) +{ + struct xfs_dqblk *dqb = xfs_buf_offset(bp, dqp->q_bufoffset); + struct xfs_disk_dquot *ddqp = &dqb->dd_diskdq; + + /* + * Ensure that we got the type and ID we were looking for. + * Everything else was checked by the dquot buffer verifier. + */ + if (!xfs_dquot_check_type(dqp, ddqp)) { + xfs_alert_tag(bp->b_mount, XFS_PTAG_VERIFIER_ERROR, + "Metadata corruption detected at %pS, quota %u", + __this_address, dqp->q_id); + xfs_alert(bp->b_mount, "Unmount and run xfs_repair"); + xfs_dquot_mark_sick(dqp); + return -EFSCORRUPTED; } /* copy everything from disk dquot to the incore dquot */ - memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t)); - xfs_qm_dquot_logitem_init(dqp); + dqp->q_type = ddqp->d_type; + dqp->q_blk.hardlimit = be64_to_cpu(ddqp->d_blk_hardlimit); + dqp->q_blk.softlimit = be64_to_cpu(ddqp->d_blk_softlimit); + dqp->q_ino.hardlimit = be64_to_cpu(ddqp->d_ino_hardlimit); + dqp->q_ino.softlimit = be64_to_cpu(ddqp->d_ino_softlimit); + dqp->q_rtb.hardlimit = be64_to_cpu(ddqp->d_rtb_hardlimit); + dqp->q_rtb.softlimit = be64_to_cpu(ddqp->d_rtb_softlimit); + + dqp->q_blk.count = be64_to_cpu(ddqp->d_bcount); + dqp->q_ino.count = be64_to_cpu(ddqp->d_icount); + dqp->q_rtb.count = be64_to_cpu(ddqp->d_rtbcount); + + dqp->q_blk.timer = xfs_dquot_from_disk_ts(ddqp, ddqp->d_btimer); + dqp->q_ino.timer = xfs_dquot_from_disk_ts(ddqp, ddqp->d_itimer); + dqp->q_rtb.timer = xfs_dquot_from_disk_ts(ddqp, ddqp->d_rtbtimer); /* * 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); + dqp->q_blk.reserved = dqp->q_blk.count; + dqp->q_ino.reserved = dqp->q_ino.count; + dqp->q_rtb.reserved = dqp->q_rtb.count; /* initialize the dquot speculative prealloc thresholds */ xfs_dquot_set_prealloc_limits(dqp); + return 0; +} - /* Mark the buf so that this will stay incore a little longer */ - xfs_buf_set_ref(bp, XFS_DQUOT_REF); +/* Copy the in-core quota fields into the on-disk buffer. */ +void +xfs_dquot_to_disk( + struct xfs_disk_dquot *ddqp, + struct xfs_dquot *dqp) +{ + ddqp->d_magic = cpu_to_be16(XFS_DQUOT_MAGIC); + ddqp->d_version = XFS_DQUOT_VERSION; + ddqp->d_type = dqp->q_type; + ddqp->d_id = cpu_to_be32(dqp->q_id); + ddqp->d_pad0 = 0; + ddqp->d_pad = 0; + + ddqp->d_blk_hardlimit = cpu_to_be64(dqp->q_blk.hardlimit); + ddqp->d_blk_softlimit = cpu_to_be64(dqp->q_blk.softlimit); + ddqp->d_ino_hardlimit = cpu_to_be64(dqp->q_ino.hardlimit); + ddqp->d_ino_softlimit = cpu_to_be64(dqp->q_ino.softlimit); + ddqp->d_rtb_hardlimit = cpu_to_be64(dqp->q_rtb.hardlimit); + ddqp->d_rtb_softlimit = cpu_to_be64(dqp->q_rtb.softlimit); + + ddqp->d_bcount = cpu_to_be64(dqp->q_blk.count); + ddqp->d_icount = cpu_to_be64(dqp->q_ino.count); + ddqp->d_rtbcount = cpu_to_be64(dqp->q_rtb.count); + + ddqp->d_bwarns = 0; + ddqp->d_iwarns = 0; + ddqp->d_rtbwarns = 0; + + ddqp->d_btimer = xfs_dquot_to_disk_ts(dqp, dqp->q_blk.timer); + ddqp->d_itimer = xfs_dquot_to_disk_ts(dqp, dqp->q_ino.timer); + ddqp->d_rtbtimer = xfs_dquot_to_disk_ts(dqp, dqp->q_rtb.timer); +} + +/* + * Read in the ondisk dquot using dqtobp() then copy it to an incore version, + * and release the buffer immediately. If @can_alloc is true, fill any + * holes in the on-disk metadata. + */ +static int +xfs_qm_dqread( + struct xfs_mount *mp, + xfs_dqid_t id, + xfs_dqtype_t type, + bool can_alloc, + struct xfs_dquot **dqpp) +{ + struct xfs_dquot *dqp; + struct xfs_buf *bp; + int error; + + dqp = xfs_dquot_alloc(mp, id, type); + trace_xfs_dqread(dqp); + + /* Try to read the buffer, allocating if necessary. */ + error = xfs_dquot_disk_read(mp, dqp, &bp); + if (error == -ENOENT && can_alloc) + error = xfs_dquot_disk_alloc(dqp, &bp); + if (error) + goto err; /* - * We got the buffer with a xfs_trans_read_buf() (in dqtobp()) - * So we need to release with xfs_trans_brelse(). - * The strategy here is identical to that of inodes; we lock - * the dquot in xfs_qm_dqget() before making it accessible to - * others. This is because dquots, like inodes, need a good level of - * concurrency, and we don't want to take locks on the entire buffers - * for dquot accesses. - * Note also that the dquot buffer may even be dirty at this point, if - * this particular dquot was repaired. We still aren't afraid to - * brelse it because we have the changes incore. + * 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. */ ASSERT(xfs_buf_islocked(bp)); - xfs_trans_brelse(tp, bp); - - if (tp) { - error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); - if (error) - goto error0; - } + error = xfs_dquot_from_disk(dqp, bp); + xfs_buf_relse(bp); + if (error) + goto err; - *O_dqpp = dqp; + *dqpp = dqp; return error; -error1: - if (tp) - xfs_trans_cancel(tp, cancelflags); -error0: +err: + trace_xfs_dqread_fail(dqp); xfs_qm_dqdestroy(dqp); - *O_dqpp = NULL; + *dqpp = NULL; return error; } /* - * Given the file system, inode OR id, and type (UDQUOT/GDQUOT), return a - * a locked dquot, doing an allocation (if requested) as needed. - * When both an inode and an id are given, the inode's id takes precedence. - * That is, if the id changes while we don't hold the ilock inside this - * function, the new dquot is returned, not necessarily the one requested - * in the id argument. + * 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. */ -int -xfs_qm_dqget( - xfs_mount_t *mp, - xfs_inode_t *ip, /* locked inode (optional) */ - xfs_dqid_t id, /* uid/projid/gid depending on type */ - uint type, /* XFS_DQ_USER/XFS_DQ_PROJ/XFS_DQ_GROUP */ - uint flags, /* DQALLOC, DQSUSER, DQREPAIR, DOWARN */ - xfs_dquot_t **O_dqpp) /* OUT : locked incore dquot */ +static int +xfs_dq_get_next_id( + struct xfs_mount *mp, + xfs_dqtype_t type, + xfs_dqid_t *id) { - struct xfs_quotainfo *qi = mp->m_quotainfo; - struct radix_tree_root *tree = xfs_dquot_tree(qi, type); - struct xfs_dquot *dqp; - int error; - - ASSERT(XFS_IS_QUOTA_RUNNING(mp)); - if ((! XFS_IS_UQUOTA_ON(mp) && type == XFS_DQ_USER) || - (! XFS_IS_PQUOTA_ON(mp) && type == XFS_DQ_PROJ) || - (! XFS_IS_GQUOTA_ON(mp) && type == XFS_DQ_GROUP)) { - return (ESRCH); + 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; + struct xfs_iext_cursor cur; + xfs_fsblock_t start; + int error = 0; + + /* If we'd wrap past the max ID, stop */ + if (next_id < *id) + return -ENOENT; + + /* If new ID is within the current chunk, advancing it sufficed */ + if (next_id % mp->m_quotainfo->qi_dqperchunk) { + *id = next_id; + return 0; } -#ifdef DEBUG - if (xfs_do_dqerror) { - if ((xfs_dqerror_target == mp->m_ddev_targp) && - (xfs_dqreq_num++ % xfs_dqerror_mod) == 0) { - xfs_debug(mp, "Returning error in dqget"); - return (EIO); - } - } + /* 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; - ASSERT(type == XFS_DQ_USER || - type == XFS_DQ_PROJ || - type == XFS_DQ_GROUP); - if (ip) { - ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); - ASSERT(xfs_inode_dquot(ip, type) == NULL); + lock_flags = xfs_ilock_data_map_shared(quotip); + error = xfs_iread_extents(NULL, quotip, XFS_DATA_FORK); + if (error) + return error; + + if (xfs_iext_lookup_extent(quotip, "ip->i_df, start, &cur, &got)) { + /* contiguous chunk, bump startoff for the id calculation */ + if (got.br_startoff < start) + got.br_startoff = start; + *id = got.br_startoff * mp->m_quotainfo->qi_dqperchunk; + } else { + error = -ENOENT; } -#endif + + xfs_iunlock(quotip, lock_flags); + + return error; +} + +/* + * 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, + xfs_dqid_t id, + xfs_dqtype_t type) +{ + struct xfs_quotainfo *qi = mp->m_quotainfo; + struct radix_tree_root *tree = xfs_dquot_tree(qi, type); + struct xfs_dquot *dqp; restart: mutex_lock(&qi->qi_tree_lock); dqp = radix_tree_lookup(tree, id); - if (dqp) { - 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++; + if (!dqp) { mutex_unlock(&qi->qi_tree_lock); + XFS_STATS_INC(mp, xs_qm_dqcachemisses); + return NULL; + } - trace_xfs_dqget_hit(dqp); - XFS_STATS_INC(xs_qm_dqcachehits); - *O_dqpp = dqp; - return 0; + if (!lockref_get_not_dead(&dqp->q_lockref)) { + mutex_unlock(&qi->qi_tree_lock); + trace_xfs_dqget_freeing(dqp); + delay(1); + goto restart; } mutex_unlock(&qi->qi_tree_lock); - XFS_STATS_INC(xs_qm_dqcachemisses); - /* - * 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. - */ - if (ip) - xfs_iunlock(ip, XFS_ILOCK_EXCL); + 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 (and held by the cache) as if there had been a cache hit. + * + * The insert needs to be done under memalloc_nofs context because the radix + * tree can do memory allocation during insert. The qi->qi_tree_lock is taken in + * memory reclaim when freeing unused dquots, so we cannot have the radix tree + * node allocation recursing into filesystem reclaim whilst we hold the + * qi_tree_lock. + */ +static int +xfs_qm_dqget_cache_insert( + struct xfs_mount *mp, + xfs_dqid_t id, + xfs_dqtype_t type, + struct xfs_dquot *dqp) +{ + struct xfs_quotainfo *qi = mp->m_quotainfo; + struct radix_tree_root *tree = xfs_dquot_tree(qi, type); + unsigned int nofs_flags; + int error; - error = xfs_qm_dqread(mp, id, type, flags, &dqp); + nofs_flags = memalloc_nofs_save(); + mutex_lock(&qi->qi_tree_lock); + error = radix_tree_insert(tree, id, dqp); + if (unlikely(error)) { + trace_xfs_dqget_dup(dqp); + goto out_unlock; + } - if (ip) - xfs_ilock(ip, XFS_ILOCK_EXCL); + lockref_init(&dqp->q_lockref); + qi->qi_dquots++; +out_unlock: + mutex_unlock(&qi->qi_tree_lock); + memalloc_nofs_restore(nofs_flags); + return error; +} + +/* Check our input parameters. */ +static int +xfs_qm_dqget_checks( + struct xfs_mount *mp, + xfs_dqtype_t type) +{ + switch (type) { + case XFS_DQTYPE_USER: + if (!XFS_IS_UQUOTA_ON(mp)) + return -ESRCH; + return 0; + case XFS_DQTYPE_GROUP: + if (!XFS_IS_GQUOTA_ON(mp)) + return -ESRCH; + return 0; + case XFS_DQTYPE_PROJ: + if (!XFS_IS_PQUOTA_ON(mp)) + return -ESRCH; + return 0; + default: + WARN_ON_ONCE(0); + return -EINVAL; + } +} + +/* + * Given the file system, id, and type (UDQUOT/GDQUOT/PDQUOT), return a + * dquot, doing an allocation (if requested) as needed. + */ +int +xfs_qm_dqget( + struct xfs_mount *mp, + xfs_dqid_t id, + xfs_dqtype_t type, + bool can_alloc, + struct xfs_dquot **O_dqpp) +{ + struct xfs_dquot *dqp; + int error; + + error = xfs_qm_dqget_checks(mp, type); if (error) return error; - if (ip) { - /* - * 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) { - xfs_qm_dqdestroy(dqp); - dqp = dqp1; - xfs_dqlock(dqp); - goto dqret; - } - } else { - /* inode stays locked on return */ - xfs_qm_dqdestroy(dqp); - return XFS_ERROR(ESRCH); - } - } +restart: + dqp = xfs_qm_dqget_cache_lookup(mp, id, type); + if (dqp) + goto found; - mutex_lock(&qi->qi_tree_lock); - error = -radix_tree_insert(tree, id, dqp); - if (unlikely(error)) { - WARN_ON(error != EEXIST); + error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp); + if (error) + return error; - /* - * Duplicate found. Just throw away the new dquot and start - * over. - */ - mutex_unlock(&qi->qi_tree_lock); - trace_xfs_dqget_dup(dqp); + error = xfs_qm_dqget_cache_insert(mp, id, type, dqp); + if (error) { xfs_qm_dqdestroy(dqp); - XFS_STATS_INC(xs_qm_dquot_dups); - goto restart; + if (error == -EEXIST) { + /* + * Duplicate found. Just throw away the new dquot and + * start over. + */ + XFS_STATS_INC(mp, xs_qm_dquot_dups); + goto restart; + } + return error; } - /* - * We 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); - - dqret: - ASSERT((ip == NULL) || xfs_isilocked(ip, XFS_ILOCK_EXCL)); trace_xfs_dqget_miss(dqp); +found: *O_dqpp = dqp; - return (0); + return 0; } - -STATIC void -xfs_qm_dqput_final( - struct xfs_dquot *dqp) +/* + * 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, + xfs_dqtype_t type, + struct xfs_dquot **dqpp) { - struct xfs_quotainfo *qi = dqp->q_mount->m_quotainfo; - struct xfs_dquot *gdqp; - struct xfs_dquot *pdqp; + int error; - trace_xfs_dqput_free(dqp); + error = xfs_qm_dqget_checks(mp, type); + if (error) + return error; + + return xfs_qm_dqread(mp, id, type, 0, dqpp); +} - mutex_lock(&qi->qi_lru_lock); - if (list_empty(&dqp->q_lru)) { - list_add_tail(&dqp->q_lru, &qi->qi_lru_list); - qi->qi_lru_count++; - XFS_STATS_INC(xs_qm_dquot_unused); +/* Return the quota id for a given inode and type. */ +xfs_dqid_t +xfs_qm_id_for_quotatype( + struct xfs_inode *ip, + xfs_dqtype_t type) +{ + switch (type) { + case XFS_DQTYPE_USER: + return i_uid_read(VFS_I(ip)); + case XFS_DQTYPE_GROUP: + return i_gid_read(VFS_I(ip)); + case XFS_DQTYPE_PROJ: + return ip->i_projid; } - mutex_unlock(&qi->qi_lru_lock); + 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, + xfs_dqtype_t type, + bool can_alloc, + struct xfs_dquot **dqpp) +{ + struct xfs_mount *mp = ip->i_mount; + struct xfs_dquot *dqp; + xfs_dqid_t id; + int error; + + ASSERT(!*dqpp); + xfs_assert_ilocked(ip, XFS_ILOCK_EXCL); + + error = xfs_qm_dqget_checks(mp, type); + if (error) + return error; + + xfs_assert_ilocked(ip, XFS_ILOCK_EXCL); + ASSERT(xfs_inode_dquot(ip, type) == NULL); + ASSERT(!xfs_is_metadir_inode(ip)); + + id = xfs_qm_id_for_quotatype(ip, type); + +restart: + dqp = xfs_qm_dqget_cache_lookup(mp, id, type); + if (dqp) + goto found; /* - * If we just added a udquot to the freelist, then we want to release - * the gdquot/pdquot reference that it (probably) has. Otherwise it'll - * keep the gdquot/pdquot from getting reclaimed. + * 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. + */ + xfs_iunlock(ip, XFS_ILOCK_EXCL); + error = xfs_qm_dqread(mp, id, type, can_alloc, &dqp); + xfs_ilock(ip, XFS_ILOCK_EXCL); + if (error) + return error; + + /* + * A dquot could be attached to this inode by now, since we had + * dropped the ilock. */ - gdqp = dqp->q_gdquot; - if (gdqp) { - xfs_dqlock(gdqp); - dqp->q_gdquot = NULL; + if (xfs_this_quota_on(mp, type)) { + struct xfs_dquot *dqp1; + + dqp1 = xfs_inode_dquot(ip, type); + if (dqp1) { + xfs_qm_dqdestroy(dqp); + dqp = dqp1; + goto dqret; + } + } else { + /* inode stays locked on return */ + xfs_qm_dqdestroy(dqp); + return -ESRCH; } - pdqp = dqp->q_pdquot; - if (pdqp) { - xfs_dqlock(pdqp); - dqp->q_pdquot = NULL; + error = xfs_qm_dqget_cache_insert(mp, id, type, dqp); + if (error) { + xfs_qm_dqdestroy(dqp); + if (error == -EEXIST) { + /* + * Duplicate found. Just throw away the new dquot and + * start over. + */ + XFS_STATS_INC(mp, xs_qm_dquot_dups); + goto restart; + } + return error; } - xfs_dqunlock(dqp); - /* - * If we had a group/project quota hint, release it now. - */ - if (gdqp) - xfs_qm_dqput(gdqp); - if (pdqp) - xfs_qm_dqput(pdqp); +dqret: + xfs_assert_ilocked(ip, XFS_ILOCK_EXCL); + trace_xfs_dqget_miss(dqp); +found: + trace_xfs_dqattach_get(dqp); + *dqpp = dqp; + return 0; } /* - * 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. + * Starting at @id and progressing upwards, look for an initialized incore + * dquot, lock it, and return it. */ -void -xfs_qm_dqput( - struct xfs_dquot *dqp) +int +xfs_qm_dqget_next( + struct xfs_mount *mp, + xfs_dqid_t id, + xfs_dqtype_t type, + struct xfs_dquot **dqpp) { - ASSERT(dqp->q_nrefs > 0); - ASSERT(XFS_DQ_IS_LOCKED(dqp)); + struct xfs_dquot *dqp; + int error = 0; + + *dqpp = NULL; + for (; !error; error = xfs_dq_get_next_id(mp, type, &id)) { + error = xfs_qm_dqget(mp, id, type, false, &dqp); + if (error == -ENOENT) + continue; + else if (error != 0) + break; + + mutex_lock(&dqp->q_qlock); + if (!XFS_IS_DQUOT_UNINITIALIZED(dqp)) { + *dqpp = dqp; + return 0; + } - trace_xfs_dqput(dqp); + mutex_unlock(&dqp->q_qlock); + xfs_qm_dqrele(dqp); + } - if (--dqp->q_nrefs > 0) - xfs_dqunlock(dqp); - else - xfs_qm_dqput_final(dqp); + return error; } /* - * Release a dquot. Flush it if dirty, then dqput() it. - * dquot must not be locked. + * Release a reference to the dquot. */ void xfs_qm_dqrele( - xfs_dquot_t *dqp) + struct xfs_dquot *dqp) { if (!dqp) return; trace_xfs_dqrele(dqp); - 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); + if (lockref_put_or_lock(&dqp->q_lockref)) + return; + if (!--dqp->q_lockref.count) { + struct xfs_quotainfo *qi = dqp->q_mount->m_quotainfo; + + trace_xfs_dqrele_free(dqp); + if (list_lru_add_obj(&qi->qi_lru, &dqp->q_lru)) + XFS_STATS_INC(dqp->q_mount, xs_qm_dquot_unused); + } + spin_unlock(&dqp->q_lockref.lock); } /* @@ -1026,14 +1139,16 @@ xfs_qm_dqrele( * 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 +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_dq_logitem *qlip = + container_of(lip, struct xfs_dq_logitem, qli_item); + struct xfs_dquot *dqp = qlip->qli_dquot; struct xfs_ail *ailp = lip->li_ailp; + struct xfs_buf *bp = NULL; + xfs_lsn_t tail_lsn; /* * We only want to pull the item from the AIL if its @@ -1043,16 +1158,33 @@ xfs_qm_dqflush_done( * since it's cheaper, and then we recheck while * holding the lock before removing the dquot from the AIL. */ - if ((lip->li_flags & XFS_LI_IN_AIL) && - lip->li_lsn == qip->qli_flush_lsn) { - - /* xfs_trans_ail_delete() drops the AIL lock. */ - spin_lock(&ailp->xa_lock); - if (lip->li_lsn == qip->qli_flush_lsn) - xfs_trans_ail_delete(ailp, lip, SHUTDOWN_CORRUPT_INCORE); - else - spin_unlock(&ailp->xa_lock); + if (test_bit(XFS_LI_IN_AIL, &lip->li_flags) && + (lip->li_lsn == qlip->qli_flush_lsn || + test_bit(XFS_LI_FAILED, &lip->li_flags))) { + spin_lock(&ailp->ail_lock); + clear_bit(XFS_LI_FAILED, &lip->li_flags); + if (lip->li_lsn == qlip->qli_flush_lsn) { + /* xfs_ail_update_finish() drops the AIL lock */ + tail_lsn = xfs_ail_delete_one(ailp, lip); + xfs_ail_update_finish(ailp, tail_lsn); + } else { + spin_unlock(&ailp->ail_lock); + } + } + + /* + * If this dquot hasn't been dirtied since initiating the last dqflush, + * release the buffer reference. We already unlinked this dquot item + * from the buffer. + */ + spin_lock(&qlip->qli_lock); + if (!qlip->qli_dirty) { + bp = lip->li_buf; + lip->li_buf = NULL; } + spin_unlock(&qlip->qli_lock); + if (bp) + xfs_buf_rele(bp); /* * Release the dq's flush lock since we're done with it. @@ -1060,6 +1192,165 @@ xfs_qm_dqflush_done( xfs_dqfunlock(dqp); } +void +xfs_buf_dquot_iodone( + struct xfs_buf *bp) +{ + struct xfs_log_item *lip, *n; + + list_for_each_entry_safe(lip, n, &bp->b_li_list, li_bio_list) { + list_del_init(&lip->li_bio_list); + xfs_qm_dqflush_done(lip); + } +} + +/* Check incore dquot for errors before we flush. */ +static xfs_failaddr_t +xfs_qm_dqflush_check( + struct xfs_dquot *dqp) +{ + xfs_dqtype_t type = xfs_dquot_type(dqp); + + if (type != XFS_DQTYPE_USER && + type != XFS_DQTYPE_GROUP && + type != XFS_DQTYPE_PROJ) + return __this_address; + + if (dqp->q_id == 0) + return NULL; + + if (dqp->q_blk.softlimit && dqp->q_blk.count > dqp->q_blk.softlimit && + !dqp->q_blk.timer) + return __this_address; + + if (dqp->q_ino.softlimit && dqp->q_ino.count > dqp->q_ino.softlimit && + !dqp->q_ino.timer) + return __this_address; + + if (dqp->q_rtb.softlimit && dqp->q_rtb.count > dqp->q_rtb.softlimit && + !dqp->q_rtb.timer) + return __this_address; + + /* bigtime flag should never be set on root dquots */ + if (dqp->q_type & XFS_DQTYPE_BIGTIME) { + if (!xfs_has_bigtime(dqp->q_mount)) + return __this_address; + if (dqp->q_id == 0) + return __this_address; + } + + return NULL; +} + +/* + * Get the buffer containing the on-disk dquot. + * + * Requires dquot flush lock, will clear the dirty flag, delete the quota log + * item from the AIL, and shut down the system if something goes wrong. + */ +static int +xfs_dquot_read_buf( + struct xfs_trans *tp, + struct xfs_dquot *dqp, + struct xfs_buf **bpp) +{ + struct xfs_mount *mp = dqp->q_mount; + struct xfs_buf *bp = NULL; + int error; + + error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, dqp->q_blkno, + mp->m_quotainfo->qi_dqchunklen, 0, + &bp, &xfs_dquot_buf_ops); + if (xfs_metadata_is_sick(error)) + xfs_dquot_mark_sick(dqp); + if (error) + goto out_abort; + + *bpp = bp; + return 0; + +out_abort: + dqp->q_flags &= ~XFS_DQFLAG_DIRTY; + xfs_trans_ail_delete(&dqp->q_logitem.qli_item, 0); + xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); + return error; +} + +/* + * Attach a dquot buffer to this dquot to avoid allocating a buffer during a + * dqflush, since dqflush can be called from reclaim context. Caller must hold + * the dqlock. + */ +int +xfs_dquot_attach_buf( + struct xfs_trans *tp, + struct xfs_dquot *dqp) +{ + struct xfs_dq_logitem *qlip = &dqp->q_logitem; + struct xfs_log_item *lip = &qlip->qli_item; + int error; + + spin_lock(&qlip->qli_lock); + if (!lip->li_buf) { + struct xfs_buf *bp = NULL; + + spin_unlock(&qlip->qli_lock); + error = xfs_dquot_read_buf(tp, dqp, &bp); + if (error) + return error; + + /* + * Hold the dquot buffer so that we retain our ref to it after + * detaching it from the transaction, then give that ref to the + * dquot log item so that the AIL does not have to read the + * dquot buffer to push this item. + */ + xfs_buf_hold(bp); + xfs_trans_brelse(tp, bp); + + spin_lock(&qlip->qli_lock); + lip->li_buf = bp; + } + qlip->qli_dirty = true; + spin_unlock(&qlip->qli_lock); + + return 0; +} + +/* + * Get a new reference the dquot buffer attached to this dquot for a dqflush + * operation. + * + * Returns 0 and a NULL bp if none was attached to the dquot; 0 and a locked + * bp; or -EAGAIN if the buffer could not be locked. + */ +int +xfs_dquot_use_attached_buf( + struct xfs_dquot *dqp, + struct xfs_buf **bpp) +{ + struct xfs_buf *bp = dqp->q_logitem.qli_item.li_buf; + + /* + * A NULL buffer can happen if the dquot dirty flag was set but the + * filesystem shut down before transaction commit happened. In that + * case we're not going to flush anyway. + */ + if (!bp) { + ASSERT(xfs_is_shutdown(dqp->q_mount)); + + *bpp = NULL; + return 0; + } + + if (!xfs_buf_trylock(bp)) + return -EAGAIN; + + xfs_buf_hold(bp); + *bpp = bp; + return 0; +} + /* * Write a modified dquot to disk. * The dquot must be locked and the flush lock too taken by caller. @@ -1071,79 +1362,47 @@ xfs_qm_dqflush_done( int xfs_qm_dqflush( struct xfs_dquot *dqp, - struct xfs_buf **bpp) + struct xfs_buf *bp) { struct xfs_mount *mp = dqp->q_mount; - struct xfs_buf *bp; - struct xfs_disk_dquot *ddqp; + struct xfs_dq_logitem *qlip = &dqp->q_logitem; + struct xfs_log_item *lip = &qlip->qli_item; + struct xfs_dqblk *dqblk; + xfs_failaddr_t fa; int error; ASSERT(XFS_DQ_IS_LOCKED(dqp)); ASSERT(!completion_done(&dqp->q_flush)); + ASSERT(atomic_read(&dqp->q_pincount) == 0); trace_xfs_dqflush(dqp); - - *bpp = NULL; - - 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 - * 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. - */ - if (XFS_FORCED_SHUTDOWN(mp)) { - struct xfs_log_item *lip = &dqp->q_logitem.qli_item; - dqp->dq_flags &= ~XFS_DQ_DIRTY; - - spin_lock(&mp->m_ail->xa_lock); - if (lip->li_flags & XFS_LI_IN_AIL) - xfs_trans_ail_delete(mp->m_ail, lip, - SHUTDOWN_CORRUPT_INCORE); - else - spin_unlock(&mp->m_ail->xa_lock); - error = XFS_ERROR(EIO); - goto out_unlock; + fa = xfs_qm_dqflush_check(dqp); + if (fa) { + xfs_alert(mp, "corrupt dquot ID 0x%x in memory at %pS", + dqp->q_id, fa); + xfs_dquot_mark_sick(dqp); + error = -EFSCORRUPTED; + goto out_abort; } - /* - * Get the buffer containing the on-disk dquot - */ - error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno, - mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL); - if (error) - goto out_unlock; - - /* - * Calculate the location of the dquot inside the buffer. - */ - ddqp = bp->b_addr + dqp->q_bufoffset; + /* Flush the incore dquot to the ondisk buffer. */ + dqblk = xfs_buf_offset(bp, dqp->q_bufoffset); + xfs_dquot_to_disk(&dqblk->dd_diskdq, dqp); /* - * A simple sanity check in case we got a corrupted dquot.. + * Clear the dirty field and remember the flush lsn for later use. */ - error = xfs_qm_dqcheck(mp, &dqp->q_core, be32_to_cpu(ddqp->d_id), 0, - XFS_QMOPT_DOWARN, "dqflush (incore copy)"); - if (error) { - xfs_buf_relse(bp); - xfs_dqfunlock(dqp); - xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); - return XFS_ERROR(EIO); - } - - /* This is the only portion of data that needs to persist */ - memcpy(ddqp, &dqp->q_core, sizeof(xfs_disk_dquot_t)); + dqp->q_flags &= ~XFS_DQFLAG_DIRTY; /* - * Clear the dirty field and remember the flush lsn for later use. + * We hold the dquot lock, so nobody can dirty it while we're + * scheduling the write out. Clear the dirty-since-flush flag. */ - dqp->dq_flags &= ~XFS_DQ_DIRTY; + spin_lock(&qlip->qli_lock); + qlip->qli_dirty = false; + spin_unlock(&qlip->qli_lock); - xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn, - &dqp->q_logitem.qli_item.li_lsn); + xfs_trans_ail_copy_lsn(mp->m_ail, &qlip->qli_flush_lsn, &lip->li_lsn); /* * copy the lsn into the on-disk dquot now while we have the in memory @@ -1154,20 +1413,18 @@ xfs_qm_dqflush( * buffer always has a valid CRC. This ensures there is no possibility * of a dquot without an up-to-date CRC getting to disk. */ - if (xfs_sb_version_hascrc(&mp->m_sb)) { - struct xfs_dqblk *dqb = (struct xfs_dqblk *)ddqp; - - dqb->dd_lsn = cpu_to_be64(dqp->q_logitem.qli_item.li_lsn); - xfs_update_cksum((char *)dqb, sizeof(struct xfs_dqblk), + if (xfs_has_crc(mp)) { + dqblk->dd_lsn = cpu_to_be64(lip->li_lsn); + xfs_update_cksum((char *)dqblk, sizeof(struct xfs_dqblk), XFS_DQUOT_CRC_OFF); } /* - * 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. + * Attach the dquot to the buffer so that we can remove this dquot from + * the AIL and release the flush lock once the dquot is synced to disk. */ - xfs_buf_attach_iodone(bp, xfs_qm_dqflush_done, - &dqp->q_logitem.qli_item); + bp->b_iodone = xfs_buf_dquot_iodone; + list_add_tail(&lip->li_bio_list, &bp->b_li_list); /* * If the buffer is pinned then push on the log so we won't @@ -1179,12 +1436,14 @@ xfs_qm_dqflush( } trace_xfs_dqflush_done(dqp); - *bpp = bp; return 0; -out_unlock: +out_abort: + dqp->q_flags &= ~XFS_DQFLAG_DIRTY; + xfs_trans_ail_delete(lip, 0); + xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); xfs_dqfunlock(dqp); - return XFS_ERROR(EIO); + return error; } /* @@ -1195,13 +1454,12 @@ out_unlock: */ void xfs_dqlock2( - xfs_dquot_t *d1, - xfs_dquot_t *d2) + struct xfs_dquot *d1, + struct xfs_dquot *d2) { if (d1 && d2) { ASSERT(d1 != d2); - if (be32_to_cpu(d1->q_core.d_id) > - be32_to_cpu(d2->q_core.d_id)) { + if (d1->q_id > d2->q_id) { mutex_lock(&d2->q_qlock); mutex_lock_nested(&d1->q_qlock, XFS_QLOCK_NESTED); } else { @@ -1215,23 +1473,66 @@ xfs_dqlock2( } } +static int +xfs_dqtrx_cmp( + const void *a, + const void *b) +{ + const struct xfs_dqtrx *qa = a; + const struct xfs_dqtrx *qb = b; + + if (qa->qt_dquot->q_id > qb->qt_dquot->q_id) + return 1; + if (qa->qt_dquot->q_id < qb->qt_dquot->q_id) + return -1; + return 0; +} + +void +xfs_dqlockn( + struct xfs_dqtrx *q) +{ + unsigned int i; + + BUILD_BUG_ON(XFS_QM_TRANS_MAXDQS > MAX_LOCKDEP_SUBCLASSES); + + /* Sort in order of dquot id, do not allow duplicates */ + for (i = 0; i < XFS_QM_TRANS_MAXDQS && q[i].qt_dquot != NULL; i++) { + unsigned int j; + + for (j = 0; j < i; j++) + ASSERT(q[i].qt_dquot != q[j].qt_dquot); + } + if (i == 0) + return; + + sort(q, i, sizeof(struct xfs_dqtrx), xfs_dqtrx_cmp, NULL); + + mutex_lock(&q[0].qt_dquot->q_qlock); + for (i = 1; i < XFS_QM_TRANS_MAXDQS && q[i].qt_dquot != NULL; i++) + mutex_lock_nested(&q[i].qt_dquot->q_qlock, + XFS_QLOCK_NESTED + i - 1); +} + int __init xfs_qm_init(void) { - xfs_qm_dqzone = - kmem_zone_init(sizeof(struct xfs_dquot), "xfs_dquot"); - if (!xfs_qm_dqzone) + xfs_dquot_cache = kmem_cache_create("xfs_dquot", + sizeof(struct xfs_dquot), + 0, 0, NULL); + if (!xfs_dquot_cache) goto out; - xfs_qm_dqtrxzone = - kmem_zone_init(sizeof(struct xfs_dquot_acct), "xfs_dqtrx"); - if (!xfs_qm_dqtrxzone) - goto out_free_dqzone; + xfs_dqtrx_cache = kmem_cache_create("xfs_dqtrx", + sizeof(struct xfs_dquot_acct), + 0, 0, NULL); + if (!xfs_dqtrx_cache) + goto out_free_dquot_cache; return 0; -out_free_dqzone: - kmem_zone_destroy(xfs_qm_dqzone); +out_free_dquot_cache: + kmem_cache_destroy(xfs_dquot_cache); out: return -ENOMEM; } @@ -1239,6 +1540,6 @@ out: void xfs_qm_exit(void) { - kmem_zone_destroy(xfs_qm_dqtrxzone); - kmem_zone_destroy(xfs_qm_dqzone); + kmem_cache_destroy(xfs_dqtrx_cache); + kmem_cache_destroy(xfs_dquot_cache); } |