xfs_qm.c

linux 内核源代码

		 * the inode is never added to the transaction.
		 */
		xfs_ilock(qip, XFS_ILOCK_SHARED);
		error = xfs_bmapi(NULL, qip, lblkno,
				  maxlblkcnt - lblkno,
				  XFS_BMAPI_METADATA,
				  NULL,
				  0, map, &nmaps, NULL, NULL);
		xfs_iunlock(qip, XFS_ILOCK_SHARED);
		if (error)
			break;

		ASSERT(nmaps <= XFS_DQITER_MAP_SIZE);
		for (i = 0; i < nmaps; i++) {
			ASSERT(map[i].br_startblock != DELAYSTARTBLOCK);
			ASSERT(map[i].br_blockcount);


			lblkno += map[i].br_blockcount;

			if (map[i].br_startblock == HOLESTARTBLOCK)
				continue;

			firstid = (xfs_dqid_t) map[i].br_startoff *
				XFS_QM_DQPERBLK(mp);
			/*
			 * Do a read-ahead on the next extent.
			 */
			if ((i+1 < nmaps) &&
			    (map[i+1].br_startblock != HOLESTARTBLOCK)) {
				rablkcnt =  map[i+1].br_blockcount;
				rablkno = map[i+1].br_startblock;
				while (rablkcnt--) {
					xfs_baread(mp->m_ddev_targp,
					       XFS_FSB_TO_DADDR(mp, rablkno),
					       (int)XFS_QI_DQCHUNKLEN(mp));
					rablkno++;
				}
			}
			/*
			 * Iterate thru all the blks in the extent and
			 * reset the counters of all the dquots inside them.
			 */
			if ((error = xfs_qm_dqiter_bufs(mp,
						       firstid,
						       map[i].br_startblock,
						       map[i].br_blockcount,
						       flags))) {
				break;
			}
		}

		if (error)
			break;
	} while (nmaps > 0);

	kmem_free(map, XFS_DQITER_MAP_SIZE * sizeof(*map));

	return error;
}

/*
 * Called by dqusage_adjust in doing a quotacheck.
 * Given the inode, and a dquot (either USR or GRP, doesn't matter),
 * this updates its incore copy as well as the buffer copy. This is
 * so that once the quotacheck is done, we can just log all the buffers,
 * as opposed to logging numerous updates to individual dquots.
 */
STATIC void
xfs_qm_quotacheck_dqadjust(
	xfs_dquot_t		*dqp,
	xfs_qcnt_t		nblks,
	xfs_qcnt_t		rtblks)
{
	ASSERT(XFS_DQ_IS_LOCKED(dqp));
	xfs_dqtrace_entry(dqp, "QCHECK DQADJUST");
	/*
	 * Adjust the inode count and the block count to reflect this inode's
	 * resource usage.
	 */
	be64_add(&dqp->q_core.d_icount, 1);
	dqp->q_res_icount++;
	if (nblks) {
		be64_add(&dqp->q_core.d_bcount, nblks);
		dqp->q_res_bcount += nblks;
	}
	if (rtblks) {
		be64_add(&dqp->q_core.d_rtbcount, rtblks);
		dqp->q_res_rtbcount += rtblks;
	}

	/*
	 * Set default limits, adjust timers (since we changed usages)
	 */
	if (! XFS_IS_SUSER_DQUOT(dqp)) {
		xfs_qm_adjust_dqlimits(dqp->q_mount, &dqp->q_core);
		xfs_qm_adjust_dqtimers(dqp->q_mount, &dqp->q_core);
	}

	dqp->dq_flags |= XFS_DQ_DIRTY;
}

STATIC int
xfs_qm_get_rtblks(
	xfs_inode_t	*ip,
	xfs_qcnt_t	*O_rtblks)
{
	xfs_filblks_t	rtblks;			/* total rt blks */
	xfs_extnum_t	idx;			/* extent record index */
	xfs_ifork_t	*ifp;			/* inode fork pointer */
	xfs_extnum_t	nextents;		/* number of extent entries */
	int		error;

	ASSERT(XFS_IS_REALTIME_INODE(ip));
	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
	if (!(ifp->if_flags & XFS_IFEXTENTS)) {
		if ((error = xfs_iread_extents(NULL, ip, XFS_DATA_FORK)))
			return error;
	}
	rtblks = 0;
	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
	for (idx = 0; idx < nextents; idx++)
		rtblks += xfs_bmbt_get_blockcount(xfs_iext_get_ext(ifp, idx));
	*O_rtblks = (xfs_qcnt_t)rtblks;
	return 0;
}

/*
 * callback routine supplied to bulkstat(). Given an inumber, find its
 * dquots and update them to account for resources taken by that inode.
 */
/* ARGSUSED */
STATIC int
xfs_qm_dqusage_adjust(
	xfs_mount_t	*mp,		/* mount point for filesystem */
	xfs_ino_t	ino,		/* inode number to get data for */
	void		__user *buffer,	/* not used */
	int		ubsize,		/* not used */
	void		*private_data,	/* not used */
	xfs_daddr_t	bno,		/* starting block of inode cluster */
	int		*ubused,	/* not used */
	void		*dip,		/* on-disk inode pointer (not used) */
	int		*res)		/* result code value */
{
	xfs_inode_t	*ip;
	xfs_dquot_t	*udqp, *gdqp;
	xfs_qcnt_t	nblks, rtblks;
	int		error;

	ASSERT(XFS_IS_QUOTA_RUNNING(mp));

	/*
	 * rootino must have its resources accounted for, not so with the quota
	 * inodes.
	 */
	if (ino == mp->m_sb.sb_uquotino || ino == mp->m_sb.sb_gquotino) {
		*res = BULKSTAT_RV_NOTHING;
		return XFS_ERROR(EINVAL);
	}

	/*
	 * We don't _need_ to take the ilock EXCL. However, the xfs_qm_dqget
	 * interface expects the inode to be exclusively locked because that's
	 * the case in all other instances. It's OK that we do this because
	 * quotacheck is done only at mount time.
	 */
	if ((error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_EXCL, &ip, bno))) {
		*res = BULKSTAT_RV_NOTHING;
		return error;
	}

	if (ip->i_d.di_mode == 0) {
		xfs_iput_new(ip, XFS_ILOCK_EXCL);
		*res = BULKSTAT_RV_NOTHING;
		return XFS_ERROR(ENOENT);
	}

	/*
	 * Obtain the locked dquots. In case of an error (eg. allocation
	 * fails for ENOSPC), we return the negative of the error number
	 * to bulkstat, so that it can get propagated to quotacheck() and
	 * making us disable quotas for the file system.
	 */
	if ((error = xfs_qm_dqget_noattach(ip, &udqp, &gdqp))) {
		xfs_iput(ip, XFS_ILOCK_EXCL);
		*res = BULKSTAT_RV_GIVEUP;
		return error;
	}

	rtblks = 0;
	if (! XFS_IS_REALTIME_INODE(ip)) {
		nblks = (xfs_qcnt_t)ip->i_d.di_nblocks;
	} else {
		/*
		 * Walk thru the extent list and count the realtime blocks.
		 */
		if ((error = xfs_qm_get_rtblks(ip, &rtblks))) {
			xfs_iput(ip, XFS_ILOCK_EXCL);
			if (udqp)
				xfs_qm_dqput(udqp);
			if (gdqp)
				xfs_qm_dqput(gdqp);
			*res = BULKSTAT_RV_GIVEUP;
			return error;
		}
		nblks = (xfs_qcnt_t)ip->i_d.di_nblocks - rtblks;
	}
	ASSERT(ip->i_delayed_blks == 0);

	/*
	 * We can't release the inode while holding its dquot locks.
	 * The inode can go into inactive and might try to acquire the dquotlocks.
	 * So, just unlock here and do a vn_rele at the end.
	 */
	xfs_iunlock(ip, XFS_ILOCK_EXCL);

	/*
	 * Add the (disk blocks and inode) resources occupied by this
	 * inode to its dquots. We do this adjustment in the incore dquot,
	 * and also copy the changes to its buffer.
	 * We don't care about putting these changes in a transaction
	 * envelope because if we crash in the middle of a 'quotacheck'
	 * we have to start from the beginning anyway.
	 * Once we're done, we'll log all the dquot bufs.
	 *
	 * The *QUOTA_ON checks below may look pretty racy, but quotachecks
	 * and quotaoffs don't race. (Quotachecks happen at mount time only).
	 */
	if (XFS_IS_UQUOTA_ON(mp)) {
		ASSERT(udqp);
		xfs_qm_quotacheck_dqadjust(udqp, nblks, rtblks);
		xfs_qm_dqput(udqp);
	}
	if (XFS_IS_OQUOTA_ON(mp)) {
		ASSERT(gdqp);
		xfs_qm_quotacheck_dqadjust(gdqp, nblks, rtblks);
		xfs_qm_dqput(gdqp);
	}
	/*
	 * Now release the inode. This will send it to 'inactive', and
	 * possibly even free blocks.
	 */
	VN_RELE(XFS_ITOV(ip));

	/*
	 * Goto next inode.
	 */
	*res = BULKSTAT_RV_DIDONE;
	return 0;
}

/*
 * Walk thru all the filesystem inodes and construct a consistent view
 * of the disk quota world. If the quotacheck fails, disable quotas.
 */
int
xfs_qm_quotacheck(
	xfs_mount_t	*mp)
{
	int		done, count, error;
	xfs_ino_t	lastino;
	size_t		structsz;
	xfs_inode_t	*uip, *gip;
	uint		flags;

	count = INT_MAX;
	structsz = 1;
	lastino = 0;
	flags = 0;

	ASSERT(XFS_QI_UQIP(mp) || XFS_QI_GQIP(mp));
	ASSERT(XFS_IS_QUOTA_RUNNING(mp));

	/*
	 * There should be no cached dquots. The (simplistic) quotacheck
	 * algorithm doesn't like that.
	 */
	ASSERT(XFS_QI_MPLNDQUOTS(mp) == 0);

	cmn_err(CE_NOTE, "XFS quotacheck %s: Please wait.", mp->m_fsname);

	/*
	 * First we go thru all the dquots on disk, USR and GRP/PRJ, and reset
	 * their counters to zero. We need a clean slate.
	 * We don't log our changes till later.
	 */
	if ((uip = XFS_QI_UQIP(mp))) {
		if ((error = xfs_qm_dqiterate(mp, uip, XFS_QMOPT_UQUOTA)))
			goto error_return;
		flags |= XFS_UQUOTA_CHKD;
	}

	if ((gip = XFS_QI_GQIP(mp))) {
		if ((error = xfs_qm_dqiterate(mp, gip, XFS_IS_GQUOTA_ON(mp) ?
					XFS_QMOPT_GQUOTA : XFS_QMOPT_PQUOTA)))
			goto error_return;
		flags |= XFS_OQUOTA_CHKD;
	}

	do {
		/*
		 * Iterate thru all the inodes in the file system,
		 * adjusting the corresponding dquot counters in core.
		 */
		if ((error = xfs_bulkstat(mp, &lastino, &count,
				     xfs_qm_dqusage_adjust, NULL,
				     structsz, NULL, BULKSTAT_FG_IGET, &done)))
			break;

	} while (! done);

	/*
	 * We can get this error if we couldn't do a dquot allocation inside
	 * xfs_qm_dqusage_adjust (via bulkstat). We don't care about the
	 * dirty dquots that might be cached, we just want to get rid of them
	 * and turn quotaoff. The dquots won't be attached to any of the inodes
	 * at this point (because we intentionally didn't in dqget_noattach).
	 */
	if (error) {
		xfs_qm_dqpurge_all(mp, XFS_QMOPT_QUOTALL | XFS_QMOPT_QUOTAOFF);
		goto error_return;
	}
	/*
	 * We've made all the changes that we need to make incore.
	 * Now flush_them down to disk buffers.
	 */
	xfs_qm_dqflush_all(mp, XFS_QMOPT_DELWRI);

	/*
	 * We didn't log anything, because if we crashed, we'll have to
	 * start the quotacheck from scratch anyway. However, we must make
	 * sure that our dquot changes are secure before we put the
	 * quotacheck'd stamp on the superblock. So, here we do a synchronous
	 * flush.
	 */
	XFS_bflush(mp->m_ddev_targp);

	/*
	 * If one type of quotas is off, then it will lose its
	 * quotachecked status, since we won't be doing accounting for
	 * that type anymore.
	 */
	mp->m_qflags &= ~(XFS_OQUOTA_CHKD | XFS_UQUOTA_CHKD);
	mp->m_qflags |= flags;

	XQM_LIST_PRINT(&(XFS_QI_MPL_LIST(mp)), MPL_NEXT, "++++ Mp list +++");

 error_return:
	if (error) {
		cmn_err(CE_WARN, "XFS quotacheck %s: Unsuccessful (Error %d): "
			"Disabling quotas.",
			mp->m_fsname, error);
		/*
		 * We must turn off quotas.
		 */
		ASSERT(mp->m_quotainfo != NULL);
		ASSERT(xfs_Gqm != NULL);
		xfs_qm_destroy_quotainfo(mp);
		(void)xfs_mount_reset_sbqflags(mp);
	} else {
		cmn_err(CE_NOTE, "XFS quotacheck %s: Done.", mp->m_fsname);
	}
	return (error);
}

/*
 * This is called after the superblock has been read in and we're ready to
 * iget the quota inodes.
 */
STATIC int
xfs_qm_init_quotainos(
	xfs_mount_t	*mp)
{
	xfs_inode_t	*uip, *gip;
	int		error;
	__int64_t	sbflags;
	uint		flags;

	ASSERT(mp->m_quotainfo);
	uip = gip = NULL;
	sbflags = 0;
	flags = 0;

	/*
	 * Get the uquota and gquota inodes
	 */
	if (XFS_SB_VERSION_HASQUOTA(&mp->m_sb)) {
		if (XFS_IS_UQUOTA_ON(mp) &&
		    mp->m_sb.sb_uquotino != NULLFSINO) {
			ASSERT(mp->m_sb.sb_uquotino > 0);
			if ((error = xfs_iget(mp, NULL, mp->m_sb.sb_uquotino,
					     0, 0, &uip, 0)))
				return XFS_ERROR(error);
		}
		if (XFS_IS_OQUOTA_ON(mp) &&
		    mp->m_sb.sb_gquotino != NULLFSINO) {
			ASSERT(mp->m_sb.sb_gquotino > 0);
			if ((error = xfs_iget(mp, NULL, mp->m_sb.sb_gquotino,
					     0, 0, &gip, 0))) {
				if (uip)
					VN_RELE(XFS_ITOV(uip));
				return XFS_ERROR(error);
			}
		}
	} else {
		flags |= XFS_QMOPT_SBVERSION;
		sbflags |= (XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO |
			    XFS_SB_GQUOTINO | XFS_SB_QFLAGS);
	}

	/*
	 * Create the two inodes, if they don't exist already. The changes
	 * made above will get added to a transaction and logged in one of
	 * the qino_alloc calls below.  If the device is readonly,
	 * temporarily switch to read-write to do this.
	 */
	if (XFS_IS_UQUOTA_ON(mp) && uip == NULL) {
		if ((error = xfs_qm_qino_alloc(mp, &uip,
					      sbflags | XFS_SB_UQUOTINO,
					      flags | XFS_QMOPT_UQUOTA)))
			return XFS_ERROR(error);

		flags &= ~XFS_QMOPT_SBVERSION;
	}
	if (XFS_IS_OQUOTA_ON(mp) && gip == NULL) {
		flags |= (XFS_IS_GQUOTA_ON(mp) ?
				XFS_QMOPT_GQUOTA : XFS_QMOPT_PQUOTA);
		error = xfs_qm_qino_alloc(mp, &gip,
					  sbflags | XFS_SB_GQUOTINO, flags);
		if (error) {
			if (uip)
				VN_RELE(XFS_ITOV(uip));

			return XFS_ERROR(error);
		}
	}

	XFS_QI_UQIP(mp) = uip;
	XFS_QI_GQIP(mp) = gip;

	return 0;
}


/*
 * Traverse the freelist of dquots and attempt to reclaim a maximum of
 * 'howmany' dquots. This operation races with dqlookup(), and attempts to
 * favor the lookup function ...
 * XXXsup merge this with qm_reclaim_one().
 */
STATIC int
xfs_qm_shake_freelist(
	int howmany)
{
	int		nreclaimed;
	xfs_dqhash_t	*hash;
	xfs_dquot_t	*dqp, *nextdqp;
	int		restarts;
	int		nflushes;

	if (howmany <= 0)
		return 0;

	nreclaimed = 0;
	restarts = 0;
	nflushes = 0;

#ifdef QUOTADEBUG
	cmn_err(CE_DEBUG, "Shake free 0x%x", howmany);
#endif
	/* lock order is : hashchainlock, freelistlock, mplistlock */
 tryagain:
	xfs_qm_freelist_lock(xfs_Gqm);

	for (dqp = xfs_Gqm->qm_dqfreelist.qh_next;
	     ((dqp != (xfs_dquot_t *) &xfs_Gqm->qm_dqfreelist) &&
	      nreclaimed < howmany); ) {
		xfs_dqlock(dqp);

		/*
		 * We are racing with dqlookup here. Naturally we don't
		 * want to reclaim a dquot that lookup wants.
		 */
		if (dqp->dq_flags & XFS_DQ_WANT) {
			xfs_dqunlock(dqp);
			xfs_qm_freelist_unlock(xfs_Gqm);
			if (++restarts >= XFS_QM_RECLAIM_MAX_RESTARTS)
				return nreclaimed;
			XQM_STATS_INC(xqmstats.xs_qm_dqwants);
			goto tryagain;
		}

		/*
		 * If the dquot is inactive, we are assured that it is
		 * not on the mplist or the hashlist, and that makes our
		 * life easier.
		 */
		if (dqp->dq_flags & XFS_DQ_INACTIVE) {
			ASSERT(dqp->q_mount == NULL);
			ASSERT(! XFS_DQ_IS_DIRTY(dqp));
			ASSERT(dqp->HL_PREVP == NULL);
			ASSERT(dqp->MPL_PREVP == NULL);
			XQM_STATS_INC(xqmstats.xs_qm_dqinact_reclaims);
			nextdqp = dqp->dq_flnext;
			goto off_freelist;
		}

		ASSERT(dqp->MPL_PREVP);
		/*
		 * Try to grab the flush lock. If this dquot is in the process of
		 * getting flushed to disk, we don't want to reclaim it.
		 */
		if (! xfs_qm_dqflock_nowait(dqp)) {
			xfs_dqunlock(dqp);
			dqp = dqp->dq_flnext;
			continue;
		}

		/*
		 * We have the flush lock so we know that this is not in the
		 * process of being flushed. So, if this is dirty, flush it
		 * DELWRI so that we don't get a freelist infested with
		 * dirty dquots.
		 */