2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2006-06-09 06:48:12 +02:00
|
|
|
* Copyright (c) 2000-2006 Silicon Graphics, Inc.
|
2005-11-02 04:58:39 +01:00
|
|
|
* All Rights Reserved.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License as
|
2005-04-17 00:20:36 +02:00
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* 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.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2005-11-02 04:58:39 +01:00
|
|
|
* 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
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2006-01-11 21:17:46 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_fs.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_types.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_bit.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_log.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_inum.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_trans.h"
|
|
|
|
#include "xfs_sb.h"
|
|
|
|
#include "xfs_ag.h"
|
|
|
|
#include "xfs_dir2.h"
|
|
|
|
#include "xfs_dmapi.h"
|
|
|
|
#include "xfs_mount.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_da_btree.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_bmap_btree.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_alloc_btree.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_ialloc_btree.h"
|
|
|
|
#include "xfs_dir2_sf.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_attr_sf.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_dinode.h"
|
|
|
|
#include "xfs_inode.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_inode_item.h"
|
|
|
|
#include "xfs_itable.h"
|
|
|
|
#include "xfs_btree.h"
|
|
|
|
#include "xfs_ialloc.h"
|
|
|
|
#include "xfs_alloc.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_bmap.h"
|
|
|
|
#include "xfs_attr.h"
|
|
|
|
#include "xfs_rw.h"
|
|
|
|
#include "xfs_error.h"
|
|
|
|
#include "xfs_quota.h"
|
|
|
|
#include "xfs_utils.h"
|
2005-11-02 04:38:42 +01:00
|
|
|
#include "xfs_rtalloc.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
#include "xfs_trans_space.h"
|
|
|
|
#include "xfs_log_priv.h"
|
2007-07-11 03:09:12 +02:00
|
|
|
#include "xfs_filestream.h"
|
2007-08-28 08:12:30 +02:00
|
|
|
#include "xfs_vnodeops.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
int
|
|
|
|
xfs_setattr(
|
2008-07-18 09:13:28 +02:00
|
|
|
struct xfs_inode *ip,
|
|
|
|
struct iattr *iattr,
|
2008-10-30 08:27:48 +01:00
|
|
|
int flags)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
2008-08-13 08:00:45 +02:00
|
|
|
struct inode *inode = VFS_I(ip);
|
2008-07-18 09:13:28 +02:00
|
|
|
int mask = iattr->ia_valid;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_t *tp;
|
|
|
|
int code;
|
|
|
|
uint lock_flags;
|
|
|
|
uint commit_flags=0;
|
|
|
|
uid_t uid=0, iuid=0;
|
|
|
|
gid_t gid=0, igid=0;
|
|
|
|
int timeflags = 0;
|
|
|
|
struct xfs_dquot *udqp, *gdqp, *olddquot1, *olddquot2;
|
2005-05-05 22:27:19 +02:00
|
|
|
int need_iolock = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-30 09:21:12 +02:00
|
|
|
if (mp->m_flags & XFS_MOUNT_RDONLY)
|
2005-04-17 00:20:36 +02:00
|
|
|
return XFS_ERROR(EROFS);
|
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2008-12-09 10:47:34 +01:00
|
|
|
code = -inode_change_ok(inode, iattr);
|
|
|
|
if (code)
|
|
|
|
return code;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
olddquot1 = olddquot2 = NULL;
|
|
|
|
udqp = gdqp = NULL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If disk quotas is on, we make sure that the dquots do exist on disk,
|
|
|
|
* before we start any other transactions. Trying to do this later
|
|
|
|
* is messy. We don't care to take a readlock to look at the ids
|
|
|
|
* in inode here, because we can't hold it across the trans_reserve.
|
|
|
|
* If the IDs do change before we take the ilock, we're covered
|
|
|
|
* because the i_*dquot fields will get updated anyway.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) {
|
2005-04-17 00:20:36 +02:00
|
|
|
uint qflags = 0;
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) {
|
|
|
|
uid = iattr->ia_uid;
|
2005-04-17 00:20:36 +02:00
|
|
|
qflags |= XFS_QMOPT_UQUOTA;
|
|
|
|
} else {
|
|
|
|
uid = ip->i_d.di_uid;
|
|
|
|
}
|
2008-07-18 09:13:28 +02:00
|
|
|
if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) {
|
|
|
|
gid = iattr->ia_gid;
|
2005-04-17 00:20:36 +02:00
|
|
|
qflags |= XFS_QMOPT_GQUOTA;
|
|
|
|
} else {
|
|
|
|
gid = ip->i_d.di_gid;
|
|
|
|
}
|
2008-07-18 09:13:20 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* We take a reference when we initialize udqp and gdqp,
|
|
|
|
* so it is important that we never blindly double trip on
|
|
|
|
* the same variable. See xfs_create() for an example.
|
|
|
|
*/
|
|
|
|
ASSERT(udqp == NULL);
|
|
|
|
ASSERT(gdqp == NULL);
|
2009-06-08 15:33:32 +02:00
|
|
|
code = xfs_qm_vop_dqalloc(ip, uid, gid, ip->i_d.di_projid,
|
2008-07-18 09:13:20 +02:00
|
|
|
qflags, &udqp, &gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (code)
|
2006-01-15 02:37:08 +01:00
|
|
|
return code;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For the other attributes, we acquire the inode lock and
|
|
|
|
* first do an error checking pass.
|
|
|
|
*/
|
|
|
|
tp = NULL;
|
|
|
|
lock_flags = XFS_ILOCK_EXCL;
|
2008-07-18 09:13:28 +02:00
|
|
|
if (flags & XFS_ATTR_NOLOCK)
|
2005-05-05 22:27:19 +02:00
|
|
|
need_iolock = 0;
|
2008-07-18 09:13:28 +02:00
|
|
|
if (!(mask & ATTR_SIZE)) {
|
|
|
|
if ((mask != (ATTR_CTIME|ATTR_ATIME|ATTR_MTIME)) ||
|
2005-04-17 00:20:36 +02:00
|
|
|
(mp->m_flags & XFS_MOUNT_WSYNC)) {
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
|
|
|
|
commit_flags = 0;
|
|
|
|
if ((code = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_ICHANGE_LOG_RES(mp), 0,
|
|
|
|
0, 0))) {
|
|
|
|
lock_flags = 0;
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(ip, DM_EVENT_TRUNCATE) &&
|
2008-07-18 09:13:28 +02:00
|
|
|
!(flags & XFS_ATTR_DMI)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
int dmflags = AT_DELAY_FLAG(flags) | DM_SEM_FLAG_WR;
|
2008-03-06 03:45:58 +01:00
|
|
|
code = XFS_SEND_DATA(mp, DM_EVENT_TRUNCATE, ip,
|
2008-07-18 09:13:28 +02:00
|
|
|
iattr->ia_size, 0, dmflags, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (code) {
|
|
|
|
lock_flags = 0;
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (need_iolock)
|
|
|
|
lock_flags |= XFS_IOLOCK_EXCL;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, lock_flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Change file ownership. Must be the owner or privileged.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & (ATTR_UID|ATTR_GID)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* These IDs could have changed since we last looked at them.
|
|
|
|
* But, we're assured that if the ownership did change
|
|
|
|
* while we didn't have the inode locked, inode's dquot(s)
|
|
|
|
* would have changed also.
|
|
|
|
*/
|
|
|
|
iuid = ip->i_d.di_uid;
|
|
|
|
igid = ip->i_d.di_gid;
|
2008-07-18 09:13:28 +02:00
|
|
|
gid = (mask & ATTR_GID) ? iattr->ia_gid : igid;
|
|
|
|
uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
2008-07-18 09:13:20 +02:00
|
|
|
* Do a quota reservation only if uid/gid is actually
|
2005-04-17 00:20:36 +02:00
|
|
|
* going to change.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
if (XFS_IS_QUOTA_RUNNING(mp) &&
|
|
|
|
((XFS_IS_UQUOTA_ON(mp) && iuid != uid) ||
|
|
|
|
(XFS_IS_GQUOTA_ON(mp) && igid != gid))) {
|
2005-04-17 00:20:36 +02:00
|
|
|
ASSERT(tp);
|
2009-06-08 15:33:32 +02:00
|
|
|
code = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp,
|
2005-04-17 00:20:36 +02:00
|
|
|
capable(CAP_FOWNER) ?
|
|
|
|
XFS_QMOPT_FORCE_RES : 0);
|
|
|
|
if (code) /* out of quota */
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Truncate file. Must have write permission and not be a directory.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & ATTR_SIZE) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/* Short circuit the truncate case for zero length files */
|
2008-07-18 09:13:28 +02:00
|
|
|
if (iattr->ia_size == 0 &&
|
|
|
|
ip->i_size == 0 && ip->i_d.di_nextents == 0) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
lock_flags &= ~XFS_ILOCK_EXCL;
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & ATTR_CTIME)
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
code = 0;
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
2008-04-22 09:33:25 +02:00
|
|
|
if (S_ISDIR(ip->i_d.di_mode)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
code = XFS_ERROR(EISDIR);
|
|
|
|
goto error_return;
|
2008-04-22 09:33:25 +02:00
|
|
|
} else if (!S_ISREG(ip->i_d.di_mode)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
code = XFS_ERROR(EINVAL);
|
|
|
|
goto error_return;
|
|
|
|
}
|
2008-12-09 10:47:34 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Make sure that the dquots are attached to the inode.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
code = xfs_qm_dqattach_locked(ip, 0);
|
2008-12-09 10:47:34 +01:00
|
|
|
if (code)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error_return;
|
|
|
|
|
2008-12-09 10:47:34 +01:00
|
|
|
/*
|
|
|
|
* Now we can make the changes. Before we join the inode
|
|
|
|
* to the transaction, if ATTR_SIZE is set then take care of
|
|
|
|
* the part of the truncation that must be done without the
|
|
|
|
* inode lock. This needs to be done before joining the inode
|
|
|
|
* to the transaction, because the inode cannot be unlocked
|
|
|
|
* once it is a part of the transaction.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (iattr->ia_size > ip->i_size) {
|
2008-05-20 03:30:46 +02:00
|
|
|
/*
|
|
|
|
* Do the first part of growing a file: zero any data
|
|
|
|
* in the last block that is beyond the old EOF. We
|
|
|
|
* need to do this before the inode is joined to the
|
|
|
|
* transaction to modify the i_size.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
code = xfs_zero_eof(ip, iattr->ia_size, ip->i_size);
|
2005-11-02 05:07:34 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2007-07-19 08:28:58 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We are going to log the inode size change in this
|
|
|
|
* transaction so any previous writes that are beyond the on
|
|
|
|
* disk EOF and the new EOF that have not been written out need
|
|
|
|
* to be written here. If we do not write the data out, we
|
|
|
|
* expose ourselves to the null files problem.
|
|
|
|
*
|
|
|
|
* Only flush from the on disk size to the smaller of the in
|
|
|
|
* memory file size or the new size as that's the range we
|
|
|
|
* really care about here and prevents waiting for other data
|
|
|
|
* not within the range we care about here.
|
|
|
|
*/
|
|
|
|
if (!code &&
|
2008-07-18 09:13:28 +02:00
|
|
|
ip->i_size != ip->i_d.di_size &&
|
|
|
|
iattr->ia_size > ip->i_d.di_size) {
|
2007-08-29 02:58:01 +02:00
|
|
|
code = xfs_flush_pages(ip,
|
2008-07-18 09:13:28 +02:00
|
|
|
ip->i_d.di_size, iattr->ia_size,
|
2007-07-19 08:28:58 +02:00
|
|
|
XFS_B_ASYNC, FI_NONE);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* wait for all I/O to complete */
|
2008-12-03 12:20:39 +01:00
|
|
|
xfs_ioend_wait(ip);
|
2007-07-19 08:28:58 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
if (!code)
|
2008-07-18 09:13:28 +02:00
|
|
|
code = xfs_itruncate_data(ip, iattr->ia_size);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (code) {
|
|
|
|
ASSERT(tp == NULL);
|
|
|
|
lock_flags &= ~XFS_ILOCK_EXCL;
|
|
|
|
ASSERT(lock_flags == XFS_IOLOCK_EXCL);
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE);
|
|
|
|
if ((code = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_ITRUNCATE_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_ITRUNCATE_LOG_COUNT))) {
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
if (need_iolock)
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return code;
|
|
|
|
}
|
|
|
|
commit_flags = XFS_TRANS_RELEASE_LOG_RES;
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
xfs_trans_ijoin(tp, ip, lock_flags);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
2008-03-06 03:45:29 +01:00
|
|
|
/*
|
|
|
|
* Only change the c/mtime if we are changing the size
|
|
|
|
* or we are explicitly asked to change it. This handles
|
|
|
|
* the semantic difference between truncate() and ftruncate()
|
|
|
|
* as implemented in the VFS.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (iattr->ia_size != ip->i_size || (mask & ATTR_CTIME))
|
2008-03-06 03:45:29 +01:00
|
|
|
timeflags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if (iattr->ia_size > ip->i_size) {
|
|
|
|
ip->i_d.di_size = iattr->ia_size;
|
|
|
|
ip->i_size = iattr->ia_size;
|
|
|
|
if (!(flags & XFS_ATTR_DMI))
|
2008-05-20 03:30:46 +02:00
|
|
|
xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
2008-07-18 09:13:28 +02:00
|
|
|
} else if (iattr->ia_size <= ip->i_size ||
|
|
|
|
(iattr->ia_size == 0 && ip->i_d.di_nextents)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* signal a sync transaction unless
|
|
|
|
* we're truncating an already unlinked
|
|
|
|
* file on a wsync filesystem
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
code = xfs_itruncate_finish(&tp, ip, iattr->ia_size,
|
2005-04-17 00:20:36 +02:00
|
|
|
XFS_DATA_FORK,
|
|
|
|
((ip->i_d.di_nlink != 0 ||
|
|
|
|
!(mp->m_flags & XFS_MOUNT_WSYNC))
|
|
|
|
? 1 : 0));
|
2006-06-09 07:27:16 +02:00
|
|
|
if (code)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto abort_return;
|
2006-06-09 07:27:16 +02:00
|
|
|
/*
|
|
|
|
* Truncated "down", so we're removing references
|
|
|
|
* to old data here - if we now delay flushing for
|
|
|
|
* a long time, we expose ourselves unduly to the
|
|
|
|
* notorious NULL files problem. So, we mark this
|
|
|
|
* vnode and flush it when the file is closed, and
|
|
|
|
* do not wait the usual (long) time for writeout.
|
|
|
|
*/
|
2007-08-29 03:44:37 +02:00
|
|
|
xfs_iflags_set(ip, XFS_ITRUNCATED);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2008-12-09 10:47:34 +01:00
|
|
|
} else if (tp) {
|
|
|
|
xfs_trans_ijoin(tp, ip, lock_flags);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Change file ownership. Must be the owner or privileged.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & (ATTR_UID|ATTR_GID)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* CAP_FSETID overrides the following restrictions:
|
|
|
|
*
|
|
|
|
* The set-user-ID and set-group-ID bits of a file will be
|
|
|
|
* cleared upon successful return from chown()
|
|
|
|
*/
|
|
|
|
if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
|
|
|
|
!capable(CAP_FSETID)) {
|
|
|
|
ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Change the ownerships and register quota modifications
|
|
|
|
* in the transaction.
|
|
|
|
*/
|
|
|
|
if (iuid != uid) {
|
2009-06-08 15:33:32 +02:00
|
|
|
if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_UQUOTA_ON(mp)) {
|
2008-07-18 09:13:28 +02:00
|
|
|
ASSERT(mask & ATTR_UID);
|
2005-04-17 00:20:36 +02:00
|
|
|
ASSERT(udqp);
|
2009-06-08 15:33:32 +02:00
|
|
|
olddquot1 = xfs_qm_vop_chown(tp, ip,
|
2005-04-17 00:20:36 +02:00
|
|
|
&ip->i_udquot, udqp);
|
|
|
|
}
|
|
|
|
ip->i_d.di_uid = uid;
|
2008-07-21 08:16:15 +02:00
|
|
|
inode->i_uid = uid;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
if (igid != gid) {
|
2009-06-08 15:33:32 +02:00
|
|
|
if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_GQUOTA_ON(mp)) {
|
2005-06-21 07:38:48 +02:00
|
|
|
ASSERT(!XFS_IS_PQUOTA_ON(mp));
|
2008-07-18 09:13:28 +02:00
|
|
|
ASSERT(mask & ATTR_GID);
|
2005-04-17 00:20:36 +02:00
|
|
|
ASSERT(gdqp);
|
2009-06-08 15:33:32 +02:00
|
|
|
olddquot2 = xfs_qm_vop_chown(tp, ip,
|
2005-04-17 00:20:36 +02:00
|
|
|
&ip->i_gdquot, gdqp);
|
|
|
|
}
|
|
|
|
ip->i_d.di_gid = gid;
|
2008-07-21 08:16:15 +02:00
|
|
|
inode->i_gid = gid;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
|
|
|
|
timeflags |= XFS_ICHGTIME_CHG;
|
|
|
|
}
|
|
|
|
|
2008-12-09 10:47:34 +01:00
|
|
|
/*
|
|
|
|
* Change file access modes.
|
|
|
|
*/
|
|
|
|
if (mask & ATTR_MODE) {
|
|
|
|
umode_t mode = iattr->ia_mode;
|
|
|
|
|
|
|
|
if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
|
|
|
|
mode &= ~S_ISGID;
|
|
|
|
|
|
|
|
ip->i_d.di_mode &= S_IFMT;
|
|
|
|
ip->i_d.di_mode |= mode & ~S_IFMT;
|
|
|
|
|
|
|
|
inode->i_mode &= S_IFMT;
|
|
|
|
inode->i_mode |= mode & ~S_IFMT;
|
|
|
|
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
timeflags |= XFS_ICHGTIME_CHG;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Change file access or modified times.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & (ATTR_ATIME|ATTR_MTIME)) {
|
|
|
|
if (mask & ATTR_ATIME) {
|
2008-07-21 08:16:15 +02:00
|
|
|
inode->i_atime = iattr->ia_atime;
|
2008-07-18 09:13:28 +02:00
|
|
|
ip->i_d.di_atime.t_sec = iattr->ia_atime.tv_sec;
|
|
|
|
ip->i_d.di_atime.t_nsec = iattr->ia_atime.tv_nsec;
|
2005-04-17 00:20:36 +02:00
|
|
|
ip->i_update_core = 1;
|
|
|
|
}
|
2008-07-18 09:13:28 +02:00
|
|
|
if (mask & ATTR_MTIME) {
|
2008-07-21 08:16:15 +02:00
|
|
|
inode->i_mtime = iattr->ia_mtime;
|
2008-07-18 09:13:28 +02:00
|
|
|
ip->i_d.di_mtime.t_sec = iattr->ia_mtime.tv_sec;
|
|
|
|
ip->i_d.di_mtime.t_nsec = iattr->ia_mtime.tv_nsec;
|
2005-04-17 00:20:36 +02:00
|
|
|
timeflags &= ~XFS_ICHGTIME_MOD;
|
|
|
|
timeflags |= XFS_ICHGTIME_CHG;
|
|
|
|
}
|
2008-07-18 09:13:28 +02:00
|
|
|
if (tp && (mask & (ATTR_MTIME_SET|ATTR_ATIME_SET)))
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2008-07-18 09:13:28 +02:00
|
|
|
* Change file inode change time only if ATTR_CTIME set
|
2005-04-17 00:20:36 +02:00
|
|
|
* AND we have been called by a DMI function.
|
|
|
|
*/
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if ((flags & XFS_ATTR_DMI) && (mask & ATTR_CTIME)) {
|
2008-07-21 08:16:15 +02:00
|
|
|
inode->i_ctime = iattr->ia_ctime;
|
2008-07-18 09:13:28 +02:00
|
|
|
ip->i_d.di_ctime.t_sec = iattr->ia_ctime.tv_sec;
|
|
|
|
ip->i_d.di_ctime.t_nsec = iattr->ia_ctime.tv_nsec;
|
2005-04-17 00:20:36 +02:00
|
|
|
ip->i_update_core = 1;
|
|
|
|
timeflags &= ~XFS_ICHGTIME_CHG;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Send out timestamp changes that need to be set to the
|
|
|
|
* current time. Not done when called by a DMI function.
|
|
|
|
*/
|
2008-07-18 09:13:28 +02:00
|
|
|
if (timeflags && !(flags & XFS_ATTR_DMI))
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ichgtime(ip, timeflags);
|
|
|
|
|
|
|
|
XFS_STATS_INC(xs_ig_attrchg);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a synchronous mount, make sure that the
|
|
|
|
* transaction goes to disk before returning to the user.
|
|
|
|
* This is slightly sub-optimal in that truncates require
|
2006-03-29 00:55:14 +02:00
|
|
|
* two sync transactions instead of one for wsync filesystems.
|
2005-04-17 00:20:36 +02:00
|
|
|
* One for the truncate and one for the timestamps since we
|
|
|
|
* don't want to change the timestamps unless we're sure the
|
|
|
|
* truncate worked. Truncates are less than 1% of the laddis
|
|
|
|
* mix so this probably isn't worth the trouble to optimize.
|
|
|
|
*/
|
|
|
|
code = 0;
|
|
|
|
if (tp) {
|
|
|
|
if (mp->m_flags & XFS_MOUNT_WSYNC)
|
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
code = xfs_trans_commit(tp, commit_flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_iunlock(ip, lock_flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Release any dquot(s) the inode had kept before chown.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(olddquot1);
|
|
|
|
xfs_qm_dqrele(olddquot2);
|
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (code) {
|
|
|
|
return code;
|
|
|
|
}
|
|
|
|
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(ip, DM_EVENT_ATTRIBUTE) &&
|
2008-07-18 09:13:28 +02:00
|
|
|
!(flags & XFS_ATTR_DMI)) {
|
2008-03-06 03:45:58 +01:00
|
|
|
(void) XFS_SEND_NAMESP(mp, DM_EVENT_ATTRIBUTE, ip, DM_RIGHT_NULL,
|
2005-04-17 00:20:36 +02:00
|
|
|
NULL, DM_RIGHT_NULL, NULL, NULL,
|
|
|
|
0, 0, AT_DELAY_FLAG(flags));
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
abort_return:
|
|
|
|
commit_flags |= XFS_TRANS_ABORT;
|
|
|
|
/* FALLTHROUGH */
|
|
|
|
error_return:
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (tp) {
|
|
|
|
xfs_trans_cancel(tp, commit_flags);
|
|
|
|
}
|
|
|
|
if (lock_flags != 0) {
|
|
|
|
xfs_iunlock(ip, lock_flags);
|
|
|
|
}
|
|
|
|
return code;
|
|
|
|
}
|
|
|
|
|
2006-06-09 07:27:16 +02:00
|
|
|
/*
|
|
|
|
* The maximum pathlen is 1024 bytes. Since the minimum file system
|
|
|
|
* blocksize is 512 bytes, we can get a max of 2 extents back from
|
|
|
|
* bmapi.
|
|
|
|
*/
|
|
|
|
#define SYMLINK_MAPS 2
|
|
|
|
|
2007-08-28 05:59:03 +02:00
|
|
|
STATIC int
|
|
|
|
xfs_readlink_bmap(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
char *link)
|
|
|
|
{
|
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
|
|
|
int pathlen = ip->i_d.di_size;
|
|
|
|
int nmaps = SYMLINK_MAPS;
|
|
|
|
xfs_bmbt_irec_t mval[SYMLINK_MAPS];
|
|
|
|
xfs_daddr_t d;
|
|
|
|
int byte_cnt;
|
|
|
|
int n;
|
|
|
|
xfs_buf_t *bp;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
error = xfs_bmapi(NULL, ip, 0, XFS_B_TO_FSB(mp, pathlen), 0, NULL, 0,
|
|
|
|
mval, &nmaps, NULL, NULL);
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
for (n = 0; n < nmaps; n++) {
|
|
|
|
d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
|
|
|
|
byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
|
|
|
|
|
|
|
|
bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
|
|
|
|
error = XFS_BUF_GETERROR(bp);
|
|
|
|
if (error) {
|
|
|
|
xfs_ioerror_alert("xfs_readlink",
|
|
|
|
ip->i_mount, bp, XFS_BUF_ADDR(bp));
|
|
|
|
xfs_buf_relse(bp);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (pathlen < byte_cnt)
|
|
|
|
byte_cnt = pathlen;
|
|
|
|
pathlen -= byte_cnt;
|
|
|
|
|
|
|
|
memcpy(link, XFS_BUF_PTR(bp), byte_cnt);
|
|
|
|
xfs_buf_relse(bp);
|
|
|
|
}
|
|
|
|
|
|
|
|
link[ip->i_d.di_size] = '\0';
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
out:
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_readlink(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *ip,
|
2007-08-28 05:59:03 +02:00
|
|
|
char *link)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 05:59:03 +02:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
int pathlen;
|
|
|
|
int error = 0;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
|
|
|
|
|
|
|
ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFLNK);
|
2007-08-28 05:59:03 +02:00
|
|
|
ASSERT(ip->i_d.di_size <= MAXPATHLEN);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-28 05:59:03 +02:00
|
|
|
pathlen = ip->i_d.di_size;
|
|
|
|
if (!pathlen)
|
|
|
|
goto out;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (ip->i_df.if_flags & XFS_IFINLINE) {
|
2007-08-28 05:59:03 +02:00
|
|
|
memcpy(link, ip->i_df.if_u1.if_data, pathlen);
|
|
|
|
link[pathlen] = '\0';
|
|
|
|
} else {
|
|
|
|
error = xfs_readlink_bmap(ip, link);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2007-08-28 05:59:03 +02:00
|
|
|
out:
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_fsync
|
|
|
|
*
|
2008-05-19 08:29:46 +02:00
|
|
|
* This is called to sync the inode and its data out to disk. We need to hold
|
|
|
|
* the I/O lock while flushing the data, and the inode lock while flushing the
|
|
|
|
* inode. The inode lock CANNOT be held while flushing the data, so acquire
|
|
|
|
* after we're done with that.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_fsync(
|
2008-05-19 08:29:46 +02:00
|
|
|
xfs_inode_t *ip)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
xfs_trans_t *tp;
|
|
|
|
int error;
|
2005-11-02 00:26:59 +01:00
|
|
|
int log_flushed = 0, changed = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2008-05-19 08:29:46 +02:00
|
|
|
/* capture size updates in I/O completion before writing the inode. */
|
2008-11-28 04:23:33 +01:00
|
|
|
error = xfs_wait_on_pages(ip, 0, -1);
|
2008-05-19 08:29:46 +02:00
|
|
|
if (error)
|
|
|
|
return XFS_ERROR(error);
|
2007-09-14 07:22:50 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2008-05-19 08:29:46 +02:00
|
|
|
* We always need to make sure that the required inode state is safe on
|
|
|
|
* disk. The vnode might be clean but we still might need to force the
|
|
|
|
* log because of committed transactions that haven't hit the disk yet.
|
|
|
|
* Likewise, there could be unflushed non-transactional changes to the
|
|
|
|
* inode core that have to go to disk and this requires us to issue
|
|
|
|
* a synchronous transaction to capture these changes correctly.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
2008-05-19 08:29:46 +02:00
|
|
|
* This code relies on the assumption that if the update_* fields
|
|
|
|
* of the inode are clear and the inode is unpinned then it is clean
|
|
|
|
* and no action is required.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
|
|
|
|
2008-05-19 08:29:46 +02:00
|
|
|
if (!(ip->i_update_size || ip->i_update_core)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2008-05-19 08:29:46 +02:00
|
|
|
* Timestamps/size haven't changed since last inode flush or
|
|
|
|
* inode transaction commit. That means either nothing got
|
|
|
|
* written or a transaction committed which caught the updates.
|
|
|
|
* If the latter happened and the transaction hasn't hit the
|
|
|
|
* disk yet, the inode will be still be pinned. If it is,
|
|
|
|
* force the log.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
|
|
|
|
if (xfs_ipincount(ip)) {
|
2008-05-19 08:29:46 +02:00
|
|
|
error = _xfs_log_force(ip->i_mount, (xfs_lsn_t)0,
|
|
|
|
XFS_LOG_FORCE | XFS_LOG_SYNC,
|
2005-11-02 00:26:59 +01:00
|
|
|
&log_flushed);
|
|
|
|
} else {
|
|
|
|
/*
|
2008-05-19 08:29:46 +02:00
|
|
|
* If the inode is not pinned and nothing has changed
|
|
|
|
* we don't need to flush the cache.
|
2005-11-02 00:26:59 +01:00
|
|
|
*/
|
|
|
|
changed = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/*
|
2008-05-19 08:29:46 +02:00
|
|
|
* Kick off a transaction to log the inode core to get the
|
|
|
|
* updates. The sync transaction will also force the log.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS);
|
2008-05-19 08:29:46 +02:00
|
|
|
error = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_FSYNC_TS_LOG_RES(ip->i_mount), 0, 0, 0);
|
|
|
|
if (error) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
/*
|
2008-05-19 08:29:46 +02:00
|
|
|
* Note - it's possible that we might have pushed ourselves out
|
|
|
|
* of the way during trans_reserve which would flush the inode.
|
|
|
|
* But there's no guarantee that the inode buffer has actually
|
|
|
|
* gone out yet (it's delwri). Plus the buffer could be pinned
|
|
|
|
* anyway if it's part of an inode in another recent
|
|
|
|
* transaction. So we play it safe and fire off the
|
|
|
|
* transaction anyway.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
2008-05-19 08:29:46 +02:00
|
|
|
xfs_trans_set_sync(tp);
|
2007-05-08 05:48:42 +02:00
|
|
|
error = _xfs_trans_commit(tp, 0, &log_flushed);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
}
|
2005-11-02 00:26:59 +01:00
|
|
|
|
|
|
|
if ((ip->i_mount->m_flags & XFS_MOUNT_BARRIER) && changed) {
|
|
|
|
/*
|
|
|
|
* If the log write didn't issue an ordered tag we need
|
|
|
|
* to flush the disk cache for the data device now.
|
|
|
|
*/
|
|
|
|
if (!log_flushed)
|
|
|
|
xfs_blkdev_issue_flush(ip->i_mount->m_ddev_targp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this inode is on the RT dev we need to flush that
|
2006-03-29 00:55:14 +02:00
|
|
|
* cache as well.
|
2005-11-02 00:26:59 +01:00
|
|
|
*/
|
2007-11-23 06:29:42 +01:00
|
|
|
if (XFS_IS_REALTIME_INODE(ip))
|
2005-11-02 00:26:59 +01:00
|
|
|
xfs_blkdev_issue_flush(ip->i_mount->m_rtdev_targp);
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2007-05-24 07:22:19 +02:00
|
|
|
* This is called by xfs_inactive to free any blocks beyond eof
|
|
|
|
* when the link count isn't zero and by xfs_dm_punch_hole() when
|
|
|
|
* punching a hole to EOF.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2007-05-24 07:22:19 +02:00
|
|
|
int
|
|
|
|
xfs_free_eofblocks(
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_mount_t *mp,
|
2007-05-24 07:22:19 +02:00
|
|
|
xfs_inode_t *ip,
|
|
|
|
int flags)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
xfs_trans_t *tp;
|
|
|
|
int error;
|
|
|
|
xfs_fileoff_t end_fsb;
|
|
|
|
xfs_fileoff_t last_fsb;
|
|
|
|
xfs_filblks_t map_len;
|
|
|
|
int nimaps;
|
|
|
|
xfs_bmbt_irec_t imap;
|
2007-05-24 07:22:19 +02:00
|
|
|
int use_iolock = (flags & XFS_FREE_EOF_LOCK);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Figure out if there are any blocks beyond the end
|
|
|
|
* of the file. If not, then there is nothing to do.
|
|
|
|
*/
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
end_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)ip->i_size));
|
2005-04-17 00:20:36 +02:00
|
|
|
last_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
|
|
|
|
map_len = last_fsb - end_fsb;
|
|
|
|
if (map_len <= 0)
|
2006-01-15 02:37:08 +01:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
nimaps = 1;
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bmapi(NULL, ip, end_fsb, map_len, 0,
|
2006-06-09 06:48:12 +02:00
|
|
|
NULL, 0, &imap, &nimaps, NULL, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
|
|
|
|
if (!error && (nimaps != 0) &&
|
2006-01-11 05:38:31 +01:00
|
|
|
(imap.br_startblock != HOLESTARTBLOCK ||
|
|
|
|
ip->i_delayed_blks)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Attach the dquots to the inode up front.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(ip, 0);
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* There are blocks after the end of file.
|
|
|
|
* Free them up now by truncating the file to
|
|
|
|
* its current size.
|
|
|
|
*/
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Do the xfs_itruncate_start() call before
|
|
|
|
* reserving any log space because
|
|
|
|
* itruncate_start will call into the buffer
|
|
|
|
* cache and we can't
|
|
|
|
* do that within a transaction.
|
|
|
|
*/
|
2007-05-24 07:22:19 +02:00
|
|
|
if (use_iolock)
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL);
|
2007-05-08 05:49:27 +02:00
|
|
|
error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE,
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
ip->i_size);
|
2007-05-08 05:49:27 +02:00
|
|
|
if (error) {
|
2007-06-28 08:43:14 +02:00
|
|
|
xfs_trans_cancel(tp, 0);
|
2007-05-24 07:22:19 +02:00
|
|
|
if (use_iolock)
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
2007-05-08 05:49:27 +02:00
|
|
|
return error;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
error = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_ITRUNCATE_LOG_RES(mp),
|
|
|
|
0, XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_ITRUNCATE_LOG_COUNT);
|
|
|
|
if (error) {
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, ip,
|
|
|
|
XFS_IOLOCK_EXCL |
|
|
|
|
XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
|
|
|
error = xfs_itruncate_finish(&tp, ip,
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
ip->i_size,
|
2005-04-17 00:20:36 +02:00
|
|
|
XFS_DATA_FORK,
|
|
|
|
0);
|
|
|
|
/*
|
|
|
|
* If we get an error at this point we
|
|
|
|
* simply don't bother truncating the file.
|
|
|
|
*/
|
|
|
|
if (error) {
|
|
|
|
xfs_trans_cancel(tp,
|
|
|
|
(XFS_TRANS_RELEASE_LOG_RES |
|
|
|
|
XFS_TRANS_ABORT));
|
|
|
|
} else {
|
|
|
|
error = xfs_trans_commit(tp,
|
2007-05-08 05:48:42 +02:00
|
|
|
XFS_TRANS_RELEASE_LOG_RES);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2007-05-24 07:22:19 +02:00
|
|
|
xfs_iunlock(ip, (use_iolock ? (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)
|
|
|
|
: XFS_ILOCK_EXCL));
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free a symlink that has blocks associated with it.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_inactive_symlink_rmt(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_trans_t **tpp)
|
|
|
|
{
|
|
|
|
xfs_buf_t *bp;
|
|
|
|
int committed;
|
|
|
|
int done;
|
|
|
|
int error;
|
|
|
|
xfs_fsblock_t first_block;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
int i;
|
|
|
|
xfs_mount_t *mp;
|
|
|
|
xfs_bmbt_irec_t mval[SYMLINK_MAPS];
|
|
|
|
int nmaps;
|
|
|
|
xfs_trans_t *ntp;
|
|
|
|
int size;
|
|
|
|
xfs_trans_t *tp;
|
|
|
|
|
|
|
|
tp = *tpp;
|
|
|
|
mp = ip->i_mount;
|
|
|
|
ASSERT(ip->i_d.di_size > XFS_IFORK_DSIZE(ip));
|
|
|
|
/*
|
|
|
|
* We're freeing a symlink that has some
|
|
|
|
* blocks allocated to it. Free the
|
|
|
|
* blocks here. We know that we've got
|
|
|
|
* either 1 or 2 extents and that we can
|
|
|
|
* free them all in one bunmapi call.
|
|
|
|
*/
|
|
|
|
ASSERT(ip->i_d.di_nextents > 0 && ip->i_d.di_nextents <= 2);
|
|
|
|
if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
*tpp = NULL;
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Lock the inode, fix the size, and join it to the transaction.
|
|
|
|
* Hold it so in the normal path, we still have it locked for
|
|
|
|
* the second transaction. In the error paths we need it
|
|
|
|
* held so the cancel won't rele it, see below.
|
|
|
|
*/
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
size = (int)ip->i_d.di_size;
|
|
|
|
ip->i_d.di_size = 0;
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
/*
|
|
|
|
* Find the block(s) so we can inval and unmap them.
|
|
|
|
*/
|
|
|
|
done = 0;
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2006-03-24 12:15:34 +01:00
|
|
|
nmaps = ARRAY_SIZE(mval);
|
2005-04-17 00:20:36 +02:00
|
|
|
if ((error = xfs_bmapi(tp, ip, 0, XFS_B_TO_FSB(mp, size),
|
|
|
|
XFS_BMAPI_METADATA, &first_block, 0, mval, &nmaps,
|
2006-06-09 06:48:12 +02:00
|
|
|
&free_list, NULL)))
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error0;
|
|
|
|
/*
|
|
|
|
* Invalidate the block(s).
|
|
|
|
*/
|
|
|
|
for (i = 0; i < nmaps; i++) {
|
|
|
|
bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
|
|
|
|
XFS_FSB_TO_DADDR(mp, mval[i].br_startblock),
|
|
|
|
XFS_FSB_TO_BB(mp, mval[i].br_blockcount), 0);
|
|
|
|
xfs_trans_binval(tp, bp);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Unmap the dead block(s) to the free_list.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_bunmapi(tp, ip, 0, size, XFS_BMAPI_METADATA, nmaps,
|
2006-06-09 06:48:12 +02:00
|
|
|
&first_block, &free_list, NULL, &done)))
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error1;
|
|
|
|
ASSERT(done);
|
|
|
|
/*
|
|
|
|
* Commit the first transaction. This logs the EFI and the inode.
|
|
|
|
*/
|
2007-02-10 08:37:16 +01:00
|
|
|
if ((error = xfs_bmap_finish(&tp, &free_list, &committed)))
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error1;
|
|
|
|
/*
|
|
|
|
* The transaction must have been committed, since there were
|
|
|
|
* actually extents freed by xfs_bunmapi. See xfs_bmap_finish.
|
|
|
|
* The new tp has the extent freeing and EFDs.
|
|
|
|
*/
|
|
|
|
ASSERT(committed);
|
|
|
|
/*
|
|
|
|
* The first xact was committed, so add the inode to the new one.
|
|
|
|
* Mark it dirty so it will be logged and moved forward in the log as
|
|
|
|
* part of every commit.
|
|
|
|
*/
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
/*
|
|
|
|
* Get a new, empty transaction to return to our caller.
|
|
|
|
*/
|
|
|
|
ntp = xfs_trans_dup(tp);
|
|
|
|
/*
|
2006-03-29 00:55:14 +02:00
|
|
|
* Commit the transaction containing extent freeing and EFDs.
|
2005-04-17 00:20:36 +02:00
|
|
|
* If we get an error on the commit here or on the reserve below,
|
|
|
|
* we need to unlock the inode since the new transaction doesn't
|
|
|
|
* have the inode attached.
|
|
|
|
*/
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
tp = ntp;
|
|
|
|
if (error) {
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
goto error0;
|
|
|
|
}
|
2008-11-17 07:37:10 +01:00
|
|
|
/*
|
|
|
|
* transaction commit worked ok so we can drop the extra ticket
|
|
|
|
* reference that we gained in xfs_trans_dup()
|
|
|
|
*/
|
|
|
|
xfs_log_ticket_put(tp->t_ticket);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Remove the memory for extent descriptions (just bookkeeping).
|
|
|
|
*/
|
|
|
|
if (ip->i_df.if_bytes)
|
|
|
|
xfs_idata_realloc(ip, -ip->i_df.if_bytes, XFS_DATA_FORK);
|
|
|
|
ASSERT(ip->i_df.if_bytes == 0);
|
|
|
|
/*
|
|
|
|
* Put an itruncate log reservation in the new transaction
|
|
|
|
* for our caller.
|
|
|
|
*/
|
|
|
|
if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Return with the inode locked but not joined to the transaction.
|
|
|
|
*/
|
|
|
|
*tpp = tp;
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
error1:
|
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
error0:
|
|
|
|
/*
|
|
|
|
* Have to come here with the inode locked and either
|
|
|
|
* (held and in the transaction) or (not in the transaction).
|
|
|
|
* If the inode isn't held then cancel would iput it, but
|
|
|
|
* that's wrong since this is inactive and the vnode ref
|
|
|
|
* count is 0 already.
|
|
|
|
* Cancel won't do anything to the inode if held, but it still
|
|
|
|
* needs to be locked until the cancel is done, if it was
|
|
|
|
* joined to the transaction.
|
|
|
|
*/
|
|
|
|
xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
*tpp = NULL;
|
|
|
|
return error;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
|
|
|
xfs_inactive_symlink_local(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_trans_t **tpp)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
ASSERT(ip->i_d.di_size <= XFS_IFORK_DSIZE(ip));
|
|
|
|
/*
|
|
|
|
* We're freeing a symlink which fit into
|
|
|
|
* the inode. Just free the memory used
|
|
|
|
* to hold the old symlink.
|
|
|
|
*/
|
|
|
|
error = xfs_trans_reserve(*tpp, 0,
|
|
|
|
XFS_ITRUNCATE_LOG_RES(ip->i_mount),
|
|
|
|
0, XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_ITRUNCATE_LOG_COUNT);
|
|
|
|
|
|
|
|
if (error) {
|
|
|
|
xfs_trans_cancel(*tpp, 0);
|
|
|
|
*tpp = NULL;
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Zero length symlinks _can_ exist.
|
|
|
|
*/
|
|
|
|
if (ip->i_df.if_bytes > 0) {
|
|
|
|
xfs_idata_realloc(ip,
|
|
|
|
-(ip->i_df.if_bytes),
|
|
|
|
XFS_DATA_FORK);
|
|
|
|
ASSERT(ip->i_df.if_bytes == 0);
|
|
|
|
}
|
2006-01-15 02:37:08 +01:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
|
|
|
xfs_inactive_attrs(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_trans_t **tpp)
|
|
|
|
{
|
|
|
|
xfs_trans_t *tp;
|
|
|
|
int error;
|
|
|
|
xfs_mount_t *mp;
|
|
|
|
|
2008-04-22 09:34:00 +02:00
|
|
|
ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
|
2005-04-17 00:20:36 +02:00
|
|
|
tp = *tpp;
|
|
|
|
mp = ip->i_mount;
|
|
|
|
ASSERT(ip->i_d.di_forkoff != 0);
|
2008-04-10 04:21:18 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2008-04-10 04:21:18 +02:00
|
|
|
if (error)
|
|
|
|
goto error_unlock;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
error = xfs_attr_inactive(ip);
|
2008-04-10 04:21:18 +02:00
|
|
|
if (error)
|
|
|
|
goto error_unlock;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
|
|
|
|
error = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_IFREE_LOG_RES(mp),
|
|
|
|
0, XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_INACTIVE_LOG_COUNT);
|
2008-04-10 04:21:18 +02:00
|
|
|
if (error)
|
|
|
|
goto error_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
xfs_idestroy_fork(ip, XFS_ATTR_FORK);
|
|
|
|
|
|
|
|
ASSERT(ip->i_d.di_anextents == 0);
|
|
|
|
|
|
|
|
*tpp = tp;
|
2006-01-15 02:37:08 +01:00
|
|
|
return 0;
|
2008-04-10 04:21:18 +02:00
|
|
|
|
|
|
|
error_cancel:
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
error_unlock:
|
|
|
|
*tpp = NULL;
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_release(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *ip)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
int error;
|
|
|
|
|
2008-04-22 09:33:25 +02:00
|
|
|
if (!S_ISREG(ip->i_d.di_mode) || (ip->i_d.di_mode == 0))
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* If this is a read-only mount, don't do this (would generate I/O) */
|
2007-08-30 09:21:12 +02:00
|
|
|
if (mp->m_flags & XFS_MOUNT_RDONLY)
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
|
2007-07-11 03:09:12 +02:00
|
|
|
if (!XFS_FORCED_SHUTDOWN(mp)) {
|
2007-08-29 03:44:37 +02:00
|
|
|
int truncated;
|
|
|
|
|
2007-07-11 03:09:12 +02:00
|
|
|
/*
|
|
|
|
* If we are using filestreams, and we have an unlinked
|
|
|
|
* file that we are processing the last close on, then nothing
|
|
|
|
* will be able to reopen and write to this file. Purge this
|
|
|
|
* inode from the filestreams cache so that it doesn't delay
|
|
|
|
* teardown of the inode.
|
|
|
|
*/
|
|
|
|
if ((ip->i_d.di_nlink == 0) && xfs_inode_is_filestream(ip))
|
|
|
|
xfs_filestream_deassociate(ip);
|
|
|
|
|
2007-06-28 08:46:47 +02:00
|
|
|
/*
|
|
|
|
* If we previously truncated this file and removed old data
|
|
|
|
* in the process, we want to initiate "early" writeout on
|
|
|
|
* the last close. This is an attempt to combat the notorious
|
|
|
|
* NULL files problem which is particularly noticable from a
|
|
|
|
* truncate down, buffered (re-)write (delalloc), followed by
|
|
|
|
* a crash. What we are effectively doing here is
|
|
|
|
* significantly reducing the time window where we'd otherwise
|
|
|
|
* be exposed to that problem.
|
|
|
|
*/
|
2007-08-29 03:44:50 +02:00
|
|
|
truncated = xfs_iflags_test_and_clear(ip, XFS_ITRUNCATED);
|
2008-08-13 08:22:09 +02:00
|
|
|
if (truncated && VN_DIRTY(VFS_I(ip)) && ip->i_delayed_blks > 0)
|
2007-08-29 02:58:01 +02:00
|
|
|
xfs_flush_pages(ip, 0, -1, XFS_B_ASYNC, FI_NONE);
|
2007-06-28 08:46:47 +02:00
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
if (ip->i_d.di_nlink != 0) {
|
|
|
|
if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
|
2008-08-13 08:22:09 +02:00
|
|
|
((ip->i_size > 0) || (VN_CACHED(VFS_I(ip)) > 0 ||
|
2006-01-11 05:38:31 +01:00
|
|
|
ip->i_delayed_blks > 0)) &&
|
2005-04-17 00:20:36 +02:00
|
|
|
(ip->i_df.if_flags & XFS_IFEXTENTS)) &&
|
2006-01-11 05:28:28 +01:00
|
|
|
(!(ip->i_d.di_flags &
|
|
|
|
(XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)))) {
|
2007-05-24 07:22:19 +02:00
|
|
|
error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_inactive
|
|
|
|
*
|
|
|
|
* This is called when the vnode reference count for the vnode
|
|
|
|
* goes to zero. If the file has been unlinked, then it must
|
|
|
|
* now be truncated. Also, we clear all of the read-ahead state
|
|
|
|
* kept for the inode here since the file is now closed.
|
|
|
|
*/
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_inactive(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *ip)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-06-09 09:00:52 +02:00
|
|
|
xfs_bmap_free_t free_list;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_fsblock_t first_block;
|
|
|
|
int committed;
|
|
|
|
xfs_trans_t *tp;
|
|
|
|
xfs_mount_t *mp;
|
|
|
|
int error;
|
|
|
|
int truncate;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the inode is already free, then there can be nothing
|
|
|
|
* to clean up here.
|
|
|
|
*/
|
2009-03-16 08:25:25 +01:00
|
|
|
if (ip->i_d.di_mode == 0 || is_bad_inode(VFS_I(ip))) {
|
2005-04-17 00:20:36 +02:00
|
|
|
ASSERT(ip->i_df.if_real_bytes == 0);
|
|
|
|
ASSERT(ip->i_df.if_broot_bytes == 0);
|
|
|
|
return VN_INACTIVE_CACHE;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Only do a truncate if it's a regular file with
|
|
|
|
* some actual space in it. It's OK to look at the
|
|
|
|
* inode's fields without the lock because we're the
|
|
|
|
* only one with a reference to the inode.
|
|
|
|
*/
|
|
|
|
truncate = ((ip->i_d.di_nlink == 0) &&
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
((ip->i_d.di_size != 0) || (ip->i_size != 0) ||
|
|
|
|
(ip->i_d.di_nextents > 0) || (ip->i_delayed_blks > 0)) &&
|
2005-04-17 00:20:36 +02:00
|
|
|
((ip->i_d.di_mode & S_IFMT) == S_IFREG));
|
|
|
|
|
|
|
|
mp = ip->i_mount;
|
|
|
|
|
2008-03-06 03:45:58 +01:00
|
|
|
if (ip->i_d.di_nlink == 0 && DM_EVENT_ENABLED(ip, DM_EVENT_DESTROY))
|
|
|
|
XFS_SEND_DESTROY(mp, ip, DM_RIGHT_NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
/* If this is a read-only mount, don't do this (would generate I/O) */
|
2007-08-30 09:21:12 +02:00
|
|
|
if (mp->m_flags & XFS_MOUNT_RDONLY)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (ip->i_d.di_nlink != 0) {
|
|
|
|
if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
|
2008-08-13 08:22:09 +02:00
|
|
|
((ip->i_size > 0) || (VN_CACHED(VFS_I(ip)) > 0 ||
|
2006-01-11 05:38:31 +01:00
|
|
|
ip->i_delayed_blks > 0)) &&
|
2006-01-11 05:28:28 +01:00
|
|
|
(ip->i_df.if_flags & XFS_IFEXTENTS) &&
|
|
|
|
(!(ip->i_d.di_flags &
|
|
|
|
(XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) ||
|
|
|
|
(ip->i_delayed_blks != 0)))) {
|
2007-05-24 07:22:19 +02:00
|
|
|
error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT(ip->i_d.di_nlink == 0);
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(ip, 0);
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
|
|
|
|
if (truncate) {
|
|
|
|
/*
|
|
|
|
* Do the xfs_itruncate_start() call before
|
|
|
|
* reserving any log space because itruncate_start
|
|
|
|
* will call into the buffer cache and we can't
|
|
|
|
* do that within a transaction.
|
|
|
|
*/
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
|
2007-05-08 05:49:27 +02:00
|
|
|
error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE, 0);
|
|
|
|
if (error) {
|
2007-06-28 08:43:14 +02:00
|
|
|
xfs_trans_cancel(tp, 0);
|
2007-05-08 05:49:27 +02:00
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return VN_INACTIVE_CACHE;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
error = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_ITRUNCATE_LOG_RES(mp),
|
|
|
|
0, XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_ITRUNCATE_LOG_COUNT);
|
|
|
|
if (error) {
|
|
|
|
/* Don't call itruncate_cleanup */
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* normally, we have to run xfs_itruncate_finish sync.
|
|
|
|
* But if filesystem is wsync and we're in the inactive
|
|
|
|
* path, then we know that nlink == 0, and that the
|
|
|
|
* xaction that made nlink == 0 is permanently committed
|
|
|
|
* since xfs_remove runs as a synchronous transaction.
|
|
|
|
*/
|
|
|
|
error = xfs_itruncate_finish(&tp, ip, 0, XFS_DATA_FORK,
|
|
|
|
(!(mp->m_flags & XFS_MOUNT_WSYNC) ? 1 : 0));
|
|
|
|
|
|
|
|
if (error) {
|
|
|
|
xfs_trans_cancel(tp,
|
|
|
|
XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
} else if ((ip->i_d.di_mode & S_IFMT) == S_IFLNK) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we get an error while cleaning up a
|
|
|
|
* symlink we bail out.
|
|
|
|
*/
|
|
|
|
error = (ip->i_d.di_size > XFS_IFORK_DSIZE(ip)) ?
|
|
|
|
xfs_inactive_symlink_rmt(ip, &tp) :
|
|
|
|
xfs_inactive_symlink_local(ip, &tp);
|
|
|
|
|
|
|
|
if (error) {
|
|
|
|
ASSERT(tp == NULL);
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
} else {
|
|
|
|
error = xfs_trans_reserve(tp, 0,
|
|
|
|
XFS_IFREE_LOG_RES(mp),
|
|
|
|
0, XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_INACTIVE_LOG_COUNT);
|
|
|
|
if (error) {
|
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there are attributes associated with the file
|
|
|
|
* then blow them away now. The code calls a routine
|
|
|
|
* that recursively deconstructs the attribute fork.
|
|
|
|
* We need to just commit the current transaction
|
|
|
|
* because we can't use it for xfs_attr_inactive().
|
|
|
|
*/
|
|
|
|
if (ip->i_d.di_anextents > 0) {
|
|
|
|
error = xfs_inactive_attrs(ip, &tp);
|
|
|
|
/*
|
|
|
|
* If we got an error, the transaction is already
|
|
|
|
* cancelled, and the inode is unlocked. Just get out.
|
|
|
|
*/
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return VN_INACTIVE_CACHE;
|
2005-04-17 00:20:36 +02:00
|
|
|
} else if (ip->i_afp) {
|
|
|
|
xfs_idestroy_fork(ip, XFS_ATTR_FORK);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free the inode.
|
|
|
|
*/
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2005-04-17 00:20:36 +02:00
|
|
|
error = xfs_ifree(tp, ip, &free_list);
|
|
|
|
if (error) {
|
|
|
|
/*
|
|
|
|
* If we fail to free the inode, shut down. The cancel
|
|
|
|
* might do that, we need to make sure. Otherwise the
|
|
|
|
* inode might be lost for a long time or forever.
|
|
|
|
*/
|
|
|
|
if (!XFS_FORCED_SHUTDOWN(mp)) {
|
|
|
|
cmn_err(CE_NOTE,
|
|
|
|
"xfs_inactive: xfs_ifree() returned an error = %d on %s",
|
|
|
|
error, mp->m_fsname);
|
2006-06-09 06:58:38 +02:00
|
|
|
xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT);
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Credit the quota account(s). The inode is gone.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_ICOUNT, -1);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
2008-04-10 04:24:17 +02:00
|
|
|
* Just ignore errors at this point. There is nothing we can
|
|
|
|
* do except to try to keep going. Make sure it's not a silent
|
|
|
|
* error.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2008-04-10 04:24:17 +02:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
|
|
|
if (error)
|
|
|
|
xfs_fs_cmn_err(CE_NOTE, mp, "xfs_inactive: "
|
|
|
|
"xfs_bmap_finish() returned error %d", error);
|
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
|
|
|
if (error)
|
|
|
|
xfs_fs_cmn_err(CE_NOTE, mp, "xfs_inactive: "
|
|
|
|
"xfs_trans_commit() returned error %d", error);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2009-06-08 15:33:32 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Release the dquots held by inode, if any.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqdetach(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
out:
|
|
|
|
return VN_INACTIVE_CACHE;
|
|
|
|
}
|
|
|
|
|
2008-05-21 08:58:22 +02:00
|
|
|
/*
|
|
|
|
* Lookups up an inode from "name". If ci_name is not NULL, then a CI match
|
|
|
|
* is allowed, otherwise it has to be an exact match. If a CI match is found,
|
|
|
|
* ci_name->name will point to a the actual name (caller must free) or
|
|
|
|
* will be set to NULL if an exact match is found.
|
|
|
|
*/
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_lookup(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *dp,
|
2008-04-10 04:22:07 +02:00
|
|
|
struct xfs_name *name,
|
2008-05-21 08:58:22 +02:00
|
|
|
xfs_inode_t **ipp,
|
|
|
|
struct xfs_name *ci_name)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2008-04-22 09:33:52 +02:00
|
|
|
xfs_ino_t inum;
|
2005-04-17 00:20:36 +02:00
|
|
|
int error;
|
|
|
|
uint lock_mode;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(dp->i_mount))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
lock_mode = xfs_ilock_map_shared(dp);
|
2008-05-21 08:58:22 +02:00
|
|
|
error = xfs_dir_lookup(NULL, dp, name, &inum, ci_name);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock_map_shared(dp, lock_mode);
|
2008-04-22 09:33:52 +02:00
|
|
|
|
|
|
|
if (error)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
error = xfs_iget(dp->i_mount, NULL, inum, 0, 0, ipp, 0);
|
|
|
|
if (error)
|
2008-05-21 08:58:22 +02:00
|
|
|
goto out_free_name;
|
2008-04-22 09:33:52 +02:00
|
|
|
|
|
|
|
xfs_itrace_ref(*ipp);
|
|
|
|
return 0;
|
|
|
|
|
2008-05-21 08:58:22 +02:00
|
|
|
out_free_name:
|
|
|
|
if (ci_name)
|
|
|
|
kmem_free(ci_name->name);
|
|
|
|
out:
|
2008-04-22 09:33:52 +02:00
|
|
|
*ipp = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_create(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *dp,
|
2008-04-10 04:22:07 +02:00
|
|
|
struct xfs_name *name,
|
2007-10-11 10:09:12 +02:00
|
|
|
mode_t mode,
|
|
|
|
xfs_dev_t rdev,
|
2008-03-06 03:46:05 +01:00
|
|
|
xfs_inode_t **ipp,
|
2005-04-17 00:20:36 +02:00
|
|
|
cred_t *credp)
|
|
|
|
{
|
2009-02-09 08:38:02 +01:00
|
|
|
int is_dir = S_ISDIR(mode);
|
|
|
|
struct xfs_mount *mp = dp->i_mount;
|
|
|
|
struct xfs_inode *ip = NULL;
|
|
|
|
struct xfs_trans *tp = NULL;
|
2008-04-10 04:22:07 +02:00
|
|
|
int error;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
xfs_fsblock_t first_block;
|
2007-08-28 08:12:30 +02:00
|
|
|
boolean_t unlock_dp_on_error = B_FALSE;
|
2005-04-17 00:20:36 +02:00
|
|
|
uint cancel_flags;
|
|
|
|
int committed;
|
|
|
|
xfs_prid_t prid;
|
2009-02-09 08:38:02 +01:00
|
|
|
struct xfs_dquot *udqp = NULL;
|
|
|
|
struct xfs_dquot *gdqp = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
uint resblks;
|
2009-02-09 08:38:02 +01:00
|
|
|
uint log_res;
|
|
|
|
uint log_count;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_CREATE)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
|
2008-03-06 03:45:58 +01:00
|
|
|
dp, DM_RIGHT_NULL, NULL,
|
2008-04-10 04:22:07 +02:00
|
|
|
DM_RIGHT_NULL, name->name, NULL,
|
2007-10-11 10:09:12 +02:00
|
|
|
mode, 0, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2005-06-21 07:39:12 +02:00
|
|
|
if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
|
|
|
|
prid = dp->i_d.di_projid;
|
2005-04-17 00:20:36 +02:00
|
|
|
else
|
2009-02-09 08:38:02 +01:00
|
|
|
prid = dfltprid;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure that we have allocated dquot(s) on disk.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_vop_dqalloc(dp, current_fsuid(), current_fsgid(), prid,
|
2009-02-09 08:38:02 +01:00
|
|
|
XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto std_return;
|
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
if (is_dir) {
|
|
|
|
rdev = 0;
|
|
|
|
resblks = XFS_MKDIR_SPACE_RES(mp, name->len);
|
|
|
|
log_res = XFS_MKDIR_LOG_RES(mp);
|
|
|
|
log_count = XFS_MKDIR_LOG_COUNT;
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_MKDIR);
|
|
|
|
} else {
|
|
|
|
resblks = XFS_CREATE_SPACE_RES(mp, name->len);
|
|
|
|
log_res = XFS_CREATE_LOG_RES(mp);
|
|
|
|
log_count = XFS_CREATE_LOG_COUNT;
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_CREATE);
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
|
2009-02-09 08:38:02 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Initially assume that the file does not exist and
|
|
|
|
* reserve the resources for that case. If that is not
|
|
|
|
* the case we'll drop the one we have and get a more
|
|
|
|
* appropriate transaction later.
|
|
|
|
*/
|
2009-02-09 08:38:02 +01:00
|
|
|
error = xfs_trans_reserve(tp, resblks, log_res, 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, log_count);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error == ENOSPC) {
|
2009-04-06 18:48:30 +02:00
|
|
|
/* flush outstanding delalloc blocks and retry */
|
|
|
|
xfs_flush_inodes(dp);
|
|
|
|
error = xfs_trans_reserve(tp, resblks, XFS_CREATE_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_CREATE_LOG_COUNT);
|
|
|
|
}
|
|
|
|
if (error == ENOSPC) {
|
|
|
|
/* No space at all so try a "no-allocation" reservation */
|
2005-04-17 00:20:36 +02:00
|
|
|
resblks = 0;
|
2009-02-09 08:38:02 +01:00
|
|
|
error = xfs_trans_reserve(tp, 0, log_res, 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, log_count);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
cancel_flags = 0;
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_trans_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2007-05-08 05:50:19 +02:00
|
|
|
xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
|
2007-08-28 08:12:30 +02:00
|
|
|
unlock_dp_on_error = B_TRUE;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
/*
|
|
|
|
* Check for directory link count overflow.
|
|
|
|
*/
|
|
|
|
if (is_dir && dp->i_d.di_nlink >= XFS_MAXLINK) {
|
|
|
|
error = XFS_ERROR(EMLINK);
|
|
|
|
goto out_trans_cancel;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Reserve disk quota and the inode.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_trans_reserve_quota(tp, mp, udqp, gdqp, resblks, 1, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_trans_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_canenter(tp, dp, name, resblks);
|
|
|
|
if (error)
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_trans_cancel;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A newly created regular or special file just has one directory
|
|
|
|
* entry pointing to them, but a directory also the "." entry
|
|
|
|
* pointing to itself.
|
|
|
|
*/
|
|
|
|
error = xfs_dir_ialloc(&tp, dp, mode, is_dir ? 2 : 1, rdev, credp,
|
|
|
|
prid, resblks > 0, &ip, &committed);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
if (error == ENOSPC)
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_trans_cancel;
|
|
|
|
goto out_trans_abort;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point, we've gotten a newly allocated inode.
|
|
|
|
* It is locked (and joined to the transaction).
|
|
|
|
*/
|
2009-02-09 08:38:02 +01:00
|
|
|
xfs_itrace_ref(ip);
|
2008-04-22 09:34:00 +02:00
|
|
|
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
2007-08-28 08:12:30 +02:00
|
|
|
* Now we join the directory inode to the transaction. We do not do it
|
|
|
|
* earlier because xfs_dir_ialloc might commit the previous transaction
|
|
|
|
* (and release all the locks). An error from here on will result in
|
|
|
|
* the transaction cancel unlocking dp so don't do it explicitly in the
|
|
|
|
* error path.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2008-03-06 03:46:05 +01:00
|
|
|
IHOLD(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
|
2007-08-28 08:12:30 +02:00
|
|
|
unlock_dp_on_error = B_FALSE;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_createname(tp, dp, name, ip->i_ino,
|
2006-06-20 05:04:51 +02:00
|
|
|
&first_block, &free_list, resblks ?
|
|
|
|
resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
ASSERT(error != ENOSPC);
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_trans_abort;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
|
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
if (is_dir) {
|
|
|
|
error = xfs_dir_init(tp, ip, dp);
|
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
|
|
|
|
|
|
|
error = xfs_bumplink(tp, dp);
|
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* If this is a synchronous mount, make sure that the
|
|
|
|
* create transaction goes to disk before returning to
|
|
|
|
* the user.
|
|
|
|
*/
|
2009-02-09 08:38:02 +01:00
|
|
|
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Attach the dquot(s) to the inodes and modify them incore.
|
|
|
|
* These ids of the inode couldn't have changed since the new
|
|
|
|
* inode has been locked ever since it was created.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_vop_create_dqattach(tp, ip, udqp, gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_trans_commit normally decrements the vnode ref count
|
|
|
|
* when it unlocks the inode. Since we want to return the
|
|
|
|
* vnode to the caller, we bump the vnode ref count now.
|
|
|
|
*/
|
|
|
|
IHOLD(ip);
|
|
|
|
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
2009-02-09 08:38:02 +01:00
|
|
|
if (error)
|
|
|
|
goto out_abort_rele;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
IRELE(ip);
|
2009-02-09 08:38:02 +01:00
|
|
|
goto out_dqrele;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-03-06 03:46:05 +01:00
|
|
|
*ipp = ip;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* Fallthrough to std_return with error = 0 */
|
2009-02-09 08:38:02 +01:00
|
|
|
std_return:
|
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTCREATE)) {
|
|
|
|
XFS_SEND_NAMESP(mp, DM_EVENT_POSTCREATE, dp, DM_RIGHT_NULL,
|
|
|
|
ip, DM_RIGHT_NULL, name->name, NULL, mode,
|
|
|
|
error, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2009-02-09 08:38:02 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
out_bmap_cancel:
|
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
out_trans_abort:
|
2005-04-17 00:20:36 +02:00
|
|
|
cancel_flags |= XFS_TRANS_ABORT;
|
2009-02-09 08:38:02 +01:00
|
|
|
out_trans_cancel:
|
|
|
|
xfs_trans_cancel(tp, cancel_flags);
|
|
|
|
out_dqrele:
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
if (unlock_dp_on_error)
|
|
|
|
xfs_iunlock(dp, XFS_ILOCK_EXCL);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
goto std_return;
|
|
|
|
|
2009-02-09 08:38:02 +01:00
|
|
|
out_abort_rele:
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Wait until after the current transaction is aborted to
|
|
|
|
* release the inode. This prevents recursive transactions
|
|
|
|
* and deadlocks from xfs_inactive.
|
|
|
|
*/
|
2009-02-09 08:38:02 +01:00
|
|
|
xfs_bmap_cancel(&free_list);
|
2005-04-17 00:20:36 +02:00
|
|
|
cancel_flags |= XFS_TRANS_ABORT;
|
|
|
|
xfs_trans_cancel(tp, cancel_flags);
|
|
|
|
IRELE(ip);
|
2009-02-09 08:38:02 +01:00
|
|
|
unlock_dp_on_error = B_FALSE;
|
|
|
|
goto out_dqrele;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef DEBUG
|
|
|
|
int xfs_locked_n;
|
|
|
|
int xfs_small_retries;
|
|
|
|
int xfs_middle_retries;
|
|
|
|
int xfs_lots_retries;
|
|
|
|
int xfs_lock_delays;
|
|
|
|
#endif
|
|
|
|
|
2007-05-08 05:50:19 +02:00
|
|
|
/*
|
|
|
|
* Bump the subclass so xfs_lock_inodes() acquires each lock with
|
|
|
|
* a different value
|
|
|
|
*/
|
|
|
|
static inline int
|
|
|
|
xfs_lock_inumorder(int lock_mode, int subclass)
|
|
|
|
{
|
|
|
|
if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL))
|
2007-06-29 09:26:09 +02:00
|
|
|
lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_IOLOCK_SHIFT;
|
2007-05-08 05:50:19 +02:00
|
|
|
if (lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL))
|
2007-06-29 09:26:09 +02:00
|
|
|
lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_ILOCK_SHIFT;
|
2007-05-08 05:50:19 +02:00
|
|
|
|
|
|
|
return lock_mode;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* The following routine will lock n inodes in exclusive mode.
|
|
|
|
* We assume the caller calls us with the inodes in i_ino order.
|
|
|
|
*
|
|
|
|
* We need to detect deadlock where an inode that we lock
|
|
|
|
* is in the AIL and we start waiting for another inode that is locked
|
|
|
|
* by a thread in a long running transaction (such as truncate). This can
|
|
|
|
* result in deadlock since the long running trans might need to wait
|
|
|
|
* for the inode we just locked in order to push the tail and free space
|
|
|
|
* in the log.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_lock_inodes(
|
|
|
|
xfs_inode_t **ips,
|
|
|
|
int inodes,
|
|
|
|
uint lock_mode)
|
|
|
|
{
|
|
|
|
int attempts = 0, i, j, try_lock;
|
|
|
|
xfs_log_item_t *lp;
|
|
|
|
|
|
|
|
ASSERT(ips && (inodes >= 2)); /* we need at least two */
|
|
|
|
|
2008-04-22 09:34:06 +02:00
|
|
|
try_lock = 0;
|
|
|
|
i = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
again:
|
|
|
|
for (; i < inodes; i++) {
|
|
|
|
ASSERT(ips[i]);
|
|
|
|
|
|
|
|
if (i && (ips[i] == ips[i-1])) /* Already locked */
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If try_lock is not set yet, make sure all locked inodes
|
|
|
|
* are not in the AIL.
|
|
|
|
* If any are, set try_lock to be used later.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (!try_lock) {
|
|
|
|
for (j = (i - 1); j >= 0 && !try_lock; j--) {
|
|
|
|
lp = (xfs_log_item_t *)ips[j]->i_itemp;
|
|
|
|
if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
|
|
|
|
try_lock++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If any of the previous locks we have locked is in the AIL,
|
|
|
|
* we must TRY to get the second and subsequent locks. If
|
|
|
|
* we can't get any, we must release all we have
|
|
|
|
* and try again.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (try_lock) {
|
|
|
|
/* try_lock must be 0 if i is 0. */
|
|
|
|
/*
|
|
|
|
* try_lock means we have an inode locked
|
|
|
|
* that is in the AIL.
|
|
|
|
*/
|
|
|
|
ASSERT(i != 0);
|
2007-05-08 05:50:19 +02:00
|
|
|
if (!xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i))) {
|
2005-04-17 00:20:36 +02:00
|
|
|
attempts++;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlock all previous guys and try again.
|
|
|
|
* xfs_iunlock will try to push the tail
|
|
|
|
* if the inode is in the AIL.
|
|
|
|
*/
|
|
|
|
|
|
|
|
for(j = i - 1; j >= 0; j--) {
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check to see if we've already
|
|
|
|
* unlocked this one.
|
|
|
|
* Not the first one going back,
|
|
|
|
* and the inode ptr is the same.
|
|
|
|
*/
|
|
|
|
if ((j != (i - 1)) && ips[j] ==
|
|
|
|
ips[j+1])
|
|
|
|
continue;
|
|
|
|
|
|
|
|
xfs_iunlock(ips[j], lock_mode);
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((attempts % 5) == 0) {
|
|
|
|
delay(1); /* Don't just spin the CPU */
|
|
|
|
#ifdef DEBUG
|
|
|
|
xfs_lock_delays++;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
i = 0;
|
|
|
|
try_lock = 0;
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
} else {
|
2007-05-08 05:50:19 +02:00
|
|
|
xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i));
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef DEBUG
|
|
|
|
if (attempts) {
|
|
|
|
if (attempts < 5) xfs_small_retries++;
|
|
|
|
else if (attempts < 100) xfs_middle_retries++;
|
|
|
|
else xfs_lots_retries++;
|
|
|
|
} else {
|
|
|
|
xfs_locked_n++;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2008-09-17 08:51:21 +02:00
|
|
|
/*
|
|
|
|
* xfs_lock_two_inodes() can only be used to lock one type of lock
|
|
|
|
* at a time - the iolock or the ilock, but not both at once. If
|
|
|
|
* we lock both at once, lockdep will report false positives saying
|
|
|
|
* we have violated locking orders.
|
|
|
|
*/
|
2008-08-13 08:18:07 +02:00
|
|
|
void
|
|
|
|
xfs_lock_two_inodes(
|
|
|
|
xfs_inode_t *ip0,
|
|
|
|
xfs_inode_t *ip1,
|
|
|
|
uint lock_mode)
|
|
|
|
{
|
|
|
|
xfs_inode_t *temp;
|
|
|
|
int attempts = 0;
|
|
|
|
xfs_log_item_t *lp;
|
|
|
|
|
2008-09-17 08:51:21 +02:00
|
|
|
if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL))
|
|
|
|
ASSERT((lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)) == 0);
|
2008-08-13 08:18:07 +02:00
|
|
|
ASSERT(ip0->i_ino != ip1->i_ino);
|
|
|
|
|
|
|
|
if (ip0->i_ino > ip1->i_ino) {
|
|
|
|
temp = ip0;
|
|
|
|
ip0 = ip1;
|
|
|
|
ip1 = temp;
|
|
|
|
}
|
|
|
|
|
|
|
|
again:
|
|
|
|
xfs_ilock(ip0, xfs_lock_inumorder(lock_mode, 0));
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the first lock we have locked is in the AIL, we must TRY to get
|
|
|
|
* the second lock. If we can't get it, we must release the first one
|
|
|
|
* and try again.
|
|
|
|
*/
|
|
|
|
lp = (xfs_log_item_t *)ip0->i_itemp;
|
|
|
|
if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
|
|
|
|
if (!xfs_ilock_nowait(ip1, xfs_lock_inumorder(lock_mode, 1))) {
|
|
|
|
xfs_iunlock(ip0, lock_mode);
|
|
|
|
if ((++attempts % 5) == 0)
|
|
|
|
delay(1); /* Don't just spin the CPU */
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
xfs_ilock(ip1, xfs_lock_inumorder(lock_mode, 1));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_remove(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *dp,
|
2008-04-10 04:22:07 +02:00
|
|
|
struct xfs_name *name,
|
|
|
|
xfs_inode_t *ip)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = dp->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_t *tp = NULL;
|
2008-06-23 05:25:17 +02:00
|
|
|
int is_dir = S_ISDIR(ip->i_d.di_mode);
|
2005-04-17 00:20:36 +02:00
|
|
|
int error = 0;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
xfs_fsblock_t first_block;
|
|
|
|
int cancel_flags;
|
|
|
|
int committed;
|
|
|
|
int link_zero;
|
|
|
|
uint resblks;
|
2008-06-23 05:25:17 +02:00
|
|
|
uint log_count;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(dp);
|
2008-06-23 05:25:17 +02:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_REMOVE)) {
|
2008-04-10 04:22:07 +02:00
|
|
|
error = XFS_SEND_NAMESP(mp, DM_EVENT_REMOVE, dp, DM_RIGHT_NULL,
|
|
|
|
NULL, DM_RIGHT_NULL, name->name, NULL,
|
|
|
|
ip->i_d.di_mode, 0, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(dp, 0);
|
2008-06-23 05:25:17 +02:00
|
|
|
if (error)
|
|
|
|
goto std_return;
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(ip, 0);
|
2008-06-23 05:25:17 +02:00
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto std_return;
|
|
|
|
|
2008-06-23 05:25:17 +02:00
|
|
|
if (is_dir) {
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_RMDIR);
|
|
|
|
log_count = XFS_DEFAULT_LOG_COUNT;
|
|
|
|
} else {
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_REMOVE);
|
|
|
|
log_count = XFS_REMOVE_LOG_COUNT;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
|
2008-06-23 05:25:17 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* We try to get the real space reservation first,
|
|
|
|
* allowing for directory btree deletion(s) implying
|
|
|
|
* possible bmap insert(s). If we can't get the space
|
|
|
|
* reservation then we use 0 instead, and avoid the bmap
|
|
|
|
* btree insert(s) in the directory code by, if the bmap
|
|
|
|
* insert tries to happen, instead trimming the LAST
|
|
|
|
* block from the directory.
|
|
|
|
*/
|
|
|
|
resblks = XFS_REMOVE_SPACE_RES(mp);
|
|
|
|
error = xfs_trans_reserve(tp, resblks, XFS_REMOVE_LOG_RES(mp), 0,
|
2008-06-23 05:25:17 +02:00
|
|
|
XFS_TRANS_PERM_LOG_RES, log_count);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error == ENOSPC) {
|
|
|
|
resblks = 0;
|
|
|
|
error = xfs_trans_reserve(tp, 0, XFS_REMOVE_LOG_RES(mp), 0,
|
2008-06-23 05:25:17 +02:00
|
|
|
XFS_TRANS_PERM_LOG_RES, log_count);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
ASSERT(error != ENOSPC);
|
2008-06-23 05:25:17 +02:00
|
|
|
cancel_flags = 0;
|
|
|
|
goto out_trans_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-08-13 08:18:07 +02:00
|
|
|
xfs_lock_two_inodes(dp, ip, XFS_ILOCK_EXCL);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point, we've gotten both the directory and the entry
|
|
|
|
* inodes locked.
|
|
|
|
*/
|
2008-04-22 09:34:24 +02:00
|
|
|
IHOLD(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
|
2008-04-22 09:34:18 +02:00
|
|
|
|
|
|
|
IHOLD(dp);
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-06-23 05:25:17 +02:00
|
|
|
/*
|
|
|
|
* If we're removing a directory perform some additional validation.
|
|
|
|
*/
|
|
|
|
if (is_dir) {
|
|
|
|
ASSERT(ip->i_d.di_nlink >= 2);
|
|
|
|
if (ip->i_d.di_nlink != 2) {
|
|
|
|
error = XFS_ERROR(ENOTEMPTY);
|
|
|
|
goto out_trans_cancel;
|
|
|
|
}
|
|
|
|
if (!xfs_dir_isempty(ip)) {
|
|
|
|
error = XFS_ERROR(ENOTEMPTY);
|
|
|
|
goto out_trans_cancel;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_removename(tp, dp, name, ip->i_ino,
|
2008-04-22 09:33:46 +02:00
|
|
|
&first_block, &free_list, resblks);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
ASSERT(error != ENOENT);
|
2008-06-23 05:25:17 +02:00
|
|
|
goto out_bmap_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
|
2008-06-23 05:25:17 +02:00
|
|
|
if (is_dir) {
|
|
|
|
/*
|
|
|
|
* Drop the link from ip's "..".
|
|
|
|
*/
|
|
|
|
error = xfs_droplink(tp, dp);
|
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
|
|
|
|
|
|
|
/*
|
2008-10-30 07:55:18 +01:00
|
|
|
* Drop the "." link from ip to self.
|
2008-06-23 05:25:17 +02:00
|
|
|
*/
|
|
|
|
error = xfs_droplink(tp, ip);
|
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* When removing a non-directory we need to log the parent
|
2008-11-28 04:23:37 +01:00
|
|
|
* inode here. For a directory this is done implicitly
|
|
|
|
* by the xfs_droplink call for the ".." entry.
|
2008-06-23 05:25:17 +02:00
|
|
|
*/
|
|
|
|
xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-06-23 05:25:17 +02:00
|
|
|
/*
|
2008-10-30 07:55:18 +01:00
|
|
|
* Drop the link from dp to ip.
|
2008-06-23 05:25:17 +02:00
|
|
|
*/
|
|
|
|
error = xfs_droplink(tp, ip);
|
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Determine if this is the last link while
|
2005-04-17 00:20:36 +02:00
|
|
|
* we are in the transaction.
|
|
|
|
*/
|
2008-06-23 05:25:17 +02:00
|
|
|
link_zero = (ip->i_d.di_nlink == 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a synchronous mount, make sure that the
|
|
|
|
* remove transaction goes to disk before returning to
|
|
|
|
* the user.
|
|
|
|
*/
|
2008-06-23 05:25:17 +02:00
|
|
|
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC))
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
2008-06-23 05:25:17 +02:00
|
|
|
if (error)
|
|
|
|
goto out_bmap_cancel;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2008-04-22 09:34:24 +02:00
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto std_return;
|
|
|
|
|
2007-07-11 03:09:12 +02:00
|
|
|
/*
|
|
|
|
* If we are using filestreams, kill the stream association.
|
|
|
|
* If the file is still open it may get a new one but that
|
|
|
|
* will get killed on last close in xfs_close() so we don't
|
|
|
|
* have to worry about that.
|
|
|
|
*/
|
2008-06-23 05:25:17 +02:00
|
|
|
if (!is_dir && link_zero && xfs_inode_is_filestream(ip))
|
2007-07-11 03:09:12 +02:00
|
|
|
xfs_filestream_deassociate(ip);
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_exit(ip);
|
2008-06-23 05:25:17 +02:00
|
|
|
xfs_itrace_exit(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
std_return:
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTREMOVE)) {
|
2008-06-23 05:25:17 +02:00
|
|
|
XFS_SEND_NAMESP(mp, DM_EVENT_POSTREMOVE, dp, DM_RIGHT_NULL,
|
|
|
|
NULL, DM_RIGHT_NULL, name->name, NULL,
|
|
|
|
ip->i_d.di_mode, error, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2008-06-23 05:25:17 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
|
2008-06-23 05:25:17 +02:00
|
|
|
out_bmap_cancel:
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
cancel_flags |= XFS_TRANS_ABORT;
|
2008-06-23 05:25:17 +02:00
|
|
|
out_trans_cancel:
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_cancel(tp, cancel_flags);
|
|
|
|
goto std_return;
|
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_link(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *tdp,
|
2008-03-06 03:46:12 +01:00
|
|
|
xfs_inode_t *sip,
|
2008-04-10 04:22:07 +02:00
|
|
|
struct xfs_name *target_name)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = tdp->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_t *tp;
|
|
|
|
int error;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
xfs_fsblock_t first_block;
|
|
|
|
int cancel_flags;
|
|
|
|
int committed;
|
|
|
|
int resblks;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(tdp);
|
2008-03-06 03:46:12 +01:00
|
|
|
xfs_itrace_entry(sip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-03-06 03:46:12 +01:00
|
|
|
ASSERT(!S_ISDIR(sip->i_d.di_mode));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(tdp, DM_EVENT_LINK)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
error = XFS_SEND_NAMESP(mp, DM_EVENT_LINK,
|
2008-03-06 03:45:58 +01:00
|
|
|
tdp, DM_RIGHT_NULL,
|
|
|
|
sip, DM_RIGHT_NULL,
|
2008-04-10 04:22:07 +02:00
|
|
|
target_name->name, NULL, 0, 0, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Return through std_return after this point. */
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(sip, 0);
|
2009-02-04 09:34:20 +01:00
|
|
|
if (error)
|
|
|
|
goto std_return;
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(tdp, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto std_return;
|
|
|
|
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_LINK);
|
|
|
|
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
|
2008-04-10 04:22:07 +02:00
|
|
|
resblks = XFS_LINK_SPACE_RES(mp, target_name->len);
|
2005-04-17 00:20:36 +02:00
|
|
|
error = xfs_trans_reserve(tp, resblks, XFS_LINK_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
|
|
|
|
if (error == ENOSPC) {
|
|
|
|
resblks = 0;
|
|
|
|
error = xfs_trans_reserve(tp, 0, XFS_LINK_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
|
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
cancel_flags = 0;
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
2008-08-13 08:18:07 +02:00
|
|
|
xfs_lock_two_inodes(sip, tdp, XFS_ILOCK_EXCL);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Increment vnode ref counts since xfs_trans_commit &
|
|
|
|
* xfs_trans_cancel will both unlock the inodes and
|
|
|
|
* decrement the associated ref counts.
|
|
|
|
*/
|
2008-03-06 03:46:12 +01:00
|
|
|
IHOLD(sip);
|
|
|
|
IHOLD(tdp);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_ijoin(tp, sip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, tdp, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the source has too many links, we can't make any more to it.
|
|
|
|
*/
|
|
|
|
if (sip->i_d.di_nlink >= XFS_MAXLINK) {
|
|
|
|
error = XFS_ERROR(EMLINK);
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
2005-06-21 07:39:12 +02:00
|
|
|
/*
|
|
|
|
* If we are using project inheritance, we only allow hard link
|
|
|
|
* creation in our tree when the project IDs are the same; else
|
|
|
|
* the tree quota mechanism could be circumvented.
|
|
|
|
*/
|
|
|
|
if (unlikely((tdp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
|
|
|
|
(tdp->i_d.di_projid != sip->i_d.di_projid))) {
|
2006-05-08 11:51:42 +02:00
|
|
|
error = XFS_ERROR(EXDEV);
|
2005-06-21 07:39:12 +02:00
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_canenter(tp, tdp, target_name, resblks);
|
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error_return;
|
|
|
|
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_createname(tp, tdp, target_name, sip->i_ino,
|
|
|
|
&first_block, &free_list, resblks);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto abort_return;
|
|
|
|
xfs_ichgtime(tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE);
|
|
|
|
|
|
|
|
error = xfs_bumplink(tp, sip);
|
2006-06-27 04:45:17 +02:00
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto abort_return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a synchronous mount, make sure that the
|
|
|
|
* link transaction goes to disk before returning to
|
|
|
|
* the user.
|
|
|
|
*/
|
|
|
|
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
|
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
}
|
|
|
|
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish (&tp, &free_list, &committed);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
goto abort_return;
|
|
|
|
}
|
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2006-06-27 04:45:17 +02:00
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto std_return;
|
|
|
|
|
|
|
|
/* Fall through to std_return with error = 0. */
|
|
|
|
std_return:
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(sip, DM_EVENT_POSTLINK)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTLINK,
|
2008-03-06 03:45:58 +01:00
|
|
|
tdp, DM_RIGHT_NULL,
|
|
|
|
sip, DM_RIGHT_NULL,
|
2008-04-10 04:22:07 +02:00
|
|
|
target_name->name, NULL, 0, error, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
return error;
|
|
|
|
|
|
|
|
abort_return:
|
|
|
|
cancel_flags |= XFS_TRANS_ABORT;
|
|
|
|
/* FALLTHROUGH */
|
2006-01-15 02:37:08 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
error_return:
|
|
|
|
xfs_trans_cancel(tp, cancel_flags);
|
|
|
|
goto std_return;
|
|
|
|
}
|
2006-06-27 04:45:17 +02:00
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_symlink(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *dp,
|
2008-04-10 04:22:07 +02:00
|
|
|
struct xfs_name *link_name,
|
|
|
|
const char *target_path,
|
2007-10-11 10:09:12 +02:00
|
|
|
mode_t mode,
|
2008-03-06 03:46:19 +01:00
|
|
|
xfs_inode_t **ipp,
|
2005-04-17 00:20:36 +02:00
|
|
|
cred_t *credp)
|
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = dp->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_t *tp;
|
|
|
|
xfs_inode_t *ip;
|
|
|
|
int error;
|
|
|
|
int pathlen;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
xfs_fsblock_t first_block;
|
2007-08-28 08:12:30 +02:00
|
|
|
boolean_t unlock_dp_on_error = B_FALSE;
|
2005-04-17 00:20:36 +02:00
|
|
|
uint cancel_flags;
|
|
|
|
int committed;
|
|
|
|
xfs_fileoff_t first_fsb;
|
|
|
|
xfs_filblks_t fs_blocks;
|
|
|
|
int nmaps;
|
|
|
|
xfs_bmbt_irec_t mval[SYMLINK_MAPS];
|
|
|
|
xfs_daddr_t d;
|
2008-04-10 04:22:07 +02:00
|
|
|
const char *cur_chunk;
|
2005-04-17 00:20:36 +02:00
|
|
|
int byte_cnt;
|
|
|
|
int n;
|
|
|
|
xfs_buf_t *bp;
|
|
|
|
xfs_prid_t prid;
|
|
|
|
struct xfs_dquot *udqp, *gdqp;
|
|
|
|
uint resblks;
|
|
|
|
|
2008-03-06 03:46:19 +01:00
|
|
|
*ipp = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
error = 0;
|
|
|
|
ip = NULL;
|
|
|
|
tp = NULL;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check component lengths of the target path name.
|
|
|
|
*/
|
|
|
|
pathlen = strlen(target_path);
|
|
|
|
if (pathlen >= MAXPATHLEN) /* total string too long */
|
|
|
|
return XFS_ERROR(ENAMETOOLONG);
|
|
|
|
|
2007-08-16 10:42:07 +02:00
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_SYMLINK)) {
|
2008-03-06 03:45:58 +01:00
|
|
|
error = XFS_SEND_NAMESP(mp, DM_EVENT_SYMLINK, dp,
|
2005-04-17 00:20:36 +02:00
|
|
|
DM_RIGHT_NULL, NULL, DM_RIGHT_NULL,
|
2008-04-10 04:22:07 +02:00
|
|
|
link_name->name, target_path, 0, 0, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Return through std_return after this point. */
|
|
|
|
|
|
|
|
udqp = gdqp = NULL;
|
2005-06-21 07:39:12 +02:00
|
|
|
if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
|
|
|
|
prid = dp->i_d.di_projid;
|
2005-04-17 00:20:36 +02:00
|
|
|
else
|
|
|
|
prid = (xfs_prid_t)dfltprid;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure that we have allocated dquot(s) on disk.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_vop_dqalloc(dp, current_fsuid(), current_fsgid(), prid,
|
2005-04-17 00:20:36 +02:00
|
|
|
XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp);
|
|
|
|
if (error)
|
|
|
|
goto std_return;
|
|
|
|
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_SYMLINK);
|
|
|
|
cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
|
|
|
|
/*
|
|
|
|
* The symlink will fit into the inode data fork?
|
|
|
|
* There can't be any attributes so we get the whole variable part.
|
|
|
|
*/
|
|
|
|
if (pathlen <= XFS_LITINO(mp))
|
|
|
|
fs_blocks = 0;
|
|
|
|
else
|
|
|
|
fs_blocks = XFS_B_TO_FSB(mp, pathlen);
|
2008-04-10 04:22:07 +02:00
|
|
|
resblks = XFS_SYMLINK_SPACE_RES(mp, link_name->len, fs_blocks);
|
2005-04-17 00:20:36 +02:00
|
|
|
error = xfs_trans_reserve(tp, resblks, XFS_SYMLINK_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_SYMLINK_LOG_COUNT);
|
|
|
|
if (error == ENOSPC && fs_blocks == 0) {
|
|
|
|
resblks = 0;
|
|
|
|
error = xfs_trans_reserve(tp, 0, XFS_SYMLINK_LOG_RES(mp), 0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES, XFS_SYMLINK_LOG_COUNT);
|
|
|
|
}
|
|
|
|
if (error) {
|
|
|
|
cancel_flags = 0;
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
2007-05-08 05:50:19 +02:00
|
|
|
xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
|
2007-08-28 08:12:30 +02:00
|
|
|
unlock_dp_on_error = B_TRUE;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check whether the directory allows new symlinks or not.
|
|
|
|
*/
|
|
|
|
if (dp->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) {
|
|
|
|
error = XFS_ERROR(EPERM);
|
|
|
|
goto error_return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reserve disk quota : blocks and inode.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_trans_reserve_quota(tp, mp, udqp, gdqp, resblks, 1, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto error_return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for ability to enter directory entry, if no space reserved.
|
|
|
|
*/
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_canenter(tp, dp, link_name, resblks);
|
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error_return;
|
|
|
|
/*
|
|
|
|
* Initialize the bmap freelist prior to calling either
|
|
|
|
* bmapi or the directory create code.
|
|
|
|
*/
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &first_block);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate an inode for the symlink.
|
|
|
|
*/
|
2007-10-11 10:09:12 +02:00
|
|
|
error = xfs_dir_ialloc(&tp, dp, S_IFLNK | (mode & ~S_IFMT),
|
2005-04-17 00:20:36 +02:00
|
|
|
1, 0, credp, prid, resblks > 0, &ip, NULL);
|
|
|
|
if (error) {
|
|
|
|
if (error == ENOSPC)
|
|
|
|
goto error_return;
|
|
|
|
goto error1;
|
|
|
|
}
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_ref(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
/*
|
|
|
|
* An error after we've joined dp to the transaction will result in the
|
|
|
|
* transaction cancel unlocking dp so don't do it explicitly in the
|
|
|
|
* error path.
|
|
|
|
*/
|
2008-03-06 03:46:19 +01:00
|
|
|
IHOLD(dp);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
|
2007-08-28 08:12:30 +02:00
|
|
|
unlock_dp_on_error = B_FALSE;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Also attach the dquot(s) to it, if applicable.
|
|
|
|
*/
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_vop_create_dqattach(tp, ip, udqp, gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (resblks)
|
|
|
|
resblks -= XFS_IALLOC_SPACE_RES(mp);
|
|
|
|
/*
|
|
|
|
* If the symlink will fit into the inode, write it inline.
|
|
|
|
*/
|
|
|
|
if (pathlen <= XFS_IFORK_DSIZE(ip)) {
|
|
|
|
xfs_idata_realloc(ip, pathlen, XFS_DATA_FORK);
|
|
|
|
memcpy(ip->i_df.if_u1.if_data, target_path, pathlen);
|
|
|
|
ip->i_d.di_size = pathlen;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The inode was initially created in extent format.
|
|
|
|
*/
|
|
|
|
ip->i_df.if_flags &= ~(XFS_IFEXTENTS | XFS_IFBROOT);
|
|
|
|
ip->i_df.if_flags |= XFS_IFINLINE;
|
|
|
|
|
|
|
|
ip->i_d.di_format = XFS_DINODE_FMT_LOCAL;
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_DDATA | XFS_ILOG_CORE);
|
|
|
|
|
|
|
|
} else {
|
|
|
|
first_fsb = 0;
|
|
|
|
nmaps = SYMLINK_MAPS;
|
|
|
|
|
|
|
|
error = xfs_bmapi(tp, ip, first_fsb, fs_blocks,
|
|
|
|
XFS_BMAPI_WRITE | XFS_BMAPI_METADATA,
|
|
|
|
&first_block, resblks, mval, &nmaps,
|
2006-06-09 06:48:12 +02:00
|
|
|
&free_list, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (resblks)
|
|
|
|
resblks -= fs_blocks;
|
|
|
|
ip->i_d.di_size = pathlen;
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
|
|
|
|
cur_chunk = target_path;
|
|
|
|
for (n = 0; n < nmaps; n++) {
|
|
|
|
d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
|
|
|
|
byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
|
|
|
|
bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, d,
|
|
|
|
BTOBB(byte_cnt), 0);
|
|
|
|
ASSERT(bp && !XFS_BUF_GETERROR(bp));
|
|
|
|
if (pathlen < byte_cnt) {
|
|
|
|
byte_cnt = pathlen;
|
|
|
|
}
|
|
|
|
pathlen -= byte_cnt;
|
|
|
|
|
|
|
|
memcpy(XFS_BUF_PTR(bp), cur_chunk, byte_cnt);
|
|
|
|
cur_chunk += byte_cnt;
|
|
|
|
|
|
|
|
xfs_trans_log_buf(tp, bp, 0, byte_cnt - 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Create the directory entry for the symlink.
|
|
|
|
*/
|
2008-04-10 04:22:07 +02:00
|
|
|
error = xfs_dir_createname(tp, dp, link_name, ip->i_ino,
|
|
|
|
&first_block, &free_list, resblks);
|
2006-06-20 05:04:51 +02:00
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto error1;
|
|
|
|
xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a synchronous mount, make sure that the
|
|
|
|
* symlink transaction goes to disk before returning to
|
|
|
|
* the user.
|
|
|
|
*/
|
|
|
|
if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
|
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_trans_commit normally decrements the vnode ref count
|
|
|
|
* when it unlocks the inode. Since we want to return the
|
|
|
|
* vnode to the caller, we bump the vnode ref count now.
|
|
|
|
*/
|
|
|
|
IHOLD(ip);
|
|
|
|
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error2;
|
|
|
|
}
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* Fall through to std_return with error = 0 or errno from
|
|
|
|
* xfs_trans_commit */
|
|
|
|
std_return:
|
2007-08-28 08:12:30 +02:00
|
|
|
if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTSYMLINK)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
(void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTSYMLINK,
|
2008-03-06 03:45:58 +01:00
|
|
|
dp, DM_RIGHT_NULL,
|
|
|
|
error ? NULL : ip,
|
2008-04-10 04:22:07 +02:00
|
|
|
DM_RIGHT_NULL, link_name->name,
|
|
|
|
target_path, 0, error, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-03-06 03:46:19 +01:00
|
|
|
if (!error)
|
|
|
|
*ipp = ip;
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
|
|
|
|
error2:
|
|
|
|
IRELE(ip);
|
|
|
|
error1:
|
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
cancel_flags |= XFS_TRANS_ABORT;
|
|
|
|
error_return:
|
|
|
|
xfs_trans_cancel(tp, cancel_flags);
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_dqrele(udqp);
|
|
|
|
xfs_qm_dqrele(gdqp);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
if (unlock_dp_on_error)
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(dp, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
goto std_return;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_set_dmattrs(
|
|
|
|
xfs_inode_t *ip,
|
2005-04-17 00:20:36 +02:00
|
|
|
u_int evmask,
|
2007-08-28 08:12:30 +02:00
|
|
|
u_int16_t state)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_t *tp;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (!capable(CAP_SYS_ADMIN))
|
|
|
|
return XFS_ERROR(EPERM);
|
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
|
|
|
|
error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES (mp), 0, 0, 0);
|
|
|
|
if (error) {
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
|
|
|
|
2007-10-11 09:44:08 +02:00
|
|
|
ip->i_d.di_dmevmask = evmask;
|
|
|
|
ip->i_d.di_dmstate = state;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
IHOLD(ip);
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_reclaim(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *ip)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-08-13 08:22:09 +02:00
|
|
|
ASSERT(!VN_MAPPED(VFS_I(ip)));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* bad inode, get out here ASAP */
|
2009-03-16 08:25:25 +01:00
|
|
|
if (is_bad_inode(VFS_I(ip))) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ireclaim(ip);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-12-03 12:20:39 +01:00
|
|
|
xfs_ioend_wait(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-09-02 08:58:38 +02:00
|
|
|
ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-01-11 05:35:17 +01:00
|
|
|
/*
|
|
|
|
* Make sure the atime in the XFS inode is correct before freeing the
|
|
|
|
* Linux inode.
|
|
|
|
*/
|
|
|
|
xfs_synchronize_atime(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-11-11 08:05:00 +01:00
|
|
|
/*
|
|
|
|
* If we have nothing to flush with this inode then complete the
|
|
|
|
* teardown now, otherwise break the link between the xfs inode and the
|
|
|
|
* linux inode and clean up the xfs inode later. This avoids flushing
|
|
|
|
* the inode to disk during the delete operation itself.
|
|
|
|
*
|
|
|
|
* When breaking the link, we need to set the XFS_IRECLAIMABLE flag
|
|
|
|
* first to ensure that xfs_iunpin() will never see an xfs inode
|
|
|
|
* that has a linux inode being reclaimed. Synchronisation is provided
|
|
|
|
* by the i_flags_lock.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
if (!ip->i_update_core && (ip->i_itemp == NULL)) {
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_iflock(ip);
|
2008-10-30 07:36:14 +01:00
|
|
|
xfs_iflags_set(ip, XFS_IRECLAIMABLE);
|
2008-10-30 07:37:15 +01:00
|
|
|
return xfs_reclaim_inode(ip, 1, XFS_IFLUSH_DELWRI_ELSE_SYNC);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2008-10-30 07:37:49 +01:00
|
|
|
xfs_inode_set_reclaim_tag(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_alloc_file_space()
|
|
|
|
* This routine allocates disk space for the given file.
|
|
|
|
*
|
|
|
|
* If alloc_type == 0, this request is for an ALLOCSP type
|
|
|
|
* request which will change the file size. In this case, no
|
|
|
|
* DMAPI event will be generated by the call. A TRUNCATE event
|
|
|
|
* will be generated later by xfs_setattr.
|
|
|
|
*
|
|
|
|
* If alloc_type != 0, this request is for a RESVSP type
|
|
|
|
* request, and a DMAPI DM_EVENT_WRITE will be generated if the
|
|
|
|
* lower block boundary byte address is less than the file's
|
|
|
|
* length.
|
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on success
|
|
|
|
* errno on error
|
|
|
|
*
|
|
|
|
*/
|
2005-06-21 07:36:52 +02:00
|
|
|
STATIC int
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_alloc_file_space(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_off_t offset,
|
|
|
|
xfs_off_t len,
|
|
|
|
int alloc_type,
|
|
|
|
int attr_flags)
|
|
|
|
{
|
2006-01-11 05:28:28 +01:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
|
|
|
xfs_off_t count;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_filblks_t allocated_fsb;
|
|
|
|
xfs_filblks_t allocatesize_fsb;
|
2006-01-11 05:28:28 +01:00
|
|
|
xfs_extlen_t extsz, temp;
|
|
|
|
xfs_fileoff_t startoffset_fsb;
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_fsblock_t firstfsb;
|
2006-01-11 05:28:28 +01:00
|
|
|
int nimaps;
|
|
|
|
int bmapi_flag;
|
|
|
|
int quota_flag;
|
2005-04-17 00:20:36 +02:00
|
|
|
int rt;
|
|
|
|
xfs_trans_t *tp;
|
2006-01-11 05:28:28 +01:00
|
|
|
xfs_bmbt_irec_t imaps[1], *imapp;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
uint qblocks, resblks, resrtextents;
|
|
|
|
int committed;
|
|
|
|
int error;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (XFS_FORCED_SHUTDOWN(mp))
|
|
|
|
return XFS_ERROR(EIO);
|
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(ip, 0);
|
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
|
|
|
|
if (len <= 0)
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
|
2007-06-18 08:50:37 +02:00
|
|
|
rt = XFS_IS_REALTIME_INODE(ip);
|
|
|
|
extsz = xfs_get_extsz_hint(ip);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
count = len;
|
|
|
|
imapp = &imaps[0];
|
2006-01-11 05:28:28 +01:00
|
|
|
nimaps = 1;
|
|
|
|
bmapi_flag = XFS_BMAPI_WRITE | (alloc_type ? XFS_BMAPI_PREALLOC : 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
startoffset_fsb = XFS_B_TO_FSBT(mp, offset);
|
|
|
|
allocatesize_fsb = XFS_B_TO_FSB(mp, count);
|
|
|
|
|
|
|
|
/* Generate a DMAPI event if needed. */
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
if (alloc_type != 0 && offset < ip->i_size &&
|
2008-07-18 09:13:28 +02:00
|
|
|
(attr_flags & XFS_ATTR_DMI) == 0 &&
|
2007-08-16 10:42:07 +02:00
|
|
|
DM_EVENT_ENABLED(ip, DM_EVENT_WRITE)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_off_t end_dmi_offset;
|
|
|
|
|
|
|
|
end_dmi_offset = offset+len;
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
if (end_dmi_offset > ip->i_size)
|
|
|
|
end_dmi_offset = ip->i_size;
|
2008-03-06 03:45:58 +01:00
|
|
|
error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, ip, offset,
|
|
|
|
end_dmi_offset - offset, 0, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-01-11 05:28:28 +01:00
|
|
|
* Allocate file space until done or until there is an error
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
retry:
|
|
|
|
while (allocatesize_fsb && !error) {
|
2006-01-11 05:28:28 +01:00
|
|
|
xfs_fileoff_t s, e;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2006-01-11 05:33:02 +01:00
|
|
|
* Determine space reservations for data/realtime.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2006-01-11 05:28:28 +01:00
|
|
|
if (unlikely(extsz)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
s = startoffset_fsb;
|
2006-01-11 05:28:28 +01:00
|
|
|
do_div(s, extsz);
|
|
|
|
s *= extsz;
|
|
|
|
e = startoffset_fsb + allocatesize_fsb;
|
|
|
|
if ((temp = do_mod(startoffset_fsb, extsz)))
|
|
|
|
e += temp;
|
|
|
|
if ((temp = do_mod(e, extsz)))
|
|
|
|
e += extsz - temp;
|
|
|
|
} else {
|
|
|
|
s = 0;
|
|
|
|
e = allocatesize_fsb;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(rt)) {
|
|
|
|
resrtextents = qblocks = (uint)(e - s);
|
|
|
|
resrtextents /= mp->m_sb.sb_rextsize;
|
|
|
|
resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
|
|
|
|
quota_flag = XFS_QMOPT_RES_RTBLKS;
|
2005-04-17 00:20:36 +02:00
|
|
|
} else {
|
2006-01-11 05:28:28 +01:00
|
|
|
resrtextents = 0;
|
|
|
|
resblks = qblocks = \
|
|
|
|
XFS_DIOSTRAT_SPACE_RES(mp, (uint)(e - s));
|
|
|
|
quota_flag = XFS_QMOPT_RES_REGBLKS;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-01-11 05:28:28 +01:00
|
|
|
* Allocate and setup the transaction.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
|
2006-01-11 05:28:28 +01:00
|
|
|
error = xfs_trans_reserve(tp, resblks,
|
|
|
|
XFS_WRITE_LOG_RES(mp), resrtextents,
|
2005-04-17 00:20:36 +02:00
|
|
|
XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_WRITE_LOG_COUNT);
|
|
|
|
/*
|
2006-01-11 05:28:28 +01:00
|
|
|
* Check for running out of space
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
if (error) {
|
|
|
|
/*
|
|
|
|
* Free the transaction structure.
|
|
|
|
*/
|
|
|
|
ASSERT(error == ENOSPC || XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_trans_reserve_quota_nblks(tp, ip, qblocks,
|
|
|
|
0, quota_flag);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto error1;
|
|
|
|
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
|
|
|
/*
|
2006-01-11 05:28:28 +01:00
|
|
|
* Issue the xfs_bmapi() call to allocate the blocks
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &firstfsb);
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bmapi(tp, ip, startoffset_fsb,
|
2006-01-11 05:28:28 +01:00
|
|
|
allocatesize_fsb, bmapi_flag,
|
|
|
|
&firstfsb, 0, imapp, &nimaps,
|
2006-06-09 06:48:12 +02:00
|
|
|
&free_list, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-01-11 05:28:28 +01:00
|
|
|
* Complete the transaction
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
if (error) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
allocated_fsb = imapp->br_blockcount;
|
|
|
|
|
2006-01-11 05:28:28 +01:00
|
|
|
if (nimaps == 0) {
|
2005-04-17 00:20:36 +02:00
|
|
|
error = XFS_ERROR(ENOSPC);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
startoffset_fsb += allocated_fsb;
|
|
|
|
allocatesize_fsb -= allocated_fsb;
|
|
|
|
}
|
|
|
|
dmapi_enospc_check:
|
2008-07-18 09:13:28 +02:00
|
|
|
if (error == ENOSPC && (attr_flags & XFS_ATTR_DMI) == 0 &&
|
2007-08-16 10:42:07 +02:00
|
|
|
DM_EVENT_ENABLED(ip, DM_EVENT_NOSPACE)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
error = XFS_SEND_NAMESP(mp, DM_EVENT_NOSPACE,
|
2008-03-06 03:45:58 +01:00
|
|
|
ip, DM_RIGHT_NULL,
|
|
|
|
ip, DM_RIGHT_NULL,
|
2005-04-17 00:20:36 +02:00
|
|
|
NULL, NULL, 0, 0, 0); /* Delay flag intentionally unused */
|
|
|
|
if (error == 0)
|
|
|
|
goto retry; /* Maybe DMAPI app. has made space */
|
|
|
|
/* else fall through with error from XFS_SEND_DATA */
|
|
|
|
}
|
|
|
|
|
|
|
|
return error;
|
|
|
|
|
2006-01-11 05:28:28 +01:00
|
|
|
error0: /* Cancel bmap, unlock inode, unreserve quota blocks, cancel trans */
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_bmap_cancel(&free_list);
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_trans_unreserve_quota_nblks(tp, ip, qblocks, 0, quota_flag);
|
2006-01-11 05:28:28 +01:00
|
|
|
|
|
|
|
error1: /* Just cancel transaction */
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
goto dmapi_enospc_check;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Zero file bytes between startoff and endoff inclusive.
|
|
|
|
* The iolock is held exclusive and no blocks are buffered.
|
2008-09-17 08:52:50 +02:00
|
|
|
*
|
|
|
|
* This function is used by xfs_free_file_space() to zero
|
|
|
|
* partial blocks when the range to free is not block aligned.
|
|
|
|
* When unreserving space with boundaries that are not block
|
|
|
|
* aligned we round up the start and round down the end
|
|
|
|
* boundaries and then use this function to zero the parts of
|
|
|
|
* the blocks that got dropped during the rounding.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_zero_remaining_bytes(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_off_t startoff,
|
|
|
|
xfs_off_t endoff)
|
|
|
|
{
|
|
|
|
xfs_bmbt_irec_t imap;
|
|
|
|
xfs_fileoff_t offset_fsb;
|
|
|
|
xfs_off_t lastoffset;
|
|
|
|
xfs_off_t offset;
|
|
|
|
xfs_buf_t *bp;
|
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
|
|
|
int nimap;
|
|
|
|
int error = 0;
|
|
|
|
|
2008-09-17 08:52:50 +02:00
|
|
|
/*
|
|
|
|
* Avoid doing I/O beyond eof - it's not necessary
|
|
|
|
* since nothing can read beyond eof. The space will
|
|
|
|
* be zeroed when the file is extended anyway.
|
|
|
|
*/
|
|
|
|
if (startoff >= ip->i_size)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (endoff > ip->i_size)
|
|
|
|
endoff = ip->i_size;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
bp = xfs_buf_get_noaddr(mp->m_sb.sb_blocksize,
|
2007-11-23 06:29:42 +01:00
|
|
|
XFS_IS_REALTIME_INODE(ip) ?
|
2005-04-17 00:20:36 +02:00
|
|
|
mp->m_rtdev_targp : mp->m_ddev_targp);
|
2008-12-05 03:16:15 +01:00
|
|
|
if (!bp)
|
|
|
|
return XFS_ERROR(ENOMEM);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
for (offset = startoff; offset <= endoff; offset = lastoffset + 1) {
|
|
|
|
offset_fsb = XFS_B_TO_FSBT(mp, offset);
|
|
|
|
nimap = 1;
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bmapi(NULL, ip, offset_fsb, 1, 0,
|
2006-06-09 06:48:12 +02:00
|
|
|
NULL, 0, &imap, &nimap, NULL, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error || nimap < 1)
|
|
|
|
break;
|
|
|
|
ASSERT(imap.br_blockcount >= 1);
|
|
|
|
ASSERT(imap.br_startoff == offset_fsb);
|
|
|
|
lastoffset = XFS_FSB_TO_B(mp, imap.br_startoff + 1) - 1;
|
|
|
|
if (lastoffset > endoff)
|
|
|
|
lastoffset = endoff;
|
|
|
|
if (imap.br_startblock == HOLESTARTBLOCK)
|
|
|
|
continue;
|
|
|
|
ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
|
|
|
|
if (imap.br_state == XFS_EXT_UNWRITTEN)
|
|
|
|
continue;
|
|
|
|
XFS_BUF_UNDONE(bp);
|
|
|
|
XFS_BUF_UNWRITE(bp);
|
|
|
|
XFS_BUF_READ(bp);
|
2009-01-15 06:22:07 +01:00
|
|
|
XFS_BUF_SET_ADDR(bp, xfs_fsb_to_db(ip, imap.br_startblock));
|
2005-04-17 00:20:36 +02:00
|
|
|
xfsbdstrat(mp, bp);
|
2008-04-10 04:22:17 +02:00
|
|
|
error = xfs_iowait(bp);
|
|
|
|
if (error) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ioerror_alert("xfs_zero_remaining_bytes(read)",
|
|
|
|
mp, bp, XFS_BUF_ADDR(bp));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
memset(XFS_BUF_PTR(bp) +
|
|
|
|
(offset - XFS_FSB_TO_B(mp, imap.br_startoff)),
|
|
|
|
0, lastoffset - offset + 1);
|
|
|
|
XFS_BUF_UNDONE(bp);
|
|
|
|
XFS_BUF_UNREAD(bp);
|
|
|
|
XFS_BUF_WRITE(bp);
|
|
|
|
xfsbdstrat(mp, bp);
|
2008-04-10 04:22:17 +02:00
|
|
|
error = xfs_iowait(bp);
|
|
|
|
if (error) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ioerror_alert("xfs_zero_remaining_bytes(write)",
|
|
|
|
mp, bp, XFS_BUF_ADDR(bp));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
xfs_buf_free(bp);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_free_file_space()
|
|
|
|
* This routine frees disk space for the given file.
|
|
|
|
*
|
|
|
|
* This routine is only called by xfs_change_file_space
|
|
|
|
* for an UNRESVSP type call.
|
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on success
|
|
|
|
* errno on error
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_free_file_space(
|
|
|
|
xfs_inode_t *ip,
|
|
|
|
xfs_off_t offset,
|
|
|
|
xfs_off_t len,
|
|
|
|
int attr_flags)
|
|
|
|
{
|
|
|
|
int committed;
|
|
|
|
int done;
|
|
|
|
xfs_off_t end_dmi_offset;
|
|
|
|
xfs_fileoff_t endoffset_fsb;
|
|
|
|
int error;
|
|
|
|
xfs_fsblock_t firstfsb;
|
|
|
|
xfs_bmap_free_t free_list;
|
|
|
|
xfs_bmbt_irec_t imap;
|
|
|
|
xfs_off_t ioffset;
|
|
|
|
xfs_extlen_t mod=0;
|
|
|
|
xfs_mount_t *mp;
|
|
|
|
int nimap;
|
|
|
|
uint resblks;
|
2006-09-28 02:56:26 +02:00
|
|
|
uint rounding;
|
2005-04-17 00:20:36 +02:00
|
|
|
int rt;
|
|
|
|
xfs_fileoff_t startoffset_fsb;
|
|
|
|
xfs_trans_t *tp;
|
2005-05-05 22:27:19 +02:00
|
|
|
int need_iolock = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
mp = ip->i_mount;
|
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-06-21 07:47:39 +02:00
|
|
|
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_qm_dqattach(ip, 0);
|
|
|
|
if (error)
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
|
|
|
|
error = 0;
|
|
|
|
if (len <= 0) /* if nothing being freed */
|
|
|
|
return error;
|
2007-11-23 06:29:42 +01:00
|
|
|
rt = XFS_IS_REALTIME_INODE(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
startoffset_fsb = XFS_B_TO_FSB(mp, offset);
|
|
|
|
end_dmi_offset = offset + len;
|
|
|
|
endoffset_fsb = XFS_B_TO_FSBT(mp, end_dmi_offset);
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if (offset < ip->i_size && (attr_flags & XFS_ATTR_DMI) == 0 &&
|
2007-08-16 10:42:07 +02:00
|
|
|
DM_EVENT_ENABLED(ip, DM_EVENT_WRITE)) {
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
if (end_dmi_offset > ip->i_size)
|
|
|
|
end_dmi_offset = ip->i_size;
|
2008-03-06 03:45:58 +01:00
|
|
|
error = XFS_SEND_DATA(mp, DM_EVENT_WRITE, ip,
|
2005-04-17 00:20:36 +02:00
|
|
|
offset, end_dmi_offset - offset,
|
|
|
|
AT_DELAY_FLAG(attr_flags), NULL);
|
|
|
|
if (error)
|
2006-01-15 02:37:08 +01:00
|
|
|
return error;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if (attr_flags & XFS_ATTR_NOLOCK)
|
2005-05-05 22:27:19 +02:00
|
|
|
need_iolock = 0;
|
2006-03-22 02:44:35 +01:00
|
|
|
if (need_iolock) {
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_ilock(ip, XFS_IOLOCK_EXCL);
|
2008-12-03 12:20:39 +01:00
|
|
|
/* wait for the completion of any pending DIOs */
|
|
|
|
xfs_ioend_wait(ip);
|
2006-03-22 02:44:35 +01:00
|
|
|
}
|
2005-05-05 22:27:19 +02:00
|
|
|
|
2007-11-23 06:30:42 +01:00
|
|
|
rounding = max_t(uint, 1 << mp->m_sb.sb_blocklog, PAGE_CACHE_SIZE);
|
2005-04-17 00:20:36 +02:00
|
|
|
ioffset = offset & ~(rounding - 1);
|
2005-06-21 07:47:39 +02:00
|
|
|
|
2008-08-13 08:22:09 +02:00
|
|
|
if (VN_CACHED(VFS_I(ip)) != 0) {
|
2007-11-23 06:30:42 +01:00
|
|
|
xfs_inval_cached_trace(ip, ioffset, -1, ioffset, -1);
|
|
|
|
error = xfs_flushinval_pages(ip, ioffset, -1, FI_REMAPF_LOCKED);
|
2007-05-08 05:49:27 +02:00
|
|
|
if (error)
|
|
|
|
goto out_unlock_iolock;
|
2005-06-21 07:47:39 +02:00
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Need to zero the stuff we're not freeing, on disk.
|
2009-03-29 09:55:42 +02:00
|
|
|
* If it's a realtime file & can't use unwritten extents then we
|
2005-04-17 00:20:36 +02:00
|
|
|
* actually need to zero the extent edges. Otherwise xfs_bunmapi
|
|
|
|
* will take care of it for us.
|
|
|
|
*/
|
2008-03-06 03:44:28 +01:00
|
|
|
if (rt && !xfs_sb_version_hasextflgbit(&mp->m_sb)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
nimap = 1;
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bmapi(NULL, ip, startoffset_fsb,
|
2006-06-09 06:48:12 +02:00
|
|
|
1, 0, NULL, 0, &imap, &nimap, NULL, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto out_unlock_iolock;
|
|
|
|
ASSERT(nimap == 0 || nimap == 1);
|
|
|
|
if (nimap && imap.br_startblock != HOLESTARTBLOCK) {
|
|
|
|
xfs_daddr_t block;
|
|
|
|
|
|
|
|
ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
|
|
|
|
block = imap.br_startblock;
|
|
|
|
mod = do_div(block, mp->m_sb.sb_rextsize);
|
|
|
|
if (mod)
|
|
|
|
startoffset_fsb += mp->m_sb.sb_rextsize - mod;
|
|
|
|
}
|
|
|
|
nimap = 1;
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bmapi(NULL, ip, endoffset_fsb - 1,
|
2006-06-09 06:48:12 +02:00
|
|
|
1, 0, NULL, 0, &imap, &nimap, NULL, NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto out_unlock_iolock;
|
|
|
|
ASSERT(nimap == 0 || nimap == 1);
|
|
|
|
if (nimap && imap.br_startblock != HOLESTARTBLOCK) {
|
|
|
|
ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
|
|
|
|
mod++;
|
|
|
|
if (mod && (mod != mp->m_sb.sb_rextsize))
|
|
|
|
endoffset_fsb -= mod;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if ((done = (endoffset_fsb <= startoffset_fsb)))
|
|
|
|
/*
|
|
|
|
* One contiguous piece to clear
|
|
|
|
*/
|
|
|
|
error = xfs_zero_remaining_bytes(ip, offset, offset + len - 1);
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* Some full blocks, possibly two pieces to clear
|
|
|
|
*/
|
|
|
|
if (offset < XFS_FSB_TO_B(mp, startoffset_fsb))
|
|
|
|
error = xfs_zero_remaining_bytes(ip, offset,
|
|
|
|
XFS_FSB_TO_B(mp, startoffset_fsb) - 1);
|
|
|
|
if (!error &&
|
|
|
|
XFS_FSB_TO_B(mp, endoffset_fsb) < offset + len)
|
|
|
|
error = xfs_zero_remaining_bytes(ip,
|
|
|
|
XFS_FSB_TO_B(mp, endoffset_fsb),
|
|
|
|
offset + len - 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* free file space until done or until there is an error
|
|
|
|
*/
|
|
|
|
resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
|
|
|
|
while (!error && !done) {
|
|
|
|
|
|
|
|
/*
|
2007-07-19 08:28:30 +02:00
|
|
|
* allocate and setup the transaction. Allow this
|
|
|
|
* transaction to dip into the reserve blocks to ensure
|
|
|
|
* the freeing of the space succeeds at ENOSPC.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
|
2007-07-19 08:28:30 +02:00
|
|
|
tp->t_flags |= XFS_TRANS_RESERVE;
|
2005-04-17 00:20:36 +02:00
|
|
|
error = xfs_trans_reserve(tp,
|
|
|
|
resblks,
|
|
|
|
XFS_WRITE_LOG_RES(mp),
|
|
|
|
0,
|
|
|
|
XFS_TRANS_PERM_LOG_RES,
|
|
|
|
XFS_WRITE_LOG_COUNT);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* check for running out of space
|
|
|
|
*/
|
|
|
|
if (error) {
|
|
|
|
/*
|
|
|
|
* Free the transaction structure.
|
|
|
|
*/
|
|
|
|
ASSERT(error == ENOSPC || XFS_FORCED_SHUTDOWN(mp));
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2009-06-08 15:33:32 +02:00
|
|
|
error = xfs_trans_reserve_quota(tp, mp,
|
|
|
|
ip->i_udquot, ip->i_gdquot,
|
|
|
|
resblks, 0, XFS_QMOPT_RES_REGBLKS);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error)
|
|
|
|
goto error1;
|
|
|
|
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* issue the bunmapi() call to free the blocks
|
|
|
|
*/
|
2009-01-15 06:22:07 +01:00
|
|
|
xfs_bmap_init(&free_list, &firstfsb);
|
2007-10-11 09:34:33 +02:00
|
|
|
error = xfs_bunmapi(tp, ip, startoffset_fsb,
|
2005-04-17 00:20:36 +02:00
|
|
|
endoffset_fsb - startoffset_fsb,
|
2006-06-09 06:48:12 +02:00
|
|
|
0, 2, &firstfsb, &free_list, NULL, &done);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* complete the transaction
|
|
|
|
*/
|
2007-02-10 08:37:16 +01:00
|
|
|
error = xfs_bmap_finish(&tp, &free_list, &committed);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (error) {
|
|
|
|
goto error0;
|
|
|
|
}
|
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
|
2005-04-17 00:20:36 +02:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock_iolock:
|
|
|
|
if (need_iolock)
|
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
|
|
|
|
return error;
|
|
|
|
|
|
|
|
error0:
|
|
|
|
xfs_bmap_cancel(&free_list);
|
|
|
|
error1:
|
|
|
|
xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
|
|
|
|
xfs_iunlock(ip, need_iolock ? (XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL) :
|
|
|
|
XFS_ILOCK_EXCL);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* xfs_change_file_space()
|
|
|
|
* This routine allocates or frees disk space for the given file.
|
|
|
|
* The user specified parameters are checked for alignment and size
|
|
|
|
* limitations.
|
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on success
|
|
|
|
* errno on error
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_change_file_space(
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_inode_t *ip,
|
2005-04-17 00:20:36 +02:00
|
|
|
int cmd,
|
|
|
|
xfs_flock64_t *bf,
|
|
|
|
xfs_off_t offset,
|
|
|
|
int attr_flags)
|
|
|
|
{
|
2007-08-28 08:12:30 +02:00
|
|
|
xfs_mount_t *mp = ip->i_mount;
|
2005-04-17 00:20:36 +02:00
|
|
|
int clrprealloc;
|
|
|
|
int error;
|
|
|
|
xfs_fsize_t fsize;
|
|
|
|
int setprealloc;
|
|
|
|
xfs_off_t startoffset;
|
|
|
|
xfs_off_t llen;
|
|
|
|
xfs_trans_t *tp;
|
2008-07-18 09:13:28 +02:00
|
|
|
struct iattr iattr;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-02-07 06:42:19 +01:00
|
|
|
xfs_itrace_entry(ip);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-08-28 08:12:30 +02:00
|
|
|
if (!S_ISREG(ip->i_d.di_mode))
|
2005-04-17 00:20:36 +02:00
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
|
|
|
|
switch (bf->l_whence) {
|
|
|
|
case 0: /*SEEK_SET*/
|
|
|
|
break;
|
|
|
|
case 1: /*SEEK_CUR*/
|
|
|
|
bf->l_start += offset;
|
|
|
|
break;
|
|
|
|
case 2: /*SEEK_END*/
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
bf->l_start += ip->i_size;
|
2005-04-17 00:20:36 +02:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
llen = bf->l_len > 0 ? bf->l_len - 1 : bf->l_len;
|
|
|
|
|
|
|
|
if ( (bf->l_start < 0)
|
|
|
|
|| (bf->l_start > XFS_MAXIOFFSET(mp))
|
|
|
|
|| (bf->l_start + llen < 0)
|
|
|
|
|| (bf->l_start + llen > XFS_MAXIOFFSET(mp)))
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
|
|
|
|
bf->l_whence = 0;
|
|
|
|
|
|
|
|
startoffset = bf->l_start;
|
[XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
The problem that has been addressed is that of synchronising updates of
the file size with writes that extend a file. Without the fix the update
of a file's size, as a result of a write beyond eof, is independent of
when the cached data is flushed to disk. Often the file size update would
be written to the filesystem log before the data is flushed to disk. When
a system crashes between these two events and the filesystem log is
replayed on mount the file's size will be set but since the contents never
made it to disk the file is full of holes. If some of the cached data was
flushed to disk then it may just be a section of the file at the end that
has holes.
There are existing fixes to help alleviate this problem, particularly in
the case where a file has been truncated, that force cached data to be
flushed to disk when the file is closed. If the system crashes while the
file(s) are still open then this flushing will never occur.
The fix that we have implemented is to introduce a second file size,
called the in-memory file size, that represents the current file size as
viewed by the user. The existing file size, called the on-disk file size,
is the one that get's written to the filesystem log and we only update it
when it is safe to do so. When we write to a file beyond eof we only
update the in- memory file size in the write operation. Later when the I/O
operation, that flushes the cached data to disk completes, an I/O
completion routine will update the on-disk file size. The on-disk file
size will be updated to the maximum offset of the I/O or to the value of
the in-memory file size if the I/O includes eof.
SGI-PV: 958522
SGI-Modid: xfs-linux-melb:xfs-kern:28322a
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-08 05:49:46 +02:00
|
|
|
fsize = ip->i_size;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* XFS_IOC_RESVSP and XFS_IOC_UNRESVSP will reserve or unreserve
|
|
|
|
* file space.
|
|
|
|
* These calls do NOT zero the data space allocated to the file,
|
|
|
|
* nor do they change the file size.
|
|
|
|
*
|
|
|
|
* XFS_IOC_ALLOCSP and XFS_IOC_FREESP will allocate and free file
|
|
|
|
* space.
|
|
|
|
* These calls cause the new file data to be zeroed and the file
|
|
|
|
* size to be changed.
|
|
|
|
*/
|
|
|
|
setprealloc = clrprealloc = 0;
|
|
|
|
|
|
|
|
switch (cmd) {
|
|
|
|
case XFS_IOC_RESVSP:
|
|
|
|
case XFS_IOC_RESVSP64:
|
|
|
|
error = xfs_alloc_file_space(ip, startoffset, bf->l_len,
|
|
|
|
1, attr_flags);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
setprealloc = 1;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case XFS_IOC_UNRESVSP:
|
|
|
|
case XFS_IOC_UNRESVSP64:
|
|
|
|
if ((error = xfs_free_file_space(ip, startoffset, bf->l_len,
|
|
|
|
attr_flags)))
|
|
|
|
return error;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case XFS_IOC_ALLOCSP:
|
|
|
|
case XFS_IOC_ALLOCSP64:
|
|
|
|
case XFS_IOC_FREESP:
|
|
|
|
case XFS_IOC_FREESP64:
|
|
|
|
if (startoffset > fsize) {
|
|
|
|
error = xfs_alloc_file_space(ip, fsize,
|
|
|
|
startoffset - fsize, 0, attr_flags);
|
|
|
|
if (error)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
iattr.ia_valid = ATTR_SIZE;
|
|
|
|
iattr.ia_size = startoffset;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-10-30 08:27:48 +01:00
|
|
|
error = xfs_setattr(ip, &iattr, attr_flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
|
|
|
|
clrprealloc = 1;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
ASSERT(0);
|
|
|
|
return XFS_ERROR(EINVAL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* update the inode timestamp, mode, and prealloc flag bits
|
|
|
|
*/
|
|
|
|
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
|
|
|
|
|
|
|
|
if ((error = xfs_trans_reserve(tp, 0, XFS_WRITEID_LOG_RES(mp),
|
|
|
|
0, 0, 0))) {
|
|
|
|
/* ASSERT(0); */
|
|
|
|
xfs_trans_cancel(tp, 0);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
|
|
|
|
xfs_trans_ihold(tp, ip);
|
|
|
|
|
2008-07-18 09:13:28 +02:00
|
|
|
if ((attr_flags & XFS_ATTR_DMI) == 0) {
|
2005-04-17 00:20:36 +02:00
|
|
|
ip->i_d.di_mode &= ~S_ISUID;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Note that we don't have to worry about mandatory
|
|
|
|
* file locking being disabled here because we only
|
|
|
|
* clear the S_ISGID bit if the Group execute bit is
|
|
|
|
* on, but if it was on then mandatory locking wouldn't
|
|
|
|
* have been enabled.
|
|
|
|
*/
|
|
|
|
if (ip->i_d.di_mode & S_IXGRP)
|
|
|
|
ip->i_d.di_mode &= ~S_ISGID;
|
|
|
|
|
|
|
|
xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
|
|
|
|
}
|
|
|
|
if (setprealloc)
|
|
|
|
ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
|
|
|
|
else if (clrprealloc)
|
|
|
|
ip->i_d.di_flags &= ~XFS_DIFLAG_PREALLOC;
|
|
|
|
|
|
|
|
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
|
|
|
|
xfs_trans_set_sync(tp);
|
|
|
|
|
2007-05-08 05:48:42 +02:00
|
|
|
error = xfs_trans_commit(tp, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|