ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct file *file,
const char *buf,
size_t size,
loff_t *offset,
int ioflags,
cred_t *credp)
{
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_iocore_t *io;
vnode_t *vp;
int iolock;
int eventsent = 0;
vrwlock_t locktype;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
if (size == 0)
return 0;
io = &xip->i_iocore;
mp = io->io_mount;
fs_check_frozen(vp->v_vfsp, SB_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(xip->i_mount)) {
return -EIO;
}
if (unlikely(ioflags & IO_ISDIRECT)) {
if (((__psint_t)buf & BBMASK) ||
(*offset & mp->m_blockmask) ||
(size & mp->m_blockmask)) {
return XFS_ERROR(-EINVAL);
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
if (ioflags & IO_ISLOCKED)
iolock = 0;
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return -EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
dmflags, &locktype);
if (error) {
if (iolock) xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflags & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
}
if ((ssize_t) size < 0) {
ret = -EINVAL;
goto error;
}
if (!access_ok(VERIFY_READ, buf, size)) {
ret = -EINVAL;
goto error;
}
retry:
if (unlikely(ioflags & IO_ISDIRECT)) {
xfs_inval_cached_pages(vp, io, *offset, 1, 1);
xfs_rw_enter_trace(XFS_DIOWR_ENTER,
io, buf, size, *offset, ioflags);
ret = do_generic_direct_write(file, buf, size, offset);
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER,
io, buf, size, *offset, ioflags);
ret = do_generic_file_write(file, buf, size, offset);
}
if (unlikely(ioflags & IO_INVIS)) {
/* generic_file_write updates the mtime/ctime but we need
* to undo that because this I/O was supposed to be
* invisible.
*/
struct inode *inode = LINVFS_GET_IP(vp);
inode->i_mtime = xip->i_d.di_mtime.t_sec;
inode->i_ctime = xip->i_d.di_ctime.t_sec;
} else {
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
}
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
return -error;
xfs_rwlock(bdp, locktype);
*offset = xip->i_d.di_size;
goto retry;
}
error:
if (ret <= 0) {
if (iolock)
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
struct inode *inode = LINVFS_GET_IP(vp);
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
mark_inode_dirty_sync(inode);
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(file->f_dentry->d_inode)) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC)
&& !(xip->i_update_size)) {
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
xfs_inode_log_item_t *iip;
xfs_lsn_t lsn;
iip = xip->i_itemp;
if (iip && iip->ili_last_lsn) {
lsn = iip->ili_last_lsn;
xfs_log_force(mp, lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, NULL);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
}
} /* (ioflags & O_SYNC) */
/*
* If we are coming from an nfsd thread then insert into the
* reference cache.
*/
if (!strcmp(current->comm, "nfsd"))
xfs_refcache_insert(xip);
/* Drop lock this way - the old refcache release is in here */
if (iolock)
xfs_rwunlock(bdp, locktype);
return(ret);
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
uio_t *uio,
int ioflag,
cred_t *credp)
{
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret = 0;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_fsize_t size;
xfs_iocore_t *io;
xfs_vnode_t *vp;
int iolock;
//int eventsent = 0;
vrwlock_t locktype;
xfs_off_t offset_c;
xfs_off_t *offset;
xfs_off_t pos;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
xip = XFS_BHVTOI(bdp);
io = &xip->i_iocore;
mp = io->io_mount;
if (XFS_FORCED_SHUTDOWN(xip->i_mount)) {
return EIO;
}
size = uio->uio_resid;
pos = offset_c = uio->uio_offset;
offset = &offset_c;
if (unlikely(ioflag & IO_ISDIRECT)) {
if (((__psint_t)buf & BBMASK) ||
(*offset & mp->m_blockmask) ||
(size & mp->m_blockmask)) {
return EINVAL;
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
if (io->io_flags & XFS_IOCORE_RT)
return EINVAL;
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (ioflag & O_APPEND)
*offset = isize;
//start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
#ifdef RMC
/* probably be a long time before if ever that we do dmapi */
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
int dmflags = FILP_DELAY_FLAG(file) | DM_SEM_FLAG_RD(ioflags);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
dmflags, &locktype);
if (error) {
if (iolock) xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
#endif
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflag & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
#if 0
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (likely(!error))
error = -remove_suid(file->f_dentry);
if (unlikely(error)) {
xfs_iunlock(xip, iolock);
goto out_unlock_mutex;
}
}
#endif
//retry:
if (unlikely(ioflag & IO_ISDIRECT)) {
#ifdef RMC
xfs_off_t pos = *offset;
struct address_space *mapping = file->f_dentry->d_inode->i_mapping;
struct inode *inode = mapping->host;
ret = precheck_file_write(file, inode, &size, &pos);
if (ret || size == 0)
goto error;
xfs_inval_cached_pages(vp, io, pos, 1, 1);
inode->i_ctime = inode->i_mtime = CURRENT_TIME;
/* mark_inode_dirty_sync(inode); - we do this later */
xfs_rw_enter_trace(XFS_DIOWR_ENTER, io, buf, size, pos, ioflags);
ret = generic_file_direct_IO(WRITE, file, (char *)buf, size, pos);
xfs_inval_cached_pages(vp, io, pos, 1, 1);
if (ret > 0)
*offset += ret;
#endif
} else {
xfs_rw_enter_trace(XFS_WRITE_ENTER, io, buf, size, *offset, ioflags);
ret = xfs_write_file(xip,uio,ioflag);
}
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
//error:
if (ret <= 0) {
if (iolock)
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
printf("xfs_write look at doing more here %s:%d\n",__FILE__,__LINE__);
#ifdef RMC
struct inode *inode = LINVFS_GET_IP(vp);
i_size_write(inode, *offset);
mark_inode_dirty_sync(inode);
#endif
xip->i_d.di_size = *offset;
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
/* Handle various SYNC-type writes */
#if 0
// if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
#endif
if (ioflag & IO_SYNC) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
!(xip->i_update_size)) {
xfs_inode_log_item_t *iip = xip->i_itemp;
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
if (iip && iip->ili_last_lsn) {
xfs_log_force(mp, iip->ili_last_lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, NULL);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
if (error)
goto out_unlock_internal;
}
xfs_rwunlock(bdp, locktype);
return ret;
} /* (ioflags & O_SYNC) */
out_unlock_internal:
xfs_rwunlock(bdp, locktype);
#if 0
out_unlock_mutex:
if (need_i_mutex)
mutex_unlock(&inode->i_mutex);
#endif
//out_nounlocks:
return -error;
}
/*
* Syssgi interface for swapext
*/
int
xfs_swapext(
xfs_swapext_t __user *sxp)
{
xfs_swapext_t sx;
xfs_inode_t *ip=NULL, *tip=NULL, *ips[2];
xfs_trans_t *tp;
xfs_mount_t *mp;
xfs_bstat_t *sbp;
struct file *fp = NULL, *tfp = NULL;
vnode_t *vp, *tvp;
bhv_desc_t *bdp, *tbdp;
vn_bhv_head_t *bhp, *tbhp;
uint lock_flags=0;
int ilf_fields, tilf_fields;
int error = 0;
xfs_ifork_t tempif, *ifp, *tifp;
__uint64_t tmp;
int aforkblks = 0;
int taforkblks = 0;
int locked = 0;
if (copy_from_user(&sx, sxp, sizeof(sx)))
return XFS_ERROR(EFAULT);
/* Pull information for the target fd */
if (((fp = fget((int)sx.sx_fdtarget)) == NULL) ||
((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
bhp = VN_BHV_HEAD(vp);
bdp = vn_bhv_lookup(bhp, &xfs_vnodeops);
if (bdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
ip = XFS_BHVTOI(bdp);
}
if (((tfp = fget((int)sx.sx_fdtmp)) == NULL) ||
((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
tbhp = VN_BHV_HEAD(tvp);
tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops);
if (tbdp == NULL) {
error = XFS_ERROR(EBADF);
goto error0;
} else {
tip = XFS_BHVTOI(tbdp);
}
if (ip->i_mount != tip->i_mount) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (ip->i_ino == tip->i_ino) {
error = XFS_ERROR(EINVAL);
goto error0;
}
mp = ip->i_mount;
sbp = &sx.sx_stat;
if (XFS_FORCED_SHUTDOWN(mp)) {
error = XFS_ERROR(EIO);
goto error0;
}
locked = 1;
/* Lock in i_ino order */
if (ip->i_ino < tip->i_ino) {
ips[0] = ip;
ips[1] = tip;
} else {
ips[0] = tip;
ips[1] = ip;
}
lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
xfs_lock_inodes(ips, 2, 0, lock_flags);
/* Check permissions */
error = xfs_iaccess(ip, S_IWUSR, NULL);
if (error)
goto error0;
error = xfs_iaccess(tip, S_IWUSR, NULL);
if (error)
goto error0;
/* Verify that both files have the same format */
if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify both files are either real-time or non-realtime */
if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
(tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Should never get a local format */
if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
error = XFS_ERROR(EINVAL);
goto error0;
}
if (VN_CACHED(tvp) != 0)
xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore),
(xfs_off_t)0, 0, 0);
/* Verify O_DIRECT for ftmp */
if (VN_CACHED(tvp) != 0) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/* Verify all data are being swapped */
if (sx.sx_offset != 0 ||
sx.sx_length != ip->i_d.di_size ||
sx.sx_length != tip->i_d.di_size) {
error = XFS_ERROR(EFAULT);
goto error0;
}
/*
* If the target has extended attributes, the tmp file
* must also in order to ensure the correct data fork
* format.
*/
if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
error = XFS_ERROR(EINVAL);
goto error0;
}
/*
* Compare the current change & modify times with that
* passed in. If they differ, we abort this swap.
* This is the mechanism used to ensure the calling
* process that the file was not changed out from
* under it.
*/
if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
(sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
(sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
(sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
/* We need to fail if the file is memory mapped. Once we have tossed
* all existing pages, the page fault will have no option
* but to go to the filesystem for pages. By making the page fault call
* VOP_READ (or write in the case of autogrow) they block on the iolock
* until we have switched the extents.
*/
if (VN_MAPPED(vp)) {
error = XFS_ERROR(EBUSY);
goto error0;
}
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs_iunlock(tip, XFS_ILOCK_EXCL);
/*
* There is a race condition here since we gave up the
* ilock. However, the data fork will not change since
* we have the iolock (locked for truncation too) so we
* are safe. We don't really care if non-io related
* fields change.
*/
VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
if ((error = xfs_trans_reserve(tp, 0,
XFS_ICHANGE_LOG_RES(mp), 0,
0, 0))) {
xfs_iunlock(ip, XFS_IOLOCK_EXCL);
xfs_iunlock(tip, XFS_IOLOCK_EXCL);
xfs_trans_cancel(tp, 0);
return error;
}
xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
/*
* Count the number of extended attribute blocks
*/
if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
(ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
(tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
&taforkblks);
if (error) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
xfs_trans_cancel(tp, 0);
return error;
}
}
/*
* Swap the data forks of the inodes
*/
ifp = &ip->i_df;
tifp = &tip->i_df;
tempif = *ifp; /* struct copy */
*ifp = *tifp; /* struct copy */
*tifp = tempif; /* struct copy */
/*
* Fix the on-disk inode values
*/
tmp = (__uint64_t)ip->i_d.di_nblocks;
ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
tmp = (__uint64_t) ip->i_d.di_nextents;
ip->i_d.di_nextents = tip->i_d.di_nextents;
tip->i_d.di_nextents = tmp;
tmp = (__uint64_t) ip->i_d.di_format;
ip->i_d.di_format = tip->i_d.di_format;
tip->i_d.di_format = tmp;
ilf_fields = XFS_ILOG_CORE;
switch(ip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
ifp->if_u1.if_extents =
ifp->if_u2.if_inline_ext;
}
ilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
ilf_fields |= XFS_ILOG_DBROOT;
break;
}
tilf_fields = XFS_ILOG_CORE;
switch(tip->i_d.di_format) {
case XFS_DINODE_FMT_EXTENTS:
/* If the extents fit in the inode, fix the
* pointer. Otherwise it's already NULL or
* pointing to the extent.
*/
if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
tifp->if_u1.if_extents =
tifp->if_u2.if_inline_ext;
}
tilf_fields |= XFS_ILOG_DEXT;
break;
case XFS_DINODE_FMT_BTREE:
tilf_fields |= XFS_ILOG_DBROOT;
break;
}
/*
* Increment vnode ref counts since xfs_trans_commit &
* xfs_trans_cancel will both unlock the inodes and
* decrement the associated ref counts.
*/
VN_HOLD(vp);
VN_HOLD(tvp);
xfs_trans_ijoin(tp, ip, lock_flags);
xfs_trans_ijoin(tp, tip, lock_flags);
xfs_trans_log_inode(tp, ip, ilf_fields);
xfs_trans_log_inode(tp, tip, tilf_fields);
/*
* If this is a synchronous mount, make sure that the
* transaction goes to disk before returning to the user.
*/
if (mp->m_flags & XFS_MOUNT_WSYNC) {
xfs_trans_set_sync(tp);
}
error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
fput(fp);
fput(tfp);
return error;
error0:
if (locked) {
xfs_iunlock(ip, lock_flags);
xfs_iunlock(tip, lock_flags);
}
if (fp != NULL) fput(fp);
if (tfp != NULL) fput(tfp);
return error;
}
ssize_t /* bytes written, or (-) error */
xfs_write(
bhv_desc_t *bdp,
struct kiocb *iocb,
const struct iovec *iovp,
unsigned int segs,
loff_t *offset,
int ioflags,
cred_t *credp)
{
struct file *file = iocb->ki_filp;
size_t size = 0;
xfs_inode_t *xip;
xfs_mount_t *mp;
ssize_t ret;
int error = 0;
xfs_fsize_t isize, new_size;
xfs_fsize_t n, limit;
xfs_iocore_t *io;
vnode_t *vp;
unsigned long seg;
int iolock;
int eventsent = 0;
vrwlock_t locktype;
XFS_STATS_INC(xs_write_calls);
vp = BHV_TO_VNODE(bdp);
vn_trace_entry(vp, "xfs_write", (inst_t *)__return_address);
xip = XFS_BHVTOI(bdp);
/* START copy & waste from filemap.c */
for (seg = 0; seg < segs; seg++) {
const struct iovec *iv = &iovp[seg];
/*
* If any segment has a negative length, or the cumulative
* length ever wraps negative then return -EINVAL.
*/
size += iv->iov_len;
if (unlikely((ssize_t)(size|iv->iov_len) < 0))
return XFS_ERROR(-EINVAL);
}
/* END copy & waste from filemap.c */
if (size == 0)
return 0;
io = &(xip->i_iocore);
mp = io->io_mount;
xfs_check_frozen(mp, bdp, XFS_FREEZE_WRITE);
if (XFS_FORCED_SHUTDOWN(mp)) {
return -EIO;
}
if (ioflags & IO_ISDIRECT) {
pb_target_t *target =
(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
if ((*offset & target->pbr_smask) ||
(size & target->pbr_smask)) {
return XFS_ERROR(-EINVAL);
}
iolock = XFS_IOLOCK_SHARED;
locktype = VRWLOCK_WRITE_DIRECT;
} else {
iolock = XFS_IOLOCK_EXCL;
locktype = VRWLOCK_WRITE;
}
xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
isize = xip->i_d.di_size;
limit = XFS_MAXIOFFSET(mp);
if (file->f_flags & O_APPEND)
*offset = isize;
start:
n = limit - *offset;
if (n <= 0) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return -EFBIG;
}
if (n < size)
size = n;
new_size = *offset + size;
if (new_size > isize) {
io->io_new_size = new_size;
}
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = *offset;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
*offset, size,
FILP_DELAY_FLAG(file), &locktype);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reaquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) &&
savedsize != xip->i_d.di_size) {
*offset = isize = xip->i_d.di_size;
goto start;
}
}
/*
* On Linux, generic_file_write updates the times even if
* no data is copied in so long as the write had a size.
*
* We must update xfs' times since revalidate will overcopy xfs.
*/
if (size && !(ioflags & IO_INVIS))
xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
/*
* If the offset is beyond the size of the file, we have a couple
* of things to do. First, if there is already space allocated
* we need to either create holes or zero the disk or ...
*
* If there is a page where the previous size lands, we need
* to zero it out up to the new size.
*/
if (!(ioflags & IO_ISDIRECT) && (*offset > isize && isize)) {
error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, *offset,
isize, *offset + size);
if (error) {
xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
return(-error);
}
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
/*
* If we're writing the file then make sure to clear the
* setuid and setgid bits if the process is not being run
* by root. This keeps people from modifying setuid and
* setgid binaries.
*/
if (((xip->i_d.di_mode & S_ISUID) ||
((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
(S_ISGID | S_IXGRP))) &&
!capable(CAP_FSETID)) {
error = xfs_write_clear_setuid(xip);
if (error) {
xfs_iunlock(xip, iolock);
return -error;
}
}
retry:
if (ioflags & IO_ISDIRECT) {
xfs_inval_cached_pages(vp, &xip->i_iocore, *offset, 1, 1);
}
ret = generic_file_aio_write_nolock(iocb, iovp, segs, offset);
if ((ret == -ENOSPC) &&
DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
!(ioflags & IO_INVIS)) {
xfs_rwunlock(bdp, locktype);
error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
0, 0, 0); /* Delay flag intentionally unused */
if (error)
return -error;
xfs_rwlock(bdp, locktype);
*offset = xip->i_d.di_size;
goto retry;
}
if (*offset > xip->i_d.di_size) {
xfs_ilock(xip, XFS_ILOCK_EXCL);
if (*offset > xip->i_d.di_size) {
struct inode *inode = LINVFS_GET_IP(vp);
xip->i_d.di_size = *offset;
i_size_write(inode, *offset);
xip->i_update_core = 1;
xip->i_update_size = 1;
}
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
if (ret <= 0) {
xfs_rwunlock(bdp, locktype);
return ret;
}
XFS_STATS_ADD(xs_write_bytes, ret);
/* Handle various SYNC-type writes */
if ((file->f_flags & O_SYNC) || IS_SYNC(file->f_dentry->d_inode)) {
/*
* If we're treating this as O_DSYNC and we have not updated the
* size, force the log.
*/
if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC)
&& !(xip->i_update_size)) {
/*
* If an allocation transaction occurred
* without extending the size, then we have to force
* the log up the proper point to ensure that the
* allocation is permanent. We can't count on
* the fact that buffered writes lock out direct I/O
* writes - the direct I/O write could have extended
* the size nontransactionally, then finished before
* we started. xfs_write_file will think that the file
* didn't grow but the update isn't safe unless the
* size change is logged.
*
* Force the log if we've committed a transaction
* against the inode or if someone else has and
* the commit record hasn't gone to disk (e.g.
* the inode is pinned). This guarantees that
* all changes affecting the inode are permanent
* when we return.
*/
xfs_inode_log_item_t *iip;
xfs_lsn_t lsn;
iip = xip->i_itemp;
if (iip && iip->ili_last_lsn) {
lsn = iip->ili_last_lsn;
xfs_log_force(mp, lsn,
XFS_LOG_FORCE | XFS_LOG_SYNC);
} else if (xfs_ipincount(xip) > 0) {
xfs_log_force(mp, (xfs_lsn_t)0,
XFS_LOG_FORCE | XFS_LOG_SYNC);
}
} else {
xfs_trans_t *tp;
/*
* O_SYNC or O_DSYNC _with_ a size update are handled
* the same way.
*
* If the write was synchronous then we need to make
* sure that the inode modification time is permanent.
* We'll have updated the timestamp above, so here
* we use a synchronous transaction to log the inode.
* It's not fast, but it's necessary.
*
* If this a dsync write and the size got changed
* non-transactionally, then we need to ensure that
* the size change gets logged in a synchronous
* transaction.
*/
tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
if ((error = xfs_trans_reserve(tp, 0,
XFS_SWRITE_LOG_RES(mp),
0, 0, 0))) {
/* Transaction reserve failed */
xfs_trans_cancel(tp, 0);
} else {
/* Transaction reserve successful */
xfs_ilock(xip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
xfs_trans_ihold(tp, xip);
xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
xfs_trans_set_sync(tp);
error = xfs_trans_commit(tp, 0, (xfs_lsn_t)0);
xfs_iunlock(xip, XFS_ILOCK_EXCL);
}
}
} /* (ioflags & O_SYNC) */
xfs_rwunlock(bdp, locktype);
return(ret);
}
