*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
*/
/* Portions Copyright 2007 Jeremy Teo */
#include <sys/sid.h>
#include <sys/mode.h>
#include "fs/fs_subr.h"
+#include <sys/zfs_ctldir.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_sa.h>
#include <sys/zfs_vnops.h>
}
EXPORT_SYMBOL(zfs_close);
+#if defined(SEEK_HOLE) && defined(SEEK_DATA)
+/*
+ * Lseek support for finding holes (cmd == SEEK_HOLE) and
+ * data (cmd == SEEK_DATA). "off" is an in/out parameter.
+ */
+static int
+zfs_holey_common(struct inode *ip, int cmd, loff_t *off)
+{
+ znode_t *zp = ITOZ(ip);
+ uint64_t noff = (uint64_t)*off; /* new offset */
+ uint64_t file_sz;
+ int error;
+ boolean_t hole;
+
+ file_sz = zp->z_size;
+ if (noff >= file_sz) {
+ return (ENXIO);
+ }
+
+ if (cmd == SEEK_HOLE)
+ hole = B_TRUE;
+ else
+ hole = B_FALSE;
+
+ error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff);
+
+ /* end of file? */
+ if ((error == ESRCH) || (noff > file_sz)) {
+ /*
+ * Handle the virtual hole at the end of file.
+ */
+ if (hole) {
+ *off = file_sz;
+ return (0);
+ }
+ return (ENXIO);
+ }
+
+ if (noff < *off)
+ return (error);
+ *off = noff;
+ return (error);
+}
+
+int
+zfs_holey(struct inode *ip, int cmd, loff_t *off)
+{
+ znode_t *zp = ITOZ(ip);
+ zfs_sb_t *zsb = ITOZSB(ip);
+ int error;
+
+ ZFS_ENTER(zsb);
+ ZFS_VERIFY_ZP(zp);
+
+ error = zfs_holey_common(ip, cmd, off);
+
+ ZFS_EXIT(zsb);
+ return (error);
+}
+EXPORT_SYMBOL(zfs_holey);
+#endif /* SEEK_HOLE && SEEK_DATA */
+
#if defined(_KERNEL)
/*
* When a file is memory mapped, we must keep the IO data synchronized
{
ASSERT(atomic_read(&ip->i_count) > 0);
if (atomic_read(&ip->i_count) == 1)
- taskq_dispatch(taskq, (task_func_t *)iput, ip, TQ_SLEEP);
+ taskq_dispatch(taskq, (task_func_t *)iput, ip, TQ_PUSHPAGE);
else
iput(ip);
}
return (ENOENT);
}
- zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
+ zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_PUSHPAGE);
zgd->zgd_zilog = zsb->z_log;
zgd->zgd_private = zp;
&xattr_obj, sizeof (xattr_obj));
if (error == 0 && xattr_obj) {
error = zfs_zget(zsb, xattr_obj, &xzp);
- ASSERT3U(error, ==, 0);
+ ASSERT0(error);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
}
out:
zfs_dirent_unlock(dl);
+ zfs_inode_update(dzp);
+ zfs_inode_update(zp);
iput(ip);
if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zilog, 0);
- zfs_inode_update(dzp);
- zfs_inode_update(zp);
ZFS_EXIT(zsb);
return (error);
}
goto update;
}
+ /*
+ * Allow multiple entries provided the first entry is
+ * the object id. Non-zpl consumers may safely make
+ * use of the additional space.
+ *
+ * XXX: This should be a feature flag for compatibility
+ */
if (zap.za_integer_length != 8 ||
- zap.za_num_integers != 1) {
+ zap.za_num_integers == 0) {
cmn_err(CE_WARN, "zap_readdir: bad directory "
- "entry, obj = %lld, offset = %lld\n",
+ "entry, obj = %lld, offset = %lld, "
+ "length = %d, num = %lld\n",
(u_longlong_t)zp->z_id,
- (u_longlong_t)*pos);
+ (u_longlong_t)*pos,
+ zap.za_integer_length,
+ (u_longlong_t)zap.za_num_integers);
error = ENXIO;
goto update;
}
dmu_prefetch(os, objnum, 0, 0);
}
- if (*pos >= 2) {
+ if (*pos > 2 || (*pos == 2 && !zfs_show_ctldir(zp))) {
zap_cursor_advance(&zc);
*pos = zap_cursor_serialize(&zc);
} else {
zp->z_mode = new_mode;
ASSERT3P(aclp, !=, NULL);
err = zfs_aclset_common(zp, aclp, cr, tx);
- ASSERT3U(err, ==, 0);
+ ASSERT0(err);
if (zp->z_acl_cached)
zfs_acl_free(zp->z_acl_cached);
zp->z_acl_cached = aclp;
error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
(void *)&szp->z_pflags, sizeof (uint64_t), tx);
- ASSERT3U(error, ==, 0);
+ ASSERT0(error);
error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL);
if (error == 0) {
zfs_sb_t *zsb = ITOZSB(ip);
loff_t offset;
loff_t pgoff;
- unsigned int pglen;
+ unsigned int pglen;
+ rl_t *rl;
dmu_tx_t *tx;
caddr_t va;
int err = 0;
uint64_t mtime[2], ctime[2];
sa_bulk_attr_t bulk[3];
int cnt = 0;
+ int sync;
+ ZFS_ENTER(zsb);
+ ZFS_VERIFY_ZP(zp);
ASSERT(PageLocked(pp));
/* Page is beyond end of file */
if (pgoff >= offset) {
unlock_page(pp);
+ ZFS_EXIT(zsb);
return (0);
}
set_page_writeback(pp);
unlock_page(pp);
+ rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
tx = dmu_tx_create(zsb->z_os);
- dmu_tx_callback_register(tx, zfs_putpage_commit_cb, pp);
+ sync = ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
+ (wbc->sync_mode == WB_SYNC_ALL));
+ if (!sync)
+ dmu_tx_callback_register(tx, zfs_putpage_commit_cb, pp);
dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen);
if (err == ERESTART)
dmu_tx_wait(tx);
+ /* Will call all registered commit callbacks */
dmu_tx_abort(tx);
+
+ /*
+ * For the synchronous case the commit callback must be
+ * explicitly called because there is no registered callback.
+ */
+ if (sync)
+ zfs_putpage_commit_cb(pp, ECANCELED);
+
+ zfs_range_unlock(rl);
+ ZFS_EXIT(zsb);
return (err);
}
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8);
- zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
- zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0);
+ /* Preserve the mtime and ctime provided by the inode */
+ ZFS_TIME_ENCODE(&ip->i_mtime, mtime);
+ ZFS_TIME_ENCODE(&ip->i_ctime, ctime);
+ zp->z_atime_dirty = 0;
+ zp->z_seq++;
+
+ err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx);
+
+ zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0);
dmu_tx_commit(tx);
- ASSERT3S(err, ==, 0);
- if ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
- (wbc->sync_mode == WB_SYNC_ALL))
+ zfs_range_unlock(rl);
+
+ if (sync) {
zil_commit(zsb->z_log, zp->z_id);
+ zfs_putpage_commit_cb(pp, err);
+ }
+ ZFS_EXIT(zsb);
return (err);
}
+/*
+ * Update the system attributes when the inode has been dirtied. For the
+ * moment we're conservative and only update the atime, mtime, and ctime.
+ */
+int
+zfs_dirty_inode(struct inode *ip, int flags)
+{
+ znode_t *zp = ITOZ(ip);
+ zfs_sb_t *zsb = ITOZSB(ip);
+ dmu_tx_t *tx;
+ uint64_t atime[2], mtime[2], ctime[2];
+ sa_bulk_attr_t bulk[3];
+ int error;
+ int cnt = 0;
+
+ ZFS_ENTER(zsb);
+ ZFS_VERIFY_ZP(zp);
+
+ tx = dmu_tx_create(zsb->z_os);
+
+ dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
+ zfs_sa_upgrade_txholds(tx, zp);
+
+ error = dmu_tx_assign(tx, TXG_WAIT);
+ if (error) {
+ dmu_tx_abort(tx);
+ goto out;
+ }
+
+ mutex_enter(&zp->z_lock);
+ SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
+ SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
+ SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
+
+ /* Preserve the mtime and ctime provided by the inode */
+ ZFS_TIME_ENCODE(&ip->i_atime, atime);
+ ZFS_TIME_ENCODE(&ip->i_mtime, mtime);
+ ZFS_TIME_ENCODE(&ip->i_ctime, ctime);
+ zp->z_atime_dirty = 0;
+
+ error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx);
+ mutex_exit(&zp->z_lock);
+
+ dmu_tx_commit(tx);
+out:
+ ZFS_EXIT(zsb);
+ return (error);
+}
+EXPORT_SYMBOL(zfs_dirty_inode);
+
/*ARGSUSED*/
void
zfs_inactive(struct inode *ip)
zfs_sb_t *zsb = ITOZSB(ip);
int error;
-#ifdef HAVE_SNAPSHOT
- /* Early return for snapshot inode? */
-#endif /* HAVE_SNAPSHOT */
+ if (zfsctl_is_node(ip)) {
+ zfsctl_inode_inactive(ip);
+ return;
+ }
rw_enter(&zsb->z_teardown_inactive_lock, RW_READER);
if (zp->z_sa_hdl == NULL) {
return (EINVAL);
}
+ /*
+ * Permissions aren't checked on Solaris because on this OS
+ * zfs_space() can only be called with an opened file handle.
+ * On Linux we can get here through truncate_range() which
+ * operates directly on inodes, so we need to check access rights.
+ */
+ if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) {
+ ZFS_EXIT(zsb);
+ return (error);
+ }
+
off = bfp->l_start;
len = bfp->l_len; /* 0 means from off to end of file */