MFC r268464: MFV r268452:

Explicitly mark file removal transactions as "presumed to result
in a net free of space" so they will not fail with ENOSPC.

Illumos issue:	4950 files sometimes can't be removed from a full
		filesystem

7 files changed, 66 insertions, 21 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c
index d82f134..a073068 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu.c
@@ -673,6 +673,12 @@ dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
 		dmu_tx_t *tx = dmu_tx_create(os);
 		dmu_tx_hold_free(tx, dn->dn_object,
 		    chunk_begin, chunk_end - chunk_begin);
+
+		/*
+		 * Mark this transaction as typically resulting in a net
+		 * reduction in space used.
+		 */
+		dmu_tx_mark_netfree(tx);
 		err = dmu_tx_assign(tx, TXG_WAIT);
 		if (err) {
 			dmu_tx_abort(tx);
@@ -724,6 +730,7 @@ dmu_free_long_object(objset_t *os, uint64_t object)
 	tx = dmu_tx_create(os);
 	dmu_tx_hold_bonus(tx, object);
 	dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
+	dmu_tx_mark_netfree(tx);
 	err = dmu_tx_assign(tx, TXG_WAIT);
 	if (err == 0) {
 		err = dmu_object_free(os, object, tx);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_tx.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_tx.c
index 1213d91..ed9757e 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_tx.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_tx.c
@@ -21,7 +21,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/dmu.h>
@@ -583,6 +583,32 @@ dmu_tx_count_free(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
 	txh->txh_space_tounref += unref;
 }
 
+/*
+ * This function marks the transaction as being a "net free".  The end
+ * result is that refquotas will be disabled for this transaction, and
+ * this transaction will be able to use half of the pool space overhead
+ * (see dsl_pool_adjustedsize()).  Therefore this function should only
+ * be called for transactions that we expect will not cause a net increase
+ * in the amount of space used (but it's OK if that is occasionally not true).
+ */
+void
+dmu_tx_mark_netfree(dmu_tx_t *tx)
+{
+	dmu_tx_hold_t *txh;
+
+	txh = dmu_tx_hold_object_impl(tx, tx->tx_objset,
+	    DMU_NEW_OBJECT, THT_FREE, 0, 0);
+
+	/*
+	 * Pretend that this operation will free 1GB of space.  This
+	 * should be large enough to cancel out the largest write.
+	 * We don't want to use something like UINT64_MAX, because that would
+	 * cause overflows when doing math with these values (e.g. in
+	 * dmu_tx_try_assign()).
+	 */
+	txh->txh_space_tofree = txh->txh_space_tounref = 1024 * 1024 * 1024;
+}
+
 void
 dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len)
 {
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
index 2fc4484..2c0b5ae 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/dmu.h
@@ -569,6 +569,7 @@ void dmu_tx_abort(dmu_tx_t *tx);
 int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how);
 void dmu_tx_wait(dmu_tx_t *tx);
 void dmu_tx_commit(dmu_tx_t *tx);
+void dmu_tx_mark_netfree(dmu_tx_t *tx);
 
 /*
  * To register a commit callback, dmu_tx_callback_register() must be called.
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_dir.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_dir.c
index be96da9..f7c2b6a 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_dir.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_dir.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/types.h>
@@ -559,6 +559,7 @@ zfs_purgedir(znode_t *dzp)
 		dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
 		/* Is this really needed ? */
 		zfs_sa_upgrade_txholds(tx, xzp);
+		dmu_tx_mark_netfree(tx);
 		error = dmu_tx_assign(tx, TXG_WAIT);
 		if (error) {
 			dmu_tx_abort(tx);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
index fd02077..4a86fb2 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  * Copyright 2013 Nexenta Systems, Inc.  All rights reserved.
  */
 
@@ -1558,7 +1558,7 @@ zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp,
  *		cr	- credentials of caller.
  *		flag	- large file flag [UNUSED].
  *		ct	- caller context
- *		vsecp 	- ACL to be set
+ *		vsecp	- ACL to be set
  *
  *	OUT:	vpp	- vnode of created or trunc'd entry.
  *
@@ -1840,7 +1840,7 @@ zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct,
 	zfsvfs_t	*zfsvfs = dzp->z_zfsvfs;
 	zilog_t		*zilog;
 	uint64_t	acl_obj, xattr_obj;
-	uint64_t 	xattr_obj_unlinked = 0;
+	uint64_t	xattr_obj_unlinked = 0;
 	uint64_t	obj = 0;
 	zfs_dirlock_t	*dl;
 	dmu_tx_t	*tx;
@@ -1940,6 +1940,14 @@ top:
 	/* charge as an update -- would be nice not to charge at all */
 	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
 
+	/*
+	 * Mark this transaction as typically resulting in a net free of
+	 * space, unless object removal will be delayed indefinitely
+	 * (due to active holds on the vnode due to the file being open).
+	 */
+	if (may_delete_now)
+		dmu_tx_mark_netfree(tx);
+
 	error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT);
 	if (error) {
 		zfs_dirent_unlock(dl);
@@ -1970,7 +1978,6 @@ top:
 	}
 
 	if (unlinked) {
-
 		/*
 		 * Hold z_lock so that we can make sure that the ACL obj
 		 * hasn't changed.  Could have been deleted due to
@@ -5039,13 +5046,13 @@ zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
  * last page is pushed.  The problem occurs when the msync() call is omitted,
  * which by far the most common case:
  *
- * 	open()
- * 	mmap()
- * 	<modify memory>
- * 	munmap()
- * 	close()
- * 	<time lapse>
- * 	putpage() via fsflush
+ *	open()
+ *	mmap()
+ *	<modify memory>
+ *	munmap()
+ *	close()
+ *	<time lapse>
+ *	putpage() via fsflush
  *
  * If we wait until fsflush to come along, we can have a modification time that
  * is some arbitrary point in the future.  In order to prevent this in the
@@ -5528,7 +5535,7 @@ const fs_operation_def_t zfs_dvnodeops_template[] = {
 	VOPNAME_PATHCONF,	{ .vop_pathconf = zfs_pathconf },
 	VOPNAME_GETSECATTR,	{ .vop_getsecattr = zfs_getsecattr },
 	VOPNAME_SETSECATTR,	{ .vop_setsecattr = zfs_setsecattr },
-	VOPNAME_VNEVENT, 	{ .vop_vnevent = fs_vnevent_support },
+	VOPNAME_VNEVENT,	{ .vop_vnevent = fs_vnevent_support },
 	NULL,			NULL
 };
 
@@ -5562,8 +5569,8 @@ const fs_operation_def_t zfs_fvnodeops_template[] = {
 	VOPNAME_GETSECATTR,	{ .vop_getsecattr = zfs_getsecattr },
 	VOPNAME_SETSECATTR,	{ .vop_setsecattr = zfs_setsecattr },
 	VOPNAME_VNEVENT,	{ .vop_vnevent = fs_vnevent_support },
-	VOPNAME_REQZCBUF, 	{ .vop_reqzcbuf = zfs_reqzcbuf },
-	VOPNAME_RETZCBUF, 	{ .vop_retzcbuf = zfs_retzcbuf },
+	VOPNAME_REQZCBUF,	{ .vop_reqzcbuf = zfs_reqzcbuf },
+	VOPNAME_RETZCBUF,	{ .vop_retzcbuf = zfs_retzcbuf },
 	NULL,			NULL
 };
 
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c
index e78b13b..6945397 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 /* Portions Copyright 2007 Jeremy Teo */
@@ -1510,7 +1510,7 @@ zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
  *	IN:	zp	- znode of file to free data in.
  *		end	- new end-of-file
  *
- * 	RETURN:	0 on success, error code on failure
+ *	RETURN:	0 on success, error code on failure
  */
 static int
 zfs_extend(znode_t *zp, uint64_t end)
@@ -1583,7 +1583,7 @@ zfs_extend(znode_t *zp, uint64_t end)
  *		off	- start of section to free.
  *		len	- length of section to free.
  *
- * 	RETURN:	0 on success, error code on failure
+ *	RETURN:	0 on success, error code on failure
  */
 static int
 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
@@ -1630,7 +1630,7 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
  *	IN:	zp	- znode of file to free data in.
  *		end	- new end-of-file.
  *
- * 	RETURN:	0 on success, error code on failure
+ *	RETURN:	0 on success, error code on failure
  */
 static int
 zfs_trunc(znode_t *zp, uint64_t end)
@@ -1664,6 +1664,7 @@ zfs_trunc(znode_t *zp, uint64_t end)
 	tx = dmu_tx_create(zfsvfs->z_os);
 	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
 	zfs_sa_upgrade_txholds(tx, zp);
+	dmu_tx_mark_netfree(tx);
 	error = dmu_tx_assign(tx, TXG_WAIT);
 	if (error) {
 		dmu_tx_abort(tx);
@@ -1706,7 +1707,7 @@ zfs_trunc(znode_t *zp, uint64_t end)
  *		flag	- current file open mode flags.
  *		log	- TRUE if this action should be logged
  *
- * 	RETURN:	0 on success, error code on failure
+ *	RETURN:	0 on success, error code on failure
  */
 int
 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c
index fa607e1..20480d1 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c
@@ -852,6 +852,7 @@ zvol_update_volsize(objset_t *os, uint64_t volsize)
 
 	tx = dmu_tx_create(os);
 	dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
+	dmu_tx_mark_netfree(tx);
 	error = dmu_tx_assign(tx, TXG_WAIT);
 	if (error) {
 		dmu_tx_abort(tx);
@@ -1988,6 +1989,7 @@ zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
 		rl = zfs_range_lock(&zv->zv_znode, df.df_start, df.df_length,
 		    RL_WRITER);
 		tx = dmu_tx_create(zv->zv_objset);
+		dmu_tx_mark_netfree(tx);
 		error = dmu_tx_assign(tx, TXG_WAIT);
 		if (error != 0) {
 			dmu_tx_abort(tx);