Illumos 5056 - ZFS deadlock on db_mtx and dn_holds

5056 ZFS deadlock on db_mtx and dn_holds
Author: Justin Gibbs <[email protected]>
Reviewed by: Will Andrews <[email protected]>
Reviewed by: Matt Ahrens <[email protected]>
Reviewed by: George Wilson <[email protected]>
Approved by: Dan McDonald <[email protected]>

References:
  https://www.illumos.org/issues/5056
  https://github.com/illumos/illumos-gate/commit/bc9014e

Porting Notes:

sa_handle_get_from_db():
  - the original patch includes an otherwise unmentioned fix for a
    possible usage of an uninitialised variable

dmu_objset_open_impl():
  - Under Illumos list_link_init() is the same as filling a list_node_t
    with NULLs, so they don't notice if they miss doing list_link_init()
    on a zero'd containing structure (e.g. allocated with kmem_zalloc as
    here). Under Linux, not so much: an uninitialised list_node_t goes
    "Boom!" some time later when it's used or destroyed.

dmu_objset_evict_dbufs():
  - reduce stack usage using kmem_alloc()

Ported-by: Chris Dunlop <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>

1 files changed, 10 insertions, 4 deletions

diff --git a/include/sys/dmu_objset.h b/include/sys/dmu_objset.h
index cbf0394e6..65ae850f4 100644
--- a/include/sys/dmu_objset.h
+++ b/include/sys/dmu_objset.h
@@ -22,6 +22,7 @@
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -74,22 +75,25 @@ struct objset {
 	arc_buf_t *os_phys_buf;
 	objset_phys_t *os_phys;
 	/*
-	 * The following "special" dnodes have no parent and are exempt from
-	 * dnode_move(), but they root their descendents in this objset using
-	 * handles anyway, so that all access to dnodes from dbufs consistently
-	 * uses handles.
+	 * The following "special" dnodes have no parent, are exempt
+	 * from dnode_move(), and are not recorded in os_dnodes, but they
+	 * root their descendents in this objset using handles anyway, so
+	 * that all access to dnodes from dbufs consistently uses handles.
 	 */
 	dnode_handle_t os_meta_dnode;
 	dnode_handle_t os_userused_dnode;
 	dnode_handle_t os_groupused_dnode;
 	zilog_t *os_zil;
 
+	list_node_t os_evicting_node;
+
 	/* can change, under dsl_dir's locks: */
 	enum zio_checksum os_checksum;
 	enum zio_compress os_compress;
 	uint8_t os_copies;
 	enum zio_checksum os_dedup_checksum;
 	boolean_t os_dedup_verify;
+	boolean_t os_evicting;
 	zfs_logbias_op_t os_logbias;
 	zfs_cache_type_t os_primary_cache;
 	zfs_cache_type_t os_secondary_cache;
@@ -168,6 +172,8 @@ int dmu_objset_userspace_upgrade(objset_t *os);
 boolean_t dmu_objset_userspace_present(objset_t *os);
 int dmu_fsname(const char *snapname, char *buf);
 
+void dmu_objset_evict_done(objset_t *os);
+
 void dmu_objset_init(void);
 void dmu_objset_fini(void);