Illumos 5497 - lock contention on arcs_mtx

Reviewed by: George Wilson <[email protected]>
Reviewed by: Matthew Ahrens <[email protected]>
Reviewed by: Richard Elling <[email protected]>
Approved by: Dan McDonald <[email protected]>

Porting notes and other significant code changes:

The illumos 5368 patch (ARC should cache more metadata), which
was never picked up by ZoL, is mostly reverted by this patch.

Since ZoL relies on the kernel asynchronously calling the shrinker to
actually reap memory, the shrinker wakes up arc_reclaim_waiters_cv every
time it runs.

The arc_adapt_thread() function no longer calls arc_do_user_evicts()
since the newly-added arc_user_evicts_thread() calls it periodically.

Notable conflicting ZoL commits which conflicted with this patch or
whose effects are either duplicated or un-done by this patch:

    302f753 - Integrate ARC more tightly with Linux
    39e055c - Adjust arc_p based on "bytes" in arc_shrink
    f521ce1 - Allow "arc_p" to drop to zero or grow to "arc_c"
    77765b5 - Remove "arc_meta_used" from arc_adjust calculation
    94520ca - Prune metadata from ghost lists in arc_adjust_meta

Trace support for multilist_insert() and multilist_remove() has been
added and produces the following output:

    fio-12498 [077] .... 112936.448324: zfs_multilist__insert: ml { offset 240 numsublists 80 sublistidx 63 }
    fio-12498 [077] .... 112936.448347: zfs_multilist__remove: ml { offset 240 numsublists 80 sublistidx 29 }

The following arcstats have been removed:

    recycle_miss - Used by arcstat.py and arc_summary.py, both of which
    have been updated appropriately.

    l2_writes_hdr_miss

The following arcstats have been added:

    evict_not_enough - Number of times arc_evict_state() was unable to
    evict enough buffers to reach its target amount.

    evict_l2_skip - Number of times arc_evict_hdr() skipped eviction
    because it was being written to the l2arc.

    l2_writes_lock_retry - Replaces l2_writes_hdr_miss.  Number of times
    l2arc_write_done() failed to acquire hash_lock (and re-tries).

    arc_meta_min - Shows the value of the zfs_arc_meta_min module
    parameter (see below).

The "index" column of the "dbuf" kstat has been removed since it doesn't
have a direct analog in the new multilist scheme.  Additional multilist-
related stats could be added in the future but would likely require
extensions to the mulilist API.

The following module parameters have been added:

    zfs_arc_evict_batch_limit - Number of ARC headers to free per sub-list
    before moving on to the next sub-list.

    zfs_arc_meta_min - Enforce a floor on the amount of metadata in
    the ARC.

    zfs_arc_num_sublists_per_state - Number of multilist sub-lists per
    ARC state.

    zfs_arc_overflow_shift - Controls amount by which the ARC must exceed
    the target size to be considered "overflowing".

Ported-by: Tim Chase <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]

6 files changed, 208 insertions, 7 deletions

diff --git a/include/sys/Makefile.am b/include/sys/Makefile.am
index 5211e6564..fd80b34ee 100644
--- a/include/sys/Makefile.am
+++ b/include/sys/Makefile.am
@@ -33,6 +33,7 @@ COMMON_H = \
 	$(top_srcdir)/include/sys/efi_partition.h \
 	$(top_srcdir)/include/sys/metaslab.h \
 	$(top_srcdir)/include/sys/metaslab_impl.h \
+	$(top_srcdir)/include/sys/multilist.h \
 	$(top_srcdir)/include/sys/nvpair.h \
 	$(top_srcdir)/include/sys/nvpair_impl.h \
 	$(top_srcdir)/include/sys/range_tree.h \
@@ -53,6 +54,7 @@ COMMON_H = \
 	$(top_srcdir)/include/sys/trace_dbuf.h \
 	$(top_srcdir)/include/sys/trace_dmu.h \
 	$(top_srcdir)/include/sys/trace_dnode.h \
+	$(top_srcdir)/include/sys/trace_multilist.h \
 	$(top_srcdir)/include/sys/trace_txg.h \
 	$(top_srcdir)/include/sys/trace_zil.h \
 	$(top_srcdir)/include/sys/trace_zrlock.h \
diff --git a/include/sys/arc.h b/include/sys/arc.h
index 903f0b413..38f9f27fe 100644
--- a/include/sys/arc.h
+++ b/include/sys/arc.h
@@ -38,6 +38,12 @@ extern "C" {
 #include <sys/spa.h>
 #include <sys/refcount.h>
 
+/*
+ * Used by arc_flush() to inform arc_evict_state() that it should evict
+ * all available buffers from the arc state being passed in.
+ */
+#define	ARC_EVICT_ALL	-1ULL
+
 typedef struct arc_buf_hdr arc_buf_hdr_t;
 typedef struct arc_buf arc_buf_t;
 typedef struct arc_prune arc_prune_t;
@@ -146,7 +152,6 @@ typedef enum arc_state_type {
 typedef struct arc_buf_info {
 	arc_state_type_t	abi_state_type;
 	arc_buf_contents_t	abi_state_contents;
-	uint64_t		abi_state_index;
 	uint32_t		abi_flags;
 	uint32_t		abi_datacnt;
 	uint64_t		abi_size;
@@ -200,7 +205,7 @@ void arc_freed(spa_t *spa, const blkptr_t *bp);
 void arc_set_callback(arc_buf_t *buf, arc_evict_func_t *func, void *private);
 boolean_t arc_clear_callback(arc_buf_t *buf);
 
-void arc_flush(spa_t *spa);
+void arc_flush(spa_t *spa, boolean_t retry);
 void arc_tempreserve_clear(uint64_t reserve);
 int arc_tempreserve_space(uint64_t reserve, uint64_t txg);
 
diff --git a/include/sys/arc_impl.h b/include/sys/arc_impl.h
index 556cc2583..54f5e9f40 100644
--- a/include/sys/arc_impl.h
+++ b/include/sys/arc_impl.h
@@ -67,10 +67,22 @@ extern "C" {
  */
 
 typedef struct arc_state {
-	list_t	arcs_list[ARC_BUFC_NUMTYPES];	/* list of evictable buffers */
-	uint64_t arcs_lsize[ARC_BUFC_NUMTYPES];	/* amount of evictable data */
-	uint64_t arcs_size;	/* total amount of data in this state */
-	kmutex_t arcs_mtx;
+	/*
+	 * list of evictable buffers
+	 */
+	multilist_t arcs_list[ARC_BUFC_NUMTYPES];
+	/*
+	 * total amount of evictable data in this state
+	 */
+	uint64_t arcs_lsize[ARC_BUFC_NUMTYPES];
+	/*
+	 * total amount of data in this state; this includes: evictable,
+	 * non-evictable, ARC_BUFC_DATA, and ARC_BUFC_METADATA.
+	 */
+	uint64_t arcs_size;
+	/*
+	 * supports the "dbufs" kstat
+	 */
 	arc_state_type_t arcs_state;
 } arc_state_t;
 
@@ -136,7 +148,7 @@ typedef struct l1arc_buf_hdr {
 
 	/* protected by arc state mutex */
 	arc_state_t		*b_state;
-	list_node_t		b_arc_node;
+	multilist_node_t	b_arc_node;
 
 	/* updated atomically */
 	clock_t			b_arc_access;
diff --git a/include/sys/multilist.h b/include/sys/multilist.h
new file mode 100644
index 000000000..98d707dd7
--- /dev/null
+++ b/include/sys/multilist.h
@@ -0,0 +1,105 @@
+/*
+ * CDDL HEADER START
+ *
+ * This file and its contents are supplied under the terms of the
+ * Common Development and Distribution License ("CDDL"), version 1.0.
+ * You may only use this file in accordance with the terms of version
+ * 1.0 of the CDDL.
+ *
+ * A full copy of the text of the CDDL should have accompanied this
+ * source.  A copy of the CDDL is also available via the Internet at
+ * http://www.illumos.org/license/CDDL.
+ *
+ * CDDL HEADER END
+ */
+/*
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
+ */
+
+#ifndef	_SYS_MULTILIST_H
+#define	_SYS_MULTILIST_H
+
+#include <sys/zfs_context.h>
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+typedef list_node_t multilist_node_t;
+typedef struct multilist multilist_t;
+typedef struct multilist_sublist multilist_sublist_t;
+typedef unsigned int multilist_sublist_index_func_t(multilist_t *, void *);
+
+struct multilist_sublist {
+	/*
+	 * The mutex used internally to implement thread safe insertions
+	 * and removals to this individual sublist. It can also be locked
+	 * by a consumer using multilist_sublist_{lock,unlock}, which is
+	 * useful if a consumer needs to traverse the list in a thread
+	 * safe manner.
+	 */
+	kmutex_t	mls_lock;
+	/*
+	 * The actual list object containing all objects in this sublist.
+	 */
+	list_t		mls_list;
+	/*
+	 * Pad to cache line, in an effort to try and prevent cache line
+	 * contention.
+	 */
+} ____cacheline_aligned;
+
+struct multilist {
+	/*
+	 * This is used to get to the multilist_node_t structure given
+	 * the void *object contained on the list.
+	 */
+	size_t				ml_offset;
+	/*
+	 * The number of sublists used internally by this multilist.
+	 */
+	uint64_t			ml_num_sublists;
+	/*
+	 * The array of pointers to the actual sublists.
+	 */
+	multilist_sublist_t		*ml_sublists;
+	/*
+	 * Pointer to function which determines the sublist to use
+	 * when inserting and removing objects from this multilist.
+	 * Please see the comment above multilist_create for details.
+	 */
+	multilist_sublist_index_func_t	*ml_index_func;
+};
+
+void multilist_destroy(multilist_t *);
+void multilist_create(multilist_t *, size_t, size_t, unsigned int,
+    multilist_sublist_index_func_t *);
+
+void multilist_insert(multilist_t *, void *);
+void multilist_remove(multilist_t *, void *);
+int  multilist_is_empty(multilist_t *);
+
+unsigned int multilist_get_num_sublists(multilist_t *);
+unsigned int multilist_get_random_index(multilist_t *);
+
+multilist_sublist_t *multilist_sublist_lock(multilist_t *, unsigned int);
+void multilist_sublist_unlock(multilist_sublist_t *);
+
+void multilist_sublist_insert_head(multilist_sublist_t *, void *);
+void multilist_sublist_insert_tail(multilist_sublist_t *, void *);
+void multilist_sublist_move_forward(multilist_sublist_t *mls, void *obj);
+void multilist_sublist_remove(multilist_sublist_t *, void *);
+
+void *multilist_sublist_head(multilist_sublist_t *);
+void *multilist_sublist_tail(multilist_sublist_t *);
+void *multilist_sublist_next(multilist_sublist_t *, void *);
+void *multilist_sublist_prev(multilist_sublist_t *, void *);
+
+void multilist_link_init(multilist_node_t *);
+int  multilist_link_active(multilist_node_t *);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* _SYS_MULTILIST_H */
diff --git a/include/sys/trace_multilist.h b/include/sys/trace_multilist.h
new file mode 100644
index 000000000..11d2f2701
--- /dev/null
+++ b/include/sys/trace_multilist.h
@@ -0,0 +1,76 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or http://www.opensolaris.org/os/licensing.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+#if defined(_KERNEL) && defined(HAVE_DECLARE_EVENT_CLASS)
+
+#undef TRACE_SYSTEM
+#define	TRACE_SYSTEM zfs
+
+#if !defined(_TRACE_MULTILIST_H) || defined(TRACE_HEADER_MULTI_READ)
+#define	_TRACE_MULTILIST_H
+
+#include <linux/tracepoint.h>
+#include <sys/types.h>
+
+/*
+ * Generic support for three argument tracepoints of the form:
+ *
+ * DTRACE_PROBE3(...,
+ *     multilist_t *, ...,
+ *     unsigned int, ...,
+ *     void *, ...);
+ */
+
+DECLARE_EVENT_CLASS(zfs_multilist_insert_remove_class,
+	TP_PROTO(multilist_t *ml, unsigned sublist_idx, void *obj),
+	TP_ARGS(ml, sublist_idx, obj),
+	TP_STRUCT__entry(
+	    __field(size_t,		ml_offset)
+	    __field(uint64_t,		ml_num_sublists)
+
+	    __field(unsigned int,	sublist_idx)
+	),
+	TP_fast_assign(
+	    __entry->ml_offset		= ml->ml_offset;
+	    __entry->ml_num_sublists	= ml->ml_num_sublists;
+
+	    __entry->sublist_idx	= sublist_idx;
+	),
+	TP_printk("ml { offset %ld numsublists %llu sublistidx %u } ",
+	    __entry->ml_offset, __entry->ml_num_sublists, __entry->sublist_idx)
+);
+
+#define	DEFINE_MULTILIST_INSERT_REMOVE_EVENT(name) \
+DEFINE_EVENT(zfs_multilist_insert_remove_class, name, \
+	TP_PROTO(multilist_t *ml, unsigned int sublist_idx, void *obj), \
+	TP_ARGS(ml, sublist_idx, obj))
+DEFINE_MULTILIST_INSERT_REMOVE_EVENT(zfs_multilist__insert);
+DEFINE_MULTILIST_INSERT_REMOVE_EVENT(zfs_multilist__remove);
+
+#endif /* _TRACE_MULTILIST_H */
+
+#undef TRACE_INCLUDE_PATH
+#undef TRACE_INCLUDE_FILE
+#define	TRACE_INCLUDE_PATH sys
+#define	TRACE_INCLUDE_FILE trace_multilist
+#include <trace/define_trace.h>
+
+#endif /* _KERNEL && HAVE_DECLARE_EVENT_CLASS */
diff --git a/include/sys/zfs_context.h b/include/sys/zfs_context.h
index 3dc54f1d7..761b1d57a 100644
--- a/include/sys/zfs_context.h
+++ b/include/sys/zfs_context.h
@@ -609,6 +609,7 @@ extern void delay(clock_t ticks);
 	} while (0);
 
 #define	max_ncpus	64
+#define	num_online_cpus() (sysconf(_SC_NPROCESSORS_ONLN))
 
 #define	minclsyspri	60
 #define	maxclsyspri	99