Skip to content

Commit

Permalink
Illumos 5497 - lock contention on arcs_mtx
Browse files Browse the repository at this point in the history
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Richard Elling <richard.elling@richardelling.com>
Approved by: Dan McDonald <danmcd@omniti.com>

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 <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov
  • Loading branch information
Prakash Surya authored and behlendorf committed Jun 11, 2015
1 parent b9541d6 commit ca0bf58
Show file tree
Hide file tree
Showing 17 changed files with 1,940 additions and 758 deletions.
4 changes: 0 additions & 4 deletions cmd/arc_summary/arc_summary.py
Original file line number Diff line number Diff line change
Expand Up @@ -191,12 +191,10 @@ def get_arc_summary(Kstat):
### ARC Misc. ###
deleted = Kstat["kstat.zfs.misc.arcstats.deleted"]
mutex_miss = Kstat["kstat.zfs.misc.arcstats.mutex_miss"]
recycle_miss = Kstat["kstat.zfs.misc.arcstats.recycle_miss"]

### ARC Misc. ###
output["arc_misc"] = {}
output["arc_misc"]["deleted"] = fHits(deleted)
output["arc_misc"]['recycle_miss'] = fHits(recycle_miss)
output["arc_misc"]['mutex_miss'] = fHits(mutex_miss)
output["arc_misc"]['evict_skips'] = fHits(mutex_miss)

Expand Down Expand Up @@ -302,8 +300,6 @@ def _arc_summary(Kstat):
### ARC Misc. ###
sys.stdout.write("ARC Misc:\n")
sys.stdout.write("\tDeleted:\t\t\t\t%s\n" % arc['arc_misc']['deleted'])
sys.stdout.write("\tRecycle Misses:\t\t\t\t%s\n" %
arc['arc_misc']['recycle_miss'])
sys.stdout.write("\tMutex Misses:\t\t\t\t%s\n" %
arc['arc_misc']['mutex_miss'])
sys.stdout.write("\tEvict Skips:\t\t\t\t%s\n" %
Expand Down
2 changes: 0 additions & 2 deletions cmd/arcstat/arcstat.py
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,6 @@
"mrug": [4, 1000, "MRU Ghost List hits per second"],
"eskip": [5, 1000, "evict_skip per second"],
"mtxmis": [6, 1000, "mutex_miss per second"],
"rmis": [4, 1000, "recycle_miss per second"],
"dread": [5, 1000, "Demand accesses per second"],
"pread": [5, 1000, "Prefetch accesses per second"],
"l2hits": [6, 1000, "L2ARC hits per second"],
Expand Down Expand Up @@ -406,7 +405,6 @@ def calculate():
v["mrug"] = d["mru_ghost_hits"] / sint
v["mfug"] = d["mfu_ghost_hits"] / sint
v["eskip"] = d["evict_skip"] / sint
v["rmis"] = d["recycle_miss"] / sint
v["mtxmis"] = d["mutex_miss"] / sint

if l2exist:
Expand Down
2 changes: 2 additions & 0 deletions include/sys/Makefile.am
Original file line number Diff line number Diff line change
Expand Up @@ -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 \
Expand All @@ -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 \
Expand Down
9 changes: 7 additions & 2 deletions include/sys/arc.h
Original file line number Diff line number Diff line change
Expand Up @@ -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;
Expand Down Expand Up @@ -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;
Expand Down Expand Up @@ -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);

Expand Down
22 changes: 17 additions & 5 deletions include/sys/arc_impl.h
Original file line number Diff line number Diff line change
Expand Up @@ -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;

Expand Down Expand Up @@ -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;
Expand Down
105 changes: 105 additions & 0 deletions include/sys/multilist.h
Original file line number Diff line number Diff line change
@@ -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 */
76 changes: 76 additions & 0 deletions include/sys/trace_multilist.h
Original file line number Diff line number Diff line change
@@ -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 */
1 change: 1 addition & 0 deletions include/sys/zfs_context.h
Original file line number Diff line number Diff line change
Expand Up @@ -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
Expand Down
1 change: 1 addition & 0 deletions lib/libzpool/Makefile.am
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,7 @@ libzpool_la_SOURCES = \
$(top_srcdir)/module/zfs/lzjb.c \
$(top_srcdir)/module/zfs/lz4.c \
$(top_srcdir)/module/zfs/metaslab.c \
$(top_srcdir)/module/zfs/multilist.c \
$(top_srcdir)/module/zfs/range_tree.c \
$(top_srcdir)/module/zfs/refcount.c \
$(top_srcdir)/module/zfs/rrwlock.c \
Expand Down
60 changes: 60 additions & 0 deletions man/man5/zfs-module-parameters.5
Original file line number Diff line number Diff line change
Expand Up @@ -347,6 +347,19 @@ increased to reduce the memory footprint.
Default value: \fB8192\fR.
.RE

.sp
.ne 2
.na
\fBzfs_arc_evict_batch_limit\fR (int)
.ad
.RS 12n
Number ARC headers to evict per sub-list before proceding to another sub-list.
This batch-style operation prevents entire sub-lists from being evicted at once
but comes at a cost of additional unlocking and locking.
.sp
Default value: \fB10\fR.
.RE

.sp
.ne 2
.na
Expand Down Expand Up @@ -395,6 +408,19 @@ for meta data.
Default value: \fB0\fR.
.RE

.sp
.ne 2
.na
\fBzfs_arc_meta_min\fR (ulong)
.ad
.RS 12n
The minimum allowed size in bytes that meta data buffers may consume in
the ARC. This value defaults to 0 which disables a floor on the amount
of the ARC devoted meta data.
.sp
Default value: \fB0\fR.
.RE

.sp
.ne 2
.na
Expand Down Expand Up @@ -447,6 +473,40 @@ Min life of prefetch block
Default value: \fB100\fR.
.RE

.sp
.ne 2
.na
\fBzfs_arc_num_sublists_per_state\fR (int)
.ad
.RS 12n
To allow more fine-grained locking, each ARC state contains a series
of lists for both data and meta data objects. Locking is performed at
the level of these "sub-lists". This parameters controls the number of
sub-lists per ARC state.
.sp
Default value: 1 or the number of on-online CPUs, whichever is greater
.RE

.sp
.ne 2
.na
\fBzfs_arc_overflow_shift\fR (int)
.ad
.RS 12n
The ARC size is considered to be overflowing if it exceeds the current
ARC target size (arc_c) by a threshold determined by this parameter.
The threshold is calculated as a fraction of arc_c using the formula
"arc_c >> \fBzfs_arc_overflow_shift\fR".

The default value of 8 causes the ARC to be considered to be overflowing
if it exceeds the target size by 1/256th (0.3%) of the target size.

When the ARC is overflowing, new buffer allocations are stalled until
the reclaim thread catches up and the overflow condition no longer exists.
.sp
Default value: \fB8\fR.
.RE

.sp
.ne 2
.na
Expand Down
Loading

0 comments on commit ca0bf58

Please sign in to comment.