sched: Implement shared runqueue in fair.c
v4 cover sched: Implement shared runqueue in fair.c
This is v4 of the shared runqueue patchset. This patch set is based off of commit 418146e39891 ("freezer,sched: Clean saved_state when restoring it during thaw") on the sched/core branch of tip.git.
In prior versions of this patch set, I was observing consistent and statistically significant wins for several benchmarks when this feature was enabled, such as kernel compile and hackbench. After rebasing onto the latest sched/core on tip.git, I'm no longer observing these wins, and in fact observe some performance loss with SHARED_RUNQ on hackbench. I ended up bisecting this to when EEVDF was merged.
As I mentioned in [0], our plan for now is to take a step back and re-evaluate how we want to proceed with this patch set. That said, I did want to send this out in the interim in case it could be of interest to anyone else who would like to continue to experiment with it.
[0]: https://lore.kernel.org/all/20231204193001.GA53255@maniforge/
v1 (RFC): https://lore.kernel.org/lkml/[email protected]/
v2: https://lore.kernel.org/lkml/[email protected]/
v3: https://lore.kernel.org/all/[email protected]/
v3 -> v4 changes:
- Ensure list is fully drained when toggling feature on and off (noticed offline by Chris Mason, and independently by Aboorva Devarajan)
- Also check !is_migration_disabled() in shared_runq_pick_next_task() if we find a task in the shard, as is_cpu_allowed() doesn't check for migration. Also do another check for is_cpu_allowed() after re-acquiring the task rq lock in case state has changed
- Statically initialize the shared_runq_node list node in the init task.
- Only try to pull a task from the shared_runq if the rq's root domain has the overload bit set (K Prateek Nayak)
- Check this_rq->ttwu_pending after trying to pull a task from a shared runqueue shard before going forward to load_balance() (K Prateek Nayak)
- Fix where we would try to skip over the lowest-level sched domain -- do the check in load_balance() instead of before, as for_each_domain() iterates over all domains starting from the beginning (K Prateek Nayak)
- Add a selftest testcase which toggles the SHARED_RUNQ feature on and off in a loop
Worth noting is that there have been a number of suggestions for improving this feature that were not included in this v4, such as:
- [1], where Chen Yu <[email protected]> suggested not putting certain tasks on a shared_runq if e.g. p->avg_runtime <= sysctl_migration_cost. I elected to not include this, as it's a heuristic that could incorrectly prevent work conservation, which is the primary goal of the feature.
- [2], where K Prateek Nayak <[email protected]> suggested adding a per-shard "overload" flag that can be set to avoid contending on the shard lock. This should be covered by checking the root domain overload flag.
- [3], where K Prateek Nayak <[email protected]> suggested also checking rq->avg_idle < sd->max_newidle_lb_cost. This is a similar suggestion to Chen Yu's above, and I elected to leave it out here for the same reason: that we want to encourage work conservation.
- [4], where Gautham Shenoy <[email protected]> suggests iterating over all tasks in a shard until one is found that can be pulled, rather than bailing out after failing to migrate the HEAD task.
None of these ideas are unreasonable, and may be worth applying if it improves the feature for more general cases following further testing. I left the patch set as is simply to keep the feature "consistent" in encouraging work conservation, but that decision can be revisited.
[1]: https://lore.kernel.org/all/ZO7e5YaS71cXVxQN@chenyu5-mobl2/
[2]: https://lore.kernel.org/all/[email protected]/
[3]: https://lore.kernel.org/all/[email protected]/
[4]: https://lore.kernel.org/lkml/[email protected]/
v2 -> v3 changes:
- Don't leave stale tasks in the lists when the SHARED_RUNQ feature is disabled (Abel Wu)
- Use raw spin lock instead of spinlock_t (Peter)
- Fix return value from shared_runq_pick_next_task() to match the semantics expected by newidle_balance() (Gautham, Abel)
- Fold patch __enqueue_entity() / __dequeue_entity() into previous patch (Peter)
- Skip <= LLC domains in newidle_balance() if SHARED_RUNQ is enabled (Peter)
- Properly support hotplug and recreating sched domains (Peter)
- Avoid unnecessary task_rq_unlock() + raw_spin_rq_lock() when src_rq == target_rq in shared_runq_pick_next_task() (Abel)
- Only issue list_del_init() in shared_runq_dequeue_task() if the task is still in the list after acquiring the lock (Aaron Lu)
- Slightly change shared_runq_shard_idx() to make it more likely to keep SMT siblings on the same bucket (Peter)
v1 -> v2 changes:
- Change name from swqueue to shared_runq (Peter)
- Shard per-LLC shared runqueues to avoid contention on scheduler-heavy workloads (Peter)
- Pull tasks from the shared_runq in newidle_balance() rather than in pick_next_task_fair() (Peter and Vincent)
- Rename a few functions to reflect their actual purpose. For example, shared_runq_dequeue_task() instead of swqueue_remove_task() (Peter)
- Expose move_queued_task() from core.c rather than migrate_task_to() (Peter)
- Properly check is_cpu_allowed() when pulling a task from a shared_runq to ensure it can actually be migrated (Peter and Gautham)
- Dropped RFC tag
David Vernet (8):
sched: Expose move_queued_task() from core.c
sched: Move is_cpu_allowed() into sched.h
sched: Tighten unpinned rq lock window in newidle_balance()
sched: Check cpu_active() earlier in newidle_balance()
sched: Enable sched_feat callbacks on enable/disable
sched: Implement shared runqueue in CFS
sched: Shard per-LLC shared runqueues
sched: Add selftest for SHARED_RUNQ
include/linux/sched.h | 2 + init/init_task.c | 1 + kernel/sched/core.c | 52 +-- kernel/sched/debug.c | 18 +- kernel/sched/fair.c | 413 +++++++++++++++++++++- kernel/sched/features.h | 2 + kernel/sched/sched.h | 60 +++- kernel/sched/topology.c | 4 +- tools/testing/selftests/sched/Makefile | 7 +- tools/testing/selftests/sched/config | 2 + tools/testing/selftests/sched/test-swq.sh | 23 ++ 11 files changed, 521 insertions(+), 63 deletions(-) create mode 100755 tools/testing/selftests/sched/test-swq.sh
Signature · 1 line
2.42.1
PATCH v4 3/8 sched: Tighten unpinned rq lock window in newidle_balance()
In newidle_balance(), we may drop and reacquire the rq lock in the load-balance phase of the function. We currently do this before we check rq->rd->overload or rq->avg_idle, which is unnecessary. Let's tighten the window where we call rq_unpin_lock().
--- kernel/sched/fair.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-)
kernel/sched/fair.c +8 -10
@@ -12296,14 +12296,6 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)if (!cpu_active (this_cpu))return 0 ; -/* -* This is OK, because current is on_cpu, which avoids it being picked -* for load-balance and preemption/IRQs are still disabled avoiding -* further scheduler activity on it and we're being very careful to -* re-start the picking loop. -*/ -rq_unpin_lock (this_rq, rf); -rcu_read_lock (); sd =rcu_dereference_check_sched_domain (this_rq->sd);
@@ -12318,6 +12310,13 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf) }rcu_read_unlock (); +/* +* This is OK, because current is on_cpu, which avoids it being picked +* for load-balance and preemption/IRQs are still disabled avoiding +* further scheduler activity on it and we're being very careful to +* re-start the picking loop. +*/ +rq_unpin_lock (this_rq, rf);raw_spin_rq_unlock (this_rq); t0 =sched_clock_cpu (this_cpu);
@@ -12358,6 +12357,7 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)rcu_read_unlock ();raw_spin_rq_lock (this_rq); +rq_repin_lock (this_rq, rf);if (curr_cost > this_rq->max_idle_balance_cost) this_rq->max_idle_balance_cost = curr_cost;
@@ -12384,8 +12384,6 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)else nohz_newidle_balance (this_rq); -rq_repin_lock (this_rq, rf); -return pulled_task; }
Signature · 1 line
2.42.1
PATCH v4 2/8 sched: Move is_cpu_allowed() into sched.h
In reply to David Vernetis_cpu_allowed() exists as a static inline function in core.c. The functionality offered by is_cpu_allowed() is useful to scheduling policies as well, e.g. to determine whether a runnable task can be migrated to another core that would otherwise go idle.
Let's move it to sched.h.
--- kernel/sched/core.c | 31 ------------------------------- kernel/sched/sched.h | 31 +++++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+), 31 deletions(-)
kernel/sched/core.c +0 -31
@@ -48,7 +48,6 @@#include <linux/kcov.h> #include <linux/kprobes.h> #include <linux/llist_api.h> -#include <linux/mmu_context.h> #include <linux/mmzone.h> #include <linux/mutex_api.h> #include <linux/nmi.h>
@@ -2469,36 +2468,6 @@ static inline bool rq_has_pinned_tasks(struct rq *rq)return rq->nr_pinned; } -/* -* Per-CPU kthreads are allowed to run on !active && online CPUs, see -* __set_cpus_allowed_ptr() and select_fallback_rq(). -*/ -static inline bool is_cpu_allowed (struct task_struct *p,int cpu) -{ -/* When not in the task's cpumask, no point in looking further. */ -if (!cpumask_test_cpu (cpu, p->cpus_ptr)) -return false; - -/* migrate_disabled() must be allowed to finish. */ -if (is_migration_disabled (p)) -return cpu_online (cpu); - -/* Non kernel threads are not allowed during either online or offline. */ -if (!(p->flags &PF_KTHREAD )) -return cpu_active (cpu) &&task_cpu_possible (cpu, p); - -/* KTHREAD_IS_PER_CPU is always allowed. */ -if (kthread_is_per_cpu (p)) -return cpu_online (cpu); - -/* Regular kernel threads don't get to stay during offline. */ -if (cpu_dying (cpu)) -return false; - -/* But are allowed during online. */ -return cpu_online (cpu); -} -/* * This is how migration works: *
kernel/sched/sched.h +31 -0
@@ -44,6 +44,7 @@#include <linux/lockdep.h> #include <linux/minmax.h> #include <linux/mm.h> +#include <linux/mmu_context.h> #include <linux/module.h> #include <linux/mutex_api.h> #include <linux/plist.h>
@@ -1206,6 +1207,36 @@ static inline bool is_migration_disabled(struct task_struct *p)#endif } +/* +* Per-CPU kthreads are allowed to run on !active && online CPUs, see +* __set_cpus_allowed_ptr() and select_fallback_rq(). +*/ +static inline bool is_cpu_allowed (struct task_struct *p,int cpu) +{ +/* When not in the task's cpumask, no point in looking further. */ +if (!cpumask_test_cpu (cpu, p->cpus_ptr)) +return false; + +/* migrate_disabled() must be allowed to finish. */ +if (is_migration_disabled (p)) +return cpu_online (cpu); + +/* Non kernel threads are not allowed during either online or offline. */ +if (!(p->flags &PF_KTHREAD )) +return cpu_active (cpu) &&task_cpu_possible (cpu, p); + +/* KTHREAD_IS_PER_CPU is always allowed. */ +if (kthread_is_per_cpu (p)) +return cpu_online (cpu); + +/* Regular kernel threads don't get to stay during offline. */ +if (cpu_dying (cpu)) +return false; + +/* But are allowed during online. */ +return cpu_online (cpu); +} + DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);#define cpu_rq (cpu) (&per_cpu(runqueues, (cpu)))
Signature · 1 line
2.42.1
PATCH v4 1/8 sched: Expose move_queued_task() from core.c
In reply to David VernetThe migrate_task_to() function exposed from kernel/sched/core.c migrates the current task, which is silently assumed to also be its first argument, to the specified CPU. The function uses stop_one_cpu() to migrate the task to the target CPU, which won't work if @p is not the current task as the stop_one_cpu() callback isn't invoked on remote CPUs.
While this operation is useful for task_numa_migrate() in fair.c, it would be useful if move_queued_task() in core.c was given external linkage, as it actually can be used to migrate any task to a CPU.
A follow-on patch will call move_queued_task() from fair.c when migrating a task in a shared runqueue to a remote CPU.
--- kernel/sched/core.c | 4 ++-- kernel/sched/sched.h | 3 +++ 2 files changed, 5 insertions(+), 2 deletions(-)
kernel/sched/core.c +2 -2
@@ -2518,8 +2518,8 @@ static inline bool is_cpu_allowed(struct task_struct *p, int cpu)* * Returns (locked) new rq. Old rq's lock is released. */ -static struct rq *move_queued_task (struct rq *rq,struct rq_flags *rf, -struct task_struct *p,int new_cpu) +struct rq *move_queued_task (struct rq *rq,struct rq_flags *rf, +struct task_struct *p,int new_cpu) {lockdep_assert_rq_held (rq);
kernel/sched/sched.h +3 -0
@@ -1787,6 +1787,9 @@ init_numa_balancing(unsigned long clone_flags, struct task_struct *p)#ifdef CONFIG_SMP + +extern struct rq *move_queued_task (struct rq *rq,struct rq_flags *rf, +struct task_struct *p,int new_cpu); static inline void queue_balance_callback(struct rq *rq, struct balance_callback *head,
Signature · 1 line
2.42.1
PATCH v4 4/8 sched: Check cpu_active() earlier in newidle_balance()
In reply to David VernetIn newidle_balance(), we check if the current CPU is inactive, and then decline to pull any remote tasks to the core if so. Before this check, however, we're currently updating rq->idle_stamp. If a core is offline, setting its idle stamp is not useful. The core won't be chosen by any task in select_task_rq_fair(), and setting the rq->idle_stamp is misleading anyways given that the core being inactive should imply that it should be idle for a very long time.
Let's set rq->idle_stamp in newidle_balance() only if the cpu is active.
--- kernel/sched/fair.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
kernel/sched/fair.c +6 -6
@@ -12284,18 +12284,18 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)if (this_rq->ttwu_pending)return 0 ; -/* -* We must set idle_stamp _before_ calling idle_balance(), such that we -* measure the duration of idle_balance() as idle time. -*/ - this_rq->idle_stamp =rq_clock (this_rq); -/* * Do not pull tasks towards !active CPUs... */ if (!cpu_active (this_cpu))return 0 ; +/* +* We must set idle_stamp _before_ calling idle_balance(), such that we +* measure the duration of idle_balance() as idle time. +*/ + this_rq->idle_stamp =rq_clock (this_rq); +rcu_read_lock (); sd =rcu_dereference_check_sched_domain (this_rq->sd);
Signature · 1 line
2.42.1
PATCH v4 5/8 sched: Enable sched_feat callbacks on enable/disable
In reply to David VernetWhen a scheduler feature is enabled or disabled, the sched_feat_enable() and sched_feat_disable() functions are invoked respectively for that feature. For features that don't require resetting any state, this works fine. However, there will be an upcoming feature called SHARED_RUNQ which needs to drain all tasks from a set of global shared runqueues in order to avoid stale tasks from staying in the queues after the feature has been disabled.
This patch therefore defines a new SCHED_FEAT_CALLBACK macro which allows scheduler features to specify a callback that should be invoked when a feature is enabled or disabled respectively. The SCHED_FEAT macro assumes a NULL callback.
--- kernel/sched/core.c | 4 ++-- kernel/sched/debug.c | 18 ++++++++++++++---- kernel/sched/sched.h | 16 ++++++++++------ 3 files changed, 26 insertions(+), 12 deletions(-)
kernel/sched/core.c +2 -2
@@ -124,12 +124,12 @@ DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);* sysctl_sched_features, defined in sched.h, to allow constants propagation * at compile time and compiler optimization based on features default. */ -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ (1UL << __SCHED_FEAT_##name) * enabled | const_debug unsigned int sysctl_sched_features =#include "features.h" 0 ; -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACK/* * Print a warning if need_resched is set for the given duration (if
kernel/sched/debug.c +14 -4
@@ -44,14 +44,14 @@ static unsigned long nsec_low(unsigned long long nsec)#define SPLIT_NS (x) nsec_high(x), nsec_low(x) -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ #name ,static const char *const sched_feat_names[] = { #include "features.h" }; -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACKstatic int sched_feat_show (struct seq_file *m,void *v) {
@@ -72,22 +72,32 @@ static int sched_feat_show(struct seq_file *m, void *v)#define jump_label_key__true STATIC_KEY_INIT_TRUE#define jump_label_key__false STATIC_KEY_INIT_FALSE -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ jump_label_key__##enabled ,struct static_key sched_feat_keys[__SCHED_FEAT_NR] = { #include "features.h" }; -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACK + +#define SCHED_FEAT_CALLBACK (name, enabled, cb) cb, +static const sched_feat_change_f sched_feat_cbs[__SCHED_FEAT_NR] = { +#include "features.h" +}; +#undef SCHED_FEAT_CALLBACKstatic void sched_feat_disable (int i) { +if (sched_feat_cbs[i]) + sched_feat_cbs[i](false);static_key_disable_cpuslocked (&sched_feat_keys[i]); }static void sched_feat_enable (int i) { +if (sched_feat_cbs[i]) + sched_feat_cbs[i](true);static_key_enable_cpuslocked (&sched_feat_keys[i]); }#else
kernel/sched/sched.h +10 -6
@@ -2091,6 +2091,8 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)#endif } +#define SCHED_FEAT (name, enabled) SCHED_FEAT_CALLBACK(name, enabled, NULL) +/* * Tunables that become constants when CONFIG_SCHED_DEBUG is off: */
@@ -2100,7 +2102,7 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)# define const_debug const#endif -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ __SCHED_FEAT_##name ,enum {
@@ -2108,7 +2110,7 @@ enum { __SCHED_FEAT_NR, }; -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACK#ifdef CONFIG_SCHED_DEBUG
@@ -2119,14 +2121,14 @@ enum { extern const_debug unsigned int sysctl_sched_features;#ifdef CONFIG_JUMP_LABEL -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ static __always_inline bool static_branch_##name(struct static_key *key) \ { \ return static_key_##enabled(key); \ }#include "features.h" -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACKextern struct static_key sched_feat_keys[__SCHED_FEAT_NR];#define sched_feat (x) (static_branch_##x(&sched_feat_keys[__SCHED_FEAT_##x]))
@@ -2144,17 +2146,19 @@ extern struct static_key sched_feat_keys[__SCHED_FEAT_NR];* constants propagation at compile time and compiler optimization based on * features default. */ -#define SCHED_FEAT (name, enabled) \ +#define SCHED_FEAT_CALLBACK (name, enabled, cb) \ (1UL << __SCHED_FEAT_##name) * enabled | static const_debug __maybe_unused unsigned int sysctl_sched_features = #include "features.h"0 ; -#undef SCHED_FEAT +#undef SCHED_FEAT_CALLBACK#define sched_feat (x) !!(sysctl_sched_features & (1UL << __SCHED_FEAT_##x))#endif /* SCHED_DEBUG */ +typedef void (*sched_feat_change_f )(bool enabling); +extern struct static_key_false sched_numa_balancing;extern struct static_key_false sched_schedstats;
Signature · 1 line
2.42.1
PATCH v4 7/8 sched: Shard per-LLC shared runqueues
In reply to David VernetThe SHARED_RUNQ scheduler feature creates a FIFO queue per LLC that tasks are put into on enqueue, and pulled from when a core in that LLC would otherwise go idle. For CPUs with large LLCs, this can sometimes cause significant contention, as illustrated in [0].
[0]: https://lore.kernel.org/all/[email protected]/
So as to try and mitigate this contention, we can instead shard the per-LLC runqueue into multiple per-LLC shards.
While this doesn't outright prevent all contention, it does somewhat mitigate it. For example, if we run the following schbench command which does almost nothing other than pound the runqueue:
schbench -L -m 52 -p 512 -r 10 -t 1
we observe with lockstats that sharding significantly decreases contention.
3 shards:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total waittime-avg acq-bounces acquisitions holdtime-min holdtime-max holdtime-total holdtime-avg
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
&shard->lock: 31510503 31510711 0.08 19.98 168932319.64 5.36 31700383 31843851 0.03 17.50 10273968.33 0.32
------------
&shard->lock 15731657 [<0000000068c0fd75>] pick_next_task_fair+0x4dd/0x510 &shard->lock 15756516 [<000000001faf84f9>] enqueue_task_fair+0x459/0x530 &shard->lock 21766 [<00000000126ec6ab>] newidle_balance+0x45a/0x650 &shard->lock 772 [<000000002886c365>] dequeue_task_fair+0x4c9/0x540 ------------ &shard->lock 23458 [<00000000126ec6ab>] newidle_balance+0x45a/0x650 &shard->lock 16505108 [<000000001faf84f9>] enqueue_task_fair+0x459/0x530 &shard->lock 14981310 [<0000000068c0fd75>] pick_next_task_fair+0x4dd/0x510 &shard->lock 835 [<000000002886c365>] dequeue_task_fair+0x4c9/0x540
No sharding:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total waittime-avg acq-bounces acquisitions holdtime-min holdtime-max holdtime-total holdtime-avg
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
&shard->lock: 117868635 118361486 0.09 393.01 1250954097.25 10.57 119345882 119780601 0.05 343.35 38313419.51 0.32
------------
&shard->lock 59169196 [<0000000060507011>] __enqueue_entity+0xdc/0x110 &shard->lock 59084239 [<00000000f1c67316>] __dequeue_entity+0x78/0xa0 &shard->lock 108051 [<00000000084a6193>] newidle_balance+0x45a/0x650 ------------ &shard->lock 60028355 [<0000000060507011>] __enqueue_entity+0xdc/0x110 &shard->lock 119882 [<00000000084a6193>] newidle_balance+0x45a/0x650 &shard->lock 58213249 [<00000000f1c67316>] __dequeue_entity+0x78/0xa0
The contention is ~3-4x worse if we don't shard at all. This roughly matches the fact that we had 3 shards on the host where this was collected. This could be addressed in future patch sets by adding a debugfs knob to control the sharding granularity. If we make the shards even smaller (what's in this patch, i.e. a size of 6), the contention goes away almost entirely:
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total waittime-avg acq-bounces acquisitions holdtime-min holdtime-max holdtime-total holdtime-avg
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
&shard->lock: 13839849 13877596 0.08 13.23 5389564.95 0.39 46910241 48069307 0.06 16.40 16534469.35 0.34
------------
&shard->lock 3559 [<00000000ea455dcc>] newidle_balance+0x45a/0x650 &shard->lock 6992418 [<000000002266f400>] __dequeue_entity+0x78/0xa0 &shard->lock 6881619 [<000000002a62f2e0>] __enqueue_entity+0xdc/0x110 ------------ &shard->lock 6640140 [<000000002266f400>] __dequeue_entity+0x78/0xa0 &shard->lock 3523 [<00000000ea455dcc>] newidle_balance+0x45a/0x650 &shard->lock 7233933 [<000000002a62f2e0>] __enqueue_entity+0xdc/0x110
Interestingly, SHARED_RUNQ performs worse than NO_SHARED_RUNQ on the schbench benchmark on Milan, but we contend even more on the rq lock:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total waittime-avg acq-bounces acquisitions holdtime-min holdtime-max holdtime-total holdtime-avg
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
&rq->__lock: 9617614 9656091 0.10 79.64 69665812.00 7.21 18092700 67652829 0.11 82.38 344524858.87 5.09
-----------
&rq->__lock 6301611 [<000000003e63bf26>] task_rq_lock+0x43/0xe0 &rq->__lock 2530807 [<00000000516703f0>] __schedule+0x72/0xaa0 &rq->__lock 109360 [<0000000011be1562>] raw_spin_rq_lock_nested+0xa/0x10 &rq->__lock 178218 [<00000000c38a30f9>] sched_ttwu_pending+0x3d/0x170 ----------- &rq->__lock 3245506 [<00000000516703f0>] __schedule+0x72/0xaa0 &rq->__lock 1294355 [<00000000c38a30f9>] sched_ttwu_pending+0x3d/0x170 &rq->__lock 2837804 [<000000003e63bf26>] task_rq_lock+0x43/0xe0 &rq->__lock 1627866 [<0000000011be1562>] raw_spin_rq_lock_nested+0xa/0x10
..................................................................................................................................................................................................
&shard->lock: 7338558 7343244 0.10 35.97 7173949.14 0.98 30200858 32679623 0.08 35.59 16270584.52 0.50
------------
&shard->lock 2004142 [<00000000f8aa2c91>] __dequeue_entity+0x78/0xa0 &shard->lock 2611264 [<00000000473978cc>] newidle_balance+0x45a/0x650 &shard->lock 2727838 [<0000000028f55bb5>] __enqueue_entity+0xdc/0x110 ------------ &shard->lock 2737232 [<00000000473978cc>] newidle_balance+0x45a/0x650 &shard->lock 1693341 [<00000000f8aa2c91>] __dequeue_entity+0x78/0xa0 &shard->lock 2912671 [<0000000028f55bb5>] __enqueue_entity+0xdc/0x110
...................................................................................................................................................................................................
If we look at the lock stats with SHARED_RUNQ disabled, the rq lock still contends the most, but it's significantly less than with it enabled:
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
class name con-bounces contentions waittime-min waittime-max waittime-total waittime-avg acq-bounces acquisitions holdtime-min holdtime-max holdtime-total holdtime-avg
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
&rq->__lock: 791277 791690 0.12 110.54 4889787.63 6.18 1575996 62390275 0.13 112.66 316262440.56 5.07
-----------
&rq->__lock 263343 [<00000000516703f0>] __schedule+0x72/0xaa0 &rq->__lock 19394 [<0000000011be1562>] raw_spin_rq_lock_nested+0xa/0x10 &rq->__lock 4143 [<000000003b542e83>] __task_rq_lock+0x51/0xf0 &rq->__lock 51094 [<00000000c38a30f9>] sched_ttwu_pending+0x3d/0x170 ----------- &rq->__lock 23756 [<0000000011be1562>] raw_spin_rq_lock_nested+0xa/0x10 &rq->__lock 379048 [<00000000516703f0>] __schedule+0x72/0xaa0 &rq->__lock 677 [<000000003b542e83>] __task_rq_lock+0x51/0xf0 &rq->__lock 47962 [<00000000c38a30f9>] sched_ttwu_pending+0x3d/0x170
In general, the takeaway here is that sharding does help with contention, but it's not necessarily one size fits all, and it's workload dependent. For now, let's include sharding to try and avoid contention, and because it doesn't seem to regress CPUs that don't need it such as the AMD 7950X.
--- kernel/sched/fair.c | 181 +++++++++++++++++++++++++++++-------------- kernel/sched/sched.h | 3 +- 2 files changed, 126 insertions(+), 58 deletions(-)
kernel/sched/fair.c +124 -57
@@ -92,45 +92,44 @@ static int __init setup_sched_thermal_decay_shift(char *str)__setup ("sched_thermal_decay_shift=" , setup_sched_thermal_decay_shift);/** -* struct shared_runq - Per-LLC queue structure for enqueuing and migrating -* runnable tasks within an LLC. -* @list: The list of tasks in the shared_runq. -* @lock: The raw spinlock that synchronizes access to the shared_runq. +* struct shared_runq_shard - A structure containing a task list and a spinlock +* for a subset of cores in a struct shared_runq. +* @list: The list of tasks in the shard. +* @lock: The raw spinlock that synchronizes access to the shard. * * WHAT * ==== * * This structure enables the scheduler to be more aggressively work -* conserving, by placing waking tasks on a per-LLC FIFO queue that can then be -* pulled from when another core in the LLC is going to go idle. +* conserving, by placing waking tasks on a per-LLC FIFO queue shard that can +* then be pulled from when another core in the LLC is going to go idle. +* +* struct rq stores two pointers in its struct cfs_rq: * -* struct rq stores a pointer to its LLC's shared_runq via struct cfs_rq. -* Waking tasks are enqueued in the calling CPU's struct shared_runq in -* __enqueue_entity(), and are opportunistically pulled from the shared_runq -* in newidle_balance(). Tasks enqueued in a shared_runq may be scheduled prior -* to being pulled from the shared_runq, in which case they're simply dequeued -* from the shared_runq in __dequeue_entity(). +* 1. The per-LLC struct shared_runq which contains one or more shards of +* enqueued tasks. +* +* 2. The shard inside of the per-LLC struct shared_runq which contains the +* list of runnable tasks for that shard. +* +* Waking tasks are enqueued in the calling CPU's struct shared_runq_shard in +* __enqueue_entity(), and are opportunistically pulled from the shared_runq in +* newidle_balance(). Pulling from shards is an O(# shards) operation. * * There is currently no task-stealing between shared_runqs in different LLCs, * which means that shared_runq is not fully work conserving. This could be * added at a later time, with tasks likely only being stolen across * shared_runqs on the same NUMA node to avoid violating NUMA affinities. * -* Note that there is a per-CPU allocation of struct shared_runq objects to -* account for the possibility that sched domains are reconfigured during e.g. -* hotplug. In practice, most of these struct shared_runq objects are unused at -* any given time, with the struct shared_runq of a single core per LLC being -* referenced by all other cores in the LLC via a pointer in their struct -* cfs_rq. -* * HOW * === * -* A shared_runq is comprised of a list, and a spinlock for synchronization. -* Given that the critical section for a shared_runq is typically a fast list -* operation, and that the shared_runq is localized to a single LLC, the -* spinlock will typically only be contended on workloads that do little else -* other than hammer the runqueue. +* A struct shared_runq_shard is comprised of a list, and a spinlock for +* synchronization. Given that the critical section for a shared_runq is +* typically a fast list operation, and that the shared_runq_shard is localized +* to a subset of cores on a single LLC (plus other cores in the LLC that pull +* from the shard in newidle_balance()), the spinlock will typically only be +* contended on workloads that do little else other than hammer the runqueue. * * WHY * ===
@@ -144,11 +143,35 @@ __setup("sched_thermal_decay_shift=", setup_sched_thermal_decay_shift);* it, as well as to strike a balance between work conservation, and L3 cache * locality. */ -struct shared_runq { +struct shared_runq_shard {struct list_head list;raw_spinlock_t lock; }____cacheline_aligned ; +/* This would likely work better as a configurable knob via debugfs */ +#define SHARED_RUNQ_SHARD_SZ 6 +#define SHARED_RUNQ_MAX_SHARDS \ + ((NR_CPUS / SHARED_RUNQ_SHARD_SZ) + (NR_CPUS % SHARED_RUNQ_SHARD_SZ != 0)) + +/** +* struct shared_runq - Per-LLC queue structure for enqueuing and migrating +* runnable tasks within an LLC. +* @num_shards: The number of shards currently active in the shared_runq. +* @shards: The shards of the shared_runq. Only @num_shards of these shards are +* active at any given time. +* +* A per-LLC shared_runq that is composed of one of more shards. There is a +* per-CPU allocation of struct shared_runq objects to account for the +* possibility that sched domains are reconfigured during e.g. hotplug. In +* practice, most of these struct shared_runq objects are unused at any given +* time, with the struct shared_runq of a single core per LLC being referenced +* by all other cores in the LLC via a pointer in their struct cfs_rq. +*/ +struct shared_runq { +unsigned int num_shards; +struct shared_runq_shard shards[SHARED_RUNQ_MAX_SHARDS ]; +}____cacheline_aligned ; +#ifdef CONFIG_SMP static DEFINE_PER_CPU(struct shared_runq, shared_runqs);
@@ -159,31 +182,61 @@ static struct shared_runq *rq_shared_runq(struct rq *rq)return rq->cfs.shared_runq; } +static struct shared_runq_shard *rq_shared_runq_shard (struct rq *rq) +{ +return rq->cfs.shard; +} + +static int shared_runq_shard_idx (const struct shared_runq *runq,int cpu) +{ +return (cpu >>1 ) % runq->num_shards; +} +static void shared_runq_reassign_domains (void ) {int i;struct shared_runq *shared_runq;struct rq *rq;struct rq_flags rf; +unsigned int num_shards, shard_idx; + +for_each_possible_cpu (i) { +if (per_cpu (sd_llc_id, i) == i) { + shared_runq = &per_cpu (shared_runqs,per_cpu (sd_llc_id, i)); + + num_shards =per_cpu (sd_llc_size, i) /SHARED_RUNQ_SHARD_SZ ; +if (per_cpu (sd_llc_size, i) %SHARED_RUNQ_SHARD_SZ ) + num_shards++; + shared_runq->num_shards = num_shards; + } + }for_each_possible_cpu (i) { rq =cpu_rq (i); shared_runq = &per_cpu (shared_runqs,per_cpu (sd_llc_id, i)); + shard_idx =shared_runq_shard_idx (shared_runq, i);rq_lock (rq, &rf); rq->cfs.shared_runq = shared_runq; + rq->cfs.shard = &shared_runq->shards[shard_idx];rq_unlock (rq, &rf); } }static void __shared_runq_drain (struct shared_runq *shared_runq) { -struct task_struct *p, *tmp; +unsigned int i; -raw_spin_lock (&shared_runq->lock); -list_for_each_entry_safe (p, tmp, &shared_runq->list, shared_runq_node) -list_del_init (&p->shared_runq_node); -raw_spin_unlock (&shared_runq->lock); +for (i =0 ; i < shared_runq->num_shards; i++) { +struct shared_runq_shard *shard; +struct task_struct *p, *tmp; + + shard = &shared_runq->shards[i]; +raw_spin_lock (&shard->lock); +list_for_each_entry_safe (p, tmp, &shard->list, shared_runq_node) +list_del_init (&p->shared_runq_node); +raw_spin_unlock (&shard->lock); + } }static void update_domains_fair (void )
@@ -237,41 +290,38 @@ void shared_runq_toggle(bool enabling)/* * Disable dequeue _after_ ensuring that all of the shared runqueues * are fully drained. Otherwise, a task could remain enqueued on a -* shared runqueue after the feature was disabled, and could exit -* before drain has completed. +* shard after the feature was disabled, and could exit before drain +* has completed. */ static_branch_disable_cpuslocked (&__shared_runq_force_dequeue); } -static struct task_struct *shared_runq_pop_task (struct rq *rq) +static struct task_struct * +shared_runq_pop_task (struct shared_runq_shard *shard,int target) {struct task_struct *p; -struct shared_runq *shared_runq; - shared_runq =rq_shared_runq (rq); -if (list_empty (&shared_runq->list)) +if (list_empty (&shard->list))return NULL ; -raw_spin_lock (&shared_runq->lock); - p = list_first_entry_or_null(&shared_runq->list, struct task_struct, +raw_spin_lock (&shard->lock); + p = list_first_entry_or_null(&shard->list, struct task_struct, shared_runq_node); -if (p &&is_cpu_allowed (p,cpu_of (rq))) +if (p &&is_cpu_allowed (p, target))list_del_init (&p->shared_runq_node);else p =NULL ; -raw_spin_unlock (&shared_runq->lock); +raw_spin_unlock (&shard->lock);return p; } -static void shared_runq_push_task (struct rq *rq,struct task_struct *p) +static void shared_runq_push_task (struct shared_runq_shard *shard, +struct task_struct *p) { -struct shared_runq *shared_runq; - - shared_runq =rq_shared_runq (rq); -raw_spin_lock (&shared_runq->lock); -list_add_tail (&p->shared_runq_node, &shared_runq->list); -raw_spin_unlock (&shared_runq->lock); +raw_spin_lock (&shard->lock); +list_add_tail (&p->shared_runq_node, &shard->list); +raw_spin_unlock (&shard->lock); }static void shared_runq_enqueue_task (struct rq *rq,struct task_struct *p)
@@ -288,7 +338,7 @@ static void shared_runq_enqueue_task(struct rq *rq, struct task_struct *p) rq->nr_running < 1)return ; -shared_runq_push_task (rq, p); +shared_runq_push_task (rq_shared_runq_shard (rq), p); }static int shared_runq_pick_next_task (struct rq *rq,struct rq_flags *rf)
@@ -296,9 +346,22 @@ static int shared_runq_pick_next_task(struct rq *rq, struct rq_flags *rf)struct task_struct *p =NULL ;struct rq *src_rq;struct rq_flags src_rf; +struct shared_runq *shared_runq; +struct shared_runq_shard *shard; +u32 i, starting_idx, curr_idx, num_shards;int ret =0 , cpu; - p =shared_runq_pop_task (rq); + shared_runq =rq_shared_runq (rq); + num_shards = shared_runq->num_shards; + starting_idx =shared_runq_shard_idx (shared_runq,cpu_of (rq)); +for (i =0 ; i < num_shards; i++) { + curr_idx = (starting_idx + i) % num_shards; + shard = &shared_runq->shards[curr_idx]; + + p =shared_runq_pop_task (shard,cpu_of (rq)); +if (p) +break ; + }if (!p)return 0 ;
@@ -332,8 +395,6 @@ static int shared_runq_pick_next_task(struct rq *rq, struct rq_flags *rf)static void shared_runq_dequeue_task (struct task_struct *p) { -struct shared_runq *shared_runq; -/* * Always dequeue a task if: * - SHARED_RUNQ is enabled
@@ -350,11 +411,13 @@ static void shared_runq_dequeue_task(struct task_struct *p)return ;if (!list_empty (&p->shared_runq_node)) { - shared_runq =rq_shared_runq (task_rq (p)); -raw_spin_lock (&shared_runq->lock); +struct shared_runq_shard *shard; + + shard =rq_shared_runq_shard (task_rq (p)); +raw_spin_lock (&shard->lock);if (likely (!list_empty (&p->shared_runq_node)))list_del_init (&p->shared_runq_node); -raw_spin_unlock (&shared_runq->lock); +raw_spin_unlock (&shard->lock); } }
@@ -13514,8 +13577,9 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)__init void init_sched_fair_class (void ) {#ifdef CONFIG_SMP -int i; +int i, j;struct shared_runq *shared_runq; +struct shared_runq_shard *shard;for_each_possible_cpu (i) {zalloc_cpumask_var_node (&per_cpu (load_balance_mask, i),GFP_KERNEL ,cpu_to_node (i));
@@ -13528,8 +13592,11 @@ __init void init_sched_fair_class(void)INIT_LIST_HEAD (&cpu_rq (i)->cfsb_csd_list);#endif shared_runq = &per_cpu (shared_runqs, i); -INIT_LIST_HEAD (&shared_runq->list); -raw_spin_lock_init (&shared_runq->lock); +for (j =0 ; j <SHARED_RUNQ_MAX_SHARDS ; j++) { + shard = &shared_runq->shards[j]; +INIT_LIST_HEAD (&shard->list); +raw_spin_lock_init (&shard->lock); + } }open_softirq (SCHED_SOFTIRQ , run_rebalance_domains);
kernel/sched/sched.h +2 -1
@@ -599,7 +599,8 @@ struct cfs_rq {#endif #ifdef CONFIG_SMP -struct shared_runq *shared_runq; +struct shared_runq *shared_runq; +struct shared_runq_shard *shard;/* * CFS load tracking */
Signature · 1 line
2.42.1
PATCH v4 6/8 sched: Implement shared runqueue in fair.c
In reply to David VernetOverview
========
The scheduler must constantly strike a balance between work conservation, and avoiding costly migrations which harm performance due to e.g. decreased cache locality. The matter is further complicated by the topology of the system. Migrating a task between cores on the same LLC may be more optimal than keeping a task local to the CPU, whereas migrating a task between LLCs or NUMA nodes may tip the balance in the other direction.
With that in mind, while CFS is by and large mostly a work conserving scheduler, there are certain instances where the scheduler will choose to keep a task local to a CPU, when it would have been more optimal to migrate it to an idle core.
An example of such a workload is the HHVM / web workload at Meta. HHVM is a VM that JITs Hack and PHP code in service of web requests. Like other JIT / compilation workloads, it tends to be heavily CPU bound, and exhibit generally poor cache locality. To try and address this, we set several debugfs (/sys/kernel/debug/sched) knobs on our HHVM workloads:
- migration_cost_ns -> 0
- latency_ns -> 20000000
- min_granularity_ns -> 10000000
- wakeup_granularity_ns -> 12000000
These knobs are intended both to encourage the scheduler to be as work conserving as possible (migration_cost_ns -> 0), and also to keep tasks running for relatively long time slices so as to avoid the overhead of context switching (the other knobs). Collectively, these knobs provide a substantial performance win; resulting in roughly a 20% improvement in throughput. Worth noting, however, is that this improvement is _not_ at full machine saturation.
That said, even with these knobs, we noticed that CPUs were still going idle even when the host was overcommitted. In response, we wrote the "shared runqueue" (SHARED_RUNQ) feature proposed in this patch set. The idea behind SHARED_RUNQ is simple: it enables the scheduler to be more aggressively work conserving by placing a waking task into a sharded per-LLC FIFO queue that can be pulled from by another core in the LLC FIFO queue which can then be pulled from before it goes idle.
With this simple change, we were able to achieve a 1 - 1.6% improvement in throughput, as well as a small, consistent improvement in p95 and p99 latencies, in HHVM. These performance improvements were in addition to the wins from the debugfs knobs mentioned above, and to other benchmarks outlined below in the Results section.
Design
======
Note that the design described here reflects sharding, which will be added in a subsequent patch. The design is described that way in this commit summary as the benchmarks described in the results section below all include sharded SHARED_RUNQ. The patches are not combined into one to ease the burden of review.
The design of SHARED_RUNQ is quite simple. A shared_runq is simply a list of struct shared_runq_shard objects, which itself is simply a struct list_head of tasks, and a spinlock:
struct shared_runq_shard {
struct list_head list;
raw_spinlock_t lock;
} ____cacheline_aligned;struct shared_runq {
u32 num_shards;
struct shared_runq_shard shards[];
} ____cacheline_aligned;We create a struct shared_runq per LLC, ensuring they're in their own cachelines to avoid false sharing between CPUs on different LLCs, and we create a number of struct shared_runq_shard objects that are housed there.
When a task first wakes up, it enqueues itself in the shared_runq_shard of its current LLC at the end of enqueue_task_fair(). Enqueues only happen if the task was not manually migrated to the current core by select_task_rq(), and is not pinned to a specific CPU.
A core will pull a task from the shards in its LLC's shared_runq at the beginning of newidle_balance().
Difference between SHARED_RUNQ and SIS_NODE
===========================================
In [0] Peter proposed a patch that addresses Tejun's observations that when workqueues are targeted towards a specific LLC on his Zen2 machine with small CCXs, that there would be significant idle time due to select_idle_sibling() not considering anything outside of the current LLC.
This patch (SIS_NODE) is essentially the complement to the proposal here. SID_NODE causes waking tasks to look for idle cores in neighboring LLCs on the same die, whereas SHARED_RUNQ causes cores about to go idle to look for enqueued tasks. That said, in its current form, the two features at are a different scope as SIS_NODE searches for idle cores between LLCs, while SHARED_RUNQ enqueues tasks within a single LLC.
The patch was since removed in [1], and we compared the results to shared_Runq (previously called "swqueue") in [2]. SIS_NODE did not outperform SHARED_RUNQ on any of the benchmarks, so we elect to not compare against it again for this v2 patch set.
[0]: https://lore.kernel.org/all/[email protected]/
[1]: https://lore.kernel.org/all/[email protected]/
[2]: https://lore.kernel.org/lkml/[email protected]/
Worth noting as well is that pointed out in [3] that the logic behind including SIS_NODE in the first place should apply to SHARED_RUNQ (meaning that e.g. very small Zen2 CPUs with only 3/4 cores per LLC should benefit from having a single shared_runq stretch across multiple LLCs). I drafted a patch that implements this by having a minimum LLC size for creating a shard, and stretches a shared_runq across multiple LLCs if they're smaller than that size, and sent it to Tejun to test on his Zen2. Tejun reported back that SIS_NODE did not seem to make a difference:
[3]: https://lore.kernel.org/lkml/[email protected]/
o____________o__________o | mean | Variance | o------------o----------o
Vanilla: | 108.84s | 0.0057 |NO_SHARED_RUNQ: | 108.82s | 0.119s |SHARED_RUNQ: | 108.17s | 0.038s |SHARED_RUNQ w/ SIS_NODE: | 108.87s | 0.111s |
o------------o----------o
I similarly tried running kcompile on SHARED_RUNQ with SIS_NODE on my 7950X Zen3, but didn't see any gain relative to plain SHARED_RUNQ.
Conclusion
==========
SHARED_RUNQ in this form provides statistically significant wins for several types of workloads, and various CPU topologies. The reason for this is roughly the same for all workloads: SHARED_RUNQ encourages work conservation inside of a CCX by having a CPU do an O(# per-LLC shards) iteration over the shared_runq shards in an LLC. We could similarly do an O(n) iteration over all of the runqueues in the current LLC when a core is going idle, but that's quite costly (especially for larger LLCs), and sharded SHARED_RUNQ seems to provide a performant middle ground between doing an O(n) walk, and doing an O(1) pull from a single per-LLC shared runq.
While SHARED_RUNQ in this form encourages work conservation, it of course does not guarantee it given that we don't implement any kind of work stealing between shared_runq's. In the future, we could potentially push CPU utilization even higher by enabling work stealing between shared_runq's, likely between CCXs on the same NUMA node.
Appendix
========
Worth noting is that various people's review feedback contributed to this patch. Most notably is likely K Prateek Nayak <[email protected]> who suggested checking this_rq->rd->overload to avoid unnecessary contention and load balancing overhead with the SHARED_RUNQ patches, and also corrected how we were doing idle cost accounting.
--- include/linux/sched.h | 2 + init/init_task.c | 3 + kernel/sched/core.c | 13 ++ kernel/sched/fair.c | 332 +++++++++++++++++++++++++++++++++++++++- kernel/sched/features.h | 4 + kernel/sched/sched.h | 9 ++ kernel/sched/topology.c | 4 +- 7 files changed, 364 insertions(+), 3 deletions(-)
include/linux/sched.h +2 -0
@@ -794,6 +794,8 @@ struct task_struct {unsigned long wakee_flip_decay_ts;struct task_struct *last_wakee; +struct list_head shared_runq_node; +/* * recent_used_cpu is initially set as the last CPU used by a task * that wakes affine another task. Waker/wakee relationships can
init/init_task.c +3 -0
@@ -75,6 +75,9 @@ struct task_struct init_task .stack = init_stack, .usage = REFCOUNT_INIT(2), .flags =PF_KTHREAD , +#ifdef CONFIG_SMP + .shared_runq_node = LIST_HEAD_INIT(init_task.shared_runq_node), +#endif .prio =MAX_PRIO -20 , .static_prio = MAX_PRIO - 20, .normal_prio = MAX_PRIO - 20,
kernel/sched/core.c +13 -0
@@ -4523,6 +4523,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)#ifdef CONFIG_SMP p->wake_entry.u_flags =CSD_TYPE_TTWU ; p->migration_pending =NULL ; +INIT_LIST_HEAD (&p->shared_runq_node);#endif init_sched_mm_cid (p); }
@@ -9713,6 +9714,18 @@ int sched_cpu_deactivate(unsigned int cpu)return 0 ; } +void sched_update_domains (void ) +{ +const struct sched_class *class; + +update_sched_domain_debugfs (); + +for_each_class (class) { +if (class->update_domains) + class->update_domains (); + } +} +static void sched_rq_cpu_starting (unsigned int cpu) {struct rq *rq =cpu_rq (cpu);
kernel/sched/fair.c +330 -2
412 lines not on this page · raw message
kernel/sched/features.h +4 -0
@@ -88,3 +88,7 @@ SCHED_FEAT(UTIL_EST_FASTUP, true) SCHED_FEAT(LATENCY_WARN, false) SCHED_FEAT(HZ_BW, true) + +#ifdef CONFIG_SMP +SCHED_FEAT_CALLBACK(SHARED_RUNQ, false, shared_runq_toggle) +#endif
kernel/sched/sched.h +9 -0
@@ -514,6 +514,12 @@ static inline bool cfs_task_bw_constrained(struct task_struct *p) { return false#endif /* CONFIG_CGROUP_SCHED */ +#ifdef CONFIG_SMP +extern void sched_update_domains (void ); +#else +static inline void sched_update_domains (void ) {} +#endif /* CONFIG_SMP */ +extern void unregister_rt_sched_group (struct task_group *tg);extern void free_rt_sched_group (struct task_group *tg);extern int alloc_rt_sched_group (struct task_group *tg,struct task_group *parent);
@@ -593,6 +599,7 @@ struct cfs_rq {#endif #ifdef CONFIG_SMP +struct shared_runq *shared_runq;/* * CFS load tracking */
@@ -2158,6 +2165,7 @@ static const_debug __maybe_unused unsigned int sysctl_sched_features =#endif /* SCHED_DEBUG */ typedef void (*sched_feat_change_f )(bool enabling); +extern void shared_runq_toggle (bool enabling);extern struct static_key_false sched_numa_balancing;extern struct static_key_false sched_schedstats;
@@ -2317,6 +2325,7 @@ struct sched_class {void (*rq_offline)(struct rq *rq);struct rq *(*find_lock_rq)(struct task_struct *p,struct rq *rq); +void (*update_domains)(void );#endif void (*task_tick)(struct rq *rq,struct task_struct *p,int queued);
kernel/sched/topology.c +3 -1
@@ -2612,6 +2612,8 @@ int __init sched_init_domains(const struct cpumask *cpu_map) doms_cur = &fallback_doms;cpumask_and (doms_cur[0 ], cpu_map,housekeeping_cpumask (HK_TYPE_DOMAIN )); err =build_sched_domains (doms_cur[0 ],NULL ); +if (!err) +sched_update_domains ();return err; }
@@ -2780,7 +2782,7 @@ void partition_sched_domains_locked(int ndoms_new, cpumask_var_t doms_new[], dattr_cur = dattr_new; ndoms_cur = ndoms_new; -update_sched_domain_debugfs (); +sched_update_domains (); }/*
Signature · 1 line
2.42.1
PATCH v4 8/8 sched: Add selftest for SHARED_RUNQ
In reply to David VernetWe want to ensure that SHARED_RUNQ works as expected. Let's add a testcase to the sched/ subdirectory containing SHARED_RUNQ which enables and disables it in a loop, while stressing the system with rcutorture.
--- tools/testing/selftests/sched/Makefile | 5 ++++- tools/testing/selftests/sched/config | 2 ++ tools/testing/selftests/sched/test-swq.sh | 23 +++++++++++++++++++++++ 3 files changed, 29 insertions(+), 1 deletion(-) create mode 100755 tools/testing/selftests/sched/test-swq.sh
tools/testing/selftests/sched/Makefile +4 -1
@@ -9,6 +9,9 @@ CFLAGS += -O2 -Wall -g -I./ $(KHDR_INCLUDES) -Wl,-rpath=./ \ LDLIBS += -lpthread TEST_GEN_FILES := cs_prctl_test -TEST_PROGS := cs_prctl_test +TEST_PROGS := \ + cs_prctl_test \ + test-srq.sh + include ../lib.mk
tools/testing/selftests/sched/config +2 -0
@@ -1 +1,3 @@ CONFIG_SCHED_DEBUG=y +CONFIG_DEBUG_KERNEL=y +CONFIG_RCU_TORTURE_TEST=m
tools/testing/selftests/sched/test-swq.sh new file +23 -0
@@ -0,0 +1,23 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0 +# Copyright (C) 2023 Meta, Inc + +echo "TEST: SHARED_RUNQ stress test ..." + +modprobe rcutorture + +for i in {1..10}; do + echo "Beginning iteration $i" + echo "SHARED_RUNQ" > /sys/kernel/debug/sched/features + sleep 2.3 + echo "NO_SHARED_RUNQ" > /sys/kernel/debug/sched/features + sleep .8 + echo "Completed iteration $i" + echo "" +done + +rmmod rcutorture + +echo "DONE: SHARED_RUNQ stress test completed" + +exit 0
Signature · 1 line
2.42.1
K Prateek Nayak
reviewerHello David,
Quoted 11 lines · David Vernet
This is v4 of the shared runqueue patchset. This patch set is based off of commit
418146e39891("freezer,sched: Cleansaved_statewhen restoring it during thaw") on the sched/core branch of tip.git.In prior versions of this patch set, I was observing consistent and statistically significant wins for several benchmarks when this feature was enabled, such as kernel compile and hackbench. After rebasing onto the latest sched/core on tip.git, I'm no longer observing these wins, and in fact observe some performance loss with
SHARED_RUNQon hackbench. I ended up bisecting this to when EEVDF was merged.
Quoted 4 lines
As I mentioned in [0], our plan for now is to take a step back and re-evaluate how we want to proceed with this patch set. That said, I did want to send this out in the interim in case it could be of interest to anyone else who would like to continue to experiment with it.
I was doing a bunch of testing prior to OSPM in case folks wanted to discuss the results. Leaving the results of SHARED_RUNQ runs from a recent-ish tip below.
tl;dr
- I haven't dug deeper into the regressions but the most prominent one seems to be hackbench with lower number of groups but the picture flips with higher number of groups.
Other benchmarks behave more or less similar to the tip. I'll leave the full results below:
o System Details
- 3rd Generation EPYC System
- 2 x 64C/128T
- NPS1 mode
o Kernels
tip: tip:sched/core at commit 8cec3dd9e593
("sched/core: Simplify code by removing
duplicate #ifdefs")
shared_runq: tip + this series
o Results
==================================================================
Test : hackbench
Units : Normalized time in seconds
Interpretation: Lower is better
Statistic : AMean
==================================================================
Case: tip[pct imp](CV) shared_runq[pct imp](CV)
1-groups 1.00 [ -0.00]( 1.80) 4.49 [-349.19](92.14) 2-groups 1.00 [ -0.00]( 1.76) 1.02 [ -2.17](19.20) 4-groups 1.00 [ -0.00]( 1.82) 0.86 [ 13.53]( 1.37) 8-groups 1.00 [ -0.00]( 1.40) 0.91 [ 8.73]( 2.39) 16-groups 1.00 [ -0.00]( 3.38) 0.91 [ 9.47]( 2.39)
==================================================================
Test : tbench
Units : Normalized throughput
Interpretation: Higher is better
Statistic : AMean
==================================================================
Clients: tip[pct imp](CV) shared_runq[pct imp](CV)
1 1.00 [ 0.00]( 0.44) 1.00 [ -0.39]( 0.53)
2 1.00 [ 0.00]( 0.39) 1.00 [ -0.16]( 0.57)
4 1.00 [ 0.00]( 0.40) 1.00 [ -0.07]( 0.69)
8 1.00 [ 0.00]( 0.16) 0.99 [ -0.67]( 0.45)
16 1.00 [ 0.00]( 3.00) 1.03 [ 2.86]( 1.23)
32 1.00 [ 0.00]( 0.84) 1.00 [ -0.32]( 1.46)
64 1.00 [ 0.00]( 1.66) 0.98 [ -1.60]( 0.79)
128 1.00 [ 0.00]( 1.04) 1.01 [ 0.57]( 0.59)
256 1.00 [ 0.00]( 0.26) 0.98 [ -1.91]( 2.48)
512 1.00 [ 0.00]( 0.15) 1.00 [ 0.22]( 0.16)
1024 1.00 [ 0.00]( 0.20) 1.00 [ -0.37]( 0.02)==================================================================
Test : stream-10
Units : Normalized Bandwidth, MB/s
Interpretation: Higher is better
Statistic : HMean
==================================================================
Test: tip[pct imp](CV) shared_runq[pct imp](CV)
Copy 1.00 [ 0.00]( 6.19) 1.10 [ 9.51]( 4.30) Scale 1.00 [ 0.00]( 6.47) 1.03 [ 2.90]( 2.82) Add 1.00 [ 0.00]( 6.50) 1.04 [ 3.82]( 3.10) Triad 1.00 [ 0.00]( 5.70) 1.01 [ 1.49]( 4.30)
==================================================================
Test : stream-100
Units : Normalized Bandwidth, MB/s
Interpretation: Higher is better
Statistic : HMean
==================================================================
Test: tip[pct imp](CV) shared_runq[pct imp](CV)
Copy 1.00 [ 0.00]( 3.22) 1.04 [ 3.67]( 2.41) Scale 1.00 [ 0.00]( 6.17) 1.03 [ 2.75]( 1.63) Add 1.00 [ 0.00]( 5.12) 1.02 [ 2.42]( 2.10) Triad 1.00 [ 0.00]( 2.29) 1.01 [ 1.11]( 1.59)
==================================================================
Test : netperf
Units : Normalized Througput
Interpretation: Higher is better
Statistic : AMean
==================================================================
Clients: tip[pct imp](CV) shared_runq[pct imp](CV)
1-clients 1.00 [ 0.00]( 0.17) 0.99 [ -0.65]( 0.40) 2-clients 1.00 [ 0.00]( 0.49) 1.00 [ -0.17]( 0.27) 4-clients 1.00 [ 0.00]( 0.65) 1.00 [ 0.09]( 0.69) 8-clients 1.00 [ 0.00]( 0.56) 1.00 [ -0.05]( 0.61) 16-clients 1.00 [ 0.00]( 0.78) 1.00 [ -0.23]( 0.58) 32-clients 1.00 [ 0.00]( 0.62) 0.98 [ -2.22]( 0.76) 64-clients 1.00 [ 0.00]( 1.41) 0.96 [ -3.75]( 1.19) 128-clients 1.00 [ 0.00]( 0.83) 0.98 [ -2.29]( 0.97) 256-clients 1.00 [ 0.00]( 4.60) 0.96 [ -4.18]( 3.02) 512-clients 1.00 [ 0.00](54.18) 0.99 [ -1.36](52.79)
==================================================================
Test : schbench
Units : Normalized 99th percentile latency in us
Interpretation: Lower is better
Statistic : Median
==================================================================
#workers: tip[pct imp](CV) shared_runq[pct imp](CV)
1 1.00 [ -0.00](34.63) 1.40 [-40.00]( 2.38) 2 1.00 [ -0.00]( 2.70) 1.08 [ -8.11]( 7.53) 4 1.00 [ -0.00]( 4.70) 0.93 [ 6.67]( 7.16) 8 1.00 [ -0.00]( 5.09) 0.92 [ 7.55](10.20) 16 1.00 [ -0.00]( 5.08) 0.97 [ 3.39]( 2.00) 32 1.00 [ -0.00]( 2.91) 1.03 [ -3.33]( 2.22) 64 1.00 [ -0.00]( 2.73) 0.99 [ 1.04]( 3.43) 128 1.00 [ -0.00]( 7.89) 0.99 [ 0.69]( 9.65) 256 1.00 [ -0.00](28.55) 0.92 [ 7.94](19.85) 512 1.00 [ -0.00]( 2.11) 1.13 [-12.69]( 6.41)
==================================================================
Test : DeathStarBench
Units : Normalized throughput
Interpretation: Higher is better
Statistic : Mean
==================================================================
Pinning scaling tip shared_runq (pct imp) 1CCD 1 1.00 1.01 (%diff: 1.45%) 2CCD 2 1.00 1.01 (%diff: 1.71%) 4CCD 4 1.00 1.01 (%diff: 1.66%) 8CCD 8 1.00 1.00 (%diff: 0.63%)
--
Quoted 9 lines
[0]: https://lore.kernel.org/all/20231204193001.GA53255@maniforge/
v1 (RFC): https://lore.kernel.org/lkml/[email protected]/
v2: https://lore.kernel.org/lkml/[email protected]/
v3: https://lore.kernel.org/all/[email protected]/[..snip..]
I'll take a deeper look at the regressions soon. I'll update the thread if I find anything interesting in the meantime.
--
Thanks and Regards,
Prateek