From: "Mike Pagano" <mpagano@gentoo.org>
To: gentoo-commits@lists.gentoo.org
Subject: [gentoo-commits] proj/linux-patches:5.14 commit in: /
Date: Sun, 3 Oct 2021 19:14:50 +0000 (UTC) [thread overview]
Message-ID: <1633288460.3840a675683c2df1aea2f9efed23617ce7eb9e01.mpagano@gentoo> (raw)
commit: 3840a675683c2df1aea2f9efed23617ce7eb9e01
Author: Mike Pagano <mpagano <AT> gentoo <DOT> org>
AuthorDate: Sun Oct 3 19:14:20 2021 +0000
Commit: Mike Pagano <mpagano <AT> gentoo <DOT> org>
CommitDate: Sun Oct 3 19:14:20 2021 +0000
URL: https://gitweb.gentoo.org/proj/linux-patches.git/commit/?id=3840a675
Upgrade BMQ and PDS io scheduler to version v5.14-r1
Signed-off-by: Mike Pagano <mpagano <AT> gentoo.org>
0000_README | 2 +-
...=> 5020_BMQ-and-PDS-io-scheduler-v5.14-r3.patch | 284 ++++++++++-----------
2 files changed, 142 insertions(+), 144 deletions(-)
diff --git a/0000_README b/0000_README
index 21444f8..2d15afd 100644
--- a/0000_README
+++ b/0000_README
@@ -115,7 +115,7 @@ Patch: 5010_enable-cpu-optimizations-universal.patch
From: https://github.com/graysky2/kernel_compiler_patch
Desc: Kernel >= 5.8 patch enables gcc = v9+ optimizations for additional CPUs.
-Patch: 5020_BMQ-and-PDS-io-scheduler-v5.14-r1.patch
+Patch: 5020_BMQ-and-PDS-io-scheduler-v5.14-r3.patch
From: https://gitlab.com/alfredchen/linux-prjc
Desc: BMQ(BitMap Queue) Scheduler. A new CPU scheduler developed from PDS(incld). Inspired by the scheduler in zircon.
diff --git a/5020_BMQ-and-PDS-io-scheduler-v5.14-r1.patch b/5020_BMQ-and-PDS-io-scheduler-v5.14-r3.patch
similarity index 98%
rename from 5020_BMQ-and-PDS-io-scheduler-v5.14-r1.patch
rename to 5020_BMQ-and-PDS-io-scheduler-v5.14-r3.patch
index 4c6f75c..99adff7 100644
--- a/5020_BMQ-and-PDS-io-scheduler-v5.14-r1.patch
+++ b/5020_BMQ-and-PDS-io-scheduler-v5.14-r3.patch
@@ -341,6 +341,20 @@ index e5af028c08b4..0a7565d0d3cf 100644
return false;
}
+diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h
+index 8f0f778b7c91..991f2280475b 100644
+--- a/include/linux/sched/topology.h
++++ b/include/linux/sched/topology.h
+@@ -225,7 +225,8 @@ static inline bool cpus_share_cache(int this_cpu, int that_cpu)
+
+ #endif /* !CONFIG_SMP */
+
+-#if defined(CONFIG_ENERGY_MODEL) && defined(CONFIG_CPU_FREQ_GOV_SCHEDUTIL)
++#if defined(CONFIG_ENERGY_MODEL) && defined(CONFIG_CPU_FREQ_GOV_SCHEDUTIL) && \
++ !defined(CONFIG_SCHED_ALT)
+ extern void rebuild_sched_domains_energy(void);
+ #else
+ static inline void rebuild_sched_domains_energy(void)
diff --git a/init/Kconfig b/init/Kconfig
index 55f9f7738ebb..9a9b244d3ca3 100644
--- a/init/Kconfig
@@ -659,10 +673,10 @@ index 978fcfca5871..0425ee149b4d 100644
obj-$(CONFIG_CPU_FREQ_GOV_SCHEDUTIL) += cpufreq_schedutil.o
diff --git a/kernel/sched/alt_core.c b/kernel/sched/alt_core.c
new file mode 100644
-index 000000000000..900889c838ea
+index 000000000000..56aed2b1e42c
--- /dev/null
+++ b/kernel/sched/alt_core.c
-@@ -0,0 +1,7248 @@
+@@ -0,0 +1,7341 @@
+/*
+ * kernel/sched/alt_core.c
+ *
@@ -732,7 +746,7 @@ index 000000000000..900889c838ea
+#define sched_feat(x) (0)
+#endif /* CONFIG_SCHED_DEBUG */
+
-+#define ALT_SCHED_VERSION "v5.14-r1"
++#define ALT_SCHED_VERSION "v5.14-r3"
+
+/* rt_prio(prio) defined in include/linux/sched/rt.h */
+#define rt_task(p) rt_prio((p)->prio)
@@ -1249,6 +1263,101 @@ index 000000000000..900889c838ea
+ update_rq_clock_task(rq, delta);
+}
+
++/*
++ * RQ Load update routine
++ */
++#define RQ_LOAD_HISTORY_BITS (sizeof(s32) * 8ULL)
++#define RQ_UTIL_SHIFT (8)
++#define RQ_LOAD_HISTORY_TO_UTIL(l) (((l) >> (RQ_LOAD_HISTORY_BITS - 1 - RQ_UTIL_SHIFT)) & 0xff)
++
++#define LOAD_BLOCK(t) ((t) >> 17)
++#define LOAD_HALF_BLOCK(t) ((t) >> 16)
++#define BLOCK_MASK(t) ((t) & ((0x01 << 18) - 1))
++#define LOAD_BLOCK_BIT(b) (1UL << (RQ_LOAD_HISTORY_BITS - 1 - (b)))
++#define CURRENT_LOAD_BIT LOAD_BLOCK_BIT(0)
++
++static inline void rq_load_update(struct rq *rq)
++{
++ u64 time = rq->clock;
++ u64 delta = min(LOAD_BLOCK(time) - LOAD_BLOCK(rq->load_stamp),
++ RQ_LOAD_HISTORY_BITS - 1);
++ u64 prev = !!(rq->load_history & CURRENT_LOAD_BIT);
++ u64 curr = !!cpu_rq(rq->cpu)->nr_running;
++
++ if (delta) {
++ rq->load_history = rq->load_history >> delta;
++
++ if (delta < RQ_UTIL_SHIFT) {
++ rq->load_block += (~BLOCK_MASK(rq->load_stamp)) * prev;
++ if (!!LOAD_HALF_BLOCK(rq->load_block) ^ curr)
++ rq->load_history ^= LOAD_BLOCK_BIT(delta);
++ }
++
++ rq->load_block = BLOCK_MASK(time) * prev;
++ } else {
++ rq->load_block += (time - rq->load_stamp) * prev;
++ }
++ if (prev ^ curr)
++ rq->load_history ^= CURRENT_LOAD_BIT;
++ rq->load_stamp = time;
++}
++
++unsigned long rq_load_util(struct rq *rq, unsigned long max)
++{
++ return RQ_LOAD_HISTORY_TO_UTIL(rq->load_history) * (max >> RQ_UTIL_SHIFT);
++}
++
++#ifdef CONFIG_SMP
++unsigned long sched_cpu_util(int cpu, unsigned long max)
++{
++ return rq_load_util(cpu_rq(cpu), max);
++}
++#endif /* CONFIG_SMP */
++
++#ifdef CONFIG_CPU_FREQ
++/**
++ * cpufreq_update_util - Take a note about CPU utilization changes.
++ * @rq: Runqueue to carry out the update for.
++ * @flags: Update reason flags.
++ *
++ * This function is called by the scheduler on the CPU whose utilization is
++ * being updated.
++ *
++ * It can only be called from RCU-sched read-side critical sections.
++ *
++ * The way cpufreq is currently arranged requires it to evaluate the CPU
++ * performance state (frequency/voltage) on a regular basis to prevent it from
++ * being stuck in a completely inadequate performance level for too long.
++ * That is not guaranteed to happen if the updates are only triggered from CFS
++ * and DL, though, because they may not be coming in if only RT tasks are
++ * active all the time (or there are RT tasks only).
++ *
++ * As a workaround for that issue, this function is called periodically by the
++ * RT sched class to trigger extra cpufreq updates to prevent it from stalling,
++ * but that really is a band-aid. Going forward it should be replaced with
++ * solutions targeted more specifically at RT tasks.
++ */
++static inline void cpufreq_update_util(struct rq *rq, unsigned int flags)
++{
++ struct update_util_data *data;
++
++#ifdef CONFIG_SMP
++ rq_load_update(rq);
++#endif
++ data = rcu_dereference_sched(*per_cpu_ptr(&cpufreq_update_util_data,
++ cpu_of(rq)));
++ if (data)
++ data->func(data, rq_clock(rq), flags);
++}
++#else
++static inline void cpufreq_update_util(struct rq *rq, unsigned int flags)
++{
++#ifdef CONFIG_SMP
++ rq_load_update(rq);
++#endif
++}
++#endif /* CONFIG_CPU_FREQ */
++
+#ifdef CONFIG_NO_HZ_FULL
+/*
+ * Tick may be needed by tasks in the runqueue depending on their policy and
@@ -4038,6 +4147,7 @@ index 000000000000..900889c838ea
+ s64 ns = rq->clock_task - p->last_ran;
+
+ p->sched_time += ns;
++ cgroup_account_cputime(p, ns);
+ account_group_exec_runtime(p, ns);
+
+ p->time_slice -= ns;
@@ -4600,6 +4710,7 @@ index 000000000000..900889c838ea
+ if (cpumask_test_cpu(dest_cpu, p->cpus_ptr)) {
+ __SCHED_DEQUEUE_TASK(p, rq, 0, );
+ set_task_cpu(p, dest_cpu);
++ sched_task_sanity_check(p, dest_rq);
+ __SCHED_ENQUEUE_TASK(p, dest_rq, 0);
+ nr_migrated++;
+ }
@@ -5753,11 +5864,7 @@ index 000000000000..900889c838ea
+ * the runqueue. This will be done when the task deboost
+ * itself.
+ */
-+ if (rt_effective_prio(p, newprio) == p->prio) {
-+ __setscheduler_params(p, attr);
-+ retval = 0;
-+ goto unlock;
-+ }
++ newprio = rt_effective_prio(p, newprio);
+ }
+
+ if (!(attr->sched_flags & SCHED_FLAG_KEEP_PARAMS)) {
@@ -6969,7 +7076,6 @@ index 000000000000..900889c838ea
+ struct task_struct *push_task = rq->curr;
+
+ lockdep_assert_held(&rq->lock);
-+ SCHED_WARN_ON(rq->cpu != smp_processor_id());
+
+ /*
+ * Ensure the thing is persistent until balance_push_set(.on = false);
@@ -6977,9 +7083,10 @@ index 000000000000..900889c838ea
+ rq->balance_callback = &balance_push_callback;
+
+ /*
-+ * Only active while going offline.
++ * Only active while going offline and when invoked on the outgoing
++ * CPU.
+ */
-+ if (!cpu_dying(rq->cpu))
++ if (!cpu_dying(rq->cpu) || rq != this_rq())
+ return;
+
+ /*
@@ -7950,10 +8057,10 @@ index 000000000000..1212a031700e
+{}
diff --git a/kernel/sched/alt_sched.h b/kernel/sched/alt_sched.h
new file mode 100644
-index 000000000000..f03af9ab9123
+index 000000000000..289058a09bd5
--- /dev/null
+++ b/kernel/sched/alt_sched.h
-@@ -0,0 +1,692 @@
+@@ -0,0 +1,666 @@
+#ifndef ALT_SCHED_H
+#define ALT_SCHED_H
+
@@ -8153,6 +8260,7 @@ index 000000000000..f03af9ab9123
+ struct rcuwait hotplug_wait;
+#endif
+ unsigned int nr_pinned;
++
+#endif /* CONFIG_SMP */
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+ u64 prev_irq_time;
@@ -8164,6 +8272,11 @@ index 000000000000..f03af9ab9123
+ u64 prev_steal_time_rq;
+#endif /* CONFIG_PARAVIRT_TIME_ACCOUNTING */
+
++ /* For genenal cpu load util */
++ s32 load_history;
++ u64 load_block;
++ u64 load_stamp;
++
+ /* calc_load related fields */
+ unsigned long calc_load_update;
+ long calc_load_active;
@@ -8216,6 +8329,8 @@ index 000000000000..f03af9ab9123
+#endif /* CONFIG_NO_HZ_COMMON */
+};
+
++extern unsigned long rq_load_util(struct rq *rq, unsigned long max);
++
+extern unsigned long calc_load_update;
+extern atomic_long_t calc_load_tasks;
+
@@ -8528,40 +8643,6 @@ index 000000000000..f03af9ab9123
+
+#ifdef CONFIG_CPU_FREQ
+DECLARE_PER_CPU(struct update_util_data __rcu *, cpufreq_update_util_data);
-+
-+/**
-+ * cpufreq_update_util - Take a note about CPU utilization changes.
-+ * @rq: Runqueue to carry out the update for.
-+ * @flags: Update reason flags.
-+ *
-+ * This function is called by the scheduler on the CPU whose utilization is
-+ * being updated.
-+ *
-+ * It can only be called from RCU-sched read-side critical sections.
-+ *
-+ * The way cpufreq is currently arranged requires it to evaluate the CPU
-+ * performance state (frequency/voltage) on a regular basis to prevent it from
-+ * being stuck in a completely inadequate performance level for too long.
-+ * That is not guaranteed to happen if the updates are only triggered from CFS
-+ * and DL, though, because they may not be coming in if only RT tasks are
-+ * active all the time (or there are RT tasks only).
-+ *
-+ * As a workaround for that issue, this function is called periodically by the
-+ * RT sched class to trigger extra cpufreq updates to prevent it from stalling,
-+ * but that really is a band-aid. Going forward it should be replaced with
-+ * solutions targeted more specifically at RT tasks.
-+ */
-+static inline void cpufreq_update_util(struct rq *rq, unsigned int flags)
-+{
-+ struct update_util_data *data;
-+
-+ data = rcu_dereference_sched(*per_cpu_ptr(&cpufreq_update_util_data,
-+ cpu_of(rq)));
-+ if (data)
-+ data->func(data, rq_clock(rq), flags);
-+}
-+#else
-+static inline void cpufreq_update_util(struct rq *rq, unsigned int flags) {}
+#endif /* CONFIG_CPU_FREQ */
+
+#ifdef CONFIG_NO_HZ_FULL
@@ -8764,88 +8845,25 @@ index 000000000000..be3ee4a553ca
+
+static inline void update_rq_time_edge(struct rq *rq) {}
diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
-index 57124614363d..4057e51cef45 100644
+index 57124614363d..f0e9c7543542 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
-@@ -57,6 +57,13 @@ struct sugov_cpu {
- unsigned long bw_dl;
- unsigned long max;
-
-+#ifdef CONFIG_SCHED_ALT
-+ /* For genenal cpu load util */
-+ s32 load_history;
-+ u64 load_block;
-+ u64 load_stamp;
-+#endif
-+
- /* The field below is for single-CPU policies only: */
- #ifdef CONFIG_NO_HZ_COMMON
- unsigned long saved_idle_calls;
-@@ -161,6 +168,7 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy,
- return cpufreq_driver_resolve_freq(policy, freq);
- }
+@@ -167,9 +167,14 @@ static void sugov_get_util(struct sugov_cpu *sg_cpu)
+ unsigned long max = arch_scale_cpu_capacity(sg_cpu->cpu);
+ sg_cpu->max = max;
+#ifndef CONFIG_SCHED_ALT
- static void sugov_get_util(struct sugov_cpu *sg_cpu)
- {
- struct rq *rq = cpu_rq(sg_cpu->cpu);
-@@ -172,6 +180,55 @@ static void sugov_get_util(struct sugov_cpu *sg_cpu)
+ sg_cpu->bw_dl = cpu_bw_dl(rq);
+ sg_cpu->util = effective_cpu_util(sg_cpu->cpu, cpu_util_cfs(rq), max,
FREQUENCY_UTIL, NULL);
- }
-
-+#else /* CONFIG_SCHED_ALT */
-+
-+#define SG_CPU_LOAD_HISTORY_BITS (sizeof(s32) * 8ULL)
-+#define SG_CPU_UTIL_SHIFT (8)
-+#define SG_CPU_LOAD_HISTORY_SHIFT (SG_CPU_LOAD_HISTORY_BITS - 1 - SG_CPU_UTIL_SHIFT)
-+#define SG_CPU_LOAD_HISTORY_TO_UTIL(l) (((l) >> SG_CPU_LOAD_HISTORY_SHIFT) & 0xff)
-+
-+#define LOAD_BLOCK(t) ((t) >> 17)
-+#define LOAD_HALF_BLOCK(t) ((t) >> 16)
-+#define BLOCK_MASK(t) ((t) & ((0x01 << 18) - 1))
-+#define LOAD_BLOCK_BIT(b) (1UL << (SG_CPU_LOAD_HISTORY_BITS - 1 - (b)))
-+#define CURRENT_LOAD_BIT LOAD_BLOCK_BIT(0)
-+
-+static void sugov_get_util(struct sugov_cpu *sg_cpu)
-+{
-+ unsigned long max = arch_scale_cpu_capacity(sg_cpu->cpu);
-+
-+ sg_cpu->max = max;
++#else
+ sg_cpu->bw_dl = 0;
-+ sg_cpu->util = SG_CPU_LOAD_HISTORY_TO_UTIL(sg_cpu->load_history) *
-+ (max >> SG_CPU_UTIL_SHIFT);
-+}
-+
-+static inline void sugov_cpu_load_update(struct sugov_cpu *sg_cpu, u64 time)
-+{
-+ u64 delta = min(LOAD_BLOCK(time) - LOAD_BLOCK(sg_cpu->load_stamp),
-+ SG_CPU_LOAD_HISTORY_BITS - 1);
-+ u64 prev = !!(sg_cpu->load_history & CURRENT_LOAD_BIT);
-+ u64 curr = !!cpu_rq(sg_cpu->cpu)->nr_running;
-+
-+ if (delta) {
-+ sg_cpu->load_history = sg_cpu->load_history >> delta;
-+
-+ if (delta <= SG_CPU_UTIL_SHIFT) {
-+ sg_cpu->load_block += (~BLOCK_MASK(sg_cpu->load_stamp)) * prev;
-+ if (!!LOAD_HALF_BLOCK(sg_cpu->load_block) ^ curr)
-+ sg_cpu->load_history ^= LOAD_BLOCK_BIT(delta);
-+ }
-+
-+ sg_cpu->load_block = BLOCK_MASK(time) * prev;
-+ } else {
-+ sg_cpu->load_block += (time - sg_cpu->load_stamp) * prev;
-+ }
-+ if (prev ^ curr)
-+ sg_cpu->load_history ^= CURRENT_LOAD_BIT;
-+ sg_cpu->load_stamp = time;
-+}
++ sg_cpu->util = rq_load_util(rq, max);
+#endif /* CONFIG_SCHED_ALT */
-+
+ }
+
/**
- * sugov_iowait_reset() - Reset the IO boost status of a CPU.
- * @sg_cpu: the sugov data for the CPU to boost
-@@ -312,13 +369,19 @@ static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; }
+@@ -312,8 +317,10 @@ static inline bool sugov_cpu_is_busy(struct sugov_cpu *sg_cpu) { return false; }
*/
static inline void ignore_dl_rate_limit(struct sugov_cpu *sg_cpu)
{
@@ -8856,27 +8874,7 @@ index 57124614363d..4057e51cef45 100644
}
static inline bool sugov_update_single_common(struct sugov_cpu *sg_cpu,
- u64 time, unsigned int flags)
- {
-+#ifdef CONFIG_SCHED_ALT
-+ sugov_cpu_load_update(sg_cpu, time);
-+#endif /* CONFIG_SCHED_ALT */
-+
- sugov_iowait_boost(sg_cpu, time, flags);
- sg_cpu->last_update = time;
-
-@@ -439,6 +502,10 @@ sugov_update_shared(struct update_util_data *hook, u64 time, unsigned int flags)
-
- raw_spin_lock(&sg_policy->update_lock);
-
-+#ifdef CONFIG_SCHED_ALT
-+ sugov_cpu_load_update(sg_cpu, time);
-+#endif /* CONFIG_SCHED_ALT */
-+
- sugov_iowait_boost(sg_cpu, time, flags);
- sg_cpu->last_update = time;
-
-@@ -599,6 +666,7 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy)
+@@ -599,6 +606,7 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy)
}
ret = sched_setattr_nocheck(thread, &attr);
@@ -8884,7 +8882,7 @@ index 57124614363d..4057e51cef45 100644
if (ret) {
kthread_stop(thread);
pr_warn("%s: failed to set SCHED_DEADLINE\n", __func__);
-@@ -833,7 +901,9 @@ cpufreq_governor_init(schedutil_gov);
+@@ -833,7 +841,9 @@ cpufreq_governor_init(schedutil_gov);
#ifdef CONFIG_ENERGY_MODEL
static void rebuild_sd_workfn(struct work_struct *work)
{
next reply other threads:[~2021-10-03 19:14 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-03 19:14 Mike Pagano [this message]
-- strict thread matches above, loose matches on Subject: below --
2021-11-21 21:14 [gentoo-commits] proj/linux-patches:5.14 commit in: / Mike Pagano
2021-11-21 20:38 Mike Pagano
2021-11-19 0:18 Mike Pagano
2021-11-18 15:32 Mike Pagano
2021-11-17 11:59 Mike Pagano
2021-11-12 14:19 Mike Pagano
2021-11-06 13:41 Mike Pagano
2021-11-02 19:30 Mike Pagano
2021-10-27 11:56 Mike Pagano
2021-10-21 12:16 Mike Pagano
2021-10-20 13:22 Mike Pagano
2021-10-18 21:17 Mike Pagano
2021-10-17 13:10 Mike Pagano
2021-10-13 16:15 Alice Ferrazzi
2021-10-09 21:30 Mike Pagano
2021-10-07 10:36 Mike Pagano
2021-09-30 10:48 Mike Pagano
2021-09-26 14:11 Mike Pagano
2021-09-22 11:37 Mike Pagano
2021-09-20 22:01 Mike Pagano
2021-09-18 16:06 Mike Pagano
2021-09-17 12:48 Mike Pagano
2021-09-17 12:40 Mike Pagano
2021-09-16 11:03 Mike Pagano
2021-09-15 11:58 Mike Pagano
2021-09-14 15:37 Mike Pagano
2021-09-12 14:36 Mike Pagano
2021-09-08 12:39 Alice Ferrazzi
2021-09-03 11:52 Mike Pagano
2021-09-03 11:17 Mike Pagano
2021-09-03 9:15 Alice Ferrazzi
2021-08-30 17:23 Mike Pagano
2021-08-25 16:30 Mike Pagano
2021-08-25 16:24 Mike Pagano
2021-08-25 16:24 Mike Pagano
2021-08-25 12:25 Mike Pagano
2021-08-18 15:03 Mike Pagano
2021-08-18 15:03 Mike Pagano
2021-08-02 22:32 Mike Pagano
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1633288460.3840a675683c2df1aea2f9efed23617ce7eb9e01.mpagano@gentoo \
--to=mpagano@gentoo.org \
--cc=gentoo-commits@lists.gentoo.org \
--cc=gentoo-dev@lists.gentoo.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox