sched_ext: Maintain an online cid mask in the scheduler arenagrafted

Schedulers on the default cid mapping treat [0, nr_online_cids) as the
online set and restart on hotplug. Schedulers that install their own mapping
with scx_bpf_cid_override() have no way to learn which cids are online: the
count no longer identifies members and the CPU-form cpumask is unusable from
cid programs. This is an obvious hole in the cid API.

Add scx_bpf_online_cmask(), a kernel-maintained cmask in the scheduler's
arena, allocated alongside the per-CPU scratch masks and populated after the
cid mapping is finalized and before ops.init(), for child schedulers too.
The pointer stays valid through ops.exit() with no reference to take. It is
the arena offset as a void pointer, the same form struct_ops arena arguments
arrive in. The verifier types the void return as a scalar for the program's
arena cast.

The mask follows the SCX hotplug notifications: seeded from cpu_active_mask
and updated before ops.cid_online/offline() runs, so it lags cpu_online_mask
only inside a hotplug transition. Updates walk the scheduler list under the
lock that also serializes unlinking. Reads are live, not atomic snapshots.
Root initialization excludes hotplug.

v2: Reworded the getter kerneldoc (Andrea Righi).

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Andrea Righi <arighi@nvidia.com>

1 files changed, 228 insertions, 0 deletions

diff --git a/drivers/opp/cpu.c b/drivers/opp/cpu.c
new file mode 100644
index 000000000..0ea69c5c5
--- /dev/null
+++ b/drivers/opp/cpu.c
@@ -0,0 +1,228 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Generic OPP helper interface for CPU device
+ *
+ * Copyright (C) 2009-2014 Texas Instruments Incorporated.
+ *	Nishanth Menon
+ *	Romit Dasgupta
+ *	Kevin Hilman
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/cpu.h>
+#include <linux/cpufreq.h>
+#include <linux/err.h>
+#include <linux/errno.h>
+#include <linux/export.h>
+#include <linux/slab.h>
+
+#include "opp.h"
+
+#ifdef CONFIG_CPU_FREQ
+
+/**
+ * dev_pm_opp_init_cpufreq_table() - create a cpufreq table for a device
+ * @dev:	device for which we do this operation
+ * @opp_table:	Cpufreq table returned back to caller
+ *
+ * Generate a cpufreq table for a provided device- this assumes that the
+ * opp table is already initialized and ready for usage.
+ *
+ * This function allocates required memory for the cpufreq table. It is
+ * expected that the caller does the required maintenance such as freeing
+ * the table as required.
+ *
+ * Returns -EINVAL for bad pointers, -ENODEV if the device is not found, -ENOMEM
+ * if no memory available for the operation (table is not populated), returns 0
+ * if successful and table is populated.
+ *
+ * WARNING: It is  important for the callers to ensure refreshing their copy of
+ * the table if any of the mentioned functions have been invoked in the interim.
+ */
+int dev_pm_opp_init_cpufreq_table(struct device *dev,
+				  struct cpufreq_frequency_table **opp_table)
+{
+	struct cpufreq_frequency_table *freq_table = NULL;
+	int i, max_opps, ret = 0;
+	unsigned long rate;
+
+	max_opps = dev_pm_opp_get_opp_count(dev);
+	if (max_opps <= 0)
+		return max_opps ? max_opps : -ENODATA;
+
+	freq_table = kzalloc_objs(*freq_table, (max_opps + 1));
+	if (!freq_table)
+		return -ENOMEM;
+
+	for (i = 0, rate = 0; i < max_opps; i++, rate++) {
+		/* find next rate */
+		struct dev_pm_opp *opp __free(put_opp) =
+			dev_pm_opp_find_freq_ceil(dev, &rate);
+
+		if (IS_ERR(opp)) {
+			ret = PTR_ERR(opp);
+			goto out;
+		}
+		freq_table[i].driver_data = i;
+		freq_table[i].frequency = rate / 1000;
+
+		/* Is Boost/turbo opp ? */
+		if (dev_pm_opp_is_turbo(opp))
+			freq_table[i].flags = CPUFREQ_BOOST_FREQ;
+	}
+
+	freq_table[i].driver_data = i;
+	freq_table[i].frequency = CPUFREQ_TABLE_END;
+
+	*opp_table = &freq_table[0];
+
+out:
+	if (ret)
+		kfree(freq_table);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_opp_init_cpufreq_table);
+
+/**
+ * dev_pm_opp_free_cpufreq_table() - free the cpufreq table
+ * @dev:	device for which we do this operation
+ * @opp_table:	table to free
+ *
+ * Free up the table allocated by dev_pm_opp_init_cpufreq_table
+ */
+void dev_pm_opp_free_cpufreq_table(struct device *dev,
+				   struct cpufreq_frequency_table **opp_table)
+{
+	if (!opp_table)
+		return;
+
+	kfree(*opp_table);
+	*opp_table = NULL;
+}
+EXPORT_SYMBOL_GPL(dev_pm_opp_free_cpufreq_table);
+#endif	/* CONFIG_CPU_FREQ */
+
+void _dev_pm_opp_cpumask_remove_table(const struct cpumask *cpumask,
+				      int last_cpu)
+{
+	struct device *cpu_dev;
+	int cpu;
+
+	WARN_ON(cpumask_empty(cpumask));
+
+	for_each_cpu(cpu, cpumask) {
+		if (cpu == last_cpu)
+			break;
+
+		cpu_dev = get_cpu_device(cpu);
+		if (!cpu_dev) {
+			pr_err("%s: failed to get cpu%d device\n", __func__,
+			       cpu);
+			continue;
+		}
+
+		dev_pm_opp_remove_table(cpu_dev);
+	}
+}
+
+/**
+ * dev_pm_opp_cpumask_remove_table() - Removes OPP table for @cpumask
+ * @cpumask:	cpumask for which OPP table needs to be removed
+ *
+ * This removes the OPP tables for CPUs present in the @cpumask.
+ * This should be used to remove all the OPPs entries associated with
+ * the cpus in @cpumask.
+ */
+void dev_pm_opp_cpumask_remove_table(const struct cpumask *cpumask)
+{
+	_dev_pm_opp_cpumask_remove_table(cpumask, -1);
+}
+EXPORT_SYMBOL_GPL(dev_pm_opp_cpumask_remove_table);
+
+/**
+ * dev_pm_opp_set_sharing_cpus() - Mark OPP table as shared by few CPUs
+ * @cpu_dev:	CPU device for which we do this operation
+ * @cpumask:	cpumask of the CPUs which share the OPP table with @cpu_dev
+ *
+ * This marks OPP table of the @cpu_dev as shared by the CPUs present in
+ * @cpumask.
+ *
+ * Returns -ENODEV if OPP table isn't already present.
+ */
+int dev_pm_opp_set_sharing_cpus(struct device *cpu_dev,
+				const struct cpumask *cpumask)
+{
+	struct opp_device *opp_dev;
+	struct device *dev;
+	int cpu;
+
+	struct opp_table *opp_table __free(put_opp_table) =
+		_find_opp_table(cpu_dev);
+
+	if (IS_ERR(opp_table))
+		return PTR_ERR(opp_table);
+
+	for_each_cpu(cpu, cpumask) {
+		if (cpu == cpu_dev->id)
+			continue;
+
+		dev = get_cpu_device(cpu);
+		if (!dev) {
+			dev_err(cpu_dev, "%s: failed to get cpu%d device\n",
+				__func__, cpu);
+			continue;
+		}
+
+		opp_dev = _add_opp_dev(dev, opp_table);
+		if (!opp_dev) {
+			dev_err(dev, "%s: failed to add opp-dev for cpu%d device\n",
+				__func__, cpu);
+			continue;
+		}
+
+		/* Mark opp-table as multiple CPUs are sharing it now */
+		opp_table->shared_opp = OPP_TABLE_ACCESS_SHARED;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dev_pm_opp_set_sharing_cpus);
+
+/**
+ * dev_pm_opp_get_sharing_cpus() - Get cpumask of CPUs sharing OPPs with @cpu_dev
+ * @cpu_dev:	CPU device for which we do this operation
+ * @cpumask:	cpumask to update with information of sharing CPUs
+ *
+ * This updates the @cpumask with CPUs that are sharing OPPs with @cpu_dev.
+ *
+ * Returns -ENODEV if OPP table isn't already present and -EINVAL if the OPP
+ * table's status is access-unknown.
+ */
+int dev_pm_opp_get_sharing_cpus(struct device *cpu_dev, struct cpumask *cpumask)
+{
+	struct opp_device *opp_dev;
+
+	struct opp_table *opp_table __free(put_opp_table) =
+		_find_opp_table(cpu_dev);
+
+	if (IS_ERR(opp_table))
+		return PTR_ERR(opp_table);
+
+	if (opp_table->shared_opp == OPP_TABLE_ACCESS_UNKNOWN)
+		return -EINVAL;
+
+	cpumask_clear(cpumask);
+
+	if (opp_table->shared_opp == OPP_TABLE_ACCESS_SHARED) {
+		guard(mutex)(&opp_table->lock);
+		list_for_each_entry(opp_dev, &opp_table->dev_list, node)
+			cpumask_set_cpu(opp_dev->dev->id, cpumask);
+	} else {
+		cpumask_set_cpu(cpu_dev->id, cpumask);
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dev_pm_opp_get_sharing_cpus);