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, 377 insertions, 0 deletions

diff --git a/lib/raid/raid6/algos.c b/lib/raid/raid6/algos.c
new file mode 100644
index 000000000..6f5c89ab2
--- /dev/null
+++ b/lib/raid/raid6/algos.c
@@ -0,0 +1,377 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2002 H. Peter Anvin - All Rights Reserved
+ *
+ * Algorithm list and algorithm selection for RAID-6
+ */
+
+#include <linux/module.h>
+#include <linux/gfp.h>
+#include <linux/raid/pq.h>
+#include <linux/slab.h>
+#include <linux/static_call.h>
+#include <kunit/visibility.h>
+#include "algos.h"
+
+#define RAID6_MAX_ALGOS		16
+static const struct raid6_calls *raid6_algos[RAID6_MAX_ALGOS];
+static unsigned int raid6_nr_algos;
+static const struct raid6_recov_calls *raid6_recov_algo;
+
+/* Selected algorithm */
+DEFINE_STATIC_CALL_NULL(raid6_gen_syndrome_impl, *raid6_intx1.gen_syndrome);
+DEFINE_STATIC_CALL_NULL(raid6_xor_syndrome_impl, *raid6_intx1.xor_syndrome);
+DEFINE_STATIC_CALL_NULL(raid6_recov_2data_impl, *raid6_recov_intx1.data2);
+DEFINE_STATIC_CALL_NULL(raid6_recov_datap_impl, *raid6_recov_intx1.datap);
+
+/**
+ * raid6_gen_syndrome - generate RAID6 P/Q parity
+ * @disks:	number of "disks" to operate on including parity
+ * @bytes:	length in bytes of each vector
+ * @ptrs:	@disks size array of memory pointers
+ *
+ * Generate @bytes worth of RAID6 P and Q parity in @ptrs[@disks - 2] and
+ * @ptrs[@disks - 1] respectively from the memory pointed to by @ptrs[0] to
+ * @ptrs[@disks - 3].
+ *
+ * @disks must be at least 4, and the memory pointed to by each member of @ptrs
+ * must be at least 64-byte aligned.  @bytes must be non-zero and a multiple of
+ * 512.
+ *
+ * See https://kernel.org/pub/linux/kernel/people/hpa/raid6.pdf for underlying
+ * algorithm.
+ */
+void raid6_gen_syndrome(int disks, size_t bytes, void **ptrs)
+{
+	WARN_ON_ONCE(!in_task() || irqs_disabled() || softirq_count());
+	WARN_ON_ONCE(bytes & 511);
+	WARN_ON_ONCE(disks < RAID6_MIN_DISKS);
+
+	static_call(raid6_gen_syndrome_impl)(disks, bytes, ptrs);
+}
+EXPORT_SYMBOL_GPL(raid6_gen_syndrome);
+
+/**
+ * raid6_xor_syndrome - update RAID6 P/Q parity
+ * @disks:	number of "disks" to operate on including parity
+ * @start:	first index into @disk to update
+ * @stop:	last index into @disk to update
+ * @bytes:	length in bytes of each vector
+ * @ptrs:	@disks size array of memory pointers
+ *
+ * Update @bytes worth of RAID6 P and Q parity in @ptrs[@disks - 2] and
+ * @ptrs[@disks - 1] respectively for the memory pointed to by
+ * @ptrs[@start..@stop].
+ *
+ * This is used to update parity in place using the following sequence:
+ *
+ * 1) call raid6_xor_syndrome(disk, start, stop, ...) for the existing data.
+ * 2) update the the data in @ptrs[@start..@stop].
+ * 3) call raid6_xor_syndrome(disk, start, stop, ...) for the new data.
+ *
+ * Data between @start and @stop that is not changed should be filled
+ * with a pointer to the kernel zero page.
+ *
+ * @disks must be at least 4, and the memory pointed to by each member of @ptrs
+ * must be at least 64-byte aligned.  @bytes must be non-zero and a multiple of
+ * 512.  @stop must be larger or equal to @start.
+ */
+void raid6_xor_syndrome(int disks, int start, int stop, size_t bytes,
+		void **ptrs)
+{
+	WARN_ON_ONCE(!in_task() || irqs_disabled() || softirq_count());
+	WARN_ON_ONCE(bytes & 511);
+	WARN_ON_ONCE(disks < RAID6_MIN_DISKS);
+	WARN_ON_ONCE(stop < start);
+
+	static_call(raid6_xor_syndrome_impl)(disks, start, stop, bytes, ptrs);
+}
+EXPORT_SYMBOL_GPL(raid6_xor_syndrome);
+
+/*
+ * raid6_can_xor_syndrome - check if raid6_xor_syndrome() can be used
+ *
+ * Returns %true if raid6_can_xor_syndrome() can be used, else %false.
+ */
+bool raid6_can_xor_syndrome(void)
+{
+	return !!static_call_query(raid6_xor_syndrome_impl);
+}
+EXPORT_SYMBOL_GPL(raid6_can_xor_syndrome);
+
+/**
+ * raid6_recov_2data - recover two missing data disks
+ * @disks:	number of "disks" to operate on including parity
+ * @bytes:	length in bytes of each vector
+ * @faila:	first failed data disk index
+ * @failb:	second failed data disk index
+ * @ptrs:	@disks size array of memory pointers
+ *
+ * Rebuild @bytes of missing data in @ptrs[@faila] and @ptrs[@failb] from the
+ * data in the remaining disks and the two parities pointed to by the other
+ * indices between 0 and @disks - 1 in @ptrs.  @disks includes the data disks
+ * and the two parities.  @faila must be smaller than @failb.
+ *
+ * Memory pointed to by each pointer in @ptrs must be page aligned and is
+ * limited to %PAGE_SIZE.
+ */
+void raid6_recov_2data(int disks, size_t bytes, int faila, int failb,
+		void **ptrs)
+{
+	WARN_ON_ONCE(!in_task() || irqs_disabled() || softirq_count());
+	WARN_ON_ONCE(bytes & 511);
+	WARN_ON_ONCE(bytes > PAGE_SIZE);
+	WARN_ON_ONCE(failb <= faila);
+
+	static_call(raid6_recov_2data_impl)(disks, bytes, faila, failb, ptrs);
+}
+EXPORT_SYMBOL_GPL(raid6_recov_2data);
+
+/**
+ * raid6_recov_datap - recover a missing data disk and missing P-parity
+ * @disks:	number of "disks" to operate on including parity
+ * @bytes:	length in bytes of each vector
+ * @faila:	failed data disk index
+ * @ptrs:	@disks size array of memory pointers
+ *
+ * Rebuild @bytes of missing data in @ptrs[@faila] and the missing P-parity in
+ * @ptrs[@disks - 2] from the data in the remaining disks and the Q-parity
+ * pointed to by the other indices between 0 and @disks - 1 in @ptrs.  @disks
+ * includes the data disks and the two parities.
+ *
+ * Memory pointed to by each pointer in @ptrs must be page aligned and is
+ * limited to %PAGE_SIZE.
+ */
+void raid6_recov_datap(int disks, size_t bytes, int faila, void **ptrs)
+{
+	WARN_ON_ONCE(!in_task() || irqs_disabled() || softirq_count());
+	WARN_ON_ONCE(bytes & 511);
+	WARN_ON_ONCE(bytes > PAGE_SIZE);
+
+	static_call(raid6_recov_datap_impl)(disks, bytes, faila, ptrs);
+}
+EXPORT_SYMBOL_GPL(raid6_recov_datap);
+
+#define RAID6_TIME_JIFFIES_LG2	4
+#define RAID6_TEST_DISKS	8
+
+static int raid6_choose_gen(void *(*const dptrs)[RAID6_TEST_DISKS],
+		const int disks)
+{
+	/* work on the second half of the disks */
+	int start = (disks >> 1) - 1, stop = disks - 3;
+	const struct raid6_calls *best = NULL;
+	unsigned long bestgenperf = 0;
+	unsigned int i;
+
+	for (i = 0; i < raid6_nr_algos; i++) {
+		const struct raid6_calls *algo = raid6_algos[i];
+		unsigned long perf = 0, j0, j1;
+
+		preempt_disable();
+		j0 = jiffies;
+		while ((j1 = jiffies) == j0)
+			cpu_relax();
+		while (time_before(jiffies,
+				    j1 + (1<<RAID6_TIME_JIFFIES_LG2))) {
+			algo->gen_syndrome(disks, PAGE_SIZE, *dptrs);
+			perf++;
+		}
+		preempt_enable();
+
+		if (perf > bestgenperf) {
+			bestgenperf = perf;
+			best = algo;
+		}
+		pr_info("raid6: %-8s gen() %5ld MB/s\n", algo->name,
+			(perf * HZ * (disks-2)) >>
+			(20 - PAGE_SHIFT + RAID6_TIME_JIFFIES_LG2));
+	}
+
+	if (!best) {
+		pr_err("raid6: Yikes! No algorithm found!\n");
+		return -EINVAL;
+	}
+
+	static_call_update(raid6_gen_syndrome_impl, best->gen_syndrome);
+	static_call_update(raid6_xor_syndrome_impl, best->xor_syndrome);
+
+	pr_info("raid6: using algorithm %s gen() %ld MB/s\n",
+		best->name,
+		(bestgenperf * HZ * (disks - 2)) >>
+		(20 - PAGE_SHIFT + RAID6_TIME_JIFFIES_LG2));
+
+	if (best->xor_syndrome) {
+		unsigned long perf = 0, j0, j1;
+
+		preempt_disable();
+		j0 = jiffies;
+		while ((j1 = jiffies) == j0)
+			cpu_relax();
+		while (time_before(jiffies,
+				   j1 + (1 << RAID6_TIME_JIFFIES_LG2))) {
+			best->xor_syndrome(disks, start, stop,
+					   PAGE_SIZE, *dptrs);
+			perf++;
+		}
+		preempt_enable();
+
+		pr_info("raid6: .... xor() %ld MB/s, rmw enabled\n",
+			(perf * HZ * (disks - 2)) >>
+			(20 - PAGE_SHIFT + RAID6_TIME_JIFFIES_LG2 + 1));
+	}
+
+	return 0;
+}
+
+
+/* Try to pick the best algorithm */
+/* This code uses the gfmul table as convenient data set to abuse */
+
+static int __init raid6_select_algo(void)
+{
+	const int disks = RAID6_TEST_DISKS;
+	char *disk_ptr, *p;
+	void *dptrs[RAID6_TEST_DISKS];
+	int i, cycle;
+	int error;
+
+	if (!IS_ENABLED(CONFIG_RAID6_PQ_BENCHMARK) || raid6_nr_algos == 1) {
+		pr_info("raid6: skipped pq benchmark and selected %s\n",
+			raid6_algos[raid6_nr_algos - 1]->name);
+		static_call_update(raid6_gen_syndrome_impl,
+				raid6_algos[raid6_nr_algos - 1]->gen_syndrome);
+		static_call_update(raid6_xor_syndrome_impl,
+				raid6_algos[raid6_nr_algos - 1]->xor_syndrome);
+		return 0;
+	}
+
+	/* prepare the buffer and fill it circularly with gfmul table */
+	disk_ptr = kmalloc(PAGE_SIZE * RAID6_TEST_DISKS, GFP_KERNEL);
+	if (!disk_ptr) {
+		pr_err("raid6: Yikes!  No memory available.\n");
+		return -ENOMEM;
+	}
+
+	p = disk_ptr;
+	for (i = 0; i < disks; i++)
+		dptrs[i] = p + PAGE_SIZE * i;
+
+	cycle = ((disks - 2) * PAGE_SIZE) / 65536;
+	for (i = 0; i < cycle; i++) {
+		memcpy(p, raid6_gfmul, 65536);
+		p += 65536;
+	}
+
+	if ((disks - 2) * PAGE_SIZE % 65536)
+		memcpy(p, raid6_gfmul, (disks - 2) * PAGE_SIZE % 65536);
+
+	/* select raid gen_syndrome function */
+	error = raid6_choose_gen(&dptrs, disks);
+
+	kfree(disk_ptr);
+
+	return error;
+}
+
+/*
+ * Register a RAID6 P/Q generation algorithm.  The most optimized/unrolled
+ * implementation should be registered last so it will be selected when the
+ * boot-time benchmark is disabled.
+ */
+void __init raid6_algo_add(const struct raid6_calls *algo)
+{
+	if (WARN_ON_ONCE(raid6_nr_algos == RAID6_MAX_ALGOS))
+		return;
+	raid6_algos[raid6_nr_algos++] = algo;
+}
+
+void __init raid6_algo_add_default(void)
+{
+	raid6_algo_add(&raid6_intx1);
+	raid6_algo_add(&raid6_intx2);
+	raid6_algo_add(&raid6_intx4);
+	raid6_algo_add(&raid6_intx8);
+}
+
+void __init raid6_recov_algo_add(const struct raid6_recov_calls *algo)
+{
+	if (WARN_ON_ONCE(raid6_recov_algo))
+		return;
+	raid6_recov_algo = algo;
+}
+
+#ifdef CONFIG_RAID6_PQ_ARCH
+#include "pq_arch.h"
+#else
+static inline void arch_raid6_init(void)
+{
+	raid6_algo_add_default();
+}
+#endif /* CONFIG_RAID6_PQ_ARCH */
+
+static int __init raid6_init(void)
+{
+	/*
+	 * Architectures providing arch_raid6_init must add all PQ generation
+	 * algorithms they want to consider in arch_raid6_init(), including
+	 * the generic ones using raid6_algo_add_default() if wanted.
+	 */
+	arch_raid6_init();
+
+	/*
+	 * Architectures don't have to set a recovery algorithm, we'll just pick
+	 * the generic integer one if none was set.
+	 */
+	if (!raid6_recov_algo)
+		raid6_recov_algo = &raid6_recov_intx1;
+	static_call_update(raid6_recov_2data_impl, raid6_recov_algo->data2);
+	static_call_update(raid6_recov_datap_impl, raid6_recov_algo->datap);
+	pr_info("raid6: using %s recovery algorithm\n", raid6_recov_algo->name);
+
+	return raid6_select_algo();
+}
+
+static void __exit raid6_exit(void)
+{
+}
+
+subsys_initcall(raid6_init);
+module_exit(raid6_exit);
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("RAID6 Q-syndrome calculations");
+
+#if IS_ENABLED(CONFIG_RAID6_PQ_KUNIT_TEST)
+const struct raid6_calls *raid6_algo_find(unsigned int idx)
+{
+	if (idx >= raid6_nr_algos) {
+		/*
+		 * Always include the simplest generic integer implementation in
+		 * the unit tests as a baseline.
+		 */
+		if (idx == raid6_nr_algos &&
+		    raid6_algos[0] != &raid6_intx1)
+			return &raid6_intx1;
+		return NULL;
+	}
+	return raid6_algos[idx];
+}
+EXPORT_SYMBOL_IF_KUNIT(raid6_algo_find);
+
+const struct raid6_recov_calls *raid6_recov_algo_find(unsigned int idx)
+{
+	switch (idx) {
+	case 0:
+		/* always test the generic integer implementation */
+		return &raid6_recov_intx1;
+	case 1:
+		/* test the optimized implementation if there is one */
+		if (raid6_recov_algo != &raid6_recov_intx1)
+			return raid6_recov_algo;
+		return NULL;
+	default:
+		return NULL;
+	}
+}
+EXPORT_SYMBOL_IF_KUNIT(raid6_recov_algo_find);
+#endif /* CONFIG_RAID6_PQ_KUNIT_TEST */