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, 139 insertions, 0 deletions

diff --git a/tools/perf/util/mem2node.c b/tools/perf/util/mem2node.c
new file mode 100644
index 000000000..51a2292cb
--- /dev/null
+++ b/tools/perf/util/mem2node.c
@@ -0,0 +1,139 @@
+#include <errno.h>
+#include <inttypes.h>
+#include <asm/bug.h>
+#include <linux/bitmap.h>
+#include <linux/kernel.h>
+#include <linux/zalloc.h>
+#include "debug.h"
+#include "env.h"
+#include "mem2node.h"
+
+struct phys_entry {
+	struct rb_node	rb_node;
+	u64	start;
+	u64	end;
+	u64	node;
+};
+
+static void phys_entry__insert(struct phys_entry *entry, struct rb_root *root)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent = NULL;
+	struct phys_entry *e;
+
+	while (*p != NULL) {
+		parent = *p;
+		e = rb_entry(parent, struct phys_entry, rb_node);
+
+		if (entry->start < e->start)
+			p = &(*p)->rb_left;
+		else
+			p = &(*p)->rb_right;
+	}
+
+	rb_link_node(&entry->rb_node, parent, p);
+	rb_insert_color(&entry->rb_node, root);
+}
+
+static void
+phys_entry__init(struct phys_entry *entry, u64 start, u64 bsize, u64 node)
+{
+	entry->start = start;
+	entry->end   = start + bsize;
+	entry->node  = node;
+	RB_CLEAR_NODE(&entry->rb_node);
+}
+
+int mem2node__init(struct mem2node *map, struct perf_env *env)
+{
+	struct memory_node *n, *nodes = &env->memory_nodes[0];
+	struct phys_entry *entries, *tmp_entries;
+	u64 bsize = env->memory_bsize;
+	int i, j = 0, max = 0;
+
+	memset(map, 0x0, sizeof(*map));
+	map->root = RB_ROOT;
+
+	for (i = 0; i < env->nr_memory_nodes; i++) {
+		n = &nodes[i];
+		max += bitmap_weight(n->set, n->size);
+	}
+
+	entries = calloc(max, sizeof(*entries));
+	if (!entries)
+		return -ENOMEM;
+
+	for (i = 0; i < env->nr_memory_nodes; i++) {
+		u64 bit;
+
+		n = &nodes[i];
+
+		for (bit = 0; bit < n->size; bit++) {
+			u64 start;
+
+			if (!test_bit(bit, n->set))
+				continue;
+
+			start = bit * bsize;
+
+			/*
+			 * Merge nearby areas, we walk in order
+			 * through the bitmap, so no need to sort.
+			 */
+			if (j > 0) {
+				struct phys_entry *prev = &entries[j - 1];
+
+				if ((prev->end == start) &&
+				    (prev->node == n->node)) {
+					prev->end += bsize;
+					continue;
+				}
+			}
+
+			phys_entry__init(&entries[j++], start, bsize, n->node);
+		}
+	}
+
+	/* Cut unused entries, due to merging. */
+	tmp_entries = realloc(entries, sizeof(*entries) * j);
+	if (tmp_entries ||
+	    WARN_ONCE(j == 0, "No memory nodes, is CONFIG_MEMORY_HOTPLUG enabled?\n"))
+		entries = tmp_entries;
+
+	for (i = 0; i < j; i++) {
+		pr_debug("mem2node %03" PRIu64 " [0x%016" PRIx64 "-0x%016" PRIx64 "]\n",
+			 entries[i].node, entries[i].start, entries[i].end);
+
+		phys_entry__insert(&entries[i], &map->root);
+	}
+
+	map->entries = entries;
+	return 0;
+}
+
+void mem2node__exit(struct mem2node *map)
+{
+	zfree(&map->entries);
+}
+
+int mem2node__node(struct mem2node *map, u64 addr)
+{
+	struct rb_node **p, *parent = NULL;
+	struct phys_entry *entry;
+
+	p = &map->root.rb_node;
+	while (*p != NULL) {
+		parent = *p;
+		entry = rb_entry(parent, struct phys_entry, rb_node);
+		if (addr < entry->start)
+			p = &(*p)->rb_left;
+		else if (addr >= entry->end)
+			p = &(*p)->rb_right;
+		else
+			goto out;
+	}
+
+	entry = NULL;
+out:
+	return entry ? (int) entry->node : -1;
+}