sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted

alb_break_llc() decides whether to break LLC preference during active
load balance. It does so by testing that every runnable fair task on the
source rq prefers its LLC:

	env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable

But the two counters cover different sets. nr_pref_llc_running is updated
in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued,
so it follows queued tasks. h_nr_runnable is updated in set_delayed()/
clear_delayed() and drops delay-dequeued tasks.

So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted
in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks,
alb_break_llc() returns false, and active balance is free to pull a task
off its preferred LLC. Active balance only moves runnable tasks, and this
is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB
skips the per-task test in can_migrate_task(). The runnable set is the one
we want.

Fix it on the counter side. A task should be counted in
nr_pref_llc_running exactly while it is both queued on its preferred LLC
(pref_llc_queued) and runnable (!sched_delayed). Define that membership
once in task_pref_llc_runnable(), and adjust the counter only through
pref_llc_running_inc()/pref_llc_running_dec() from the four sites that
change either input: account_llc_enqueue(), account_llc_dequeue(),
set_delayed() and clear_delayed(). Gating every update on the same
predicate keeps the delay, wake and dequeue paths from double-counting
or underflowing; see the comments at those sites for the ordering.

nr_llc_running and sd->llc_counts are not touched and stay on queued
semantics.

Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC")
Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/
Reported-by: Zhan Xusheng <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com

1 files changed, 195 insertions, 0 deletions

diff --git a/samples/bpf/fds_example.c b/samples/bpf/fds_example.c
new file mode 100644
index 000000000..88a26f3ce
--- /dev/null
+++ b/samples/bpf/fds_example.c
@@ -0,0 +1,195 @@
+#include <linux/unistd.h>
+#include <linux/bpf.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <string.h>
+#include <assert.h>
+#include <errno.h>
+
+#include <sys/types.h>
+#include <sys/socket.h>
+
+#include <bpf/bpf.h>
+
+#include <bpf/libbpf.h>
+#include "bpf_insn.h"
+#include "sock_example.h"
+#include "bpf_util.h"
+
+#define BPF_F_PIN	(1 << 0)
+#define BPF_F_GET	(1 << 1)
+#define BPF_F_PIN_GET	(BPF_F_PIN | BPF_F_GET)
+
+#define BPF_F_KEY	(1 << 2)
+#define BPF_F_VAL	(1 << 3)
+#define BPF_F_KEY_VAL	(BPF_F_KEY | BPF_F_VAL)
+
+#define BPF_M_UNSPEC	0
+#define BPF_M_MAP	1
+#define BPF_M_PROG	2
+
+char bpf_log_buf[BPF_LOG_BUF_SIZE];
+
+static void usage(void)
+{
+	printf("Usage: fds_example [...]\n");
+	printf("       -F <file>   File to pin/get object\n");
+	printf("       -P          |- pin object\n");
+	printf("       -G          `- get object\n");
+	printf("       -m          eBPF map mode\n");
+	printf("       -k <key>    |- map key\n");
+	printf("       -v <value>  `- map value\n");
+	printf("       -p          eBPF prog mode\n");
+	printf("       -o <object> `- object file\n");
+	printf("       -h          Display this help.\n");
+}
+
+static int bpf_prog_create(const char *object)
+{
+	static struct bpf_insn insns[] = {
+		BPF_MOV64_IMM(BPF_REG_0, 1),
+		BPF_EXIT_INSN(),
+	};
+	size_t insns_cnt = ARRAY_SIZE(insns);
+	struct bpf_object *obj;
+	int err;
+
+	if (object) {
+		obj = bpf_object__open_file(object, NULL);
+		assert(!libbpf_get_error(obj));
+		err = bpf_object__load(obj);
+		assert(!err);
+		return bpf_program__fd(bpf_object__next_program(obj, NULL));
+	} else {
+		LIBBPF_OPTS(bpf_prog_load_opts, opts,
+			.log_buf = bpf_log_buf,
+			.log_size = BPF_LOG_BUF_SIZE,
+		);
+
+		return bpf_prog_load(BPF_PROG_TYPE_SOCKET_FILTER, NULL, "GPL",
+				     insns, insns_cnt, &opts);
+	}
+}
+
+static int bpf_do_map(const char *file, uint32_t flags, uint32_t key,
+		      uint32_t value)
+{
+	int fd, ret;
+
+	if (flags & BPF_F_PIN) {
+		fd = bpf_map_create(BPF_MAP_TYPE_ARRAY, NULL, sizeof(uint32_t),
+				    sizeof(uint32_t), 1024, NULL);
+		printf("bpf: map fd:%d (%s)\n", fd, strerror(errno));
+		assert(fd > 0);
+
+		ret = bpf_obj_pin(fd, file);
+		printf("bpf: pin ret:(%d,%s)\n", ret, strerror(errno));
+		assert(ret == 0);
+	} else {
+		fd = bpf_obj_get(file);
+		printf("bpf: get fd:%d (%s)\n", fd, strerror(errno));
+		assert(fd > 0);
+	}
+
+	if ((flags & BPF_F_KEY_VAL) == BPF_F_KEY_VAL) {
+		ret = bpf_map_update_elem(fd, &key, &value, 0);
+		printf("bpf: fd:%d u->(%u:%u) ret:(%d,%s)\n", fd, key, value,
+		       ret, strerror(errno));
+		assert(ret == 0);
+	} else if (flags & BPF_F_KEY) {
+		ret = bpf_map_lookup_elem(fd, &key, &value);
+		printf("bpf: fd:%d l->(%u):%u ret:(%d,%s)\n", fd, key, value,
+		       ret, strerror(errno));
+		assert(ret == 0);
+	}
+
+	return 0;
+}
+
+static int bpf_do_prog(const char *file, uint32_t flags, const char *object)
+{
+	int fd, sock, ret;
+
+	if (flags & BPF_F_PIN) {
+		fd = bpf_prog_create(object);
+		printf("bpf: prog fd:%d (%s)\n", fd, strerror(errno));
+		assert(fd > 0);
+
+		ret = bpf_obj_pin(fd, file);
+		printf("bpf: pin ret:(%d,%s)\n", ret, strerror(errno));
+		assert(ret == 0);
+	} else {
+		fd = bpf_obj_get(file);
+		printf("bpf: get fd:%d (%s)\n", fd, strerror(errno));
+		assert(fd > 0);
+	}
+
+	sock = open_raw_sock("lo");
+	assert(sock > 0);
+
+	ret = setsockopt(sock, SOL_SOCKET, SO_ATTACH_BPF, &fd, sizeof(fd));
+	printf("bpf: sock:%d <- fd:%d attached ret:(%d,%s)\n", sock, fd,
+	       ret, strerror(errno));
+	assert(ret == 0);
+
+	return 0;
+}
+
+int main(int argc, char **argv)
+{
+	const char *file = NULL, *object = NULL;
+	uint32_t key = 0, value = 0, flags = 0;
+	int opt, mode = BPF_M_UNSPEC;
+
+	while ((opt = getopt(argc, argv, "F:PGmk:v:po:")) != -1) {
+		switch (opt) {
+		/* General args */
+		case 'F':
+			file = optarg;
+			break;
+		case 'P':
+			flags |= BPF_F_PIN;
+			break;
+		case 'G':
+			flags |= BPF_F_GET;
+			break;
+		/* Map-related args */
+		case 'm':
+			mode = BPF_M_MAP;
+			break;
+		case 'k':
+			key = strtoul(optarg, NULL, 0);
+			flags |= BPF_F_KEY;
+			break;
+		case 'v':
+			value = strtoul(optarg, NULL, 0);
+			flags |= BPF_F_VAL;
+			break;
+		/* Prog-related args */
+		case 'p':
+			mode = BPF_M_PROG;
+			break;
+		case 'o':
+			object = optarg;
+			break;
+		default:
+			goto out;
+		}
+	}
+
+	if (!(flags & BPF_F_PIN_GET) || !file)
+		goto out;
+
+	switch (mode) {
+	case BPF_M_MAP:
+		return bpf_do_map(file, flags, key, value);
+	case BPF_M_PROG:
+		return bpf_do_prog(file, flags, object);
+	}
+out:
+	usage();
+	return -1;
+}