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, 231 insertions, 0 deletions

diff --git a/drivers/net/netdevsim/health.c b/drivers/net/netdevsim/health.c
new file mode 100644
index 000000000..3bd0e7a48
--- /dev/null
+++ b/drivers/net/netdevsim/health.c
@@ -0,0 +1,231 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019 Mellanox Technologies. All rights reserved */
+
+#include <linux/debugfs.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+
+#include "netdevsim.h"
+
+static int
+nsim_dev_empty_reporter_dump(struct devlink_health_reporter *reporter,
+			     struct devlink_fmsg *fmsg, void *priv_ctx,
+			     struct netlink_ext_ack *extack)
+{
+	return 0;
+}
+
+static int
+nsim_dev_empty_reporter_diagnose(struct devlink_health_reporter *reporter,
+				 struct devlink_fmsg *fmsg,
+				 struct netlink_ext_ack *extack)
+{
+	return 0;
+}
+
+static const
+struct devlink_health_reporter_ops nsim_dev_empty_reporter_ops = {
+	.name = "empty",
+	.dump = nsim_dev_empty_reporter_dump,
+	.diagnose = nsim_dev_empty_reporter_diagnose,
+};
+
+struct nsim_dev_dummy_reporter_ctx {
+	char *break_msg;
+};
+
+static int
+nsim_dev_dummy_reporter_recover(struct devlink_health_reporter *reporter,
+				void *priv_ctx,
+				struct netlink_ext_ack *extack)
+{
+	struct nsim_dev_health *health = devlink_health_reporter_priv(reporter);
+	struct nsim_dev_dummy_reporter_ctx *ctx = priv_ctx;
+
+	if (health->fail_recover) {
+		/* For testing purposes, user set debugfs fail_recover
+		 * value to true. Fail right away.
+		 */
+		NL_SET_ERR_MSG_MOD(extack, "User setup the recover to fail for testing purposes");
+		return -EINVAL;
+	}
+	if (ctx) {
+		kfree(health->recovered_break_msg);
+		health->recovered_break_msg = kstrdup(ctx->break_msg,
+						      GFP_KERNEL);
+		if (!health->recovered_break_msg)
+			return -ENOMEM;
+	}
+	return 0;
+}
+
+static int nsim_dev_dummy_fmsg_put(struct devlink_fmsg *fmsg, u32 binary_len)
+{
+	char *binary;
+	int i;
+
+	devlink_fmsg_bool_pair_put(fmsg, "test_bool", true);
+	devlink_fmsg_u8_pair_put(fmsg, "test_u8", 1);
+	devlink_fmsg_u32_pair_put(fmsg, "test_u32", 3);
+	devlink_fmsg_u64_pair_put(fmsg, "test_u64", 4);
+	devlink_fmsg_string_pair_put(fmsg, "test_string", "somestring");
+
+	binary = kmalloc(binary_len, GFP_KERNEL | __GFP_NOWARN);
+	if (!binary)
+		return -ENOMEM;
+	get_random_bytes(binary, binary_len);
+	devlink_fmsg_binary_pair_put(fmsg, "test_binary", binary, binary_len);
+	kfree(binary);
+
+	devlink_fmsg_pair_nest_start(fmsg, "test_nest");
+	devlink_fmsg_obj_nest_start(fmsg);
+	devlink_fmsg_bool_pair_put(fmsg, "nested_test_bool", false);
+	devlink_fmsg_u8_pair_put(fmsg, "nested_test_u8", false);
+	devlink_fmsg_obj_nest_end(fmsg);
+	devlink_fmsg_pair_nest_end(fmsg);
+	devlink_fmsg_arr_pair_nest_end(fmsg);
+	devlink_fmsg_arr_pair_nest_start(fmsg, "test_u32_array");
+
+	for (i = 0; i < 10; i++)
+		devlink_fmsg_u32_put(fmsg, i);
+	devlink_fmsg_arr_pair_nest_end(fmsg);
+	devlink_fmsg_arr_pair_nest_start(fmsg, "test_array_of_objects");
+
+	for (i = 0; i < 10; i++) {
+		devlink_fmsg_obj_nest_start(fmsg);
+		devlink_fmsg_bool_pair_put(fmsg, "in_array_nested_test_bool",
+					   false);
+		devlink_fmsg_u8_pair_put(fmsg, "in_array_nested_test_u8", i);
+		devlink_fmsg_obj_nest_end(fmsg);
+	}
+	devlink_fmsg_arr_pair_nest_end(fmsg);
+
+	return 0;
+}
+
+static int
+nsim_dev_dummy_reporter_dump(struct devlink_health_reporter *reporter,
+			     struct devlink_fmsg *fmsg, void *priv_ctx,
+			     struct netlink_ext_ack *extack)
+{
+	struct nsim_dev_health *health = devlink_health_reporter_priv(reporter);
+	struct nsim_dev_dummy_reporter_ctx *ctx = priv_ctx;
+
+	if (ctx)
+		devlink_fmsg_string_pair_put(fmsg, "break_message", ctx->break_msg);
+
+	return nsim_dev_dummy_fmsg_put(fmsg, health->binary_len);
+}
+
+static int
+nsim_dev_dummy_reporter_diagnose(struct devlink_health_reporter *reporter,
+				 struct devlink_fmsg *fmsg,
+				 struct netlink_ext_ack *extack)
+{
+	struct nsim_dev_health *health = devlink_health_reporter_priv(reporter);
+
+	if (health->recovered_break_msg)
+		devlink_fmsg_string_pair_put(fmsg, "recovered_break_message",
+					     health->recovered_break_msg);
+
+	return nsim_dev_dummy_fmsg_put(fmsg, health->binary_len);
+}
+
+static const
+struct devlink_health_reporter_ops nsim_dev_dummy_reporter_ops = {
+	.name = "dummy",
+	.recover = nsim_dev_dummy_reporter_recover,
+	.dump = nsim_dev_dummy_reporter_dump,
+	.diagnose = nsim_dev_dummy_reporter_diagnose,
+};
+
+static ssize_t nsim_dev_health_break_write(struct file *file,
+					   const char __user *data,
+					   size_t count, loff_t *ppos)
+{
+	struct nsim_dev_health *health = file->private_data;
+	struct nsim_dev_dummy_reporter_ctx ctx;
+	char *break_msg;
+	int err;
+
+	if (count == 0 || count > PAGE_SIZE)
+		return -EINVAL;
+	break_msg = memdup_user_nul(data, count);
+	if (IS_ERR(break_msg))
+		return PTR_ERR(break_msg);
+
+	if (break_msg[count - 1] == '\n')
+		break_msg[count - 1] = '\0';
+
+	ctx.break_msg = break_msg;
+	err = devlink_health_report(health->dummy_reporter, break_msg, &ctx);
+	if (err)
+		goto out;
+
+out:
+	kfree(break_msg);
+	return err ?: count;
+}
+
+static const struct file_operations nsim_dev_health_break_fops = {
+	.open = simple_open,
+	.write = nsim_dev_health_break_write,
+	.llseek = generic_file_llseek,
+	.owner = THIS_MODULE,
+};
+
+int nsim_dev_health_init(struct nsim_dev *nsim_dev, struct devlink *devlink)
+{
+	struct nsim_dev_health *health = &nsim_dev->health;
+	int err;
+
+	health->empty_reporter =
+		devl_health_reporter_create(devlink,
+					    &nsim_dev_empty_reporter_ops,
+					    health);
+	if (IS_ERR(health->empty_reporter))
+		return PTR_ERR(health->empty_reporter);
+
+	health->dummy_reporter =
+		devl_health_reporter_create(devlink,
+					    &nsim_dev_dummy_reporter_ops,
+					    health);
+	if (IS_ERR(health->dummy_reporter)) {
+		err = PTR_ERR(health->dummy_reporter);
+		goto err_empty_reporter_destroy;
+	}
+
+	health->ddir = debugfs_create_dir("health", nsim_dev->ddir);
+	if (IS_ERR(health->ddir)) {
+		err = PTR_ERR(health->ddir);
+		goto err_dummy_reporter_destroy;
+	}
+
+	health->recovered_break_msg = NULL;
+	debugfs_create_file("break_health", 0200, health->ddir, health,
+			    &nsim_dev_health_break_fops);
+	health->binary_len = 16;
+	debugfs_create_u32("binary_len", 0600, health->ddir,
+			   &health->binary_len);
+	health->fail_recover = false;
+	debugfs_create_bool("fail_recover", 0600, health->ddir,
+			    &health->fail_recover);
+	return 0;
+
+err_dummy_reporter_destroy:
+	devl_health_reporter_destroy(health->dummy_reporter);
+err_empty_reporter_destroy:
+	devl_health_reporter_destroy(health->empty_reporter);
+	return err;
+}
+
+void nsim_dev_health_exit(struct nsim_dev *nsim_dev)
+{
+	struct nsim_dev_health *health = &nsim_dev->health;
+
+	debugfs_remove_recursive(health->ddir);
+	kfree(health->recovered_break_msg);
+	devl_health_reporter_destroy(health->dummy_reporter);
+	devl_health_reporter_destroy(health->empty_reporter);
+}