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, 359 insertions, 0 deletions

diff --git a/drivers/clk/clk-devres.c b/drivers/clk/clk-devres.c
new file mode 100644
index 000000000..4cc3863bd
--- /dev/null
+++ b/drivers/clk/clk-devres.c
@@ -0,0 +1,359 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/export.h>
+#include <linux/gfp.h>
+
+struct devm_clk_state {
+	struct clk *clk;
+	void (*exit)(struct clk *clk);
+};
+
+static void devm_clk_release(struct device *dev, void *res)
+{
+	struct devm_clk_state *state = res;
+
+	if (state->exit)
+		state->exit(state->clk);
+
+	clk_put(state->clk);
+}
+
+static struct clk *__devm_clk_get(struct device *dev, const char *id,
+				  struct clk *(*get)(struct device *dev, const char *id),
+				  int (*init)(struct clk *clk),
+				  void (*exit)(struct clk *clk))
+{
+	struct devm_clk_state *state;
+	struct clk *clk;
+	int ret;
+
+	state = devres_alloc(devm_clk_release, sizeof(*state), GFP_KERNEL);
+	if (!state)
+		return ERR_PTR(-ENOMEM);
+
+	clk = get(dev, id);
+	if (IS_ERR(clk)) {
+		ret = PTR_ERR(clk);
+		goto err_clk_get;
+	}
+
+	if (init) {
+		ret = init(clk);
+		if (ret)
+			goto err_clk_init;
+	}
+
+	state->clk = clk;
+	state->exit = exit;
+
+	devres_add(dev, state);
+
+	return clk;
+
+err_clk_init:
+
+	clk_put(clk);
+err_clk_get:
+
+	devres_free(state);
+	return ERR_PTR(ret);
+}
+
+struct clk *devm_clk_get(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get, NULL, NULL);
+}
+EXPORT_SYMBOL(devm_clk_get);
+
+struct clk *devm_clk_get_prepared(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get, clk_prepare, clk_unprepare);
+}
+EXPORT_SYMBOL_GPL(devm_clk_get_prepared);
+
+struct clk *devm_clk_get_enabled(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get,
+			      clk_prepare_enable, clk_disable_unprepare);
+}
+EXPORT_SYMBOL_GPL(devm_clk_get_enabled);
+
+struct clk *devm_clk_get_optional(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get_optional, NULL, NULL);
+}
+EXPORT_SYMBOL(devm_clk_get_optional);
+
+struct clk *devm_clk_get_optional_prepared(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get_optional,
+			      clk_prepare, clk_unprepare);
+}
+EXPORT_SYMBOL_GPL(devm_clk_get_optional_prepared);
+
+struct clk *devm_clk_get_optional_enabled(struct device *dev, const char *id)
+{
+	return __devm_clk_get(dev, id, clk_get_optional,
+			      clk_prepare_enable, clk_disable_unprepare);
+}
+EXPORT_SYMBOL_GPL(devm_clk_get_optional_enabled);
+
+static void devm_clk_disable_unprepare(void *data)
+{
+	clk_disable_unprepare(data);
+}
+
+struct clk *devm_clk_get_optional_enabled_with_rate(struct device *dev,
+						    const char *id,
+						    unsigned long rate)
+{
+	struct clk *clk;
+	int ret;
+
+	clk = devm_clk_get_optional(dev, id);
+	if (IS_ERR(clk))
+		return ERR_CAST(clk);
+
+	ret = clk_set_rate(clk, rate);
+	if (ret)
+		goto out_put_clk;
+
+	ret = clk_prepare_enable(clk);
+	if (ret)
+		goto out_put_clk;
+
+	ret = devm_add_action_or_reset(dev, devm_clk_disable_unprepare, clk);
+	if (ret)
+		goto out_put_clk;
+
+	return clk;
+
+out_put_clk:
+	devm_clk_put(dev, clk);
+	return ERR_PTR(ret);
+}
+EXPORT_SYMBOL_GPL(devm_clk_get_optional_enabled_with_rate);
+
+struct clk_bulk_devres {
+	struct clk_bulk_data *clks;
+	int num_clks;
+};
+
+static void devm_clk_bulk_release(struct device *dev, void *res)
+{
+	struct clk_bulk_devres *devres = res;
+
+	clk_bulk_put(devres->num_clks, devres->clks);
+}
+
+static int __devm_clk_bulk_get(struct device *dev, int num_clks,
+			       struct clk_bulk_data *clks, bool optional)
+{
+	struct clk_bulk_devres *devres;
+	int ret;
+
+	devres = devres_alloc(devm_clk_bulk_release,
+			      sizeof(*devres), GFP_KERNEL);
+	if (!devres)
+		return -ENOMEM;
+
+	if (optional)
+		ret = clk_bulk_get_optional(dev, num_clks, clks);
+	else
+		ret = clk_bulk_get(dev, num_clks, clks);
+	if (!ret) {
+		devres->clks = clks;
+		devres->num_clks = num_clks;
+		devres_add(dev, devres);
+	} else {
+		devres_free(devres);
+	}
+
+	return ret;
+}
+
+int __must_check devm_clk_bulk_get(struct device *dev, int num_clks,
+		      struct clk_bulk_data *clks)
+{
+	return __devm_clk_bulk_get(dev, num_clks, clks, false);
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get);
+
+int __must_check devm_clk_bulk_get_optional(struct device *dev, int num_clks,
+		      struct clk_bulk_data *clks)
+{
+	return __devm_clk_bulk_get(dev, num_clks, clks, true);
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get_optional);
+
+static void devm_clk_bulk_release_enable(struct device *dev, void *res)
+{
+	struct clk_bulk_devres *devres = res;
+
+	clk_bulk_disable_unprepare(devres->num_clks, devres->clks);
+	clk_bulk_put(devres->num_clks, devres->clks);
+}
+
+static int __devm_clk_bulk_get_enable(struct device *dev, int num_clks,
+				      struct clk_bulk_data *clks, bool optional)
+{
+	struct clk_bulk_devres *devres;
+	int ret;
+
+	devres = devres_alloc(devm_clk_bulk_release_enable,
+			      sizeof(*devres), GFP_KERNEL);
+	if (!devres)
+		return -ENOMEM;
+
+	if (optional)
+		ret = clk_bulk_get_optional(dev, num_clks, clks);
+	else
+		ret = clk_bulk_get(dev, num_clks, clks);
+	if (ret)
+		goto err_clk_get;
+
+	ret = clk_bulk_prepare_enable(num_clks, clks);
+	if (ret)
+		goto err_clk_prepare;
+
+	devres->clks = clks;
+	devres->num_clks = num_clks;
+	devres_add(dev, devres);
+
+	return 0;
+
+err_clk_prepare:
+	clk_bulk_put(num_clks, clks);
+err_clk_get:
+	devres_free(devres);
+	return ret;
+}
+
+int __must_check devm_clk_bulk_get_enable(struct device *dev, int num_clks,
+					  struct clk_bulk_data *clks)
+{
+	return __devm_clk_bulk_get_enable(dev, num_clks, clks, false);
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get_enable);
+
+int __must_check devm_clk_bulk_get_optional_enable(struct device *dev, int num_clks,
+						   struct clk_bulk_data *clks)
+{
+	return __devm_clk_bulk_get_enable(dev, num_clks, clks, true);
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get_optional_enable);
+
+static void devm_clk_bulk_release_all(struct device *dev, void *res)
+{
+	struct clk_bulk_devres *devres = res;
+
+	clk_bulk_put_all(devres->num_clks, devres->clks);
+}
+
+int __must_check devm_clk_bulk_get_all(struct device *dev,
+				       struct clk_bulk_data **clks)
+{
+	struct clk_bulk_devres *devres;
+	int ret;
+
+	devres = devres_alloc(devm_clk_bulk_release_all,
+			      sizeof(*devres), GFP_KERNEL);
+	if (!devres)
+		return -ENOMEM;
+
+	ret = clk_bulk_get_all(dev, &devres->clks);
+	if (ret > 0) {
+		*clks = devres->clks;
+		devres->num_clks = ret;
+		devres_add(dev, devres);
+	} else {
+		devres_free(devres);
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get_all);
+
+static void devm_clk_bulk_release_all_enable(struct device *dev, void *res)
+{
+	struct clk_bulk_devres *devres = res;
+
+	clk_bulk_disable_unprepare(devres->num_clks, devres->clks);
+	clk_bulk_put_all(devres->num_clks, devres->clks);
+}
+
+int __must_check devm_clk_bulk_get_all_enabled(struct device *dev,
+					       struct clk_bulk_data **clks)
+{
+	struct clk_bulk_devres *devres;
+	int ret;
+
+	devres = devres_alloc(devm_clk_bulk_release_all_enable,
+			      sizeof(*devres), GFP_KERNEL);
+	if (!devres)
+		return -ENOMEM;
+
+	ret = clk_bulk_get_all(dev, &devres->clks);
+	if (ret > 0) {
+		*clks = devres->clks;
+		devres->num_clks = ret;
+	} else {
+		devres_free(devres);
+		return ret;
+	}
+
+	ret = clk_bulk_prepare_enable(devres->num_clks, *clks);
+	if (!ret) {
+		devres_add(dev, devres);
+	} else {
+		clk_bulk_put_all(devres->num_clks, devres->clks);
+		devres_free(devres);
+		return ret;
+	}
+
+	return devres->num_clks;
+}
+EXPORT_SYMBOL_GPL(devm_clk_bulk_get_all_enabled);
+
+static int devm_clk_match(struct device *dev, void *res, void *data)
+{
+	struct clk **c = res;
+	if (!c || !*c) {
+		WARN_ON(!c || !*c);
+		return 0;
+	}
+	return *c == data;
+}
+
+void devm_clk_put(struct device *dev, struct clk *clk)
+{
+	int ret;
+
+	ret = devres_release(dev, devm_clk_release, devm_clk_match, clk);
+
+	WARN_ON(ret);
+}
+EXPORT_SYMBOL(devm_clk_put);
+
+struct clk *devm_get_clk_from_child(struct device *dev,
+				    struct device_node *np, const char *con_id)
+{
+	struct devm_clk_state *state;
+	struct clk *clk;
+
+	state = devres_alloc(devm_clk_release, sizeof(*state), GFP_KERNEL);
+	if (!state)
+		return ERR_PTR(-ENOMEM);
+
+	clk = of_clk_get_by_name(np, con_id);
+	if (!IS_ERR(clk)) {
+		state->clk = clk;
+		devres_add(dev, state);
+	} else {
+		devres_free(state);
+	}
+
+	return clk;
+}
+EXPORT_SYMBOL(devm_get_clk_from_child);