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, 153 insertions, 0 deletions

diff --git a/include/drm/drm_managed.h b/include/drm/drm_managed.h
new file mode 100644
index 000000000..72d0d68be
--- /dev/null
+++ b/include/drm/drm_managed.h
@@ -0,0 +1,153 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#ifndef _DRM_MANAGED_H_
+#define _DRM_MANAGED_H_
+
+#include <linux/gfp.h>
+#include <linux/overflow.h>
+#include <linux/types.h>
+
+struct drm_device;
+struct mutex;
+
+typedef void (*drmres_release_t)(struct drm_device *dev, void *res);
+
+/**
+ * drmm_add_action - add a managed release action to a &drm_device
+ * @dev: DRM device
+ * @action: function which should be called when @dev is released
+ * @data: opaque pointer, passed to @action
+ *
+ * This function adds the release @action with optional parameter @data to the
+ * list of cleanup actions for @dev. The cleanup actions will be run in reverse
+ * order in the final drm_dev_put() call for @dev.
+ */
+#define drmm_add_action(dev, action, data) \
+	__drmm_add_action(dev, action, data, #action)
+
+int __must_check __drmm_add_action(struct drm_device *dev,
+				   drmres_release_t action,
+				   void *data, const char *name);
+
+/**
+ * drmm_add_action_or_reset - add a managed release action to a &drm_device
+ * @dev: DRM device
+ * @action: function which should be called when @dev is released
+ * @data: opaque pointer, passed to @action
+ *
+ * Similar to drmm_add_action(), with the only difference that upon failure
+ * @action is directly called for any cleanup work necessary on failures.
+ */
+#define drmm_add_action_or_reset(dev, action, data) \
+	__drmm_add_action_or_reset(dev, action, data, #action)
+
+int __must_check __drmm_add_action_or_reset(struct drm_device *dev,
+					    drmres_release_t action,
+					    void *data, const char *name);
+
+void drmm_release_action(struct drm_device *dev,
+			 drmres_release_t action,
+			 void *data);
+
+void *drmm_kmalloc(struct drm_device *dev, size_t size, gfp_t gfp) __malloc;
+
+/**
+ * drmm_kzalloc - &drm_device managed kzalloc()
+ * @dev: DRM device
+ * @size: size of the memory allocation
+ * @gfp: GFP allocation flags
+ *
+ * This is a &drm_device managed version of kzalloc(). The allocated memory is
+ * automatically freed on the final drm_dev_put(). Memory can also be freed
+ * before the final drm_dev_put() by calling drmm_kfree().
+ */
+static inline void *drmm_kzalloc(struct drm_device *dev, size_t size, gfp_t gfp)
+{
+	return drmm_kmalloc(dev, size, gfp | __GFP_ZERO);
+}
+
+/**
+ * drmm_kmalloc_array - &drm_device managed kmalloc_array()
+ * @dev: DRM device
+ * @n: number of array elements to allocate
+ * @size: size of array member
+ * @flags: GFP allocation flags
+ *
+ * This is a &drm_device managed version of kmalloc_array(). The allocated
+ * memory is automatically freed on the final drm_dev_put() and works exactly
+ * like a memory allocation obtained by drmm_kmalloc().
+ */
+static inline void *drmm_kmalloc_array(struct drm_device *dev,
+				       size_t n, size_t size, gfp_t flags)
+{
+	size_t bytes;
+
+	if (unlikely(check_mul_overflow(n, size, &bytes)))
+		return NULL;
+
+	return drmm_kmalloc(dev, bytes, flags);
+}
+
+/**
+ * drmm_kcalloc - &drm_device managed kcalloc()
+ * @dev: DRM device
+ * @n: number of array elements to allocate
+ * @size: size of array member
+ * @flags: GFP allocation flags
+ *
+ * This is a &drm_device managed version of kcalloc(). The allocated memory is
+ * automatically freed on the final drm_dev_put() and works exactly like a
+ * memory allocation obtained by drmm_kmalloc().
+ */
+static inline void *drmm_kcalloc(struct drm_device *dev,
+				 size_t n, size_t size, gfp_t flags)
+{
+	return drmm_kmalloc_array(dev, n, size, flags | __GFP_ZERO);
+}
+
+char *drmm_kstrdup(struct drm_device *dev, const char *s, gfp_t gfp);
+
+void drmm_kfree(struct drm_device *dev, void *data);
+
+void __drmm_mutex_release(struct drm_device *dev, void *res);
+
+/**
+ * drmm_mutex_init - &drm_device-managed mutex_init()
+ * @dev: DRM device
+ * @lock: lock to be initialized
+ *
+ * Returns:
+ * 0 on success, or a negative errno code otherwise.
+ *
+ * This is a &drm_device-managed version of mutex_init(). The initialized
+ * lock is automatically destroyed on the final drm_dev_put().
+ */
+#define drmm_mutex_init(dev, lock) ({					     \
+	mutex_init(lock);						     \
+	drmm_add_action_or_reset(dev, __drmm_mutex_release, lock);	     \
+})									     \
+
+void __drmm_workqueue_release(struct drm_device *device, void *wq);
+
+/**
+ * drmm_alloc_ordered_workqueue - &drm_device managed alloc_ordered_workqueue()
+ * @dev: DRM device
+ * @fmt: printf format for the name of the workqueue
+ * @flags: WQ_* flags (only WQ_FREEZABLE and WQ_MEM_RECLAIM are meaningful)
+ * @args: args for @fmt
+ *
+ * This is a &drm_device-managed version of alloc_ordered_workqueue(). The
+ * allocated workqueue is automatically destroyed on the final drm_dev_put().
+ *
+ * Returns: workqueue on success, negative ERR_PTR otherwise.
+ */
+#define drmm_alloc_ordered_workqueue(dev, fmt, flags, args...)					\
+	({											\
+		struct workqueue_struct *wq = alloc_ordered_workqueue(fmt, flags, ##args);	\
+		wq ? ({										\
+			int ret = drmm_add_action_or_reset(dev, __drmm_workqueue_release, wq);	\
+			ret ? ERR_PTR(ret) : wq;						\
+		}) : ERR_PTR(-ENOMEM);								\
+	})
+
+#endif