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, 137 insertions, 0 deletions

diff --git a/include/trace/events/percpu.h b/include/trace/events/percpu.h
new file mode 100644
index 000000000..5b8211ca8
--- /dev/null
+++ b/include/trace/events/percpu.h
@@ -0,0 +1,137 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM percpu
+
+#if !defined(_TRACE_PERCPU_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_PERCPU_H
+
+#include <linux/tracepoint.h>
+#include <trace/events/mmflags.h>
+
+TRACE_EVENT(percpu_alloc_percpu,
+
+	TP_PROTO(unsigned long call_site,
+		 bool reserved, bool is_atomic, size_t size,
+		 size_t align, void *base_addr, int off,
+		 void __percpu *ptr, size_t bytes_alloc, gfp_t gfp_flags),
+
+	TP_ARGS(call_site, reserved, is_atomic, size, align, base_addr, off,
+		ptr, bytes_alloc, gfp_flags),
+
+	TP_STRUCT__entry(
+		__field(	unsigned long,		call_site	)
+		__field(	bool,			reserved	)
+		__field(	bool,			is_atomic	)
+		__field(	size_t,			size		)
+		__field(	size_t,			align		)
+		__field(	void *,			base_addr	)
+		__field(	int,			off		)
+		__field(	void __percpu *,	ptr		)
+		__field(	size_t,			bytes_alloc	)
+		__field(	unsigned long,		gfp_flags	)
+	),
+	TP_fast_assign(
+		__entry->call_site	= call_site;
+		__entry->reserved	= reserved;
+		__entry->is_atomic	= is_atomic;
+		__entry->size		= size;
+		__entry->align		= align;
+		__entry->base_addr	= base_addr;
+		__entry->off		= off;
+		__entry->ptr		= ptr;
+		__entry->bytes_alloc	= bytes_alloc;
+		__entry->gfp_flags	= (__force unsigned long)gfp_flags;
+	),
+
+	TP_printk("call_site=%pS reserved=%d is_atomic=%d size=%zu align=%zu base_addr=%p off=%d ptr=%p bytes_alloc=%zu gfp_flags=%s",
+		  (void *)__entry->call_site,
+		  __entry->reserved, __entry->is_atomic,
+		  __entry->size, __entry->align,
+		  __entry->base_addr, __entry->off, __entry->ptr,
+		  __entry->bytes_alloc, show_gfp_flags(__entry->gfp_flags))
+);
+
+TRACE_EVENT(percpu_free_percpu,
+
+	TP_PROTO(void *base_addr, int off, void __percpu *ptr),
+
+	TP_ARGS(base_addr, off, ptr),
+
+	TP_STRUCT__entry(
+		__field(	void *,			base_addr	)
+		__field(	int,			off		)
+		__field(	void __percpu *,	ptr		)
+	),
+
+	TP_fast_assign(
+		__entry->base_addr	= base_addr;
+		__entry->off		= off;
+		__entry->ptr		= ptr;
+	),
+
+	TP_printk("base_addr=%p off=%d ptr=%p",
+		__entry->base_addr, __entry->off, __entry->ptr)
+);
+
+TRACE_EVENT(percpu_alloc_percpu_fail,
+
+	TP_PROTO(bool reserved, bool is_atomic, size_t size, size_t align),
+
+	TP_ARGS(reserved, is_atomic, size, align),
+
+	TP_STRUCT__entry(
+		__field(	bool,	reserved	)
+		__field(	bool,	is_atomic	)
+		__field(	size_t,	size		)
+		__field(	size_t, align		)
+	),
+
+	TP_fast_assign(
+		__entry->reserved	= reserved;
+		__entry->is_atomic	= is_atomic;
+		__entry->size		= size;
+		__entry->align		= align;
+	),
+
+	TP_printk("reserved=%d is_atomic=%d size=%zu align=%zu",
+		  __entry->reserved, __entry->is_atomic,
+		  __entry->size, __entry->align)
+);
+
+TRACE_EVENT(percpu_create_chunk,
+
+	TP_PROTO(void *base_addr),
+
+	TP_ARGS(base_addr),
+
+	TP_STRUCT__entry(
+		__field(	void *, base_addr	)
+	),
+
+	TP_fast_assign(
+		__entry->base_addr	= base_addr;
+	),
+
+	TP_printk("base_addr=%p", __entry->base_addr)
+);
+
+TRACE_EVENT(percpu_destroy_chunk,
+
+	TP_PROTO(void *base_addr),
+
+	TP_ARGS(base_addr),
+
+	TP_STRUCT__entry(
+		__field(	void *,	base_addr	)
+	),
+
+	TP_fast_assign(
+		__entry->base_addr	= base_addr;
+	),
+
+	TP_printk("base_addr=%p", __entry->base_addr)
+);
+
+#endif /* _TRACE_PERCPU_H */
+
+#include <trace/define_trace.h>