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, 104 insertions, 0 deletions

diff --git a/scripts/dtc/srcpos.h b/scripts/dtc/srcpos.h
new file mode 100644
index 000000000..4d60b50e3
--- /dev/null
+++ b/scripts/dtc/srcpos.h
@@ -0,0 +1,104 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright 2007 Jon Loeliger, Freescale Semiconductor, Inc.
+ */
+
+#ifndef SRCPOS_H
+#define SRCPOS_H
+
+#include <stdio.h>
+#include <stdbool.h>
+#include "util.h"
+
+struct srcfile_state {
+	FILE *f;
+	char *name;
+	char *dir;
+	int lineno, colno;
+	struct srcfile_state *prev;
+};
+
+extern FILE *depfile; /* = NULL */
+extern struct srcfile_state *current_srcfile; /* = NULL */
+
+/**
+ * Open a source file.
+ *
+ * If the source file is a relative pathname, then it is searched for in the
+ * current directory (the directory of the last source file read) and after
+ * that in the search path.
+ *
+ * We work through the search path in order from the first path specified to
+ * the last.
+ *
+ * If the file is not found, then this function does not return, but calls
+ * die().
+ *
+ * @param fname		Filename to search
+ * @param fullnamep	If non-NULL, it is set to the allocated filename of the
+ *			file that was opened. The caller is then responsible
+ *			for freeing the pointer.
+ * @return pointer to opened FILE
+ */
+FILE *srcfile_relative_open(const char *fname, char **fullnamep);
+
+void srcfile_push(const char *fname);
+bool srcfile_pop(void);
+
+/**
+ * Add a new directory to the search path for input files
+ *
+ * The new path is added at the end of the list.
+ *
+ * @param dirname	Directory to add
+ */
+void srcfile_add_search_path(const char *dirname);
+
+struct srcpos {
+    int first_line;
+    int first_column;
+    int last_line;
+    int last_column;
+    struct srcfile_state *file;
+    struct srcpos *next;
+};
+
+#define YYLTYPE struct srcpos
+
+#define YYLLOC_DEFAULT(Current, Rhs, N)						\
+	do {									\
+		if (N) {							\
+			(Current).first_line = YYRHSLOC(Rhs, 1).first_line;	\
+			(Current).first_column = YYRHSLOC(Rhs, 1).first_column;	\
+			(Current).last_line = YYRHSLOC(Rhs, N).last_line;	\
+			(Current).last_column  = YYRHSLOC (Rhs, N).last_column;	\
+			(Current).file = YYRHSLOC(Rhs, N).file;			\
+		} else {							\
+			(Current).first_line = (Current).last_line =		\
+				YYRHSLOC(Rhs, 0).last_line;			\
+			(Current).first_column = (Current).last_column =	\
+				YYRHSLOC(Rhs, 0).last_column;			\
+			(Current).file = YYRHSLOC (Rhs, 0).file;		\
+		}								\
+		(Current).next = NULL;						\
+	} while (0)
+
+
+extern void srcpos_update(struct srcpos *pos, const char *text, int len);
+extern struct srcpos *srcpos_copy(struct srcpos *pos);
+extern struct srcpos *srcpos_extend(struct srcpos *new_srcpos,
+				    struct srcpos *old_srcpos);
+extern void srcpos_free(struct srcpos *pos);
+extern char *srcpos_string(struct srcpos *pos);
+extern char *srcpos_string_first(struct srcpos *pos, int level);
+extern char *srcpos_string_last(struct srcpos *pos, int level);
+
+
+extern void PRINTF(3, 0) srcpos_verror(struct srcpos *pos, const char *prefix,
+					const char *fmt, va_list va);
+extern void PRINTF(3, 4) srcpos_error(struct srcpos *pos, const char *prefix,
+				      const char *fmt, ...);
+
+extern void srcpos_set_line(char *f, int l);
+
+#endif /* SRCPOS_H */