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, 126 insertions, 0 deletions

diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h
new file mode 100644
index 000000000..87a0818a8
--- /dev/null
+++ b/tools/perf/util/util.h
@@ -0,0 +1,126 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __PERF_UTIL_H
+#define __PERF_UTIL_H
+
+#define _BSD_SOURCE 1
+/* glibc 2.20 deprecates _BSD_SOURCE in favour of _DEFAULT_SOURCE */
+#define _DEFAULT_SOURCE 1
+
+#include <dirent.h>
+#include <fcntl.h>
+#include <stdbool.h>
+#include <stddef.h>
+#include <linux/compiler.h>
+#include <linux/bitmap.h>
+#include <sys/types.h>
+#ifndef __cplusplus
+#include <internal/cpumap.h>
+#endif
+
+extern const char perf_usage_string[];
+extern const char perf_more_info_string[];
+
+extern const char *input_name;
+
+/* This will control if perf_{host,guest} will set attr.exclude_{host,guest}. */
+extern bool exclude_GH_default;
+
+extern bool perf_host;
+extern bool perf_guest;
+
+/* General helper functions */
+void usage(const char *err) __noreturn;
+
+struct dirent;
+struct strlist;
+
+int mkdir_p(char *path, mode_t mode);
+int rm_rf(const char *path);
+int rm_rf_perf_data(const char *path);
+struct strlist *lsdir(const char *name, bool (*filter)(const char *, struct dirent *));
+bool lsdir_no_dot_filter(const char *name, struct dirent *d);
+
+size_t hex_width(u64 v);
+
+int sysctl__max_stack(void);
+
+bool sysctl__nmi_watchdog_enabled(void);
+
+int perf_tip(char **strp, const char *dirpath);
+
+void cpumask_to_cpulist(char *cpumask, char *cpulist);
+
+void print_separator2(int pre_dash_cnt, const char *s, int post_dash_cnt);
+
+#ifndef HAVE_SCHED_GETCPU_SUPPORT
+int sched_getcpu(void);
+#endif
+
+#ifndef HAVE_SCANDIRAT_SUPPORT
+int scandirat(int dirfd, const char *dirp,
+	      struct dirent ***namelist,
+	      int (*filter)(const struct dirent *),
+	      int (*compar)(const struct dirent **, const struct dirent **));
+#endif
+
+extern bool perf_singlethreaded;
+
+void perf_set_singlethreaded(void);
+void perf_set_multithreaded(void);
+
+char *perf_exe(char *buf, int len);
+
+#ifndef O_CLOEXEC
+#ifdef __sparc__
+#define O_CLOEXEC      0x400000
+#elif defined(__alpha__) || defined(__hppa__)
+#define O_CLOEXEC      010000000
+#else
+#define O_CLOEXEC      02000000
+#endif
+#endif
+
+struct perf_debuginfod {
+	const char	*urls;
+	bool		 set;
+};
+void perf_debuginfod_setup(struct perf_debuginfod *di);
+
+const char *perf_basename(const char *path);
+
+char *filename_with_chroot(int pid, const char *filename);
+
+int do_realloc_array_as_needed(void **arr, size_t *arr_sz, size_t x,
+			       size_t msz, const void *init_val);
+
+#define realloc_array_as_needed(a, n, x, v) ({			\
+	typeof(x) __x = (x);					\
+	__x >= (n) ?						\
+		do_realloc_array_as_needed((void **)&(a),	\
+					   &(n),		\
+					   __x,			\
+					   sizeof(*(a)),	\
+					   (const void *)(v)) :	\
+		0;						\
+	})
+
+static inline bool host_is_bigendian(void)
+{
+#ifdef __BYTE_ORDER__
+#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+	return false;
+#elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
+	return true;
+#else
+#error "Unrecognized __BYTE_ORDER__"
+#endif
+#else /* !__BYTE_ORDER__ */
+	unsigned char str[] = { 0x1, 0x2, 0x3, 0x4, 0x0, 0x0, 0x0, 0x0};
+	unsigned int *ptr;
+
+	ptr = (unsigned int *)(void *)str;
+	return *ptr == 0x01020304;
+#endif
+}
+
+#endif /* __PERF_UTIL_H */