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, 131 insertions, 0 deletions

diff --git a/lib/dump_stack.c b/lib/dump_stack.c
new file mode 100644
index 000000000..f0c78b5b5
--- /dev/null
+++ b/lib/dump_stack.c
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Provide a default dump_stack() function for architectures
+ * which don't implement their own.
+ */
+
+#include <linux/kernel.h>
+#include <linux/buildid.h>
+#include <linux/export.h>
+#include <linux/sched.h>
+#include <linux/sched/debug.h>
+#include <linux/smp.h>
+#include <linux/atomic.h>
+#include <linux/kexec.h>
+#include <linux/utsname.h>
+#include <linux/stop_machine.h>
+
+static char dump_stack_arch_desc_str[128];
+
+/**
+ * dump_stack_set_arch_desc - set arch-specific str to show with task dumps
+ * @fmt: printf-style format string
+ * @...: arguments for the format string
+ *
+ * The configured string will be printed right after utsname during task
+ * dumps.  Usually used to add arch-specific system identifiers.  If an
+ * arch wants to make use of such an ID string, it should initialize this
+ * as soon as possible during boot.
+ */
+void __init dump_stack_set_arch_desc(const char *fmt, ...)
+{
+	va_list args;
+
+	va_start(args, fmt);
+	vsnprintf(dump_stack_arch_desc_str, sizeof(dump_stack_arch_desc_str),
+		  fmt, args);
+	va_end(args);
+}
+
+#if IS_ENABLED(CONFIG_STACKTRACE_BUILD_ID)
+#define BUILD_ID_FMT " %20phN"
+#define BUILD_ID_VAL vmlinux_build_id
+#else
+#define BUILD_ID_FMT "%s"
+#define BUILD_ID_VAL ""
+#endif
+
+/**
+ * dump_stack_print_info - print generic debug info for dump_stack()
+ * @log_lvl: log level
+ *
+ * Arch-specific dump_stack() implementations can use this function to
+ * print out the same debug information as the generic dump_stack().
+ */
+void dump_stack_print_info(const char *log_lvl)
+{
+	printk("%sCPU: %d UID: %u PID: %d Comm: %.20s %s%s %s %.*s %s " BUILD_ID_FMT "\n",
+	       log_lvl, raw_smp_processor_id(),
+	       __kuid_val(current_real_cred()->euid),
+	       current->pid, current->comm,
+	       kexec_crash_loaded() ? "Kdump: loaded " : "",
+	       print_tainted(),
+	       init_utsname()->release,
+	       (int)strcspn(init_utsname()->version, " "),
+	       init_utsname()->version, preempt_model_str(), BUILD_ID_VAL);
+
+	if (get_taint())
+		printk("%s%s\n", log_lvl, print_tainted_verbose());
+
+	if (dump_stack_arch_desc_str[0] != '\0')
+		printk("%sHardware name: %s\n",
+		       log_lvl, dump_stack_arch_desc_str);
+
+	print_worker_info(log_lvl, current);
+	print_stop_info(log_lvl, current);
+	print_scx_info(log_lvl, current);
+}
+
+/**
+ * show_regs_print_info - print generic debug info for show_regs()
+ * @log_lvl: log level
+ *
+ * show_regs() implementations can use this function to print out generic
+ * debug information.
+ */
+void show_regs_print_info(const char *log_lvl)
+{
+	dump_stack_print_info(log_lvl);
+}
+
+static void __dump_stack(const char *log_lvl)
+{
+	dump_stack_print_info(log_lvl);
+	show_stack(NULL, NULL, log_lvl);
+}
+
+/**
+ * dump_stack_lvl - dump the current task information and its stack trace
+ * @log_lvl: log level
+ *
+ * Architectures can override this implementation by implementing its own.
+ */
+asmlinkage __visible void dump_stack_lvl(const char *log_lvl)
+{
+	bool in_panic = panic_on_this_cpu();
+	unsigned long flags;
+
+	/*
+	 * Permit this cpu to perform nested stack dumps while serialising
+	 * against other CPUs, unless this CPU is in panic.
+	 *
+	 * When in panic, non-panic CPUs are not permitted to store new
+	 * printk messages so there is no need to synchronize the output.
+	 * This avoids potential deadlock in panic() if another CPU is
+	 * holding and unable to release the printk_cpu_sync.
+	 */
+	if (!in_panic)
+		printk_cpu_sync_get_irqsave(flags);
+
+	__dump_stack(log_lvl);
+
+	if (!in_panic)
+		printk_cpu_sync_put_irqrestore(flags);
+}
+EXPORT_SYMBOL(dump_stack_lvl);
+
+asmlinkage __visible void dump_stack(void)
+{
+	dump_stack_lvl(KERN_DEFAULT);
+}
+EXPORT_SYMBOL(dump_stack);