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, 73 insertions, 0 deletions

diff --git a/include/clocksource/timer-xilinx.h b/include/clocksource/timer-xilinx.h
new file mode 100644
index 000000000..d116f18de
--- /dev/null
+++ b/include/clocksource/timer-xilinx.h
@@ -0,0 +1,73 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Copyright (C) 2021 Sean Anderson <sean.anderson@seco.com>
+ */
+
+#ifndef XILINX_TIMER_H
+#define XILINX_TIMER_H
+
+#include <linux/compiler.h>
+
+#define TCSR0	0x00
+#define TLR0	0x04
+#define TCR0	0x08
+#define TCSR1	0x10
+#define TLR1	0x14
+#define TCR1	0x18
+
+#define TCSR_MDT	BIT(0)
+#define TCSR_UDT	BIT(1)
+#define TCSR_GENT	BIT(2)
+#define TCSR_CAPT	BIT(3)
+#define TCSR_ARHT	BIT(4)
+#define TCSR_LOAD	BIT(5)
+#define TCSR_ENIT	BIT(6)
+#define TCSR_ENT	BIT(7)
+#define TCSR_TINT	BIT(8)
+#define TCSR_PWMA	BIT(9)
+#define TCSR_ENALL	BIT(10)
+#define TCSR_CASC	BIT(11)
+
+struct clk;
+struct device_node;
+struct regmap;
+
+/**
+ * struct xilinx_timer_priv - Private data for Xilinx AXI timer drivers
+ * @map: Regmap of the device, possibly with an offset
+ * @clk: Parent clock
+ * @max: Maximum value of the counters
+ */
+struct xilinx_timer_priv {
+	struct regmap *map;
+	struct clk *clk;
+	u64 max;
+};
+
+/**
+ * xilinx_timer_tlr_cycles() - Calculate the TLR for a period specified
+ *                             in clock cycles
+ * @priv: The timer's private data
+ * @tcsr: The value of the TCSR register for this counter
+ * @cycles: The number of cycles in this period
+ *
+ * Callers of this function MUST ensure that @cycles is representable as
+ * a TLR.
+ *
+ * Return: The calculated value for TLR
+ */
+u32 xilinx_timer_tlr_cycles(struct xilinx_timer_priv *priv, u32 tcsr,
+			    u64 cycles);
+
+/**
+ * xilinx_timer_get_period() - Get the current period of a counter
+ * @priv: The timer's private data
+ * @tlr: The value of TLR for this counter
+ * @tcsr: The value of TCSR for this counter
+ *
+ * Return: The period, in ns
+ */
+unsigned int xilinx_timer_get_period(struct xilinx_timer_priv *priv,
+				     u32 tlr, u32 tcsr);
+
+#endif /* XILINX_TIMER_H */