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, 101 insertions, 0 deletions

diff --git a/drivers/dma/dw/internal.h b/drivers/dma/dw/internal.h
new file mode 100644
index 000000000..f1bd06a20
--- /dev/null
+++ b/drivers/dma/dw/internal.h
@@ -0,0 +1,101 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Driver for the Synopsys DesignWare DMA Controller
+ *
+ * Copyright (C) 2013 Intel Corporation
+ */
+
+#ifndef _DMA_DW_INTERNAL_H
+#define _DMA_DW_INTERNAL_H
+
+#include <linux/dma/dw.h>
+
+#include "regs.h"
+
+int do_dma_probe(struct dw_dma_chip *chip);
+int do_dma_remove(struct dw_dma_chip *chip);
+
+void do_dw_dma_on(struct dw_dma *dw);
+void do_dw_dma_off(struct dw_dma *dw);
+
+int do_dw_dma_disable(struct dw_dma_chip *chip);
+int do_dw_dma_enable(struct dw_dma_chip *chip);
+
+extern bool dw_dma_filter(struct dma_chan *chan, void *param);
+
+#ifdef CONFIG_ACPI
+void dw_dma_acpi_controller_register(struct dw_dma *dw);
+void dw_dma_acpi_controller_free(struct dw_dma *dw);
+#else /* !CONFIG_ACPI */
+static inline void dw_dma_acpi_controller_register(struct dw_dma *dw) {}
+static inline void dw_dma_acpi_controller_free(struct dw_dma *dw) {}
+#endif /* !CONFIG_ACPI */
+
+struct platform_device;
+
+#ifdef CONFIG_OF
+struct dw_dma_platform_data *dw_dma_parse_dt(struct platform_device *pdev);
+void dw_dma_of_controller_register(struct dw_dma *dw);
+void dw_dma_of_controller_free(struct dw_dma *dw);
+#else
+static inline struct dw_dma_platform_data *dw_dma_parse_dt(struct platform_device *pdev)
+{
+	return NULL;
+}
+static inline void dw_dma_of_controller_register(struct dw_dma *dw) {}
+static inline void dw_dma_of_controller_free(struct dw_dma *dw) {}
+#endif
+
+struct dw_dma_chip_pdata {
+	const struct dw_dma_platform_data *pdata;
+	int (*probe)(struct dw_dma_chip *chip);
+	int (*remove)(struct dw_dma_chip *chip);
+	struct dw_dma_chip *chip;
+	u8 m_master;
+	u8 p_master;
+};
+
+static __maybe_unused const struct dw_dma_chip_pdata dw_dma_chip_pdata = {
+	.probe = dw_dma_probe,
+	.remove = dw_dma_remove,
+	.m_master = 0,
+	.p_master = 1,
+};
+
+static const struct dw_dma_platform_data idma32_pdata = {
+	.nr_channels = 8,
+	.chan_allocation_order = CHAN_ALLOCATION_ASCENDING,
+	.chan_priority = CHAN_PRIORITY_ASCENDING,
+	.block_size = 131071,
+	.nr_masters = 1,
+	.data_width = {4},
+	.multi_block = {1, 1, 1, 1, 1, 1, 1, 1},
+};
+
+static __maybe_unused const struct dw_dma_chip_pdata idma32_chip_pdata = {
+	.pdata = &idma32_pdata,
+	.probe = idma32_dma_probe,
+	.remove = idma32_dma_remove,
+	.m_master = 0,
+	.p_master = 0,
+};
+
+static const struct dw_dma_platform_data xbar_pdata = {
+	.nr_channels = 8,
+	.chan_allocation_order = CHAN_ALLOCATION_ASCENDING,
+	.chan_priority = CHAN_PRIORITY_ASCENDING,
+	.block_size = 131071,
+	.nr_masters = 1,
+	.data_width = {4},
+	.quirks = DW_DMA_QUIRK_XBAR_PRESENT,
+};
+
+static __maybe_unused const struct dw_dma_chip_pdata xbar_chip_pdata = {
+	.pdata = &xbar_pdata,
+	.probe = idma32_dma_probe,
+	.remove = idma32_dma_remove,
+	.m_master = 0,
+	.p_master = 0,
+};
+
+#endif /* _DMA_DW_INTERNAL_H */