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

4 files changed, 117 insertions, 0 deletions

diff --git a/arch/mips/power/Makefile b/arch/mips/power/Makefile
new file mode 100644
index 000000000..14b7d9ee6
--- /dev/null
+++ b/arch/mips/power/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+obj-$(CONFIG_HIBERNATION) += cpu.o hibernate.o hibernate_asm.o
diff --git a/arch/mips/power/cpu.c b/arch/mips/power/cpu.c
new file mode 100644
index 000000000..d8ef7778e
--- /dev/null
+++ b/arch/mips/power/cpu.c
@@ -0,0 +1,43 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Suspend support specific for mips.
+ *
+ * Copyright (C) 2009 Lemote Inc.
+ * Author: Hu Hongbing <huhb@lemote.com>
+ *	   Wu Zhangjin <wuzhangjin@gmail.com>
+ */
+#include <linux/suspend.h>
+#include <asm/sections.h>
+#include <asm/fpu.h>
+#include <asm/dsp.h>
+
+static u32 saved_status;
+struct pt_regs saved_regs;
+
+void save_processor_state(void)
+{
+	saved_status = read_c0_status();
+
+	if (is_fpu_owner())
+		save_fp(current);
+
+	save_dsp(current);
+}
+
+void restore_processor_state(void)
+{
+	write_c0_status(saved_status);
+
+	if (is_fpu_owner())
+		restore_fp(current);
+
+	restore_dsp(current);
+}
+
+int pfn_is_nosave(unsigned long pfn)
+{
+	unsigned long nosave_begin_pfn = PFN_DOWN(__pa(&__nosave_begin));
+	unsigned long nosave_end_pfn = PFN_UP(__pa(&__nosave_end));
+
+	return	(pfn >= nosave_begin_pfn) && (pfn < nosave_end_pfn);
+}
diff --git a/arch/mips/power/hibernate.c b/arch/mips/power/hibernate.c
new file mode 100644
index 000000000..192879e76
--- /dev/null
+++ b/arch/mips/power/hibernate.c
@@ -0,0 +1,12 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/suspend.h>
+#include <asm/tlbflush.h>
+
+extern int restore_image(void);
+
+int swsusp_arch_resume(void)
+{
+	/* Avoid TLB mismatch during and after kernel resume */
+	local_flush_tlb_all();
+	return restore_image();
+}
diff --git a/arch/mips/power/hibernate_asm.S b/arch/mips/power/hibernate_asm.S
new file mode 100644
index 000000000..e62538734
--- /dev/null
+++ b/arch/mips/power/hibernate_asm.S
@@ -0,0 +1,60 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Hibernation support specific for mips - temporary page tables
+ *
+ * Copyright (C) 2009 Lemote Inc.
+ * Author: Hu Hongbing <huhb@lemote.com>
+ *	   Wu Zhangjin <wuzhangjin@gmail.com>
+ */
+#include <asm/asm-offsets.h>
+#include <asm/regdef.h>
+#include <asm/asm.h>
+
+.text
+LEAF(swsusp_arch_suspend)
+	PTR_LA t0, saved_regs
+	PTR_S ra, PT_R31(t0)
+	PTR_S sp, PT_R29(t0)
+	PTR_S fp, PT_R30(t0)
+	PTR_S gp, PT_R28(t0)
+	PTR_S s0, PT_R16(t0)
+	PTR_S s1, PT_R17(t0)
+	PTR_S s2, PT_R18(t0)
+	PTR_S s3, PT_R19(t0)
+	PTR_S s4, PT_R20(t0)
+	PTR_S s5, PT_R21(t0)
+	PTR_S s6, PT_R22(t0)
+	PTR_S s7, PT_R23(t0)
+	j swsusp_save
+END(swsusp_arch_suspend)
+
+LEAF(restore_image)
+	PTR_L t0, restore_pblist
+0:
+	PTR_L t1, PBE_ADDRESS(t0)   /* source */
+	PTR_L t2, PBE_ORIG_ADDRESS(t0) /* destination */
+	PTR_ADDU t3, t1, _PAGE_SIZE
+1:
+	REG_L t8, (t1)
+	REG_S t8, (t2)
+	PTR_ADDIU t1, t1, SZREG
+	PTR_ADDIU t2, t2, SZREG
+	bne t1, t3, 1b
+	PTR_L t0, PBE_NEXT(t0)
+	bnez t0, 0b
+	PTR_LA t0, saved_regs
+	PTR_L ra, PT_R31(t0)
+	PTR_L sp, PT_R29(t0)
+	PTR_L fp, PT_R30(t0)
+	PTR_L gp, PT_R28(t0)
+	PTR_L s0, PT_R16(t0)
+	PTR_L s1, PT_R17(t0)
+	PTR_L s2, PT_R18(t0)
+	PTR_L s3, PT_R19(t0)
+	PTR_L s4, PT_R20(t0)
+	PTR_L s5, PT_R21(t0)
+	PTR_L s6, PT_R22(t0)
+	PTR_L s7, PT_R23(t0)
+	PTR_LI v0, 0x0
+	jr ra
+END(restore_image)