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, 124 insertions, 0 deletions

diff --git a/arch/alpha/boot/head.S b/arch/alpha/boot/head.S
new file mode 100644
index 000000000..06a7c95fe
--- /dev/null
+++ b/arch/alpha/boot/head.S
@@ -0,0 +1,124 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * arch/alpha/boot/head.S
+ *
+ * initial bootloader stuff..
+ */
+
+#include <asm/pal.h>
+
+	.set noreorder
+	.globl	__start
+	.ent	__start
+__start:
+	br	$29,2f
+2:	ldgp	$29,0($29)
+	jsr	$26,start_kernel
+	call_pal PAL_halt
+	.end __start
+
+	.align 5
+	.globl	wrent
+	.ent	wrent
+wrent:
+	.prologue 0
+	call_pal PAL_wrent
+	ret ($26)
+	.end wrent
+
+	.align 5
+	.globl	wrkgp
+	.ent	wrkgp
+wrkgp:
+	.prologue 0
+	call_pal PAL_wrkgp
+	ret ($26)
+	.end wrkgp
+
+	.align 5
+	.globl	switch_to_osf_pal
+	.ent	switch_to_osf_pal
+switch_to_osf_pal:
+	subq	$30,128,$30
+	.frame	$30,128,$26
+	stq	$26,0($30)
+	stq	$1,8($30)
+	stq	$2,16($30)
+	stq	$3,24($30)
+	stq	$4,32($30)
+	stq	$5,40($30)
+	stq	$6,48($30)
+	stq	$7,56($30)
+	stq	$8,64($30)
+	stq	$9,72($30)
+	stq	$10,80($30)
+	stq	$11,88($30)
+	stq	$12,96($30)
+	stq	$13,104($30)
+	stq	$14,112($30)
+	stq	$15,120($30)
+	.prologue 0
+
+	stq	$30,0($17)	/* save KSP in PCB */
+
+	bis	$30,$30,$20	/* a4 = KSP */
+	br	$17,1f
+
+	ldq	$26,0($30)
+	ldq	$1,8($30)
+	ldq	$2,16($30)
+	ldq	$3,24($30)
+	ldq	$4,32($30)
+	ldq	$5,40($30)
+	ldq	$6,48($30)
+	ldq	$7,56($30)
+	ldq	$8,64($30)
+	ldq	$9,72($30)
+	ldq	$10,80($30)
+	ldq	$11,88($30)
+	ldq	$12,96($30)
+	ldq	$13,104($30)
+	ldq	$14,112($30)
+	ldq	$15,120($30)
+	addq	$30,128,$30
+	ret ($26)
+1:	call_pal PAL_swppal
+	.end	switch_to_osf_pal
+
+	.align 3
+	.globl	tbi
+	.ent	tbi
+tbi:
+	.prologue 0
+	call_pal PAL_tbi
+	ret	($26)
+	.end tbi
+
+	.align 3
+	.globl	halt
+	.ent	halt
+halt:
+	.prologue 0
+	call_pal PAL_halt
+	.end halt
+
+/* $16 - new stack page */
+	.align 3
+	.globl	move_stack
+	.ent	move_stack
+move_stack:
+	.prologue 0
+	lda	$0, 0x1fff($31)
+	and	$0, $30, $1			/* Stack offset */
+	or	$1, $16, $16			/* New stack pointer */
+	mov	$30, $1
+	mov	$16, $2
+1:	ldq	$3, 0($1)			/* Move the stack */
+	addq	$1, 8, $1
+	stq	$3, 0($2)
+	and	$0, $1, $4
+	addq	$2, 8, $2
+	bne	$4, 1b
+	mov	$16, $30
+	ret	($26)
+	.end move_stack