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, 252 insertions, 0 deletions

diff --git a/arch/powerpc/boot/string.S b/arch/powerpc/boot/string.S
new file mode 100644
index 000000000..76801eb6d
--- /dev/null
+++ b/arch/powerpc/boot/string.S
@@ -0,0 +1,252 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (C) Paul Mackerras 1997.
+ *
+ * NOTE: this code runs in 32 bit mode and is packaged as ELF32.
+ */
+
+#include "ppc_asm.h"
+
+	.text
+	.globl	strcpy
+strcpy:
+	addi	r5,r3,-1
+	addi	r4,r4,-1
+1:	lbzu	r0,1(r4)
+	cmpwi	0,r0,0
+	stbu	r0,1(r5)
+	bne	1b
+	blr
+
+	.globl	strcat
+strcat:
+	addi	r5,r3,-1
+	addi	r4,r4,-1
+1:	lbzu	r0,1(r5)
+	cmpwi	0,r0,0
+	bne	1b
+	addi	r5,r5,-1
+1:	lbzu	r0,1(r4)
+	cmpwi	0,r0,0
+	stbu	r0,1(r5)
+	bne	1b
+	blr
+
+	.globl	strchr
+strchr:
+	addi	r3,r3,-1
+1:	lbzu	r0,1(r3)
+	cmpw	0,r0,r4
+	beqlr
+	cmpwi	0,r0,0
+	bne	1b
+	li	r3,0
+	blr
+
+	.globl	strcmp
+strcmp:
+	addi	r5,r3,-1
+	addi	r4,r4,-1
+1:	lbzu	r3,1(r5)
+	cmpwi	1,r3,0
+	lbzu	r0,1(r4)
+	subf.	r3,r0,r3
+	beqlr	1
+	beq	1b
+	blr
+
+	.globl	strncmp
+strncmp:
+	mtctr	r5
+	addi	r5,r3,-1
+	addi	r4,r4,-1
+1:	lbzu	r3,1(r5)
+	cmpwi	1,r3,0
+	lbzu	r0,1(r4)
+	subf.	r3,r0,r3
+	beqlr	1
+	bdnzt	eq,1b
+	blr
+
+	.globl	strlen
+strlen:
+	addi	r4,r3,-1
+1:	lbzu	r0,1(r4)
+	cmpwi	0,r0,0
+	bne	1b
+	subf	r3,r3,r4
+	blr
+
+	.globl	memset
+memset:
+	rlwimi	r4,r4,8,16,23
+	rlwimi	r4,r4,16,0,15
+	addi	r6,r3,-4
+	cmplwi	0,r5,4
+	blt	7f
+	stwu	r4,4(r6)
+	beqlr
+	andi.	r0,r6,3
+	add	r5,r0,r5
+	subf	r6,r0,r6
+	rlwinm	r0,r5,32-2,2,31
+	mtctr	r0
+	bdz	6f
+1:	stwu	r4,4(r6)
+	bdnz	1b
+6:	andi.	r5,r5,3
+7:	cmpwi	0,r5,0
+	beqlr
+	mtctr	r5
+	addi	r6,r6,3
+8:	stbu	r4,1(r6)
+	bdnz	8b
+	blr
+
+	.globl	memmove
+memmove:
+	cmplw	0,r3,r4
+	bgt	backwards_memcpy
+	/* fall through */
+
+	.globl	memcpy
+memcpy:
+	rlwinm.	r7,r5,32-3,3,31		/* r7 = r5 >> 3 */
+	addi	r6,r3,-4
+	addi	r4,r4,-4
+	beq	3f			/* if less than 8 bytes to do */
+	andi.	r0,r6,3			/* get dest word aligned */
+	mtctr	r7
+	bne	5f
+	andi.	r0,r4,3			/* check src word aligned too */
+	bne	3f
+1:	lwz	r7,4(r4)
+	lwzu	r8,8(r4)
+	stw	r7,4(r6)
+	stwu	r8,8(r6)
+	bdnz	1b
+	andi.	r5,r5,7
+2:	cmplwi	0,r5,4
+	blt	3f
+	lwzu	r0,4(r4)
+	addi	r5,r5,-4
+	stwu	r0,4(r6)
+3:	cmpwi	0,r5,0
+	beqlr
+	mtctr	r5
+	addi	r4,r4,3
+	addi	r6,r6,3
+4:	lbzu	r0,1(r4)
+	stbu	r0,1(r6)
+	bdnz	4b
+	blr
+5:	subfic	r0,r0,4
+	cmpw	cr1,r0,r5
+	add	r7,r0,r4
+	andi.	r7,r7,3			/* will source be word-aligned too? */
+	ble	cr1,3b
+	bne	3b			/* do byte-by-byte if not */
+	mtctr	r0
+6:	lbz	r7,4(r4)
+	addi	r4,r4,1
+	stb	r7,4(r6)
+	addi	r6,r6,1
+	bdnz	6b
+	subf	r5,r0,r5
+	rlwinm.	r7,r5,32-3,3,31
+	beq	2b
+	mtctr	r7
+	b	1b
+
+	.globl	backwards_memcpy
+backwards_memcpy:
+	rlwinm.	r7,r5,32-3,3,31		/* r7 = r5 >> 3 */
+	add	r6,r3,r5
+	add	r4,r4,r5
+	beq	3f
+	andi.	r0,r6,3
+	mtctr	r7
+	bne	5f
+	andi.	r0,r4,3
+	bne	3f
+1:	lwz	r7,-4(r4)
+	lwzu	r8,-8(r4)
+	stw	r7,-4(r6)
+	stwu	r8,-8(r6)
+	bdnz	1b
+	andi.	r5,r5,7
+2:	cmplwi	0,r5,4
+	blt	3f
+	lwzu	r0,-4(r4)
+	subi	r5,r5,4
+	stwu	r0,-4(r6)
+3:	cmpwi	0,r5,0
+	beqlr
+	mtctr	r5
+4:	lbzu	r0,-1(r4)
+	stbu	r0,-1(r6)
+	bdnz	4b
+	blr
+5:	cmpw	cr1,r0,r5
+	subf	r7,r0,r4
+	andi.	r7,r7,3
+	ble	cr1,3b
+	bne	3b
+	mtctr	r0
+6:	lbzu	r7,-1(r4)
+	stbu	r7,-1(r6)
+	bdnz	6b
+	subf	r5,r0,r5
+	rlwinm.	r7,r5,32-3,3,31
+	beq	2b
+	mtctr	r7
+	b	1b
+
+	.globl	memchr
+memchr:
+	cmpwi	0,r5,0
+	blelr
+	mtctr	r5
+	addi	r3,r3,-1
+1:	lbzu	r0,1(r3)
+	cmpw	r0,r4
+	beqlr
+	bdnz	1b
+	li	r3,0
+	blr
+
+	.globl	memcmp
+memcmp:
+	cmpwi	0,r5,0
+	ble	2f
+	mtctr	r5
+	addi	r6,r3,-1
+	addi	r4,r4,-1
+1:	lbzu	r3,1(r6)
+	lbzu	r0,1(r4)
+	subf.	r3,r0,r3
+	bdnzt	2,1b
+	blr
+2:	li	r3,0
+	blr
+
+
+/*
+ * Flush the dcache and invalidate the icache for a range of addresses.
+ *
+ * flush_cache(addr, len)
+ */
+	.global	flush_cache
+flush_cache:
+	addi	4,4,0x1f	/* len = (len + 0x1f) / 0x20 */
+	rlwinm.	4,4,27,5,31
+	mtctr	4
+	beqlr
+1:	dcbf	0,3
+	icbi	0,3
+	addi	3,3,0x20
+	bdnz	1b
+	sync
+	isync
+	blr
+