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, 88 insertions, 0 deletions

diff --git a/arch/xtensa/lib/modsi3.S b/arch/xtensa/lib/modsi3.S
new file mode 100644
index 000000000..c5f4295c6
--- /dev/null
+++ b/arch/xtensa/lib/modsi3.S
@@ -0,0 +1,88 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later WITH GCC-exception-2.0 */
+#include <linux/linkage.h>
+#include <asm/asmmacro.h>
+#include <asm/core.h>
+
+ENTRY(__modsi3)
+
+	abi_entry_default
+#if XCHAL_HAVE_DIV32
+	rems	a2, a2, a3
+#else
+	mov	a7, a2		/* save original (signed) dividend */
+	do_abs	a2, a2, a4	/* udividend = abs (dividend) */
+	do_abs	a3, a3, a4	/* udivisor = abs (divisor) */
+	bltui	a3, 2, .Lle_one	/* check if udivisor <= 1 */
+	do_nsau	a5, a2, a6, a8	/* udividend_shift = nsau (udividend) */
+	do_nsau	a4, a3, a6, a8	/* udivisor_shift = nsau (udivisor) */
+	bgeu	a5, a4, .Lspecial
+
+	sub	a4, a4, a5	/* count = udivisor_shift - udividend_shift */
+	ssl	a4
+	sll	a3, a3		/* udivisor <<= count */
+
+	/* test-subtract-and-shift loop */
+#if XCHAL_HAVE_LOOPS
+	loopnez	a4, .Lloopend
+#endif /* XCHAL_HAVE_LOOPS */
+.Lloop:
+	bltu	a2, a3, .Lzerobit
+	sub	a2, a2, a3
+.Lzerobit:
+	srli	a3, a3, 1
+#if !XCHAL_HAVE_LOOPS
+	addi	a4, a4, -1
+	bnez	a4, .Lloop
+#endif /* !XCHAL_HAVE_LOOPS */
+.Lloopend:
+
+.Lspecial:
+	bltu	a2, a3, .Lreturn
+	sub	a2, a2, a3	/* subtract again if udividend >= udivisor */
+.Lreturn:
+	bgez	a7, .Lpositive
+	neg	a2, a2		/* if (dividend < 0), return -udividend */
+.Lpositive:
+	abi_ret_default
+
+.Lle_one:
+	bnez	a3, .Lreturn0
+
+	/* Divide by zero: Use an illegal instruction to force an exception.
+	   The subsequent "DIV0" string can be recognized by the exception
+	   handler to identify the real cause of the exception.  */
+	ill
+	.ascii	"DIV0"
+
+.Lreturn0:
+	movi	a2, 0
+#endif /* XCHAL_HAVE_DIV32 */
+	abi_ret_default
+
+ENDPROC(__modsi3)
+EXPORT_SYMBOL(__modsi3)
+
+#if !XCHAL_HAVE_NSA
+	.section .rodata
+	.align	4
+	.global	__nsau_data
+	.type	__nsau_data, @object
+__nsau_data:
+	.byte	8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4
+	.byte	3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3
+	.byte	2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2
+	.byte	2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2
+	.byte	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
+	.byte	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
+	.byte	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
+	.byte	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.byte	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+	.size	__nsau_data, . - __nsau_data
+#endif /* !XCHAL_HAVE_NSA */