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

6 files changed, 82 insertions, 0 deletions

diff --git a/Documentation/devicetree/bindings/arc/axs101.txt b/Documentation/devicetree/bindings/arc/axs101.txt
new file mode 100644
index 000000000..48290d517
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/axs101.txt
@@ -0,0 +1,7 @@
+Synopsys DesignWare ARC Software Development Platforms Device Tree Bindings
+---------------------------------------------------------------------------
+
+SDP Main Board with an AXC001 CPU Card hoisting ARC700 core in silicon
+
+Required root node properties:
+    - compatible = "snps,axs101", "snps,arc-sdp";
diff --git a/Documentation/devicetree/bindings/arc/axs103.txt b/Documentation/devicetree/bindings/arc/axs103.txt
new file mode 100644
index 000000000..6eea862e7
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/axs103.txt
@@ -0,0 +1,8 @@
+Synopsys DesignWare ARC Software Development Platforms Device Tree Bindings
+---------------------------------------------------------------------------
+
+SDP Main Board with an AXC003 FPGA Card which can contain various flavours of
+HS38x cores.
+
+Required root node properties:
+    - compatible = "snps,axs103", "snps,arc-sdp";
diff --git a/Documentation/devicetree/bindings/arc/eznps.txt b/Documentation/devicetree/bindings/arc/eznps.txt
new file mode 100644
index 000000000..1aa50c640
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/eznps.txt
@@ -0,0 +1,7 @@
+EZchip NPS Network Processor Platforms Device Tree Bindings
+---------------------------------------------------------------------------
+
+Appliance main board with NPS400 ASIC.
+
+Required root node properties:
+    - compatible = "ezchip,arc-nps";
diff --git a/Documentation/devicetree/bindings/arc/hsdk.txt b/Documentation/devicetree/bindings/arc/hsdk.txt
new file mode 100644
index 000000000..be50654bb
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/hsdk.txt
@@ -0,0 +1,7 @@
+Synopsys DesignWare ARC HS Development Kit Device Tree Bindings
+---------------------------------------------------------------------------
+
+ARC HSDK Board with quad-core ARC HS38x4 in silicon.
+
+Required root node properties:
+    - compatible = "snps,hsdk";
diff --git a/Documentation/devicetree/bindings/arc/pct.txt b/Documentation/devicetree/bindings/arc/pct.txt
new file mode 100644
index 000000000..4e874d9a3
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/pct.txt
@@ -0,0 +1,20 @@
+* ARC Performance Counters
+
+The ARC700 can be configured with a pipeline performance monitor for counting
+CPU and cache events like cache misses and hits. Like conventional PCT there
+are 100+ hardware conditions dynamically mapped to up to 32 counters
+
+Note that:
+ * The ARC 700 PCT does not support interrupts; although HW events may be
+   counted, the HW events themselves cannot serve as a trigger for a sample.
+
+Required properties:
+
+- compatible : should contain
+	"snps,arc700-pct"
+
+Example:
+
+pmu {
+        compatible = "snps,arc700-pct";
+};
diff --git a/Documentation/devicetree/bindings/arc/snps,archs-pct.yaml b/Documentation/devicetree/bindings/arc/snps,archs-pct.yaml
new file mode 100644
index 000000000..532f7584f
--- /dev/null
+++ b/Documentation/devicetree/bindings/arc/snps,archs-pct.yaml
@@ -0,0 +1,33 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/arc/snps,archs-pct.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: ARC HS Performance Counters
+
+maintainers:
+  - Aryabhatta Dey <aryabhattadey35@gmail.com>
+
+description:
+  The ARC HS can be configured with a pipeline performance monitor for counting
+  CPU and cache events like cache misses and hits. Like conventional PCT there
+  are 100+ hardware conditions dynamically mapped to up to 32 counters.
+  It also supports overflow interrupts.
+
+properties:
+  compatible:
+    const: snps,archs-pct
+
+  reg:
+    maxItems: 1
+
+  clocks:
+    maxItems: 1
+
+required:
+  - compatible
+  - reg
+  - clocks
+
+additionalProperties: false