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

3 files changed, 222 insertions, 0 deletions

diff --git a/Documentation/devicetree/bindings/timestamp/hardware-timestamps-common.yaml b/Documentation/devicetree/bindings/timestamp/hardware-timestamps-common.yaml
new file mode 100644
index 000000000..95f42acd0
--- /dev/null
+++ b/Documentation/devicetree/bindings/timestamp/hardware-timestamps-common.yaml
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/timestamp/hardware-timestamps-common.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Hardware timestamp providers
+
+maintainers:
+  - Dipen Patel <dipenp@nvidia.com>
+
+description:
+  Some devices/SoCs have hardware timestamp engines (HTE) which can use
+  hardware means to timestamp entity in realtime. The entity could be anything
+  from GPIOs, IRQs, Bus and so on. The hardware timestamp engine present
+  itself as a provider with the bindings described in this document.
+
+properties:
+  $nodename:
+    pattern: "^timestamp(@.*|-([0-9]|[1-9][0-9]+))?$"
+
+  "#timestamp-cells":
+    description:
+      Number of cells in a HTE specifier.
+
+required:
+  - "#timestamp-cells"
+
+additionalProperties: true
diff --git a/Documentation/devicetree/bindings/timestamp/hte-consumer.yaml b/Documentation/devicetree/bindings/timestamp/hte-consumer.yaml
new file mode 100644
index 000000000..5142d6d4f
--- /dev/null
+++ b/Documentation/devicetree/bindings/timestamp/hte-consumer.yaml
@@ -0,0 +1,39 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/timestamp/hte-consumer.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: HTE Consumer
+
+maintainers:
+  - Dipen Patel <dipenp@nvidia.com>
+
+select: true
+
+properties:
+  timestamps:
+    $ref: /schemas/types.yaml#/definitions/phandle-array
+    description:
+      The list of HTE provider phandle. The first cell must represent the
+      provider phandle followed by the line identifiers. The meaning of the
+      line identifier and exact number of arguments must be specified in the
+      HTE provider device tree binding document.
+
+  timestamp-names:
+    $ref: /schemas/types.yaml#/definitions/string-array
+    description:
+      An optional string property to label each line specifier present in the
+      timestamp property.
+
+dependencies:
+  timestamp-names: [ timestamps ]
+
+additionalProperties: true
+
+examples:
+  - |
+    hte_tegra_consumer {
+              timestamps = <&tegra_hte_aon 0x9>, <&tegra_hte_lic 0x19>;
+              timestamp-names = "hte-gpio", "hte-i2c";
+    };
diff --git a/Documentation/devicetree/bindings/timestamp/nvidia,tegra194-hte.yaml b/Documentation/devicetree/bindings/timestamp/nvidia,tegra194-hte.yaml
new file mode 100644
index 000000000..a96d6cd23
--- /dev/null
+++ b/Documentation/devicetree/bindings/timestamp/nvidia,tegra194-hte.yaml
@@ -0,0 +1,154 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/timestamp/nvidia,tegra194-hte.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Tegra on chip generic hardware timestamping engine (HTE) provider
+
+maintainers:
+  - Dipen Patel <dipenp@nvidia.com>
+
+description:
+  Tegra SoC has two instances of generic hardware timestamping engines (GTE)
+  known as GTE GPIO and GTE IRQ, which can monitor subset of GPIO and on chip
+  IRQ lines for the state change respectively, upon detection it will record
+  timestamp (taken from system counter) in its internal hardware FIFO. It has
+  a bitmap array arranged in 32bit slices where each bit represent signal/line
+  to enable or disable for the hardware timestamping. The GTE GPIO monitors
+  GPIO lines from the AON (always on) GPIO controller.
+
+properties:
+  compatible:
+    enum:
+      - nvidia,tegra194-gte-aon
+      - nvidia,tegra194-gte-lic
+      - nvidia,tegra234-gte-aon
+      - nvidia,tegra234-gte-lic
+      - nvidia,tegra264-gte-aon
+      - nvidia,tegra264-gte-lic
+
+  reg:
+    maxItems: 1
+
+  interrupts:
+    maxItems: 1
+
+  nvidia,int-threshold:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    description:
+      HTE device generates its interrupt based on this u32 FIFO threshold
+      value. The recommended value is 1.
+    minimum: 1
+    maximum: 256
+
+  nvidia,slices:
+    $ref: /schemas/types.yaml#/definitions/uint32
+    deprecated: true
+    description:
+      HTE lines are arranged in 32 bit slice where each bit represents different
+      line/signal that it can enable/configure for the timestamp. It is u32
+      property and the value depends on the HTE instance in the chip. The AON
+      GTE instances for both Tegra194 and Tegra234 has 3 slices. The Tegra194
+      LIC instance has 11 slices and Tegra234 LIC has 17 slices.
+    enum: [3, 11, 17]
+
+  nvidia,gpio-controller:
+    $ref: /schemas/types.yaml#/definitions/phandle
+    description:
+      The phandle to AON gpio controller instance. This is required to handle
+      namespace conversion between GPIO and GTE.
+
+  '#timestamp-cells':
+    description:
+      This represents number of line id arguments as specified by the
+      consumers. For the GTE IRQ, this is IRQ number as mentioned in the
+      SoC technical reference manual. For the GTE GPIO, its value is same as
+      mentioned in the nvidia GPIO device tree binding document.
+    const: 1
+
+required:
+  - compatible
+  - reg
+  - interrupts
+  - "#timestamp-cells"
+
+allOf:
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - nvidia,tegra194-gte-aon
+              - nvidia,tegra234-gte-aon
+    then:
+      properties:
+        nvidia,slices:
+          const: 3
+
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - nvidia,tegra194-gte-lic
+    then:
+      properties:
+        nvidia,slices:
+          const: 11
+
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - nvidia,tegra234-gte-lic
+    then:
+      properties:
+        nvidia,slices:
+          const: 17
+
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - nvidia,tegra234-gte-aon
+              - nvidia,tegra264-gte-aon
+    then:
+      required:
+        - nvidia,gpio-controller
+
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - nvidia,tegra264-gte-aon
+              - nvidia,tegra264-gte-lic
+    then:
+      properties:
+        nvidia,slices: false
+
+additionalProperties: false
+
+examples:
+  - |
+    tegra_hte_aon: timestamp@c1e0000 {
+              compatible = "nvidia,tegra194-gte-aon";
+              reg = <0xc1e0000 0x10000>;
+              interrupts = <0 13 0x4>;
+              nvidia,int-threshold = <1>;
+              #timestamp-cells = <1>;
+    };
+
+  - |
+    tegra_hte_lic: timestamp@3aa0000 {
+              compatible = "nvidia,tegra194-gte-lic";
+              reg = <0x3aa0000 0x10000>;
+              interrupts = <0 11 0x4>;
+              nvidia,int-threshold = <1>;
+              #timestamp-cells = <1>;
+    };
+
+...