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, 155 insertions, 0 deletions

diff --git a/drivers/acpi/riscv/cppc.c b/drivers/acpi/riscv/cppc.c
new file mode 100644
index 000000000..42c1a9052
--- /dev/null
+++ b/drivers/acpi/riscv/cppc.c
@@ -0,0 +1,155 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Implement CPPC FFH helper routines for RISC-V.
+ *
+ * Copyright (C) 2024 Ventana Micro Systems Inc.
+ */
+
+#include <acpi/cppc_acpi.h>
+#include <asm/csr.h>
+#include <asm/sbi.h>
+
+#define SBI_EXT_CPPC 0x43505043
+
+/* CPPC interfaces defined in SBI spec */
+#define SBI_CPPC_PROBE			0x0
+#define SBI_CPPC_READ			0x1
+#define SBI_CPPC_READ_HI		0x2
+#define SBI_CPPC_WRITE			0x3
+
+/* RISC-V FFH definitions from RISC-V FFH spec */
+#define FFH_CPPC_TYPE(r)		(((r) & GENMASK_ULL(63, 60)) >> 60)
+#define FFH_CPPC_SBI_REG(r)		((r) & GENMASK(31, 0))
+#define FFH_CPPC_CSR_NUM(r)		((r) & GENMASK(11, 0))
+
+#define FFH_CPPC_SBI			0x1
+#define FFH_CPPC_CSR			0x2
+
+struct sbi_cppc_data {
+	u64 val;
+	u32 reg;
+	struct sbiret ret;
+};
+
+static bool cppc_ext_present;
+
+static int __init sbi_cppc_init(void)
+{
+	if (sbi_spec_version >= sbi_mk_version(2, 0) &&
+	    sbi_probe_extension(SBI_EXT_CPPC) > 0) {
+		cppc_ext_present = true;
+	} else {
+		cppc_ext_present = false;
+	}
+
+	return 0;
+}
+device_initcall(sbi_cppc_init);
+
+static void sbi_cppc_read(void *read_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)read_data;
+
+	data->ret = sbi_ecall(SBI_EXT_CPPC, SBI_CPPC_READ,
+			      data->reg, 0, 0, 0, 0, 0);
+}
+
+static void sbi_cppc_write(void *write_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)write_data;
+
+	data->ret = sbi_ecall(SBI_EXT_CPPC, SBI_CPPC_WRITE,
+			      data->reg, data->val, 0, 0, 0, 0);
+}
+
+static void cppc_ffh_csr_read(void *read_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)read_data;
+
+	switch (data->reg) {
+	/* Support only TIME CSR for now */
+	case CSR_TIME:
+		data->ret.value = csr_read(CSR_TIME);
+		data->ret.error = 0;
+		break;
+	default:
+		data->ret.error = -EINVAL;
+		break;
+	}
+}
+
+static void cppc_ffh_csr_write(void *write_data)
+{
+	struct sbi_cppc_data *data = (struct sbi_cppc_data *)write_data;
+
+	data->ret.error = -EINVAL;
+}
+
+/*
+ * Refer to drivers/acpi/cppc_acpi.c for the description of the functions
+ * below.
+ */
+bool cpc_ffh_supported(void)
+{
+	return true;
+}
+
+int cpc_read_ffh(int cpu, struct cpc_reg *reg, u64 *val)
+{
+	struct sbi_cppc_data data;
+
+	if (WARN_ON_ONCE(irqs_disabled()))
+		return -EPERM;
+
+	if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_SBI) {
+		if (!cppc_ext_present)
+			return -EINVAL;
+
+		data.reg = FFH_CPPC_SBI_REG(reg->address);
+
+		smp_call_function_single(cpu, sbi_cppc_read, &data, 1);
+
+		*val = data.ret.value;
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	} else if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_CSR) {
+		data.reg = FFH_CPPC_CSR_NUM(reg->address);
+
+		smp_call_function_single(cpu, cppc_ffh_csr_read, &data, 1);
+
+		*val = data.ret.value;
+
+		return data.ret.error;
+	}
+
+	return -EINVAL;
+}
+
+int cpc_write_ffh(int cpu, struct cpc_reg *reg, u64 val)
+{
+	struct sbi_cppc_data data;
+
+	if (WARN_ON_ONCE(irqs_disabled()))
+		return -EPERM;
+
+	if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_SBI) {
+		if (!cppc_ext_present)
+			return -EINVAL;
+
+		data.reg = FFH_CPPC_SBI_REG(reg->address);
+		data.val = val;
+
+		smp_call_function_single(cpu, sbi_cppc_write, &data, 1);
+
+		return (data.ret.error) ? sbi_err_map_linux_errno(data.ret.error) : 0;
+	} else if (FFH_CPPC_TYPE(reg->address) == FFH_CPPC_CSR) {
+		data.reg = FFH_CPPC_CSR_NUM(reg->address);
+		data.val = val;
+
+		smp_call_function_single(cpu, cppc_ffh_csr_write, &data, 1);
+
+		return data.ret.error;
+	}
+
+	return -EINVAL;
+}