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, 116 insertions, 0 deletions

diff --git a/drivers/crypto/ccp/tee-dev.h b/drivers/crypto/ccp/tee-dev.h
new file mode 100644
index 000000000..c23416cb7
--- /dev/null
+++ b/drivers/crypto/ccp/tee-dev.h
@@ -0,0 +1,116 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2019,2021 Advanced Micro Devices, Inc.
+ *
+ * Author: Rijo Thomas <Rijo-john.Thomas@amd.com>
+ * Author: Devaraj Rangasamy <Devaraj.Rangasamy@amd.com>
+ *
+ */
+
+/* This file describes the TEE communication interface between host and AMD
+ * Secure Processor
+ */
+
+#ifndef __TEE_DEV_H__
+#define __TEE_DEV_H__
+
+#include <linux/device.h>
+#include <linux/mutex.h>
+
+#define TEE_DEFAULT_CMD_TIMEOUT		(10 * MSEC_PER_SEC)
+#define TEE_DEFAULT_RING_TIMEOUT	10
+#define MAX_BUFFER_SIZE			988
+
+/**
+ * struct tee_init_ring_cmd - Command to init TEE ring buffer
+ * @low_addr:  bits [31:0] of the physical address of ring buffer
+ * @hi_addr:   bits [63:32] of the physical address of ring buffer
+ * @size:      size of ring buffer in bytes
+ */
+struct tee_init_ring_cmd {
+	u32 low_addr;
+	u32 hi_addr;
+	u32 size;
+};
+
+#define MAX_RING_BUFFER_ENTRIES		32
+
+/**
+ * struct ring_buf_manager - Helper structure to manage ring buffer.
+ * @ring_start:  starting address of ring buffer
+ * @ring_size:   size of ring buffer in bytes
+ * @ring_pa:     physical address of ring buffer
+ * @wptr:        index to the last written entry in ring buffer
+ */
+struct ring_buf_manager {
+	struct mutex mutex;	/* synchronizes access to ring buffer */
+	void *ring_start;
+	u32 ring_size;
+	phys_addr_t ring_pa;
+	u32 wptr;
+};
+
+struct psp_tee_device {
+	struct device *dev;
+	struct psp_device *psp;
+	void __iomem *io_regs;
+	struct tee_vdata *vdata;
+	struct ring_buf_manager rb_mgr;
+};
+
+/**
+ * enum tee_cmd_state - TEE command states for the ring buffer interface
+ * @TEE_CMD_STATE_INIT:      initial state of command when sent from host
+ * @TEE_CMD_STATE_PROCESS:   command being processed by TEE environment
+ * @TEE_CMD_STATE_COMPLETED: command processing completed
+ */
+enum tee_cmd_state {
+	TEE_CMD_STATE_INIT,
+	TEE_CMD_STATE_PROCESS,
+	TEE_CMD_STATE_COMPLETED,
+};
+
+/**
+ * enum cmd_resp_state - TEE command's response status maintained by driver
+ * @CMD_RESPONSE_INVALID:      initial state when no command is written to ring
+ * @CMD_WAITING_FOR_RESPONSE:  driver waiting for response from TEE
+ * @CMD_RESPONSE_TIMEDOUT:     failed to get response from TEE
+ * @CMD_RESPONSE_COPIED:       driver has copied response from TEE
+ */
+enum cmd_resp_state {
+	CMD_RESPONSE_INVALID,
+	CMD_WAITING_FOR_RESPONSE,
+	CMD_RESPONSE_TIMEDOUT,
+	CMD_RESPONSE_COPIED,
+};
+
+/**
+ * struct tee_ring_cmd - Structure of the command buffer in TEE ring
+ * @cmd_id:      refers to &enum tee_cmd_id. Command id for the ring buffer
+ *               interface
+ * @cmd_state:   refers to &enum tee_cmd_state
+ * @status:      status of TEE command execution
+ * @res0:        reserved region
+ * @pdata:       private data (currently unused)
+ * @res1:        reserved region
+ * @buf:         TEE command specific buffer
+ * @flag:	 refers to &enum cmd_resp_state
+ */
+struct tee_ring_cmd {
+	u32 cmd_id;
+	u32 cmd_state;
+	u32 status;
+	u32 res0[1];
+	u64 pdata;
+	u32 res1[2];
+	u8 buf[MAX_BUFFER_SIZE];
+	u32 flag;
+
+	/* Total size: 1024 bytes */
+} __packed;
+
+int tee_dev_init(struct psp_device *psp);
+void tee_dev_destroy(struct psp_device *psp);
+int tee_restore(struct psp_device *psp);
+
+#endif /* __TEE_DEV_H__ */