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, 190 insertions, 0 deletions

diff --git a/drivers/crypto/ccp/sp-dev.h b/drivers/crypto/ccp/sp-dev.h
new file mode 100644
index 000000000..a83751cfd
--- /dev/null
+++ b/drivers/crypto/ccp/sp-dev.h
@@ -0,0 +1,190 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * AMD Secure Processor driver
+ *
+ * Copyright (C) 2017-2019 Advanced Micro Devices, Inc.
+ *
+ * Author: Tom Lendacky <thomas.lendacky@amd.com>
+ * Author: Gary R Hook <gary.hook@amd.com>
+ * Author: Brijesh Singh <brijesh.singh@amd.com>
+ */
+
+#ifndef __SP_DEV_H__
+#define __SP_DEV_H__
+
+#include <linux/device.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <linux/wait.h>
+#include <linux/dmapool.h>
+#include <linux/hw_random.h>
+#include <linux/bitops.h>
+#include <linux/interrupt.h>
+#include <linux/irqreturn.h>
+
+#define SP_MAX_NAME_LEN		32
+
+#define CACHE_NONE			0x00
+#define CACHE_WB_NO_ALLOC		0xb7
+
+#define PLATFORM_FEATURE_DBC		0x1
+#define PLATFORM_FEATURE_HSTI		0x2
+
+#define PSP_FEATURE(psp, feat)	(psp->vdata && psp->vdata->platform_features & PLATFORM_FEATURE_##feat)
+
+/* Structure to hold CCP device data */
+struct ccp_device;
+struct ccp_vdata {
+	const unsigned int version;
+	const unsigned int dma_chan_attr;
+	void (*setup)(struct ccp_device *);
+	const struct ccp_actions *perform;
+	const unsigned int offset;
+	const unsigned int rsamax;
+};
+
+struct sev_vdata {
+	const unsigned int cmdresp_reg;
+	const unsigned int cmdbuff_addr_lo_reg;
+	const unsigned int cmdbuff_addr_hi_reg;
+};
+
+struct tee_vdata {
+	const unsigned int cmdresp_reg;
+	const unsigned int cmdbuff_addr_lo_reg;
+	const unsigned int cmdbuff_addr_hi_reg;
+	const unsigned int ring_wptr_reg;
+	const unsigned int ring_rptr_reg;
+	const unsigned int info_reg;
+};
+
+struct platform_access_vdata {
+	const unsigned int cmdresp_reg;
+	const unsigned int cmdbuff_addr_lo_reg;
+	const unsigned int cmdbuff_addr_hi_reg;
+	const unsigned int doorbell_button_reg;
+	const unsigned int doorbell_cmd_reg;
+
+};
+
+struct psp_vdata {
+	const struct sev_vdata *sev;
+	const struct tee_vdata *tee;
+	const struct platform_access_vdata *platform_access;
+	const unsigned int cmdresp_reg;
+	const unsigned int cmdbuff_addr_lo_reg;
+	const unsigned int cmdbuff_addr_hi_reg;
+	const unsigned int feature_reg;
+	const unsigned int inten_reg;
+	const unsigned int intsts_reg;
+	const unsigned int bootloader_info_reg;
+	const unsigned int platform_features;
+};
+
+/* Structure to hold SP device data */
+struct sp_dev_vdata {
+	const unsigned int bar;
+
+	const struct ccp_vdata *ccp_vdata;
+	const struct psp_vdata *psp_vdata;
+};
+
+struct sp_device {
+	struct list_head entry;
+
+	struct device *dev;
+
+	const struct sp_dev_vdata *dev_vdata;
+	unsigned int ord;
+	char name[SP_MAX_NAME_LEN];
+
+	/* Bus specific device information */
+	void *dev_specific;
+
+	/* I/O area used for device communication. */
+	void __iomem *io_map;
+
+	/* DMA caching attribute support */
+	unsigned int axcache;
+
+	/* get and set master device */
+	struct sp_device*(*get_psp_master_device)(void);
+	void (*set_psp_master_device)(struct sp_device *);
+	void (*clear_psp_master_device)(struct sp_device *);
+
+	bool irq_registered;
+	bool use_tasklet;
+
+	unsigned int ccp_irq;
+	irq_handler_t ccp_irq_handler;
+	void *ccp_irq_data;
+
+	unsigned int psp_irq;
+	irq_handler_t psp_irq_handler;
+	void *psp_irq_data;
+
+	void *ccp_data;
+	void *psp_data;
+};
+
+int sp_pci_init(void);
+void sp_pci_exit(void);
+
+int sp_platform_init(void);
+void sp_platform_exit(void);
+
+struct sp_device *sp_alloc_struct(struct device *dev);
+
+int sp_init(struct sp_device *sp);
+void sp_destroy(struct sp_device *sp);
+
+int sp_suspend(struct sp_device *sp);
+int sp_resume(struct sp_device *sp);
+int sp_restore(struct sp_device *sp);
+int sp_request_ccp_irq(struct sp_device *sp, irq_handler_t handler,
+		       const char *name, void *data);
+void sp_free_ccp_irq(struct sp_device *sp, void *data);
+int sp_request_psp_irq(struct sp_device *sp, irq_handler_t handler,
+		       const char *name, void *data);
+void sp_free_psp_irq(struct sp_device *sp, void *data);
+struct sp_device *sp_get_psp_master_device(void);
+
+#ifdef CONFIG_CRYPTO_DEV_SP_CCP
+
+int ccp_dev_init(struct sp_device *sp);
+void ccp_dev_destroy(struct sp_device *sp);
+
+void ccp_dev_suspend(struct sp_device *sp);
+void ccp_dev_resume(struct sp_device *sp);
+
+#else	/* !CONFIG_CRYPTO_DEV_SP_CCP */
+
+static inline int ccp_dev_init(struct sp_device *sp)
+{
+	return 0;
+}
+static inline void ccp_dev_destroy(struct sp_device *sp) { }
+static inline void ccp_dev_suspend(struct sp_device *sp) { }
+static inline void ccp_dev_resume(struct sp_device *sp) { }
+#endif	/* CONFIG_CRYPTO_DEV_SP_CCP */
+
+#ifdef CONFIG_CRYPTO_DEV_SP_PSP
+
+int psp_dev_init(struct sp_device *sp);
+void psp_pci_init(void);
+void psp_dev_destroy(struct sp_device *sp);
+void psp_pci_exit(void);
+int psp_restore(struct sp_device *sp);
+
+#else /* !CONFIG_CRYPTO_DEV_SP_PSP */
+
+static inline int psp_dev_init(struct sp_device *sp) { return 0; }
+static inline void psp_pci_init(void) { }
+static inline void psp_dev_destroy(struct sp_device *sp) { }
+static inline void psp_pci_exit(void) { }
+static inline int psp_restore(struct sp_device *sp) { return 0; }
+
+#endif /* CONFIG_CRYPTO_DEV_SP_PSP */
+
+#endif