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, 119 insertions, 0 deletions

diff --git a/drivers/usb/class/cdc-acm.h b/drivers/usb/class/cdc-acm.h
new file mode 100644
index 000000000..01f448a78
--- /dev/null
+++ b/drivers/usb/class/cdc-acm.h
@@ -0,0 +1,119 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ *
+ * Includes for cdc-acm.c
+ *
+ * Mainly take from usbnet's cdc-ether part
+ *
+ */
+
+/*
+ * Major and minor numbers.
+ */
+
+#define ACM_TTY_MAJOR		166
+#define ACM_TTY_MINORS		256
+
+#define ACM_MINOR_INVALID	ACM_TTY_MINORS
+
+/*
+ * Requests.
+ */
+
+#define USB_RT_ACM		(USB_TYPE_CLASS | USB_RECIP_INTERFACE)
+
+/*
+ * Internal driver structures.
+ */
+
+/*
+ * The only reason to have several buffers is to accommodate assumptions
+ * in line disciplines. They ask for empty space amount, receive our URB size,
+ * and proceed to issue several 1-character writes, assuming they will fit.
+ * The very first write takes a complete URB. Fortunately, this only happens
+ * when processing onlcr, so we only need 2 buffers. These values must be
+ * powers of 2.
+ */
+#define ACM_NW  16
+#define ACM_NR  16
+
+struct acm_wb {
+	u8 *buf;
+	dma_addr_t dmah;
+	unsigned int len;
+	struct urb		*urb;
+	struct acm		*instance;
+	bool use;
+};
+
+struct acm_rb {
+	int			size;
+	unsigned char		*base;
+	dma_addr_t		dma;
+	int			index;
+	struct acm		*instance;
+};
+
+struct acm {
+	struct usb_device *dev;				/* the corresponding usb device */
+	struct usb_interface *control;			/* control interface */
+	struct usb_interface *data;			/* data interface */
+	unsigned in, out;				/* i/o pipes */
+	struct tty_port port;			 	/* our tty port data */
+	struct urb *ctrlurb;				/* urbs */
+	u8 *ctrl_buffer;				/* buffers of urbs */
+	dma_addr_t ctrl_dma;				/* dma handles of buffers */
+	u8 *country_codes;				/* country codes from device */
+	unsigned int country_code_size;			/* size of this buffer */
+	unsigned int country_rel_date;			/* release date of version */
+	struct acm_wb wb[ACM_NW];
+	unsigned long read_urbs_free;
+	struct urb *read_urbs[ACM_NR];
+	struct acm_rb read_buffers[ACM_NR];
+	int rx_buflimit;
+	spinlock_t read_lock;
+	u8 *notification_buffer;			/* to reassemble fragmented notifications */
+	unsigned int nb_index;
+	unsigned int nb_size;
+	int transmitting;
+	spinlock_t write_lock;
+	struct mutex mutex;
+	bool disconnected;
+	unsigned long flags;
+#		define EVENT_TTY_WAKEUP	0
+#		define EVENT_RX_STALL	1
+#		define ACM_THROTTLED	2
+#		define ACM_ERROR_DELAY	3
+	unsigned long urbs_in_error_delay;		/* these need to be restarted after a delay */
+	struct usb_cdc_line_coding line;		/* bits, stop, parity */
+	struct delayed_work dwork;		        /* work queue entry for various purposes */
+	unsigned int ctrlin;				/* input control lines (DCD, DSR, RI, break, overruns) */
+	unsigned int ctrlout;				/* output control lines (DTR, RTS) */
+	struct async_icount iocount;			/* counters for control line changes */
+	struct async_icount oldcount;			/* for comparison of counter */
+	wait_queue_head_t wioctl;			/* for ioctl */
+	unsigned int writesize;				/* max packet size for the output bulk endpoint */
+	unsigned int readsize,ctrlsize;			/* buffer sizes for freeing */
+	unsigned int minor;				/* acm minor number */
+	unsigned char clocal;				/* termios CLOCAL */
+	unsigned int ctrl_caps;				/* control capabilities from the class specific header */
+	unsigned int susp_count;			/* number of suspended interfaces */
+	unsigned int combined_interfaces:1;		/* control and data collapsed */
+	u8 bInterval;
+	struct usb_anchor delayed;			/* writes queued for a device about to be woken */
+	unsigned long quirks;
+};
+
+/* constants describing various quirks and errors */
+#define NO_UNION_NORMAL			BIT(0)
+#define SINGLE_RX_URB			BIT(1)
+#define NO_CAP_LINE			BIT(2)
+#define IGNORE_DEVICE			BIT(3)
+#define QUIRK_CONTROL_LINE_STATE	BIT(4)
+#define CLEAR_HALT_CONDITIONS		BIT(5)
+#define SEND_ZERO_PACKET		BIT(6)
+#define DISABLE_ECHO			BIT(7)
+#define MISSING_CAP_BRK			BIT(8)
+#define NO_UNION_12			BIT(9)
+#define VENDOR_CLASS_DATA_IFACE		BIT(10)  /* data interface uses vendor-specific class */
+#define ALWAYS_POLL_CTRL		BIT(11) /* keep ctrl URB active even without an open TTY */