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, 103 insertions, 0 deletions

diff --git a/Documentation/PCI/endpoint/pci-test-function.rst b/Documentation/PCI/endpoint/pci-test-function.rst
new file mode 100644
index 000000000..3c8521d7a
--- /dev/null
+++ b/Documentation/PCI/endpoint/pci-test-function.rst
@@ -0,0 +1,103 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+=================
+PCI Test Function
+=================
+
+:Author: Kishon Vijay Abraham I <kishon@ti.com>
+
+Traditionally PCI RC has always been validated by using standard
+PCI cards like ethernet PCI cards or USB PCI cards or SATA PCI cards.
+However with the addition of EP-core in linux kernel, it is possible
+to configure a PCI controller that can operate in EP mode to work as
+a test device.
+
+The PCI endpoint test device is a virtual device (defined in software)
+used to test the endpoint functionality and serve as a sample driver
+for other PCI endpoint devices (to use the EP framework).
+
+The PCI endpoint test device has the following registers:
+
+	1) PCI_ENDPOINT_TEST_MAGIC
+	2) PCI_ENDPOINT_TEST_COMMAND
+	3) PCI_ENDPOINT_TEST_STATUS
+	4) PCI_ENDPOINT_TEST_SRC_ADDR
+	5) PCI_ENDPOINT_TEST_DST_ADDR
+	6) PCI_ENDPOINT_TEST_SIZE
+	7) PCI_ENDPOINT_TEST_CHECKSUM
+	8) PCI_ENDPOINT_TEST_IRQ_TYPE
+	9) PCI_ENDPOINT_TEST_IRQ_NUMBER
+
+* PCI_ENDPOINT_TEST_MAGIC
+
+This register will be used to test BAR0. A known pattern will be written
+and read back from MAGIC register to verify BAR0.
+
+* PCI_ENDPOINT_TEST_COMMAND
+
+This register will be used by the host driver to indicate the function
+that the endpoint device must perform.
+
+========	================================================================
+Bitfield	Description
+========	================================================================
+Bit 0		raise legacy IRQ
+Bit 1		raise MSI IRQ
+Bit 2		raise MSI-X IRQ
+Bit 3		read command (read data from RC buffer)
+Bit 4		write command (write data to RC buffer)
+Bit 5		copy command (copy data from one RC buffer to another RC buffer)
+========	================================================================
+
+* PCI_ENDPOINT_TEST_STATUS
+
+This register reflects the status of the PCI endpoint device.
+
+========	==============================
+Bitfield	Description
+========	==============================
+Bit 0		read success
+Bit 1		read fail
+Bit 2		write success
+Bit 3		write fail
+Bit 4		copy success
+Bit 5		copy fail
+Bit 6		IRQ raised
+Bit 7		source address is invalid
+Bit 8		destination address is invalid
+========	==============================
+
+* PCI_ENDPOINT_TEST_SRC_ADDR
+
+This register contains the source address (RC buffer address) for the
+COPY/READ command.
+
+* PCI_ENDPOINT_TEST_DST_ADDR
+
+This register contains the destination address (RC buffer address) for
+the COPY/WRITE command.
+
+* PCI_ENDPOINT_TEST_IRQ_TYPE
+
+This register contains the interrupt type (Legacy/MSI) triggered
+for the READ/WRITE/COPY and raise IRQ (Legacy/MSI) commands.
+
+Possible types:
+
+======	==
+Legacy	0
+MSI	1
+MSI-X	2
+======	==
+
+* PCI_ENDPOINT_TEST_IRQ_NUMBER
+
+This register contains the triggered ID interrupt.
+
+Admissible values:
+
+======	===========
+Legacy	0
+MSI	[1 .. 32]
+MSI-X	[1 .. 2048]
+======	===========