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, 104 insertions, 0 deletions

diff --git a/Documentation/driver-api/zorro.rst b/Documentation/driver-api/zorro.rst
new file mode 100644
index 000000000..664072b01
--- /dev/null
+++ b/Documentation/driver-api/zorro.rst
@@ -0,0 +1,104 @@
+========================================
+Writing Device Drivers for Zorro Devices
+========================================
+
+:Author: Written by Geert Uytterhoeven <geert@linux-m68k.org>
+:Last revised: September 5, 2003
+
+
+Introduction
+------------
+
+The Zorro bus is the bus used in the Amiga family of computers. Thanks to
+AutoConfig(tm), it's 100% Plug-and-Play.
+
+There are two types of Zorro buses, Zorro II and Zorro III:
+
+  - The Zorro II address space is 24-bit and lies within the first 16 MB of the
+    Amiga's address map.
+
+  - Zorro III is a 32-bit extension of Zorro II, which is backwards compatible
+    with Zorro II. The Zorro III address space lies outside the first 16 MB.
+
+
+Probing for Zorro Devices
+-------------------------
+
+Zorro devices are found by calling ``zorro_find_device()``, which returns a
+pointer to the ``next`` Zorro device with the specified Zorro ID. A probe loop
+for the board with Zorro ID ``ZORRO_PROD_xxx`` looks like::
+
+    struct zorro_dev *z = NULL;
+
+    while ((z = zorro_find_device(ZORRO_PROD_xxx, z))) {
+	if (!zorro_request_region(z->resource.start+MY_START, MY_SIZE,
+				  "My explanation"))
+	...
+    }
+
+``ZORRO_WILDCARD`` acts as a wildcard and finds any Zorro device. If your driver
+supports different types of boards, you can use a construct like::
+
+    struct zorro_dev *z = NULL;
+
+    while ((z = zorro_find_device(ZORRO_WILDCARD, z))) {
+	if (z->id != ZORRO_PROD_xxx1 && z->id != ZORRO_PROD_xxx2 && ...)
+	    continue;
+	if (!zorro_request_region(z->resource.start+MY_START, MY_SIZE,
+				  "My explanation"))
+	...
+    }
+
+
+Zorro Resources
+---------------
+
+Before you can access a Zorro device's registers, you have to make sure it's
+not yet in use. This is done using the I/O memory space resource management
+functions::
+
+    request_mem_region()
+    release_mem_region()
+
+Shortcuts to claim the whole device's address space are provided as well::
+
+    zorro_request_device
+    zorro_release_device
+
+
+Accessing the Zorro Address Space
+---------------------------------
+
+The address regions in the Zorro device resources are Zorro bus address
+regions. Due to the identity bus-physical address mapping on the Zorro bus,
+they are CPU physical addresses as well.
+
+The treatment of these regions depends on the type of Zorro space:
+
+  - Zorro II address space is always mapped and does not have to be mapped
+    explicitly using z_ioremap().
+    
+    Conversion from bus/physical Zorro II addresses to kernel virtual addresses
+    and vice versa is done using::
+
+	virt_addr = ZTWO_VADDR(bus_addr);
+	bus_addr = ZTWO_PADDR(virt_addr);
+
+  - Zorro III address space must be mapped explicitly using z_ioremap() first
+    before it can be accessed::
+ 
+	virt_addr = z_ioremap(bus_addr, size);
+	...
+	z_iounmap(virt_addr);
+
+
+References
+----------
+
+#. linux/include/linux/zorro.h
+#. linux/include/uapi/linux/zorro.h
+#. linux/include/uapi/linux/zorro_ids.h
+#. linux/arch/m68k/include/asm/zorro.h
+#. linux/drivers/zorro
+#. /proc/bus/zorro
+