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, 85 insertions, 0 deletions

diff --git a/drivers/firmware/google/Kconfig b/drivers/firmware/google/Kconfig
new file mode 100644
index 000000000..b78c644fa
--- /dev/null
+++ b/drivers/firmware/google/Kconfig
@@ -0,0 +1,85 @@
+# SPDX-License-Identifier: GPL-2.0-only
+menuconfig GOOGLE_FIRMWARE
+	bool "Google Firmware Drivers"
+	default n
+	help
+	  These firmware drivers are used by Google servers,
+	  Chromebooks and other devices using coreboot firmware.
+	  If in doubt, say "N".
+
+if GOOGLE_FIRMWARE
+
+config GOOGLE_SMI
+	tristate "SMI interface for Google platforms"
+	depends on X86 && ACPI && DMI
+	help
+	  Say Y here if you want to enable SMI callbacks for Google
+	  platforms.  This provides an interface for writing to and
+	  clearing the event log.  If CONFIG_EFI is also enabled this
+	  driver provides an interface for reading and writing NVRAM
+	  variables.
+
+config GOOGLE_CBMEM
+	tristate "CBMEM entries in sysfs"
+	depends on GOOGLE_COREBOOT_TABLE
+	help
+	  CBMEM is a downwards-growing memory region created by the
+	  Coreboot BIOS containing tagged data structures from the
+	  BIOS.  These data structures expose things like the verified
+	  boot firmware variables, flash layout, firmware event log,
+	  and more.
+
+	  This option enables the cbmem module, which causes the
+	  kernel to search for Coreboot CBMEM entries, and expose the
+	  memory for each entry in sysfs under
+	  /sys/bus/coreboot/devices/cbmem-<id>.
+
+config GOOGLE_COREBOOT_TABLE
+	tristate "Coreboot Table Access"
+	depends on HAS_IOMEM && (ACPI || OF)
+	help
+	  This option enables the coreboot_table module, which provides other
+	  firmware modules access to the coreboot table. The coreboot table
+	  pointer is accessed through the ACPI "GOOGCB00" object or the
+	  device tree node /firmware/coreboot.
+	  If unsure say N.
+
+config GOOGLE_MEMCONSOLE
+	tristate
+	depends on GOOGLE_MEMCONSOLE_X86_LEGACY || GOOGLE_MEMCONSOLE_COREBOOT
+
+config GOOGLE_MEMCONSOLE_X86_LEGACY
+	tristate "Firmware Memory Console - X86 Legacy support"
+	depends on X86 && ACPI && DMI
+	select GOOGLE_MEMCONSOLE
+	help
+	  This option enables the kernel to search for a firmware log in
+	  the EBDA on Google servers.  If found, this log is exported to
+	  userland in the file /sys/firmware/log.
+
+config GOOGLE_FRAMEBUFFER_COREBOOT
+	tristate "Coreboot Framebuffer"
+	depends on GOOGLE_COREBOOT_TABLE
+	help
+	  This option enables the kernel to search for a framebuffer in
+	  the coreboot table.  If found, it is registered with a platform
+	  device of type coreboot-framebuffer. Using the old device of
+	  type simple-framebuffer is deprecated.
+
+config GOOGLE_MEMCONSOLE_COREBOOT
+	tristate "Firmware Memory Console"
+	depends on GOOGLE_COREBOOT_TABLE
+	select GOOGLE_MEMCONSOLE
+	help
+	  This option enables the kernel to search for a firmware log in
+	  the coreboot table.  If found, this log is exported to userland
+	  in the file /sys/firmware/log.
+
+config GOOGLE_VPD
+	tristate "Vital Product Data"
+	depends on GOOGLE_COREBOOT_TABLE
+	help
+	  This option enables the kernel to expose the content of Google VPD
+	  under /sys/firmware/vpd.
+
+endif # GOOGLE_FIRMWARE