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, 97 insertions, 0 deletions

diff --git a/drivers/firmware/qcom/Kconfig b/drivers/firmware/qcom/Kconfig
new file mode 100644
index 000000000..c7f8413ab
--- /dev/null
+++ b/drivers/firmware/qcom/Kconfig
@@ -0,0 +1,97 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# For a description of the syntax of this configuration file,
+# see Documentation/kbuild/kconfig-language.rst.
+#
+
+menu "Qualcomm firmware drivers"
+
+config QCOM_PAS
+	tristate "Qualcomm generic PAS interface driver"
+	help
+	  Enable the generic Peripheral Authentication Service (PAS) provided
+	  by the firmware. It acts as the common layer with different TZ
+	  backends plugged in whether it's an SCM implementation or a proper
+	  TEE bus based PAS service implementation.
+
+config QCOM_PAS_TEE
+	tristate "Qualcomm PAS TEE interface driver"
+	select QCOM_PAS
+	depends on TEE
+	depends on !CPU_BIG_ENDIAN
+	default m if ARCH_QCOM
+	help
+	  Enable the generic Peripheral Authentication Service (PAS) provided
+	  by the firmware TEE implementation as the backend.
+
+config QCOM_SCM
+	tristate "Qualcomm PAS SCM interface driver"
+	select QCOM_PAS
+	select QCOM_TZMEM
+	default y if ARCH_QCOM
+
+config QCOM_TZMEM
+	tristate
+	select GENERIC_ALLOCATOR
+
+choice
+	prompt "TrustZone interface memory allocator mode"
+	depends on QCOM_TZMEM
+	default QCOM_TZMEM_MODE_GENERIC
+	help
+	  Selects the mode of the memory allocator providing memory buffers of
+	  suitable format for sharing with the TrustZone. If in doubt, select
+	  'Generic'.
+
+config QCOM_TZMEM_MODE_GENERIC
+	bool "Generic"
+	help
+	  Use the generic allocator mode. The memory is page-aligned, non-cachable
+	  and physically contiguous.
+
+config QCOM_TZMEM_MODE_SHMBRIDGE
+	bool "SHM Bridge"
+	help
+	  Use Qualcomm Shared Memory Bridge. The memory has the same alignment as
+	  in the 'Generic' allocator but is also explicitly marked as an SHM Bridge
+	  buffer.
+
+	  With this selected, all buffers passed to the TrustZone must be allocated
+	  using the TZMem allocator or else the TrustZone will refuse to use them.
+
+endchoice
+
+config QCOM_QSEECOM
+	bool "Qualcomm QSEECOM interface driver"
+	depends on QCOM_SCM=y
+	select AUXILIARY_BUS
+	help
+	  Various Qualcomm SoCs have a Secure Execution Environment (SEE) running
+	  in the Trust Zone. This module provides an interface to that via the
+	  QSEECOM mechanism, using SCM calls.
+
+	  The QSEECOM interface allows, among other things, access to applications
+	  running in the SEE. An example of such an application is 'uefisecapp',
+	  which is required to access UEFI variables on certain systems. If
+	  selected, the interface will also attempt to detect and register client
+	  devices for supported applications.
+
+	  Select Y here to enable the QSEECOM interface driver.
+
+config QCOM_QSEECOM_UEFISECAPP
+	bool "Qualcomm SEE UEFI Secure App client driver"
+	depends on QCOM_QSEECOM
+	depends on EFI
+	help
+	  Various Qualcomm SoCs do not allow direct access to EFI variables.
+	  Instead, these need to be accessed via the UEFI Secure Application
+	  (uefisecapp), residing in the Secure Execution Environment (SEE).
+
+	  This module provides a client driver for uefisecapp, installing efivar
+	  operations to allow the kernel accessing EFI variables, and via that also
+	  provide user-space with access to EFI variables via efivarfs.
+
+	  Select Y here to provide access to EFI variables on the aforementioned
+	  platforms.
+
+endmenu