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, 65 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-secvar b/Documentation/ABI/testing/sysfs-secvar
new file mode 100644
index 000000000..c52a5fd15
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-secvar
@@ -0,0 +1,65 @@
+What:		/sys/firmware/secvar
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	This directory is created if the POWER firmware supports OS
+		secureboot, thereby secure variables. It exposes interface
+		for reading/writing the secure variables
+
+What:		/sys/firmware/secvar/vars
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	This directory lists all the secure variables that are supported
+		by the firmware.
+
+What:		/sys/firmware/secvar/format
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	A string indicating which backend is in use by the firmware.
+		This determines the format of the variable and the accepted
+		format of variable updates.
+
+		On powernv/OPAL, this value is provided by the OPAL firmware
+		and is expected to be "ibm,edk2-compat-v1".
+
+		On pseries/PLPKS, this is generated by the kernel based on the
+		version number in the SB_VERSION variable in the keystore. The
+		version numbering in the SB_VERSION variable starts from 1. The
+		format string takes the form "ibm,plpks-sb-v<version>" in the
+		case of dynamic key management mode. If the SB_VERSION variable
+		does not exist (or there is an error while reading it), it takes
+		the form "ibm,plpks-sb-v0", indicating that the key management
+		mode is static.
+
+What:		/sys/firmware/secvar/vars/<variable name>
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	Each secure variable is represented as a directory named as
+		<variable_name>. The variable name is unique and is in ASCII
+		representation. The data and size can be determined by reading
+		their respective attribute files.
+
+		Only secvars relevant to the key management mode are exposed.
+		Only in the dynamic key management mode should the user have
+		access (read and write) to the secure boot secvars db, dbx,
+		grubdb, grubdbx, and sbat. These secvars are not consumed in the
+		static key management mode. PK, trustedcadb and moduledb are the
+		secvars common to both static and dynamic key management modes.
+
+What:		/sys/firmware/secvar/vars/<variable_name>/size
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	An integer representation of the size of the content of the
+		variable. In other words, it represents the size of the data.
+
+What:		/sys/firmware/secvar/vars/<variable_name>/data
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	A read-only file containing the value of the variable. The size
+		of the file represents the maximum size of the variable data.
+
+What:		/sys/firmware/secvar/vars/<variable_name>/update
+Date:		August 2019
+Contact:	Nayna Jain <nayna@linux.ibm.com>
+Description:	A write-only file that is used to submit the new value for the
+		variable. The size of the file represents the maximum size of
+		the variable data that can be written.