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, 31 insertions, 0 deletions

diff --git a/Documentation/admin-guide/LSM/LoadPin.rst b/Documentation/admin-guide/LSM/LoadPin.rst
new file mode 100644
index 000000000..dd3ca68b5
--- /dev/null
+++ b/Documentation/admin-guide/LSM/LoadPin.rst
@@ -0,0 +1,31 @@
+=======
+LoadPin
+=======
+
+LoadPin is a Linux Security Module that ensures all kernel-loaded files
+(modules, firmware, etc) all originate from the same filesystem, with
+the expectation that such a filesystem is backed by a read-only device
+such as dm-verity or CDROM. This allows systems that have a verified
+and/or unchangeable filesystem to enforce module and firmware loading
+restrictions without needing to sign the files individually.
+
+The LSM is selectable at build-time with ``CONFIG_SECURITY_LOADPIN``, and
+can be controlled at boot-time with the kernel command line option
+"``loadpin.enforce``". By default, it is enabled, but can be disabled at
+boot ("``loadpin.enforce=0``").
+
+LoadPin starts pinning when it sees the first file loaded. If the
+block device backing the filesystem is not read-only, a sysctl is
+created to toggle pinning: ``/proc/sys/kernel/loadpin/enabled``. (Having
+a mutable filesystem means pinning is mutable too, but having the
+sysctl allows for easy testing on systems with a mutable filesystem.)
+
+It's also possible to exclude specific file types from LoadPin using kernel
+command line option "``loadpin.exclude``". By default, all files are
+included, but they can be excluded using kernel command line option such
+as "``loadpin.exclude=kernel-module,kexec-image``". This allows to use
+different mechanisms such as ``CONFIG_MODULE_SIG`` and
+``CONFIG_KEXEC_VERIFY_SIG`` to verify kernel module and kernel image while
+still use LoadPin to protect the integrity of other files kernel loads. The
+full list of valid file types can be found in ``kernel_read_file_str``
+defined in ``include/linux/kernel_read_file.h``.