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, 57 insertions, 0 deletions

diff --git a/Documentation/networking/mpls-sysctl.rst b/Documentation/networking/mpls-sysctl.rst
new file mode 100644
index 000000000..0a2ac8840
--- /dev/null
+++ b/Documentation/networking/mpls-sysctl.rst
@@ -0,0 +1,57 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+====================
+MPLS Sysfs variables
+====================
+
+/proc/sys/net/mpls/* Variables:
+===============================
+
+platform_labels - INTEGER
+	Number of entries in the platform label table.  It is not
+	possible to configure forwarding for label values equal to or
+	greater than the number of platform labels.
+
+	A dense utilization of the entries in the platform label table
+	is possible and expected as the platform labels are locally
+	allocated.
+
+	If the number of platform label table entries is set to 0 no
+	label will be recognized by the kernel and mpls forwarding
+	will be disabled.
+
+	Reducing this value will remove all label routing entries that
+	no longer fit in the table.
+
+	Possible values: 0 - 1048575
+
+	Default: 0
+
+ip_ttl_propagate - BOOL
+	Control whether TTL is propagated from the IPv4/IPv6 header to
+	the MPLS header on imposing labels and propagated from the
+	MPLS header to the IPv4/IPv6 header on popping the last label.
+
+	If disabled, the MPLS transport network will appear as a
+	single hop to transit traffic.
+
+	* 0 - disabled / RFC 3443 [Short] Pipe Model
+	* 1 - enabled / RFC 3443 Uniform Model (default)
+
+default_ttl - INTEGER
+	Default TTL value to use for MPLS packets where it cannot be
+	propagated from an IP header, either because one isn't present
+	or ip_ttl_propagate has been disabled.
+
+	Possible values: 1 - 255
+
+	Default: 255
+
+conf/<interface>/input - BOOL
+	Control whether packets can be input on this interface.
+
+	If disabled, packets will be discarded without further
+	processing.
+
+	* 0 - disabled (default)
+	* not 0 - enabled