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, 121 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-class-net-queues b/Documentation/ABI/testing/sysfs-class-net-queues
new file mode 100644
index 000000000..84aa25e0d
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-class-net-queues
@@ -0,0 +1,121 @@
+What:		/sys/class/net/<iface>/queues/rx-<queue>/rps_cpus
+Date:		March 2010
+KernelVersion:	2.6.35
+Contact:	netdev@vger.kernel.org
+Description:
+		Mask of the CPU(s) currently enabled to participate into the
+		Receive Packet Steering packet processing flow for this
+		network device queue. Possible values depend on the number
+		of available CPU(s) in the system.
+
+What:		/sys/class/net/<iface>/queues/rx-<queue>/rps_flow_cnt
+Date:		April 2010
+KernelVersion:	2.6.35
+Contact:	netdev@vger.kernel.org
+Description:
+		Number of Receive Packet Steering flows being currently
+		processed by this particular network device receive queue.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/tx_timeout
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the number of transmit timeout events seen by this
+		network interface transmit queue.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/tx_maxrate
+Date:		March 2015
+KernelVersion:	4.1
+Contact:	netdev@vger.kernel.org
+Description:
+		A Mbps max-rate set for the queue, a value of zero means disabled,
+		default is disabled.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/xps_cpus
+Date:		November 2010
+KernelVersion:	2.6.38
+Contact:	netdev@vger.kernel.org
+Description:
+		Mask of the CPU(s) currently enabled to participate into the
+		Transmit Packet Steering packet processing flow for this
+		network device transmit queue. Possible values depend on the
+		number of available CPU(s) in the system.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/xps_rxqs
+Date:		June 2018
+KernelVersion:	4.18.0
+Contact:	netdev@vger.kernel.org
+Description:
+		Mask of the receive queue(s) currently enabled to participate
+		into the Transmit Packet Steering packet processing flow for this
+		network device transmit queue. Possible values depend on the
+		number of available receive queue(s) in the network device.
+		Default is disabled.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/hold_time
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the hold time in milliseconds to measure the slack
+		of this particular network device transmit queue.
+		Default value is 1000.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/inflight
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the number of bytes (objects) in flight on this
+		network device transmit queue.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/limit
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the current limit of bytes allowed to be queued
+		on this network device transmit queue. This value is clamped
+		to be within the bounds defined by limit_max and limit_min.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/limit_max
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the absolute maximum limit of bytes allowed to be
+		queued on this network device transmit queue. See
+		include/linux/dynamic_queue_limits.h for the default value.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/limit_min
+Date:		November 2011
+KernelVersion:	3.3
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the absolute minimum limit of bytes allowed to be
+		queued on this network device transmit queue. Default value is
+		0.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/stall_thrs
+Date:		Jan 2024
+KernelVersion:	6.9
+Contact:	netdev@vger.kernel.org
+Description:
+		Tx completion stall detection threshold in ms. Kernel will
+		guarantee to detect all stalls longer than this threshold but
+		may also detect stalls longer than half of the threshold.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/stall_cnt
+Date:		Jan 2024
+KernelVersion:	6.9
+Contact:	netdev@vger.kernel.org
+Description:
+		Number of detected Tx completion stalls.
+
+What:		/sys/class/net/<iface>/queues/tx-<queue>/byte_queue_limits/stall_max
+Date:		Jan 2024
+KernelVersion:	6.9
+Contact:	netdev@vger.kernel.org
+Description:
+		Longest detected Tx completion stall. Write 0 to clear.