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, 131 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/debugfs-amd-iommu b/Documentation/ABI/testing/debugfs-amd-iommu
new file mode 100644
index 000000000..5621a66aa
--- /dev/null
+++ b/Documentation/ABI/testing/debugfs-amd-iommu
@@ -0,0 +1,131 @@
+What:		/sys/kernel/debug/iommu/amd/iommu<x>/mmio
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file provides read/write access for user input. Users specify the
+		MMIO register offset for iommu<x>, and the file outputs the corresponding
+		MMIO register value of iommu<x>
+
+		Example::
+
+		  $ echo "0x18" > /sys/kernel/debug/iommu/amd/iommu00/mmio
+		  $ cat /sys/kernel/debug/iommu/amd/iommu00/mmio
+
+		Output::
+
+		  Offset:0x18 Value:0x000c22000003f48d
+
+What:		/sys/kernel/debug/iommu/amd/iommu<x>/capability
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file provides read/write access for user input. Users specify the
+		capability register offset for iommu<x>, and the file outputs the
+		corresponding capability register value of iommu<x>.
+
+		Example::
+
+		  $ echo "0x10" > /sys/kernel/debug/iommu/amd/iommu00/capability
+		  $ cat /sys/kernel/debug/iommu/amd/iommu00/capability
+
+		Output::
+
+		  Offset:0x10 Value:0x00203040
+
+What:		/sys/kernel/debug/iommu/amd/iommu<x>/cmdbuf
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file is a read-only output file containing iommu<x> command
+		buffer entries.
+
+		Examples::
+
+		  $ cat /sys/kernel/debug/iommu/amd/iommu<x>/cmdbuf
+
+		Output::
+
+		  CMD Buffer Head Offset:339 Tail Offset:339
+		    0: 00835001 10000001 00003c00 00000000
+		    1: 00000000 30000005 fffff003 7fffffff
+		    2: 00835001 10000001 00003c01 00000000
+		    3: 00000000 30000005 fffff003 7fffffff
+		    4: 00835001 10000001 00003c02 00000000
+		    5: 00000000 30000005 fffff003 7fffffff
+		    6: 00835001 10000001 00003c03 00000000
+		    7: 00000000 30000005 fffff003 7fffffff
+		    8: 00835001 10000001 00003c04 00000000
+		    9: 00000000 30000005 fffff003 7fffffff
+		   10: 00835001 10000001 00003c05 00000000
+		   11: 00000000 30000005 fffff003 7fffffff
+		  [...]
+
+What:		/sys/kernel/debug/iommu/amd/devid
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file provides read/write access for user input. Users specify the
+		device ID, which can be used to dump IOMMU data structures such as the
+		interrupt remapping table and device table.
+
+		Example:
+
+		1.
+		  ::
+
+		    $ echo 0000:01:00.0 > /sys/kernel/debug/iommu/amd/devid
+		    $ cat /sys/kernel/debug/iommu/amd/devid
+
+		  Output::
+
+		    0000:01:00.0
+
+		2.
+		  ::
+
+		    $ echo 01:00.0 > /sys/kernel/debug/iommu/amd/devid
+		    $ cat /sys/kernel/debug/iommu/amd/devid
+
+		  Output::
+
+		    0000:01:00.0
+
+What:		/sys/kernel/debug/iommu/amd/devtbl
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file is a read-only output file containing the device table entry
+		for the device ID provided in /sys/kernel/debug/iommu/amd/devid.
+
+		Example::
+
+		  $ cat /sys/kernel/debug/iommu/amd/devtbl
+
+		Output::
+
+		  DeviceId             QWORD[3]         QWORD[2]         QWORD[1]         QWORD[0] iommu
+		  0000:01:00.0 0000000000000000 20000001373b8013 0000000000000038 6000000114d7b603 iommu3
+
+What:		/sys/kernel/debug/iommu/amd/irqtbl
+Date:		January 2025
+Contact:	Dheeraj Kumar Srivastava <dheerajkumar.srivastava@amd.com>
+Description:
+		This file is a read-only output file containing valid IRT table entries
+		for the device ID provided in /sys/kernel/debug/iommu/amd/devid.
+
+		Example::
+
+		  $ cat /sys/kernel/debug/iommu/amd/irqtbl
+
+		Output::
+
+		  DeviceId 0000:01:00.0
+		  IRT[0000] 0000000000000020 0000000000000241
+		  IRT[0001] 0000000000000020 0000000000000841
+		  IRT[0002] 0000000000000020 0000000000002041
+		  IRT[0003] 0000000000000020 0000000000008041
+		  IRT[0004] 0000000000000020 0000000000020041
+		  IRT[0005] 0000000000000020 0000000000080041
+		  IRT[0006] 0000000000000020 0000000000200041
+		  IRT[0007] 0000000000000020 0000000000800041
+		  [...]