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, 354 insertions, 0 deletions

diff --git a/Documentation/ABI/testing/sysfs-class-net b/Documentation/ABI/testing/sysfs-class-net
new file mode 100644
index 000000000..ebf21beba
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-class-net
@@ -0,0 +1,354 @@
+What:		/sys/class/net/<iface>/name_assign_type
+Date:		July 2014
+KernelVersion:	3.17
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the name assignment type. Possible values are:
+
+		== ==========================================================
+		1  enumerated by the kernel, possibly in an unpredictable way
+		2  predictably named by the kernel
+		3  named by userspace
+		4  renamed
+		== ==========================================================
+
+What:		/sys/class/net/<iface>/addr_assign_type
+Date:		July 2010
+KernelVersion:	3.2
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the address assignment type. Possible values are:
+
+		== =============================
+		0  permanent address
+		1  randomly generated
+		2  stolen from another device
+		3  set using dev_set_mac_address
+		== =============================
+
+What:		/sys/class/net/<iface>/addr_len
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the hardware address size in bytes.
+		Values vary based on the lower-level protocol used by the
+		interface (Ethernet, FDDI, ATM, IEEE 802.15.4...). See
+		include/uapi/linux/if_*.h for actual values.
+
+What:		/sys/class/net/<iface>/address
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Hardware address currently assigned to this interface.
+		Format is a string, e.g: 00:11:22:33:44:55 for an Ethernet MAC
+		address.
+
+What:		/sys/class/net/<bridge iface>/bridge/group_fwd_mask
+Date:		January 2012
+KernelVersion:	3.2
+Contact:	netdev@vger.kernel.org
+Description:
+		Bitmask to allow forwarding of link local frames with address
+		01-80-C2-00-00-0X on a bridge device. Only values that set bits
+		not matching BR_GROUPFWD_RESTRICTED in net/bridge/br_private.h
+		allowed.
+		Default value 0 does not forward any link local frames.
+
+		Restricted bits:
+
+		== ========================================================
+		0  01-80-C2-00-00-00 Bridge Group Address used for STP
+		1  01-80-C2-00-00-01 (MAC Control) 802.3 used for MAC PAUSE
+		2  01-80-C2-00-00-02 (Link Aggregation) 802.3ad
+		== ========================================================
+
+		Any values not setting these bits can be used. Take special
+		care when forwarding control frames e.g. 802.1X-PAE or LLDP.
+
+What:		/sys/class/net/<iface>/broadcast
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Hardware broadcast address for this interface. Format is a
+		string, e.g: ff:ff:ff:ff:ff:ff for an Ethernet broadcast MAC
+		address.
+
+What:		/sys/class/net/<iface>/carrier
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the current physical link state of the interface.
+		Possible values are:
+
+		== =====================
+		0  physical link is down
+		1  physical link is up
+		== =====================
+
+		Note: some special devices, e.g: bonding and team drivers will
+		allow this attribute to be written to force a link state for
+		operating correctly and designating another fallback interface.
+
+What:		/sys/class/net/<iface>/dev_id
+Date:		April 2008
+KernelVersion:	2.6.26
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the device unique identifier. Format is an hexadecimal
+		value. This is used to disambiguate interfaces which might be
+		stacked (e.g: VLAN interfaces) but still have the same MAC
+		address as their parent device.
+
+What:		/sys/class/net/<iface>/dev_port
+Date:		February 2014
+KernelVersion:	3.15
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the port number of this network device, formatted
+		as a decimal value. Some NICs have multiple independent ports
+		on the same PCI bus, device and function. This attribute allows
+		userspace to distinguish the respective interfaces.
+
+		Note: some device drivers started to use 'dev_id' for this
+		purpose since long before 3.15 and have not adopted the new
+		attribute ever since. To query the port number, some tools look
+		exclusively at 'dev_port', while others only consult 'dev_id'.
+		If a network device has multiple client adapter ports as
+		described in the previous paragraph and does not set this
+		attribute to its port number, it's a kernel bug.
+
+What:		/sys/class/net/<iface>/dormant
+Date:		March 2006
+KernelVersion:	2.6.17
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates whether the interface is in dormant state. Possible
+		values are:
+		0: interface is not dormant
+		1: interface is dormant
+
+		This attribute can be used by supplicant software to signal that
+		the device is not usable unless some supplicant-based
+		authentication is performed (e.g: 802.1x). 'link_mode' attribute
+		will also reflect the dormant state.
+
+What:		/sys/class/net/<iface>/testing
+Date:		April 2002
+KernelVersion:	5.8
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates whether the interface is under test. Possible
+		values are:
+
+		== =============================
+		0  interface is not being tested
+		1  interface is being tested
+		== =============================
+
+		When an interface is under test, it cannot be expected
+		to pass packets as normal.
+
+What:		/sys/class/net/<iface>/duplex
+Date:		October 2009
+KernelVersion:	2.6.33
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface latest or current duplex value. Possible
+		values are:
+
+		====  ===========
+		half  half duplex
+		full  full duplex
+		====  ===========
+
+		Note: This attribute is only valid for interfaces that implement
+		the ethtool get_link_ksettings method (mostly Ethernet).
+
+What:		/sys/class/net/<iface>/flags
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface flags as a bitmask in hexadecimal. See
+		include/uapi/linux/if.h for a list of all possible values and
+		the flags semantics.
+
+What:		/sys/class/net/<iface>/ifalias
+Date:		September 2008
+KernelVersion:	2.6.28
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates/stores an interface alias name as a string. This can
+		be used for system management purposes.
+
+What:		/sys/class/net/<iface>/ifindex
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the system-wide interface unique index identifier as a
+		decimal number. This attribute is used for mapping an interface
+		identifier to an interface name. It is used throughout the
+		networking stack for specifying the interface specific
+		requests/events.
+
+What:		/sys/class/net/<iface>/iflink
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the system-wide interface unique index identifier a
+		the interface is linked to. Format is decimal. This attribute is
+		used to resolve interfaces chaining, linking and stacking.
+		Physical interfaces have the same 'ifindex' and 'iflink' values.
+
+What:		/sys/class/net/<iface>/link_mode
+Date:		March 2006
+KernelVersion:	2.6.17
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface link mode, as a decimal number. This
+		attribute should be used in conjunction with 'dormant' attribute
+		to determine the interface usability. Possible values:
+
+		==  =================
+		0   default link mode
+		1   dormant link mode
+		==  =================
+
+What:		/sys/class/net/<iface>/mtu
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface currently configured MTU value, in
+		bytes, and in decimal format. Specific values depends on the
+		lower-level interface protocol used. Ethernet devices will show
+		a 'mtu' attribute value of 1500 unless changed.
+
+What:		/sys/class/net/<iface>/netdev_group
+Date:		January 2011
+KernelVersion:	2.6.39
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface network device group, as a decimal
+		integer. Default value is 0 which corresponds to the initial
+		network devices group. The group can be changed to affect
+		routing decisions (see: net/ipv4/fib_rules and
+		net/ipv6/fib6_rules.c).
+
+What:		/sys/class/net/<iface>/operstate
+Date:		March 2006
+KernelVersion:	2.6.17
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface RFC2863 operational state as a string.
+
+		Possible values are:
+
+		"unknown", "notpresent", "down", "lowerlayerdown", "testing",
+		"dormant", "up".
+
+What:		/sys/class/net/<iface>/phys_port_id
+Date:		July 2013
+KernelVersion:	3.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface unique physical port identifier within
+		the NIC, as a string.
+
+What:		/sys/class/net/<iface>/phys_port_name
+Date:		March 2015
+KernelVersion:	4.0
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface physical port name within the NIC,
+		as a string.
+
+What:		/sys/class/net/<iface>/speed
+Date:		October 2009
+KernelVersion:	2.6.33
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface latest or current speed value. Value is
+		an integer representing the link speed in Mbits/sec.
+
+		Note: this attribute is only valid for interfaces that implement
+		the ethtool get_link_ksettings method (mostly Ethernet).
+
+What:		/sys/class/net/<iface>/tx_queue_len
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface transmit queue len in number of packets,
+		as an integer value. Value depend on the type of interface,
+		Ethernet network adapters have a default value of 1000 unless
+		configured otherwise
+
+What:		/sys/class/net/<iface>/type
+Date:		April 2005
+KernelVersion:	2.6.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the interface protocol type as a decimal value. See
+		include/uapi/linux/if_arp.h for all possible values.
+
+What:		/sys/class/net/<iface>/phys_switch_id
+Date:		November 2014
+KernelVersion:	3.19
+Contact:	netdev@vger.kernel.org
+Description:
+		Indicates the unique physical switch identifier of a switch this
+		port belongs to, as a string.
+
+What:		/sys/class/net/<iface>/phydev
+Date:		May 2017
+KernelVersion:	4.13
+Contact:	netdev@vger.kernel.org
+Description:
+		Symbolic link to the PHY device this network device is attached
+		to.
+
+What:		/sys/class/net/<iface>/carrier_changes
+Date:		Mar 2014
+KernelVersion:	3.15
+Contact:	netdev@vger.kernel.org
+Description:
+		32-bit unsigned integer counting the number of times the link has
+		seen a change from UP to DOWN and vice versa
+
+What:		/sys/class/net/<iface>/carrier_up_count
+Date:		Jan 2018
+KernelVersion:	4.16
+Contact:	netdev@vger.kernel.org
+Description:
+		32-bit unsigned integer counting the number of times the link has
+		been up
+
+What:		/sys/class/net/<iface>/carrier_down_count
+Date:		Jan 2018
+KernelVersion:	4.16
+Contact:	netdev@vger.kernel.org
+Description:
+		32-bit unsigned integer counting the number of times the link has
+		been down
+
+What:		/sys/class/net/<iface>/threaded
+Date:		Jan 2021
+KernelVersion:	5.12
+Contact:	netdev@vger.kernel.org
+Description:
+		Boolean value to control the threaded mode per device. User could
+		set this value to enable/disable threaded mode for all napi
+		belonging to this device, without the need to do device up/down.
+
+		Possible values:
+		== ==================================
+		0  threaded mode disabled for this dev
+		1  threaded mode enabled for this dev
+		== ==================================