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

5 files changed, 193 insertions, 0 deletions

diff --git a/scripts/kconfig/tests/transitional/Kconfig b/scripts/kconfig/tests/transitional/Kconfig
new file mode 100644
index 000000000..faa4d396f
--- /dev/null
+++ b/scripts/kconfig/tests/transitional/Kconfig
@@ -0,0 +1,132 @@
+# SPDX-License-Identifier: GPL-2.0
+# Test transitional symbols for config migration with all Kconfig types
+
+# Enable module support for tristate testing
+config MODULES
+	bool "Enable loadable module support"
+	modules
+	default y
+
+# Basic migration tests for all types
+config NEW_BOOL
+	bool "New bool option"
+	default OLD_BOOL
+
+config OLD_BOOL
+	bool
+	transitional
+
+config NEW_TRISTATE
+	tristate "New tristate option"
+	default OLD_TRISTATE
+
+config OLD_TRISTATE
+	tristate
+	transitional
+
+config NEW_STRING
+	string "New string option"
+	default OLD_STRING
+
+config OLD_STRING
+	string
+	transitional
+
+config NEW_HEX
+	hex "New hex option"
+	default OLD_HEX
+
+config OLD_HEX
+	hex
+	transitional
+
+config NEW_INT
+	int "New int option"
+	default OLD_INT
+
+config OLD_INT
+	int
+	transitional
+
+# Precedence tests for all types
+config NEW_BOOL_PRECEDENCE
+	bool "New bool option with precedence"
+	default OLD_BOOL_PRECEDENCE
+
+config OLD_BOOL_PRECEDENCE
+	bool
+	transitional
+
+config NEW_STRING_PRECEDENCE
+	string "New string option with precedence"
+	default OLD_STRING_PRECEDENCE
+
+config OLD_STRING_PRECEDENCE
+	string
+	transitional
+
+config NEW_TRISTATE_PRECEDENCE
+	tristate "New tristate option with precedence"
+	default OLD_TRISTATE_PRECEDENCE
+
+config OLD_TRISTATE_PRECEDENCE
+	tristate
+	transitional
+
+config NEW_HEX_PRECEDENCE
+	hex "New hex option with precedence"
+	default OLD_HEX_PRECEDENCE
+
+config OLD_HEX_PRECEDENCE
+	hex
+	transitional
+
+config NEW_INT_PRECEDENCE
+	int "New int option with precedence"
+	default OLD_INT_PRECEDENCE
+
+config OLD_INT_PRECEDENCE
+	int
+	transitional
+
+# Test that help sections are allowed for transitional symbols
+config OLD_WITH_HELP
+	bool
+	transitional
+	help
+	  This transitional symbol has a help section to validate that help is allowed.
+
+# Test that we can set something to =n via transitional symbol
+config NEW_DISABLED
+	tristate "Check for setting to disabled"
+	default OLD_DISABLED
+
+config OLD_DISABLED
+	tristate
+	transitional
+
+# Test that a potential new value disappears if it lacks a prompt
+config NEW_DISABLED_UNSAVED
+	tristate
+	default OLD_DISABLED
+
+config OLD_DISABLED_UNSAVED
+	tristate
+	transitional
+
+# Test conditional default: transitional value should not prevent prompting
+# when default visibility makes the expression evaluate to 'no'
+config DEPENDENCY_TEST
+	bool "Dependency for testing"
+	default n
+
+config NEW_CONDITIONAL_DEFAULT
+	bool "New option with conditional default"
+	default OLD_CONDITIONAL_DEFAULT if DEPENDENCY_TEST
+
+config OLD_CONDITIONAL_DEFAULT
+	bool
+	transitional
+
+config REGULAR_OPTION
+	bool "Regular option"
diff --git a/scripts/kconfig/tests/transitional/__init__.py b/scripts/kconfig/tests/transitional/__init__.py
new file mode 100644
index 000000000..b50ba2397
--- /dev/null
+++ b/scripts/kconfig/tests/transitional/__init__.py
@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: GPL-2.0
+"""
+Test transitional symbol migration functionality for all Kconfig types.
+
+This tests that:
+- OLD_* options in existing .config cause NEW_* options to be set
+- OLD_* options are not written to the new .config file
+- NEW_* options appear in the new .config file with correct values
+- NEW_* options with defaults from transitional symbols are not prompted
+- All Kconfig types work correctly: bool, tristate, string, hex, int
+- User-set NEW values take precedence over conflicting OLD transitional values
+"""
+
+def test(conf):
+    # Run olddefconfig to process the migration with the initial config
+    assert conf.olddefconfig(dot_config='initial_config') == 0
+
+    # Check that the configuration matches expected output
+    assert conf.config_contains('expected_config')
+
+    # Test oldconfig to ensure symbols with transitional defaults are not prompted
+    assert conf.oldconfig(dot_config='initial_config', in_keys='n\n') == 0
+
+    # Except for when conditional default evaluates to 'no'
+    assert conf.stdout_contains('expected_stdout')
diff --git a/scripts/kconfig/tests/transitional/expected_config b/scripts/kconfig/tests/transitional/expected_config
new file mode 100644
index 000000000..e01f5f070
--- /dev/null
+++ b/scripts/kconfig/tests/transitional/expected_config
@@ -0,0 +1,15 @@
+CONFIG_MODULES=y
+CONFIG_NEW_BOOL=y
+CONFIG_NEW_TRISTATE=m
+CONFIG_NEW_STRING="test string"
+CONFIG_NEW_HEX=0x1234
+CONFIG_NEW_INT=42
+# CONFIG_NEW_BOOL_PRECEDENCE is not set
+CONFIG_NEW_STRING_PRECEDENCE="user value"
+CONFIG_NEW_TRISTATE_PRECEDENCE=y
+CONFIG_NEW_HEX_PRECEDENCE=0xABCD
+CONFIG_NEW_INT_PRECEDENCE=100
+# CONFIG_NEW_DISABLED is not set
+# CONFIG_DEPENDENCY_TEST is not set
+# CONFIG_NEW_CONDITIONAL_DEFAULT is not set
+# CONFIG_REGULAR_OPTION is not set
diff --git a/scripts/kconfig/tests/transitional/expected_stdout b/scripts/kconfig/tests/transitional/expected_stdout
new file mode 100644
index 000000000..6f0b285d6
--- /dev/null
+++ b/scripts/kconfig/tests/transitional/expected_stdout
@@ -0,0 +1 @@
+New option with conditional default (NEW_CONDITIONAL_DEFAULT) [N/y/?] (NEW) n
diff --git a/scripts/kconfig/tests/transitional/initial_config b/scripts/kconfig/tests/transitional/initial_config
new file mode 100644
index 000000000..68b7da672
--- /dev/null
+++ b/scripts/kconfig/tests/transitional/initial_config
@@ -0,0 +1,20 @@
+CONFIG_MODULES=y
+CONFIG_OLD_BOOL=y
+CONFIG_OLD_TRISTATE=m
+CONFIG_OLD_STRING="test string"
+CONFIG_OLD_HEX=0x1234
+CONFIG_OLD_INT=42
+# CONFIG_NEW_BOOL_PRECEDENCE is not set
+CONFIG_OLD_BOOL_PRECEDENCE=y
+CONFIG_NEW_STRING_PRECEDENCE="user value"
+CONFIG_OLD_STRING_PRECEDENCE="old value"
+CONFIG_NEW_TRISTATE_PRECEDENCE=y
+CONFIG_OLD_TRISTATE_PRECEDENCE=m
+CONFIG_NEW_HEX_PRECEDENCE=0xABCD
+CONFIG_OLD_HEX_PRECEDENCE=0x5678
+CONFIG_NEW_INT_PRECEDENCE=100
+CONFIG_OLD_INT_PRECEDENCE=200
+# CONFIG_OLD_DISABLED is not set
+# CONFIG_OLD_DISABLED_UNSAVED is not set
+# CONFIG_DEPENDENCY_TEST is not set
+CONFIG_OLD_CONDITIONAL_DEFAULT=y