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, 98 insertions, 0 deletions

diff --git a/scripts/Makefile.vmlinux_o b/scripts/Makefile.vmlinux_o
new file mode 100644
index 000000000..24a3a4fd2
--- /dev/null
+++ b/scripts/Makefile.vmlinux_o
@@ -0,0 +1,98 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+PHONY := __default
+__default: vmlinux.o
+
+include include/config/auto.conf
+include $(srctree)/scripts/Kbuild.include
+
+# for objtool
+include $(srctree)/scripts/Makefile.lib
+
+# Generate a linker script to ensure correct ordering of initcalls for Clang LTO
+# ---------------------------------------------------------------------------
+
+quiet_cmd_gen_initcalls_lds = GEN     $@
+      cmd_gen_initcalls_lds = \
+	$(PYTHON3) $(srctree)/scripts/jobserver-exec \
+	$(PERL) $(real-prereqs) > $@
+
+.tmp_initcalls.lds: $(srctree)/scripts/generate_initcall_order.pl \
+		vmlinux.a $(KBUILD_VMLINUX_LIBS) FORCE
+	$(call if_changed,gen_initcalls_lds)
+
+targets := .tmp_initcalls.lds
+
+ifdef CONFIG_LTO_CLANG
+initcalls-lds := .tmp_initcalls.lds
+endif
+
+# objtool for vmlinux.o
+# ---------------------------------------------------------------------------
+#
+# For delay-objtool (IBT or LTO), objtool doesn't run on individual translation
+# units.  Instead it runs on vmlinux.o.
+#
+# For !delay-objtool + CONFIG_NOINSTR_VALIDATION, it runs on both translation
+# units and vmlinux.o, with the latter only used for noinstr/unret validation.
+
+objtool-enabled := $(or $(delay-objtool),$(CONFIG_NOINSTR_VALIDATION))
+
+ifeq ($(delay-objtool),y)
+vmlinux-objtool-args-y					+= $(objtool-args-y)
+else
+vmlinux-objtool-args-$(CONFIG_OBJTOOL_WERROR)		+= --werror
+endif
+
+vmlinux-objtool-args-$(CONFIG_NOINSTR_VALIDATION)	+= --noinstr \
+							   $(if $(or $(CONFIG_MITIGATION_UNRET_ENTRY),$(CONFIG_MITIGATION_SRSO)), --unret)
+
+# Only used for builds initiated by klp-build
+vmlinux-objtool-args-$(if $(KLP_SYMIDS),y)		+= --klp-symids
+
+objtool-args = $(vmlinux-objtool-args-y) --link
+
+# Link of vmlinux.o used for section mismatch analysis
+# ---------------------------------------------------------------------------
+
+vmlinux-o-ld-args-$(CONFIG_BUILTIN_MODULE_RANGES)	+= -Map=$@.map
+
+quiet_cmd_ld_vmlinux.o = LD      $@
+      cmd_ld_vmlinux.o = \
+	$(LD) ${KBUILD_LDFLAGS} -r -o $@ \
+	$(vmlinux-o-ld-args-y) \
+	$(addprefix -T , $(initcalls-lds)) \
+	--whole-archive vmlinux.a --no-whole-archive \
+	--start-group $(KBUILD_VMLINUX_LIBS) --end-group \
+	$(cmd_objtool)
+
+cmd_check_function_names = $(srctree)/scripts/check-function-names.sh $@
+
+define rule_ld_vmlinux.o
+	$(call cmd_and_savecmd,ld_vmlinux.o)
+	$(call cmd,gen_objtooldep)
+	$(call cmd,check_function_names)
+endef
+
+
+vmlinux.o: $(initcalls-lds) vmlinux.a $(KBUILD_VMLINUX_LIBS) FORCE
+	$(call if_changed_rule,ld_vmlinux.o)
+
+targets += vmlinux.o
+
+# Add FORCE to the prerequisites of a target to force it to be always rebuilt.
+# ---------------------------------------------------------------------------
+
+PHONY += FORCE
+FORCE:
+
+# Read all saved command lines and dependencies for the $(targets) we
+# may be building above, using $(if_changed{,_dep}). As an
+# optimization, we don't need to read them if the target does not
+# exist, we will rebuild anyway in that case.
+
+existing-targets := $(wildcard $(sort $(targets)))
+
+-include $(foreach f,$(existing-targets),$(dir $(f)).$(notdir $(f)).cmd)
+
+.PHONY: $(PHONY)