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, 197 insertions, 0 deletions

diff --git a/arch/mips/vdso/Makefile b/arch/mips/vdso/Makefile
new file mode 100644
index 000000000..00d3ba2c4
--- /dev/null
+++ b/arch/mips/vdso/Makefile
@@ -0,0 +1,197 @@
+# SPDX-License-Identifier: GPL-2.0
+# Objects to go into the VDSO.
+
+# Include the generic Makefile to check the built vdso.
+include $(srctree)/lib/vdso/Makefile.include
+
+obj-vdso-y := elf.o sigreturn.o
+
+# Common compiler flags between ABIs.
+ccflags-vdso := \
+	$(filter -I%,$(KBUILD_CFLAGS)) \
+	$(filter -E%,$(KBUILD_CFLAGS)) \
+	$(filter -mmicromips,$(KBUILD_CFLAGS)) \
+	$(filter -march=%,$(KBUILD_CFLAGS)) \
+	$(filter -m%-float,$(KBUILD_CFLAGS)) \
+	$(filter -mno-loongson-%,$(KBUILD_CFLAGS)) \
+	$(CLANG_FLAGS) \
+	-D__VDSO__
+
+ifndef CONFIG_64BIT
+ccflags-vdso += -DBUILD_VDSO32
+endif
+
+#
+# The -fno-jump-tables flag only prevents the compiler from generating
+# jump tables but does not prevent the compiler from emitting absolute
+# offsets.
+cflags-vdso := $(ccflags-vdso) \
+	$(filter -W%,$(filter-out -Wa$(comma)%,$(KBUILD_CFLAGS))) \
+	$(filter -std=%,$(KBUILD_CFLAGS)) \
+	-O3 -g -fPIC -fno-strict-aliasing -fno-common -fno-builtin -G 0 \
+	-mrelax-pic-calls $(call cc-option, -mexplicit-relocs) \
+	-fno-stack-protector -fno-jump-tables -DDISABLE_BRANCH_PROFILING \
+	$(call cc-option, -fno-asynchronous-unwind-tables)
+aflags-vdso := $(ccflags-vdso) \
+	-D__ASSEMBLY__ -Wa,-gdwarf-2
+
+ifneq ($(c-gettimeofday-y),)
+obj-vdso-y += vgettimeofday.o
+CFLAGS_vgettimeofday.o = -include $(c-gettimeofday-y)
+
+# config-n32-o32-env.c prepares the environment to build a 32bit vDSO
+# library on a 64bit kernel.
+# Note: Needs to be included before than the generic library.
+CFLAGS_vgettimeofday-o32.o = -include $(src)/config-n32-o32-env.c -include $(c-gettimeofday-y)
+CFLAGS_vgettimeofday-n32.o = -include $(src)/config-n32-o32-env.c -include $(c-gettimeofday-y)
+endif
+
+CFLAGS_REMOVE_vgettimeofday.o = $(CC_FLAGS_FTRACE)
+
+# VDSO linker flags.
+ldflags-y := -Bsymbolic --no-undefined -soname=linux-vdso.so.1 \
+	$(filter -E%,$(KBUILD_CFLAGS)) -shared \
+	-G 0 --eh-frame-hdr --hash-style=sysv --build-id=sha1 -T
+
+CFLAGS_REMOVE_vdso.o = $(CC_FLAGS_FTRACE)
+
+# Check that we don't have PIC 'jalr t9' calls left
+quiet_cmd_vdso_mips_check = VDSOCHK $@
+      cmd_vdso_mips_check = if $(OBJDUMP) --disassemble $@ | grep -E -h "jalr.*t9" > /dev/null; \
+		       then (echo >&2 "$@: PIC 'jalr t9' calls are not supported"; \
+			     rm -f $@; /bin/false); fi
+
+#
+# Shared build commands.
+#
+
+quiet_cmd_vdsold_and_vdso_check = LD      $@
+      cmd_vdsold_and_vdso_check = $(cmd_ld); $(cmd_vdso_check); $(cmd_vdso_mips_check)
+
+quiet_cmd_vdsoas_o_S = AS      $@
+      cmd_vdsoas_o_S = $(CC) $(a_flags) -c -o $@ $<
+
+# Strip rule for the raw .so files
+$(obj)/%.so.raw: OBJCOPYFLAGS := -S
+$(obj)/%.so.raw: $(obj)/%.so.dbg.raw FORCE
+	$(call if_changed,objcopy)
+
+hostprogs := genvdso
+
+quiet_cmd_genvdso = GENVDSO $@
+define cmd_genvdso
+	$(foreach file,$(filter %.raw,$^),cp $(file) $(file:%.raw=%) &&) \
+	$(obj)/genvdso $(<:%.raw=%) $(<:%.dbg.raw=%) $@ $(VDSO_NAME)
+endef
+
+#
+# Build native VDSO.
+#
+
+native-abi := $(filter -mabi=%,$(KBUILD_CFLAGS))
+
+targets += $(obj-vdso-y)
+targets += vdso.lds
+targets += vdso.so.dbg.raw vdso.so.raw
+targets += vdso.so.dbg vdso.so
+targets += vdso-image.c
+
+obj-vdso := $(obj-vdso-y:%.o=$(obj)/%.o)
+
+$(obj-vdso): KBUILD_CFLAGS := $(cflags-vdso) $(native-abi)
+$(obj-vdso): KBUILD_AFLAGS := $(aflags-vdso) $(native-abi)
+
+$(obj)/vdso.lds: KBUILD_CPPFLAGS := $(ccflags-vdso) $(native-abi)
+
+$(obj)/vdso.so.dbg.raw: $(obj)/vdso.lds $(obj-vdso) FORCE
+	$(call if_changed,vdsold_and_vdso_check)
+
+$(obj)/vdso-image.c: $(obj)/vdso.so.dbg.raw $(obj)/vdso.so.raw \
+                     $(obj)/genvdso FORCE
+	$(call if_changed,genvdso)
+
+obj-y += vdso-image.o
+
+#
+# Build O32 VDSO.
+#
+
+# Define these outside the ifdef to ensure they are picked up by clean.
+targets += $(obj-vdso-y:%.o=%-o32.o)
+targets += vdso-o32.lds
+targets += vdso-o32.so.dbg.raw vdso-o32.so.raw
+targets += vdso-o32.so.dbg vdso-o32.so
+targets += vdso-o32-image.c
+
+ifdef CONFIG_MIPS32_O32
+
+obj-vdso-o32 := $(obj-vdso-y:%.o=$(obj)/%-o32.o)
+
+$(obj-vdso-o32): KBUILD_CFLAGS := $(cflags-vdso) -mabi=32
+$(obj-vdso-o32): KBUILD_AFLAGS := $(aflags-vdso) -mabi=32
+
+$(obj)/%-o32.o: $(src)/%.S FORCE
+	$(call if_changed_dep,vdsoas_o_S)
+
+$(obj)/%-o32.o: $(src)/%.c FORCE
+	$(call cmd,force_checksrc)
+	$(call if_changed_rule,cc_o_c)
+
+$(obj)/vdso-o32.lds: KBUILD_CPPFLAGS := $(ccflags-vdso) -mabi=32
+$(obj)/vdso-o32.lds: $(src)/vdso.lds.S FORCE
+	$(call if_changed_dep,cpp_lds_S)
+
+$(obj)/vdso-o32.so.dbg.raw: $(obj)/vdso-o32.lds $(obj-vdso-o32) FORCE
+	$(call if_changed,vdsold_and_vdso_check)
+
+$(obj)/vdso-o32-image.c: VDSO_NAME := o32
+$(obj)/vdso-o32-image.c: $(obj)/vdso-o32.so.dbg.raw $(obj)/vdso-o32.so.raw \
+                         $(obj)/genvdso FORCE
+	$(call if_changed,genvdso)
+
+obj-y += vdso-o32-image.o
+
+endif
+
+#
+# Build N32 VDSO.
+#
+
+targets += $(obj-vdso-y:%.o=%-n32.o)
+targets += vdso-n32.lds
+targets += vdso-n32.so.dbg.raw vdso-n32.so.raw
+targets += vdso-n32.so.dbg vdso-n32.so
+targets += vdso-n32-image.c
+
+ifdef CONFIG_MIPS32_N32
+
+obj-vdso-n32 := $(obj-vdso-y:%.o=$(obj)/%-n32.o)
+
+$(obj-vdso-n32): KBUILD_CFLAGS := $(cflags-vdso) -mabi=n32
+$(obj-vdso-n32): KBUILD_AFLAGS := $(aflags-vdso) -mabi=n32
+
+$(obj)/%-n32.o: $(src)/%.S FORCE
+	$(call if_changed_dep,vdsoas_o_S)
+
+$(obj)/%-n32.o: $(src)/%.c FORCE
+	$(call cmd,force_checksrc)
+	$(call if_changed_rule,cc_o_c)
+
+$(obj)/vdso-n32.lds: KBUILD_CPPFLAGS := $(ccflags-vdso) -mabi=n32
+$(obj)/vdso-n32.lds: $(src)/vdso.lds.S FORCE
+	$(call if_changed_dep,cpp_lds_S)
+
+$(obj)/vdso-n32.so.dbg.raw: $(obj)/vdso-n32.lds $(obj-vdso-n32) FORCE
+	$(call if_changed,vdsold_and_vdso_check)
+
+$(obj)/vdso-n32-image.c: VDSO_NAME := n32
+$(obj)/vdso-n32-image.c: $(obj)/vdso-n32.so.dbg.raw $(obj)/vdso-n32.so.raw \
+                         $(obj)/genvdso FORCE
+	$(call if_changed,genvdso)
+
+obj-y += vdso-n32-image.o
+
+endif
+
+# FIXME: Need install rule for debug.
+# Needs to deal with dependency for generation of dbg by cmd_genvdso...