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, 130 insertions, 0 deletions

diff --git a/arch/x86/hyperv/nested.c b/arch/x86/hyperv/nested.c
new file mode 100644
index 000000000..8ccbb7c4f
--- /dev/null
+++ b/arch/x86/hyperv/nested.c
@@ -0,0 +1,130 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * Hyper-V nested virtualization code.
+ *
+ * Copyright (C) 2018, Microsoft, Inc.
+ *
+ * Author : Lan Tianyu <Tianyu.Lan@microsoft.com>
+ */
+#define pr_fmt(fmt)  "Hyper-V: " fmt
+
+
+#include <linux/types.h>
+#include <linux/export.h>
+#include <hyperv/hvhdk.h>
+#include <asm/mshyperv.h>
+#include <asm/tlbflush.h>
+
+#include <asm/trace/hyperv.h>
+
+int hyperv_flush_guest_mapping(u64 as)
+{
+	struct hv_guest_mapping_flush *flush;
+	u64 status;
+	unsigned long flags;
+	int ret = -ENOTSUPP;
+
+	if (!hv_hypercall_pg)
+		goto fault;
+
+	local_irq_save(flags);
+
+	flush = *this_cpu_ptr(hyperv_pcpu_input_arg);
+
+	if (unlikely(!flush)) {
+		local_irq_restore(flags);
+		goto fault;
+	}
+
+	flush->address_space = as;
+	flush->flags = 0;
+
+	status = hv_do_hypercall(HVCALL_FLUSH_GUEST_PHYSICAL_ADDRESS_SPACE,
+				 flush, NULL);
+	local_irq_restore(flags);
+
+	if (hv_result_success(status))
+		ret = 0;
+
+fault:
+	trace_hyperv_nested_flush_guest_mapping(as, ret);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(hyperv_flush_guest_mapping);
+
+int hyperv_fill_flush_guest_mapping_list(
+		struct hv_guest_mapping_flush_list *flush,
+		u64 start_gfn, u64 pages)
+{
+	u64 cur = start_gfn;
+	u64 additional_pages;
+	int gpa_n = 0;
+
+	do {
+		/*
+		 * If flush requests exceed max flush count, go back to
+		 * flush tlbs without range.
+		 */
+		if (gpa_n >= HV_MAX_FLUSH_REP_COUNT)
+			return -ENOSPC;
+
+		additional_pages = min_t(u64, pages, HV_MAX_FLUSH_PAGES) - 1;
+
+		flush->gpa_list[gpa_n].page.additional_pages = additional_pages;
+		flush->gpa_list[gpa_n].page.largepage = false;
+		flush->gpa_list[gpa_n].page.basepfn = cur;
+
+		pages -= additional_pages + 1;
+		cur += additional_pages + 1;
+		gpa_n++;
+	} while (pages > 0);
+
+	return gpa_n;
+}
+EXPORT_SYMBOL_GPL(hyperv_fill_flush_guest_mapping_list);
+
+int hyperv_flush_guest_mapping_range(u64 as,
+		hyperv_fill_flush_list_func fill_flush_list_func, void *data)
+{
+	struct hv_guest_mapping_flush_list *flush;
+	u64 status;
+	unsigned long flags;
+	int ret = -ENOTSUPP;
+	int gpa_n = 0;
+
+	if (!hv_hypercall_pg || !fill_flush_list_func)
+		goto fault;
+
+	local_irq_save(flags);
+
+	flush = *this_cpu_ptr(hyperv_pcpu_input_arg);
+
+	if (unlikely(!flush)) {
+		local_irq_restore(flags);
+		goto fault;
+	}
+
+	flush->address_space = as;
+	flush->flags = 0;
+
+	gpa_n = fill_flush_list_func(flush, data);
+	if (gpa_n < 0) {
+		local_irq_restore(flags);
+		goto fault;
+	}
+
+	status = hv_do_rep_hypercall(HVCALL_FLUSH_GUEST_PHYSICAL_ADDRESS_LIST,
+				     gpa_n, 0, flush, NULL);
+
+	local_irq_restore(flags);
+
+	if (hv_result_success(status))
+		ret = 0;
+	else
+		ret = hv_result(status);
+fault:
+	trace_hyperv_nested_flush_guest_mapping_range(as, ret);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(hyperv_flush_guest_mapping_range);