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, 91 insertions, 0 deletions

diff --git a/include/rdma/opa_addr.h b/include/rdma/opa_addr.h
new file mode 100644
index 000000000..c6bfa2640
--- /dev/null
+++ b/include/rdma/opa_addr.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */
+/*
+ * Copyright(c) 2017 Intel Corporation.
+ */
+
+#ifndef OPA_ADDR_H
+#define OPA_ADDR_H
+
+#include <rdma/opa_smi.h>
+
+#define	OPA_SPECIAL_OUI		(0x00066AULL)
+#define OPA_MAKE_ID(x)          (cpu_to_be64(OPA_SPECIAL_OUI << 40 | (x)))
+#define OPA_TO_IB_UCAST_LID(x) (((x) >= be16_to_cpu(IB_MULTICAST_LID_BASE)) \
+				? 0 : x)
+#define OPA_GID_INDEX		0x1
+/**
+ * 0xF8 - 4 bits of multicast range and 1 bit for collective range
+ * Example: For 24 bit LID space,
+ * Multicast range: 0xF00000 to 0xF7FFFF
+ * Collective range: 0xF80000 to 0xFFFFFE
+ */
+#define OPA_MCAST_NR 0x4 /* Number of top bits set */
+#define OPA_COLLECTIVE_NR 0x1 /* Number of bits after MCAST_NR */
+
+/**
+ * ib_is_opa_gid: Returns true if the top 24 bits of the gid
+ * contains the OPA_STL_OUI identifier. This identifies that
+ * the provided gid is a special purpose GID meant to carry
+ * extended LID information.
+ *
+ * @gid: The Global identifier
+ */
+static inline bool ib_is_opa_gid(const union ib_gid *gid)
+{
+	return ((be64_to_cpu(gid->global.interface_id) >> 40) ==
+		OPA_SPECIAL_OUI);
+}
+
+/**
+ * opa_get_lid_from_gid: Returns the last 32 bits of the gid.
+ * OPA devices use one of the gids in the gid table to also
+ * store the lid.
+ *
+ * @gid: The Global identifier
+ */
+static inline u32 opa_get_lid_from_gid(const union ib_gid *gid)
+{
+	return be64_to_cpu(gid->global.interface_id) & 0xFFFFFFFF;
+}
+
+/**
+ * opa_is_extended_lid: Returns true if dlid or slid are
+ * extended.
+ *
+ * @dlid: The DLID
+ * @slid: The SLID
+ */
+static inline bool opa_is_extended_lid(__be32 dlid, __be32 slid)
+{
+	if ((be32_to_cpu(dlid) >=
+	     be16_to_cpu(IB_MULTICAST_LID_BASE)) ||
+	    (be32_to_cpu(slid) >=
+	     be16_to_cpu(IB_MULTICAST_LID_BASE)))
+		return true;
+
+	return false;
+}
+
+/* Get multicast lid base */
+static inline u32 opa_get_mcast_base(u32 nr_top_bits)
+{
+	return (be32_to_cpu(OPA_LID_PERMISSIVE) << (32 - nr_top_bits));
+}
+
+/* Check for a valid unicast LID for non-SM traffic types */
+static inline bool rdma_is_valid_unicast_lid(struct rdma_ah_attr *attr)
+{
+	if (attr->type == RDMA_AH_ATTR_TYPE_IB) {
+		if (!rdma_ah_get_dlid(attr) ||
+		    rdma_ah_get_dlid(attr) >=
+		    be16_to_cpu(IB_MULTICAST_LID_BASE))
+			return false;
+	} else if (attr->type == RDMA_AH_ATTR_TYPE_OPA) {
+		if (!rdma_ah_get_dlid(attr) ||
+		    rdma_ah_get_dlid(attr) >=
+		    opa_get_mcast_base(OPA_MCAST_NR))
+			return false;
+	}
+	return true;
+}
+#endif /* OPA_ADDR_H */