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, 215 insertions, 0 deletions

diff --git a/net/mpls/internal.h b/net/mpls/internal.h
new file mode 100644
index 000000000..80cb5bbcd
--- /dev/null
+++ b/net/mpls/internal.h
@@ -0,0 +1,215 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef MPLS_INTERNAL_H
+#define MPLS_INTERNAL_H
+#include <net/mpls.h>
+
+/* put a reasonable limit on the number of labels
+ * we will accept from userspace
+ */
+#define MAX_NEW_LABELS 30
+
+struct mpls_entry_decoded {
+	u32 label;
+	u8 ttl;
+	u8 tc;
+	u8 bos;
+};
+
+struct mpls_pcpu_stats {
+	struct mpls_link_stats	stats;
+	struct u64_stats_sync	syncp;
+};
+
+struct mpls_dev {
+	int				input_enabled;
+	struct net_device		*dev;
+	struct mpls_pcpu_stats __percpu	*stats;
+
+	struct ctl_table_header		*sysctl;
+	struct rcu_head			rcu;
+};
+
+#if BITS_PER_LONG == 32
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		TYPEOF_UNQUAL(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.pkts_field++;				\
+		ptr->stats.bytes_field += (len);			\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)					\
+	do {								\
+		TYPEOF_UNQUAL(*(mdev)->stats) *ptr =			\
+			raw_cpu_ptr((mdev)->stats);			\
+		local_bh_disable();					\
+		u64_stats_update_begin(&ptr->syncp);			\
+		ptr->stats.field++;					\
+		u64_stats_update_end(&ptr->syncp);			\
+		local_bh_enable();					\
+	} while (0)
+
+#else
+
+#define MPLS_INC_STATS_LEN(mdev, len, pkts_field, bytes_field)		\
+	do {								\
+		this_cpu_inc((mdev)->stats->stats.pkts_field);		\
+		this_cpu_add((mdev)->stats->stats.bytes_field, (len));	\
+	} while (0)
+
+#define MPLS_INC_STATS(mdev, field)			\
+	this_cpu_inc((mdev)->stats->stats.field)
+
+#endif
+
+struct sk_buff;
+
+#define LABEL_NOT_SPECIFIED (1 << 20)
+
+/* This maximum ha length copied from the definition of struct neighbour */
+#define VIA_ALEN_ALIGN sizeof(unsigned long)
+#define MAX_VIA_ALEN (ALIGN(MAX_ADDR_LEN, VIA_ALEN_ALIGN))
+
+enum mpls_payload_type {
+	MPT_UNSPEC, /* IPv4 or IPv6 */
+	MPT_IPV4 = 4,
+	MPT_IPV6 = 6,
+
+	/* Other types not implemented:
+	 *  - Pseudo-wire with or without control word (RFC4385)
+	 *  - GAL (RFC5586)
+	 */
+};
+
+struct mpls_nh { /* next hop label forwarding entry */
+	struct net_device	*nh_dev;
+	netdevice_tracker	nh_dev_tracker;
+
+	/* nh_flags is accessed under RCU in the packet path; it is
+	 * modified handling netdev events with rtnl lock held
+	 */
+	unsigned int		nh_flags;
+	u8			nh_labels;
+	u8			nh_via_alen;
+	u8			nh_via_table;
+	u8			nh_reserved1;
+
+	u32			nh_label[];
+};
+
+/* offset of via from beginning of mpls_nh */
+#define MPLS_NH_VIA_OFF(num_labels) \
+		ALIGN(sizeof(struct mpls_nh) + (num_labels) * sizeof(u32), \
+		      VIA_ALEN_ALIGN)
+
+/* all nexthops within a route have the same size based on the
+ * max number of labels and max via length across all nexthops
+ */
+#define MPLS_NH_SIZE(num_labels, max_via_alen)		\
+		(MPLS_NH_VIA_OFF((num_labels)) +	\
+		ALIGN((max_via_alen), VIA_ALEN_ALIGN))
+
+enum mpls_ttl_propagation {
+	MPLS_TTL_PROP_DEFAULT,
+	MPLS_TTL_PROP_ENABLED,
+	MPLS_TTL_PROP_DISABLED,
+};
+
+/* The route, nexthops and vias are stored together in the same memory
+ * block:
+ *
+ * +----------------------+
+ * | mpls_route           |
+ * +----------------------+
+ * | mpls_nh 0            |
+ * +----------------------+
+ * | alignment padding    |   4 bytes for odd number of labels
+ * +----------------------+
+ * | via[rt_max_alen] 0   |
+ * +----------------------+
+ * | alignment padding    |   via's aligned on sizeof(unsigned long)
+ * +----------------------+
+ * | ...                  |
+ * +----------------------+
+ * | mpls_nh n-1          |
+ * +----------------------+
+ * | via[rt_max_alen] n-1 |
+ * +----------------------+
+ */
+struct mpls_route { /* next hop label forwarding entry */
+	struct rcu_head		rt_rcu;
+	u8			rt_protocol;
+	u8			rt_payload_type;
+	u8			rt_max_alen;
+	u8			rt_ttl_propagate;
+	u8			rt_nhn;
+	/* rt_nhn_alive is accessed under RCU in the packet path; it
+	 * is modified handling netdev events with rtnl lock held
+	 */
+	u8			rt_nhn_alive;
+	u8			rt_nh_size;
+	u8			rt_via_offset;
+	u8			rt_reserved1;
+	struct mpls_nh		rt_nh[];
+};
+
+#define for_nexthops(rt) {						\
+	int nhsel; const struct mpls_nh *nh;				\
+	for (nhsel = 0, nh = (rt)->rt_nh;				\
+	     nhsel < (rt)->rt_nhn;					\
+	     nh = (void *)nh + (rt)->rt_nh_size, nhsel++)
+
+#define change_nexthops(rt) {						\
+	int nhsel; struct mpls_nh *nh;					\
+	for (nhsel = 0, nh = (rt)->rt_nh;				\
+	     nhsel < (rt)->rt_nhn;					\
+	     nh = (void *)nh + (rt)->rt_nh_size, nhsel++)
+
+#define endfor_nexthops(rt) }
+
+static inline struct mpls_entry_decoded mpls_entry_decode(struct mpls_shim_hdr *hdr)
+{
+	struct mpls_entry_decoded result;
+	unsigned entry = be32_to_cpu(hdr->label_stack_entry);
+
+	result.label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
+	result.ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
+	result.tc =  (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
+	result.bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
+
+	return result;
+}
+
+#define mpls_dereference(net, p)					\
+	rcu_dereference_protected(					\
+		(p),							\
+		lockdep_is_held(&(net)->mpls.platform_mutex))
+
+static inline struct mpls_dev *mpls_dev_rcu(const struct net_device *dev)
+{
+	return rcu_dereference(dev->mpls_ptr);
+}
+
+static inline struct mpls_dev *mpls_dev_get(const struct net *net,
+					    const struct net_device *dev)
+{
+	return mpls_dereference(net, dev->mpls_ptr);
+}
+
+int nla_put_labels(struct sk_buff *skb, int attrtype,  u8 labels,
+		   const u32 label[]);
+int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
+		   u32 label[], struct netlink_ext_ack *extack);
+bool mpls_output_possible(const struct net_device *dev);
+unsigned int mpls_dev_mtu(const struct net_device *dev);
+bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu);
+void mpls_stats_inc_outucastpkts(struct net *net,
+				 struct net_device *dev,
+				 const struct sk_buff *skb);
+
+#endif /* MPLS_INTERNAL_H */