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, 354 insertions, 0 deletions

diff --git a/net/mac802154/driver-ops.h b/net/mac802154/driver-ops.h
new file mode 100644
index 000000000..a7af3f0dd
--- /dev/null
+++ b/net/mac802154/driver-ops.h
@@ -0,0 +1,354 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __MAC802154_DRIVER_OPS
+#define __MAC802154_DRIVER_OPS
+
+#include <linux/types.h>
+#include <linux/rtnetlink.h>
+
+#include <net/mac802154.h>
+
+#include "ieee802154_i.h"
+#include "trace.h"
+
+static inline int
+drv_xmit_async(struct ieee802154_local *local, struct sk_buff *skb)
+{
+	return local->ops->xmit_async(&local->hw, skb);
+}
+
+static inline int
+drv_xmit_sync(struct ieee802154_local *local, struct sk_buff *skb)
+{
+	might_sleep();
+
+	return local->ops->xmit_sync(&local->hw, skb);
+}
+
+static inline int drv_set_pan_id(struct ieee802154_local *local, __le16 pan_id)
+{
+	struct ieee802154_hw_addr_filt filt;
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_hw_addr_filt) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	filt.pan_id = pan_id;
+
+	trace_802154_drv_set_pan_id(local, pan_id);
+	ret = local->ops->set_hw_addr_filt(&local->hw, &filt,
+					    IEEE802154_AFILT_PANID_CHANGED);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_extended_addr(struct ieee802154_local *local, __le64 extended_addr)
+{
+	struct ieee802154_hw_addr_filt filt;
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_hw_addr_filt) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	filt.ieee_addr = extended_addr;
+
+	trace_802154_drv_set_extended_addr(local, extended_addr);
+	ret = local->ops->set_hw_addr_filt(&local->hw, &filt,
+					    IEEE802154_AFILT_IEEEADDR_CHANGED);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_short_addr(struct ieee802154_local *local, __le16 short_addr)
+{
+	struct ieee802154_hw_addr_filt filt;
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_hw_addr_filt) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	filt.short_addr = short_addr;
+
+	trace_802154_drv_set_short_addr(local, short_addr);
+	ret = local->ops->set_hw_addr_filt(&local->hw, &filt,
+					    IEEE802154_AFILT_SADDR_CHANGED);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_pan_coord(struct ieee802154_local *local, bool is_coord)
+{
+	struct ieee802154_hw_addr_filt filt;
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_hw_addr_filt) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	filt.pan_coord = is_coord;
+
+	trace_802154_drv_set_pan_coord(local, is_coord);
+	ret = local->ops->set_hw_addr_filt(&local->hw, &filt,
+					    IEEE802154_AFILT_PANC_CHANGED);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_promiscuous_mode(struct ieee802154_local *local, bool on)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_promiscuous_mode) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_promiscuous_mode(local, on);
+	ret = local->ops->set_promiscuous_mode(&local->hw, on);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int drv_start(struct ieee802154_local *local,
+			    enum ieee802154_filtering_level level,
+			    const struct ieee802154_hw_addr_filt *addr_filt)
+{
+	int ret;
+
+	might_sleep();
+
+	/* setup receive mode parameters e.g. address mode */
+	if (local->hw.flags & IEEE802154_HW_AFILT) {
+		ret = drv_set_pan_id(local, addr_filt->pan_id);
+		if (ret < 0)
+			return ret;
+
+		ret = drv_set_short_addr(local, addr_filt->short_addr);
+		if (ret < 0)
+			return ret;
+
+		ret = drv_set_extended_addr(local, addr_filt->ieee_addr);
+		if (ret < 0)
+			return ret;
+	}
+
+	switch (level) {
+	case IEEE802154_FILTERING_NONE:
+		fallthrough;
+	case IEEE802154_FILTERING_1_FCS:
+		fallthrough;
+	case IEEE802154_FILTERING_2_PROMISCUOUS:
+		/* TODO: Requires a different receive mode setup e.g.
+		 * at86rf233 hardware.
+		 */
+		fallthrough;
+	case IEEE802154_FILTERING_3_SCAN:
+		if (local->hw.flags & IEEE802154_HW_PROMISCUOUS) {
+			ret = drv_set_promiscuous_mode(local, true);
+			if (ret < 0)
+				return ret;
+		} else {
+			return -EOPNOTSUPP;
+		}
+
+		/* In practice other filtering levels can be requested, but as
+		 * for now most hardware/drivers only support
+		 * IEEE802154_FILTERING_NONE, we fallback to this actual
+		 * filtering level in hardware and make our own additional
+		 * filtering in mac802154 receive path.
+		 *
+		 * TODO: Move this logic to the device drivers as hardware may
+		 * support more higher level filters. Hardware may also require
+		 * a different order how register are set, which could currently
+		 * be buggy, so all received parameters need to be moved to the
+		 * start() callback and let the driver go into the mode before
+		 * it will turn on receive handling.
+		 */
+		local->phy->filtering = IEEE802154_FILTERING_NONE;
+		break;
+	case IEEE802154_FILTERING_4_FRAME_FIELDS:
+		/* Do not error out if IEEE802154_HW_PROMISCUOUS because we
+		 * expect the hardware to operate at the level
+		 * IEEE802154_FILTERING_4_FRAME_FIELDS anyway.
+		 */
+		if (local->hw.flags & IEEE802154_HW_PROMISCUOUS) {
+			ret = drv_set_promiscuous_mode(local, false);
+			if (ret < 0)
+				return ret;
+		}
+
+		local->phy->filtering = IEEE802154_FILTERING_4_FRAME_FIELDS;
+		break;
+	default:
+		WARN_ON(1);
+		return -EINVAL;
+	}
+
+	trace_802154_drv_start(local);
+	local->started = true;
+	smp_mb();
+	ret = local->ops->start(&local->hw);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline void drv_stop(struct ieee802154_local *local)
+{
+	might_sleep();
+
+	trace_802154_drv_stop(local);
+	local->ops->stop(&local->hw);
+	trace_802154_drv_return_void(local);
+
+	/* sync away all work on the tasklet before clearing started */
+	tasklet_disable(&local->tasklet);
+	tasklet_enable(&local->tasklet);
+
+	barrier();
+
+	local->started = false;
+}
+
+static inline int
+drv_set_channel(struct ieee802154_local *local, u8 page, u8 channel)
+{
+	int ret;
+
+	might_sleep();
+
+	trace_802154_drv_set_channel(local, page, channel);
+	ret = local->ops->set_channel(&local->hw, page, channel);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int drv_set_tx_power(struct ieee802154_local *local, s32 mbm)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_txpower) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_tx_power(local, mbm);
+	ret = local->ops->set_txpower(&local->hw, mbm);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int drv_set_cca_mode(struct ieee802154_local *local,
+				   const struct wpan_phy_cca *cca)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_cca_mode) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_cca_mode(local, cca);
+	ret = local->ops->set_cca_mode(&local->hw, cca);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int drv_set_lbt_mode(struct ieee802154_local *local, bool mode)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_lbt) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_lbt_mode(local, mode);
+	ret = local->ops->set_lbt(&local->hw, mode);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_cca_ed_level(struct ieee802154_local *local, s32 mbm)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_cca_ed_level) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_cca_ed_level(local, mbm);
+	ret = local->ops->set_cca_ed_level(&local->hw, mbm);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_csma_params(struct ieee802154_local *local, u8 min_be, u8 max_be,
+		    u8 max_csma_backoffs)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_csma_params) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_csma_params(local, min_be, max_be,
+					 max_csma_backoffs);
+	ret = local->ops->set_csma_params(&local->hw, min_be, max_be,
+					   max_csma_backoffs);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+static inline int
+drv_set_max_frame_retries(struct ieee802154_local *local, s8 max_frame_retries)
+{
+	int ret;
+
+	might_sleep();
+
+	if (!local->ops->set_frame_retries) {
+		WARN_ON(1);
+		return -EOPNOTSUPP;
+	}
+
+	trace_802154_drv_set_max_frame_retries(local, max_frame_retries);
+	ret = local->ops->set_frame_retries(&local->hw, max_frame_retries);
+	trace_802154_drv_return_int(local, ret);
+	return ret;
+}
+
+#endif /* __MAC802154_DRIVER_OPS */