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, 141 insertions, 0 deletions

diff --git a/io_uring/notif.c b/io_uring/notif.c
new file mode 100644
index 000000000..efce8ae12
--- /dev/null
+++ b/io_uring/notif.c
@@ -0,0 +1,141 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/file.h>
+#include <linux/slab.h>
+#include <linux/net.h>
+#include <linux/io_uring.h>
+
+#include "io_uring.h"
+#include "notif.h"
+#include "rsrc.h"
+
+static const struct ubuf_info_ops io_ubuf_ops;
+
+static void io_notif_tw_complete(struct io_tw_req tw_req, io_tw_token_t tw)
+{
+	struct io_kiocb *notif = tw_req.req;
+	struct io_notif_data *nd = io_notif_to_data(notif);
+	struct io_ring_ctx *ctx = notif->ctx;
+
+	lockdep_assert_held(&ctx->uring_lock);
+
+	do {
+		notif = cmd_to_io_kiocb(nd);
+
+		if (WARN_ON_ONCE(ctx != notif->ctx))
+			return;
+		lockdep_assert(refcount_read(&nd->uarg.refcnt) == 0);
+
+		if (unlikely(nd->zc_report) && (nd->zc_copied || !nd->zc_used))
+			notif->cqe.res |= IORING_NOTIF_USAGE_ZC_COPIED;
+
+		if (nd->account_pages && notif->ctx->user) {
+			__io_unaccount_mem(notif->ctx->user, nd->account_pages);
+			nd->account_pages = 0;
+		}
+
+		nd = nd->next;
+		io_req_task_complete((struct io_tw_req){notif}, tw);
+	} while (nd);
+}
+
+void io_tx_ubuf_complete(struct sk_buff *skb, struct ubuf_info *uarg,
+			 bool success)
+{
+	struct io_notif_data *nd = container_of(uarg, struct io_notif_data, uarg);
+	struct io_kiocb *notif = cmd_to_io_kiocb(nd);
+	unsigned tw_flags;
+
+	if (nd->zc_report) {
+		if (success && !nd->zc_used && skb)
+			WRITE_ONCE(nd->zc_used, true);
+		else if (!success && !nd->zc_copied)
+			WRITE_ONCE(nd->zc_copied, true);
+	}
+
+	if (!refcount_dec_and_test(&uarg->refcnt))
+		return;
+
+	if (nd->head != nd) {
+		io_tx_ubuf_complete(skb, &nd->head->uarg, success);
+		return;
+	}
+
+	tw_flags = nd->next ? 0 : IOU_F_TWQ_LAZY_WAKE;
+	notif->io_task_work.func = io_notif_tw_complete;
+	__io_req_task_work_add(notif, tw_flags);
+}
+
+static int io_link_skb(struct sk_buff *skb, struct ubuf_info *uarg)
+{
+	struct io_notif_data *nd, *prev_nd;
+	struct io_kiocb *prev_notif, *notif;
+	struct ubuf_info *prev_uarg = skb_zcopy(skb);
+
+	nd = container_of(uarg, struct io_notif_data, uarg);
+	notif = cmd_to_io_kiocb(nd);
+
+	if (!prev_uarg) {
+		net_zcopy_get(&nd->uarg);
+		skb_zcopy_init(skb, &nd->uarg);
+		return 0;
+	}
+	/* handle it separately as we can't link a notif to itself */
+	if (unlikely(prev_uarg == &nd->uarg))
+		return 0;
+	/* we can't join two links together, just request a fresh skb */
+	if (unlikely(nd->head != nd || nd->next))
+		return -EEXIST;
+	/* don't mix zc providers */
+	if (unlikely(prev_uarg->ops != &io_ubuf_ops))
+		return -EEXIST;
+
+	prev_nd = container_of(prev_uarg, struct io_notif_data, uarg);
+	prev_notif = cmd_to_io_kiocb(prev_nd);
+
+	/* make sure all notifications can be finished in the same task_work */
+	if (unlikely(notif->ctx != prev_notif->ctx ||
+		     notif->tctx != prev_notif->tctx))
+		return -EEXIST;
+
+	nd->head = prev_nd->head;
+	nd->next = prev_nd->next;
+	prev_nd->next = nd;
+	net_zcopy_get(&nd->head->uarg);
+	return 0;
+}
+
+static const struct ubuf_info_ops io_ubuf_ops = {
+	.complete = io_tx_ubuf_complete,
+	.link_skb = io_link_skb,
+};
+
+struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
+	__must_hold(&ctx->uring_lock)
+{
+	struct io_kiocb *notif;
+	struct io_notif_data *nd;
+
+	if (unlikely(!io_alloc_req(ctx, &notif)))
+		return NULL;
+	notif->ctx = ctx;
+	notif->opcode = IORING_OP_NOP;
+	notif->flags = 0;
+	notif->file = NULL;
+	notif->tctx = current->io_uring;
+	io_get_task_refs(1);
+	notif->file_node = NULL;
+	notif->buf_node = NULL;
+
+	nd = io_notif_to_data(notif);
+	nd->zc_report = false;
+	nd->account_pages = 0;
+	nd->next = NULL;
+	nd->head = nd;
+
+	nd->uarg.flags = IO_NOTIF_UBUF_FLAGS;
+	nd->uarg.ops = &io_ubuf_ops;
+	refcount_set(&nd->uarg.refcnt, 1);
+	return notif;
+}