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, 193 insertions, 0 deletions

diff --git a/net/ceph/decode.c b/net/ceph/decode.c
new file mode 100644
index 000000000..2f21af38c
--- /dev/null
+++ b/net/ceph/decode.c
@@ -0,0 +1,193 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/ceph/ceph_debug.h>
+
+#include <linux/inet.h>
+
+#include <linux/ceph/decode.h>
+#include <linux/ceph/messenger.h>  /* for ceph_pr_addr() */
+
+static int
+ceph_decode_entity_addr_versioned(void **p, void *end,
+				  struct ceph_entity_addr *addr)
+{
+	int ret;
+	u8 struct_v;
+	u32 struct_len, addr_len;
+	void *struct_end;
+
+	ret = ceph_start_decoding(p, end, 1, "entity_addr_t", &struct_v,
+				  &struct_len);
+	if (ret)
+		goto bad;
+
+	ret = -EINVAL;
+	struct_end = *p + struct_len;
+
+	ceph_decode_copy_safe(p, end, &addr->type, sizeof(addr->type), bad);
+
+	ceph_decode_copy_safe(p, end, &addr->nonce, sizeof(addr->nonce), bad);
+
+	ceph_decode_32_safe(p, end, addr_len, bad);
+	if (addr_len > sizeof(addr->in_addr))
+		goto bad;
+
+	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
+	if (addr_len) {
+		ceph_decode_copy_safe(p, end, &addr->in_addr, addr_len, bad);
+
+		addr->in_addr.ss_family =
+			le16_to_cpu((__force __le16)addr->in_addr.ss_family);
+	}
+
+	/* Advance past anything the client doesn't yet understand */
+	*p = struct_end;
+	ret = 0;
+bad:
+	return ret;
+}
+
+static int
+ceph_decode_entity_addr_legacy(void **p, void *end,
+			       struct ceph_entity_addr *addr)
+{
+	int ret = -EINVAL;
+
+	/* Skip rest of type field */
+	ceph_decode_skip_n(p, end, 3, bad);
+
+	/*
+	 * Clients that don't support ADDR2 always send TYPE_NONE, change it
+	 * to TYPE_LEGACY for forward compatibility.
+	 */
+	addr->type = CEPH_ENTITY_ADDR_TYPE_LEGACY;
+	ceph_decode_copy_safe(p, end, &addr->nonce, sizeof(addr->nonce), bad);
+	memset(&addr->in_addr, 0, sizeof(addr->in_addr));
+	ceph_decode_copy_safe(p, end, &addr->in_addr,
+			      sizeof(addr->in_addr), bad);
+	addr->in_addr.ss_family =
+			be16_to_cpu((__force __be16)addr->in_addr.ss_family);
+	ret = 0;
+bad:
+	return ret;
+}
+
+int
+ceph_decode_entity_addr(void **p, void *end, struct ceph_entity_addr *addr)
+{
+	u8 marker;
+
+	ceph_decode_8_safe(p, end, marker, bad);
+	if (marker == 1)
+		return ceph_decode_entity_addr_versioned(p, end, addr);
+	else if (marker == 0)
+		return ceph_decode_entity_addr_legacy(p, end, addr);
+bad:
+	return -EINVAL;
+}
+EXPORT_SYMBOL(ceph_decode_entity_addr);
+
+/*
+ * Return addr of desired type (MSGR2 or LEGACY) or error.  In case of
+ * multiple matches, use the first one for compatibility with userspace
+ * messenger.
+ *
+ * Assume encoding with MSG_ADDR2.
+ */
+int ceph_decode_entity_addrvec(void **p, void *end, bool msgr2,
+			       struct ceph_entity_addr *addr)
+{
+	__le32 my_type = msgr2 ? CEPH_ENTITY_ADDR_TYPE_MSGR2 :
+				 CEPH_ENTITY_ADDR_TYPE_LEGACY;
+	struct ceph_entity_addr tmp_addr;
+	int addr_cnt;
+	bool found;
+	u8 marker;
+	int ret;
+	int i;
+
+	ceph_decode_8_safe(p, end, marker, e_inval);
+	if (marker != 2) {
+		pr_err("bad addrvec marker %d\n", marker);
+		return -EINVAL;
+	}
+
+	ceph_decode_32_safe(p, end, addr_cnt, e_inval);
+	dout("%s addr_cnt %d\n", __func__, addr_cnt);
+
+	found = false;
+	for (i = 0; i < addr_cnt; i++) {
+		ret = ceph_decode_entity_addr(p, end, &tmp_addr);
+		if (ret)
+			return ret;
+
+		dout("%s i %d addr %s\n", __func__, i, ceph_pr_addr(&tmp_addr));
+		if (tmp_addr.type == my_type) {
+			if (!found) {
+				memcpy(addr, &tmp_addr, sizeof(*addr));
+				found = true;
+			} else {
+				dout("%s skipping extra match of type %d in addrvec\n",
+				     __func__, le32_to_cpu(my_type));
+			}
+		}
+	}
+
+	if (found)
+		return 0;
+
+	if (!addr_cnt)
+		return 0;  /* normal -- e.g. unused OSD id/slot */
+
+	if (addr_cnt == 1 && !memchr_inv(&tmp_addr, 0, sizeof(tmp_addr)))
+		return 0;  /* weird but effectively the same as !addr_cnt */
+
+	pr_err("no match of type %d in addrvec\n", le32_to_cpu(my_type));
+	return -ENOENT;
+
+e_inval:
+	return -EINVAL;
+}
+EXPORT_SYMBOL(ceph_decode_entity_addrvec);
+
+static int get_sockaddr_encoding_len(sa_family_t family)
+{
+	union {
+		struct sockaddr sa;
+		struct sockaddr_in sin;
+		struct sockaddr_in6 sin6;
+	} u;
+
+	switch (family) {
+	case AF_INET:
+		return sizeof(u.sin);
+	case AF_INET6:
+		return sizeof(u.sin6);
+	default:
+		return sizeof(u);
+	}
+}
+
+int ceph_entity_addr_encoding_len(const struct ceph_entity_addr *addr)
+{
+	sa_family_t family = get_unaligned(&addr->in_addr.ss_family);
+	int addr_len = get_sockaddr_encoding_len(family);
+
+	return 1 + CEPH_ENCODING_START_BLK_LEN + 4 + 4 + 4 + addr_len;
+}
+
+void ceph_encode_entity_addr(void **p, const struct ceph_entity_addr *addr)
+{
+	sa_family_t family = get_unaligned(&addr->in_addr.ss_family);
+	int addr_len = get_sockaddr_encoding_len(family);
+
+	ceph_encode_8(p, 1);  /* marker */
+	ceph_start_encoding(p, 1, 1, sizeof(addr->type) +
+				     sizeof(addr->nonce) +
+				     sizeof(u32) + addr_len);
+	ceph_encode_copy(p, &addr->type, sizeof(addr->type));
+	ceph_encode_copy(p, &addr->nonce, sizeof(addr->nonce));
+
+	ceph_encode_32(p, addr_len);
+	ceph_encode_16(p, family);
+	ceph_encode_copy(p, addr->in_addr.__data, addr_len - sizeof(family));
+}