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, 157 insertions, 0 deletions

diff --git a/block/partitions/mac.c b/block/partitions/mac.c
new file mode 100644
index 000000000..df03ca428
--- /dev/null
+++ b/block/partitions/mac.c
@@ -0,0 +1,157 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  fs/partitions/mac.c
+ *
+ *  Code extracted from drivers/block/genhd.c
+ *  Copyright (C) 1991-1998  Linus Torvalds
+ *  Re-organised Feb 1998 Russell King
+ */
+
+#include <linux/ctype.h>
+#include "check.h"
+#include "mac.h"
+
+#ifdef CONFIG_PPC_PMAC
+#include <asm/machdep.h>
+extern void note_bootable_part(dev_t dev, int part, int goodness);
+#endif
+
+/*
+ * Code to understand MacOS partition tables.
+ */
+
+#ifdef CONFIG_PPC_PMAC
+static inline void mac_fix_string(char *stg, int len)
+{
+	int i;
+
+	for (i = len - 1; i >= 0 && stg[i] == ' '; i--)
+		stg[i] = 0;
+}
+#endif
+
+int mac_partition(struct parsed_partitions *state)
+{
+	Sector sect;
+	unsigned char *data;
+	int slot, blocks_in_map;
+	unsigned secsize, datasize, partoffset;
+#ifdef CONFIG_PPC_PMAC
+	int found_root = 0;
+	int found_root_goodness = 0;
+#endif
+	struct mac_partition *part;
+	struct mac_driver_desc *md;
+
+	/* Get 0th block and look at the first partition map entry. */
+	md = read_part_sector(state, 0, &sect);
+	if (!md)
+		return -1;
+	if (be16_to_cpu(md->signature) != MAC_DRIVER_MAGIC) {
+		put_dev_sector(sect);
+		return 0;
+	}
+	secsize = be16_to_cpu(md->block_size);
+	put_dev_sector(sect);
+
+	/*
+	 * If the "block size" is not a power of 2, things get weird - we might
+	 * end up with a partition straddling a sector boundary, so we wouldn't
+	 * be able to read a partition entry with read_part_sector().
+	 * Real block sizes are probably (?) powers of two, so just require
+	 * that.
+	 */
+	if (!is_power_of_2(secsize))
+		return -1;
+	datasize = round_down(secsize, 512);
+	data = read_part_sector(state, datasize / 512, &sect);
+	if (!data)
+		return -1;
+	partoffset = secsize % 512;
+	if (partoffset + sizeof(*part) > datasize) {
+		put_dev_sector(sect);
+		return -1;
+	}
+	part = (struct mac_partition *) (data + partoffset);
+	if (be16_to_cpu(part->signature) != MAC_PARTITION_MAGIC) {
+		put_dev_sector(sect);
+		return 0;		/* not a MacOS disk */
+	}
+	blocks_in_map = be32_to_cpu(part->map_count);
+	if (blocks_in_map < 0 || blocks_in_map >= DISK_MAX_PARTS) {
+		put_dev_sector(sect);
+		return 0;
+	}
+
+	if (blocks_in_map >= state->limit)
+		blocks_in_map = state->limit - 1;
+
+	seq_buf_puts(&state->pp_buf, " [mac]");
+	for (slot = 1; slot <= blocks_in_map; ++slot) {
+		int pos = slot * secsize;
+		put_dev_sector(sect);
+		data = read_part_sector(state, pos/512, &sect);
+		if (!data)
+			return -1;
+		part = (struct mac_partition *) (data + pos%512);
+		if (be16_to_cpu(part->signature) != MAC_PARTITION_MAGIC)
+			break;
+		put_partition(state, slot,
+			be32_to_cpu(part->start_block) * (secsize/512),
+			be32_to_cpu(part->block_count) * (secsize/512));
+
+		if (!strncasecmp(part->type, "Linux_RAID", 10))
+			state->parts[slot].flags = ADDPART_FLAG_RAID;
+#ifdef CONFIG_PPC_PMAC
+		/*
+		 * If this is the first bootable partition, tell the
+		 * setup code, in case it wants to make this the root.
+		 */
+		if (machine_is(powermac)) {
+			int goodness = 0;
+
+			mac_fix_string(part->processor, 16);
+			mac_fix_string(part->name, 32);
+			mac_fix_string(part->type, 32);					
+		    
+			if ((be32_to_cpu(part->status) & MAC_STATUS_BOOTABLE)
+			    && strcasecmp(part->processor, "powerpc") == 0)
+				goodness++;
+
+			if (strcasecmp(part->type, "Apple_UNIX_SVR2") == 0
+			    || (strncasecmp(part->type, "Linux", 5) == 0
+			        && strcasecmp(part->type, "Linux_swap") != 0)) {
+				int i, l;
+
+				goodness++;
+				l = strnlen(part->name, sizeof(part->name));
+				if (strncmp(part->name, "/", sizeof(part->name)) == 0)
+					goodness++;
+				for (i = 0; i <= l - 4; ++i) {
+					if (strncasecmp(part->name + i, "root",
+						     4) == 0) {
+						goodness += 2;
+						break;
+					}
+				}
+				if (strncasecmp(part->name, "swap", 4) == 0)
+					goodness--;
+			}
+
+			if (goodness > found_root_goodness) {
+				found_root = slot;
+				found_root_goodness = goodness;
+			}
+		}
+#endif /* CONFIG_PPC_PMAC */
+	}
+#ifdef CONFIG_PPC_PMAC
+	if (found_root_goodness)
+		note_bootable_part(state->disk->part0->bd_dev, found_root,
+				   found_root_goodness);
+#endif
+
+	put_dev_sector(sect);
+	seq_buf_puts(&state->pp_buf, "\n");
+	return 1;
+}