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

4 files changed, 146 insertions, 0 deletions

diff --git a/scripts/ipe/Makefile b/scripts/ipe/Makefile
new file mode 100644
index 000000000..e87553fbb
--- /dev/null
+++ b/scripts/ipe/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+subdir-y := polgen
diff --git a/scripts/ipe/polgen/.gitignore b/scripts/ipe/polgen/.gitignore
new file mode 100644
index 000000000..b6f05cf3d
--- /dev/null
+++ b/scripts/ipe/polgen/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+polgen
diff --git a/scripts/ipe/polgen/Makefile b/scripts/ipe/polgen/Makefile
new file mode 100644
index 000000000..c20456a2f
--- /dev/null
+++ b/scripts/ipe/polgen/Makefile
@@ -0,0 +1,5 @@
+# SPDX-License-Identifier: GPL-2.0
+hostprogs-always-y	:= polgen
+HOST_EXTRACFLAGS += \
+	-I$(srctree)/include \
+	-I$(srctree)/include/uapi \
diff --git a/scripts/ipe/polgen/polgen.c b/scripts/ipe/polgen/polgen.c
new file mode 100644
index 000000000..01134cf89
--- /dev/null
+++ b/scripts/ipe/polgen/polgen.c
@@ -0,0 +1,137 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020-2024 Microsoft Corporation. All rights reserved.
+ */
+
+#include <stdlib.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <errno.h>
+
+static void usage(const char *const name)
+{
+	printf("Usage: %s OutputFile (PolicyFile)\n", name);
+	exit(EINVAL);
+}
+
+static int policy_to_buffer(const char *pathname, char **buffer, size_t *size)
+{
+	size_t fsize;
+	size_t read;
+	char *lbuf;
+	int rc = 0;
+	FILE *fd;
+
+	fd = fopen(pathname, "r");
+	if (!fd) {
+		rc = errno;
+		goto out;
+	}
+
+	fseek(fd, 0, SEEK_END);
+	fsize = ftell(fd);
+	rewind(fd);
+
+	lbuf = malloc(fsize);
+	if (!lbuf) {
+		rc = ENOMEM;
+		goto out_close;
+	}
+
+	read = fread((void *)lbuf, sizeof(*lbuf), fsize, fd);
+	if (read != fsize) {
+		rc = -1;
+		goto out_free;
+	}
+
+	*buffer = lbuf;
+	*size = fsize;
+	fclose(fd);
+
+	return rc;
+
+out_free:
+	free(lbuf);
+out_close:
+	fclose(fd);
+out:
+	return rc;
+}
+
+static int write_boot_policy(const char *pathname, const char *buf, size_t size)
+{
+	FILE *fd;
+	size_t i;
+
+	fd = fopen(pathname, "w");
+	if (!fd)
+		return errno;
+
+	fprintf(fd, "/* This file is automatically generated.");
+	fprintf(fd, " Do not edit. */\n");
+	fprintf(fd, "#include <linux/stddef.h>\n");
+	fprintf(fd, "\nextern const char *const ipe_boot_policy;\n\n");
+	fprintf(fd, "const char *const ipe_boot_policy =\n");
+
+	if (!buf || size == 0) {
+		fprintf(fd, "\tNULL;\n");
+		fclose(fd);
+		return 0;
+	}
+
+	fprintf(fd, "\t\"");
+
+	for (i = 0; i < size; ++i) {
+		switch (buf[i]) {
+		case '"':
+			fprintf(fd, "\\\"");
+			break;
+		case '\'':
+			fprintf(fd, "'");
+			break;
+		case '\n':
+			fprintf(fd, "\\n\"\n\t\"");
+			break;
+		case '\\':
+			fprintf(fd, "\\\\");
+			break;
+		case '\t':
+			fprintf(fd, "\\t");
+			break;
+		case '\?':
+			fprintf(fd, "\\?");
+			break;
+		default:
+			fprintf(fd, "%c", buf[i]);
+		}
+	}
+	fprintf(fd, "\";\n");
+	fclose(fd);
+
+	return 0;
+}
+
+int main(int argc, const char *const argv[])
+{
+	char *policy = NULL;
+	size_t len = 0;
+	int rc = 0;
+
+	if (argc < 2)
+		usage(argv[0]);
+
+	if (argc > 2) {
+		rc = policy_to_buffer(argv[2], &policy, &len);
+		if (rc != 0)
+			goto cleanup;
+	}
+
+	rc = write_boot_policy(argv[1], policy, len);
+cleanup:
+	if (policy)
+		free(policy);
+	if (rc != 0)
+		perror("An error occurred during policy conversion: ");
+	return rc;
+}