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

3 files changed, 126 insertions, 0 deletions

diff --git a/tools/testing/selftests/filelock/.gitignore b/tools/testing/selftests/filelock/.gitignore
new file mode 100644
index 000000000..825e899a1
--- /dev/null
+++ b/tools/testing/selftests/filelock/.gitignore
@@ -0,0 +1 @@
+ofdlocks
diff --git a/tools/testing/selftests/filelock/Makefile b/tools/testing/selftests/filelock/Makefile
new file mode 100644
index 000000000..478e82f8b
--- /dev/null
+++ b/tools/testing/selftests/filelock/Makefile
@@ -0,0 +1,5 @@
+# SPDX-License-Identifier: GPL-2.0
+
+TEST_GEN_PROGS := ofdlocks
+
+include ../lib.mk
diff --git a/tools/testing/selftests/filelock/ofdlocks.c b/tools/testing/selftests/filelock/ofdlocks.c
new file mode 100644
index 000000000..68bac28b2
--- /dev/null
+++ b/tools/testing/selftests/filelock/ofdlocks.c
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <assert.h>
+#include <stdio.h>
+#include <unistd.h>
+#include <string.h>
+#include "kselftest.h"
+
+static int lock_set(int fd, struct flock *fl)
+{
+	int ret;
+
+	fl->l_pid = 0;		// needed for OFD locks
+	fl->l_whence = SEEK_SET;
+	ret = fcntl(fd, F_OFD_SETLK, fl);
+	if (ret)
+		ksft_perror("fcntl()");
+	return ret;
+}
+
+static int lock_get(int fd, struct flock *fl)
+{
+	int ret;
+
+	fl->l_pid = 0;		// needed for OFD locks
+	fl->l_whence = SEEK_SET;
+	ret = fcntl(fd, F_OFD_GETLK, fl);
+	if (ret)
+		ksft_perror("fcntl()");
+	return ret;
+}
+
+int main(void)
+{
+	int rc;
+	struct flock fl, fl2;
+	int fd = open("/tmp/aa", O_RDWR | O_CREAT | O_EXCL, 0600);
+	int fd2 = open("/tmp/aa", O_RDONLY);
+
+	ksft_print_header();
+	ksft_set_plan(4);
+
+	unlink("/tmp/aa");
+	assert(fd != -1);
+	assert(fd2 != -1);
+	ksft_print_msg("opened fds %i %i\n", fd, fd2);
+
+	/* Set some read lock */
+	fl.l_type = F_RDLCK;
+	fl.l_start = 5;
+	fl.l_len = 3;
+	rc = lock_set(fd, &fl);
+	ksft_test_result(rc == 0, "set OFD read lock on first fd\n");
+	if (rc != 0)
+		ksft_finished();
+
+	/* Make sure read locks do not conflict on different fds. */
+	fl.l_type = F_RDLCK;
+	fl.l_start = 5;
+	fl.l_len = 1;
+	rc = lock_get(fd2, &fl);
+	if (rc != 0)
+		ksft_finished();
+	if (fl.l_type != F_UNLCK)
+		ksft_exit_fail_msg("read locks conflicted\n");
+
+	/* Make sure read/write locks do conflict on different fds. */
+	fl.l_type = F_WRLCK;
+	fl.l_start = 5;
+	fl.l_len = 1;
+	rc = lock_get(fd2, &fl);
+	if (rc != 0)
+		ksft_finished();
+	ksft_test_result(fl.l_type != F_UNLCK,
+			 "read and write locks conflicted\n");
+	if (fl.l_type == F_UNLCK)
+		ksft_finished();
+
+	/* Get info about the lock on first fd. */
+	fl.l_type = F_UNLCK;
+	fl.l_start = 5;
+	fl.l_len = 1;
+	rc = lock_get(fd, &fl);
+	if (rc != 0)
+		ksft_exit_fail_msg("F_OFD_GETLK with F_UNLCK not supported\n");
+	ksft_test_result(fl.l_type != F_UNLCK,
+			 "F_OFD_GETLK with F_UNLCK returned lock info\n");
+	if (fl.l_type == F_UNLCK)
+		ksft_exit_fail();
+	ksft_print_msg("F_UNLCK test returns: locked, type %i pid %i len %zi\n",
+		       fl.l_type, fl.l_pid, fl.l_len);
+
+	/* Try the same but by locking everything by len==0. */
+	fl2.l_type = F_UNLCK;
+	fl2.l_start = 0;
+	fl2.l_len = 0;
+	rc = lock_get(fd, &fl2);
+	if (rc != 0)
+		ksft_exit_fail_msg
+		    ("F_OFD_GETLK with F_UNLCK not supported\n");
+	ksft_test_result(memcmp(&fl, &fl2, sizeof(fl)) == 0,
+			 "F_UNLCK with len==0 returned the same\n");
+	if (memcmp(&fl, &fl2, sizeof(fl))) {
+		ksft_exit_fail_msg
+		    ("F_UNLCK test returns: locked, type %i pid %i len %zi\n",
+		     fl.l_type, fl.l_pid, fl.l_len);
+	}
+
+	/* Get info about the lock on second fd - no locks on it. */
+	fl.l_type = F_UNLCK;
+	fl.l_start = 0;
+	fl.l_len = 0;
+	lock_get(fd2, &fl);
+	ksft_test_result(fl.l_type == F_UNLCK,
+			 "F_OFD_GETLK with F_UNLCK return lock info from another fd\n");
+
+	ksft_finished();
+}