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/samples/pidfd/.gitignore b/samples/pidfd/.gitignore
new file mode 100644
index 000000000..d4cfa3176
--- /dev/null
+++ b/samples/pidfd/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+/pidfd-metadata
diff --git a/samples/pidfd/Makefile b/samples/pidfd/Makefile
new file mode 100644
index 000000000..9754e2d81
--- /dev/null
+++ b/samples/pidfd/Makefile
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0
+usertprogs-always-y += pidfd-metadata
+
+userccflags += -I usr/include
diff --git a/samples/pidfd/pidfd-metadata.c b/samples/pidfd/pidfd-metadata.c
new file mode 100644
index 000000000..c459155da
--- /dev/null
+++ b/samples/pidfd/pidfd-metadata.c
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#define _GNU_SOURCE
+#include <err.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <limits.h>
+#include <sched.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/syscall.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+#ifndef CLONE_PIDFD
+#define CLONE_PIDFD 0x00001000
+#endif
+
+#ifndef __NR_pidfd_send_signal
+#define __NR_pidfd_send_signal -1
+#endif
+
+static int do_child(void *args)
+{
+	printf("%d\n", getpid());
+	_exit(EXIT_SUCCESS);
+}
+
+static pid_t pidfd_clone(int flags, int *pidfd)
+{
+	size_t stack_size = 1024;
+	char *stack[1024] = { 0 };
+
+#ifdef __ia64__
+	return __clone2(do_child, stack, stack_size, flags | SIGCHLD, NULL, pidfd);
+#else
+	return clone(do_child, stack + stack_size, flags | SIGCHLD, NULL, pidfd);
+#endif
+}
+
+static inline int sys_pidfd_send_signal(int pidfd, int sig, siginfo_t *info,
+					unsigned int flags)
+{
+	return syscall(__NR_pidfd_send_signal, pidfd, sig, info, flags);
+}
+
+static int pidfd_metadata_fd(pid_t pid, int pidfd)
+{
+	int procfd, ret;
+	char path[100];
+
+	snprintf(path, sizeof(path), "/proc/%d", pid);
+	procfd = open(path, O_DIRECTORY | O_RDONLY | O_CLOEXEC);
+	if (procfd < 0) {
+		warn("Failed to open %s\n", path);
+		return -1;
+	}
+
+	/*
+	 * Verify that the pid has not been recycled and our /proc/<pid> handle
+	 * is still valid.
+	 */
+	ret = sys_pidfd_send_signal(pidfd, 0, NULL, 0);
+	if (ret < 0) {
+		switch (errno) {
+		case EPERM:
+			/* Process exists, just not allowed to signal it. */
+			break;
+		default:
+			warn("Failed to signal process\n");
+			close(procfd);
+			procfd = -1;
+		}
+	}
+
+	return procfd;
+}
+
+int main(int argc, char *argv[])
+{
+	int pidfd = -1, ret = EXIT_FAILURE;
+	char buf[4096] = { 0 };
+	pid_t pid;
+	int procfd, statusfd;
+	ssize_t bytes;
+
+	pid = pidfd_clone(CLONE_PIDFD, &pidfd);
+	if (pid < 0)
+		err(ret, "CLONE_PIDFD");
+	if (pidfd == -1) {
+		warnx("CLONE_PIDFD is not supported by the kernel");
+		goto out;
+	}
+
+	procfd = pidfd_metadata_fd(pid, pidfd);
+	close(pidfd);
+	if (procfd < 0)
+		goto out;
+
+	statusfd = openat(procfd, "status", O_RDONLY | O_CLOEXEC);
+	close(procfd);
+	if (statusfd < 0)
+		goto out;
+
+	bytes = read(statusfd, buf, sizeof(buf));
+	if (bytes > 0)
+		bytes = write(STDOUT_FILENO, buf, bytes);
+	close(statusfd);
+	ret = EXIT_SUCCESS;
+
+out:
+	(void)wait(NULL);
+
+	exit(ret);
+}