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, 84 insertions, 0 deletions

diff --git a/include/uapi/asm-generic/statfs.h b/include/uapi/asm-generic/statfs.h
new file mode 100644
index 000000000..bd35c7619
--- /dev/null
+++ b/include/uapi/asm-generic/statfs.h
@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+#ifndef _UAPI_GENERIC_STATFS_H
+#define _UAPI_GENERIC_STATFS_H
+
+#include <linux/types.h>
+
+
+/*
+ * Most 64-bit platforms use 'long', while most 32-bit platforms use '__u32'.
+ * Yes, they differ in signedness as well as size.
+ * Special cases can override it for themselves -- except for S390x, which
+ * is just a little too special for us. And MIPS, which I'm not touching
+ * with a 10' pole.
+ */
+#ifndef __statfs_word
+#if __BITS_PER_LONG == 64
+#define __statfs_word __kernel_long_t
+#else
+#define __statfs_word __u32
+#endif
+#endif
+
+struct statfs {
+	__statfs_word f_type;
+	__statfs_word f_bsize;
+	__statfs_word f_blocks;
+	__statfs_word f_bfree;
+	__statfs_word f_bavail;
+	__statfs_word f_files;
+	__statfs_word f_ffree;
+	__kernel_fsid_t f_fsid;
+	__statfs_word f_namelen;
+	__statfs_word f_frsize;
+	__statfs_word f_flags;
+	__statfs_word f_spare[4];
+};
+
+/*
+ * ARM needs to avoid the 32-bit padding at the end, for consistency
+ * between EABI and OABI 
+ */
+#ifndef ARCH_PACK_STATFS64
+#define ARCH_PACK_STATFS64
+#endif
+
+struct statfs64 {
+	__statfs_word f_type;
+	__statfs_word f_bsize;
+	__u64 f_blocks;
+	__u64 f_bfree;
+	__u64 f_bavail;
+	__u64 f_files;
+	__u64 f_ffree;
+	__kernel_fsid_t f_fsid;
+	__statfs_word f_namelen;
+	__statfs_word f_frsize;
+	__statfs_word f_flags;
+	__statfs_word f_spare[4];
+} ARCH_PACK_STATFS64;
+
+/* 
+ * IA64 and x86_64 need to avoid the 32-bit padding at the end,
+ * to be compatible with the i386 ABI
+ */
+#ifndef ARCH_PACK_COMPAT_STATFS64
+#define ARCH_PACK_COMPAT_STATFS64
+#endif
+
+struct compat_statfs64 {
+	__u32 f_type;
+	__u32 f_bsize;
+	__u64 f_blocks;
+	__u64 f_bfree;
+	__u64 f_bavail;
+	__u64 f_files;
+	__u64 f_ffree;
+	__kernel_fsid_t f_fsid;
+	__u32 f_namelen;
+	__u32 f_frsize;
+	__u32 f_flags;
+	__u32 f_spare[4];
+} ARCH_PACK_COMPAT_STATFS64;
+
+#endif /* _UAPI_GENERIC_STATFS_H */