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, 191 insertions, 0 deletions

diff --git a/drivers/mtd/ubi/misc.c b/drivers/mtd/ubi/misc.c
new file mode 100644
index 000000000..1794d66b6
--- /dev/null
+++ b/drivers/mtd/ubi/misc.c
@@ -0,0 +1,191 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) International Business Machines Corp., 2006
+ *
+ * Author: Artem Bityutskiy (Битюцкий Артём)
+ */
+
+/* Here we keep miscellaneous functions which are used all over the UBI code */
+
+#include "ubi.h"
+
+/**
+ * ubi_calc_data_len - calculate how much real data is stored in a buffer.
+ * @ubi: UBI device description object
+ * @buf: a buffer with the contents of the physical eraseblock
+ * @length: the buffer length
+ *
+ * This function calculates how much "real data" is stored in @buf and returnes
+ * the length. Continuous 0xFF bytes at the end of the buffer are not
+ * considered as "real data".
+ */
+int ubi_calc_data_len(const struct ubi_device *ubi, const void *buf,
+		      int length)
+{
+	int i;
+
+	ubi_assert(!(length & (ubi->min_io_size - 1)));
+
+	for (i = length - 1; i >= 0; i--)
+		if (((const uint8_t *)buf)[i] != 0xFF)
+			break;
+
+	/* The resulting length must be aligned to the minimum flash I/O size */
+	length = ALIGN(i + 1, ubi->min_io_size);
+	return length;
+}
+
+/**
+ * ubi_check_volume - check the contents of a static volume.
+ * @ubi: UBI device description object
+ * @vol_id: ID of the volume to check
+ *
+ * This function checks if static volume @vol_id is corrupted by fully reading
+ * it and checking data CRC. This function returns %0 if the volume is not
+ * corrupted, %1 if it is corrupted and a negative error code in case of
+ * failure. Dynamic volumes are not checked and zero is returned immediately.
+ */
+int ubi_check_volume(struct ubi_device *ubi, int vol_id)
+{
+	void *buf;
+	int err = 0, i;
+	struct ubi_volume *vol = ubi->volumes[vol_id];
+
+	if (vol->vol_type != UBI_STATIC_VOLUME)
+		return 0;
+
+	buf = vmalloc(vol->usable_leb_size);
+	if (!buf)
+		return -ENOMEM;
+
+	for (i = 0; i < vol->used_ebs; i++) {
+		int size;
+
+		cond_resched();
+
+		if (i == vol->used_ebs - 1)
+			size = vol->last_eb_bytes;
+		else
+			size = vol->usable_leb_size;
+
+		err = ubi_eba_read_leb(ubi, vol, i, buf, 0, size, 1);
+		if (err) {
+			if (mtd_is_eccerr(err))
+				err = 1;
+			break;
+		}
+	}
+
+	vfree(buf);
+	return err;
+}
+
+/**
+ * ubi_update_reserved - update bad eraseblock handling accounting data.
+ * @ubi: UBI device description object
+ *
+ * This function calculates the gap between current number of PEBs reserved for
+ * bad eraseblock handling and the required level of PEBs that must be
+ * reserved, and if necessary, reserves more PEBs to fill that gap, according
+ * to availability. Should be called with ubi->volumes_lock held.
+ */
+void ubi_update_reserved(struct ubi_device *ubi)
+{
+	int need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
+
+	if (need <= 0 || ubi->avail_pebs == 0)
+		return;
+
+	need = min_t(int, need, ubi->avail_pebs);
+	ubi->avail_pebs -= need;
+	ubi->rsvd_pebs += need;
+	ubi->beb_rsvd_pebs += need;
+	ubi_msg(ubi, "reserved more %d PEBs for bad PEB handling", need);
+}
+
+/**
+ * ubi_calculate_reserved - calculate how many PEBs must be reserved for bad
+ * eraseblock handling.
+ * @ubi: UBI device description object
+ */
+void ubi_calculate_reserved(struct ubi_device *ubi)
+{
+	/*
+	 * Calculate the actual number of PEBs currently needed to be reserved
+	 * for future bad eraseblock handling.
+	 */
+	ubi->beb_rsvd_level = ubi->bad_peb_limit - ubi->bad_peb_count;
+	if (ubi->beb_rsvd_level < 0) {
+		ubi->beb_rsvd_level = 0;
+		ubi_warn(ubi, "number of bad PEBs (%d) is above the expected limit (%d), not reserving any PEBs for bad PEB handling, will use available PEBs (if any)",
+			 ubi->bad_peb_count, ubi->bad_peb_limit);
+	}
+}
+
+/**
+ * ubi_check_pattern - check if buffer contains only a certain byte pattern.
+ * @buf: buffer to check
+ * @patt: the pattern to check
+ * @size: buffer size in bytes
+ *
+ * This function returns %1 in there are only @patt bytes in @buf, and %0 if
+ * something else was also found.
+ */
+int ubi_check_pattern(const void *buf, uint8_t patt, int size)
+{
+	int i;
+
+	for (i = 0; i < size; i++)
+		if (((const uint8_t *)buf)[i] != patt)
+			return 0;
+	return 1;
+}
+
+/* Normal UBI messages */
+void ubi_msg(const struct ubi_device *ubi, const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+
+	va_start(args, fmt);
+
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	pr_notice(UBI_NAME_STR "%d: %pV\n", ubi->ubi_num, &vaf);
+
+	va_end(args);
+}
+
+/* UBI warning messages */
+void ubi_warn(const struct ubi_device *ubi, const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+
+	va_start(args, fmt);
+
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	pr_warn(UBI_NAME_STR "%d warning: %ps: %pV\n",
+		ubi->ubi_num, __builtin_return_address(0), &vaf);
+
+	va_end(args);
+}
+
+/* UBI error messages */
+void ubi_err(const struct ubi_device *ubi, const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+
+	va_start(args, fmt);
+
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	pr_err(UBI_NAME_STR "%d error: %ps: %pV\n",
+	       ubi->ubi_num, __builtin_return_address(0), &vaf);
+	va_end(args);
+}