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, 96 insertions, 0 deletions

diff --git a/drivers/md/dm-vdo/status-codes.c b/drivers/md/dm-vdo/status-codes.c
new file mode 100644
index 000000000..9df5e4d7f
--- /dev/null
+++ b/drivers/md/dm-vdo/status-codes.c
@@ -0,0 +1,96 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2023 Red Hat
+ */
+
+#include "status-codes.h"
+
+#include "errors.h"
+#include "logger.h"
+#include "permassert.h"
+#include "thread-utils.h"
+
+const struct error_info vdo_status_list[] = {
+	{ "VDO_NOT_IMPLEMENTED", "Not implemented" },
+	{ "VDO_OUT_OF_RANGE", "Out of range" },
+	{ "VDO_REF_COUNT_INVALID", "Reference count would become invalid" },
+	{ "VDO_NO_SPACE", "Out of space" },
+	{ "VDO_BAD_CONFIGURATION", "Bad configuration option" },
+	{ "VDO_COMPONENT_BUSY", "Prior operation still in progress" },
+	{ "VDO_BAD_PAGE", "Corrupt or incorrect page" },
+	{ "VDO_UNSUPPORTED_VERSION", "Unsupported component version" },
+	{ "VDO_INCORRECT_COMPONENT", "Component id mismatch in decoder" },
+	{ "VDO_PARAMETER_MISMATCH", "Parameters have conflicting values" },
+	{ "VDO_UNKNOWN_PARTITION", "No partition exists with a given id" },
+	{ "VDO_PARTITION_EXISTS", "A partition already exists with a given id" },
+	{ "VDO_INCREMENT_TOO_SMALL", "Physical block growth of too few blocks" },
+	{ "VDO_CHECKSUM_MISMATCH", "Incorrect checksum" },
+	{ "VDO_LOCK_ERROR", "A lock is held incorrectly" },
+	{ "VDO_READ_ONLY", "The device is in read-only mode" },
+	{ "VDO_SHUTTING_DOWN", "The device is shutting down" },
+	{ "VDO_CORRUPT_JOURNAL", "Recovery journal corrupted" },
+	{ "VDO_TOO_MANY_SLABS", "Exceeds maximum number of slabs supported" },
+	{ "VDO_INVALID_FRAGMENT", "Compressed block fragment is invalid" },
+	{ "VDO_RETRY_AFTER_REBUILD", "Retry operation after rebuilding finishes" },
+	{ "VDO_BAD_MAPPING", "Invalid page mapping" },
+	{ "VDO_BIO_CREATION_FAILED", "Bio creation failed" },
+	{ "VDO_BAD_MAGIC", "Bad magic number" },
+	{ "VDO_BAD_NONCE", "Bad nonce" },
+	{ "VDO_JOURNAL_OVERFLOW", "Journal sequence number overflow" },
+	{ "VDO_INVALID_ADMIN_STATE", "Invalid operation for current state" },
+};
+
+/**
+ * vdo_register_status_codes() - Register the VDO status codes.
+ * Return: A success or error code.
+ */
+int vdo_register_status_codes(void)
+{
+	int result;
+
+	BUILD_BUG_ON((VDO_STATUS_CODE_LAST - VDO_STATUS_CODE_BASE) !=
+		     ARRAY_SIZE(vdo_status_list));
+
+	result = uds_register_error_block("VDO Status", VDO_STATUS_CODE_BASE,
+					  VDO_STATUS_CODE_BLOCK_END, vdo_status_list,
+					  sizeof(vdo_status_list));
+	return (result == UDS_SUCCESS) ? VDO_SUCCESS : result;
+}
+
+/**
+ * vdo_status_to_errno() - Given an error code, return a value we can return to the OS.
+ * @error: The error code to convert.
+ *
+ * The input error code may be a system-generated value (such as -EIO), an errno macro used in our
+ * code (such as EIO), or a UDS or VDO status code; the result must be something the rest of the OS
+ * can consume (negative errno values such as -EIO, in the case of the kernel).
+ *
+ * Return: A system error code value.
+ */
+int vdo_status_to_errno(int error)
+{
+	char error_name[VDO_MAX_ERROR_NAME_SIZE];
+	char error_message[VDO_MAX_ERROR_MESSAGE_SIZE];
+
+	/* 0 is success, negative a system error code */
+	if (likely(error <= 0))
+		return error;
+	if (error < 1024)
+		return -error;
+
+	/* VDO or UDS error */
+	switch (error) {
+	case VDO_BAD_CONFIGURATION:
+		return -EINVAL;
+	case VDO_NO_SPACE:
+		return -ENOSPC;
+	case VDO_READ_ONLY:
+		return -EIO;
+	default:
+		vdo_log_info("%s: mapping internal status code %d (%s: %s) to EIO",
+			     __func__, error,
+			     uds_string_error_name(error, error_name, sizeof(error_name)),
+			     uds_string_error(error, error_message, sizeof(error_message)));
+		return -EIO;
+	}
+}