From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug 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 Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- drivers/md/dm-sysfs.c | 145 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 145 insertions(+) create mode 100644 drivers/md/dm-sysfs.c (limited to 'drivers/md/dm-sysfs.c') diff --git a/drivers/md/dm-sysfs.c b/drivers/md/dm-sysfs.c new file mode 100644 index 000000000..22bc70923 --- /dev/null +++ b/drivers/md/dm-sysfs.c @@ -0,0 +1,145 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2008 Red Hat, Inc. All rights reserved. + * + * This file is released under the GPL. + */ + +#include +#include +#include "dm-core.h" +#include "dm-rq.h" + +struct dm_sysfs_attr { + struct attribute attr; + ssize_t (*show)(struct mapped_device *md, char *p); + ssize_t (*store)(struct mapped_device *md, const char *p, size_t count); +}; + +#define DM_ATTR_RO(_name) \ +struct dm_sysfs_attr dm_attr_##_name = \ + __ATTR(_name, 0444, dm_attr_##_name##_show, NULL) + +static ssize_t dm_attr_show(struct kobject *kobj, struct attribute *attr, + char *page) +{ + struct dm_sysfs_attr *dm_attr; + struct mapped_device *md; + ssize_t ret; + + dm_attr = container_of(attr, struct dm_sysfs_attr, attr); + if (!dm_attr->show) + return -EIO; + + md = dm_get_from_kobject(kobj); + if (!md) + return -EINVAL; + + ret = dm_attr->show(md, page); + dm_put(md); + + return ret; +} + +#define DM_ATTR_RW(_name) \ +struct dm_sysfs_attr dm_attr_##_name = \ + __ATTR(_name, 0644, dm_attr_##_name##_show, dm_attr_##_name##_store) + +static ssize_t dm_attr_store(struct kobject *kobj, struct attribute *attr, + const char *page, size_t count) +{ + struct dm_sysfs_attr *dm_attr; + struct mapped_device *md; + ssize_t ret; + + dm_attr = container_of(attr, struct dm_sysfs_attr, attr); + if (!dm_attr->store) + return -EIO; + + md = dm_get_from_kobject(kobj); + if (!md) + return -EINVAL; + + ret = dm_attr->store(md, page, count); + dm_put(md); + + return ret; +} + +static ssize_t dm_attr_name_show(struct mapped_device *md, char *buf) +{ + if (dm_copy_name_and_uuid(md, buf, NULL)) + return -EIO; + + strcat(buf, "\n"); + return strlen(buf); +} + +static ssize_t dm_attr_uuid_show(struct mapped_device *md, char *buf) +{ + if (dm_copy_name_and_uuid(md, NULL, buf)) + return -EIO; + + strcat(buf, "\n"); + return strlen(buf); +} + +static ssize_t dm_attr_suspended_show(struct mapped_device *md, char *buf) +{ + return sysfs_emit(buf, "%d\n", dm_suspended_md(md)); +} + +static ssize_t dm_attr_use_blk_mq_show(struct mapped_device *md, char *buf) +{ + /* Purely for userspace compatibility */ + return sysfs_emit(buf, "%d\n", true); +} + +static DM_ATTR_RO(name); +static DM_ATTR_RO(uuid); +static DM_ATTR_RO(suspended); +static DM_ATTR_RO(use_blk_mq); +static DM_ATTR_RW(rq_based_seq_io_merge_deadline); + +static struct attribute *dm_attrs[] = { + &dm_attr_name.attr, + &dm_attr_uuid.attr, + &dm_attr_suspended.attr, + &dm_attr_use_blk_mq.attr, + &dm_attr_rq_based_seq_io_merge_deadline.attr, + NULL, +}; +ATTRIBUTE_GROUPS(dm); + +static const struct sysfs_ops dm_sysfs_ops = { + .show = dm_attr_show, + .store = dm_attr_store, +}; + +static const struct kobj_type dm_ktype = { + .sysfs_ops = &dm_sysfs_ops, + .default_groups = dm_groups, + .release = dm_kobject_release, +}; + +/* + * Initialize kobj + * because nobody using md yet, no need to call explicit dm_get/put + */ +int dm_sysfs_init(struct mapped_device *md) +{ + return kobject_init_and_add(dm_kobject(md), &dm_ktype, + &disk_to_dev(dm_disk(md))->kobj, + "%s", "dm"); +} + +/* + * Remove kobj, called after all references removed + */ +void dm_sysfs_exit(struct mapped_device *md) +{ + struct kobject *kobj = dm_kobject(md); + + kobject_put(kobj); + wait_for_completion(dm_get_completion_from_kobject(kobj)); +} -- cgit v1.3.1