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 --- block/blk-crypto-sysfs.c | 208 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 208 insertions(+) create mode 100644 block/blk-crypto-sysfs.c (limited to 'block/blk-crypto-sysfs.c') diff --git a/block/blk-crypto-sysfs.c b/block/blk-crypto-sysfs.c new file mode 100644 index 000000000..b069c418b --- /dev/null +++ b/block/blk-crypto-sysfs.c @@ -0,0 +1,208 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright 2021 Google LLC + * + * sysfs support for blk-crypto. This file contains the code which exports the + * crypto capabilities of devices via /sys/block/$disk/queue/crypto/. + */ + +#include + +#include "blk-crypto-internal.h" + +struct blk_crypto_kobj { + struct kobject kobj; + struct blk_crypto_profile *profile; +}; + +struct blk_crypto_attr { + struct attribute attr; + ssize_t (*show)(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page); +}; + +static struct blk_crypto_profile *kobj_to_crypto_profile(struct kobject *kobj) +{ + return container_of(kobj, struct blk_crypto_kobj, kobj)->profile; +} + +static const struct blk_crypto_attr *attr_to_crypto_attr(const struct attribute *attr) +{ + return container_of_const(attr, struct blk_crypto_attr, attr); +} + +static ssize_t hw_wrapped_keys_show(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page) +{ + /* Always show supported, since the file doesn't exist otherwise. */ + return sysfs_emit(page, "supported\n"); +} + +static ssize_t max_dun_bits_show(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page) +{ + return sysfs_emit(page, "%u\n", 8 * profile->max_dun_bytes_supported); +} + +static ssize_t num_keyslots_show(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page) +{ + return sysfs_emit(page, "%u\n", profile->num_slots); +} + +static ssize_t raw_keys_show(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page) +{ + /* Always show supported, since the file doesn't exist otherwise. */ + return sysfs_emit(page, "supported\n"); +} + +#define BLK_CRYPTO_RO_ATTR(_name) \ + static const struct blk_crypto_attr _name##_attr = __ATTR_RO(_name) + +BLK_CRYPTO_RO_ATTR(hw_wrapped_keys); +BLK_CRYPTO_RO_ATTR(max_dun_bits); +BLK_CRYPTO_RO_ATTR(num_keyslots); +BLK_CRYPTO_RO_ATTR(raw_keys); + +static umode_t blk_crypto_is_visible(struct kobject *kobj, + const struct attribute *attr, int n) +{ + struct blk_crypto_profile *profile = kobj_to_crypto_profile(kobj); + const struct blk_crypto_attr *a = attr_to_crypto_attr(attr); + + if (a == &hw_wrapped_keys_attr && + !(profile->key_types_supported & BLK_CRYPTO_KEY_TYPE_HW_WRAPPED)) + return 0; + if (a == &raw_keys_attr && + !(profile->key_types_supported & BLK_CRYPTO_KEY_TYPE_RAW)) + return 0; + + return 0444; +} + +static const struct attribute *const blk_crypto_attrs[] = { + &hw_wrapped_keys_attr.attr, + &max_dun_bits_attr.attr, + &num_keyslots_attr.attr, + &raw_keys_attr.attr, + NULL, +}; + +static const struct attribute_group blk_crypto_attr_group = { + .attrs_const = blk_crypto_attrs, + .is_visible_const = blk_crypto_is_visible, +}; + +/* + * The encryption mode attributes. To avoid hard-coding the list of encryption + * modes, these are initialized at boot time by blk_crypto_sysfs_init(). + */ +static struct blk_crypto_attr __blk_crypto_mode_attrs[BLK_ENCRYPTION_MODE_MAX]; +static const struct attribute *blk_crypto_mode_attrs[BLK_ENCRYPTION_MODE_MAX + 1]; + +static umode_t blk_crypto_mode_is_visible(struct kobject *kobj, + const struct attribute *attr, int n) +{ + struct blk_crypto_profile *profile = kobj_to_crypto_profile(kobj); + const struct blk_crypto_attr *a = attr_to_crypto_attr(attr); + int mode_num = a - __blk_crypto_mode_attrs; + + if (profile->modes_supported[mode_num]) + return 0444; + return 0; +} + +static ssize_t blk_crypto_mode_show(struct blk_crypto_profile *profile, + const struct blk_crypto_attr *attr, char *page) +{ + int mode_num = attr - __blk_crypto_mode_attrs; + + return sysfs_emit(page, "0x%x\n", profile->modes_supported[mode_num]); +} + +static const struct attribute_group blk_crypto_modes_attr_group = { + .name = "modes", + .attrs_const = blk_crypto_mode_attrs, + .is_visible_const = blk_crypto_mode_is_visible, +}; + +static const struct attribute_group *blk_crypto_attr_groups[] = { + &blk_crypto_attr_group, + &blk_crypto_modes_attr_group, + NULL, +}; + +static ssize_t blk_crypto_attr_show(struct kobject *kobj, + struct attribute *attr, char *page) +{ + struct blk_crypto_profile *profile = kobj_to_crypto_profile(kobj); + const struct blk_crypto_attr *a = attr_to_crypto_attr(attr); + + return a->show(profile, a, page); +} + +static const struct sysfs_ops blk_crypto_attr_ops = { + .show = blk_crypto_attr_show, +}; + +static void blk_crypto_release(struct kobject *kobj) +{ + kfree(container_of(kobj, struct blk_crypto_kobj, kobj)); +} + +static const struct kobj_type blk_crypto_ktype = { + .default_groups = blk_crypto_attr_groups, + .sysfs_ops = &blk_crypto_attr_ops, + .release = blk_crypto_release, +}; + +/* + * If the request_queue has a blk_crypto_profile, create the "crypto" + * subdirectory in sysfs (/sys/block/$disk/queue/crypto/). + */ +int blk_crypto_sysfs_register(struct gendisk *disk) +{ + struct request_queue *q = disk->queue; + struct blk_crypto_kobj *obj; + int err; + + if (!q->crypto_profile) + return 0; + + obj = kzalloc_obj(*obj); + if (!obj) + return -ENOMEM; + obj->profile = q->crypto_profile; + + err = kobject_init_and_add(&obj->kobj, &blk_crypto_ktype, + &disk->queue_kobj, "crypto"); + if (err) { + kobject_put(&obj->kobj); + return err; + } + q->crypto_kobject = &obj->kobj; + return 0; +} + +void blk_crypto_sysfs_unregister(struct gendisk *disk) +{ + kobject_put(disk->queue->crypto_kobject); +} + +static int __init blk_crypto_sysfs_init(void) +{ + int i; + + BUILD_BUG_ON(BLK_ENCRYPTION_MODE_INVALID != 0); + for (i = 1; i < BLK_ENCRYPTION_MODE_MAX; i++) { + struct blk_crypto_attr *attr = &__blk_crypto_mode_attrs[i]; + + attr->attr.name = blk_crypto_modes[i].name; + attr->attr.mode = 0444; + attr->show = blk_crypto_mode_show; + blk_crypto_mode_attrs[i - 1] = &attr->attr; + } + return 0; +} +subsys_initcall(blk_crypto_sysfs_init); -- cgit v1.3.1