- Kernel - mailweb.openeuler.org

[PATCH] sched: Introduce priority load balance for qos scheduler
by Song Zhang 18 Dec '23

18 Dec '23

euleros inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8MWDD CVE: NA Add new sysctl interface: /proc/sys/kernel/sched_prio_load_balance_enabled 0: default behavior 1: enable priority load balance for qos scheduler For tasks co-location with qos scheduler, when CFS do load balance, it is reasonable to prefer migrating online(Latency Sensitive) tasks. So the CFS load balance can be changed to below: cfs_tasks list is owned by online tasks. Add new cfs_offline_tasks list which is owned by offline tasks. Prefer to migrate the online tasks of cfs_tasks list to dst rq. In the scenario of hyperthread interference, if the smt expeller feature enabled, CPU A and CPU B are two hyperthreads on a physical core, CPU A runs online tasks while CPU B only has offline tasks, The offline tasks on CPU B are expelled by the online tasks on CPU A and cannot be scheduled. However, when load balance is triggered, before CPU B can migrate some online tasks from CPU A, the load on the two cpus is already balanced. As a result, CPU B cannot run online tasks and online tasks cannot be evenly distributed among different cpus. Signed-off-by: Song Zhang <zhangsong34(a)huawei.com> --- arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + include/linux/sched/sysctl.h | 4 ++ init/Kconfig | 9 +++ kernel/sched/core.c | 3 + kernel/sched/fair.c | 90 +++++++++++++++++++++++++- kernel/sched/sched.h | 3 + 7 files changed, 110 insertions(+), 1 deletion(-) diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index 039f3496af..296421c0d5 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -162,6 +162,7 @@ CONFIG_BLK_CGROUP=y CONFIG_CGROUP_WRITEBACK=y CONFIG_CGROUP_SCHED=y CONFIG_QOS_SCHED=y +CONFIG_QOS_SCHED_PRIO_LB=y CONFIG_FAIR_GROUP_SCHED=y CONFIG_CFS_BANDWIDTH=y CONFIG_RT_GROUP_SCHED=y diff --git a/arch/x86/configs/openeuler_defconfig b/arch/x86/configs/openeuler_defconfig index f481c1c8b4..efb9a1efa7 100644 --- a/arch/x86/configs/openeuler_defconfig +++ b/arch/x86/configs/openeuler_defconfig @@ -184,6 +184,7 @@ CONFIG_BLK_CGROUP=y CONFIG_CGROUP_WRITEBACK=y CONFIG_CGROUP_SCHED=y CONFIG_QOS_SCHED=y +CONFIG_QOS_SCHED_PRIO_LB=y CONFIG_FAIR_GROUP_SCHED=y CONFIG_CFS_BANDWIDTH=y CONFIG_RT_GROUP_SCHED=y diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h index 5a64582b08..efee7213a0 100644 --- a/include/linux/sched/sysctl.h +++ b/include/linux/sched/sysctl.h @@ -29,4 +29,8 @@ extern int sysctl_numa_balancing_mode; #define sysctl_numa_balancing_mode 0 #endif +#ifdef CONFIG_QOS_SCHED_PRIO_LB +extern unsigned int sysctl_sched_prio_load_balance_enabled; +#endif + #endif /* _LINUX_SCHED_SYSCTL_H */ diff --git a/init/Kconfig b/init/Kconfig index b6952df34e..c177b7591a 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -997,6 +997,15 @@ config QOS_SCHED If in doubt, say N. +config QOS_SCHED_PRIO_LB + bool "Priority load balance for Qos scheduler" + depends on QOS_SCHED + default n + help + This feature enable priority load balance + for Qos scheduler, which prefer migrating online tasks + and migrating offline tasks secondly between CPUs. + config FAIR_GROUP_SCHED bool "Group scheduling for SCHED_OTHER" depends on CGROUP_SCHED diff --git a/kernel/sched/core.c b/kernel/sched/core.c index a1c73dea1f..d68ddb4d28 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -10045,6 +10045,9 @@ void __init sched_init(void) rq->max_idle_balance_cost = sysctl_sched_migration_cost; INIT_LIST_HEAD(&rq->cfs_tasks); +#ifdef CONFIG_QOS_SCHED_PRIO_LB + INIT_LIST_HEAD(&rq->cfs_offline_tasks); +#endif rq_attach_root(rq, &def_root_domain); #ifdef CONFIG_NO_HZ_COMMON diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 8ae0d65713..3b35448448 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -147,6 +147,10 @@ static int hundred_thousand = 100000; static int unthrottle_qos_cfs_rqs(int cpu); #endif +#ifdef CONFIG_QOS_SCHED_PRIO_LB +unsigned int sysctl_sched_prio_load_balance_enabled; +#endif + #ifdef CONFIG_CFS_BANDWIDTH /* * Amount of runtime to allocate from global (tg) to local (per-cfs_rq) pool @@ -214,6 +218,17 @@ static struct ctl_table sched_fair_sysctls[] = { .extra1 = SYSCTL_ONE_HUNDRED, .extra2 = &one_thousand, }, +#endif +#ifdef CONFIG_QOS_SCHED_PRIO_LB + { + .procname = "sched_prio_load_balance_enabled", + .data = &sysctl_sched_prio_load_balance_enabled, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = proc_dointvec_minmax, + .extra1 = SYSCTL_ZERO, + .extra2 = SYSCTL_ONE, + }, #endif {} }; @@ -3565,6 +3580,21 @@ static inline void update_scan_period(struct task_struct *p, int new_cpu) #endif /* CONFIG_NUMA_BALANCING */ +#ifdef CONFIG_QOS_SCHED_PRIO_LB +static void +adjust_rq_cfs_tasks(void (*list_op)(struct list_head *, struct list_head *), + struct rq *rq, + struct sched_entity *se) +{ + struct task_group *tg = task_group(task_of(se)); + + if (sysctl_sched_prio_load_balance_enabled && tg->qos_level == -1) + (*list_op)(&se->group_node, &rq->cfs_offline_tasks); + else + (*list_op)(&se->group_node, &rq->cfs_tasks); +} +#endif + static void account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) { @@ -3574,7 +3604,11 @@ account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se) struct rq *rq = rq_of(cfs_rq); account_numa_enqueue(rq, task_of(se)); +#ifdef CONFIG_QOS_SCHED_PRIO_LB + adjust_rq_cfs_tasks(list_add, rq, se); +#else list_add(&se->group_node, &rq->cfs_tasks); +#endif } #endif cfs_rq->nr_running++; @@ -8611,7 +8645,11 @@ done: __maybe_unused; * the list, so our cfs_tasks list becomes MRU * one. */ +#ifdef CONFIG_QOS_SCHED_PRIO_LB + adjust_rq_cfs_tasks(list_move, rq, &p->se); +#else list_move(&p->se.group_node, &rq->cfs_tasks); +#endif #endif if (hrtick_enabled_fair(rq)) @@ -8959,6 +8997,14 @@ static int task_hot(struct task_struct *p, struct lb_env *env) (&p->se == cfs_rq_of(&p->se)->next)) return 1; +#ifdef CONFIG_QOS_SCHED_PRIO_LB + /* Preempt sched idle cpu do not consider migration cost */ + if (sysctl_sched_prio_load_balance_enabled && + cpus_share_cache(env->src_cpu, env->dst_cpu) && + sched_idle_cpu(env->dst_cpu)) + return 0; +#endif + if (sysctl_sched_migration_cost == -1) return 1; @@ -9149,11 +9195,18 @@ static void detach_task(struct task_struct *p, struct lb_env *env) static struct task_struct *detach_one_task(struct lb_env *env) { struct task_struct *p; + struct list_head *tasks = &env->src_rq->cfs_tasks; +#ifdef CONFIG_QOS_SCHED_PRIO_LB + int loop = 0; +#endif lockdep_assert_rq_held(env->src_rq); +#ifdef CONFIG_QOS_SCHED_PRIO_LB +again: +#endif list_for_each_entry_reverse(p, - &env->src_rq->cfs_tasks, se.group_node) { + tasks, se.group_node) { if (!can_migrate_task(p, env)) continue; @@ -9168,6 +9221,15 @@ static struct task_struct *detach_one_task(struct lb_env *env) schedstat_inc(env->sd->lb_gained[env->idle]); return p; } +#ifdef CONFIG_QOS_SCHED_PRIO_LB + if (sysctl_sched_prio_load_balance_enabled) { + loop++; + if (loop == 1) { + tasks = &env->src_rq->cfs_offline_tasks; + goto again; + } + } +#endif return NULL; } @@ -9183,6 +9245,9 @@ static int detach_tasks(struct lb_env *env) unsigned long util, load; struct task_struct *p; int detached = 0; +#ifdef CONFIG_QOS_SCHED_PRIO_LB + int loop = 0; +#endif lockdep_assert_rq_held(env->src_rq); @@ -9198,6 +9263,9 @@ static int detach_tasks(struct lb_env *env) if (env->imbalance <= 0) return 0; +#ifdef CONFIG_QOS_SCHED_PRIO_LB +again: +#endif while (!list_empty(tasks)) { /* * We don't want to steal all, otherwise we may be treated likewise, @@ -9303,6 +9371,22 @@ static int detach_tasks(struct lb_env *env) list_move(&p->se.group_node, tasks); } +#ifdef CONFIG_QOS_SCHED_PRIO_LB + if (sysctl_sched_prio_load_balance_enabled && env->imbalance > 0) { + /* + * Avoid offline tasks starve to death if env->loop exceed + * env->loop_max, so we should set env->loop to 0 and detach + * offline tasks again. + */ + if (env->loop > env->loop_max) + env->loop = 0; + loop++; + if (loop == 1) { + tasks = &env->src_rq->cfs_offline_tasks; + goto again; + } + } +#endif /* * Right now, this is one of only two places we collect this stat * so we can safely collect detach_one_task() stats here rather @@ -12920,7 +13004,11 @@ static void set_next_task_fair(struct rq *rq, struct task_struct *p, bool first) * Move the next running task to the front of the list, so our * cfs_tasks list becomes MRU one. */ +#ifdef CONFIG_QOS_SCHED_PRIO_LB + adjust_rq_cfs_tasks(list_move, rq, se); +#else list_move(&se->group_node, &rq->cfs_tasks); +#endif } #endif diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 3de84e95ba..76951226b6 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1075,6 +1075,9 @@ struct rq { int online; struct list_head cfs_tasks; +#ifdef CONFIG_QOS_SCHED_PRIO_LB + struct list_head cfs_offline_tasks; +#endif struct sched_avg avg_rt; struct sched_avg avg_dl; -- 2.33.0

1 0

[PATCH OLK-6.6 0/2] Support iocost for cgroup v1
by Li Nan 18 Dec '23

18 Dec '23

Li Nan (2): cgroup: Support iocost for cgroup v1 iocost: add config option for iocost in cgroup v1 block/blk-iocost.c | 27 +++++++++++++++++++++++++++ block/Kconfig | 8 ++++++++ 2 files changed, 35 insertions(+) -- 2.39.2

2 3

[PATCH OLK-6.6 0/4] Backport nbd bugfix patch
by Li Nan 18 Dec '23

18 Dec '23

Li Nan (4): nbd: fold nbd config initialization into nbd_alloc_config() nbd: factor out a helper to get nbd_config without holding 'config_lock' nbd: fix null-ptr-dereference while accessing 'nbd->config' nbd: pass nbd_sock to nbd_read_reply() instead of index drivers/block/nbd.c | 117 ++++++++++++++++++++++++++++---------------- 1 file changed, 75 insertions(+), 42 deletions(-) -- 2.39.2

2 5

[PATCH openEuler-1.0-LTS v2 0/2] Fix data-races around
by Zhengchao Shao 18 Dec '23

18 Dec '23

Fix data-races around sysctl_ip_fwd_update_priority and sysctl_ip_no_pmtu_disc. Kuniyuki Iwashima (2): ip: Fix data-races around sysctl_ip_fwd_update_priority. ip: Fix data-races around sysctl_ip_no_pmtu_disc. drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 3 ++- net/ipv4/af_inet.c | 2 +- net/ipv4/icmp.c | 2 +- net/ipv4/ip_forward.c | 2 +- net/ipv6/af_inet6.c | 2 +- 5 files changed, 6 insertions(+), 5 deletions(-) -- 2.34.1

2 3

[PATCH OLK-6.6 v3] ascend: export interfaces required by ascend drivers
by Yuan Can 18 Dec '23

18 Dec '23

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OX3Q CVE: NA -------------------------------- Export oom_type_notifier_call and __get_vm_area_caller. Export pm_autosleep_set_state and __vmalloc_node_range. Export alloc_workqueue_attrs, free_workqueue_attrs and apply_workqueue_attrs. Export find_get_task_by_vpid. Signed-off-by: Yuan Can <yuancan(a)huawei.com> --- kernel/pid.c | 1 + kernel/power/autosleep.c | 1 + kernel/workqueue.c | 3 +++ mm/oom_kill.c | 1 + mm/vmalloc.c | 2 ++ 5 files changed, 8 insertions(+) diff --git a/kernel/pid.c b/kernel/pid.c index 6500ef956f2f..383abde0c208 100644 --- a/kernel/pid.c +++ b/kernel/pid.c @@ -437,6 +437,7 @@ struct task_struct *find_get_task_by_vpid(pid_t nr) return task; } +EXPORT_SYMBOL_GPL(find_get_task_by_vpid); struct pid *get_task_pid(struct task_struct *task, enum pid_type type) { diff --git a/kernel/power/autosleep.c b/kernel/power/autosleep.c index b29c8aca7486..80ba474daa40 100644 --- a/kernel/power/autosleep.c +++ b/kernel/power/autosleep.c @@ -113,6 +113,7 @@ int pm_autosleep_set_state(suspend_state_t state) mutex_unlock(&autosleep_lock); return 0; } +EXPORT_SYMBOL_GPL(pm_autosleep_set_state); int __init pm_autosleep_init(void) { diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 0f682da96e1c..3eabf97c4e9a 100644 --- a/kernel/workqueue.c +++ b/kernel/workqueue.c @@ -3713,6 +3713,7 @@ void free_workqueue_attrs(struct workqueue_attrs *attrs) kfree(attrs); } } +EXPORT_SYMBOL_GPL(free_workqueue_attrs); /** * alloc_workqueue_attrs - allocate a workqueue_attrs @@ -3741,6 +3742,7 @@ struct workqueue_attrs *alloc_workqueue_attrs(void) free_workqueue_attrs(attrs); return NULL; } +EXPORT_SYMBOL_GPL(alloc_workqueue_attrs); static void copy_workqueue_attrs(struct workqueue_attrs *to, const struct workqueue_attrs *from) @@ -4482,6 +4484,7 @@ int apply_workqueue_attrs(struct workqueue_struct *wq, return ret; } +EXPORT_SYMBOL_GPL(apply_workqueue_attrs); /** * wq_update_pod - update pod affinity of a wq for CPU hot[un]plug diff --git a/mm/oom_kill.c b/mm/oom_kill.c index 601ee56cc7d7..81b79f82f601 100644 --- a/mm/oom_kill.c +++ b/mm/oom_kill.c @@ -1141,6 +1141,7 @@ int oom_type_notifier_call(unsigned int type, struct oom_control *oc) return blocking_notifier_call_chain(&oom_type_notify_list, type, NULL); } +EXPORT_SYMBOL_GPL(oom_type_notifier_call); #endif /** diff --git a/mm/vmalloc.c b/mm/vmalloc.c index 30665fb33589..719539b32488 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -2622,6 +2622,7 @@ struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags, return __get_vm_area_node(size, 1, PAGE_SHIFT, flags, start, end, NUMA_NO_NODE, GFP_KERNEL, caller); } +EXPORT_SYMBOL(__get_vm_area_caller); /** * get_vm_area - reserve a contiguous kernel virtual area @@ -3362,6 +3363,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align, return NULL; } +EXPORT_SYMBOL_GPL(__vmalloc_node_range); /** * __vmalloc_node - allocate virtually contiguous memory -- 2.17.1

2 1

[PATCH OLK-6.6] ubi: block: Fix use-after-free in ubiblock_cleanup
by Li Nan 18 Dec '23

18 Dec '23

From: ZhaoLong Wang <wangzhaolong1(a)huawei.com> mainline inclusion from mainline-v6.7-rc1 commit d07cec9c238ae8fc6c1a9f3f5d30a2f8ec6cdc71 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I8P68G Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… -------------------------------- The following BUG is reported when a ubiblock is removed: ================================================================== BUG: KASAN: slab-use-after-free in ubiblock_cleanup+0x88/0xa0 [ubi] Read of size 4 at addr ffff88810c8f3804 by task ubiblock/1716 CPU: 5 PID: 1716 Comm: ubiblock Not tainted 6.6.0-rc2+ #135 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ?-20190727_073836-buildvm-ppc64le-16.ppc.fedoraproject.org-3.fc31 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x37/0x50 print_report+0xd0/0x620 kasan_report+0xb6/0xf0 ubiblock_cleanup+0x88/0xa0 [ubi] ubiblock_remove+0x121/0x190 [ubi] vol_cdev_ioctl+0x355/0x630 [ubi] __x64_sys_ioctl+0xc7/0x100 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 RIP: 0033:0x7f08d7445577 Code: b3 66 90 48 8b 05 11 89 2c 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d e1 8 RSP: 002b:00007ffde05a3018 EFLAGS: 00000206 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 00000000ffffffff RCX: 00007f08d7445577 RDX: 0000000000000000 RSI: 0000000000004f08 RDI: 0000000000000003 RBP: 0000000000816010 R08: 00000000008163a7 R09: 0000000000000000 R10: 0000000000000003 R11: 0000000000000206 R12: 0000000000000003 R13: 00007ffde05a3130 R14: 0000000000000000 R15: 0000000000000000 </TASK> Allocated by task 1715: kasan_save_stack+0x22/0x50 kasan_set_track+0x25/0x30 __kasan_kmalloc+0x7f/0x90 __alloc_disk_node+0x40/0x2b0 __blk_mq_alloc_disk+0x3e/0xb0 ubiblock_create+0x2ba/0x620 [ubi] vol_cdev_ioctl+0x581/0x630 [ubi] __x64_sys_ioctl+0xc7/0x100 do_syscall_64+0x3f/0x90 entry_SYSCALL_64_after_hwframe+0x6e/0xd8 Freed by task 0: kasan_save_stack+0x22/0x50 kasan_set_track+0x25/0x30 kasan_save_free_info+0x2b/0x50 __kasan_slab_free+0x10e/0x190 __kmem_cache_free+0x96/0x220 bdev_free_inode+0xa4/0xf0 rcu_core+0x496/0xec0 __do_softirq+0xeb/0x384 The buggy address belongs to the object at ffff88810c8f3800 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 4 bytes inside of freed 1024-byte region [ffff88810c8f3800, ffff88810c8f3c00) The buggy address belongs to the physical page: page:00000000d03de848 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x10c8f0 head:00000000d03de848 order:3 entire_mapcount:0 nr_pages_mapped:0 pincount:0 flags: 0x200000000000840(slab|head|node=0|zone=2) page_type: 0xffffffff() raw: 0200000000000840 ffff888100042dc0 ffffea0004244400 dead000000000002 raw: 0000000000000000 0000000080100010 00000001ffffffff 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff88810c8f3700: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff88810c8f3780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc >ffff88810c8f3800: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff88810c8f3880: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff88810c8f3900: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== Fix it by using a local variable to record the gendisk ID. Fixes: 77567b25ab9f ("ubi: use blk_mq_alloc_disk and blk_cleanup_disk") Signed-off-by: ZhaoLong Wang <wangzhaolong1(a)huawei.com> Reviewed-by: Zhihao Cheng <chengzhihao1(a)huawei.com> Signed-off-by: Richard Weinberger <richard(a)nod.at> Signed-off-by: Li Nan <linan122(a)huawei.com> --- drivers/mtd/ubi/block.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/mtd/ubi/block.c b/drivers/mtd/ubi/block.c index 437c5b83ffe5..309a42aeaa4c 100644 --- a/drivers/mtd/ubi/block.c +++ b/drivers/mtd/ubi/block.c @@ -447,13 +447,15 @@ int ubiblock_create(struct ubi_volume_info *vi) static void ubiblock_cleanup(struct ubiblock *dev) { + int id = dev->gd->first_minor; + /* Stop new requests to arrive */ del_gendisk(dev->gd); /* Finally destroy the blk queue */ dev_info(disk_to_dev(dev->gd), "released"); put_disk(dev->gd); blk_mq_free_tag_set(&dev->tag_set); - idr_remove(&ubiblock_minor_idr, dev->gd->first_minor); + idr_remove(&ubiblock_minor_idr, id); } int ubiblock_remove(struct ubi_volume_info *vi) -- 2.39.2

2 1

[PATCH OLK-5.10] dm: add error handling support for add_disk()
by Li Lingfeng 18 Dec '23

18 Dec '23

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I81XCK CVE: NA --------------------------- NULL pointer dereference and uaf will be triggered in del_gendisk() if add_disk has not complete. Clear request_queue and gendisk early if add_disk fail. Signed-off-by: Li Lingfeng <lilingfeng3(a)huawei.com> --- drivers/md/dm.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) diff --git a/drivers/md/dm.c b/drivers/md/dm.c index e2df27c373f9..654aae1dd777 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -1853,13 +1853,11 @@ static struct mapped_device *alloc_dev(int minor) * established. If request-based table is loaded: blk-mq will * override accordingly. */ - md->queue = blk_alloc_queue(numa_node_id); - if (!md->queue) - goto bad; - - md->disk = alloc_disk_node(1, md->numa_node_id); + md->disk = blk_alloc_disk(md->numa_node_id); if (!md->disk) goto bad; + md->disk->minors = 1; + md->queue = md->disk->queue; init_waitqueue_head(&md->wait); INIT_WORK(&md->work, dm_wq_work); @@ -1873,11 +1871,16 @@ static struct mapped_device *alloc_dev(int minor) md->disk->major = _major; md->disk->first_minor = minor; md->disk->fops = &dm_blk_dops; - md->disk->queue = md->queue; md->disk->private_data = md; sprintf(md->disk->disk_name, "dm-%d", minor); - add_disk_no_queue_reg(md->disk); + r = add_disk_no_queue_reg_safe(md->disk); + if (r) { + blk_cleanup_disk(md->disk); + md->queue = NULL; + md->disk = NULL; + goto bad; + } if (IS_ENABLED(CONFIG_DAX_DRIVER)) { md->dax_dev = alloc_dax(md, md->disk->disk_name, -- 2.31.1

2 1

[PATCH OLK-6.6 v3] blk-throttle: enable hierarchical throttle in cgroup v1
by Yu Kuai 18 Dec '23

18 Dec '23

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I4R7RE CVE: NA ----------------------------------------------- blkio subsytem is not under default hierarchy in cgroup v1 by default, which means configurations will only be effective on current cgroup for io throttle. This patch introduces a new feature that enable default hierarchy for io throttle, which means configurations will be effective on child cgroups. Such feature is disabled by default, and can be enabled by: 1) set CONFIG_BLK_DEV_SUPPORT_LEGACY_GLOBAL_LIMIT, then 2) adding "blkcg_global_limit=1" or "blkcg_global_limit=Y" or "blkcg_global_limit=y" in boot cmd; Signed-off-by: Yu Kuai <yukuai3(a)huawei.com> --- arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + block/Kconfig | 11 ++++++++++ block/blk-throttle.c | 28 +++++++++++++++++++++++++- 4 files changed, 40 insertions(+), 1 deletion(-) diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index 039f3496af78..3e76da491416 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -855,6 +855,7 @@ CONFIG_BLK_DEV_INTEGRITY_T10=m CONFIG_BLK_DEV_ZONED=y CONFIG_BLK_DEV_THROTTLING=y # CONFIG_BLK_DEV_THROTTLING_LOW is not set +CONFIG_BLK_DEV_SUPPORT_LEGACY_GLOBAL_LIMIT=y CONFIG_BLK_WBT=y CONFIG_BLK_WBT_MQ=y # CONFIG_BLK_CGROUP_IOLATENCY is not set diff --git a/arch/x86/configs/openeuler_defconfig b/arch/x86/configs/openeuler_defconfig index f481c1c8b4d1..a4bd203e6db5 100644 --- a/arch/x86/configs/openeuler_defconfig +++ b/arch/x86/configs/openeuler_defconfig @@ -925,6 +925,7 @@ CONFIG_BLK_DEV_INTEGRITY_T10=m CONFIG_BLK_DEV_ZONED=y CONFIG_BLK_DEV_THROTTLING=y # CONFIG_BLK_DEV_THROTTLING_LOW is not set +CONFIG_BLK_DEV_SUPPORT_LEGACY_GLOBAL_LIMIT=y CONFIG_BLK_WBT=y CONFIG_BLK_WBT_MQ=y # CONFIG_BLK_CGROUP_IOLATENCY is not set diff --git a/block/Kconfig b/block/Kconfig index f1364d1c0d93..95378563a05a 100644 --- a/block/Kconfig +++ b/block/Kconfig @@ -111,6 +111,17 @@ config BLK_DEV_THROTTLING_LOW Note, this is an experimental interface and could be changed someday. +config BLK_DEV_SUPPORT_LEGACY_GLOBAL_LIMIT + bool "Block layer global limit in cgroup v1" + depends on BLK_DEV_THROTTLING + default n + help + Blkio subsytem is not under default hierarchy in cgroup v1 by default, + enabling this will support globlal limit in cgroup v1. + + Note, a cmdline "blkcg_global_limit=1" is still required to enabled this + feature. + config BLK_WBT bool "Enable support for block device writeback throttling" help diff --git a/block/blk-throttle.c b/block/blk-throttle.c index 13e4377a8b28..76a54f43135c 100644 --- a/block/blk-throttle.c +++ b/block/blk-throttle.c @@ -41,6 +41,31 @@ /* A workqueue to queue throttle related work */ static struct workqueue_struct *kthrotld_workqueue; +#ifdef CONFIG_BLK_DEV_SUPPORT_LEGACY_GLOBAL_LIMIT +/* True if global limit is enabled in cgroup v1 */ +static bool global_limit; + +static inline bool blkcg_global_limit_enabled(void) +{ + return global_limit; +} + +static int __init setup_global_limit(char *str) +{ + if (!strcmp(str, "1") || !strcmp(str, "Y") || !strcmp(str, "y")) + global_limit = true; + + return 1; +} + +__setup("blkcg_global_limit=", setup_global_limit); + +#else +static inline bool blkcg_global_limit_enabled(void) +{ + return false; +} +#endif #define rb_entry_tg(node) rb_entry((node), struct throtl_grp, rb_node) /* We measure latency for request size from <= 4k to >= 1M */ @@ -405,7 +430,8 @@ static void throtl_pd_init(struct blkg_policy_data *pd) * regardless of the position of the group in the hierarchy. */ sq->parent_sq = &td->service_queue; - if (cgroup_subsys_on_dfl(io_cgrp_subsys) && blkg->parent) + if ((cgroup_subsys_on_dfl(io_cgrp_subsys) || + blkcg_global_limit_enabled()) && blkg->parent) sq->parent_sq = &blkg_to_tg(blkg->parent)->service_queue; tg->td = td; } -- 2.39.2

2 1

[PATCH OLK-6.6] fs/dirty_pages: dump the number of dirty pages for each inode
by Zizhi Wo 18 Dec '23

18 Dec '23

From: yu kuai <yukuai3(a)huawei.com> hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OHM4?from=project-issue CVE: NA --------------------------- In order to analyse the IO performance when using buffer IO, it's useful to obtain the number of dirty pages for each inode in the filesystem. This feature depends on 'CONFIG_DIRTY_PAGES'. It creates 3 interfaces by using procfs. /proc/dirty/buffer_size for buffer allocation and release, /proc/dirty/page_threshold to filter result and /proc/dirty/dirty_list to get dirty pages. Signed-off-by: yu kuai <yukuai3(a)huawei.com> Signed-off-by: Zizhi Wo <wozizhi(a)huawei.com> --- arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + fs/Kconfig | 13 + fs/Makefile | 1 + fs/dirty_pages.c | 457 +++++++++++++++++++++++++ 5 files changed, 473 insertions(+) create mode 100644 fs/dirty_pages.c diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index 039f3496af78..66b4812050df 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -6883,6 +6883,7 @@ CONFIG_PROC_PAGE_MONITOR=y CONFIG_PROC_CHILDREN=y CONFIG_KERNFS=y CONFIG_SYSFS=y +CONFIG_DIRTY_PAGES=y CONFIG_TMPFS=y CONFIG_TMPFS_POSIX_ACL=y CONFIG_TMPFS_XATTR=y diff --git a/arch/x86/configs/openeuler_defconfig b/arch/x86/configs/openeuler_defconfig index f481c1c8b4d1..6133f66fde61 100644 --- a/arch/x86/configs/openeuler_defconfig +++ b/arch/x86/configs/openeuler_defconfig @@ -8090,6 +8090,7 @@ CONFIG_PROC_PID_ARCH_STATUS=y CONFIG_PROC_CPU_RESCTRL=y CONFIG_KERNFS=y CONFIG_SYSFS=y +CONFIG_DIRTY_PAGES=y CONFIG_TMPFS=y CONFIG_TMPFS_POSIX_ACL=y CONFIG_TMPFS_XATTR=y diff --git a/fs/Kconfig b/fs/Kconfig index aa7e03cc1941..a2280cf98729 100644 --- a/fs/Kconfig +++ b/fs/Kconfig @@ -170,6 +170,19 @@ source "fs/proc/Kconfig" source "fs/kernfs/Kconfig" source "fs/sysfs/Kconfig" +config DIRTY_PAGES + bool "Dumps the number of dirty pages of each file" + depends on PROC_FS + default y + help + This config supports the rendering of dirty page data to the user, + which may be useful to analyze the IO performance when using buffer + IO. + + It create 3 interfaces by using procfs. /proc/dirty/buffer_size for + buffer allocation and release; /proc/dirty/page_threshold to filter + result; /proc/dirty/dirty_list to get dirty pages. + config TMPFS bool "Tmpfs virtual memory file system support (former shm fs)" depends on SHMEM diff --git a/fs/Makefile b/fs/Makefile index f9541f40be4e..6246e173e1e4 100644 --- a/fs/Makefile +++ b/fs/Makefile @@ -44,6 +44,7 @@ obj-$(CONFIG_FS_POSIX_ACL) += posix_acl.o obj-$(CONFIG_NFS_COMMON) += nfs_common/ obj-$(CONFIG_COREDUMP) += coredump.o obj-$(CONFIG_SYSCTL) += drop_caches.o sysctls.o +obj-$(CONFIG_DIRTY_PAGES) += dirty_pages.o obj-$(CONFIG_FHANDLE) += fhandle.o obj-y += iomap/ diff --git a/fs/dirty_pages.c b/fs/dirty_pages.c new file mode 100644 index 000000000000..884e448d0c6c --- /dev/null +++ b/fs/dirty_pages.c @@ -0,0 +1,457 @@ +#include <linux/proc_fs.h> +#include <linux/seq_file.h> +#include <linux/uaccess.h> +#include <linux/pagemap.h> +#include <linux/pagevec.h> +#include <linux/fs.h> +#include <linux/mm.h> +#include <linux/slab.h> +#include <linux/sched.h> +#include <linux/proc_fs.h> +#include <linux/kdev_t.h> +#include <linux/vmalloc.h> +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/init.h> +#include "internal.h" + +static char *buf_dirty; /* buffer to store number of dirty pages */ +static unsigned long buf_size; /* size of buffer in bytes */ +static long buff_num; /* size of buffer in number of pages */ +static int buff_limit; /* filter threshold of dirty pages*/ + +static struct proc_dir_entry *dirty_dir; + +static struct mutex buff_used; /* buffer is in used */ +static struct mutex buff_lock; /* lock when buffer is changed */ + +/* proc root directory */ +#define DIRTY_ROOT "dirty" +/* proc file for buffer allocation and release */ +#define DIRTY_SWITCH "buffer_size" +/* proc file to obtain diry pages of each inode */ +#define DIRTY_PAGES "dirty_list" +/* proc file to filter result */ +#define DIRTY_LIMIT "page_threshold" + +#define MAX_BUFF_SIZE 102400 +static void seq_set_overflow(struct seq_file *m) +{ + m->count = m->size; +} + +static unsigned long dump_dirtypages_inode(struct inode *inode) +{ + struct folio_batch fbatch; + unsigned long nr_dirtys = 0; + unsigned int nr_folios; + pgoff_t index = 0; + + folio_batch_init(&fbatch); + + while (1) { + nr_folios = filemap_get_folios_tag(inode->i_mapping, &index, + (pgoff_t)-1, PAGECACHE_TAG_DIRTY, &fbatch); + if (!nr_folios) + break; + + folio_batch_release(&fbatch); + cond_resched(); + + nr_dirtys += nr_folios; + } + + return nr_dirtys; +} + +static char *inode_filename(struct inode *inode, char *tmpname) +{ + struct dentry *dentry; + char *filename; + + dentry = d_find_alias(inode); + if (!dentry) + return ERR_PTR(-ENOENT); + + tmpname[PATH_MAX-1] = '\0'; + filename = dentry_path_raw(dentry, tmpname, PATH_MAX); + + dput(dentry); + + return filename; +} + +static inline bool is_sb_writable(struct super_block *sb) +{ + if (sb_rdonly(sb)) + return false; + + if (sb->s_writers.frozen == SB_FREEZE_COMPLETE) + return false; + + return true; +} + +/* + * dump_dirtypages_sb - dump the dirty pages of each inode in the sb + * @sb the super block + * @m the seq_file witch is initialized in proc_dpages_open + * + * For each inode in the sb, call dump_dirtypages_pages to get the number + * of dirty pages. And use seq_printf to store the result in the buffer + * if it's not less than the threshold. The inode in unusual state will + * be skipped. + */ +static void dump_dirtypages_sb(struct super_block *sb, struct seq_file *m) +{ + struct inode *inode, *toput_inode = NULL; + unsigned long nr_dirtys; + const char *fstype; + char *filename; + char *tmpname; + int limit = READ_ONCE(buff_limit); + + if (!is_sb_writable(sb)) + return; + + tmpname = kmalloc(PATH_MAX, GFP_KERNEL); + if (!tmpname) + return; + + spin_lock(&sb->s_inode_list_lock); + list_for_each_entry(inode, &sb->s_inodes, i_sb_list) { + spin_lock(&inode->i_lock); + + /* + * We must skip inodes in unusual state. We may also skip + * inodes without pages but we deliberately won't in case + * we need to reschedule to avoid softlockups. + */ + if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) || + (inode->i_mapping->nrpages == 0 && !need_resched())) { + spin_unlock(&inode->i_lock); + continue; + } + __iget(inode); + spin_unlock(&inode->i_lock); + spin_unlock(&sb->s_inode_list_lock); + + cond_resched(); + + nr_dirtys = dump_dirtypages_inode(inode); + if (!nr_dirtys || nr_dirtys < limit) + goto skip; + + filename = inode_filename(inode, tmpname); + if (IS_ERR_OR_NULL(filename)) + filename = "unknown"; + + if (sb->s_type && sb->s_type->name) + fstype = sb->s_type->name; + else + fstype = "unknown"; + /* + * seq_printf return nothing, if the buffer is exhausted + * (m->size <= m->count), seq_printf will not store + * anything, just set m->count = m->size and return. In + * that case, log a warn message in buffer to remind users. + */ + if (m->size <= m->count) { + seq_set_overflow(m); + strncpy(m->buf + m->count - 12, "terminated\n\0", 12); + iput(inode); + goto done; + } + seq_printf(m, "FSType: %s, Dev ID: %u(%u:%u) ino %lu, dirty pages %lu, path %s\n", + fstype, sb->s_dev, MAJOR(sb->s_dev), + MINOR(sb->s_dev), inode->i_ino, + nr_dirtys, filename); +skip: + iput(toput_inode); + toput_inode = inode; + spin_lock(&sb->s_inode_list_lock); + } + spin_unlock(&sb->s_inode_list_lock); +done: + iput(toput_inode); + kfree(tmpname); +} + +static int proc_dpages_show(struct seq_file *m, void *v) +{ + iterate_supers((void *)dump_dirtypages_sb, (void *)m); + return 0; +} + + +static ssize_t seq_read_dirty( + struct file *file, + char __user *buf, + size_t size, + loff_t *ppos) +{ + struct seq_file *m = (struct seq_file *)file->private_data; + size_t copied = 0; + size_t n; + int err = 0; + + if (!mutex_trylock(&buff_used)) + return -EBUSY; + /* don't allow buffer to change during read */ + mutex_lock(&buff_lock); + if (m->count == 0) { + memset(buf_dirty, 0, buf_size); + if (!m->buf) { + m->size = buf_size; + m->buf = buf_dirty; + } + err = m->op->show(m, NULL); + if (err < 0) + goto done; + } + + /* if buffer changed somehow */ + if (m->size != buf_size) { + mutex_unlock(&buff_lock); + mutex_unlock(&buff_used); + return -EFAULT; + } + + n = min(m->count - m->from, size); + /* check if this is the last read */ + if (n == 0) + goto done; + n -= copy_to_user(buf, m->buf + m->from, n); + if (unlikely(!n)) { + err = -EFAULT; + goto done; + } + m->from += n; + copied += n; +done: + if (!copied) + copied = err; + else + *ppos += copied; + + mutex_unlock(&buff_lock); + mutex_unlock(&buff_used); + return copied; +} + +static void free_buf_dirty(void) +{ + if (buf_dirty != NULL) { + vfree(buf_dirty); + buf_dirty = NULL; + buf_size = 0; + } +} + +static ssize_t write_proc( + struct file *filp, + const char *buf, + size_t count, + loff_t *offp) +{ + char *msg; + int ret = 0; + long old_buff_num; + + if (!mutex_trylock(&buff_used)) + return -EBUSY; + + if (count > PAGE_SIZE) { + ret = -EINVAL; + goto error; + } + + msg = kzalloc(PAGE_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto error; + } + + if (copy_from_user(msg, buf, count)) { + ret = -EINVAL; + goto free; + } + old_buff_num = buff_num; + ret = kstrtol(msg, 10, &buff_num); + if (ret != 0 || buff_num < 0 || buff_num > MAX_BUFF_SIZE) { + buff_num = old_buff_num; + ret = -EINVAL; + goto free; + } + + mutex_lock(&buff_lock); + + ret = count; + if (buff_num == 0) { + free_buf_dirty(); + goto out; + } + if (buff_num == old_buff_num) + goto out; + + free_buf_dirty(); + buf_size = PAGE_SIZE * buff_num; + buf_dirty = vzalloc(buf_size); + + if (!buf_dirty) { + ret = -ENOMEM; + goto out; + } +out: + mutex_unlock(&buff_lock); +free: + kfree(msg); +error: + mutex_unlock(&buff_used); + return ret; +} + +static int proc_dpages_open(struct inode *inode, struct file *filp) +{ + if (buf_dirty == NULL || buf_size == 0) { + pr_warn("please allocate buffer before getting dirty pages\n"); + return -ENOMEM; + } + + return single_open(filp, proc_dpages_show, NULL); +} + +static int seq_release_dirty(struct inode *inode, struct file *file) +{ + struct seq_file *m = file->private_data; + + /* we don't want to free the buf */ + m->buf = NULL; + single_release(inode, file); + return 0; +} + +static const struct proc_ops proc_dpages_operations = { + .proc_open = proc_dpages_open, + .proc_read = seq_read_dirty, + .proc_release = seq_release_dirty, +}; + +static int proc_switch_show(struct seq_file *m, void *v) +{ + seq_printf(m, "%ld\n", buff_num); + return 0; +} + +static int proc_limit_show(struct seq_file *m, void *v) +{ + seq_printf(m, "%d\n", READ_ONCE(buff_limit)); + return 0; +} + +static int proc_switch_open(struct inode *inode, struct file *filp) +{ + return single_open(filp, proc_switch_show, NULL); +} + +static int proc_limit_open(struct inode *inode, struct file *filp) +{ + return single_open(filp, proc_limit_show, NULL); +} + +static ssize_t write_limit_proc( + struct file *filp, + const char *buf, + size_t count, + loff_t *offp) +{ + char *msg; + int ret = 0; + long temp; + + if (count > PAGE_SIZE) { + ret = -EINVAL; + goto error; + } + + msg = kzalloc(PAGE_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto error; + } + + if (copy_from_user(msg, buf, count)) { + ret = -EINVAL; + goto free; + } + ret = kstrtol(msg, 10, &temp); + if (ret != 0 || temp < 0 || temp > INT_MAX) { + ret = -EINVAL; + goto free; + } + + WRITE_ONCE(buff_limit, temp); + ret = count; + +free: + kfree(msg); +error: + return ret; +} + + +static const struct proc_ops proc_switch_operations = { + .proc_open = proc_switch_open, + .proc_read = seq_read, + .proc_write = write_proc, + .proc_lseek = seq_lseek, + .proc_release = single_release, +}; + +static const struct proc_ops proc_limit_operations = { + .proc_open = proc_limit_open, + .proc_read = seq_read, + .proc_write = write_limit_proc, + .proc_lseek = seq_lseek, + .proc_release = single_release, +}; + + +static int __init dpages_proc_init(void) +{ + static struct proc_dir_entry *proc_file; + + dirty_dir = proc_mkdir(DIRTY_ROOT, NULL); + if (!dirty_dir) + goto fail_dir; + + proc_file = proc_create(DIRTY_PAGES, 0440, + dirty_dir, &proc_dpages_operations); + if (!proc_file) + goto fail_pages; + + proc_file = proc_create(DIRTY_SWITCH, 0640, + dirty_dir, &proc_switch_operations); + if (!proc_file) + goto fail_switch; + + proc_file = proc_create(DIRTY_LIMIT, 0640, + dirty_dir, &proc_limit_operations); + if (!proc_file) + goto fail_limit; + + mutex_init(&buff_used); + mutex_init(&buff_lock); + return 0; + +fail_limit: + remove_proc_entry(DIRTY_SWITCH, dirty_dir); +fail_switch: + remove_proc_entry(DIRTY_PAGES, dirty_dir); +fail_pages: + remove_proc_entry(DIRTY_ROOT, NULL); +fail_dir: + return -ENOMEM; +} + +subsys_initcall(dpages_proc_init); -- 2.39.2

2 1

[PATCH OLK-6.6 00/11] LoongArch: fix some pci problems
by Hongchen Zhang 18 Dec '23

18 Dec '23

Baoqi Zhang (2): LS7A2000: Add quirk for OHCI device rev 0x02 PCI: LS7A2000: fix GPU card error Hongchen Zhang (1): PCI: fix kabi error caused by pm_suspend_target_state Jianmin Lv (2): PCI: PM: fix pcie mrrs restoring PCI: LS7A2000: fix pm transition of devices under pcie port Juxin Gao (1): PCI: irq: Add early_param pci_irq_limit to limit pci irq numbers Tianli Xiong (4): PCI: Check if entry->offset already exist for mem resource PCI: Check if the pci controller can use both CFG0 and CFG1 mode to access configuration space LoongArch: fix some PCIE card not scanning properly LoongArch: pci root bridige set acpi companion only when not acpi_disabled. suijingfeng (1): PCI: fix X server auto probe fail when both ast and etnaviv drm present arch/loongarch/pci/acpi.c | 23 ++-- drivers/pci/controller/pci-loongson.c | 147 +++++++++++++++++++++++++- drivers/pci/msi/msi.c | 25 +++++ drivers/pci/pci.c | 20 +++- 4 files changed, 200 insertions(+), 15 deletions(-) -- 2.33.0

2 12