[linux-2.6-block.git] / fs / btrfs / async-thread.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 * Copyright (C) 2014 Fujitsu.  All rights reserved.
 */

#include <linux/kthread.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/freezer.h>
#include "async-thread.h"
#include "ctree.h"

#define WORK_DONE_BIT 0
#define WORK_ORDER_DONE_BIT 1
#define WORK_HIGH_PRIO_BIT 2

#define NO_THRESHOLD (-1)
#define DFT_THRESHOLD (32)

struct __btrfs_workqueue {
	struct workqueue_struct *normal_wq;

	/* File system this workqueue services */
	struct btrfs_fs_info *fs_info;

	/* List head pointing to ordered work list */
	struct list_head ordered_list;

	/* Spinlock for ordered_list */
	spinlock_t list_lock;

	/* Thresholding related variants */
	atomic_t pending;

	/* Up limit of concurrency workers */
	int limit_active;

	/* Current number of concurrency workers */
	int current_active;

	/* Threshold to change current_active */
	int thresh;
	unsigned int count;
	spinlock_t thres_lock;
};

struct btrfs_workqueue {
	struct __btrfs_workqueue *normal;
	struct __btrfs_workqueue *high;
};

static void normal_work_helper(struct btrfs_work *work);

#define BTRFS_WORK_HELPER(name)					\
noinline_for_stack void btrfs_##name(struct work_struct *arg)		\
{									\
	struct btrfs_work *work = container_of(arg, struct btrfs_work,	\
					       normal_work);		\
	normal_work_helper(work);					\
}

struct btrfs_fs_info *
btrfs_workqueue_owner(const struct __btrfs_workqueue *wq)
{
	return wq->fs_info;
}

struct btrfs_fs_info *
btrfs_work_owner(const struct btrfs_work *work)
{
	return work->wq->fs_info;
}

bool btrfs_workqueue_normal_congested(const struct btrfs_workqueue *wq)
{
	/*
	 * We could compare wq->normal->pending with num_online_cpus()
	 * to support "thresh == NO_THRESHOLD" case, but it requires
	 * moving up atomic_inc/dec in thresh_queue/exec_hook. Let's
	 * postpone it until someone needs the support of that case.
	 */
	if (wq->normal->thresh == NO_THRESHOLD)
		return false;

	return atomic_read(&wq->normal->pending) > wq->normal->thresh * 2;
}

BTRFS_WORK_HELPER(worker_helper);
BTRFS_WORK_HELPER(delalloc_helper);
BTRFS_WORK_HELPER(flush_delalloc_helper);
BTRFS_WORK_HELPER(cache_helper);
BTRFS_WORK_HELPER(submit_helper);
BTRFS_WORK_HELPER(fixup_helper);
BTRFS_WORK_HELPER(endio_helper);
BTRFS_WORK_HELPER(endio_meta_helper);
BTRFS_WORK_HELPER(endio_meta_write_helper);
BTRFS_WORK_HELPER(endio_raid56_helper);
BTRFS_WORK_HELPER(endio_repair_helper);
BTRFS_WORK_HELPER(rmw_helper);
BTRFS_WORK_HELPER(endio_write_helper);
BTRFS_WORK_HELPER(freespace_write_helper);
BTRFS_WORK_HELPER(delayed_meta_helper);
BTRFS_WORK_HELPER(readahead_helper);
BTRFS_WORK_HELPER(qgroup_rescan_helper);
BTRFS_WORK_HELPER(extent_refs_helper);
BTRFS_WORK_HELPER(scrub_helper);
BTRFS_WORK_HELPER(scrubwrc_helper);
BTRFS_WORK_HELPER(scrubnc_helper);
BTRFS_WORK_HELPER(scrubparity_helper);

static struct __btrfs_workqueue *
__btrfs_alloc_workqueue(struct btrfs_fs_info *fs_info, const char *name,
			unsigned int flags, int limit_active, int thresh)
{
	struct __btrfs_workqueue *ret = kzalloc(sizeof(*ret), GFP_KERNEL);

	if (!ret)
		return NULL;

	ret->fs_info = fs_info;
	ret->limit_active = limit_active;
	atomic_set(&ret->pending, 0);
	if (thresh == 0)
		thresh = DFT_THRESHOLD;
	/* For low threshold, disabling threshold is a better choice */
	if (thresh < DFT_THRESHOLD) {
		ret->current_active = limit_active;
		ret->thresh = NO_THRESHOLD;
	} else {
		/*
		 * For threshold-able wq, let its concurrency grow on demand.
		 * Use minimal max_active at alloc time to reduce resource
		 * usage.
		 */
		ret->current_active = 1;
		ret->thresh = thresh;
	}

	if (flags & WQ_HIGHPRI)
		ret->normal_wq = alloc_workqueue("%s-%s-high", flags,
						 ret->current_active, "btrfs",
						 name);
	else
		ret->normal_wq = alloc_workqueue("%s-%s", flags,
						 ret->current_active, "btrfs",
						 name);
	if (!ret->normal_wq) {
		kfree(ret);
		return NULL;
	}

	INIT_LIST_HEAD(&ret->ordered_list);
	spin_lock_init(&ret->list_lock);
	spin_lock_init(&ret->thres_lock);
	trace_btrfs_workqueue_alloc(ret, name, flags & WQ_HIGHPRI);
	return ret;
}

static inline void
__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq);

struct btrfs_workqueue *btrfs_alloc_workqueue(struct btrfs_fs_info *fs_info,
					      const char *name,
					      unsigned int flags,
					      int limit_active,
					      int thresh)
{
	struct btrfs_workqueue *ret = kzalloc(sizeof(*ret), GFP_KERNEL);

	if (!ret)
		return NULL;

	ret->normal = __btrfs_alloc_workqueue(fs_info, name,
					      flags & ~WQ_HIGHPRI,
					      limit_active, thresh);
	if (!ret->normal) {
		kfree(ret);
		return NULL;
	}

	if (flags & WQ_HIGHPRI) {
		ret->high = __btrfs_alloc_workqueue(fs_info, name, flags,
						    limit_active, thresh);
		if (!ret->high) {
			__btrfs_destroy_workqueue(ret->normal);
			kfree(ret);
			return NULL;
		}
	}
	return ret;
}

/*
 * Hook for threshold which will be called in btrfs_queue_work.
 * This hook WILL be called in IRQ handler context,
 * so workqueue_set_max_active MUST NOT be called in this hook
 */
static inline void thresh_queue_hook(struct __btrfs_workqueue *wq)
{
	if (wq->thresh == NO_THRESHOLD)
		return;
	atomic_inc(&wq->pending);
}

/*
 * Hook for threshold which will be called before executing the work,
 * This hook is called in kthread content.
 * So workqueue_set_max_active is called here.
 */
static inline void thresh_exec_hook(struct __btrfs_workqueue *wq)
{
	int new_current_active;
	long pending;
	int need_change = 0;

	if (wq->thresh == NO_THRESHOLD)
		return;

	atomic_dec(&wq->pending);
	spin_lock(&wq->thres_lock);
	/*
	 * Use wq->count to limit the calling frequency of
	 * workqueue_set_max_active.
	 */
	wq->count++;
	wq->count %= (wq->thresh / 4);
	if (!wq->count)
		goto  out;
	new_current_active = wq->current_active;

	/*
	 * pending may be changed later, but it's OK since we really
	 * don't need it so accurate to calculate new_max_active.
	 */
	pending = atomic_read(&wq->pending);
	if (pending > wq->thresh)
		new_current_active++;
	if (pending < wq->thresh / 2)
		new_current_active--;
	new_current_active = clamp_val(new_current_active, 1, wq->limit_active);
	if (new_current_active != wq->current_active)  {
		need_change = 1;
		wq->current_active = new_current_active;
	}
out:
	spin_unlock(&wq->thres_lock);

	if (need_change) {
		workqueue_set_max_active(wq->normal_wq, wq->current_active);
	}
}

static void run_ordered_work(struct __btrfs_workqueue *wq)
{
	struct list_head *list = &wq->ordered_list;
	struct btrfs_work *work;
	spinlock_t *lock = &wq->list_lock;
	unsigned long flags;

	while (1) {
		void *wtag;

		spin_lock_irqsave(lock, flags);
		if (list_empty(list))
			break;
		work = list_entry(list->next, struct btrfs_work,
				  ordered_list);
		if (!test_bit(WORK_DONE_BIT, &work->flags))
			break;

		/*
		 * we are going to call the ordered done function, but
		 * we leave the work item on the list as a barrier so
		 * that later work items that are done don't have their
		 * functions called before this one returns
		 */
		if (test_and_set_bit(WORK_ORDER_DONE_BIT, &work->flags))
			break;
		trace_btrfs_ordered_sched(work);
		spin_unlock_irqrestore(lock, flags);
		work->ordered_func(work);

		/* now take the lock again and drop our item from the list */
		spin_lock_irqsave(lock, flags);
		list_del(&work->ordered_list);
		spin_unlock_irqrestore(lock, flags);

		/*
		 * We don't want to call the ordered free functions with the
		 * lock held though. Save the work as tag for the trace event,
		 * because the callback could free the structure.
		 */
		wtag = work;
		work->ordered_free(work);
		trace_btrfs_all_work_done(wq->fs_info, wtag);
	}
	spin_unlock_irqrestore(lock, flags);
}

static void normal_work_helper(struct btrfs_work *work)
{
	struct __btrfs_workqueue *wq;
	void *wtag;
	int need_order = 0;

	/*
	 * We should not touch things inside work in the following cases:
	 * 1) after work->func() if it has no ordered_free
	 *    Since the struct is freed in work->func().
	 * 2) after setting WORK_DONE_BIT
	 *    The work may be freed in other threads almost instantly.
	 * So we save the needed things here.
	 */
	if (work->ordered_func)
		need_order = 1;
	wq = work->wq;
	/* Safe for tracepoints in case work gets freed by the callback */
	wtag = work;

	trace_btrfs_work_sched(work);
	thresh_exec_hook(wq);
	work->func(work);
	if (need_order) {
		set_bit(WORK_DONE_BIT, &work->flags);
		run_ordered_work(wq);
	}
	if (!need_order)
		trace_btrfs_all_work_done(wq->fs_info, wtag);
}

void btrfs_init_work(struct btrfs_work *work, btrfs_work_func_t uniq_func,
		     btrfs_func_t func,
		     btrfs_func_t ordered_func,
		     btrfs_func_t ordered_free)
{
	work->func = func;
	work->ordered_func = ordered_func;
	work->ordered_free = ordered_free;
	INIT_WORK(&work->normal_work, uniq_func);
	INIT_LIST_HEAD(&work->ordered_list);
	work->flags = 0;
}

static inline void __btrfs_queue_work(struct __btrfs_workqueue *wq,
				      struct btrfs_work *work)
{
	unsigned long flags;

	work->wq = wq;
	thresh_queue_hook(wq);
	if (work->ordered_func) {
		spin_lock_irqsave(&wq->list_lock, flags);
		list_add_tail(&work->ordered_list, &wq->ordered_list);
		spin_unlock_irqrestore(&wq->list_lock, flags);
	}
	trace_btrfs_work_queued(work);
	queue_work(wq->normal_wq, &work->normal_work);
}

void btrfs_queue_work(struct btrfs_workqueue *wq,
		      struct btrfs_work *work)
{
	struct __btrfs_workqueue *dest_wq;

	if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags) && wq->high)
		dest_wq = wq->high;
	else
		dest_wq = wq->normal;
	__btrfs_queue_work(dest_wq, work);
}

static inline void
__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq)
{
	destroy_workqueue(wq->normal_wq);
	trace_btrfs_workqueue_destroy(wq);
	kfree(wq);
}

void btrfs_destroy_workqueue(struct btrfs_workqueue *wq)
{
	if (!wq)
		return;
	if (wq->high)
		__btrfs_destroy_workqueue(wq->high);
	__btrfs_destroy_workqueue(wq->normal);
	kfree(wq);
}

void btrfs_workqueue_set_max(struct btrfs_workqueue *wq, int limit_active)
{
	if (!wq)
		return;
	wq->normal->limit_active = limit_active;
	if (wq->high)
		wq->high->limit_active = limit_active;
}

void btrfs_set_work_high_priority(struct btrfs_work *work)
{
	set_bit(WORK_HIGH_PRIO_BIT, &work->flags);
}
Commit	Line	Data
c1d7c514	1	// SPDX-License-Identifier: GPL-2.0
8b712842 CM	2	/*
8b712842 CM	3	* Copyright (C) 2007 Oracle. All rights reserved.
08a9ff32	4	* Copyright (C) 2014 Fujitsu. All rights reserved.
8b712842 CM	5	*/
	6
	7	#include <linux/kthread.h>
5a0e3ad6	8	#include <linux/slab.h>
8b712842 CM	9	#include <linux/list.h>
8b712842 CM	10	#include <linux/spinlock.h>
b51912c9	11	#include <linux/freezer.h>
8b712842	12	#include "async-thread.h"
52483bc2	13	#include "ctree.h"
8b712842	14
a046e9c8 QW	15	#define WORK_DONE_BIT 0
	16	#define WORK_ORDER_DONE_BIT 1
	17	#define WORK_HIGH_PRIO_BIT 2
4a69a410	18
0bd9289c QW	19	#define NO_THRESHOLD (-1)
	20	#define DFT_THRESHOLD (32)
	21
d458b054	22	struct __btrfs_workqueue {
08a9ff32	23	struct workqueue_struct *normal_wq;
cb001095 JM	24
	25	/* File system this workqueue services */
	26	struct btrfs_fs_info *fs_info;
	27
08a9ff32 QW	28	/* List head pointing to ordered work list */
	29	struct list_head ordered_list;
	30
	31	/* Spinlock for ordered_list */
	32	spinlock_t list_lock;
0bd9289c QW	33
	34	/* Thresholding related variants */
	35	atomic_t pending;
c6dd6ea5 QW	36
	37	/* Up limit of concurrency workers */
	38	int limit_active;
	39
	40	/* Current number of concurrency workers */
	41	int current_active;
	42
	43	/* Threshold to change current_active */
0bd9289c QW	44	int thresh;
	45	unsigned int count;
	46	spinlock_t thres_lock;
08a9ff32 QW	47	};
08a9ff32 QW	48
d458b054 QW	49	struct btrfs_workqueue {
	50	struct __btrfs_workqueue *normal;
	51	struct __btrfs_workqueue *high;
1ca08976 QW	52	};
1ca08976 QW	53
9e0af237 LB	54	static void normal_work_helper(struct btrfs_work *work);
	55
	56	#define BTRFS_WORK_HELPER(name) \
6939f667	57	noinline_for_stack void btrfs_##name(struct work_struct *arg) \
9e0af237 LB	58	{ \
	59	struct btrfs_work *work = container_of(arg, struct btrfs_work, \
	60	normal_work); \
	61	normal_work_helper(work); \
	62	}
	63
cb001095	64	struct btrfs_fs_info *
9a35b637	65	btrfs_workqueue_owner(const struct __btrfs_workqueue *wq)
cb001095 JM	66	{
	67	return wq->fs_info;
	68	}
	69
	70	struct btrfs_fs_info *
9a35b637	71	btrfs_work_owner(const struct btrfs_work *work)
cb001095 JM	72	{
	73	return work->wq->fs_info;
	74	}
	75
9a35b637	76	bool btrfs_workqueue_normal_congested(const struct btrfs_workqueue *wq)
2939e1a8 MP	77	{
	78	/*
	79	* We could compare wq->normal->pending with num_online_cpus()
	80	* to support "thresh == NO_THRESHOLD" case, but it requires
	81	* moving up atomic_inc/dec in thresh_queue/exec_hook. Let's
	82	* postpone it until someone needs the support of that case.
	83	*/
	84	if (wq->normal->thresh == NO_THRESHOLD)
	85	return false;
	86
	87	return atomic_read(&wq->normal->pending) > wq->normal->thresh * 2;
	88	}
	89
9e0af237 LB	90	BTRFS_WORK_HELPER(worker_helper);
	91	BTRFS_WORK_HELPER(delalloc_helper);
	92	BTRFS_WORK_HELPER(flush_delalloc_helper);
	93	BTRFS_WORK_HELPER(cache_helper);
	94	BTRFS_WORK_HELPER(submit_helper);
	95	BTRFS_WORK_HELPER(fixup_helper);
	96	BTRFS_WORK_HELPER(endio_helper);
	97	BTRFS_WORK_HELPER(endio_meta_helper);
	98	BTRFS_WORK_HELPER(endio_meta_write_helper);
	99	BTRFS_WORK_HELPER(endio_raid56_helper);
8b110e39	100	BTRFS_WORK_HELPER(endio_repair_helper);
9e0af237 LB	101	BTRFS_WORK_HELPER(rmw_helper);
	102	BTRFS_WORK_HELPER(endio_write_helper);
	103	BTRFS_WORK_HELPER(freespace_write_helper);
	104	BTRFS_WORK_HELPER(delayed_meta_helper);
	105	BTRFS_WORK_HELPER(readahead_helper);
	106	BTRFS_WORK_HELPER(qgroup_rescan_helper);
	107	BTRFS_WORK_HELPER(extent_refs_helper);
	108	BTRFS_WORK_HELPER(scrub_helper);
	109	BTRFS_WORK_HELPER(scrubwrc_helper);
	110	BTRFS_WORK_HELPER(scrubnc_helper);
20b2e302	111	BTRFS_WORK_HELPER(scrubparity_helper);
9e0af237 LB	112
9e0af237 LB	113	static struct __btrfs_workqueue *
cb001095 JM	114	__btrfs_alloc_workqueue(struct btrfs_fs_info fs_info, const char name,
cb001095 JM	115	unsigned int flags, int limit_active, int thresh)
1ca08976	116	{
61dd5ae6	117	struct __btrfs_workqueue ret = kzalloc(sizeof(ret), GFP_KERNEL);
1ca08976	118
5d99a998	119	if (!ret)
1ca08976 QW	120	return NULL;
1ca08976 QW	121
cb001095	122	ret->fs_info = fs_info;
c6dd6ea5	123	ret->limit_active = limit_active;
0bd9289c QW	124	atomic_set(&ret->pending, 0);
	125	if (thresh == 0)
	126	thresh = DFT_THRESHOLD;
	127	/* For low threshold, disabling threshold is a better choice */
	128	if (thresh < DFT_THRESHOLD) {
c6dd6ea5	129	ret->current_active = limit_active;
0bd9289c QW	130	ret->thresh = NO_THRESHOLD;
0bd9289c QW	131	} else {
c6dd6ea5 QW	132	/*
	133	* For threshold-able wq, let its concurrency grow on demand.
	134	* Use minimal max_active at alloc time to reduce resource
	135	* usage.
	136	*/
	137	ret->current_active = 1;
0bd9289c QW	138	ret->thresh = thresh;
	139	}
	140
1ca08976 QW	141	if (flags & WQ_HIGHPRI)
1ca08976 QW	142	ret->normal_wq = alloc_workqueue("%s-%s-high", flags,
c6dd6ea5 QW	143	ret->current_active, "btrfs",
c6dd6ea5 QW	144	name);
1ca08976 QW	145	else
1ca08976 QW	146	ret->normal_wq = alloc_workqueue("%s-%s", flags,
c6dd6ea5	147	ret->current_active, "btrfs",
0bd9289c	148	name);
5d99a998	149	if (!ret->normal_wq) {
1ca08976 QW	150	kfree(ret);
	151	return NULL;
	152	}
	153
	154	INIT_LIST_HEAD(&ret->ordered_list);
	155	spin_lock_init(&ret->list_lock);
0bd9289c	156	spin_lock_init(&ret->thres_lock);
c3a46891	157	trace_btrfs_workqueue_alloc(ret, name, flags & WQ_HIGHPRI);
1ca08976 QW	158	return ret;
	159	}
	160
	161	static inline void
d458b054	162	__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq);
1ca08976	163
cb001095 JM	164	struct btrfs_workqueue btrfs_alloc_workqueue(struct btrfs_fs_info fs_info,
cb001095 JM	165	const char *name,
6f011058	166	unsigned int flags,
c6dd6ea5	167	int limit_active,
d458b054	168	int thresh)
08a9ff32	169	{
61dd5ae6	170	struct btrfs_workqueue ret = kzalloc(sizeof(ret), GFP_KERNEL);
08a9ff32	171
5d99a998	172	if (!ret)
08a9ff32 QW	173	return NULL;
08a9ff32 QW	174
cb001095 JM	175	ret->normal = __btrfs_alloc_workqueue(fs_info, name,
cb001095 JM	176	flags & ~WQ_HIGHPRI,
c6dd6ea5	177	limit_active, thresh);
5d99a998	178	if (!ret->normal) {
08a9ff32 QW	179	kfree(ret);
	180	return NULL;
	181	}
	182
1ca08976	183	if (flags & WQ_HIGHPRI) {
cb001095 JM	184	ret->high = __btrfs_alloc_workqueue(fs_info, name, flags,
cb001095 JM	185	limit_active, thresh);
5d99a998	186	if (!ret->high) {
1ca08976 QW	187	__btrfs_destroy_workqueue(ret->normal);
	188	kfree(ret);
	189	return NULL;
	190	}
	191	}
08a9ff32 QW	192	return ret;
	193	}
	194
0bd9289c QW	195	/*
	196	* Hook for threshold which will be called in btrfs_queue_work.
	197	* This hook WILL be called in IRQ handler context,
	198	* so workqueue_set_max_active MUST NOT be called in this hook
	199	*/
d458b054	200	static inline void thresh_queue_hook(struct __btrfs_workqueue *wq)
0bd9289c QW	201	{
	202	if (wq->thresh == NO_THRESHOLD)
	203	return;
	204	atomic_inc(&wq->pending);
	205	}
	206
	207	/*
	208	* Hook for threshold which will be called before executing the work,
	209	* This hook is called in kthread content.
	210	* So workqueue_set_max_active is called here.
	211	*/
d458b054	212	static inline void thresh_exec_hook(struct __btrfs_workqueue *wq)
0bd9289c	213	{
c6dd6ea5	214	int new_current_active;
0bd9289c QW	215	long pending;
	216	int need_change = 0;
	217
	218	if (wq->thresh == NO_THRESHOLD)
	219	return;
	220
	221	atomic_dec(&wq->pending);
	222	spin_lock(&wq->thres_lock);
	223	/*
	224	* Use wq->count to limit the calling frequency of
	225	* workqueue_set_max_active.
	226	*/
	227	wq->count++;
	228	wq->count %= (wq->thresh / 4);
	229	if (!wq->count)
	230	goto out;
c6dd6ea5	231	new_current_active = wq->current_active;
0bd9289c QW	232
	233	/*
	234	* pending may be changed later, but it's OK since we really
	235	* don't need it so accurate to calculate new_max_active.
	236	*/
	237	pending = atomic_read(&wq->pending);
	238	if (pending > wq->thresh)
c6dd6ea5	239	new_current_active++;
0bd9289c	240	if (pending < wq->thresh / 2)
c6dd6ea5 QW	241	new_current_active--;
	242	new_current_active = clamp_val(new_current_active, 1, wq->limit_active);
	243	if (new_current_active != wq->current_active) {
0bd9289c	244	need_change = 1;
c6dd6ea5	245	wq->current_active = new_current_active;
0bd9289c QW	246	}
	247	out:
	248	spin_unlock(&wq->thres_lock);
	249
	250	if (need_change) {
c6dd6ea5	251	workqueue_set_max_active(wq->normal_wq, wq->current_active);
0bd9289c QW	252	}
	253	}
	254
d458b054	255	static void run_ordered_work(struct __btrfs_workqueue *wq)
08a9ff32 QW	256	{
08a9ff32 QW	257	struct list_head *list = &wq->ordered_list;
d458b054	258	struct btrfs_work *work;
08a9ff32 QW	259	spinlock_t *lock = &wq->list_lock;
	260	unsigned long flags;
	261
	262	while (1) {
ac0c7cf8 DS	263	void *wtag;
ac0c7cf8 DS	264
08a9ff32 QW	265	spin_lock_irqsave(lock, flags);
	266	if (list_empty(list))
	267	break;
d458b054	268	work = list_entry(list->next, struct btrfs_work,
08a9ff32 QW	269	ordered_list);
	270	if (!test_bit(WORK_DONE_BIT, &work->flags))
	271	break;
	272
	273	/*
	274	* we are going to call the ordered done function, but
	275	* we leave the work item on the list as a barrier so
	276	* that later work items that are done don't have their
	277	* functions called before this one returns
	278	*/
	279	if (test_and_set_bit(WORK_ORDER_DONE_BIT, &work->flags))
	280	break;
52483bc2	281	trace_btrfs_ordered_sched(work);
08a9ff32 QW	282	spin_unlock_irqrestore(lock, flags);
	283	work->ordered_func(work);
	284
	285	/* now take the lock again and drop our item from the list */
	286	spin_lock_irqsave(lock, flags);
	287	list_del(&work->ordered_list);
	288	spin_unlock_irqrestore(lock, flags);
	289
	290	/*
ac0c7cf8 DS	291	* We don't want to call the ordered free functions with the
	292	* lock held though. Save the work as tag for the trace event,
	293	* because the callback could free the structure.
08a9ff32	294	*/
ac0c7cf8	295	wtag = work;
08a9ff32	296	work->ordered_free(work);
ac0c7cf8	297	trace_btrfs_all_work_done(wq->fs_info, wtag);
08a9ff32 QW	298	}
	299	spin_unlock_irqrestore(lock, flags);
	300	}
	301
9e0af237	302	static void normal_work_helper(struct btrfs_work *work)
08a9ff32	303	{
d458b054	304	struct __btrfs_workqueue *wq;
ac0c7cf8	305	void *wtag;
08a9ff32 QW	306	int need_order = 0;
08a9ff32 QW	307
08a9ff32 QW	308	/*
	309	* We should not touch things inside work in the following cases:
	310	* 1) after work->func() if it has no ordered_free
	311	* Since the struct is freed in work->func().
	312	* 2) after setting WORK_DONE_BIT
	313	* The work may be freed in other threads almost instantly.
	314	* So we save the needed things here.
	315	*/
	316	if (work->ordered_func)
	317	need_order = 1;
	318	wq = work->wq;
ac0c7cf8 DS	319	/* Safe for tracepoints in case work gets freed by the callback */
ac0c7cf8 DS	320	wtag = work;
08a9ff32	321
52483bc2	322	trace_btrfs_work_sched(work);
0bd9289c	323	thresh_exec_hook(wq);
08a9ff32 QW	324	work->func(work);
	325	if (need_order) {
	326	set_bit(WORK_DONE_BIT, &work->flags);
	327	run_ordered_work(wq);
	328	}
52483bc2	329	if (!need_order)
ac0c7cf8	330	trace_btrfs_all_work_done(wq->fs_info, wtag);
08a9ff32 QW	331	}
08a9ff32 QW	332
9e0af237	333	void btrfs_init_work(struct btrfs_work *work, btrfs_work_func_t uniq_func,
6db8914f QW	334	btrfs_func_t func,
	335	btrfs_func_t ordered_func,
	336	btrfs_func_t ordered_free)
08a9ff32 QW	337	{
	338	work->func = func;
	339	work->ordered_func = ordered_func;
	340	work->ordered_free = ordered_free;
9e0af237	341	INIT_WORK(&work->normal_work, uniq_func);
08a9ff32 QW	342	INIT_LIST_HEAD(&work->ordered_list);
	343	work->flags = 0;
	344	}
	345
d458b054 QW	346	static inline void __btrfs_queue_work(struct __btrfs_workqueue *wq,
d458b054 QW	347	struct btrfs_work *work)
08a9ff32 QW	348	{
	349	unsigned long flags;
	350
	351	work->wq = wq;
0bd9289c	352	thresh_queue_hook(wq);
08a9ff32 QW	353	if (work->ordered_func) {
	354	spin_lock_irqsave(&wq->list_lock, flags);
	355	list_add_tail(&work->ordered_list, &wq->ordered_list);
	356	spin_unlock_irqrestore(&wq->list_lock, flags);
	357	}
52483bc2	358	trace_btrfs_work_queued(work);
0a95b851	359	queue_work(wq->normal_wq, &work->normal_work);
08a9ff32 QW	360	}
08a9ff32 QW	361
d458b054 QW	362	void btrfs_queue_work(struct btrfs_workqueue *wq,
d458b054 QW	363	struct btrfs_work *work)
1ca08976	364	{
d458b054	365	struct __btrfs_workqueue *dest_wq;
1ca08976 QW	366
	367	if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags) && wq->high)
	368	dest_wq = wq->high;
	369	else
	370	dest_wq = wq->normal;
	371	__btrfs_queue_work(dest_wq, work);
	372	}
	373
	374	static inline void
d458b054	375	__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq)
08a9ff32 QW	376	{
08a9ff32 QW	377	destroy_workqueue(wq->normal_wq);
c3a46891	378	trace_btrfs_workqueue_destroy(wq);
08a9ff32 QW	379	kfree(wq);
	380	}
	381
d458b054	382	void btrfs_destroy_workqueue(struct btrfs_workqueue *wq)
1ca08976 QW	383	{
	384	if (!wq)
	385	return;
	386	if (wq->high)
	387	__btrfs_destroy_workqueue(wq->high);
	388	__btrfs_destroy_workqueue(wq->normal);
ef66af10	389	kfree(wq);
1ca08976 QW	390	}
1ca08976 QW	391
c6dd6ea5	392	void btrfs_workqueue_set_max(struct btrfs_workqueue *wq, int limit_active)
08a9ff32	393	{
800ee224 ST	394	if (!wq)
800ee224 ST	395	return;
c6dd6ea5	396	wq->normal->limit_active = limit_active;
1ca08976	397	if (wq->high)
c6dd6ea5	398	wq->high->limit_active = limit_active;
1ca08976 QW	399	}
1ca08976 QW	400
d458b054	401	void btrfs_set_work_high_priority(struct btrfs_work *work)
1ca08976 QW	402	{
1ca08976 QW	403	set_bit(WORK_HIGH_PRIO_BIT, &work->flags);
08a9ff32	404	}