xref: /linux/fs/xfs/xfs_pwork.c (revision 3839a7460721b87501134697b7b90c45dcc7825d)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright (C) 2019 Oracle.  All Rights Reserved.
4  * Author: Darrick J. Wong <darrick.wong@oracle.com>
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_trace.h"
14 #include "xfs_sysctl.h"
15 #include "xfs_pwork.h"
16 #include <linux/nmi.h>
17 
18 /*
19  * Parallel Work Queue
20  * ===================
21  *
22  * Abstract away the details of running a large and "obviously" parallelizable
23  * task across multiple CPUs.  Callers initialize the pwork control object with
24  * a desired level of parallelization and a work function.  Next, they embed
25  * struct xfs_pwork in whatever structure they use to pass work context to a
26  * worker thread and queue that pwork.  The work function will be passed the
27  * pwork item when it is run (from process context) and any returned error will
28  * be recorded in xfs_pwork_ctl.error.  Work functions should check for errors
29  * and abort if necessary; the non-zeroness of xfs_pwork_ctl.error does not
30  * stop workqueue item processing.
31  *
32  * This is the rough equivalent of the xfsprogs workqueue code, though we can't
33  * reuse that name here.
34  */
35 
36 /* Invoke our caller's function. */
37 static void
38 xfs_pwork_work(
39 	struct work_struct	*work)
40 {
41 	struct xfs_pwork	*pwork;
42 	struct xfs_pwork_ctl	*pctl;
43 	int			error;
44 
45 	pwork = container_of(work, struct xfs_pwork, work);
46 	pctl = pwork->pctl;
47 	error = pctl->work_fn(pctl->mp, pwork);
48 	if (error && !pctl->error)
49 		pctl->error = error;
50 	if (atomic_dec_and_test(&pctl->nr_work))
51 		wake_up(&pctl->poll_wait);
52 }
53 
54 /*
55  * Set up control data for parallel work.  @work_fn is the function that will
56  * be called.  @tag will be written into the kernel threads.  @nr_threads is
57  * the level of parallelism desired, or 0 for no limit.
58  */
59 int
60 xfs_pwork_init(
61 	struct xfs_mount	*mp,
62 	struct xfs_pwork_ctl	*pctl,
63 	xfs_pwork_work_fn	work_fn,
64 	const char		*tag,
65 	unsigned int		nr_threads)
66 {
67 #ifdef DEBUG
68 	if (xfs_globals.pwork_threads >= 0)
69 		nr_threads = xfs_globals.pwork_threads;
70 #endif
71 	trace_xfs_pwork_init(mp, nr_threads, current->pid);
72 
73 	pctl->wq = alloc_workqueue("%s-%d", WQ_FREEZABLE, nr_threads, tag,
74 			current->pid);
75 	if (!pctl->wq)
76 		return -ENOMEM;
77 	pctl->work_fn = work_fn;
78 	pctl->error = 0;
79 	pctl->mp = mp;
80 	atomic_set(&pctl->nr_work, 0);
81 	init_waitqueue_head(&pctl->poll_wait);
82 
83 	return 0;
84 }
85 
86 /* Queue some parallel work. */
87 void
88 xfs_pwork_queue(
89 	struct xfs_pwork_ctl	*pctl,
90 	struct xfs_pwork	*pwork)
91 {
92 	INIT_WORK(&pwork->work, xfs_pwork_work);
93 	pwork->pctl = pctl;
94 	atomic_inc(&pctl->nr_work);
95 	queue_work(pctl->wq, &pwork->work);
96 }
97 
98 /* Wait for the work to finish and tear down the control structure. */
99 int
100 xfs_pwork_destroy(
101 	struct xfs_pwork_ctl	*pctl)
102 {
103 	destroy_workqueue(pctl->wq);
104 	pctl->wq = NULL;
105 	return pctl->error;
106 }
107 
108 /*
109  * Wait for the work to finish by polling completion status and touch the soft
110  * lockup watchdog.  This is for callers such as mount which hold locks.
111  */
112 void
113 xfs_pwork_poll(
114 	struct xfs_pwork_ctl	*pctl)
115 {
116 	while (wait_event_timeout(pctl->poll_wait,
117 				atomic_read(&pctl->nr_work) == 0, HZ) == 0)
118 		touch_softlockup_watchdog();
119 }
120 
121 /*
122  * Return the amount of parallelism that the data device can handle, or 0 for
123  * no limit.
124  */
125 unsigned int
126 xfs_pwork_guess_datadev_parallelism(
127 	struct xfs_mount	*mp)
128 {
129 	struct xfs_buftarg	*btp = mp->m_ddev_targp;
130 
131 	/*
132 	 * For now we'll go with the most conservative setting possible,
133 	 * which is two threads for an SSD and 1 thread everywhere else.
134 	 */
135 	return blk_queue_nonrot(btp->bt_bdev->bd_queue) ? 2 : 1;
136 }
137