ioprio.c 4.8 KB
Newer Older
1 2 3
/*
 * fs/ioprio.c
 *
4
 * Copyright (C) 2004 Jens Axboe <axboe@kernel.dk>
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
 *
 * Helper functions for setting/querying io priorities of processes. The
 * system calls closely mimmick getpriority/setpriority, see the man page for
 * those. The prio argument is a composite of prio class and prio data, where
 * the data argument has meaning within that class. The standard scheduling
 * classes have 8 distinct prio levels, with 0 being the highest prio and 7
 * being the lowest.
 *
 * IOW, setting BE scheduling class with prio 2 is done ala:
 *
 * unsigned int prio = (IOPRIO_CLASS_BE << IOPRIO_CLASS_SHIFT) | 2;
 *
 * ioprio_set(PRIO_PROCESS, pid, prio);
 *
 * See also Documentation/block/ioprio.txt
 *
 */
22
#include <linux/gfp.h>
23
#include <linux/kernel.h>
24
#include <linux/export.h>
25 26
#include <linux/ioprio.h>
#include <linux/blkdev.h>
27
#include <linux/capability.h>
28
#include <linux/syscalls.h>
29
#include <linux/security.h>
30
#include <linux/pid_namespace.h>
31

32
int set_task_ioprio(struct task_struct *task, int ioprio)
33
{
34
	int err;
35
	struct io_context *ioc;
36
	const struct cred *cred = current_cred(), *tcred;
37

38 39 40 41 42
	rcu_read_lock();
	tcred = __task_cred(task);
	if (tcred->uid != cred->euid &&
	    tcred->uid != cred->uid && !capable(CAP_SYS_NICE)) {
		rcu_read_unlock();
43
		return -EPERM;
44 45
	}
	rcu_read_unlock();
46

47 48 49 50
	err = security_task_setioprio(task, ioprio);
	if (err)
		return err;

51 52
	ioc = get_task_io_context(task, GFP_ATOMIC, NUMA_NO_NODE);
	if (ioc) {
53
		ioc_ioprio_changed(ioc, ioprio);
54
		put_io_context(ioc);
55
	}
56

57
	return err;
58
}
59
EXPORT_SYMBOL_GPL(set_task_ioprio);
60

61
SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio)
62 63 64 65 66
{
	int class = IOPRIO_PRIO_CLASS(ioprio);
	int data = IOPRIO_PRIO_DATA(ioprio);
	struct task_struct *p, *g;
	struct user_struct *user;
67
	struct pid *pgrp;
68 69 70 71 72 73 74 75 76 77 78 79 80 81
	int ret;

	switch (class) {
		case IOPRIO_CLASS_RT:
			if (!capable(CAP_SYS_ADMIN))
				return -EPERM;
			/* fall through, rt has prio field too */
		case IOPRIO_CLASS_BE:
			if (data >= IOPRIO_BE_NR || data < 0)
				return -EINVAL;

			break;
		case IOPRIO_CLASS_IDLE:
			break;
82 83 84 85
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
86 87 88 89 90
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
91
	rcu_read_lock();
92 93 94 95 96
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
97
				p = find_task_by_vpid(who);
98 99 100 101 102
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
103 104
				pgrp = task_pgrp(current);
			else
105
				pgrp = find_vpid(who);
106
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
107 108 109
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
110
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
111 112 113
			break;
		case IOPRIO_WHO_USER:
			if (!who)
114
				user = current_user();
115 116 117 118 119 120 121
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
122
				if (__task_cred(p)->uid != who)
123 124 125
					continue;
				ret = set_task_ioprio(p, ioprio);
				if (ret)
126
					goto free_uid;
127
			} while_each_thread(g, p);
128
free_uid:
129 130 131 132 133 134 135
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

136
	rcu_read_unlock();
137 138 139
	return ret;
}

140 141 142 143 144 145 146
static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
147 148 149
	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	if (p->io_context)
		ret = p->io_context->ioprio;
150 151 152 153
out:
	return ret;
}

O
Oleg Nesterov 已提交
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
int ioprio_best(unsigned short aprio, unsigned short bprio)
{
	unsigned short aclass = IOPRIO_PRIO_CLASS(aprio);
	unsigned short bclass = IOPRIO_PRIO_CLASS(bprio);

	if (aclass == IOPRIO_CLASS_NONE)
		aclass = IOPRIO_CLASS_BE;
	if (bclass == IOPRIO_CLASS_NONE)
		bclass = IOPRIO_CLASS_BE;

	if (aclass == bclass)
		return min(aprio, bprio);
	if (aclass > bclass)
		return bprio;
	else
		return aprio;
}

172
SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
173 174 175
{
	struct task_struct *g, *p;
	struct user_struct *user;
176
	struct pid *pgrp;
177
	int ret = -ESRCH;
178
	int tmpio;
179

180
	rcu_read_lock();
181 182 183 184 185
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
186
				p = find_task_by_vpid(who);
187
			if (p)
188
				ret = get_task_ioprio(p);
189 190 191
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
192 193
				pgrp = task_pgrp(current);
			else
194
				pgrp = find_vpid(who);
195
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
196 197 198
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
199
				if (ret == -ESRCH)
200
					ret = tmpio;
201
				else
202
					ret = ioprio_best(ret, tmpio);
203
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
204 205 206
			break;
		case IOPRIO_WHO_USER:
			if (!who)
207
				user = current_user();
208 209 210 211 212 213 214
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
215
				if (__task_cred(p)->uid != user->uid)
216
					continue;
217 218 219
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
220
				if (ret == -ESRCH)
221
					ret = tmpio;
222
				else
223
					ret = ioprio_best(ret, tmpio);
224 225 226 227 228 229 230 231 232
			} while_each_thread(g, p);

			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

233
	rcu_read_unlock();
234 235
	return ret;
}