ioprio.c 5.0 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
	rcu_read_lock();
	tcred = __task_cred(task);
40 41
	if (!uid_eq(tcred->uid, cred->euid) &&
	    !uid_eq(tcred->uid, cred->uid) && !capable(CAP_SYS_NICE)) {
42
		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) {
T
Tejun Heo 已提交
53
		ioc->ioprio = 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
	kuid_t uid;
69 70 71 72 73 74 75 76 77 78 79 80 81 82
	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;
83 84 85 86
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
87 88 89 90 91
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
92
	rcu_read_lock();
93 94 95 96 97
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
98
				p = find_task_by_vpid(who);
99 100 101 102 103
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
104 105
				pgrp = task_pgrp(current);
			else
106
				pgrp = find_vpid(who);
107
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
108 109 110
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
111
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
112 113
			break;
		case IOPRIO_WHO_USER:
114 115 116
			uid = make_kuid(current_user_ns(), who);
			if (!uid_valid(uid))
				break;
117
			if (!who)
118
				user = current_user();
119
			else
120
				user = find_user(uid);
121 122 123 124 125

			if (!user)
				break;

			do_each_thread(g, p) {
126
				if (!uid_eq(task_uid(p), uid))
127 128 129
					continue;
				ret = set_task_ioprio(p, ioprio);
				if (ret)
130
					goto free_uid;
131
			} while_each_thread(g, p);
132
free_uid:
133 134 135 136 137 138 139
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

140
	rcu_read_unlock();
141 142 143
	return ret;
}

144 145 146 147 148 149 150
static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
151 152 153
	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	if (p->io_context)
		ret = p->io_context->ioprio;
154 155 156 157
out:
	return ret;
}

O
Oleg Nesterov 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
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;
}

176
SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
177 178 179
{
	struct task_struct *g, *p;
	struct user_struct *user;
180
	struct pid *pgrp;
181
	kuid_t uid;
182
	int ret = -ESRCH;
183
	int tmpio;
184

185
	rcu_read_lock();
186 187 188 189 190
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
191
				p = find_task_by_vpid(who);
192
			if (p)
193
				ret = get_task_ioprio(p);
194 195 196
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
197 198
				pgrp = task_pgrp(current);
			else
199
				pgrp = find_vpid(who);
200
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
201 202 203
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
204
				if (ret == -ESRCH)
205
					ret = tmpio;
206
				else
207
					ret = ioprio_best(ret, tmpio);
208
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
209 210
			break;
		case IOPRIO_WHO_USER:
211
			uid = make_kuid(current_user_ns(), who);
212
			if (!who)
213
				user = current_user();
214
			else
215
				user = find_user(uid);
216 217 218 219 220

			if (!user)
				break;

			do_each_thread(g, p) {
221
				if (!uid_eq(task_uid(p), user->uid))
222
					continue;
223 224 225
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
226
				if (ret == -ESRCH)
227
					ret = tmpio;
228
				else
229
					ret = ioprio_best(ret, tmpio);
230 231 232 233 234 235 236 237 238
			} while_each_thread(g, p);

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

239
	rcu_read_unlock();
240 241
	return ret;
}