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 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
	task_lock(task);
52 53 54 55 56 57
	do {
		ioc = task->io_context;
		/* see wmb() in current_io_context() */
		smp_read_barrier_depends();
		if (ioc)
			break;
58

59 60 61 62 63 64 65
		ioc = alloc_io_context(GFP_ATOMIC, -1);
		if (!ioc) {
			err = -ENOMEM;
			break;
		}
		task->io_context = ioc;
	} while (1);
66

67 68
	if (!err) {
		ioc->ioprio = ioprio;
69
		ioc->ioprio_changed = 1;
70
	}
71 72

	task_unlock(task);
73
	return err;
74
}
75
EXPORT_SYMBOL_GPL(set_task_ioprio);
76

77
SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio)
78 79 80 81 82
{
	int class = IOPRIO_PRIO_CLASS(ioprio);
	int data = IOPRIO_PRIO_DATA(ioprio);
	struct task_struct *p, *g;
	struct user_struct *user;
83
	struct pid *pgrp;
84 85 86 87 88 89 90 91 92 93 94 95 96 97
	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;
98 99 100 101
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
102 103 104 105 106
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
107
	rcu_read_lock();
108 109 110 111 112
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
113
				p = find_task_by_vpid(who);
114 115 116 117 118
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
119 120
				pgrp = task_pgrp(current);
			else
121
				pgrp = find_vpid(who);
122
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
123 124 125
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
126
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
127 128 129
			break;
		case IOPRIO_WHO_USER:
			if (!who)
130
				user = current_user();
131 132 133 134 135 136 137
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
138
				if (__task_cred(p)->uid != who)
139 140 141
					continue;
				ret = set_task_ioprio(p, ioprio);
				if (ret)
142
					goto free_uid;
143
			} while_each_thread(g, p);
144
free_uid:
145 146 147 148 149 150 151
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

152
	rcu_read_unlock();
153 154 155
	return ret;
}

156 157 158 159 160 161 162
static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
163 164 165
	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	if (p->io_context)
		ret = p->io_context->ioprio;
166 167 168 169
out:
	return ret;
}

O
Oleg Nesterov 已提交
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
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;
}

188
SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
189 190 191
{
	struct task_struct *g, *p;
	struct user_struct *user;
192
	struct pid *pgrp;
193
	int ret = -ESRCH;
194
	int tmpio;
195

196
	rcu_read_lock();
197 198 199 200 201
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
202
				p = find_task_by_vpid(who);
203
			if (p)
204
				ret = get_task_ioprio(p);
205 206 207
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
208 209
				pgrp = task_pgrp(current);
			else
210
				pgrp = find_vpid(who);
211
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
212 213 214
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
215
				if (ret == -ESRCH)
216
					ret = tmpio;
217
				else
218
					ret = ioprio_best(ret, tmpio);
219
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
220 221 222
			break;
		case IOPRIO_WHO_USER:
			if (!who)
223
				user = current_user();
224 225 226 227 228 229 230
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
231
				if (__task_cred(p)->uid != user->uid)
232
					continue;
233 234 235
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
236
				if (ret == -ESRCH)
237
					ret = tmpio;
238
				else
239
					ret = ioprio_best(ret, tmpio);
240 241 242 243 244 245 246 247 248
			} while_each_thread(g, p);

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

249
	rcu_read_unlock();
250 251
	return ret;
}