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 22 23 24
 *
 * 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
 *
 */
#include <linux/kernel.h>
#include <linux/ioprio.h>
#include <linux/blkdev.h>
25
#include <linux/capability.h>
26
#include <linux/syscalls.h>
27
#include <linux/security.h>
28
#include <linux/pid_namespace.h>
29 30 31

static int set_task_ioprio(struct task_struct *task, int ioprio)
{
32
	int err;
33 34 35 36 37 38
	struct io_context *ioc;

	if (task->uid != current->euid &&
	    task->uid != current->uid && !capable(CAP_SYS_NICE))
		return -EPERM;

39 40 41 42
	err = security_task_setioprio(task, ioprio);
	if (err)
		return err;

43 44 45 46 47
	task_lock(task);

	task->ioprio = ioprio;

	ioc = task->io_context;
48 49 50
	/* see wmb() in current_io_context() */
	smp_read_barrier_depends();

51 52
	if (ioc)
		ioc->ioprio_changed = 1;
53 54 55 56 57

	task_unlock(task);
	return 0;
}

58
asmlinkage long sys_ioprio_set(int which, int who, int ioprio)
59 60 61 62 63
{
	int class = IOPRIO_PRIO_CLASS(ioprio);
	int data = IOPRIO_PRIO_DATA(ioprio);
	struct task_struct *p, *g;
	struct user_struct *user;
64
	struct pid *pgrp;
65 66 67 68 69 70 71 72 73 74 75 76 77
	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:
78 79
			if (!capable(CAP_SYS_ADMIN))
				return -EPERM;
80
			break;
81 82 83 84
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
85 86 87 88 89
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
90 91 92 93 94 95
	/*
	 * We want IOPRIO_WHO_PGRP/IOPRIO_WHO_USER to be "atomic",
	 * so we can't use rcu_read_lock(). See re-copy of ->ioprio
	 * in copy_process().
	 */
	read_lock(&tasklist_lock);
96 97 98 99 100
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
101
				p = find_task_by_vpid(who);
102 103 104 105 106
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
107 108
				pgrp = task_pgrp(current);
			else
109
				pgrp = find_vpid(who);
110
			do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
111 112 113
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
114
			} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
			break;
		case IOPRIO_WHO_USER:
			if (!who)
				user = current->user;
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				if (p->uid != who)
					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
	read_unlock(&tasklist_lock);
141 142 143
	return ret;
}

144 145 146 147 148 149 150 151 152 153 154 155
static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
	ret = p->ioprio;
out:
	return ret;
}

O
Oleg Nesterov 已提交
156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173
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;
}

174
asmlinkage long sys_ioprio_get(int which, int who)
175 176 177
{
	struct task_struct *g, *p;
	struct user_struct *user;
178
	struct pid *pgrp;
179
	int ret = -ESRCH;
180
	int tmpio;
181

182
	read_lock(&tasklist_lock);
183 184 185 186 187
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
188
				p = find_task_by_vpid(who);
189
			if (p)
190
				ret = get_task_ioprio(p);
191 192 193
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
194 195
				pgrp = task_pgrp(current);
			else
196
				pgrp = find_vpid(who);
197
			do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
198 199 200
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
201
				if (ret == -ESRCH)
202
					ret = tmpio;
203
				else
204
					ret = ioprio_best(ret, tmpio);
205
			} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
206 207 208 209 210 211 212 213 214 215 216 217 218
			break;
		case IOPRIO_WHO_USER:
			if (!who)
				user = current->user;
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				if (p->uid != user->uid)
					continue;
219 220 221
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
222
				if (ret == -ESRCH)
223
					ret = tmpio;
224
				else
225
					ret = ioprio_best(ret, tmpio);
226 227 228 229 230 231 232 233 234
			} while_each_thread(g, p);

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

235
	read_unlock(&tasklist_lock);
236 237 238
	return ret;
}