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 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
	struct io_context *ioc;

35 36
	if (task->uid != current_euid() &&
	    task->uid != current_uid() && !capable(CAP_SYS_NICE))
37 38
		return -EPERM;

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

43
	task_lock(task);
44 45 46 47 48 49
	do {
		ioc = task->io_context;
		/* see wmb() in current_io_context() */
		smp_read_barrier_depends();
		if (ioc)
			break;
50

51 52 53 54 55 56 57
		ioc = alloc_io_context(GFP_ATOMIC, -1);
		if (!ioc) {
			err = -ENOMEM;
			break;
		}
		task->io_context = ioc;
	} while (1);
58

59 60
	if (!err) {
		ioc->ioprio = ioprio;
61
		ioc->ioprio_changed = 1;
62
	}
63 64

	task_unlock(task);
65
	return err;
66 67
}

68
asmlinkage long sys_ioprio_set(int which, int who, int ioprio)
69 70 71 72 73
{
	int class = IOPRIO_PRIO_CLASS(ioprio);
	int data = IOPRIO_PRIO_DATA(ioprio);
	struct task_struct *p, *g;
	struct user_struct *user;
74
	struct pid *pgrp;
75 76 77 78 79 80 81 82 83 84 85 86 87 88
	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;
89 90 91 92
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
93 94 95 96 97
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
98 99 100 101 102 103
	/*
	 * 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);
104 105 106 107 108
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
109
				p = find_task_by_vpid(who);
110 111 112 113 114
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
115 116
				pgrp = task_pgrp(current);
			else
117
				pgrp = find_vpid(who);
118
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
119 120 121
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
122
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
			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)
138
					goto free_uid;
139
			} while_each_thread(g, p);
140
free_uid:
141 142 143 144 145 146 147
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

148
	read_unlock(&tasklist_lock);
149 150 151
	return ret;
}

152 153 154 155 156 157 158
static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
159 160 161
	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	if (p->io_context)
		ret = p->io_context->ioprio;
162 163 164 165
out:
	return ret;
}

O
Oleg Nesterov 已提交
166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
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;
}

184
asmlinkage long sys_ioprio_get(int which, int who)
185 186 187
{
	struct task_struct *g, *p;
	struct user_struct *user;
188
	struct pid *pgrp;
189
	int ret = -ESRCH;
190
	int tmpio;
191

192
	read_lock(&tasklist_lock);
193 194 195 196 197
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
198
				p = find_task_by_vpid(who);
199
			if (p)
200
				ret = get_task_ioprio(p);
201 202 203
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
204 205
				pgrp = task_pgrp(current);
			else
206
				pgrp = find_vpid(who);
207
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
208 209 210
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
211
				if (ret == -ESRCH)
212
					ret = tmpio;
213
				else
214
					ret = ioprio_best(ret, tmpio);
215
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
216 217 218 219 220 221 222 223 224 225 226 227 228
			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;
229 230 231
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
232
				if (ret == -ESRCH)
233
					ret = tmpio;
234
				else
235
					ret = ioprio_best(ret, tmpio);
236 237 238 239 240 241 242 243 244
			} while_each_thread(g, p);

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

245
	read_unlock(&tasklist_lock);
246 247 248
	return ret;
}