[net-next-2.6.git] / fs / ioprio.c

/*
 * fs/ioprio.c
 *
 * Copyright (C) 2004 Jens Axboe <axboe@kernel.dk>
 *
 * 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>
#include <linux/capability.h>
#include <linux/syscalls.h>
#include <linux/security.h>
#include <linux/pid_namespace.h>

static int set_task_ioprio(struct task_struct *task, int ioprio)
{
	int err;
	struct io_context *ioc;

	if (task->cred->uid != current_euid() &&
	    task->cred->uid != current_uid() && !capable(CAP_SYS_NICE))
		return -EPERM;

	err = security_task_setioprio(task, ioprio);
	if (err)
		return err;

	task_lock(task);
	do {
		ioc = task->io_context;
		/* see wmb() in current_io_context() */
		smp_read_barrier_depends();
		if (ioc)
			break;

		ioc = alloc_io_context(GFP_ATOMIC, -1);
		if (!ioc) {
			err = -ENOMEM;
			break;
		}
		task->io_context = ioc;
	} while (1);

	if (!err) {
		ioc->ioprio = ioprio;
		ioc->ioprio_changed = 1;
	}

	task_unlock(task);
	return err;
}

asmlinkage long sys_ioprio_set(int which, int who, int ioprio)
{
	int class = IOPRIO_PRIO_CLASS(ioprio);
	int data = IOPRIO_PRIO_DATA(ioprio);
	struct task_struct *p, *g;
	struct user_struct *user;
	struct pid *pgrp;
	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;
		case IOPRIO_CLASS_NONE:
			if (data)
				return -EINVAL;
			break;
		default:
			return -EINVAL;
	}

	ret = -ESRCH;
	/*
	 * 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);
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
				p = find_task_by_vpid(who);
			if (p)
				ret = set_task_ioprio(p, ioprio);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
				pgrp = task_pgrp(current);
			else
				pgrp = find_vpid(who);
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					break;
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
			break;
		case IOPRIO_WHO_USER:
			if (!who)
				user = current->cred->user;
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				if (p->cred->uid != who)
					continue;
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					goto free_uid;
			} while_each_thread(g, p);
free_uid:
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

	read_unlock(&tasklist_lock);
	return ret;
}

static int get_task_ioprio(struct task_struct *p)
{
	int ret;

	ret = security_task_getioprio(p);
	if (ret)
		goto out;
	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	if (p->io_context)
		ret = p->io_context->ioprio;
out:
	return ret;
}

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;
}

asmlinkage long sys_ioprio_get(int which, int who)
{
	struct task_struct *g, *p;
	struct user_struct *user;
	struct pid *pgrp;
	int ret = -ESRCH;
	int tmpio;

	read_lock(&tasklist_lock);
	switch (which) {
		case IOPRIO_WHO_PROCESS:
			if (!who)
				p = current;
			else
				p = find_task_by_vpid(who);
			if (p)
				ret = get_task_ioprio(p);
			break;
		case IOPRIO_WHO_PGRP:
			if (!who)
				pgrp = task_pgrp(current);
			else
				pgrp = find_vpid(who);
			do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
				if (ret == -ESRCH)
					ret = tmpio;
				else
					ret = ioprio_best(ret, tmpio);
			} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
			break;
		case IOPRIO_WHO_USER:
			if (!who)
				user = current->cred->user;
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				if (p->cred->uid != user->uid)
					continue;
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
				if (ret == -ESRCH)
					ret = tmpio;
				else
					ret = ioprio_best(ret, tmpio);
			} while_each_thread(g, p);

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

	read_unlock(&tasklist_lock);
	return ret;
}
Commit	Line	Data
22e2c507 JA	1	/*
	2	* fs/ioprio.c
	3	*
0fe23479	4	* Copyright (C) 2004 Jens Axboe <axboe@kernel.dk>
22e2c507 JA	5	*
	6	* Helper functions for setting/querying io priorities of processes. The
	7	* system calls closely mimmick getpriority/setpriority, see the man page for
	8	* those. The prio argument is a composite of prio class and prio data, where
	9	* the data argument has meaning within that class. The standard scheduling
	10	* classes have 8 distinct prio levels, with 0 being the highest prio and 7
	11	* being the lowest.
	12	*
	13	* IOW, setting BE scheduling class with prio 2 is done ala:
	14	*
	15	* unsigned int prio = (IOPRIO_CLASS_BE << IOPRIO_CLASS_SHIFT) \| 2;
	16	*
	17	* ioprio_set(PRIO_PROCESS, pid, prio);
	18	*
	19	* See also Documentation/block/ioprio.txt
	20	*
	21	*/
	22	#include <linux/kernel.h>
	23	#include <linux/ioprio.h>
	24	#include <linux/blkdev.h>
16f7e0fe	25	#include <linux/capability.h>
9abdc4cd	26	#include <linux/syscalls.h>
03e68060	27	#include <linux/security.h>
b488893a	28	#include <linux/pid_namespace.h>
22e2c507 JA	29
	30	static int set_task_ioprio(struct task_struct *task, int ioprio)
	31	{
03e68060	32	int err;
22e2c507 JA	33	struct io_context *ioc;
22e2c507 JA	34
b6dff3ec DH	35	if (task->cred->uid != current_euid() &&
b6dff3ec DH	36	task->cred->uid != current_uid() && !capable(CAP_SYS_NICE))
22e2c507 JA	37	return -EPERM;
22e2c507 JA	38
03e68060 JM	39	err = security_task_setioprio(task, ioprio);
	40	if (err)
	41	return err;
	42
22e2c507	43	task_lock(task);
fd0928df JA	44	do {
	45	ioc = task->io_context;
	46	/* see wmb() in current_io_context() */
	47	smp_read_barrier_depends();
	48	if (ioc)
	49	break;
22e2c507	50
fd0928df JA	51	ioc = alloc_io_context(GFP_ATOMIC, -1);
	52	if (!ioc) {
	53	err = -ENOMEM;
	54	break;
	55	}
	56	task->io_context = ioc;
fd0928df	57	} while (1);
9f83e45e	58
fd0928df JA	59	if (!err) {
fd0928df JA	60	ioc->ioprio = ioprio;
fc46379d	61	ioc->ioprio_changed = 1;
fd0928df	62	}
22e2c507 JA	63
22e2c507 JA	64	task_unlock(task);
fd0928df	65	return err;
22e2c507 JA	66	}
22e2c507 JA	67
cf366808	68	asmlinkage long sys_ioprio_set(int which, int who, int ioprio)
22e2c507 JA	69	{
	70	int class = IOPRIO_PRIO_CLASS(ioprio);
	71	int data = IOPRIO_PRIO_DATA(ioprio);
	72	struct task_struct p, g;
	73	struct user_struct *user;
41487c65	74	struct pid *pgrp;
22e2c507 JA	75	int ret;
	76
	77	switch (class) {
	78	case IOPRIO_CLASS_RT:
	79	if (!capable(CAP_SYS_ADMIN))
	80	return -EPERM;
	81	/* fall through, rt has prio field too */
	82	case IOPRIO_CLASS_BE:
	83	if (data >= IOPRIO_BE_NR \|\| data < 0)
	84	return -EINVAL;
	85
	86	break;
	87	case IOPRIO_CLASS_IDLE:
	88	break;
8ec680e4 JA	89	case IOPRIO_CLASS_NONE:
	90	if (data)
	91	return -EINVAL;
	92	break;
22e2c507 JA	93	default:
	94	return -EINVAL;
	95	}
	96
	97	ret = -ESRCH;
cf342e52 ON	98	/*
	99	* We want IOPRIO_WHO_PGRP/IOPRIO_WHO_USER to be "atomic",
	100	* so we can't use rcu_read_lock(). See re-copy of ->ioprio
	101	* in copy_process().
	102	*/
	103	read_lock(&tasklist_lock);
22e2c507 JA	104	switch (which) {
	105	case IOPRIO_WHO_PROCESS:
	106	if (!who)
	107	p = current;
	108	else
228ebcbe	109	p = find_task_by_vpid(who);
22e2c507 JA	110	if (p)
	111	ret = set_task_ioprio(p, ioprio);
	112	break;
	113	case IOPRIO_WHO_PGRP:
	114	if (!who)
41487c65 EB	115	pgrp = task_pgrp(current);
41487c65 EB	116	else
b488893a	117	pgrp = find_vpid(who);
2d70b68d	118	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
22e2c507 JA	119	ret = set_task_ioprio(p, ioprio);
	120	if (ret)
	121	break;
2d70b68d	122	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
22e2c507 JA	123	break;
	124	case IOPRIO_WHO_USER:
	125	if (!who)
b6dff3ec	126	user = current->cred->user;
22e2c507 JA	127	else
	128	user = find_user(who);
	129
	130	if (!user)
	131	break;
	132
	133	do_each_thread(g, p) {
b6dff3ec	134	if (p->cred->uid != who)
22e2c507 JA	135	continue;
	136	ret = set_task_ioprio(p, ioprio);
	137	if (ret)
78bd4d48	138	goto free_uid;
22e2c507	139	} while_each_thread(g, p);
78bd4d48	140	free_uid:
22e2c507 JA	141	if (who)
	142	free_uid(user);
	143	break;
	144	default:
	145	ret = -EINVAL;
	146	}
	147
cf342e52	148	read_unlock(&tasklist_lock);
22e2c507 JA	149	return ret;
	150	}
	151
a1836a42 DQ	152	static int get_task_ioprio(struct task_struct *p)
	153	{
	154	int ret;
	155
	156	ret = security_task_getioprio(p);
	157	if (ret)
	158	goto out;
fd0928df JA	159	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	160	if (p->io_context)
	161	ret = p->io_context->ioprio;
a1836a42 DQ	162	out:
	163	return ret;
	164	}
	165
e014ff8d ON	166	int ioprio_best(unsigned short aprio, unsigned short bprio)
	167	{
	168	unsigned short aclass = IOPRIO_PRIO_CLASS(aprio);
	169	unsigned short bclass = IOPRIO_PRIO_CLASS(bprio);
	170
e014ff8d ON	171	if (aclass == IOPRIO_CLASS_NONE)
	172	aclass = IOPRIO_CLASS_BE;
	173	if (bclass == IOPRIO_CLASS_NONE)
	174	bclass = IOPRIO_CLASS_BE;
	175
	176	if (aclass == bclass)
	177	return min(aprio, bprio);
	178	if (aclass > bclass)
	179	return bprio;
	180	else
	181	return aprio;
	182	}
	183
cf366808	184	asmlinkage long sys_ioprio_get(int which, int who)
22e2c507 JA	185	{
	186	struct task_struct g, p;
	187	struct user_struct *user;
41487c65	188	struct pid *pgrp;
22e2c507	189	int ret = -ESRCH;
a1836a42	190	int tmpio;
22e2c507	191
cf342e52	192	read_lock(&tasklist_lock);
22e2c507 JA	193	switch (which) {
	194	case IOPRIO_WHO_PROCESS:
	195	if (!who)
	196	p = current;
	197	else
228ebcbe	198	p = find_task_by_vpid(who);
22e2c507	199	if (p)
a1836a42	200	ret = get_task_ioprio(p);
22e2c507 JA	201	break;
	202	case IOPRIO_WHO_PGRP:
	203	if (!who)
41487c65 EB	204	pgrp = task_pgrp(current);
41487c65 EB	205	else
b488893a	206	pgrp = find_vpid(who);
2d70b68d	207	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
a1836a42 DQ	208	tmpio = get_task_ioprio(p);
	209	if (tmpio < 0)
	210	continue;
22e2c507	211	if (ret == -ESRCH)
a1836a42	212	ret = tmpio;
22e2c507	213	else
a1836a42	214	ret = ioprio_best(ret, tmpio);
2d70b68d	215	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
22e2c507 JA	216	break;
	217	case IOPRIO_WHO_USER:
	218	if (!who)
b6dff3ec	219	user = current->cred->user;
22e2c507 JA	220	else
	221	user = find_user(who);
	222
	223	if (!user)
	224	break;
	225
	226	do_each_thread(g, p) {
b6dff3ec	227	if (p->cred->uid != user->uid)
22e2c507	228	continue;
a1836a42 DQ	229	tmpio = get_task_ioprio(p);
	230	if (tmpio < 0)
	231	continue;
22e2c507	232	if (ret == -ESRCH)
a1836a42	233	ret = tmpio;
22e2c507	234	else
a1836a42	235	ret = ioprio_best(ret, tmpio);
22e2c507 JA	236	} while_each_thread(g, p);
	237
	238	if (who)
	239	free_uid(user);
	240	break;
	241	default:
	242	ret = -EINVAL;
	243	}
	244
cf342e52	245	read_unlock(&tasklist_lock);
22e2c507 JA	246	return ret;
	247	}
	248