[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/gfp.h>
#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>

int set_task_ioprio(struct task_struct *task, int ioprio)
{
	int err;
	struct io_context *ioc;
	const struct cred *cred = current_cred(), *tcred;

	rcu_read_lock();
	tcred = __task_cred(task);
	if (tcred->uid != cred->euid &&
	    tcred->uid != cred->uid && !capable(CAP_SYS_NICE)) {
		rcu_read_unlock();
		return -EPERM;
	}
	rcu_read_unlock();

	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;
}
EXPORT_SYMBOL_GPL(set_task_ioprio);

SYSCALL_DEFINE3(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:
			rcu_read_lock();
			if (!who)
				p = current;
			else
				p = find_task_by_vpid(who);
			if (p)
				ret = set_task_ioprio(p, ioprio);
			rcu_read_unlock();
			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_user();
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				int match;

				rcu_read_lock();
				match = __task_cred(p)->uid == who;
				rcu_read_unlock();
				if (!match)
					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;
}

SYSCALL_DEFINE2(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:
			rcu_read_lock();
			if (!who)
				p = current;
			else
				p = find_task_by_vpid(who);
			if (p)
				ret = get_task_ioprio(p);
			rcu_read_unlock();
			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_user();
			else
				user = find_user(who);

			if (!user)
				break;

			do_each_thread(g, p) {
				int match;

				rcu_read_lock();
				match = __task_cred(p)->uid == user->uid;
				rcu_read_unlock();
				if (!match)
					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
	1	/*
	2	* fs/ioprio.c
	3	*
	4	* Copyright (C) 2004 Jens Axboe <axboe@kernel.dk>
	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/gfp.h>
	23	#include <linux/kernel.h>
	24	#include <linux/ioprio.h>
	25	#include <linux/blkdev.h>
	26	#include <linux/capability.h>
	27	#include <linux/syscalls.h>
	28	#include <linux/security.h>
	29	#include <linux/pid_namespace.h>
	30
	31	int set_task_ioprio(struct task_struct *task, int ioprio)
	32	{
	33	int err;
	34	struct io_context *ioc;
	35	const struct cred cred = current_cred(), tcred;
	36
	37	rcu_read_lock();
	38	tcred = __task_cred(task);
	39	if (tcred->uid != cred->euid &&
	40	tcred->uid != cred->uid && !capable(CAP_SYS_NICE)) {
	41	rcu_read_unlock();
	42	return -EPERM;
	43	}
	44	rcu_read_unlock();
	45
	46	err = security_task_setioprio(task, ioprio);
	47	if (err)
	48	return err;
	49
	50	task_lock(task);
	51	do {
	52	ioc = task->io_context;
	53	/* see wmb() in current_io_context() */
	54	smp_read_barrier_depends();
	55	if (ioc)
	56	break;
	57
	58	ioc = alloc_io_context(GFP_ATOMIC, -1);
	59	if (!ioc) {
	60	err = -ENOMEM;
	61	break;
	62	}
	63	task->io_context = ioc;
	64	} while (1);
	65
	66	if (!err) {
	67	ioc->ioprio = ioprio;
	68	ioc->ioprio_changed = 1;
	69	}
	70
	71	task_unlock(task);
	72	return err;
	73	}
	74	EXPORT_SYMBOL_GPL(set_task_ioprio);
	75
	76	SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio)
	77	{
	78	int class = IOPRIO_PRIO_CLASS(ioprio);
	79	int data = IOPRIO_PRIO_DATA(ioprio);
	80	struct task_struct p, g;
	81	struct user_struct *user;
	82	struct pid *pgrp;
	83	int ret;
	84
	85	switch (class) {
	86	case IOPRIO_CLASS_RT:
	87	if (!capable(CAP_SYS_ADMIN))
	88	return -EPERM;
	89	/* fall through, rt has prio field too */
	90	case IOPRIO_CLASS_BE:
	91	if (data >= IOPRIO_BE_NR \|\| data < 0)
	92	return -EINVAL;
	93
	94	break;
	95	case IOPRIO_CLASS_IDLE:
	96	break;
	97	case IOPRIO_CLASS_NONE:
	98	if (data)
	99	return -EINVAL;
	100	break;
	101	default:
	102	return -EINVAL;
	103	}
	104
	105	ret = -ESRCH;
	106	/*
	107	* We want IOPRIO_WHO_PGRP/IOPRIO_WHO_USER to be "atomic",
	108	* so we can't use rcu_read_lock(). See re-copy of ->ioprio
	109	* in copy_process().
	110	*/
	111	read_lock(&tasklist_lock);
	112	switch (which) {
	113	case IOPRIO_WHO_PROCESS:
	114	rcu_read_lock();
	115	if (!who)
	116	p = current;
	117	else
	118	p = find_task_by_vpid(who);
	119	if (p)
	120	ret = set_task_ioprio(p, ioprio);
	121	rcu_read_unlock();
	122	break;
	123	case IOPRIO_WHO_PGRP:
	124	if (!who)
	125	pgrp = task_pgrp(current);
	126	else
	127	pgrp = find_vpid(who);
	128	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
	129	ret = set_task_ioprio(p, ioprio);
	130	if (ret)
	131	break;
	132	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
	133	break;
	134	case IOPRIO_WHO_USER:
	135	if (!who)
	136	user = current_user();
	137	else
	138	user = find_user(who);
	139
	140	if (!user)
	141	break;
	142
	143	do_each_thread(g, p) {
	144	int match;
	145
	146	rcu_read_lock();
	147	match = __task_cred(p)->uid == who;
	148	rcu_read_unlock();
	149	if (!match)
	150	continue;
	151	ret = set_task_ioprio(p, ioprio);
	152	if (ret)
	153	goto free_uid;
	154	} while_each_thread(g, p);
	155	free_uid:
	156	if (who)
	157	free_uid(user);
	158	break;
	159	default:
	160	ret = -EINVAL;
	161	}
	162
	163	read_unlock(&tasklist_lock);
	164	return ret;
	165	}
	166
	167	static int get_task_ioprio(struct task_struct *p)
	168	{
	169	int ret;
	170
	171	ret = security_task_getioprio(p);
	172	if (ret)
	173	goto out;
	174	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	175	if (p->io_context)
	176	ret = p->io_context->ioprio;
	177	out:
	178	return ret;
	179	}
	180
	181	int ioprio_best(unsigned short aprio, unsigned short bprio)
	182	{
	183	unsigned short aclass = IOPRIO_PRIO_CLASS(aprio);
	184	unsigned short bclass = IOPRIO_PRIO_CLASS(bprio);
	185
	186	if (aclass == IOPRIO_CLASS_NONE)
	187	aclass = IOPRIO_CLASS_BE;
	188	if (bclass == IOPRIO_CLASS_NONE)
	189	bclass = IOPRIO_CLASS_BE;
	190
	191	if (aclass == bclass)
	192	return min(aprio, bprio);
	193	if (aclass > bclass)
	194	return bprio;
	195	else
	196	return aprio;
	197	}
	198
	199	SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
	200	{
	201	struct task_struct g, p;
	202	struct user_struct *user;
	203	struct pid *pgrp;
	204	int ret = -ESRCH;
	205	int tmpio;
	206
	207	read_lock(&tasklist_lock);
	208	switch (which) {
	209	case IOPRIO_WHO_PROCESS:
	210	rcu_read_lock();
	211	if (!who)
	212	p = current;
	213	else
	214	p = find_task_by_vpid(who);
	215	if (p)
	216	ret = get_task_ioprio(p);
	217	rcu_read_unlock();
	218	break;
	219	case IOPRIO_WHO_PGRP:
	220	if (!who)
	221	pgrp = task_pgrp(current);
	222	else
	223	pgrp = find_vpid(who);
	224	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
	225	tmpio = get_task_ioprio(p);
	226	if (tmpio < 0)
	227	continue;
	228	if (ret == -ESRCH)
	229	ret = tmpio;
	230	else
	231	ret = ioprio_best(ret, tmpio);
	232	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
	233	break;
	234	case IOPRIO_WHO_USER:
	235	if (!who)
	236	user = current_user();
	237	else
	238	user = find_user(who);
	239
	240	if (!user)
	241	break;
	242
	243	do_each_thread(g, p) {
	244	int match;
	245
	246	rcu_read_lock();
	247	match = __task_cred(p)->uid == user->uid;
	248	rcu_read_unlock();
	249	if (!match)
	250	continue;
	251	tmpio = get_task_ioprio(p);
	252	if (tmpio < 0)
	253	continue;
	254	if (ret == -ESRCH)
	255	ret = tmpio;
	256	else
	257	ret = ioprio_best(ret, tmpio);
	258	} while_each_thread(g, p);
	259
	260	if (who)
	261	free_uid(user);
	262	break;
	263	default:
	264	ret = -EINVAL;
	265	}
	266
	267	read_unlock(&tasklist_lock);
	268	return ret;
	269	}