[mirror_ubuntu-artful-kernel.git] / block / 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/export.h>
#include <linux/ioprio.h>
#include <linux/blkdev.h>
#include <linux/capability.h>
#include <linux/sched/user.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 (!uid_eq(tcred->uid, cred->euid) &&
	    !uid_eq(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;

	ioc = get_task_io_context(task, GFP_ATOMIC, NUMA_NO_NODE);
	if (ioc) {
		ioc->ioprio = ioprio;
		put_io_context(ioc);
	}

	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;
	kuid_t uid;
	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;
	rcu_read_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:
			uid = make_kuid(current_user_ns(), who);
			if (!uid_valid(uid))
				break;
			if (!who)
				user = current_user();
			else
				user = find_user(uid);

			if (!user)
				break;

			for_each_process_thread(g, p) {
				if (!uid_eq(task_uid(p), uid) ||
				    !task_pid_vnr(p))
					continue;
				ret = set_task_ioprio(p, ioprio);
				if (ret)
					goto free_uid;
			}
free_uid:
			if (who)
				free_uid(user);
			break;
		default:
			ret = -EINVAL;
	}

	rcu_read_unlock();
	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);
	task_lock(p);
	if (p->io_context)
		ret = p->io_context->ioprio;
	task_unlock(p);
out:
	return ret;
}

int ioprio_best(unsigned short aprio, unsigned short bprio)
{
	unsigned short aclass;
	unsigned short bclass;

	if (!ioprio_valid(aprio))
		aprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);
	if (!ioprio_valid(bprio))
		bprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);

	aclass = IOPRIO_PRIO_CLASS(aprio);
	bclass = IOPRIO_PRIO_CLASS(bprio);
	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;
	kuid_t uid;
	int ret = -ESRCH;
	int tmpio;

	rcu_read_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:
			uid = make_kuid(current_user_ns(), who);
			if (!who)
				user = current_user();
			else
				user = find_user(uid);

			if (!user)
				break;

			for_each_process_thread(g, p) {
				if (!uid_eq(task_uid(p), user->uid) ||
				    !task_pid_vnr(p))
					continue;
				tmpio = get_task_ioprio(p);
				if (tmpio < 0)
					continue;
				if (ret == -ESRCH)
					ret = tmpio;
				else
					ret = ioprio_best(ret, tmpio);
			}

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

	rcu_read_unlock();
	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/export.h>
	25	#include <linux/ioprio.h>
	26	#include <linux/blkdev.h>
	27	#include <linux/capability.h>
	28	#include <linux/sched/user.h>
	29	#include <linux/syscalls.h>
	30	#include <linux/security.h>
	31	#include <linux/pid_namespace.h>
	32
	33	int set_task_ioprio(struct task_struct *task, int ioprio)
	34	{
	35	int err;
	36	struct io_context *ioc;
	37	const struct cred cred = current_cred(), tcred;
	38
	39	rcu_read_lock();
	40	tcred = __task_cred(task);
	41	if (!uid_eq(tcred->uid, cred->euid) &&
	42	!uid_eq(tcred->uid, cred->uid) && !capable(CAP_SYS_NICE)) {
	43	rcu_read_unlock();
	44	return -EPERM;
	45	}
	46	rcu_read_unlock();
	47
	48	err = security_task_setioprio(task, ioprio);
	49	if (err)
	50	return err;
	51
	52	ioc = get_task_io_context(task, GFP_ATOMIC, NUMA_NO_NODE);
	53	if (ioc) {
	54	ioc->ioprio = ioprio;
	55	put_io_context(ioc);
	56	}
	57
	58	return err;
	59	}
	60	EXPORT_SYMBOL_GPL(set_task_ioprio);
	61
	62	SYSCALL_DEFINE3(ioprio_set, int, which, int, who, int, ioprio)
	63	{
	64	int class = IOPRIO_PRIO_CLASS(ioprio);
	65	int data = IOPRIO_PRIO_DATA(ioprio);
	66	struct task_struct p, g;
	67	struct user_struct *user;
	68	struct pid *pgrp;
	69	kuid_t uid;
	70	int ret;
	71
	72	switch (class) {
	73	case IOPRIO_CLASS_RT:
	74	if (!capable(CAP_SYS_ADMIN))
	75	return -EPERM;
	76	/* fall through, rt has prio field too */
	77	case IOPRIO_CLASS_BE:
	78	if (data >= IOPRIO_BE_NR \|\| data < 0)
	79	return -EINVAL;
	80
	81	break;
	82	case IOPRIO_CLASS_IDLE:
	83	break;
	84	case IOPRIO_CLASS_NONE:
	85	if (data)
	86	return -EINVAL;
	87	break;
	88	default:
	89	return -EINVAL;
	90	}
	91
	92	ret = -ESRCH;
	93	rcu_read_lock();
	94	switch (which) {
	95	case IOPRIO_WHO_PROCESS:
	96	if (!who)
	97	p = current;
	98	else
	99	p = find_task_by_vpid(who);
	100	if (p)
	101	ret = set_task_ioprio(p, ioprio);
	102	break;
	103	case IOPRIO_WHO_PGRP:
	104	if (!who)
	105	pgrp = task_pgrp(current);
	106	else
	107	pgrp = find_vpid(who);
	108	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
	109	ret = set_task_ioprio(p, ioprio);
	110	if (ret)
	111	break;
	112	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
	113	break;
	114	case IOPRIO_WHO_USER:
	115	uid = make_kuid(current_user_ns(), who);
	116	if (!uid_valid(uid))
	117	break;
	118	if (!who)
	119	user = current_user();
	120	else
	121	user = find_user(uid);
	122
	123	if (!user)
	124	break;
	125
	126	for_each_process_thread(g, p) {
	127	if (!uid_eq(task_uid(p), uid) \|\|
	128	!task_pid_vnr(p))
	129	continue;
	130	ret = set_task_ioprio(p, ioprio);
	131	if (ret)
	132	goto free_uid;
	133	}
	134	free_uid:
	135	if (who)
	136	free_uid(user);
	137	break;
	138	default:
	139	ret = -EINVAL;
	140	}
	141
	142	rcu_read_unlock();
	143	return ret;
	144	}
	145
	146	static int get_task_ioprio(struct task_struct *p)
	147	{
	148	int ret;
	149
	150	ret = security_task_getioprio(p);
	151	if (ret)
	152	goto out;
	153	ret = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_NONE, IOPRIO_NORM);
	154	task_lock(p);
	155	if (p->io_context)
	156	ret = p->io_context->ioprio;
	157	task_unlock(p);
	158	out:
	159	return ret;
	160	}
	161
	162	int ioprio_best(unsigned short aprio, unsigned short bprio)
	163	{
	164	unsigned short aclass;
	165	unsigned short bclass;
	166
	167	if (!ioprio_valid(aprio))
	168	aprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);
	169	if (!ioprio_valid(bprio))
	170	bprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, IOPRIO_NORM);
	171
	172	aclass = IOPRIO_PRIO_CLASS(aprio);
	173	bclass = IOPRIO_PRIO_CLASS(bprio);
	174	if (aclass == bclass)
	175	return min(aprio, bprio);
	176	if (aclass > bclass)
	177	return bprio;
	178	else
	179	return aprio;
	180	}
	181
	182	SYSCALL_DEFINE2(ioprio_get, int, which, int, who)
	183	{
	184	struct task_struct g, p;
	185	struct user_struct *user;
	186	struct pid *pgrp;
	187	kuid_t uid;
	188	int ret = -ESRCH;
	189	int tmpio;
	190
	191	rcu_read_lock();
	192	switch (which) {
	193	case IOPRIO_WHO_PROCESS:
	194	if (!who)
	195	p = current;
	196	else
	197	p = find_task_by_vpid(who);
	198	if (p)
	199	ret = get_task_ioprio(p);
	200	break;
	201	case IOPRIO_WHO_PGRP:
	202	if (!who)
	203	pgrp = task_pgrp(current);
	204	else
	205	pgrp = find_vpid(who);
	206	do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
	207	tmpio = get_task_ioprio(p);
	208	if (tmpio < 0)
	209	continue;
	210	if (ret == -ESRCH)
	211	ret = tmpio;
	212	else
	213	ret = ioprio_best(ret, tmpio);
	214	} while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
	215	break;
	216	case IOPRIO_WHO_USER:
	217	uid = make_kuid(current_user_ns(), who);
	218	if (!who)
	219	user = current_user();
	220	else
	221	user = find_user(uid);
	222
	223	if (!user)
	224	break;
	225
	226	for_each_process_thread(g, p) {
	227	if (!uid_eq(task_uid(p), user->uid) \|\|
	228	!task_pid_vnr(p))
	229	continue;
	230	tmpio = get_task_ioprio(p);
	231	if (tmpio < 0)
	232	continue;
	233	if (ret == -ESRCH)
	234	ret = tmpio;
	235	else
	236	ret = ioprio_best(ret, tmpio);
	237	}
	238
	239	if (who)
	240	free_uid(user);
	241	break;
	242	default:
	243	ret = -EINVAL;
	244	}
	245
	246	rcu_read_unlock();
	247	return ret;
	248	}