[mirror_ubuntu-hirsute-kernel.git] / kernel / cgroup / pids.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Process number limiting controller for cgroups.
 *
 * Used to allow a cgroup hierarchy to stop any new processes from fork()ing
 * after a certain limit is reached.
 *
 * Since it is trivial to hit the task limit without hitting any kmemcg limits
 * in place, PIDs are a fundamental resource. As such, PID exhaustion must be
 * preventable in the scope of a cgroup hierarchy by allowing resource limiting
 * of the number of tasks in a cgroup.
 *
 * In order to use the `pids` controller, set the maximum number of tasks in
 * pids.max (this is not available in the root cgroup for obvious reasons). The
 * number of processes currently in the cgroup is given by pids.current.
 * Organisational operations are not blocked by cgroup policies, so it is
 * possible to have pids.current > pids.max. However, it is not possible to
 * violate a cgroup policy through fork(). fork() will return -EAGAIN if forking
 * would cause a cgroup policy to be violated.
 *
 * To set a cgroup to have no limit, set pids.max to "max". This is the default
 * for all new cgroups (N.B. that PID limits are hierarchical, so the most
 * stringent limit in the hierarchy is followed).
 *
 * pids.current tracks all child cgroup hierarchies, so parent/pids.current is
 * a superset of parent/child/pids.current.
 *
 * Copyright (C) 2015 Aleksa Sarai <cyphar@cyphar.com>
 */

#include <linux/kernel.h>
#include <linux/threads.h>
#include <linux/atomic.h>
#include <linux/cgroup.h>
#include <linux/slab.h>

#define PIDS_MAX (PID_MAX_LIMIT + 1ULL)
#define PIDS_MAX_STR "max"

struct pids_cgroup {
	struct cgroup_subsys_state	css;

	/*
	 * Use 64-bit types so that we can safely represent "max" as
	 * %PIDS_MAX = (%PID_MAX_LIMIT + 1).
	 */
	atomic64_t			counter;
	int64_t				limit;

	/* Handle for "pids.events" */
	struct cgroup_file		events_file;

	/* Number of times fork failed because limit was hit. */
	atomic64_t			events_limit;
};

static struct pids_cgroup *css_pids(struct cgroup_subsys_state *css)
{
	return container_of(css, struct pids_cgroup, css);
}

static struct pids_cgroup *parent_pids(struct pids_cgroup *pids)
{
	return css_pids(pids->css.parent);
}

static struct cgroup_subsys_state *
pids_css_alloc(struct cgroup_subsys_state *parent)
{
	struct pids_cgroup *pids;

	pids = kzalloc(sizeof(struct pids_cgroup), GFP_KERNEL);
	if (!pids)
		return ERR_PTR(-ENOMEM);

	pids->limit = PIDS_MAX;
	atomic64_set(&pids->counter, 0);
	atomic64_set(&pids->events_limit, 0);
	return &pids->css;
}

static void pids_css_free(struct cgroup_subsys_state *css)
{
	kfree(css_pids(css));
}

/**
 * pids_cancel - uncharge the local pid count
 * @pids: the pid cgroup state
 * @num: the number of pids to cancel
 *
 * This function will WARN if the pid count goes under 0, because such a case is
 * a bug in the pids controller proper.
 */
static void pids_cancel(struct pids_cgroup *pids, int num)
{
	/*
	 * A negative count (or overflow for that matter) is invalid,
	 * and indicates a bug in the `pids` controller proper.
	 */
	WARN_ON_ONCE(atomic64_add_negative(-num, &pids->counter));
}

/**
 * pids_uncharge - hierarchically uncharge the pid count
 * @pids: the pid cgroup state
 * @num: the number of pids to uncharge
 */
static void pids_uncharge(struct pids_cgroup *pids, int num)
{
	struct pids_cgroup *p;

	for (p = pids; parent_pids(p); p = parent_pids(p))
		pids_cancel(p, num);
}

/**
 * pids_charge - hierarchically charge the pid count
 * @pids: the pid cgroup state
 * @num: the number of pids to charge
 *
 * This function does *not* follow the pid limit set. It cannot fail and the new
 * pid count may exceed the limit. This is only used for reverting failed
 * attaches, where there is no other way out than violating the limit.
 */
static void pids_charge(struct pids_cgroup *pids, int num)
{
	struct pids_cgroup *p;

	for (p = pids; parent_pids(p); p = parent_pids(p))
		atomic64_add(num, &p->counter);
}

/**
 * pids_try_charge - hierarchically try to charge the pid count
 * @pids: the pid cgroup state
 * @num: the number of pids to charge
 *
 * This function follows the set limit. It will fail if the charge would cause
 * the new value to exceed the hierarchical limit. Returns 0 if the charge
 * succeeded, otherwise -EAGAIN.
 */
static int pids_try_charge(struct pids_cgroup *pids, int num)
{
	struct pids_cgroup *p, *q;

	for (p = pids; parent_pids(p); p = parent_pids(p)) {
		int64_t new = atomic64_add_return(num, &p->counter);

		/*
		 * Since new is capped to the maximum number of pid_t, if
		 * p->limit is %PIDS_MAX then we know that this test will never
		 * fail.
		 */
		if (new > p->limit)
			goto revert;
	}

	return 0;

revert:
	for (q = pids; q != p; q = parent_pids(q))
		pids_cancel(q, num);
	pids_cancel(p, num);

	return -EAGAIN;
}

static int pids_can_attach(struct cgroup_taskset *tset)
{
	struct task_struct *task;
	struct cgroup_subsys_state *dst_css;

	cgroup_taskset_for_each(task, dst_css, tset) {
		struct pids_cgroup *pids = css_pids(dst_css);
		struct cgroup_subsys_state *old_css;
		struct pids_cgroup *old_pids;

		/*
		 * No need to pin @old_css between here and cancel_attach()
		 * because cgroup core protects it from being freed before
		 * the migration completes or fails.
		 */
		old_css = task_css(task, pids_cgrp_id);
		old_pids = css_pids(old_css);

		pids_charge(pids, 1);
		pids_uncharge(old_pids, 1);
	}

	return 0;
}

static void pids_cancel_attach(struct cgroup_taskset *tset)
{
	struct task_struct *task;
	struct cgroup_subsys_state *dst_css;

	cgroup_taskset_for_each(task, dst_css, tset) {
		struct pids_cgroup *pids = css_pids(dst_css);
		struct cgroup_subsys_state *old_css;
		struct pids_cgroup *old_pids;

		old_css = task_css(task, pids_cgrp_id);
		old_pids = css_pids(old_css);

		pids_charge(old_pids, 1);
		pids_uncharge(pids, 1);
	}
}

/*
 * task_css_check(true) in pids_can_fork() and pids_cancel_fork() relies
 * on cgroup_threadgroup_change_begin() held by the copy_process().
 */
static int pids_can_fork(struct task_struct *task)
{
	struct cgroup_subsys_state *css;
	struct pids_cgroup *pids;
	int err;

	css = task_css_check(current, pids_cgrp_id, true);
	pids = css_pids(css);
	err = pids_try_charge(pids, 1);
	if (err) {
		/* Only log the first time events_limit is incremented. */
		if (atomic64_inc_return(&pids->events_limit) == 1) {
			pr_info("cgroup: fork rejected by pids controller in ");
			pr_cont_cgroup_path(css->cgroup);
			pr_cont("\n");
		}
		cgroup_file_notify(&pids->events_file);
	}
	return err;
}

static void pids_cancel_fork(struct task_struct *task)
{
	struct cgroup_subsys_state *css;
	struct pids_cgroup *pids;

	css = task_css_check(current, pids_cgrp_id, true);
	pids = css_pids(css);
	pids_uncharge(pids, 1);
}

static void pids_release(struct task_struct *task)
{
	struct pids_cgroup *pids = css_pids(task_css(task, pids_cgrp_id));

	pids_uncharge(pids, 1);
}

static ssize_t pids_max_write(struct kernfs_open_file *of, char *buf,
			      size_t nbytes, loff_t off)
{
	struct cgroup_subsys_state *css = of_css(of);
	struct pids_cgroup *pids = css_pids(css);
	int64_t limit;
	int err;

	buf = strstrip(buf);
	if (!strcmp(buf, PIDS_MAX_STR)) {
		limit = PIDS_MAX;
		goto set_limit;
	}

	err = kstrtoll(buf, 0, &limit);
	if (err)
		return err;

	if (limit < 0 || limit >= PIDS_MAX)
		return -EINVAL;

set_limit:
	/*
	 * Limit updates don't need to be mutex'd, since it isn't
	 * critical that any racing fork()s follow the new limit.
	 */
	pids->limit = limit;
	return nbytes;
}

static int pids_max_show(struct seq_file *sf, void *v)
{
	struct cgroup_subsys_state *css = seq_css(sf);
	struct pids_cgroup *pids = css_pids(css);
	int64_t limit = pids->limit;

	if (limit >= PIDS_MAX)
		seq_printf(sf, "%s\n", PIDS_MAX_STR);
	else
		seq_printf(sf, "%lld\n", limit);

	return 0;
}

static s64 pids_current_read(struct cgroup_subsys_state *css,
			     struct cftype *cft)
{
	struct pids_cgroup *pids = css_pids(css);

	return atomic64_read(&pids->counter);
}

static int pids_events_show(struct seq_file *sf, void *v)
{
	struct pids_cgroup *pids = css_pids(seq_css(sf));

	seq_printf(sf, "max %lld\n", (s64)atomic64_read(&pids->events_limit));
	return 0;
}

static struct cftype pids_files[] = {
	{
		.name = "max",
		.write = pids_max_write,
		.seq_show = pids_max_show,
		.flags = CFTYPE_NOT_ON_ROOT,
	},
	{
		.name = "current",
		.read_s64 = pids_current_read,
		.flags = CFTYPE_NOT_ON_ROOT,
	},
	{
		.name = "events",
		.seq_show = pids_events_show,
		.file_offset = offsetof(struct pids_cgroup, events_file),
		.flags = CFTYPE_NOT_ON_ROOT,
	},
	{ }	/* terminate */
};

struct cgroup_subsys pids_cgrp_subsys = {
	.css_alloc	= pids_css_alloc,
	.css_free	= pids_css_free,
	.can_attach 	= pids_can_attach,
	.cancel_attach 	= pids_cancel_attach,
	.can_fork	= pids_can_fork,
	.cancel_fork	= pids_cancel_fork,
	.release	= pids_release,
	.legacy_cftypes	= pids_files,
	.dfl_cftypes	= pids_files,
	.threaded	= true,
};
Commit	Line	Data
f85d2086	1	// SPDX-License-Identifier: GPL-2.0-only
49b786ea AS	2	/*
	3	* Process number limiting controller for cgroups.
	4	*
	5	* Used to allow a cgroup hierarchy to stop any new processes from fork()ing
	6	* after a certain limit is reached.
	7	*
	8	* Since it is trivial to hit the task limit without hitting any kmemcg limits
	9	* in place, PIDs are a fundamental resource. As such, PID exhaustion must be
	10	* preventable in the scope of a cgroup hierarchy by allowing resource limiting
	11	* of the number of tasks in a cgroup.
	12	*
	13	* In order to use the `pids` controller, set the maximum number of tasks in
	14	* pids.max (this is not available in the root cgroup for obvious reasons). The
	15	* number of processes currently in the cgroup is given by pids.current.
	16	* Organisational operations are not blocked by cgroup policies, so it is
	17	* possible to have pids.current > pids.max. However, it is not possible to
	18	* violate a cgroup policy through fork(). fork() will return -EAGAIN if forking
	19	* would cause a cgroup policy to be violated.
	20	*
	21	* To set a cgroup to have no limit, set pids.max to "max". This is the default
	22	* for all new cgroups (N.B. that PID limits are hierarchical, so the most
	23	* stringent limit in the hierarchy is followed).
	24	*
	25	* pids.current tracks all child cgroup hierarchies, so parent/pids.current is
	26	* a superset of parent/child/pids.current.
	27	*
	28	* Copyright (C) 2015 Aleksa Sarai <cyphar@cyphar.com>
49b786ea AS	29	*/
	30
	31	#include <linux/kernel.h>
	32	#include <linux/threads.h>
	33	#include <linux/atomic.h>
	34	#include <linux/cgroup.h>
	35	#include <linux/slab.h>
	36
	37	#define PIDS_MAX (PID_MAX_LIMIT + 1ULL)
	38	#define PIDS_MAX_STR "max"
	39
	40	struct pids_cgroup {
	41	struct cgroup_subsys_state css;
	42
	43	/*
	44	* Use 64-bit types so that we can safely represent "max" as
	45	* %PIDS_MAX = (%PID_MAX_LIMIT + 1).
	46	*/
	47	atomic64_t counter;
	48	int64_t limit;
135b8b37 KY	49
	50	/* Handle for "pids.events" */
	51	struct cgroup_file events_file;
	52
	53	/* Number of times fork failed because limit was hit. */
	54	atomic64_t events_limit;
49b786ea AS	55	};
	56
	57	static struct pids_cgroup css_pids(struct cgroup_subsys_state css)
	58	{
	59	return container_of(css, struct pids_cgroup, css);
	60	}
	61
	62	static struct pids_cgroup parent_pids(struct pids_cgroup pids)
	63	{
	64	return css_pids(pids->css.parent);
	65	}
	66
	67	static struct cgroup_subsys_state *
	68	pids_css_alloc(struct cgroup_subsys_state *parent)
	69	{
	70	struct pids_cgroup *pids;
	71
	72	pids = kzalloc(sizeof(struct pids_cgroup), GFP_KERNEL);
	73	if (!pids)
	74	return ERR_PTR(-ENOMEM);
	75
	76	pids->limit = PIDS_MAX;
	77	atomic64_set(&pids->counter, 0);
135b8b37	78	atomic64_set(&pids->events_limit, 0);
49b786ea AS	79	return &pids->css;
	80	}
	81
	82	static void pids_css_free(struct cgroup_subsys_state *css)
	83	{
	84	kfree(css_pids(css));
	85	}
	86
	87	/**
	88	* pids_cancel - uncharge the local pid count
	89	* @pids: the pid cgroup state
	90	* @num: the number of pids to cancel
	91	*
	92	* This function will WARN if the pid count goes under 0, because such a case is
	93	* a bug in the pids controller proper.
	94	*/
	95	static void pids_cancel(struct pids_cgroup *pids, int num)
	96	{
	97	/*
	98	* A negative count (or overflow for that matter) is invalid,
	99	* and indicates a bug in the `pids` controller proper.
	100	*/
	101	WARN_ON_ONCE(atomic64_add_negative(-num, &pids->counter));
	102	}
	103
	104	/**
	105	* pids_uncharge - hierarchically uncharge the pid count
	106	* @pids: the pid cgroup state
	107	* @num: the number of pids to uncharge
	108	*/
	109	static void pids_uncharge(struct pids_cgroup *pids, int num)
	110	{
	111	struct pids_cgroup *p;
	112
67cde9c4	113	for (p = pids; parent_pids(p); p = parent_pids(p))
49b786ea AS	114	pids_cancel(p, num);
	115	}
	116
	117	/**
	118	* pids_charge - hierarchically charge the pid count
	119	* @pids: the pid cgroup state
	120	* @num: the number of pids to charge
	121	*
	122	* This function does not follow the pid limit set. It cannot fail and the new
	123	* pid count may exceed the limit. This is only used for reverting failed
	124	* attaches, where there is no other way out than violating the limit.
	125	*/
	126	static void pids_charge(struct pids_cgroup *pids, int num)
	127	{
	128	struct pids_cgroup *p;
	129
67cde9c4	130	for (p = pids; parent_pids(p); p = parent_pids(p))
49b786ea AS	131	atomic64_add(num, &p->counter);
	132	}
	133
	134	/**
	135	* pids_try_charge - hierarchically try to charge the pid count
	136	* @pids: the pid cgroup state
	137	* @num: the number of pids to charge
	138	*
	139	* This function follows the set limit. It will fail if the charge would cause
	140	* the new value to exceed the hierarchical limit. Returns 0 if the charge
fccd3af5	141	* succeeded, otherwise -EAGAIN.
49b786ea AS	142	*/
	143	static int pids_try_charge(struct pids_cgroup *pids, int num)
	144	{
	145	struct pids_cgroup p, q;
	146
67cde9c4	147	for (p = pids; parent_pids(p); p = parent_pids(p)) {
49b786ea AS	148	int64_t new = atomic64_add_return(num, &p->counter);
	149
	150	/*
	151	* Since new is capped to the maximum number of pid_t, if
	152	* p->limit is %PIDS_MAX then we know that this test will never
	153	* fail.
	154	*/
	155	if (new > p->limit)
	156	goto revert;
	157	}
	158
	159	return 0;
	160
	161	revert:
	162	for (q = pids; q != p; q = parent_pids(q))
	163	pids_cancel(q, num);
	164	pids_cancel(p, num);
	165
	166	return -EAGAIN;
	167	}
	168
1f7dd3e5	169	static int pids_can_attach(struct cgroup_taskset *tset)
49b786ea	170	{
49b786ea	171	struct task_struct *task;
1f7dd3e5	172	struct cgroup_subsys_state *dst_css;
49b786ea	173
1f7dd3e5 TH	174	cgroup_taskset_for_each(task, dst_css, tset) {
1f7dd3e5 TH	175	struct pids_cgroup *pids = css_pids(dst_css);
49b786ea AS	176	struct cgroup_subsys_state *old_css;
	177	struct pids_cgroup *old_pids;
	178
	179	/*
ce523995 AS	180	* No need to pin @old_css between here and cancel_attach()
	181	* because cgroup core protects it from being freed before
	182	* the migration completes or fails.
49b786ea	183	*/
ce523995	184	old_css = task_css(task, pids_cgrp_id);
49b786ea AS	185	old_pids = css_pids(old_css);
	186
	187	pids_charge(pids, 1);
	188	pids_uncharge(old_pids, 1);
	189	}
	190
	191	return 0;
	192	}
	193
1f7dd3e5	194	static void pids_cancel_attach(struct cgroup_taskset *tset)
49b786ea	195	{
49b786ea	196	struct task_struct *task;
1f7dd3e5	197	struct cgroup_subsys_state *dst_css;
49b786ea	198
1f7dd3e5 TH	199	cgroup_taskset_for_each(task, dst_css, tset) {
1f7dd3e5 TH	200	struct pids_cgroup *pids = css_pids(dst_css);
49b786ea AS	201	struct cgroup_subsys_state *old_css;
	202	struct pids_cgroup *old_pids;
	203
	204	old_css = task_css(task, pids_cgrp_id);
	205	old_pids = css_pids(old_css);
	206
	207	pids_charge(old_pids, 1);
	208	pids_uncharge(pids, 1);
49b786ea AS	209	}
	210	}
	211
afbcb364 ON	212	/*
afbcb364 ON	213	* task_css_check(true) in pids_can_fork() and pids_cancel_fork() relies
780de9dd	214	* on cgroup_threadgroup_change_begin() held by the copy_process().
afbcb364	215	*/
b53202e6	216	static int pids_can_fork(struct task_struct *task)
49b786ea AS	217	{
	218	struct cgroup_subsys_state *css;
	219	struct pids_cgroup *pids;
135b8b37	220	int err;
49b786ea	221
afbcb364	222	css = task_css_check(current, pids_cgrp_id, true);
49b786ea	223	pids = css_pids(css);
135b8b37 KY	224	err = pids_try_charge(pids, 1);
	225	if (err) {
	226	/* Only log the first time events_limit is incremented. */
	227	if (atomic64_inc_return(&pids->events_limit) == 1) {
	228	pr_info("cgroup: fork rejected by pids controller in ");
1d18c274	229	pr_cont_cgroup_path(css->cgroup);
135b8b37 KY	230	pr_cont("\n");
	231	}
	232	cgroup_file_notify(&pids->events_file);
	233	}
	234	return err;
49b786ea AS	235	}
49b786ea AS	236
b53202e6	237	static void pids_cancel_fork(struct task_struct *task)
49b786ea	238	{
afbcb364 ON	239	struct cgroup_subsys_state *css;
afbcb364 ON	240	struct pids_cgroup *pids;
49b786ea	241
afbcb364 ON	242	css = task_css_check(current, pids_cgrp_id, true);
afbcb364 ON	243	pids = css_pids(css);
49b786ea	244	pids_uncharge(pids, 1);
49b786ea AS	245	}
49b786ea AS	246
51bee5ab	247	static void pids_release(struct task_struct *task)
49b786ea	248	{
2e91fa7f	249	struct pids_cgroup *pids = css_pids(task_css(task, pids_cgrp_id));
49b786ea AS	250
	251	pids_uncharge(pids, 1);
	252	}
	253
	254	static ssize_t pids_max_write(struct kernfs_open_file of, char buf,
	255	size_t nbytes, loff_t off)
	256	{
	257	struct cgroup_subsys_state *css = of_css(of);
	258	struct pids_cgroup *pids = css_pids(css);
	259	int64_t limit;
	260	int err;
	261
	262	buf = strstrip(buf);
	263	if (!strcmp(buf, PIDS_MAX_STR)) {
	264	limit = PIDS_MAX;
	265	goto set_limit;
	266	}
	267
	268	err = kstrtoll(buf, 0, &limit);
	269	if (err)
	270	return err;
	271
	272	if (limit < 0 \|\| limit >= PIDS_MAX)
	273	return -EINVAL;
	274
	275	set_limit:
	276	/*
	277	* Limit updates don't need to be mutex'd, since it isn't
	278	* critical that any racing fork()s follow the new limit.
	279	*/
	280	pids->limit = limit;
	281	return nbytes;
	282	}
	283
	284	static int pids_max_show(struct seq_file sf, void v)
	285	{
	286	struct cgroup_subsys_state *css = seq_css(sf);
	287	struct pids_cgroup *pids = css_pids(css);
	288	int64_t limit = pids->limit;
	289
	290	if (limit >= PIDS_MAX)
	291	seq_printf(sf, "%s\n", PIDS_MAX_STR);
	292	else
	293	seq_printf(sf, "%lld\n", limit);
	294
	295	return 0;
	296	}
	297
	298	static s64 pids_current_read(struct cgroup_subsys_state *css,
	299	struct cftype *cft)
	300	{
	301	struct pids_cgroup *pids = css_pids(css);
	302
	303	return atomic64_read(&pids->counter);
	304	}
	305
135b8b37 KY	306	static int pids_events_show(struct seq_file sf, void v)
	307	{
	308	struct pids_cgroup *pids = css_pids(seq_css(sf));
	309
9f6870dd	310	seq_printf(sf, "max %lld\n", (s64)atomic64_read(&pids->events_limit));
135b8b37 KY	311	return 0;
	312	}
	313
49b786ea AS	314	static struct cftype pids_files[] = {
	315	{
	316	.name = "max",
	317	.write = pids_max_write,
	318	.seq_show = pids_max_show,
	319	.flags = CFTYPE_NOT_ON_ROOT,
	320	},
	321	{
	322	.name = "current",
	323	.read_s64 = pids_current_read,
67cde9c4	324	.flags = CFTYPE_NOT_ON_ROOT,
49b786ea	325	},
135b8b37 KY	326	{
	327	.name = "events",
	328	.seq_show = pids_events_show,
	329	.file_offset = offsetof(struct pids_cgroup, events_file),
	330	.flags = CFTYPE_NOT_ON_ROOT,
	331	},
49b786ea AS	332	{ } /* terminate */
	333	};
	334
	335	struct cgroup_subsys pids_cgrp_subsys = {
	336	.css_alloc = pids_css_alloc,
	337	.css_free = pids_css_free,
49b786ea AS	338	.can_attach = pids_can_attach,
	339	.cancel_attach = pids_cancel_attach,
	340	.can_fork = pids_can_fork,
	341	.cancel_fork = pids_cancel_fork,
51bee5ab	342	.release = pids_release,
49b786ea AS	343	.legacy_cftypes = pids_files,
49b786ea AS	344	.dfl_cftypes = pids_files,
8cfd8147	345	.threaded = true,
49b786ea	346	};