[mirror_ubuntu-jammy-kernel.git] / kernel / bpf / trampoline.c

// SPDX-License-Identifier: GPL-2.0-only
/* Copyright (c) 2019 Facebook */
#include <linux/hash.h>
#include <linux/bpf.h>
#include <linux/filter.h>
#include <linux/ftrace.h>

/* btf_vmlinux has ~22k attachable functions. 1k htab is enough. */
#define TRAMPOLINE_HASH_BITS 10
#define TRAMPOLINE_TABLE_SIZE (1 << TRAMPOLINE_HASH_BITS)

static struct hlist_head trampoline_table[TRAMPOLINE_TABLE_SIZE];

/* serializes access to trampoline_table */
static DEFINE_MUTEX(trampoline_mutex);

struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
{
	struct bpf_trampoline *tr;
	struct hlist_head *head;
	void *image;
	int i;

	mutex_lock(&trampoline_mutex);
	head = &trampoline_table[hash_64(key, TRAMPOLINE_HASH_BITS)];
	hlist_for_each_entry(tr, head, hlist) {
		if (tr->key == key) {
			refcount_inc(&tr->refcnt);
			goto out;
		}
	}
	tr = kzalloc(sizeof(*tr), GFP_KERNEL);
	if (!tr)
		goto out;

	/* is_root was checked earlier. No need for bpf_jit_charge_modmem() */
	image = bpf_jit_alloc_exec(PAGE_SIZE);
	if (!image) {
		kfree(tr);
		tr = NULL;
		goto out;
	}

	tr->key = key;
	INIT_HLIST_NODE(&tr->hlist);
	hlist_add_head(&tr->hlist, head);
	refcount_set(&tr->refcnt, 1);
	mutex_init(&tr->mutex);
	for (i = 0; i < BPF_TRAMP_MAX; i++)
		INIT_HLIST_HEAD(&tr->progs_hlist[i]);

	set_vm_flush_reset_perms(image);
	/* Keep image as writeable. The alternative is to keep flipping ro/rw
	 * everytime new program is attached or detached.
	 */
	set_memory_x((long)image, 1);
	tr->image = image;
out:
	mutex_unlock(&trampoline_mutex);
	return tr;
}

static int is_ftrace_location(void *ip)
{
	long addr;

	addr = ftrace_location((long)ip);
	if (!addr)
		return 0;
	if (WARN_ON_ONCE(addr != (long)ip))
		return -EFAULT;
	return 1;
}

static int unregister_fentry(struct bpf_trampoline *tr, void *old_addr)
{
	void *ip = tr->func.addr;
	int ret;

	if (tr->func.ftrace_managed)
		ret = unregister_ftrace_direct((long)ip, (long)old_addr);
	else
		ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, old_addr, NULL);
	return ret;
}

static int modify_fentry(struct bpf_trampoline *tr, void *old_addr, void *new_addr)
{
	void *ip = tr->func.addr;
	int ret;

	if (tr->func.ftrace_managed)
		ret = modify_ftrace_direct((long)ip, (long)old_addr, (long)new_addr);
	else
		ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, old_addr, new_addr);
	return ret;
}

/* first time registering */
static int register_fentry(struct bpf_trampoline *tr, void *new_addr)
{
	void *ip = tr->func.addr;
	int ret;

	ret = is_ftrace_location(ip);
	if (ret < 0)
		return ret;
	tr->func.ftrace_managed = ret;

	if (tr->func.ftrace_managed)
		ret = register_ftrace_direct((long)ip, (long)new_addr);
	else
		ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, NULL, new_addr);
	return ret;
}

/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
 * bytes on x86.  Pick a number to fit into PAGE_SIZE / 2
 */
#define BPF_MAX_TRAMP_PROGS 40

static int bpf_trampoline_update(struct bpf_trampoline *tr)
{
	void *old_image = tr->image + ((tr->selector + 1) & 1) * PAGE_SIZE/2;
	void *new_image = tr->image + (tr->selector & 1) * PAGE_SIZE/2;
	struct bpf_prog *progs_to_run[BPF_MAX_TRAMP_PROGS];
	int fentry_cnt = tr->progs_cnt[BPF_TRAMP_FENTRY];
	int fexit_cnt = tr->progs_cnt[BPF_TRAMP_FEXIT];
	struct bpf_prog **progs, **fentry, **fexit;
	u32 flags = BPF_TRAMP_F_RESTORE_REGS;
	struct bpf_prog_aux *aux;
	int err;

	if (fentry_cnt + fexit_cnt == 0) {
		err = unregister_fentry(tr, old_image);
		tr->selector = 0;
		goto out;
	}

	/* populate fentry progs */
	fentry = progs = progs_to_run;
	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FENTRY], tramp_hlist)
		*progs++ = aux->prog;

	/* populate fexit progs */
	fexit = progs;
	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FEXIT], tramp_hlist)
		*progs++ = aux->prog;

	if (fexit_cnt)
		flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME;

	err = arch_prepare_bpf_trampoline(new_image, &tr->func.model, flags,
					  fentry, fentry_cnt,
					  fexit, fexit_cnt,
					  tr->func.addr);
	if (err)
		goto out;

	if (tr->selector)
		/* progs already running at this address */
		err = modify_fentry(tr, old_image, new_image);
	else
		/* first time registering */
		err = register_fentry(tr, new_image);
	if (err)
		goto out;
	tr->selector++;
out:
	return err;
}

static enum bpf_tramp_prog_type bpf_attach_type_to_tramp(enum bpf_attach_type t)
{
	switch (t) {
	case BPF_TRACE_FENTRY:
		return BPF_TRAMP_FENTRY;
	default:
		return BPF_TRAMP_FEXIT;
	}
}

int bpf_trampoline_link_prog(struct bpf_prog *prog)
{
	enum bpf_tramp_prog_type kind;
	struct bpf_trampoline *tr;
	int err = 0;

	tr = prog->aux->trampoline;
	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
	mutex_lock(&tr->mutex);
	if (tr->progs_cnt[BPF_TRAMP_FENTRY] + tr->progs_cnt[BPF_TRAMP_FEXIT]
	    >= BPF_MAX_TRAMP_PROGS) {
		err = -E2BIG;
		goto out;
	}
	if (!hlist_unhashed(&prog->aux->tramp_hlist)) {
		/* prog already linked */
		err = -EBUSY;
		goto out;
	}
	hlist_add_head(&prog->aux->tramp_hlist, &tr->progs_hlist[kind]);
	tr->progs_cnt[kind]++;
	err = bpf_trampoline_update(prog->aux->trampoline);
	if (err) {
		hlist_del(&prog->aux->tramp_hlist);
		tr->progs_cnt[kind]--;
	}
out:
	mutex_unlock(&tr->mutex);
	return err;
}

/* bpf_trampoline_unlink_prog() should never fail. */
int bpf_trampoline_unlink_prog(struct bpf_prog *prog)
{
	enum bpf_tramp_prog_type kind;
	struct bpf_trampoline *tr;
	int err;

	tr = prog->aux->trampoline;
	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
	mutex_lock(&tr->mutex);
	hlist_del(&prog->aux->tramp_hlist);
	tr->progs_cnt[kind]--;
	err = bpf_trampoline_update(prog->aux->trampoline);
	mutex_unlock(&tr->mutex);
	return err;
}

void bpf_trampoline_put(struct bpf_trampoline *tr)
{
	if (!tr)
		return;
	mutex_lock(&trampoline_mutex);
	if (!refcount_dec_and_test(&tr->refcnt))
		goto out;
	WARN_ON_ONCE(mutex_is_locked(&tr->mutex));
	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FENTRY])))
		goto out;
	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FEXIT])))
		goto out;
	bpf_jit_free_exec(tr->image);
	hlist_del(&tr->hlist);
	kfree(tr);
out:
	mutex_unlock(&trampoline_mutex);
}

/* The logic is similar to BPF_PROG_RUN, but with explicit rcu and preempt that
 * are needed for trampoline. The macro is split into
 * call _bpf_prog_enter
 * call prog->bpf_func
 * call __bpf_prog_exit
 */
u64 notrace __bpf_prog_enter(void)
{
	u64 start = 0;

	rcu_read_lock();
	preempt_disable();
	if (static_branch_unlikely(&bpf_stats_enabled_key))
		start = sched_clock();
	return start;
}

void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start)
{
	struct bpf_prog_stats *stats;

	if (static_branch_unlikely(&bpf_stats_enabled_key) &&
	    /* static_key could be enabled in __bpf_prog_enter
	     * and disabled in __bpf_prog_exit.
	     * And vice versa.
	     * Hence check that 'start' is not zero.
	     */
	    start) {
		stats = this_cpu_ptr(prog->aux->stats);
		u64_stats_update_begin(&stats->syncp);
		stats->cnt++;
		stats->nsecs += sched_clock() - start;
		u64_stats_update_end(&stats->syncp);
	}
	preempt_enable();
	rcu_read_unlock();
}

int __weak
arch_prepare_bpf_trampoline(void *image, struct btf_func_model *m, u32 flags,
			    struct bpf_prog **fentry_progs, int fentry_cnt,
			    struct bpf_prog **fexit_progs, int fexit_cnt,
			    void *orig_call)
{
	return -ENOTSUPP;
}

static int __init init_trampolines(void)
{
	int i;

	for (i = 0; i < TRAMPOLINE_TABLE_SIZE; i++)
		INIT_HLIST_HEAD(&trampoline_table[i]);
	return 0;
}
late_initcall(init_trampolines);
Commit	Line	Data
fec56f58 AS	1	// SPDX-License-Identifier: GPL-2.0-only
	2	/* Copyright (c) 2019 Facebook */
	3	#include <linux/hash.h>
	4	#include <linux/bpf.h>
	5	#include <linux/filter.h>
b91e014f	6	#include <linux/ftrace.h>
fec56f58 AS	7
	8	/* btf_vmlinux has ~22k attachable functions. 1k htab is enough. */
	9	#define TRAMPOLINE_HASH_BITS 10
	10	#define TRAMPOLINE_TABLE_SIZE (1 << TRAMPOLINE_HASH_BITS)
	11
	12	static struct hlist_head trampoline_table[TRAMPOLINE_TABLE_SIZE];
	13
	14	/* serializes access to trampoline_table */
	15	static DEFINE_MUTEX(trampoline_mutex);
	16
	17	struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
	18	{
	19	struct bpf_trampoline *tr;
	20	struct hlist_head *head;
	21	void *image;
	22	int i;
	23
	24	mutex_lock(&trampoline_mutex);
	25	head = &trampoline_table[hash_64(key, TRAMPOLINE_HASH_BITS)];
	26	hlist_for_each_entry(tr, head, hlist) {
	27	if (tr->key == key) {
	28	refcount_inc(&tr->refcnt);
	29	goto out;
	30	}
	31	}
	32	tr = kzalloc(sizeof(*tr), GFP_KERNEL);
	33	if (!tr)
	34	goto out;
	35
	36	/* is_root was checked earlier. No need for bpf_jit_charge_modmem() */
	37	image = bpf_jit_alloc_exec(PAGE_SIZE);
	38	if (!image) {
	39	kfree(tr);
	40	tr = NULL;
	41	goto out;
	42	}
	43
	44	tr->key = key;
	45	INIT_HLIST_NODE(&tr->hlist);
	46	hlist_add_head(&tr->hlist, head);
	47	refcount_set(&tr->refcnt, 1);
	48	mutex_init(&tr->mutex);
	49	for (i = 0; i < BPF_TRAMP_MAX; i++)
	50	INIT_HLIST_HEAD(&tr->progs_hlist[i]);
	51
	52	set_vm_flush_reset_perms(image);
	53	/* Keep image as writeable. The alternative is to keep flipping ro/rw
	54	* everytime new program is attached or detached.
	55	*/
	56	set_memory_x((long)image, 1);
	57	tr->image = image;
	58	out:
	59	mutex_unlock(&trampoline_mutex);
	60	return tr;
	61	}
	62
b91e014f AS	63	static int is_ftrace_location(void *ip)
	64	{
	65	long addr;
	66
	67	addr = ftrace_location((long)ip);
	68	if (!addr)
	69	return 0;
	70	if (WARN_ON_ONCE(addr != (long)ip))
	71	return -EFAULT;
	72	return 1;
	73	}
	74
	75	static int unregister_fentry(struct bpf_trampoline tr, void old_addr)
	76	{
	77	void *ip = tr->func.addr;
	78	int ret;
	79
	80	if (tr->func.ftrace_managed)
	81	ret = unregister_ftrace_direct((long)ip, (long)old_addr);
	82	else
	83	ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, old_addr, NULL);
	84	return ret;
	85	}
	86
	87	static int modify_fentry(struct bpf_trampoline tr, void old_addr, void *new_addr)
	88	{
	89	void *ip = tr->func.addr;
	90	int ret;
	91
	92	if (tr->func.ftrace_managed)
	93	ret = modify_ftrace_direct((long)ip, (long)old_addr, (long)new_addr);
	94	else
	95	ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, old_addr, new_addr);
	96	return ret;
	97	}
	98
	99	/* first time registering */
	100	static int register_fentry(struct bpf_trampoline tr, void new_addr)
	101	{
	102	void *ip = tr->func.addr;
	103	int ret;
	104
	105	ret = is_ftrace_location(ip);
	106	if (ret < 0)
	107	return ret;
	108	tr->func.ftrace_managed = ret;
	109
	110	if (tr->func.ftrace_managed)
	111	ret = register_ftrace_direct((long)ip, (long)new_addr);
	112	else
	113	ret = bpf_arch_text_poke(ip, BPF_MOD_CALL, NULL, new_addr);
	114	return ret;
	115	}
	116
fec56f58 AS	117	/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
	118	* bytes on x86. Pick a number to fit into PAGE_SIZE / 2
	119	*/
	120	#define BPF_MAX_TRAMP_PROGS 40
	121
	122	static int bpf_trampoline_update(struct bpf_trampoline *tr)
	123	{
	124	void old_image = tr->image + ((tr->selector + 1) & 1) PAGE_SIZE/2;
	125	void new_image = tr->image + (tr->selector & 1) PAGE_SIZE/2;
	126	struct bpf_prog *progs_to_run[BPF_MAX_TRAMP_PROGS];
	127	int fentry_cnt = tr->progs_cnt[BPF_TRAMP_FENTRY];
	128	int fexit_cnt = tr->progs_cnt[BPF_TRAMP_FEXIT];
	129	struct bpf_prog progs, fentry, **fexit;
	130	u32 flags = BPF_TRAMP_F_RESTORE_REGS;
	131	struct bpf_prog_aux *aux;
	132	int err;
	133
	134	if (fentry_cnt + fexit_cnt == 0) {
b91e014f	135	err = unregister_fentry(tr, old_image);
fec56f58 AS	136	tr->selector = 0;
	137	goto out;
	138	}
	139
	140	/* populate fentry progs */
	141	fentry = progs = progs_to_run;
	142	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FENTRY], tramp_hlist)
	143	*progs++ = aux->prog;
	144
	145	/* populate fexit progs */
	146	fexit = progs;
	147	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FEXIT], tramp_hlist)
	148	*progs++ = aux->prog;
	149
	150	if (fexit_cnt)
	151	flags = BPF_TRAMP_F_CALL_ORIG \| BPF_TRAMP_F_SKIP_FRAME;
	152
	153	err = arch_prepare_bpf_trampoline(new_image, &tr->func.model, flags,
	154	fentry, fentry_cnt,
	155	fexit, fexit_cnt,
	156	tr->func.addr);
	157	if (err)
	158	goto out;
	159
	160	if (tr->selector)
	161	/* progs already running at this address */
b91e014f	162	err = modify_fentry(tr, old_image, new_image);
fec56f58 AS	163	else
fec56f58 AS	164	/* first time registering */
b91e014f	165	err = register_fentry(tr, new_image);
fec56f58 AS	166	if (err)
	167	goto out;
	168	tr->selector++;
	169	out:
	170	return err;
	171	}
	172
	173	static enum bpf_tramp_prog_type bpf_attach_type_to_tramp(enum bpf_attach_type t)
	174	{
	175	switch (t) {
	176	case BPF_TRACE_FENTRY:
	177	return BPF_TRAMP_FENTRY;
	178	default:
	179	return BPF_TRAMP_FEXIT;
	180	}
	181	}
	182
	183	int bpf_trampoline_link_prog(struct bpf_prog *prog)
	184	{
	185	enum bpf_tramp_prog_type kind;
	186	struct bpf_trampoline *tr;
	187	int err = 0;
	188
	189	tr = prog->aux->trampoline;
	190	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
	191	mutex_lock(&tr->mutex);
	192	if (tr->progs_cnt[BPF_TRAMP_FENTRY] + tr->progs_cnt[BPF_TRAMP_FEXIT]
	193	>= BPF_MAX_TRAMP_PROGS) {
	194	err = -E2BIG;
	195	goto out;
	196	}
	197	if (!hlist_unhashed(&prog->aux->tramp_hlist)) {
	198	/* prog already linked */
	199	err = -EBUSY;
	200	goto out;
	201	}
	202	hlist_add_head(&prog->aux->tramp_hlist, &tr->progs_hlist[kind]);
	203	tr->progs_cnt[kind]++;
	204	err = bpf_trampoline_update(prog->aux->trampoline);
	205	if (err) {
	206	hlist_del(&prog->aux->tramp_hlist);
	207	tr->progs_cnt[kind]--;
	208	}
	209	out:
	210	mutex_unlock(&tr->mutex);
	211	return err;
	212	}
	213
	214	/* bpf_trampoline_unlink_prog() should never fail. */
	215	int bpf_trampoline_unlink_prog(struct bpf_prog *prog)
	216	{
	217	enum bpf_tramp_prog_type kind;
	218	struct bpf_trampoline *tr;
	219	int err;
	220
	221	tr = prog->aux->trampoline;
	222	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
	223	mutex_lock(&tr->mutex);
	224	hlist_del(&prog->aux->tramp_hlist);
	225	tr->progs_cnt[kind]--;
	226	err = bpf_trampoline_update(prog->aux->trampoline);
	227	mutex_unlock(&tr->mutex);
	228	return err;
	229	}
230
231	void bpf_trampoline_put(struct bpf_trampoline *tr)
232	{
233	if (!tr)
234	return;
235	mutex_lock(&trampoline_mutex);
236	if (!refcount_dec_and_test(&tr->refcnt))
237	goto out;
238	WARN_ON_ONCE(mutex_is_locked(&tr->mutex));
239	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FENTRY])))
240	goto out;
241	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FEXIT])))
242	goto out;
243	bpf_jit_free_exec(tr->image);
244	hlist_del(&tr->hlist);
245	kfree(tr);
246	out:
247	mutex_unlock(&trampoline_mutex);
248	}
249
250	/* The logic is similar to BPF_PROG_RUN, but with explicit rcu and preempt that
251	* are needed for trampoline. The macro is split into
252	* call _bpf_prog_enter
253	* call prog->bpf_func
254	* call __bpf_prog_exit
255	*/
256	u64 notrace __bpf_prog_enter(void)
257	{
258	u64 start = 0;
259
260	rcu_read_lock();
261	preempt_disable();
262	if (static_branch_unlikely(&bpf_stats_enabled_key))
263	start = sched_clock();
264	return start;
265	}
266
267	void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start)
268	{
269	struct bpf_prog_stats *stats;
270
271	if (static_branch_unlikely(&bpf_stats_enabled_key) &&
272	/* static_key could be enabled in __bpf_prog_enter
273	* and disabled in __bpf_prog_exit.
274	* And vice versa.
275	* Hence check that 'start' is not zero.
276	*/
277	start) {
278	stats = this_cpu_ptr(prog->aux->stats);
279	u64_stats_update_begin(&stats->syncp);
280	stats->cnt++;
281	stats->nsecs += sched_clock() - start;
282	u64_stats_update_end(&stats->syncp);
283	}
284	preempt_enable();
285	rcu_read_unlock();
286	}
287
288	int __weak
289	arch_prepare_bpf_trampoline(void image, struct btf_func_model m, u32 flags,
290	struct bpf_prog **fentry_progs, int fentry_cnt,
291	struct bpf_prog **fexit_progs, int fexit_cnt,
292	void *orig_call)
293	{
294	return -ENOTSUPP;
295	}
296
297	static int __init init_trampolines(void)
298	{
299	int i;
300
301	for (i = 0; i < TRAMPOLINE_TABLE_SIZE; i++)
302	INIT_HLIST_HEAD(&trampoline_table[i]);
303	return 0;
304	}
305	late_initcall(init_trampolines);