[linux-block.git] / net / core / xdp.c

// SPDX-License-Identifier: GPL-2.0-only
/* net/core/xdp.c
 *
 * Copyright (c) 2017 Jesper Dangaard Brouer, Red Hat Inc.
 */
#include <linux/bpf.h>
#include <linux/filter.h>
#include <linux/types.h>
#include <linux/mm.h>
#include <linux/netdevice.h>
#include <linux/slab.h>
#include <linux/idr.h>
#include <linux/rhashtable.h>
#include <linux/bug.h>
#include <net/page_pool.h>

#include <net/xdp.h>
#include <net/xdp_priv.h> /* struct xdp_mem_allocator */
#include <trace/events/xdp.h>
#include <net/xdp_sock_drv.h>

#define REG_STATE_NEW		0x0
#define REG_STATE_REGISTERED	0x1
#define REG_STATE_UNREGISTERED	0x2
#define REG_STATE_UNUSED	0x3

static DEFINE_IDA(mem_id_pool);
static DEFINE_MUTEX(mem_id_lock);
#define MEM_ID_MAX 0xFFFE
#define MEM_ID_MIN 1
static int mem_id_next = MEM_ID_MIN;

static bool mem_id_init; /* false */
static struct rhashtable *mem_id_ht;

static u32 xdp_mem_id_hashfn(const void *data, u32 len, u32 seed)
{
	const u32 *k = data;
	const u32 key = *k;

	BUILD_BUG_ON(sizeof_field(struct xdp_mem_allocator, mem.id)
		     != sizeof(u32));

	/* Use cyclic increasing ID as direct hash key */
	return key;
}

static int xdp_mem_id_cmp(struct rhashtable_compare_arg *arg,
			  const void *ptr)
{
	const struct xdp_mem_allocator *xa = ptr;
	u32 mem_id = *(u32 *)arg->key;

	return xa->mem.id != mem_id;
}

static const struct rhashtable_params mem_id_rht_params = {
	.nelem_hint = 64,
	.head_offset = offsetof(struct xdp_mem_allocator, node),
	.key_offset  = offsetof(struct xdp_mem_allocator, mem.id),
	.key_len = sizeof_field(struct xdp_mem_allocator, mem.id),
	.max_size = MEM_ID_MAX,
	.min_size = 8,
	.automatic_shrinking = true,
	.hashfn    = xdp_mem_id_hashfn,
	.obj_cmpfn = xdp_mem_id_cmp,
};

static void __xdp_mem_allocator_rcu_free(struct rcu_head *rcu)
{
	struct xdp_mem_allocator *xa;

	xa = container_of(rcu, struct xdp_mem_allocator, rcu);

	/* Allow this ID to be reused */
	ida_simple_remove(&mem_id_pool, xa->mem.id);

	kfree(xa);
}

static void mem_xa_remove(struct xdp_mem_allocator *xa)
{
	trace_mem_disconnect(xa);

	if (!rhashtable_remove_fast(mem_id_ht, &xa->node, mem_id_rht_params))
		call_rcu(&xa->rcu, __xdp_mem_allocator_rcu_free);
}

static void mem_allocator_disconnect(void *allocator)
{
	struct xdp_mem_allocator *xa;
	struct rhashtable_iter iter;

	mutex_lock(&mem_id_lock);

	rhashtable_walk_enter(mem_id_ht, &iter);
	do {
		rhashtable_walk_start(&iter);

		while ((xa = rhashtable_walk_next(&iter)) && !IS_ERR(xa)) {
			if (xa->allocator == allocator)
				mem_xa_remove(xa);
		}

		rhashtable_walk_stop(&iter);

	} while (xa == ERR_PTR(-EAGAIN));
	rhashtable_walk_exit(&iter);

	mutex_unlock(&mem_id_lock);
}

void xdp_rxq_info_unreg_mem_model(struct xdp_rxq_info *xdp_rxq)
{
	struct xdp_mem_allocator *xa;
	int id = xdp_rxq->mem.id;

	if (xdp_rxq->reg_state != REG_STATE_REGISTERED) {
		WARN(1, "Missing register, driver bug");
		return;
	}

	if (id == 0)
		return;

	if (xdp_rxq->mem.type == MEM_TYPE_PAGE_POOL) {
		rcu_read_lock();
		xa = rhashtable_lookup(mem_id_ht, &id, mem_id_rht_params);
		page_pool_destroy(xa->page_pool);
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_unreg_mem_model);

void xdp_rxq_info_unreg(struct xdp_rxq_info *xdp_rxq)
{
	/* Simplify driver cleanup code paths, allow unreg "unused" */
	if (xdp_rxq->reg_state == REG_STATE_UNUSED)
		return;

	WARN(!(xdp_rxq->reg_state == REG_STATE_REGISTERED), "Driver BUG");

	xdp_rxq_info_unreg_mem_model(xdp_rxq);

	xdp_rxq->reg_state = REG_STATE_UNREGISTERED;
	xdp_rxq->dev = NULL;

	/* Reset mem info to defaults */
	xdp_rxq->mem.id = 0;
	xdp_rxq->mem.type = 0;
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_unreg);

static void xdp_rxq_info_init(struct xdp_rxq_info *xdp_rxq)
{
	memset(xdp_rxq, 0, sizeof(*xdp_rxq));
}

/* Returns 0 on success, negative on failure */
int xdp_rxq_info_reg(struct xdp_rxq_info *xdp_rxq,
		     struct net_device *dev, u32 queue_index, unsigned int napi_id)
{
	if (xdp_rxq->reg_state == REG_STATE_UNUSED) {
		WARN(1, "Driver promised not to register this");
		return -EINVAL;
	}

	if (xdp_rxq->reg_state == REG_STATE_REGISTERED) {
		WARN(1, "Missing unregister, handled but fix driver");
		xdp_rxq_info_unreg(xdp_rxq);
	}

	if (!dev) {
		WARN(1, "Missing net_device from driver");
		return -ENODEV;
	}

	/* State either UNREGISTERED or NEW */
	xdp_rxq_info_init(xdp_rxq);
	xdp_rxq->dev = dev;
	xdp_rxq->queue_index = queue_index;
	xdp_rxq->napi_id = napi_id;

	xdp_rxq->reg_state = REG_STATE_REGISTERED;
	return 0;
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_reg);

void xdp_rxq_info_unused(struct xdp_rxq_info *xdp_rxq)
{
	xdp_rxq->reg_state = REG_STATE_UNUSED;
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_unused);

bool xdp_rxq_info_is_reg(struct xdp_rxq_info *xdp_rxq)
{
	return (xdp_rxq->reg_state == REG_STATE_REGISTERED);
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_is_reg);

static int __mem_id_init_hash_table(void)
{
	struct rhashtable *rht;
	int ret;

	if (unlikely(mem_id_init))
		return 0;

	rht = kzalloc(sizeof(*rht), GFP_KERNEL);
	if (!rht)
		return -ENOMEM;

	ret = rhashtable_init(rht, &mem_id_rht_params);
	if (ret < 0) {
		kfree(rht);
		return ret;
	}
	mem_id_ht = rht;
	smp_mb(); /* mutex lock should provide enough pairing */
	mem_id_init = true;

	return 0;
}

/* Allocate a cyclic ID that maps to allocator pointer.
 * See: https://www.kernel.org/doc/html/latest/core-api/idr.html
 *
 * Caller must lock mem_id_lock.
 */
static int __mem_id_cyclic_get(gfp_t gfp)
{
	int retries = 1;
	int id;

again:
	id = ida_simple_get(&mem_id_pool, mem_id_next, MEM_ID_MAX, gfp);
	if (id < 0) {
		if (id == -ENOSPC) {
			/* Cyclic allocator, reset next id */
			if (retries--) {
				mem_id_next = MEM_ID_MIN;
				goto again;
			}
		}
		return id; /* errno */
	}
	mem_id_next = id + 1;

	return id;
}

static bool __is_supported_mem_type(enum xdp_mem_type type)
{
	if (type == MEM_TYPE_PAGE_POOL)
		return is_page_pool_compiled_in();

	if (type >= MEM_TYPE_MAX)
		return false;

	return true;
}

int xdp_rxq_info_reg_mem_model(struct xdp_rxq_info *xdp_rxq,
			       enum xdp_mem_type type, void *allocator)
{
	struct xdp_mem_allocator *xdp_alloc;
	gfp_t gfp = GFP_KERNEL;
	int id, errno, ret;
	void *ptr;

	if (xdp_rxq->reg_state != REG_STATE_REGISTERED) {
		WARN(1, "Missing register, driver bug");
		return -EFAULT;
	}

	if (!__is_supported_mem_type(type))
		return -EOPNOTSUPP;

	xdp_rxq->mem.type = type;

	if (!allocator) {
		if (type == MEM_TYPE_PAGE_POOL)
			return -EINVAL; /* Setup time check page_pool req */
		return 0;
	}

	/* Delay init of rhashtable to save memory if feature isn't used */
	if (!mem_id_init) {
		mutex_lock(&mem_id_lock);
		ret = __mem_id_init_hash_table();
		mutex_unlock(&mem_id_lock);
		if (ret < 0) {
			WARN_ON(1);
			return ret;
		}
	}

	xdp_alloc = kzalloc(sizeof(*xdp_alloc), gfp);
	if (!xdp_alloc)
		return -ENOMEM;

	mutex_lock(&mem_id_lock);
	id = __mem_id_cyclic_get(gfp);
	if (id < 0) {
		errno = id;
		goto err;
	}
	xdp_rxq->mem.id = id;
	xdp_alloc->mem  = xdp_rxq->mem;
	xdp_alloc->allocator = allocator;

	/* Insert allocator into ID lookup table */
	ptr = rhashtable_insert_slow(mem_id_ht, &id, &xdp_alloc->node);
	if (IS_ERR(ptr)) {
		ida_simple_remove(&mem_id_pool, xdp_rxq->mem.id);
		xdp_rxq->mem.id = 0;
		errno = PTR_ERR(ptr);
		goto err;
	}

	if (type == MEM_TYPE_PAGE_POOL)
		page_pool_use_xdp_mem(allocator, mem_allocator_disconnect);

	mutex_unlock(&mem_id_lock);

	trace_mem_connect(xdp_alloc, xdp_rxq);
	return 0;
err:
	mutex_unlock(&mem_id_lock);
	kfree(xdp_alloc);
	return errno;
}
EXPORT_SYMBOL_GPL(xdp_rxq_info_reg_mem_model);

/* XDP RX runs under NAPI protection, and in different delivery error
 * scenarios (e.g. queue full), it is possible to return the xdp_frame
 * while still leveraging this protection.  The @napi_direct boolean
 * is used for those calls sites.  Thus, allowing for faster recycling
 * of xdp_frames/pages in those cases.
 */
static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
			 struct xdp_buff *xdp)
{
	struct xdp_mem_allocator *xa;
	struct page *page;

	switch (mem->type) {
	case MEM_TYPE_PAGE_POOL:
		rcu_read_lock();
		/* mem->id is valid, checked in xdp_rxq_info_reg_mem_model() */
		xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
		page = virt_to_head_page(data);
		if (napi_direct && xdp_return_frame_no_direct())
			napi_direct = false;
		page_pool_put_full_page(xa->page_pool, page, napi_direct);
		rcu_read_unlock();
		break;
	case MEM_TYPE_PAGE_SHARED:
		page_frag_free(data);
		break;
	case MEM_TYPE_PAGE_ORDER0:
		page = virt_to_page(data); /* Assumes order0 page*/
		put_page(page);
		break;
	case MEM_TYPE_XSK_BUFF_POOL:
		/* NB! Only valid from an xdp_buff! */
		xsk_buff_free(xdp);
		break;
	default:
		/* Not possible, checked in xdp_rxq_info_reg_mem_model() */
		WARN(1, "Incorrect XDP memory type (%d) usage", mem->type);
		break;
	}
}

void xdp_return_frame(struct xdp_frame *xdpf)
{
	__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
}
EXPORT_SYMBOL_GPL(xdp_return_frame);

void xdp_return_frame_rx_napi(struct xdp_frame *xdpf)
{
	__xdp_return(xdpf->data, &xdpf->mem, true, NULL);
}
EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi);

/* XDP bulk APIs introduce a defer/flush mechanism to return
 * pages belonging to the same xdp_mem_allocator object
 * (identified via the mem.id field) in bulk to optimize
 * I-cache and D-cache.
 * The bulk queue size is set to 16 to be aligned to how
 * XDP_REDIRECT bulking works. The bulk is flushed when
 * it is full or when mem.id changes.
 * xdp_frame_bulk is usually stored/allocated on the function
 * call-stack to avoid locking penalties.
 */
void xdp_flush_frame_bulk(struct xdp_frame_bulk *bq)
{
	struct xdp_mem_allocator *xa = bq->xa;

	if (unlikely(!xa || !bq->count))
		return;

	page_pool_put_page_bulk(xa->page_pool, bq->q, bq->count);
	/* bq->xa is not cleared to save lookup, if mem.id same in next bulk */
	bq->count = 0;
}
EXPORT_SYMBOL_GPL(xdp_flush_frame_bulk);

/* Must be called with rcu_read_lock held */
void xdp_return_frame_bulk(struct xdp_frame *xdpf,
			   struct xdp_frame_bulk *bq)
{
	struct xdp_mem_info *mem = &xdpf->mem;
	struct xdp_mem_allocator *xa;

	if (mem->type != MEM_TYPE_PAGE_POOL) {
		__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
		return;
	}

	xa = bq->xa;
	if (unlikely(!xa)) {
		xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
		bq->count = 0;
		bq->xa = xa;
	}

	if (bq->count == XDP_BULK_QUEUE_SIZE)
		xdp_flush_frame_bulk(bq);

	if (unlikely(mem->id != xa->mem.id)) {
		xdp_flush_frame_bulk(bq);
		bq->xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	}

	bq->q[bq->count++] = xdpf->data;
}
EXPORT_SYMBOL_GPL(xdp_return_frame_bulk);

void xdp_return_buff(struct xdp_buff *xdp)
{
	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp);
}

/* Only called for MEM_TYPE_PAGE_POOL see xdp.h */
void __xdp_release_frame(void *data, struct xdp_mem_info *mem)
{
	struct xdp_mem_allocator *xa;
	struct page *page;

	rcu_read_lock();
	xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	page = virt_to_head_page(data);
	if (xa)
		page_pool_release_page(xa->page_pool, page);
	rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(__xdp_release_frame);

void xdp_attachment_setup(struct xdp_attachment_info *info,
			  struct netdev_bpf *bpf)
{
	if (info->prog)
		bpf_prog_put(info->prog);
	info->prog = bpf->prog;
	info->flags = bpf->flags;
}
EXPORT_SYMBOL_GPL(xdp_attachment_setup);

struct xdp_frame *xdp_convert_zc_to_xdp_frame(struct xdp_buff *xdp)
{
	unsigned int metasize, totsize;
	void *addr, *data_to_copy;
	struct xdp_frame *xdpf;
	struct page *page;

	/* Clone into a MEM_TYPE_PAGE_ORDER0 xdp_frame. */
	metasize = xdp_data_meta_unsupported(xdp) ? 0 :
		   xdp->data - xdp->data_meta;
	totsize = xdp->data_end - xdp->data + metasize;

	if (sizeof(*xdpf) + totsize > PAGE_SIZE)
		return NULL;

	page = dev_alloc_page();
	if (!page)
		return NULL;

	addr = page_to_virt(page);
	xdpf = addr;
	memset(xdpf, 0, sizeof(*xdpf));

	addr += sizeof(*xdpf);
	data_to_copy = metasize ? xdp->data_meta : xdp->data;
	memcpy(addr, data_to_copy, totsize);

	xdpf->data = addr + metasize;
	xdpf->len = totsize - metasize;
	xdpf->headroom = 0;
	xdpf->metasize = metasize;
	xdpf->frame_sz = PAGE_SIZE;
	xdpf->mem.type = MEM_TYPE_PAGE_ORDER0;

	xsk_buff_free(xdp);
	return xdpf;
}
EXPORT_SYMBOL_GPL(xdp_convert_zc_to_xdp_frame);

/* Used by XDP_WARN macro, to avoid inlining WARN() in fast-path */
void xdp_warn(const char *msg, const char *func, const int line)
{
	WARN(1, "XDP_WARN: %s(line:%d): %s\n", func, line, msg);
};
EXPORT_SYMBOL_GPL(xdp_warn);

int xdp_alloc_skb_bulk(void **skbs, int n_skb, gfp_t gfp)
{
	n_skb = kmem_cache_alloc_bulk(skbuff_head_cache, gfp,
				      n_skb, skbs);
	if (unlikely(!n_skb))
		return -ENOMEM;

	return 0;
}
EXPORT_SYMBOL_GPL(xdp_alloc_skb_bulk);

struct sk_buff *__xdp_build_skb_from_frame(struct xdp_frame *xdpf,
					   struct sk_buff *skb,
					   struct net_device *dev)
{
	unsigned int headroom, frame_size;
	void *hard_start;

	/* Part of headroom was reserved to xdpf */
	headroom = sizeof(*xdpf) + xdpf->headroom;

	/* Memory size backing xdp_frame data already have reserved
	 * room for build_skb to place skb_shared_info in tailroom.
	 */
	frame_size = xdpf->frame_sz;

	hard_start = xdpf->data - headroom;
	skb = build_skb_around(skb, hard_start, frame_size);
	if (unlikely(!skb))
		return NULL;

	skb_reserve(skb, headroom);
	__skb_put(skb, xdpf->len);
	if (xdpf->metasize)
		skb_metadata_set(skb, xdpf->metasize);

	/* Essential SKB info: protocol and skb->dev */
	skb->protocol = eth_type_trans(skb, dev);

	/* Optional SKB info, currently missing:
	 * - HW checksum info		(skb->ip_summed)
	 * - HW RX hash			(skb_set_hash)
	 * - RX ring dev queue index	(skb_record_rx_queue)
	 */

	/* Until page_pool get SKB return path, release DMA here */
	xdp_release_frame(xdpf);

	/* Allow SKB to reuse area used by xdp_frame */
	xdp_scrub_frame(xdpf);

	return skb;
}
EXPORT_SYMBOL_GPL(__xdp_build_skb_from_frame);

struct sk_buff *xdp_build_skb_from_frame(struct xdp_frame *xdpf,
					 struct net_device *dev)
{
	struct sk_buff *skb;

	skb = kmem_cache_alloc(skbuff_head_cache, GFP_ATOMIC);
	if (unlikely(!skb))
		return NULL;

	memset(skb, 0, offsetof(struct sk_buff, tail));

	return __xdp_build_skb_from_frame(xdpf, skb, dev);
}
EXPORT_SYMBOL_GPL(xdp_build_skb_from_frame);
Commit	Line	Data
ddc64d0a	1	// SPDX-License-Identifier: GPL-2.0-only
aecd67b6 JDB	2	/* net/core/xdp.c
	3	*
	4	* Copyright (c) 2017 Jesper Dangaard Brouer, Red Hat Inc.
aecd67b6	5	*/
05296620 JK	6	#include <linux/bpf.h>
05296620 JK	7	#include <linux/filter.h>
aecd67b6 JDB	8	#include <linux/types.h>
aecd67b6 JDB	9	#include <linux/mm.h>
05296620	10	#include <linux/netdevice.h>
8d5d8852 JDB	11	#include <linux/slab.h>
	12	#include <linux/idr.h>
	13	#include <linux/rhashtable.h>
34cc0b33	14	#include <linux/bug.h>
57d0a1c1	15	#include <net/page_pool.h>
aecd67b6 JDB	16
aecd67b6 JDB	17	#include <net/xdp.h>
f033b688 JDB	18	#include <net/xdp_priv.h> /* struct xdp_mem_allocator */
f033b688 JDB	19	#include <trace/events/xdp.h>
2b43470a	20	#include <net/xdp_sock_drv.h>
aecd67b6 JDB	21
	22	#define REG_STATE_NEW 0x0
	23	#define REG_STATE_REGISTERED 0x1
	24	#define REG_STATE_UNREGISTERED 0x2
	25	#define REG_STATE_UNUSED 0x3
	26
8d5d8852 JDB	27	static DEFINE_IDA(mem_id_pool);
	28	static DEFINE_MUTEX(mem_id_lock);
	29	#define MEM_ID_MAX 0xFFFE
	30	#define MEM_ID_MIN 1
	31	static int mem_id_next = MEM_ID_MIN;
	32
	33	static bool mem_id_init; /* false */
	34	static struct rhashtable *mem_id_ht;
	35
8d5d8852 JDB	36	static u32 xdp_mem_id_hashfn(const void *data, u32 len, u32 seed)
	37	{
	38	const u32 *k = data;
	39	const u32 key = *k;
	40
c593642c	41	BUILD_BUG_ON(sizeof_field(struct xdp_mem_allocator, mem.id)
8d5d8852 JDB	42	!= sizeof(u32));
8d5d8852 JDB	43
9f9a7077 N	44	/* Use cyclic increasing ID as direct hash key */
9f9a7077 N	45	return key;
8d5d8852 JDB	46	}
	47
	48	static int xdp_mem_id_cmp(struct rhashtable_compare_arg *arg,
	49	const void *ptr)
	50	{
	51	const struct xdp_mem_allocator *xa = ptr;
	52	u32 mem_id = (u32 )arg->key;
	53
	54	return xa->mem.id != mem_id;
	55	}
	56
	57	static const struct rhashtable_params mem_id_rht_params = {
	58	.nelem_hint = 64,
	59	.head_offset = offsetof(struct xdp_mem_allocator, node),
	60	.key_offset = offsetof(struct xdp_mem_allocator, mem.id),
c593642c	61	.key_len = sizeof_field(struct xdp_mem_allocator, mem.id),
8d5d8852 JDB	62	.max_size = MEM_ID_MAX,
	63	.min_size = 8,
	64	.automatic_shrinking = true,
	65	.hashfn = xdp_mem_id_hashfn,
	66	.obj_cmpfn = xdp_mem_id_cmp,
	67	};
	68
	69	static void __xdp_mem_allocator_rcu_free(struct rcu_head *rcu)
	70	{
	71	struct xdp_mem_allocator *xa;
	72
	73	xa = container_of(rcu, struct xdp_mem_allocator, rcu);
	74
	75	/* Allow this ID to be reused */
	76	ida_simple_remove(&mem_id_pool, xa->mem.id);
	77
8d5d8852 JDB	78	kfree(xa);
	79	}
	80
c3f812ce	81	static void mem_xa_remove(struct xdp_mem_allocator *xa)
99c07c43	82	{
c3f812ce	83	trace_mem_disconnect(xa);
99c07c43	84
c3f812ce	85	if (!rhashtable_remove_fast(mem_id_ht, &xa->node, mem_id_rht_params))
99c07c43	86	call_rcu(&xa->rcu, __xdp_mem_allocator_rcu_free);
99c07c43 JDB	87	}
99c07c43 JDB	88
c3f812ce JL	89	static void mem_allocator_disconnect(void *allocator)
	90	{
	91	struct xdp_mem_allocator *xa;
	92	struct rhashtable_iter iter;
	93
86c76c09 JL	94	mutex_lock(&mem_id_lock);
86c76c09 JL	95
c3f812ce JL	96	rhashtable_walk_enter(mem_id_ht, &iter);
	97	do {
	98	rhashtable_walk_start(&iter);
	99
	100	while ((xa = rhashtable_walk_next(&iter)) && !IS_ERR(xa)) {
	101	if (xa->allocator == allocator)
	102	mem_xa_remove(xa);
	103	}
	104
	105	rhashtable_walk_stop(&iter);
99c07c43	106
c3f812ce JL	107	} while (xa == ERR_PTR(-EAGAIN));
c3f812ce JL	108	rhashtable_walk_exit(&iter);
86c76c09 JL	109
86c76c09 JL	110	mutex_unlock(&mem_id_lock);
c3f812ce JL	111	}
c3f812ce JL	112
dce5bd61	113	void xdp_rxq_info_unreg_mem_model(struct xdp_rxq_info *xdp_rxq)
8d5d8852 JDB	114	{
	115	struct xdp_mem_allocator *xa;
	116	int id = xdp_rxq->mem.id;
8d5d8852	117
dce5bd61 BT	118	if (xdp_rxq->reg_state != REG_STATE_REGISTERED) {
	119	WARN(1, "Missing register, driver bug");
	120	return;
	121	}
	122
8d5d8852 JDB	123	if (id == 0)
	124	return;
	125
c3f812ce JL	126	if (xdp_rxq->mem.type == MEM_TYPE_PAGE_POOL) {
	127	rcu_read_lock();
	128	xa = rhashtable_lookup(mem_id_ht, &id, mem_id_rht_params);
	129	page_pool_destroy(xa->page_pool);
	130	rcu_read_unlock();
99c07c43	131	}
8d5d8852	132	}
dce5bd61	133	EXPORT_SYMBOL_GPL(xdp_rxq_info_unreg_mem_model);
8d5d8852	134
aecd67b6 JDB	135	void xdp_rxq_info_unreg(struct xdp_rxq_info *xdp_rxq)
	136	{
	137	/* Simplify driver cleanup code paths, allow unreg "unused" */
	138	if (xdp_rxq->reg_state == REG_STATE_UNUSED)
	139	return;
	140
	141	WARN(!(xdp_rxq->reg_state == REG_STATE_REGISTERED), "Driver BUG");
	142
dce5bd61	143	xdp_rxq_info_unreg_mem_model(xdp_rxq);
8d5d8852	144
aecd67b6 JDB	145	xdp_rxq->reg_state = REG_STATE_UNREGISTERED;
aecd67b6 JDB	146	xdp_rxq->dev = NULL;
8d5d8852 JDB	147
	148	/* Reset mem info to defaults */
	149	xdp_rxq->mem.id = 0;
	150	xdp_rxq->mem.type = 0;
aecd67b6 JDB	151	}
	152	EXPORT_SYMBOL_GPL(xdp_rxq_info_unreg);
	153
	154	static void xdp_rxq_info_init(struct xdp_rxq_info *xdp_rxq)
	155	{
	156	memset(xdp_rxq, 0, sizeof(*xdp_rxq));
	157	}
	158
	159	/* Returns 0 on success, negative on failure */
	160	int xdp_rxq_info_reg(struct xdp_rxq_info *xdp_rxq,
b02e5a0e	161	struct net_device *dev, u32 queue_index, unsigned int napi_id)
aecd67b6 JDB	162	{
	163	if (xdp_rxq->reg_state == REG_STATE_UNUSED) {
	164	WARN(1, "Driver promised not to register this");
	165	return -EINVAL;
	166	}
	167
	168	if (xdp_rxq->reg_state == REG_STATE_REGISTERED) {
	169	WARN(1, "Missing unregister, handled but fix driver");
	170	xdp_rxq_info_unreg(xdp_rxq);
	171	}
	172
	173	if (!dev) {
	174	WARN(1, "Missing net_device from driver");
	175	return -ENODEV;
	176	}
	177
	178	/* State either UNREGISTERED or NEW */
	179	xdp_rxq_info_init(xdp_rxq);
	180	xdp_rxq->dev = dev;
	181	xdp_rxq->queue_index = queue_index;
b02e5a0e	182	xdp_rxq->napi_id = napi_id;
aecd67b6 JDB	183
	184	xdp_rxq->reg_state = REG_STATE_REGISTERED;
	185	return 0;
	186	}
	187	EXPORT_SYMBOL_GPL(xdp_rxq_info_reg);
	188
	189	void xdp_rxq_info_unused(struct xdp_rxq_info *xdp_rxq)
	190	{
	191	xdp_rxq->reg_state = REG_STATE_UNUSED;
	192	}
	193	EXPORT_SYMBOL_GPL(xdp_rxq_info_unused);
c0124f32 JDB	194
	195	bool xdp_rxq_info_is_reg(struct xdp_rxq_info *xdp_rxq)
	196	{
	197	return (xdp_rxq->reg_state == REG_STATE_REGISTERED);
	198	}
	199	EXPORT_SYMBOL_GPL(xdp_rxq_info_is_reg);
5ab073ff	200
8d5d8852 JDB	201	static int __mem_id_init_hash_table(void)
	202	{
	203	struct rhashtable *rht;
	204	int ret;
	205
	206	if (unlikely(mem_id_init))
	207	return 0;
	208
	209	rht = kzalloc(sizeof(*rht), GFP_KERNEL);
	210	if (!rht)
	211	return -ENOMEM;
	212
	213	ret = rhashtable_init(rht, &mem_id_rht_params);
	214	if (ret < 0) {
	215	kfree(rht);
	216	return ret;
	217	}
	218	mem_id_ht = rht;
	219	smp_mb(); /* mutex lock should provide enough pairing */
	220	mem_id_init = true;
	221
	222	return 0;
	223	}
	224
	225	/* Allocate a cyclic ID that maps to allocator pointer.
	226	* See: https://www.kernel.org/doc/html/latest/core-api/idr.html
	227	*
	228	* Caller must lock mem_id_lock.
	229	*/
	230	static int __mem_id_cyclic_get(gfp_t gfp)
	231	{
	232	int retries = 1;
	233	int id;
	234
	235	again:
	236	id = ida_simple_get(&mem_id_pool, mem_id_next, MEM_ID_MAX, gfp);
	237	if (id < 0) {
	238	if (id == -ENOSPC) {
	239	/* Cyclic allocator, reset next id */
	240	if (retries--) {
	241	mem_id_next = MEM_ID_MIN;
	242	goto again;
	243	}
	244	}
	245	return id; /* errno */
	246	}
	247	mem_id_next = id + 1;
	248
	249	return id;
	250	}
	251
57d0a1c1 JDB	252	static bool __is_supported_mem_type(enum xdp_mem_type type)
	253	{
	254	if (type == MEM_TYPE_PAGE_POOL)
	255	return is_page_pool_compiled_in();
	256
	257	if (type >= MEM_TYPE_MAX)
	258	return false;
	259
	260	return true;
	261	}
	262
5ab073ff JDB	263	int xdp_rxq_info_reg_mem_model(struct xdp_rxq_info *xdp_rxq,
	264	enum xdp_mem_type type, void *allocator)
	265	{
8d5d8852 JDB	266	struct xdp_mem_allocator *xdp_alloc;
	267	gfp_t gfp = GFP_KERNEL;
	268	int id, errno, ret;
	269	void *ptr;
	270
	271	if (xdp_rxq->reg_state != REG_STATE_REGISTERED) {
	272	WARN(1, "Missing register, driver bug");
	273	return -EFAULT;
	274	}
	275
57d0a1c1 JDB	276	if (!__is_supported_mem_type(type))
57d0a1c1 JDB	277	return -EOPNOTSUPP;
5ab073ff JDB	278
	279	xdp_rxq->mem.type = type;
	280
57d0a1c1	281	if (!allocator) {
0807892e	282	if (type == MEM_TYPE_PAGE_POOL)
57d0a1c1	283	return -EINVAL; /* Setup time check page_pool req */
8d5d8852	284	return 0;
57d0a1c1	285	}
8d5d8852 JDB	286
	287	/* Delay init of rhashtable to save memory if feature isn't used */
	288	if (!mem_id_init) {
	289	mutex_lock(&mem_id_lock);
	290	ret = __mem_id_init_hash_table();
	291	mutex_unlock(&mem_id_lock);
	292	if (ret < 0) {
	293	WARN_ON(1);
	294	return ret;
	295	}
	296	}
	297
	298	xdp_alloc = kzalloc(sizeof(*xdp_alloc), gfp);
	299	if (!xdp_alloc)
	300	return -ENOMEM;
	301
	302	mutex_lock(&mem_id_lock);
	303	id = __mem_id_cyclic_get(gfp);
	304	if (id < 0) {
	305	errno = id;
	306	goto err;
	307	}
	308	xdp_rxq->mem.id = id;
	309	xdp_alloc->mem = xdp_rxq->mem;
	310	xdp_alloc->allocator = allocator;
	311
	312	/* Insert allocator into ID lookup table */
	313	ptr = rhashtable_insert_slow(mem_id_ht, &id, &xdp_alloc->node);
	314	if (IS_ERR(ptr)) {
516a7593 JDB	315	ida_simple_remove(&mem_id_pool, xdp_rxq->mem.id);
516a7593 JDB	316	xdp_rxq->mem.id = 0;
8d5d8852 JDB	317	errno = PTR_ERR(ptr);
	318	goto err;
	319	}
	320
1da4bbef	321	if (type == MEM_TYPE_PAGE_POOL)
c3f812ce	322	page_pool_use_xdp_mem(allocator, mem_allocator_disconnect);
1da4bbef	323
8d5d8852	324	mutex_unlock(&mem_id_lock);
5ab073ff	325
f033b688	326	trace_mem_connect(xdp_alloc, xdp_rxq);
5ab073ff	327	return 0;
8d5d8852 JDB	328	err:
	329	mutex_unlock(&mem_id_lock);
	330	kfree(xdp_alloc);
	331	return errno;
5ab073ff JDB	332	}
5ab073ff JDB	333	EXPORT_SYMBOL_GPL(xdp_rxq_info_reg_mem_model);
8d5d8852	334
389ab7f0 JDB	335	/* XDP RX runs under NAPI protection, and in different delivery error
389ab7f0 JDB	336	* scenarios (e.g. queue full), it is possible to return the xdp_frame
baead859	337	* while still leveraging this protection. The @napi_direct boolean
389ab7f0	338	* is used for those calls sites. Thus, allowing for faster recycling
ed1182dc	339	* of xdp_frames/pages in those cases.
389ab7f0	340	*/
ed1182dc BT	341	static void __xdp_return(void data, struct xdp_mem_info mem, bool napi_direct,
ed1182dc BT	342	struct xdp_buff *xdp)
8d5d8852	343	{
57d0a1c1 JDB	344	struct xdp_mem_allocator *xa;
	345	struct page *page;
	346
	347	switch (mem->type) {
	348	case MEM_TYPE_PAGE_POOL:
	349	rcu_read_lock();
	350	/* mem->id is valid, checked in xdp_rxq_info_reg_mem_model() */
	351	xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	352	page = virt_to_head_page(data);
622d1369 OBL	353	if (napi_direct && xdp_return_frame_no_direct())
622d1369 OBL	354	napi_direct = false;
458de8a9	355	page_pool_put_full_page(xa->page_pool, page, napi_direct);
57d0a1c1 JDB	356	rcu_read_unlock();
	357	break;
	358	case MEM_TYPE_PAGE_SHARED:
8d5d8852	359	page_frag_free(data);
57d0a1c1 JDB	360	break;
	361	case MEM_TYPE_PAGE_ORDER0:
	362	page = virt_to_page(data); /* Assumes order0 page*/
8d5d8852	363	put_page(page);
57d0a1c1	364	break;
ed1182dc BT	365	case MEM_TYPE_XSK_BUFF_POOL:
	366	/* NB! Only valid from an xdp_buff! */
	367	xsk_buff_free(xdp);
	368	break;
57d0a1c1 JDB	369	default:
57d0a1c1 JDB	370	/* Not possible, checked in xdp_rxq_info_reg_mem_model() */
82c41671	371	WARN(1, "Incorrect XDP memory type (%d) usage", mem->type);
57d0a1c1	372	break;
8d5d8852 JDB	373	}
8d5d8852 JDB	374	}
c497176c BT	375
	376	void xdp_return_frame(struct xdp_frame *xdpf)
	377	{
ed1182dc	378	__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
c497176c	379	}
8d5d8852	380	EXPORT_SYMBOL_GPL(xdp_return_frame);
c497176c	381
389ab7f0 JDB	382	void xdp_return_frame_rx_napi(struct xdp_frame *xdpf)
389ab7f0 JDB	383	{
ed1182dc	384	__xdp_return(xdpf->data, &xdpf->mem, true, NULL);
389ab7f0 JDB	385	}
	386	EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi);
	387
89653987 LB	388	/* XDP bulk APIs introduce a defer/flush mechanism to return
	389	* pages belonging to the same xdp_mem_allocator object
	390	* (identified via the mem.id field) in bulk to optimize
	391	* I-cache and D-cache.
	392	* The bulk queue size is set to 16 to be aligned to how
	393	* XDP_REDIRECT bulking works. The bulk is flushed when
	394	* it is full or when mem.id changes.
	395	* xdp_frame_bulk is usually stored/allocated on the function
	396	* call-stack to avoid locking penalties.
	397	*/
	398	void xdp_flush_frame_bulk(struct xdp_frame_bulk *bq)
	399	{
	400	struct xdp_mem_allocator *xa = bq->xa;
89653987	401
78862447	402	if (unlikely(!xa \|\| !bq->count))
89653987 LB	403	return;
89653987 LB	404
78862447	405	page_pool_put_page_bulk(xa->page_pool, bq->q, bq->count);
89653987 LB	406	/* bq->xa is not cleared to save lookup, if mem.id same in next bulk */
	407	bq->count = 0;
	408	}
	409	EXPORT_SYMBOL_GPL(xdp_flush_frame_bulk);
	410
	411	/* Must be called with rcu_read_lock held */
	412	void xdp_return_frame_bulk(struct xdp_frame *xdpf,
	413	struct xdp_frame_bulk *bq)
	414	{
	415	struct xdp_mem_info *mem = &xdpf->mem;
	416	struct xdp_mem_allocator *xa;
	417
	418	if (mem->type != MEM_TYPE_PAGE_POOL) {
46d5e62d	419	__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
89653987 LB	420	return;
	421	}
	422
	423	xa = bq->xa;
	424	if (unlikely(!xa)) {
	425	xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	426	bq->count = 0;
	427	bq->xa = xa;
	428	}
	429
	430	if (bq->count == XDP_BULK_QUEUE_SIZE)
	431	xdp_flush_frame_bulk(bq);
	432
	433	if (unlikely(mem->id != xa->mem.id)) {
	434	xdp_flush_frame_bulk(bq);
	435	bq->xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	436	}
	437
	438	bq->q[bq->count++] = xdpf->data;
	439	}
	440	EXPORT_SYMBOL_GPL(xdp_return_frame_bulk);
	441
c497176c BT	442	void xdp_return_buff(struct xdp_buff *xdp)
c497176c BT	443	{
ed1182dc	444	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp);
c497176c	445	}
05296620	446
6bf071bf JDB	447	/* Only called for MEM_TYPE_PAGE_POOL see xdp.h */
	448	void __xdp_release_frame(void data, struct xdp_mem_info mem)
	449	{
	450	struct xdp_mem_allocator *xa;
	451	struct page *page;
	452
	453	rcu_read_lock();
	454	xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
	455	page = virt_to_head_page(data);
	456	if (xa)
	457	page_pool_release_page(xa->page_pool, page);
	458	rcu_read_unlock();
	459	}
	460	EXPORT_SYMBOL_GPL(__xdp_release_frame);
	461
05296620 JK	462	void xdp_attachment_setup(struct xdp_attachment_info *info,
	463	struct netdev_bpf *bpf)
	464	{
	465	if (info->prog)
	466	bpf_prog_put(info->prog);
	467	info->prog = bpf->prog;
	468	info->flags = bpf->flags;
	469	}
	470	EXPORT_SYMBOL_GPL(xdp_attachment_setup);
b0d1beef BT	471
	472	struct xdp_frame xdp_convert_zc_to_xdp_frame(struct xdp_buff xdp)
	473	{
72962167	474	unsigned int metasize, totsize;
b0d1beef BT	475	void addr, data_to_copy;
	476	struct xdp_frame *xdpf;
	477	struct page *page;
	478
	479	/* Clone into a MEM_TYPE_PAGE_ORDER0 xdp_frame. */
	480	metasize = xdp_data_meta_unsupported(xdp) ? 0 :
	481	xdp->data - xdp->data_meta;
b0d1beef BT	482	totsize = xdp->data_end - xdp->data + metasize;
	483
	484	if (sizeof(*xdpf) + totsize > PAGE_SIZE)
	485	return NULL;
	486
	487	page = dev_alloc_page();
	488	if (!page)
	489	return NULL;
	490
	491	addr = page_to_virt(page);
	492	xdpf = addr;
	493	memset(xdpf, 0, sizeof(*xdpf));
	494
	495	addr += sizeof(*xdpf);
	496	data_to_copy = metasize ? xdp->data_meta : xdp->data;
	497	memcpy(addr, data_to_copy, totsize);
	498
	499	xdpf->data = addr + metasize;
	500	xdpf->len = totsize - metasize;
	501	xdpf->headroom = 0;
	502	xdpf->metasize = metasize;
3ff23516	503	xdpf->frame_sz = PAGE_SIZE;
b0d1beef BT	504	xdpf->mem.type = MEM_TYPE_PAGE_ORDER0;
b0d1beef BT	505
82c41671	506	xsk_buff_free(xdp);
b0d1beef BT	507	return xdpf;
	508	}
	509	EXPORT_SYMBOL_GPL(xdp_convert_zc_to_xdp_frame);
34cc0b33 JDB	510
	511	/* Used by XDP_WARN macro, to avoid inlining WARN() in fast-path */
	512	void xdp_warn(const char msg, const char func, const int line)
	513	{
	514	WARN(1, "XDP_WARN: %s(line:%d): %s\n", func, line, msg);
	515	};
	516	EXPORT_SYMBOL_GPL(xdp_warn);
97a0e1ea	517
65e6dcf7 LB	518	int xdp_alloc_skb_bulk(void **skbs, int n_skb, gfp_t gfp)
	519	{
	520	n_skb = kmem_cache_alloc_bulk(skbuff_head_cache, gfp,
	521	n_skb, skbs);
	522	if (unlikely(!n_skb))
	523	return -ENOMEM;
	524
	525	return 0;
	526	}
	527	EXPORT_SYMBOL_GPL(xdp_alloc_skb_bulk);
	528
97a0e1ea LB	529	struct sk_buff __xdp_build_skb_from_frame(struct xdp_frame xdpf,
	530	struct sk_buff *skb,
	531	struct net_device *dev)
	532	{
	533	unsigned int headroom, frame_size;
	534	void *hard_start;
	535
	536	/* Part of headroom was reserved to xdpf */
	537	headroom = sizeof(*xdpf) + xdpf->headroom;
	538
	539	/* Memory size backing xdp_frame data already have reserved
	540	* room for build_skb to place skb_shared_info in tailroom.
	541	*/
	542	frame_size = xdpf->frame_sz;
	543
	544	hard_start = xdpf->data - headroom;
	545	skb = build_skb_around(skb, hard_start, frame_size);
	546	if (unlikely(!skb))
	547	return NULL;
	548
	549	skb_reserve(skb, headroom);
	550	__skb_put(skb, xdpf->len);
	551	if (xdpf->metasize)
	552	skb_metadata_set(skb, xdpf->metasize);
	553
	554	/* Essential SKB info: protocol and skb->dev */
	555	skb->protocol = eth_type_trans(skb, dev);
	556
	557	/* Optional SKB info, currently missing:
	558	* - HW checksum info (skb->ip_summed)
	559	* - HW RX hash (skb_set_hash)
	560	* - RX ring dev queue index (skb_record_rx_queue)
	561	*/
	562
	563	/* Until page_pool get SKB return path, release DMA here */
	564	xdp_release_frame(xdpf);
	565
	566	/* Allow SKB to reuse area used by xdp_frame */
	567	xdp_scrub_frame(xdpf);
	568
	569	return skb;
	570	}
	571	EXPORT_SYMBOL_GPL(__xdp_build_skb_from_frame);
89f479f0 LB	572
	573	struct sk_buff xdp_build_skb_from_frame(struct xdp_frame xdpf,
	574	struct net_device *dev)
	575	{
	576	struct sk_buff *skb;
	577
	578	skb = kmem_cache_alloc(skbuff_head_cache, GFP_ATOMIC);
	579	if (unlikely(!skb))
	580	return NULL;
	581
	582	memset(skb, 0, offsetof(struct sk_buff, tail));
	583
	584	return __xdp_build_skb_from_frame(xdpf, skb, dev);
	585	}
	586	EXPORT_SYMBOL_GPL(xdp_build_skb_from_frame);