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

// SPDX-License-Identifier: GPL-2.0

#include <net/xsk_buff_pool.h>
#include <net/xdp_sock.h>
#include <net/xdp_sock_drv.h>

#include "xsk_queue.h"
#include "xdp_umem.h"
#include "xsk.h"

void xp_add_xsk(struct xsk_buff_pool *pool, struct xdp_sock *xs)
{
	unsigned long flags;

	if (!xs->tx)
		return;

	spin_lock_irqsave(&pool->xsk_tx_list_lock, flags);
	list_add_rcu(&xs->tx_list, &pool->xsk_tx_list);
	spin_unlock_irqrestore(&pool->xsk_tx_list_lock, flags);
}

void xp_del_xsk(struct xsk_buff_pool *pool, struct xdp_sock *xs)
{
	unsigned long flags;

	if (!xs->tx)
		return;

	spin_lock_irqsave(&pool->xsk_tx_list_lock, flags);
	list_del_rcu(&xs->tx_list);
	spin_unlock_irqrestore(&pool->xsk_tx_list_lock, flags);
}

void xp_destroy(struct xsk_buff_pool *pool)
{
	if (!pool)
		return;

	kvfree(pool->heads);
	kvfree(pool);
}

struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
						struct xdp_umem *umem)
{
	struct xsk_buff_pool *pool;
	struct xdp_buff_xsk *xskb;
	u32 i;

	pool = kvzalloc(struct_size(pool, free_heads, umem->chunks),
			GFP_KERNEL);
	if (!pool)
		goto out;

	pool->heads = kvcalloc(umem->chunks, sizeof(*pool->heads), GFP_KERNEL);
	if (!pool->heads)
		goto out;

	pool->chunk_mask = ~((u64)umem->chunk_size - 1);
	pool->addrs_cnt = umem->size;
	pool->heads_cnt = umem->chunks;
	pool->free_heads_cnt = umem->chunks;
	pool->headroom = umem->headroom;
	pool->chunk_size = umem->chunk_size;
	pool->unaligned = umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG;
	pool->frame_len = umem->chunk_size - umem->headroom -
		XDP_PACKET_HEADROOM;
	pool->umem = umem;
	pool->addrs = umem->addrs;
	INIT_LIST_HEAD(&pool->free_list);
	INIT_LIST_HEAD(&pool->xsk_tx_list);
	spin_lock_init(&pool->xsk_tx_list_lock);
	refcount_set(&pool->users, 1);

	pool->fq = xs->fq_tmp;
	pool->cq = xs->cq_tmp;
	xs->fq_tmp = NULL;
	xs->cq_tmp = NULL;

	for (i = 0; i < pool->free_heads_cnt; i++) {
		xskb = &pool->heads[i];
		xskb->pool = pool;
		xskb->xdp.frame_sz = umem->chunk_size - umem->headroom;
		pool->free_heads[i] = xskb;
	}

	return pool;

out:
	xp_destroy(pool);
	return NULL;
}

void xp_set_rxq_info(struct xsk_buff_pool *pool, struct xdp_rxq_info *rxq)
{
	u32 i;

	for (i = 0; i < pool->heads_cnt; i++)
		pool->heads[i].xdp.rxq = rxq;
}
EXPORT_SYMBOL(xp_set_rxq_info);

static void xp_disable_drv_zc(struct xsk_buff_pool *pool)
{
	struct netdev_bpf bpf;
	int err;

	ASSERT_RTNL();

	if (pool->umem->zc) {
		bpf.command = XDP_SETUP_XSK_POOL;
		bpf.xsk.pool = NULL;
		bpf.xsk.queue_id = pool->queue_id;

		err = pool->netdev->netdev_ops->ndo_bpf(pool->netdev, &bpf);

		if (err)
			WARN(1, "Failed to disable zero-copy!\n");
	}
}

static int __xp_assign_dev(struct xsk_buff_pool *pool,
			   struct net_device *netdev, u16 queue_id, u16 flags)
{
	bool force_zc, force_copy;
	struct netdev_bpf bpf;
	int err = 0;

	ASSERT_RTNL();

	force_zc = flags & XDP_ZEROCOPY;
	force_copy = flags & XDP_COPY;

	if (force_zc && force_copy)
		return -EINVAL;

	if (xsk_get_pool_from_qid(netdev, queue_id))
		return -EBUSY;

	pool->netdev = netdev;
	pool->queue_id = queue_id;
	err = xsk_reg_pool_at_qid(netdev, pool, queue_id);
	if (err)
		return err;

	if (flags & XDP_USE_NEED_WAKEUP) {
		pool->uses_need_wakeup = true;
		/* Tx needs to be explicitly woken up the first time.
		 * Also for supporting drivers that do not implement this
		 * feature. They will always have to call sendto().
		 */
		pool->cached_need_wakeup = XDP_WAKEUP_TX;
	}

	dev_hold(netdev);

	if (force_copy)
		/* For copy-mode, we are done. */
		return 0;

	if (!netdev->netdev_ops->ndo_bpf ||
	    !netdev->netdev_ops->ndo_xsk_wakeup) {
		err = -EOPNOTSUPP;
		goto err_unreg_pool;
	}

	bpf.command = XDP_SETUP_XSK_POOL;
	bpf.xsk.pool = pool;
	bpf.xsk.queue_id = queue_id;

	err = netdev->netdev_ops->ndo_bpf(netdev, &bpf);
	if (err)
		goto err_unreg_pool;

	if (!pool->dma_pages) {
		WARN(1, "Driver did not DMA map zero-copy buffers");
		goto err_unreg_xsk;
	}
	pool->umem->zc = true;
	return 0;

err_unreg_xsk:
	xp_disable_drv_zc(pool);
err_unreg_pool:
	if (!force_zc)
		err = 0; /* fallback to copy mode */
	if (err)
		xsk_clear_pool_at_qid(netdev, queue_id);
	return err;
}

int xp_assign_dev(struct xsk_buff_pool *pool, struct net_device *dev,
		  u16 queue_id, u16 flags)
{
	return __xp_assign_dev(pool, dev, queue_id, flags);
}

int xp_assign_dev_shared(struct xsk_buff_pool *pool, struct xdp_umem *umem,
			 struct net_device *dev, u16 queue_id)
{
	u16 flags;

	/* One fill and completion ring required for each queue id. */
	if (!pool->fq || !pool->cq)
		return -EINVAL;

	flags = umem->zc ? XDP_ZEROCOPY : XDP_COPY;
	if (pool->uses_need_wakeup)
		flags |= XDP_USE_NEED_WAKEUP;

	return __xp_assign_dev(pool, dev, queue_id, flags);
}

void xp_clear_dev(struct xsk_buff_pool *pool)
{
	if (!pool->netdev)
		return;

	xp_disable_drv_zc(pool);
	xsk_clear_pool_at_qid(pool->netdev, pool->queue_id);
	dev_put(pool->netdev);
	pool->netdev = NULL;
}

static void xp_release_deferred(struct work_struct *work)
{
	struct xsk_buff_pool *pool = container_of(work, struct xsk_buff_pool,
						  work);

	rtnl_lock();
	xp_clear_dev(pool);
	rtnl_unlock();

	if (pool->fq) {
		xskq_destroy(pool->fq);
		pool->fq = NULL;
	}

	if (pool->cq) {
		xskq_destroy(pool->cq);
		pool->cq = NULL;
	}

	xdp_put_umem(pool->umem);
	xp_destroy(pool);
}

void xp_get_pool(struct xsk_buff_pool *pool)
{
	refcount_inc(&pool->users);
}

bool xp_put_pool(struct xsk_buff_pool *pool)
{
	if (!pool)
		return false;

	if (refcount_dec_and_test(&pool->users)) {
		INIT_WORK(&pool->work, xp_release_deferred);
		schedule_work(&pool->work);
		return true;
	}

	return false;
}

static struct xsk_dma_map *xp_find_dma_map(struct xsk_buff_pool *pool)
{
	struct xsk_dma_map *dma_map;

	list_for_each_entry(dma_map, &pool->umem->xsk_dma_list, list) {
		if (dma_map->netdev == pool->netdev)
			return dma_map;
	}

	return NULL;
}

static struct xsk_dma_map *xp_create_dma_map(struct device *dev, struct net_device *netdev,
					     u32 nr_pages, struct xdp_umem *umem)
{
	struct xsk_dma_map *dma_map;

	dma_map = kzalloc(sizeof(*dma_map), GFP_KERNEL);
	if (!dma_map)
		return NULL;

	dma_map->dma_pages = kvcalloc(nr_pages, sizeof(*dma_map->dma_pages), GFP_KERNEL);
	if (!dma_map->dma_pages) {
		kfree(dma_map);
		return NULL;
	}

	dma_map->netdev = netdev;
	dma_map->dev = dev;
	dma_map->dma_need_sync = false;
	dma_map->dma_pages_cnt = nr_pages;
	refcount_set(&dma_map->users, 1);
	list_add(&dma_map->list, &umem->xsk_dma_list);
	return dma_map;
}

static void xp_destroy_dma_map(struct xsk_dma_map *dma_map)
{
	list_del(&dma_map->list);
	kvfree(dma_map->dma_pages);
	kfree(dma_map);
}

static void __xp_dma_unmap(struct xsk_dma_map *dma_map, unsigned long attrs)
{
	dma_addr_t *dma;
	u32 i;

	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
		dma = &dma_map->dma_pages[i];
		if (*dma) {
			dma_unmap_page_attrs(dma_map->dev, *dma, PAGE_SIZE,
					     DMA_BIDIRECTIONAL, attrs);
			*dma = 0;
		}
	}

	xp_destroy_dma_map(dma_map);
}

void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs)
{
	struct xsk_dma_map *dma_map;

	if (pool->dma_pages_cnt == 0)
		return;

	dma_map = xp_find_dma_map(pool);
	if (!dma_map) {
		WARN(1, "Could not find dma_map for device");
		return;
	}

	if (!refcount_dec_and_test(&dma_map->users))
		return;

	__xp_dma_unmap(dma_map, attrs);
	kvfree(pool->dma_pages);
	pool->dma_pages_cnt = 0;
	pool->dev = NULL;
}
EXPORT_SYMBOL(xp_dma_unmap);

static void xp_check_dma_contiguity(struct xsk_dma_map *dma_map)
{
	u32 i;

	for (i = 0; i < dma_map->dma_pages_cnt - 1; i++) {
		if (dma_map->dma_pages[i] + PAGE_SIZE == dma_map->dma_pages[i + 1])
			dma_map->dma_pages[i] |= XSK_NEXT_PG_CONTIG_MASK;
		else
			dma_map->dma_pages[i] &= ~XSK_NEXT_PG_CONTIG_MASK;
	}
}

static int xp_init_dma_info(struct xsk_buff_pool *pool, struct xsk_dma_map *dma_map)
{
	pool->dma_pages = kvcalloc(dma_map->dma_pages_cnt, sizeof(*pool->dma_pages), GFP_KERNEL);
	if (!pool->dma_pages)
		return -ENOMEM;

	pool->dev = dma_map->dev;
	pool->dma_pages_cnt = dma_map->dma_pages_cnt;
	pool->dma_need_sync = dma_map->dma_need_sync;
	memcpy(pool->dma_pages, dma_map->dma_pages,
	       pool->dma_pages_cnt * sizeof(*pool->dma_pages));

	return 0;
}

int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
	       unsigned long attrs, struct page **pages, u32 nr_pages)
{
	struct xsk_dma_map *dma_map;
	dma_addr_t dma;
	int err;
	u32 i;

	dma_map = xp_find_dma_map(pool);
	if (dma_map) {
		err = xp_init_dma_info(pool, dma_map);
		if (err)
			return err;

		refcount_inc(&dma_map->users);
		return 0;
	}

	dma_map = xp_create_dma_map(dev, pool->netdev, nr_pages, pool->umem);
	if (!dma_map)
		return -ENOMEM;

	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
		dma = dma_map_page_attrs(dev, pages[i], 0, PAGE_SIZE,
					 DMA_BIDIRECTIONAL, attrs);
		if (dma_mapping_error(dev, dma)) {
			__xp_dma_unmap(dma_map, attrs);
			return -ENOMEM;
		}
		if (dma_need_sync(dev, dma))
			dma_map->dma_need_sync = true;
		dma_map->dma_pages[i] = dma;
	}

	if (pool->unaligned)
		xp_check_dma_contiguity(dma_map);

	err = xp_init_dma_info(pool, dma_map);
	if (err) {
		__xp_dma_unmap(dma_map, attrs);
		return err;
	}

	return 0;
}
EXPORT_SYMBOL(xp_dma_map);

static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
					  u64 addr)
{
	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
}

static bool xp_check_unaligned(struct xsk_buff_pool *pool, u64 *addr)
{
	*addr = xp_unaligned_extract_addr(*addr);
	if (*addr >= pool->addrs_cnt ||
	    *addr + pool->chunk_size > pool->addrs_cnt ||
	    xp_addr_crosses_non_contig_pg(pool, *addr))
		return false;
	return true;
}

static bool xp_check_aligned(struct xsk_buff_pool *pool, u64 *addr)
{
	*addr = xp_aligned_extract_addr(pool, *addr);
	return *addr < pool->addrs_cnt;
}

static struct xdp_buff_xsk *__xp_alloc(struct xsk_buff_pool *pool)
{
	struct xdp_buff_xsk *xskb;
	u64 addr;
	bool ok;

	if (pool->free_heads_cnt == 0)
		return NULL;

	xskb = pool->free_heads[--pool->free_heads_cnt];

	for (;;) {
		if (!xskq_cons_peek_addr_unchecked(pool->fq, &addr)) {
			pool->fq->queue_empty_descs++;
			xp_release(xskb);
			return NULL;
		}

		ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
		     xp_check_aligned(pool, &addr);
		if (!ok) {
			pool->fq->invalid_descs++;
			xskq_cons_release(pool->fq);
			continue;
		}
		break;
	}
	xskq_cons_release(pool->fq);

	xskb->orig_addr = addr;
	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
	if (pool->dma_pages_cnt) {
		xskb->frame_dma = (pool->dma_pages[addr >> PAGE_SHIFT] &
				   ~XSK_NEXT_PG_CONTIG_MASK) +
				  (addr & ~PAGE_MASK);
		xskb->dma = xskb->frame_dma + pool->headroom +
			    XDP_PACKET_HEADROOM;
	}
	return xskb;
}

struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool)
{
	struct xdp_buff_xsk *xskb;

	if (!pool->free_list_cnt) {
		xskb = __xp_alloc(pool);
		if (!xskb)
			return NULL;
	} else {
		pool->free_list_cnt--;
		xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk,
					free_list_node);
		list_del(&xskb->free_list_node);
	}

	xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM;
	xskb->xdp.data_meta = xskb->xdp.data;

	if (pool->dma_need_sync) {
		dma_sync_single_range_for_device(pool->dev, xskb->dma, 0,
						 pool->frame_len,
						 DMA_BIDIRECTIONAL);
	}
	return &xskb->xdp;
}
EXPORT_SYMBOL(xp_alloc);

bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count)
{
	if (pool->free_list_cnt >= count)
		return true;
	return xskq_cons_has_entries(pool->fq, count - pool->free_list_cnt);
}
EXPORT_SYMBOL(xp_can_alloc);

void xp_free(struct xdp_buff_xsk *xskb)
{
	xskb->pool->free_list_cnt++;
	list_add(&xskb->free_list_node, &xskb->pool->free_list);
}
EXPORT_SYMBOL(xp_free);

void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr)
{
	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	return pool->addrs + addr;
}
EXPORT_SYMBOL(xp_raw_get_data);

dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
{
	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	return (pool->dma_pages[addr >> PAGE_SHIFT] &
		~XSK_NEXT_PG_CONTIG_MASK) +
		(addr & ~PAGE_MASK);
}
EXPORT_SYMBOL(xp_raw_get_dma);

void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
{
	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
				      xskb->pool->frame_len, DMA_BIDIRECTIONAL);
}
EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);

void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
				 size_t size)
{
	dma_sync_single_range_for_device(pool->dev, dma, 0,
					 size, DMA_BIDIRECTIONAL);
}
EXPORT_SYMBOL(xp_dma_sync_for_device_slow);
Commit	Line	Data
2b43470a BT	1	// SPDX-License-Identifier: GPL-2.0
	2
	3	#include <net/xsk_buff_pool.h>
	4	#include <net/xdp_sock.h>
1c1efc2a	5	#include <net/xdp_sock_drv.h>
2b43470a BT	6
2b43470a BT	7	#include "xsk_queue.h"
1c1efc2a MK	8	#include "xdp_umem.h"
1c1efc2a MK	9	#include "xsk.h"
2b43470a	10
a5aa8e52 MK	11	void xp_add_xsk(struct xsk_buff_pool pool, struct xdp_sock xs)
	12	{
	13	unsigned long flags;
	14
	15	if (!xs->tx)
	16	return;
	17
	18	spin_lock_irqsave(&pool->xsk_tx_list_lock, flags);
	19	list_add_rcu(&xs->tx_list, &pool->xsk_tx_list);
	20	spin_unlock_irqrestore(&pool->xsk_tx_list_lock, flags);
	21	}
	22
	23	void xp_del_xsk(struct xsk_buff_pool pool, struct xdp_sock xs)
	24	{
	25	unsigned long flags;
	26
	27	if (!xs->tx)
	28	return;
	29
	30	spin_lock_irqsave(&pool->xsk_tx_list_lock, flags);
	31	list_del_rcu(&xs->tx_list);
	32	spin_unlock_irqrestore(&pool->xsk_tx_list_lock, flags);
	33	}
	34
2b43470a BT	35	void xp_destroy(struct xsk_buff_pool *pool)
	36	{
	37	if (!pool)
	38	return;
	39
2b43470a BT	40	kvfree(pool->heads);
	41	kvfree(pool);
	42	}
	43
1c1efc2a MK	44	struct xsk_buff_pool xp_create_and_assign_umem(struct xdp_sock xs,
1c1efc2a MK	45	struct xdp_umem *umem)
2b43470a BT	46	{
	47	struct xsk_buff_pool *pool;
	48	struct xdp_buff_xsk *xskb;
2b43470a BT	49	u32 i;
2b43470a BT	50
1c1efc2a MK	51	pool = kvzalloc(struct_size(pool, free_heads, umem->chunks),
1c1efc2a MK	52	GFP_KERNEL);
2b43470a BT	53	if (!pool)
	54	goto out;
	55
1c1efc2a	56	pool->heads = kvcalloc(umem->chunks, sizeof(*pool->heads), GFP_KERNEL);
2b43470a BT	57	if (!pool->heads)
	58	goto out;
	59
1c1efc2a MK	60	pool->chunk_mask = ~((u64)umem->chunk_size - 1);
	61	pool->addrs_cnt = umem->size;
	62	pool->heads_cnt = umem->chunks;
	63	pool->free_heads_cnt = umem->chunks;
	64	pool->headroom = umem->headroom;
	65	pool->chunk_size = umem->chunk_size;
	66	pool->unaligned = umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG;
	67	pool->frame_len = umem->chunk_size - umem->headroom -
	68	XDP_PACKET_HEADROOM;
1742b3d5	69	pool->umem = umem;
7f7ffa4e	70	pool->addrs = umem->addrs;
2b43470a	71	INIT_LIST_HEAD(&pool->free_list);
a5aa8e52 MK	72	INIT_LIST_HEAD(&pool->xsk_tx_list);
a5aa8e52 MK	73	spin_lock_init(&pool->xsk_tx_list_lock);
1c1efc2a	74	refcount_set(&pool->users, 1);
2b43470a	75
7361f9c3 MK	76	pool->fq = xs->fq_tmp;
	77	pool->cq = xs->cq_tmp;
	78	xs->fq_tmp = NULL;
	79	xs->cq_tmp = NULL;
	80
2b43470a BT	81	for (i = 0; i < pool->free_heads_cnt; i++) {
	82	xskb = &pool->heads[i];
	83	xskb->pool = pool;
1c1efc2a	84	xskb->xdp.frame_sz = umem->chunk_size - umem->headroom;
2b43470a BT	85	pool->free_heads[i] = xskb;
	86	}
	87
7f7ffa4e	88	return pool;
2b43470a BT	89
	90	out:
	91	xp_destroy(pool);
	92	return NULL;
	93	}
	94
2b43470a BT	95	void xp_set_rxq_info(struct xsk_buff_pool pool, struct xdp_rxq_info rxq)
	96	{
	97	u32 i;
	98
	99	for (i = 0; i < pool->heads_cnt; i++)
	100	pool->heads[i].xdp.rxq = rxq;
	101	}
	102	EXPORT_SYMBOL(xp_set_rxq_info);
	103
921b6869 MK	104	static void xp_disable_drv_zc(struct xsk_buff_pool *pool)
	105	{
	106	struct netdev_bpf bpf;
	107	int err;
	108
	109	ASSERT_RTNL();
	110
	111	if (pool->umem->zc) {
	112	bpf.command = XDP_SETUP_XSK_POOL;
	113	bpf.xsk.pool = NULL;
	114	bpf.xsk.queue_id = pool->queue_id;
	115
	116	err = pool->netdev->netdev_ops->ndo_bpf(pool->netdev, &bpf);
	117
	118	if (err)
	119	WARN(1, "Failed to disable zero-copy!\n");
	120	}
	121	}
	122
b5aea28d MK	123	static int __xp_assign_dev(struct xsk_buff_pool *pool,
b5aea28d MK	124	struct net_device *netdev, u16 queue_id, u16 flags)
1c1efc2a	125	{
1c1efc2a MK	126	bool force_zc, force_copy;
	127	struct netdev_bpf bpf;
	128	int err = 0;
	129
	130	ASSERT_RTNL();
	131
	132	force_zc = flags & XDP_ZEROCOPY;
	133	force_copy = flags & XDP_COPY;
	134
	135	if (force_zc && force_copy)
	136	return -EINVAL;
	137
c2d3d6a4	138	if (xsk_get_pool_from_qid(netdev, queue_id))
1c1efc2a MK	139	return -EBUSY;
1c1efc2a MK	140
921b6869 MK	141	pool->netdev = netdev;
921b6869 MK	142	pool->queue_id = queue_id;
c2d3d6a4	143	err = xsk_reg_pool_at_qid(netdev, pool, queue_id);
1c1efc2a MK	144	if (err)
	145	return err;
	146
	147	if (flags & XDP_USE_NEED_WAKEUP) {
c2d3d6a4	148	pool->uses_need_wakeup = true;
1c1efc2a MK	149	/* Tx needs to be explicitly woken up the first time.
	150	* Also for supporting drivers that do not implement this
	151	* feature. They will always have to call sendto().
	152	*/
c2d3d6a4	153	pool->cached_need_wakeup = XDP_WAKEUP_TX;
1c1efc2a MK	154	}
1c1efc2a MK	155
c2d3d6a4 MK	156	dev_hold(netdev);
c2d3d6a4 MK	157
1c1efc2a MK	158	if (force_copy)
	159	/* For copy-mode, we are done. */
	160	return 0;
	161
c2d3d6a4 MK	162	if (!netdev->netdev_ops->ndo_bpf \|\|
c2d3d6a4 MK	163	!netdev->netdev_ops->ndo_xsk_wakeup) {
1c1efc2a MK	164	err = -EOPNOTSUPP;
	165	goto err_unreg_pool;
	166	}
	167
	168	bpf.command = XDP_SETUP_XSK_POOL;
	169	bpf.xsk.pool = pool;
	170	bpf.xsk.queue_id = queue_id;
	171
c2d3d6a4	172	err = netdev->netdev_ops->ndo_bpf(netdev, &bpf);
1c1efc2a MK	173	if (err)
	174	goto err_unreg_pool;
	175
921b6869 MK	176	if (!pool->dma_pages) {
	177	WARN(1, "Driver did not DMA map zero-copy buffers");
	178	goto err_unreg_xsk;
	179	}
c2d3d6a4	180	pool->umem->zc = true;
1c1efc2a MK	181	return 0;
1c1efc2a MK	182
921b6869 MK	183	err_unreg_xsk:
921b6869 MK	184	xp_disable_drv_zc(pool);
1c1efc2a MK	185	err_unreg_pool:
	186	if (!force_zc)
	187	err = 0; /* fallback to copy mode */
	188	if (err)
c2d3d6a4	189	xsk_clear_pool_at_qid(netdev, queue_id);
1c1efc2a MK	190	return err;
	191	}
	192
b5aea28d MK	193	int xp_assign_dev(struct xsk_buff_pool pool, struct net_device dev,
	194	u16 queue_id, u16 flags)
	195	{
	196	return __xp_assign_dev(pool, dev, queue_id, flags);
	197	}
	198
	199	int xp_assign_dev_shared(struct xsk_buff_pool pool, struct xdp_umem umem,
	200	struct net_device *dev, u16 queue_id)
	201	{
	202	u16 flags;
	203
	204	/* One fill and completion ring required for each queue id. */
	205	if (!pool->fq \|\| !pool->cq)
	206	return -EINVAL;
	207
	208	flags = umem->zc ? XDP_ZEROCOPY : XDP_COPY;
	209	if (pool->uses_need_wakeup)
	210	flags \|= XDP_USE_NEED_WAKEUP;
	211
	212	return __xp_assign_dev(pool, dev, queue_id, flags);
	213	}
	214
1c1efc2a MK	215	void xp_clear_dev(struct xsk_buff_pool *pool)
1c1efc2a MK	216	{
c2d3d6a4	217	if (!pool->netdev)
1c1efc2a MK	218	return;
1c1efc2a MK	219
921b6869	220	xp_disable_drv_zc(pool);
c2d3d6a4 MK	221	xsk_clear_pool_at_qid(pool->netdev, pool->queue_id);
	222	dev_put(pool->netdev);
	223	pool->netdev = NULL;
1c1efc2a MK	224	}
	225
	226	static void xp_release_deferred(struct work_struct *work)
	227	{
	228	struct xsk_buff_pool *pool = container_of(work, struct xsk_buff_pool,
	229	work);
	230
	231	rtnl_lock();
	232	xp_clear_dev(pool);
	233	rtnl_unlock();
	234
7361f9c3 MK	235	if (pool->fq) {
	236	xskq_destroy(pool->fq);
	237	pool->fq = NULL;
	238	}
	239
	240	if (pool->cq) {
	241	xskq_destroy(pool->cq);
	242	pool->cq = NULL;
	243	}
	244
1c1efc2a MK	245	xdp_put_umem(pool->umem);
	246	xp_destroy(pool);
	247	}
	248
	249	void xp_get_pool(struct xsk_buff_pool *pool)
	250	{
	251	refcount_inc(&pool->users);
	252	}
	253
e5e1a4bc	254	bool xp_put_pool(struct xsk_buff_pool *pool)
1c1efc2a MK	255	{
1c1efc2a MK	256	if (!pool)
e5e1a4bc	257	return false;
1c1efc2a MK	258
	259	if (refcount_dec_and_test(&pool->users)) {
	260	INIT_WORK(&pool->work, xp_release_deferred);
	261	schedule_work(&pool->work);
e5e1a4bc	262	return true;
1c1efc2a	263	}
e5e1a4bc MK	264
e5e1a4bc MK	265	return false;
1c1efc2a MK	266	}
1c1efc2a MK	267
921b6869 MK	268	static struct xsk_dma_map xp_find_dma_map(struct xsk_buff_pool pool)
	269	{
	270	struct xsk_dma_map *dma_map;
	271
	272	list_for_each_entry(dma_map, &pool->umem->xsk_dma_list, list) {
	273	if (dma_map->netdev == pool->netdev)
	274	return dma_map;
	275	}
	276
	277	return NULL;
	278	}
	279
	280	static struct xsk_dma_map xp_create_dma_map(struct device dev, struct net_device *netdev,
	281	u32 nr_pages, struct xdp_umem *umem)
	282	{
	283	struct xsk_dma_map *dma_map;
	284
	285	dma_map = kzalloc(sizeof(*dma_map), GFP_KERNEL);
	286	if (!dma_map)
	287	return NULL;
	288
	289	dma_map->dma_pages = kvcalloc(nr_pages, sizeof(*dma_map->dma_pages), GFP_KERNEL);
1d6fd78a	290	if (!dma_map->dma_pages) {
921b6869 MK	291	kfree(dma_map);
	292	return NULL;
	293	}
	294
	295	dma_map->netdev = netdev;
	296	dma_map->dev = dev;
	297	dma_map->dma_need_sync = false;
	298	dma_map->dma_pages_cnt = nr_pages;
bf74a370	299	refcount_set(&dma_map->users, 1);
921b6869 MK	300	list_add(&dma_map->list, &umem->xsk_dma_list);
	301	return dma_map;
	302	}
	303
	304	static void xp_destroy_dma_map(struct xsk_dma_map *dma_map)
	305	{
	306	list_del(&dma_map->list);
	307	kvfree(dma_map->dma_pages);
	308	kfree(dma_map);
	309	}
	310
	311	static void __xp_dma_unmap(struct xsk_dma_map *dma_map, unsigned long attrs)
2b43470a BT	312	{
	313	dma_addr_t *dma;
	314	u32 i;
	315
921b6869 MK	316	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
921b6869 MK	317	dma = &dma_map->dma_pages[i];
2b43470a	318	if (*dma) {
921b6869	319	dma_unmap_page_attrs(dma_map->dev, *dma, PAGE_SIZE,
2b43470a BT	320	DMA_BIDIRECTIONAL, attrs);
	321	*dma = 0;
	322	}
	323	}
	324
921b6869 MK	325	xp_destroy_dma_map(dma_map);
	326	}
	327
	328	void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs)
	329	{
	330	struct xsk_dma_map *dma_map;
	331
	332	if (pool->dma_pages_cnt == 0)
	333	return;
	334
	335	dma_map = xp_find_dma_map(pool);
	336	if (!dma_map) {
	337	WARN(1, "Could not find dma_map for device");
	338	return;
	339	}
	340
	341	if (!refcount_dec_and_test(&dma_map->users))
	342	return;
	343
	344	__xp_dma_unmap(dma_map, attrs);
2b43470a BT	345	kvfree(pool->dma_pages);
	346	pool->dma_pages_cnt = 0;
	347	pool->dev = NULL;
	348	}
	349	EXPORT_SYMBOL(xp_dma_unmap);
	350
921b6869	351	static void xp_check_dma_contiguity(struct xsk_dma_map *dma_map)
2b43470a BT	352	{
	353	u32 i;
	354
921b6869 MK	355	for (i = 0; i < dma_map->dma_pages_cnt - 1; i++) {
	356	if (dma_map->dma_pages[i] + PAGE_SIZE == dma_map->dma_pages[i + 1])
	357	dma_map->dma_pages[i] \|= XSK_NEXT_PG_CONTIG_MASK;
2b43470a	358	else
921b6869	359	dma_map->dma_pages[i] &= ~XSK_NEXT_PG_CONTIG_MASK;
2b43470a BT	360	}
	361	}
	362
921b6869 MK	363	static int xp_init_dma_info(struct xsk_buff_pool pool, struct xsk_dma_map dma_map)
	364	{
	365	pool->dma_pages = kvcalloc(dma_map->dma_pages_cnt, sizeof(*pool->dma_pages), GFP_KERNEL);
	366	if (!pool->dma_pages)
	367	return -ENOMEM;
	368
	369	pool->dev = dma_map->dev;
	370	pool->dma_pages_cnt = dma_map->dma_pages_cnt;
	371	pool->dma_need_sync = dma_map->dma_need_sync;
921b6869 MK	372	memcpy(pool->dma_pages, dma_map->dma_pages,
	373	pool->dma_pages_cnt * sizeof(*pool->dma_pages));
	374
	375	return 0;
	376	}
	377
2b43470a BT	378	int xp_dma_map(struct xsk_buff_pool pool, struct device dev,
	379	unsigned long attrs, struct page **pages, u32 nr_pages)
	380	{
921b6869	381	struct xsk_dma_map *dma_map;
2b43470a	382	dma_addr_t dma;
921b6869	383	int err;
2b43470a BT	384	u32 i;
2b43470a BT	385
921b6869 MK	386	dma_map = xp_find_dma_map(pool);
	387	if (dma_map) {
	388	err = xp_init_dma_info(pool, dma_map);
	389	if (err)
	390	return err;
2b43470a	391
bf74a370	392	refcount_inc(&dma_map->users);
921b6869 MK	393	return 0;
921b6869 MK	394	}
2b43470a	395
921b6869 MK	396	dma_map = xp_create_dma_map(dev, pool->netdev, nr_pages, pool->umem);
	397	if (!dma_map)
	398	return -ENOMEM;
	399
	400	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
2b43470a BT	401	dma = dma_map_page_attrs(dev, pages[i], 0, PAGE_SIZE,
	402	DMA_BIDIRECTIONAL, attrs);
	403	if (dma_mapping_error(dev, dma)) {
921b6869	404	__xp_dma_unmap(dma_map, attrs);
2b43470a BT	405	return -ENOMEM;
2b43470a BT	406	}
7e024575	407	if (dma_need_sync(dev, dma))
921b6869 MK	408	dma_map->dma_need_sync = true;
921b6869 MK	409	dma_map->dma_pages[i] = dma;
2b43470a BT	410	}
	411
	412	if (pool->unaligned)
921b6869 MK	413	xp_check_dma_contiguity(dma_map);
	414
	415	err = xp_init_dma_info(pool, dma_map);
	416	if (err) {
	417	__xp_dma_unmap(dma_map, attrs);
	418	return err;
	419	}
	420
2b43470a BT	421	return 0;
	422	}
	423	EXPORT_SYMBOL(xp_dma_map);
	424
2b43470a BT	425	static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
	426	u64 addr)
	427	{
	428	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
	429	}
	430
2b43470a BT	431	static bool xp_check_unaligned(struct xsk_buff_pool pool, u64 addr)
	432	{
	433	addr = xp_unaligned_extract_addr(addr);
	434	if (*addr >= pool->addrs_cnt \|\|
	435	*addr + pool->chunk_size > pool->addrs_cnt \|\|
	436	xp_addr_crosses_non_contig_pg(pool, *addr))
	437	return false;
	438	return true;
	439	}
	440
	441	static bool xp_check_aligned(struct xsk_buff_pool pool, u64 addr)
	442	{
	443	addr = xp_aligned_extract_addr(pool, addr);
	444	return *addr < pool->addrs_cnt;
	445	}
	446
	447	static struct xdp_buff_xsk __xp_alloc(struct xsk_buff_pool pool)
	448	{
	449	struct xdp_buff_xsk *xskb;
	450	u64 addr;
	451	bool ok;
	452
	453	if (pool->free_heads_cnt == 0)
	454	return NULL;
	455
	456	xskb = pool->free_heads[--pool->free_heads_cnt];
	457
	458	for (;;) {
	459	if (!xskq_cons_peek_addr_unchecked(pool->fq, &addr)) {
8aa5a335	460	pool->fq->queue_empty_descs++;
2b43470a BT	461	xp_release(xskb);
	462	return NULL;
	463	}
	464
	465	ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
	466	xp_check_aligned(pool, &addr);
	467	if (!ok) {
	468	pool->fq->invalid_descs++;
	469	xskq_cons_release(pool->fq);
	470	continue;
	471	}
	472	break;
	473	}
	474	xskq_cons_release(pool->fq);
	475
	476	xskb->orig_addr = addr;
	477	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
	478	if (pool->dma_pages_cnt) {
	479	xskb->frame_dma = (pool->dma_pages[addr >> PAGE_SHIFT] &
	480	~XSK_NEXT_PG_CONTIG_MASK) +
	481	(addr & ~PAGE_MASK);
	482	xskb->dma = xskb->frame_dma + pool->headroom +
	483	XDP_PACKET_HEADROOM;
	484	}
	485	return xskb;
	486	}
	487
	488	struct xdp_buff xp_alloc(struct xsk_buff_pool pool)
	489	{
	490	struct xdp_buff_xsk *xskb;
	491
	492	if (!pool->free_list_cnt) {
	493	xskb = __xp_alloc(pool);
	494	if (!xskb)
	495	return NULL;
	496	} else {
	497	pool->free_list_cnt--;
	498	xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk,
	499	free_list_node);
	500	list_del(&xskb->free_list_node);
	501	}
	502
	503	xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM;
	504	xskb->xdp.data_meta = xskb->xdp.data;
	505
91d5b702	506	if (pool->dma_need_sync) {
2b43470a BT	507	dma_sync_single_range_for_device(pool->dev, xskb->dma, 0,
	508	pool->frame_len,
	509	DMA_BIDIRECTIONAL);
	510	}
	511	return &xskb->xdp;
	512	}
	513	EXPORT_SYMBOL(xp_alloc);
	514
	515	bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count)
	516	{
	517	if (pool->free_list_cnt >= count)
	518	return true;
	519	return xskq_cons_has_entries(pool->fq, count - pool->free_list_cnt);
	520	}
	521	EXPORT_SYMBOL(xp_can_alloc);
	522
	523	void xp_free(struct xdp_buff_xsk *xskb)
	524	{
	525	xskb->pool->free_list_cnt++;
	526	list_add(&xskb->free_list_node, &xskb->pool->free_list);
	527	}
	528	EXPORT_SYMBOL(xp_free);
	529
2b43470a BT	530	void xp_raw_get_data(struct xsk_buff_pool pool, u64 addr)
	531	{
	532	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	533	return pool->addrs + addr;
	534	}
	535	EXPORT_SYMBOL(xp_raw_get_data);
	536
	537	dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
	538	{
	539	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	540	return (pool->dma_pages[addr >> PAGE_SHIFT] &
	541	~XSK_NEXT_PG_CONTIG_MASK) +
	542	(addr & ~PAGE_MASK);
	543	}
	544	EXPORT_SYMBOL(xp_raw_get_dma);
	545
26062b18	546	void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
2b43470a	547	{
2b43470a BT	548	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
	549	xskb->pool->frame_len, DMA_BIDIRECTIONAL);
	550	}
26062b18	551	EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);
2b43470a	552
26062b18 BT	553	void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
26062b18 BT	554	size_t size)
2b43470a	555	{
2b43470a BT	556	dma_sync_single_range_for_device(pool->dev, dma, 0,
	557	size, DMA_BIDIRECTIONAL);
	558	}
26062b18	559	EXPORT_SYMBOL(xp_dma_sync_for_device_slow);