[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);
	spin_lock_init(&pool->cq_lock);
	refcount_set(&pool->users, 1);

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

	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() or poll().
	 */
	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");
		err = -EINVAL;
		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);
		dev_put(netdev);
	}
	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, false);
	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);
f09ced40	74	spin_lock_init(&pool->cq_lock);
1c1efc2a	75	refcount_set(&pool->users, 1);
2b43470a	76
7361f9c3 MK	77	pool->fq = xs->fq_tmp;
7361f9c3 MK	78	pool->cq = xs->cq_tmp;
7361f9c3	79
2b43470a BT	80	for (i = 0; i < pool->free_heads_cnt; i++) {
	81	xskb = &pool->heads[i];
	82	xskb->pool = pool;
1c1efc2a	83	xskb->xdp.frame_sz = umem->chunk_size - umem->headroom;
2b43470a BT	84	pool->free_heads[i] = xskb;
	85	}
	86
7f7ffa4e	87	return pool;
2b43470a BT	88
	89	out:
	90	xp_destroy(pool);
	91	return NULL;
	92	}
	93
2b43470a BT	94	void xp_set_rxq_info(struct xsk_buff_pool pool, struct xdp_rxq_info rxq)
	95	{
	96	u32 i;
	97
	98	for (i = 0; i < pool->heads_cnt; i++)
	99	pool->heads[i].xdp.rxq = rxq;
	100	}
	101	EXPORT_SYMBOL(xp_set_rxq_info);
	102
921b6869 MK	103	static void xp_disable_drv_zc(struct xsk_buff_pool *pool)
	104	{
	105	struct netdev_bpf bpf;
	106	int err;
	107
	108	ASSERT_RTNL();
	109
	110	if (pool->umem->zc) {
	111	bpf.command = XDP_SETUP_XSK_POOL;
	112	bpf.xsk.pool = NULL;
	113	bpf.xsk.queue_id = pool->queue_id;
	114
	115	err = pool->netdev->netdev_ops->ndo_bpf(pool->netdev, &bpf);
	116
	117	if (err)
	118	WARN(1, "Failed to disable zero-copy!\n");
	119	}
	120	}
	121
b5aea28d MK	122	static int __xp_assign_dev(struct xsk_buff_pool *pool,
b5aea28d MK	123	struct net_device *netdev, u16 queue_id, u16 flags)
1c1efc2a	124	{
1c1efc2a MK	125	bool force_zc, force_copy;
	126	struct netdev_bpf bpf;
	127	int err = 0;
	128
	129	ASSERT_RTNL();
	130
	131	force_zc = flags & XDP_ZEROCOPY;
	132	force_copy = flags & XDP_COPY;
	133
	134	if (force_zc && force_copy)
	135	return -EINVAL;
	136
c2d3d6a4	137	if (xsk_get_pool_from_qid(netdev, queue_id))
1c1efc2a MK	138	return -EBUSY;
1c1efc2a MK	139
921b6869 MK	140	pool->netdev = netdev;
921b6869 MK	141	pool->queue_id = queue_id;
c2d3d6a4	142	err = xsk_reg_pool_at_qid(netdev, pool, queue_id);
1c1efc2a MK	143	if (err)
	144	return err;
	145
e3920818	146	if (flags & XDP_USE_NEED_WAKEUP)
c2d3d6a4	147	pool->uses_need_wakeup = true;
e3920818 BT	148	/* Tx needs to be explicitly woken up the first time. Also
	149	* for supporting drivers that do not implement this
	150	* feature. They will always have to call sendto() or poll().
	151	*/
	152	pool->cached_need_wakeup = XDP_WAKEUP_TX;
1c1efc2a	153
c2d3d6a4 MK	154	dev_hold(netdev);
c2d3d6a4 MK	155
1c1efc2a MK	156	if (force_copy)
	157	/* For copy-mode, we are done. */
	158	return 0;
	159
c2d3d6a4 MK	160	if (!netdev->netdev_ops->ndo_bpf \|\|
c2d3d6a4 MK	161	!netdev->netdev_ops->ndo_xsk_wakeup) {
1c1efc2a MK	162	err = -EOPNOTSUPP;
	163	goto err_unreg_pool;
	164	}
	165
	166	bpf.command = XDP_SETUP_XSK_POOL;
	167	bpf.xsk.pool = pool;
	168	bpf.xsk.queue_id = queue_id;
	169
c2d3d6a4	170	err = netdev->netdev_ops->ndo_bpf(netdev, &bpf);
1c1efc2a MK	171	if (err)
	172	goto err_unreg_pool;
	173
921b6869 MK	174	if (!pool->dma_pages) {
921b6869 MK	175	WARN(1, "Driver did not DMA map zero-copy buffers");
12c8a8ca	176	err = -EINVAL;
921b6869 MK	177	goto err_unreg_xsk;
921b6869 MK	178	}
c2d3d6a4	179	pool->umem->zc = true;
1c1efc2a MK	180	return 0;
1c1efc2a MK	181
921b6869 MK	182	err_unreg_xsk:
921b6869 MK	183	xp_disable_drv_zc(pool);
1c1efc2a MK	184	err_unreg_pool:
	185	if (!force_zc)
	186	err = 0; /* fallback to copy mode */
17864891	187	if (err) {
c2d3d6a4	188	xsk_clear_pool_at_qid(netdev, queue_id);
17864891 MM	189	dev_put(netdev);
17864891 MM	190	}
1c1efc2a MK	191	return err;
	192	}
	193
b5aea28d MK	194	int xp_assign_dev(struct xsk_buff_pool pool, struct net_device dev,
	195	u16 queue_id, u16 flags)
	196	{
	197	return __xp_assign_dev(pool, dev, queue_id, flags);
	198	}
	199
	200	int xp_assign_dev_shared(struct xsk_buff_pool pool, struct xdp_umem umem,
	201	struct net_device *dev, u16 queue_id)
	202	{
	203	u16 flags;
	204
	205	/* One fill and completion ring required for each queue id. */
	206	if (!pool->fq \|\| !pool->cq)
	207	return -EINVAL;
	208
	209	flags = umem->zc ? XDP_ZEROCOPY : XDP_COPY;
	210	if (pool->uses_need_wakeup)
	211	flags \|= XDP_USE_NEED_WAKEUP;
	212
	213	return __xp_assign_dev(pool, dev, queue_id, flags);
	214	}
	215
1c1efc2a MK	216	void xp_clear_dev(struct xsk_buff_pool *pool)
1c1efc2a MK	217	{
c2d3d6a4	218	if (!pool->netdev)
1c1efc2a MK	219	return;
1c1efc2a MK	220
921b6869	221	xp_disable_drv_zc(pool);
c2d3d6a4 MK	222	xsk_clear_pool_at_qid(pool->netdev, pool->queue_id);
	223	dev_put(pool->netdev);
	224	pool->netdev = NULL;
1c1efc2a MK	225	}
	226
	227	static void xp_release_deferred(struct work_struct *work)
	228	{
	229	struct xsk_buff_pool *pool = container_of(work, struct xsk_buff_pool,
	230	work);
	231
	232	rtnl_lock();
	233	xp_clear_dev(pool);
	234	rtnl_unlock();
	235
7361f9c3 MK	236	if (pool->fq) {
	237	xskq_destroy(pool->fq);
	238	pool->fq = NULL;
	239	}
	240
	241	if (pool->cq) {
	242	xskq_destroy(pool->cq);
	243	pool->cq = NULL;
	244	}
	245
537cf4e3	246	xdp_put_umem(pool->umem, false);
1c1efc2a MK	247	xp_destroy(pool);
	248	}
	249
	250	void xp_get_pool(struct xsk_buff_pool *pool)
	251	{
	252	refcount_inc(&pool->users);
	253	}
	254
e5e1a4bc	255	bool xp_put_pool(struct xsk_buff_pool *pool)
1c1efc2a MK	256	{
1c1efc2a MK	257	if (!pool)
e5e1a4bc	258	return false;
1c1efc2a MK	259
	260	if (refcount_dec_and_test(&pool->users)) {
	261	INIT_WORK(&pool->work, xp_release_deferred);
	262	schedule_work(&pool->work);
e5e1a4bc	263	return true;
1c1efc2a	264	}
e5e1a4bc MK	265
e5e1a4bc MK	266	return false;
1c1efc2a MK	267	}
1c1efc2a MK	268
921b6869 MK	269	static struct xsk_dma_map xp_find_dma_map(struct xsk_buff_pool pool)
	270	{
	271	struct xsk_dma_map *dma_map;
	272
	273	list_for_each_entry(dma_map, &pool->umem->xsk_dma_list, list) {
	274	if (dma_map->netdev == pool->netdev)
	275	return dma_map;
	276	}
	277
	278	return NULL;
	279	}
	280
	281	static struct xsk_dma_map xp_create_dma_map(struct device dev, struct net_device *netdev,
	282	u32 nr_pages, struct xdp_umem *umem)
	283	{
	284	struct xsk_dma_map *dma_map;
	285
	286	dma_map = kzalloc(sizeof(*dma_map), GFP_KERNEL);
	287	if (!dma_map)
	288	return NULL;
	289
	290	dma_map->dma_pages = kvcalloc(nr_pages, sizeof(*dma_map->dma_pages), GFP_KERNEL);
1d6fd78a	291	if (!dma_map->dma_pages) {
921b6869 MK	292	kfree(dma_map);
	293	return NULL;
	294	}
	295
	296	dma_map->netdev = netdev;
	297	dma_map->dev = dev;
	298	dma_map->dma_need_sync = false;
	299	dma_map->dma_pages_cnt = nr_pages;
bf74a370	300	refcount_set(&dma_map->users, 1);
921b6869 MK	301	list_add(&dma_map->list, &umem->xsk_dma_list);
	302	return dma_map;
	303	}
	304
	305	static void xp_destroy_dma_map(struct xsk_dma_map *dma_map)
	306	{
	307	list_del(&dma_map->list);
	308	kvfree(dma_map->dma_pages);
	309	kfree(dma_map);
	310	}
	311
	312	static void __xp_dma_unmap(struct xsk_dma_map *dma_map, unsigned long attrs)
2b43470a BT	313	{
	314	dma_addr_t *dma;
	315	u32 i;
	316
921b6869 MK	317	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
921b6869 MK	318	dma = &dma_map->dma_pages[i];
2b43470a	319	if (*dma) {
921b6869	320	dma_unmap_page_attrs(dma_map->dev, *dma, PAGE_SIZE,
2b43470a BT	321	DMA_BIDIRECTIONAL, attrs);
	322	*dma = 0;
	323	}
	324	}
	325
921b6869 MK	326	xp_destroy_dma_map(dma_map);
	327	}
	328
	329	void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs)
	330	{
	331	struct xsk_dma_map *dma_map;
	332
	333	if (pool->dma_pages_cnt == 0)
	334	return;
	335
	336	dma_map = xp_find_dma_map(pool);
	337	if (!dma_map) {
	338	WARN(1, "Could not find dma_map for device");
	339	return;
	340	}
	341
	342	if (!refcount_dec_and_test(&dma_map->users))
	343	return;
	344
	345	__xp_dma_unmap(dma_map, attrs);
2b43470a BT	346	kvfree(pool->dma_pages);
	347	pool->dma_pages_cnt = 0;
	348	pool->dev = NULL;
	349	}
	350	EXPORT_SYMBOL(xp_dma_unmap);
	351
921b6869	352	static void xp_check_dma_contiguity(struct xsk_dma_map *dma_map)
2b43470a BT	353	{
	354	u32 i;
	355
921b6869 MK	356	for (i = 0; i < dma_map->dma_pages_cnt - 1; i++) {
	357	if (dma_map->dma_pages[i] + PAGE_SIZE == dma_map->dma_pages[i + 1])
	358	dma_map->dma_pages[i] \|= XSK_NEXT_PG_CONTIG_MASK;
2b43470a	359	else
921b6869	360	dma_map->dma_pages[i] &= ~XSK_NEXT_PG_CONTIG_MASK;
2b43470a BT	361	}
	362	}
	363
921b6869 MK	364	static int xp_init_dma_info(struct xsk_buff_pool pool, struct xsk_dma_map dma_map)
	365	{
	366	pool->dma_pages = kvcalloc(dma_map->dma_pages_cnt, sizeof(*pool->dma_pages), GFP_KERNEL);
	367	if (!pool->dma_pages)
	368	return -ENOMEM;
	369
	370	pool->dev = dma_map->dev;
	371	pool->dma_pages_cnt = dma_map->dma_pages_cnt;
	372	pool->dma_need_sync = dma_map->dma_need_sync;
921b6869 MK	373	memcpy(pool->dma_pages, dma_map->dma_pages,
	374	pool->dma_pages_cnt * sizeof(*pool->dma_pages));
	375
	376	return 0;
	377	}
	378
2b43470a BT	379	int xp_dma_map(struct xsk_buff_pool pool, struct device dev,
	380	unsigned long attrs, struct page **pages, u32 nr_pages)
	381	{
921b6869	382	struct xsk_dma_map *dma_map;
2b43470a	383	dma_addr_t dma;
921b6869	384	int err;
2b43470a BT	385	u32 i;
2b43470a BT	386
921b6869 MK	387	dma_map = xp_find_dma_map(pool);
	388	if (dma_map) {
	389	err = xp_init_dma_info(pool, dma_map);
	390	if (err)
	391	return err;
2b43470a	392
bf74a370	393	refcount_inc(&dma_map->users);
921b6869 MK	394	return 0;
921b6869 MK	395	}
2b43470a	396
921b6869 MK	397	dma_map = xp_create_dma_map(dev, pool->netdev, nr_pages, pool->umem);
	398	if (!dma_map)
	399	return -ENOMEM;
	400
	401	for (i = 0; i < dma_map->dma_pages_cnt; i++) {
2b43470a BT	402	dma = dma_map_page_attrs(dev, pages[i], 0, PAGE_SIZE,
	403	DMA_BIDIRECTIONAL, attrs);
	404	if (dma_mapping_error(dev, dma)) {
921b6869	405	__xp_dma_unmap(dma_map, attrs);
2b43470a BT	406	return -ENOMEM;
2b43470a BT	407	}
7e024575	408	if (dma_need_sync(dev, dma))
921b6869 MK	409	dma_map->dma_need_sync = true;
921b6869 MK	410	dma_map->dma_pages[i] = dma;
2b43470a BT	411	}
	412
	413	if (pool->unaligned)
921b6869 MK	414	xp_check_dma_contiguity(dma_map);
	415
	416	err = xp_init_dma_info(pool, dma_map);
	417	if (err) {
	418	__xp_dma_unmap(dma_map, attrs);
	419	return err;
	420	}
	421
2b43470a BT	422	return 0;
	423	}
	424	EXPORT_SYMBOL(xp_dma_map);
	425
2b43470a BT	426	static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
	427	u64 addr)
	428	{
	429	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
	430	}
	431
2b43470a BT	432	static bool xp_check_unaligned(struct xsk_buff_pool pool, u64 addr)
	433	{
	434	addr = xp_unaligned_extract_addr(addr);
	435	if (*addr >= pool->addrs_cnt \|\|
	436	*addr + pool->chunk_size > pool->addrs_cnt \|\|
	437	xp_addr_crosses_non_contig_pg(pool, *addr))
	438	return false;
	439	return true;
	440	}
	441
	442	static bool xp_check_aligned(struct xsk_buff_pool pool, u64 addr)
	443	{
	444	addr = xp_aligned_extract_addr(pool, addr);
	445	return *addr < pool->addrs_cnt;
	446	}
	447
	448	static struct xdp_buff_xsk __xp_alloc(struct xsk_buff_pool pool)
	449	{
	450	struct xdp_buff_xsk *xskb;
	451	u64 addr;
	452	bool ok;
	453
	454	if (pool->free_heads_cnt == 0)
	455	return NULL;
	456
	457	xskb = pool->free_heads[--pool->free_heads_cnt];
	458
	459	for (;;) {
	460	if (!xskq_cons_peek_addr_unchecked(pool->fq, &addr)) {
8aa5a335	461	pool->fq->queue_empty_descs++;
2b43470a BT	462	xp_release(xskb);
	463	return NULL;
	464	}
	465
	466	ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
	467	xp_check_aligned(pool, &addr);
	468	if (!ok) {
	469	pool->fq->invalid_descs++;
	470	xskq_cons_release(pool->fq);
	471	continue;
	472	}
	473	break;
	474	}
	475	xskq_cons_release(pool->fq);
	476
	477	xskb->orig_addr = addr;
	478	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
	479	if (pool->dma_pages_cnt) {
	480	xskb->frame_dma = (pool->dma_pages[addr >> PAGE_SHIFT] &
	481	~XSK_NEXT_PG_CONTIG_MASK) +
	482	(addr & ~PAGE_MASK);
	483	xskb->dma = xskb->frame_dma + pool->headroom +
	484	XDP_PACKET_HEADROOM;
	485	}
	486	return xskb;
	487	}
	488
	489	struct xdp_buff xp_alloc(struct xsk_buff_pool pool)
	490	{
	491	struct xdp_buff_xsk *xskb;
	492
	493	if (!pool->free_list_cnt) {
	494	xskb = __xp_alloc(pool);
	495	if (!xskb)
	496	return NULL;
	497	} else {
	498	pool->free_list_cnt--;
	499	xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk,
	500	free_list_node);
	501	list_del(&xskb->free_list_node);
	502	}
	503
	504	xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM;
	505	xskb->xdp.data_meta = xskb->xdp.data;
	506
91d5b702	507	if (pool->dma_need_sync) {
2b43470a BT	508	dma_sync_single_range_for_device(pool->dev, xskb->dma, 0,
	509	pool->frame_len,
	510	DMA_BIDIRECTIONAL);
	511	}
	512	return &xskb->xdp;
	513	}
	514	EXPORT_SYMBOL(xp_alloc);
	515
	516	bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count)
	517	{
	518	if (pool->free_list_cnt >= count)
	519	return true;
	520	return xskq_cons_has_entries(pool->fq, count - pool->free_list_cnt);
	521	}
	522	EXPORT_SYMBOL(xp_can_alloc);
	523
	524	void xp_free(struct xdp_buff_xsk *xskb)
	525	{
	526	xskb->pool->free_list_cnt++;
	527	list_add(&xskb->free_list_node, &xskb->pool->free_list);
	528	}
	529	EXPORT_SYMBOL(xp_free);
	530
2b43470a BT	531	void xp_raw_get_data(struct xsk_buff_pool pool, u64 addr)
	532	{
	533	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	534	return pool->addrs + addr;
	535	}
	536	EXPORT_SYMBOL(xp_raw_get_data);
	537
	538	dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
	539	{
	540	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	541	return (pool->dma_pages[addr >> PAGE_SHIFT] &
	542	~XSK_NEXT_PG_CONTIG_MASK) +
	543	(addr & ~PAGE_MASK);
	544	}
	545	EXPORT_SYMBOL(xp_raw_get_dma);
	546
26062b18	547	void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
2b43470a	548	{
2b43470a BT	549	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
	550	xskb->pool->frame_len, DMA_BIDIRECTIONAL);
	551	}
26062b18	552	EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);
2b43470a	553
26062b18 BT	554	void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
26062b18 BT	555	size_t size)
2b43470a	556	{
2b43470a BT	557	dma_sync_single_range_for_device(pool->dev, dma, 0,
	558	size, DMA_BIDIRECTIONAL);
	559	}
26062b18	560	EXPORT_SYMBOL(xp_dma_sync_for_device_slow);