[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)
{
	bool unaligned = umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG;
	struct xsk_buff_pool *pool;
	struct xdp_buff_xsk *xskb;
	u32 i, entries;

	entries = unaligned ? umem->chunks : 0;
	pool = kvzalloc(struct_size(pool, free_heads, entries),	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->chunk_shift = ffs(umem->chunk_size) - 1;
	pool->unaligned = unaligned;
	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;
		INIT_LIST_HEAD(&xskb->free_list_node);
		if (pool->unaligned)
			pool->free_heads[i] = xskb;
		else
			xp_init_xskb_addr(xskb, pool, i * pool->chunk_size);
	}

	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");
	}
}

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_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);
	else
		for (i = 0; i < pool->heads_cnt; i++) {
			struct xdp_buff_xsk *xskb = &pool->heads[i];

			xp_init_xskb_dma(xskb, pool, dma_map->dma_pages, xskb->orig_addr);
		}

	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;

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

	if (pool->unaligned) {
		xskb = pool->free_heads[--pool->free_heads_cnt];
		xp_init_xskb_addr(xskb, pool, addr);
		if (pool->dma_pages_cnt)
			xp_init_xskb_dma(xskb, pool, pool->dma_pages, addr);
	} else {
		xskb = &pool->heads[xp_aligned_extract_idx(pool, addr)];
	}

	xskq_cons_release(pool->fq);
	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_init(&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);

static u32 xp_alloc_new_from_fq(struct xsk_buff_pool *pool, struct xdp_buff **xdp, u32 max)
{
	u32 i, cached_cons, nb_entries;

	if (max > pool->free_heads_cnt)
		max = pool->free_heads_cnt;
	max = xskq_cons_nb_entries(pool->fq, max);

	cached_cons = pool->fq->cached_cons;
	nb_entries = max;
	i = max;
	while (i--) {
		struct xdp_buff_xsk *xskb;
		u64 addr;
		bool ok;

		__xskq_cons_read_addr_unchecked(pool->fq, cached_cons++, &addr);

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

		if (pool->unaligned) {
			xskb = pool->free_heads[--pool->free_heads_cnt];
			xp_init_xskb_addr(xskb, pool, addr);
			if (pool->dma_pages_cnt)
				xp_init_xskb_dma(xskb, pool, pool->dma_pages, addr);
		} else {
			xskb = &pool->heads[xp_aligned_extract_idx(pool, addr)];
		}

		*xdp = &xskb->xdp;
		xdp++;
	}

	xskq_cons_release_n(pool->fq, max);
	return nb_entries;
}

static u32 xp_alloc_reused(struct xsk_buff_pool *pool, struct xdp_buff **xdp, u32 nb_entries)
{
	struct xdp_buff_xsk *xskb;
	u32 i;

	nb_entries = min_t(u32, nb_entries, pool->free_list_cnt);

	i = nb_entries;
	while (i--) {
		xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk, free_list_node);
		list_del_init(&xskb->free_list_node);

		*xdp = &xskb->xdp;
		xdp++;
	}
	pool->free_list_cnt -= nb_entries;

	return nb_entries;
}

u32 xp_alloc_batch(struct xsk_buff_pool *pool, struct xdp_buff **xdp, u32 max)
{
	u32 nb_entries1 = 0, nb_entries2;

	if (unlikely(pool->dma_need_sync)) {
		/* Slow path */
		*xdp = xp_alloc(pool);
		return !!*xdp;
	}

	if (unlikely(pool->free_list_cnt)) {
		nb_entries1 = xp_alloc_reused(pool, xdp, max);
		if (nb_entries1 == max)
			return nb_entries1;

		max -= nb_entries1;
		xdp += nb_entries1;
	}

	nb_entries2 = xp_alloc_new_from_fq(pool, xdp, max);
	if (!nb_entries2)
		pool->fq->queue_empty_descs++;

	return nb_entries1 + nb_entries2;
}
EXPORT_SYMBOL(xp_alloc_batch);

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)
{
	if (!list_empty(&xskb->free_list_node))
		return;

	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	46	{
94033cd8	47	bool unaligned = umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG;
2b43470a BT	48	struct xsk_buff_pool *pool;
2b43470a BT	49	struct xdp_buff_xsk *xskb;
94033cd8	50	u32 i, entries;
2b43470a	51
94033cd8 MK	52	entries = unaligned ? umem->chunks : 0;
94033cd8 MK	53	pool = kvzalloc(struct_size(pool, free_heads, entries), GFP_KERNEL);
2b43470a BT	54	if (!pool)
	55	goto out;
	56
1c1efc2a	57	pool->heads = kvcalloc(umem->chunks, sizeof(*pool->heads), GFP_KERNEL);
2b43470a BT	58	if (!pool->heads)
	59	goto out;
	60
1c1efc2a MK	61	pool->chunk_mask = ~((u64)umem->chunk_size - 1);
	62	pool->addrs_cnt = umem->size;
	63	pool->heads_cnt = umem->chunks;
	64	pool->free_heads_cnt = umem->chunks;
	65	pool->headroom = umem->headroom;
	66	pool->chunk_size = umem->chunk_size;
94033cd8 MK	67	pool->chunk_shift = ffs(umem->chunk_size) - 1;
94033cd8 MK	68	pool->unaligned = unaligned;
1c1efc2a MK	69	pool->frame_len = umem->chunk_size - umem->headroom -
1c1efc2a MK	70	XDP_PACKET_HEADROOM;
1742b3d5	71	pool->umem = umem;
7f7ffa4e	72	pool->addrs = umem->addrs;
2b43470a	73	INIT_LIST_HEAD(&pool->free_list);
a5aa8e52 MK	74	INIT_LIST_HEAD(&pool->xsk_tx_list);
a5aa8e52 MK	75	spin_lock_init(&pool->xsk_tx_list_lock);
f09ced40	76	spin_lock_init(&pool->cq_lock);
1c1efc2a	77	refcount_set(&pool->users, 1);
2b43470a	78
7361f9c3 MK	79	pool->fq = xs->fq_tmp;
7361f9c3 MK	80	pool->cq = xs->cq_tmp;
7361f9c3	81
2b43470a BT	82	for (i = 0; i < pool->free_heads_cnt; i++) {
	83	xskb = &pool->heads[i];
	84	xskb->pool = pool;
1c1efc2a	85	xskb->xdp.frame_sz = umem->chunk_size - umem->headroom;
5bec7ca2	86	INIT_LIST_HEAD(&xskb->free_list_node);
94033cd8 MK	87	if (pool->unaligned)
	88	pool->free_heads[i] = xskb;
	89	else
	90	xp_init_xskb_addr(xskb, pool, i * pool->chunk_size);
2b43470a BT	91	}
2b43470a BT	92
7f7ffa4e	93	return pool;
2b43470a BT	94
	95	out:
	96	xp_destroy(pool);
	97	return NULL;
	98	}
	99
2b43470a BT	100	void xp_set_rxq_info(struct xsk_buff_pool pool, struct xdp_rxq_info rxq)
	101	{
	102	u32 i;
	103
	104	for (i = 0; i < pool->heads_cnt; i++)
	105	pool->heads[i].xdp.rxq = rxq;
	106	}
	107	EXPORT_SYMBOL(xp_set_rxq_info);
	108
921b6869 MK	109	static void xp_disable_drv_zc(struct xsk_buff_pool *pool)
	110	{
	111	struct netdev_bpf bpf;
	112	int err;
	113
	114	ASSERT_RTNL();
	115
	116	if (pool->umem->zc) {
	117	bpf.command = XDP_SETUP_XSK_POOL;
	118	bpf.xsk.pool = NULL;
	119	bpf.xsk.queue_id = pool->queue_id;
	120
	121	err = pool->netdev->netdev_ops->ndo_bpf(pool->netdev, &bpf);
	122
	123	if (err)
	124	WARN(1, "Failed to disable zero-copy!\n");
	125	}
	126	}
	127
f0863eab BT	128	int xp_assign_dev(struct xsk_buff_pool *pool,
f0863eab BT	129	struct net_device *netdev, u16 queue_id, u16 flags)
1c1efc2a	130	{
1c1efc2a MK	131	bool force_zc, force_copy;
	132	struct netdev_bpf bpf;
	133	int err = 0;
	134
	135	ASSERT_RTNL();
	136
	137	force_zc = flags & XDP_ZEROCOPY;
	138	force_copy = flags & XDP_COPY;
	139
	140	if (force_zc && force_copy)
	141	return -EINVAL;
	142
c2d3d6a4	143	if (xsk_get_pool_from_qid(netdev, queue_id))
1c1efc2a MK	144	return -EBUSY;
1c1efc2a MK	145
921b6869 MK	146	pool->netdev = netdev;
921b6869 MK	147	pool->queue_id = queue_id;
c2d3d6a4	148	err = xsk_reg_pool_at_qid(netdev, pool, queue_id);
1c1efc2a MK	149	if (err)
	150	return err;
	151
e3920818	152	if (flags & XDP_USE_NEED_WAKEUP)
c2d3d6a4	153	pool->uses_need_wakeup = true;
e3920818 BT	154	/* Tx needs to be explicitly woken up the first time. Also
	155	* for supporting drivers that do not implement this
	156	* feature. They will always have to call sendto() or poll().
	157	*/
	158	pool->cached_need_wakeup = XDP_WAKEUP_TX;
1c1efc2a	159
c2d3d6a4 MK	160	dev_hold(netdev);
c2d3d6a4 MK	161
1c1efc2a MK	162	if (force_copy)
	163	/* For copy-mode, we are done. */
	164	return 0;
	165
c2d3d6a4 MK	166	if (!netdev->netdev_ops->ndo_bpf \|\|
c2d3d6a4 MK	167	!netdev->netdev_ops->ndo_xsk_wakeup) {
1c1efc2a MK	168	err = -EOPNOTSUPP;
	169	goto err_unreg_pool;
	170	}
	171
	172	bpf.command = XDP_SETUP_XSK_POOL;
	173	bpf.xsk.pool = pool;
	174	bpf.xsk.queue_id = queue_id;
	175
c2d3d6a4	176	err = netdev->netdev_ops->ndo_bpf(netdev, &bpf);
1c1efc2a MK	177	if (err)
	178	goto err_unreg_pool;
	179
921b6869 MK	180	if (!pool->dma_pages) {
921b6869 MK	181	WARN(1, "Driver did not DMA map zero-copy buffers");
12c8a8ca	182	err = -EINVAL;
921b6869 MK	183	goto err_unreg_xsk;
921b6869 MK	184	}
c2d3d6a4	185	pool->umem->zc = true;
1c1efc2a MK	186	return 0;
1c1efc2a MK	187
921b6869 MK	188	err_unreg_xsk:
921b6869 MK	189	xp_disable_drv_zc(pool);
1c1efc2a MK	190	err_unreg_pool:
	191	if (!force_zc)
	192	err = 0; /* fallback to copy mode */
17864891	193	if (err) {
c2d3d6a4	194	xsk_clear_pool_at_qid(netdev, queue_id);
17864891 MM	195	dev_put(netdev);
17864891 MM	196	}
1c1efc2a MK	197	return err;
	198	}
	199
b5aea28d MK	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
f0863eab	213	return xp_assign_dev(pool, dev, queue_id, flags);
b5aea28d MK	214	}
b5aea28d MK	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	414	xp_check_dma_contiguity(dma_map);
94033cd8 MK	415	else
	416	for (i = 0; i < pool->heads_cnt; i++) {
	417	struct xdp_buff_xsk *xskb = &pool->heads[i];
	418
	419	xp_init_xskb_dma(xskb, pool, dma_map->dma_pages, xskb->orig_addr);
	420	}
921b6869 MK	421
	422	err = xp_init_dma_info(pool, dma_map);
	423	if (err) {
	424	__xp_dma_unmap(dma_map, attrs);
	425	return err;
	426	}
	427
2b43470a BT	428	return 0;
	429	}
	430	EXPORT_SYMBOL(xp_dma_map);
	431
2b43470a BT	432	static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
	433	u64 addr)
	434	{
	435	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
	436	}
	437
2b43470a BT	438	static bool xp_check_unaligned(struct xsk_buff_pool pool, u64 addr)
	439	{
	440	addr = xp_unaligned_extract_addr(addr);
	441	if (*addr >= pool->addrs_cnt \|\|
	442	*addr + pool->chunk_size > pool->addrs_cnt \|\|
	443	xp_addr_crosses_non_contig_pg(pool, *addr))
	444	return false;
	445	return true;
	446	}
	447
	448	static bool xp_check_aligned(struct xsk_buff_pool pool, u64 addr)
	449	{
	450	addr = xp_aligned_extract_addr(pool, addr);
	451	return *addr < pool->addrs_cnt;
	452	}
	453
	454	static struct xdp_buff_xsk __xp_alloc(struct xsk_buff_pool pool)
	455	{
	456	struct xdp_buff_xsk *xskb;
	457	u64 addr;
	458	bool ok;
	459
	460	if (pool->free_heads_cnt == 0)
	461	return NULL;
	462
2b43470a BT	463	for (;;) {
2b43470a BT	464	if (!xskq_cons_peek_addr_unchecked(pool->fq, &addr)) {
8aa5a335	465	pool->fq->queue_empty_descs++;
2b43470a BT	466	return NULL;
	467	}
	468
	469	ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
	470	xp_check_aligned(pool, &addr);
	471	if (!ok) {
	472	pool->fq->invalid_descs++;
	473	xskq_cons_release(pool->fq);
	474	continue;
	475	}
	476	break;
	477	}
2b43470a	478
94033cd8 MK	479	if (pool->unaligned) {
	480	xskb = pool->free_heads[--pool->free_heads_cnt];
	481	xp_init_xskb_addr(xskb, pool, addr);
	482	if (pool->dma_pages_cnt)
	483	xp_init_xskb_dma(xskb, pool, pool->dma_pages, addr);
	484	} else {
	485	xskb = &pool->heads[xp_aligned_extract_idx(pool, addr)];
2b43470a	486	}
94033cd8 MK	487
94033cd8 MK	488	xskq_cons_release(pool->fq);
2b43470a BT	489	return xskb;
	490	}
	491
	492	struct xdp_buff xp_alloc(struct xsk_buff_pool pool)
	493	{
	494	struct xdp_buff_xsk *xskb;
	495
	496	if (!pool->free_list_cnt) {
	497	xskb = __xp_alloc(pool);
	498	if (!xskb)
	499	return NULL;
	500	} else {
	501	pool->free_list_cnt--;
	502	xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk,
	503	free_list_node);
199d983b	504	list_del_init(&xskb->free_list_node);
2b43470a BT	505	}
	506
	507	xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM;
	508	xskb->xdp.data_meta = xskb->xdp.data;
	509
91d5b702	510	if (pool->dma_need_sync) {
2b43470a BT	511	dma_sync_single_range_for_device(pool->dev, xskb->dma, 0,
	512	pool->frame_len,
	513	DMA_BIDIRECTIONAL);
	514	}
	515	return &xskb->xdp;
	516	}
	517	EXPORT_SYMBOL(xp_alloc);
	518
47e4075d MK	519	static u32 xp_alloc_new_from_fq(struct xsk_buff_pool pool, struct xdp_buff *xdp, u32 max)
	520	{
	521	u32 i, cached_cons, nb_entries;
	522
	523	if (max > pool->free_heads_cnt)
	524	max = pool->free_heads_cnt;
	525	max = xskq_cons_nb_entries(pool->fq, max);
	526
	527	cached_cons = pool->fq->cached_cons;
	528	nb_entries = max;
	529	i = max;
	530	while (i--) {
	531	struct xdp_buff_xsk *xskb;
	532	u64 addr;
	533	bool ok;
	534
	535	__xskq_cons_read_addr_unchecked(pool->fq, cached_cons++, &addr);
	536
	537	ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
	538	xp_check_aligned(pool, &addr);
	539	if (unlikely(!ok)) {
	540	pool->fq->invalid_descs++;
	541	nb_entries--;
	542	continue;
	543	}
	544
94033cd8 MK	545	if (pool->unaligned) {
	546	xskb = pool->free_heads[--pool->free_heads_cnt];
	547	xp_init_xskb_addr(xskb, pool, addr);
	548	if (pool->dma_pages_cnt)
	549	xp_init_xskb_dma(xskb, pool, pool->dma_pages, addr);
	550	} else {
	551	xskb = &pool->heads[xp_aligned_extract_idx(pool, addr)];
	552	}
	553
47e4075d	554	*xdp = &xskb->xdp;
47e4075d MK	555	xdp++;
	556	}
	557
	558	xskq_cons_release_n(pool->fq, max);
	559	return nb_entries;
	560	}
	561
	562	static u32 xp_alloc_reused(struct xsk_buff_pool pool, struct xdp_buff *xdp, u32 nb_entries)
	563	{
	564	struct xdp_buff_xsk *xskb;
	565	u32 i;
	566
	567	nb_entries = min_t(u32, nb_entries, pool->free_list_cnt);
	568
	569	i = nb_entries;
	570	while (i--) {
	571	xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk, free_list_node);
199d983b	572	list_del_init(&xskb->free_list_node);
47e4075d MK	573
	574	*xdp = &xskb->xdp;
	575	xdp++;
	576	}
	577	pool->free_list_cnt -= nb_entries;
	578
	579	return nb_entries;
	580	}
	581
	582	u32 xp_alloc_batch(struct xsk_buff_pool pool, struct xdp_buff *xdp, u32 max)
	583	{
	584	u32 nb_entries1 = 0, nb_entries2;
	585
	586	if (unlikely(pool->dma_need_sync)) {
	587	/* Slow path */
	588	*xdp = xp_alloc(pool);
	589	return !!*xdp;
	590	}
	591
	592	if (unlikely(pool->free_list_cnt)) {
	593	nb_entries1 = xp_alloc_reused(pool, xdp, max);
	594	if (nb_entries1 == max)
	595	return nb_entries1;
	596
	597	max -= nb_entries1;
	598	xdp += nb_entries1;
	599	}
	600
	601	nb_entries2 = xp_alloc_new_from_fq(pool, xdp, max);
	602	if (!nb_entries2)
	603	pool->fq->queue_empty_descs++;
	604
	605	return nb_entries1 + nb_entries2;
	606	}
	607	EXPORT_SYMBOL(xp_alloc_batch);
	608
2b43470a BT	609	bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count)
	610	{
	611	if (pool->free_list_cnt >= count)
	612	return true;
	613	return xskq_cons_has_entries(pool->fq, count - pool->free_list_cnt);
	614	}
	615	EXPORT_SYMBOL(xp_can_alloc);
	616
	617	void xp_free(struct xdp_buff_xsk *xskb)
	618	{
199d983b MK	619	if (!list_empty(&xskb->free_list_node))
	620	return;
	621
2b43470a BT	622	xskb->pool->free_list_cnt++;
	623	list_add(&xskb->free_list_node, &xskb->pool->free_list);
	624	}
	625	EXPORT_SYMBOL(xp_free);
	626
2b43470a BT	627	void xp_raw_get_data(struct xsk_buff_pool pool, u64 addr)
	628	{
	629	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	630	return pool->addrs + addr;
	631	}
	632	EXPORT_SYMBOL(xp_raw_get_data);
	633
	634	dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
	635	{
	636	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
	637	return (pool->dma_pages[addr >> PAGE_SHIFT] &
	638	~XSK_NEXT_PG_CONTIG_MASK) +
	639	(addr & ~PAGE_MASK);
	640	}
	641	EXPORT_SYMBOL(xp_raw_get_dma);
	642
26062b18	643	void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
2b43470a	644	{
2b43470a BT	645	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
	646	xskb->pool->frame_len, DMA_BIDIRECTIONAL);
	647	}
26062b18	648	EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);
2b43470a	649
26062b18 BT	650	void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
26062b18 BT	651	size_t size)
2b43470a	652	{
2b43470a BT	653	dma_sync_single_range_for_device(pool->dev, dma, 0,
	654	size, DMA_BIDIRECTIONAL);
	655	}
26062b18	656	EXPORT_SYMBOL(xp_dma_sync_for_device_slow);