[linux-block.git] / include / net / xsk_buff_pool.h

/* SPDX-License-Identifier: GPL-2.0 */
/* Copyright(c) 2020 Intel Corporation. */

#ifndef XSK_BUFF_POOL_H_
#define XSK_BUFF_POOL_H_

#include <linux/if_xdp.h>
#include <linux/types.h>
#include <linux/dma-mapping.h>
#include <linux/bpf.h>
#include <net/xdp.h>

struct xsk_buff_pool;
struct xdp_rxq_info;
struct xsk_cb_desc;
struct xsk_queue;
struct xdp_desc;
struct xdp_umem;
struct xdp_sock;
struct device;
struct page;

#define XSK_PRIV_MAX 24

struct xdp_buff_xsk {
	struct xdp_buff xdp;
	u8 cb[XSK_PRIV_MAX];
	dma_addr_t dma;
	dma_addr_t frame_dma;
	struct xsk_buff_pool *pool;
	u64 orig_addr;
	struct list_head free_list_node;
	struct list_head xskb_list_node;
};

#define XSK_CHECK_PRIV_TYPE(t) BUILD_BUG_ON(sizeof(t) > offsetofend(struct xdp_buff_xsk, cb))
#define XSK_TX_COMPL_FITS(t) BUILD_BUG_ON(sizeof(struct xsk_tx_metadata_compl) > sizeof(t))

struct xsk_dma_map {
	dma_addr_t *dma_pages;
	struct device *dev;
	struct net_device *netdev;
	refcount_t users;
	struct list_head list; /* Protected by the RTNL_LOCK */
	u32 dma_pages_cnt;
	bool dma_need_sync;
};

struct xsk_buff_pool {
	/* Members only used in the control path first. */
	struct device *dev;
	struct net_device *netdev;
	struct list_head xsk_tx_list;
	/* Protects modifications to the xsk_tx_list */
	spinlock_t xsk_tx_list_lock;
	refcount_t users;
	struct xdp_umem *umem;
	struct work_struct work;
	struct list_head free_list;
	struct list_head xskb_list;
	u32 heads_cnt;
	u16 queue_id;

	/* Data path members as close to free_heads at the end as possible. */
	struct xsk_queue *fq ____cacheline_aligned_in_smp;
	struct xsk_queue *cq;
	/* For performance reasons, each buff pool has its own array of dma_pages
	 * even when they are identical.
	 */
	dma_addr_t *dma_pages;
	struct xdp_buff_xsk *heads;
	struct xdp_desc *tx_descs;
	u64 chunk_mask;
	u64 addrs_cnt;
	u32 free_list_cnt;
	u32 dma_pages_cnt;
	u32 free_heads_cnt;
	u32 headroom;
	u32 chunk_size;
	u32 chunk_shift;
	u32 frame_len;
	u8 tx_metadata_len; /* inherited from umem */
	u8 cached_need_wakeup;
	bool uses_need_wakeup;
	bool dma_need_sync;
	bool unaligned;
	bool tx_sw_csum;
	void *addrs;
	/* Mutual exclusion of the completion ring in the SKB mode. Two cases to protect:
	 * NAPI TX thread and sendmsg error paths in the SKB destructor callback and when
	 * sockets share a single cq when the same netdev and queue id is shared.
	 */
	spinlock_t cq_lock;
	struct xdp_buff_xsk *free_heads[];
};

/* Masks for xdp_umem_page flags.
 * The low 12-bits of the addr will be 0 since this is the page address, so we
 * can use them for flags.
 */
#define XSK_NEXT_PG_CONTIG_SHIFT 0
#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)

/* AF_XDP core. */
struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
						struct xdp_umem *umem);
int xp_assign_dev(struct xsk_buff_pool *pool, struct net_device *dev,
		  u16 queue_id, u16 flags);
int xp_assign_dev_shared(struct xsk_buff_pool *pool, struct xdp_sock *umem_xs,
			 struct net_device *dev, u16 queue_id);
int xp_alloc_tx_descs(struct xsk_buff_pool *pool, struct xdp_sock *xs);
void xp_destroy(struct xsk_buff_pool *pool);
void xp_get_pool(struct xsk_buff_pool *pool);
bool xp_put_pool(struct xsk_buff_pool *pool);
void xp_clear_dev(struct xsk_buff_pool *pool);
void xp_add_xsk(struct xsk_buff_pool *pool, struct xdp_sock *xs);
void xp_del_xsk(struct xsk_buff_pool *pool, struct xdp_sock *xs);

/* AF_XDP, and XDP core. */
void xp_free(struct xdp_buff_xsk *xskb);

static inline void xp_init_xskb_addr(struct xdp_buff_xsk *xskb, struct xsk_buff_pool *pool,
				     u64 addr)
{
	xskb->orig_addr = addr;
	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
}

static inline void xp_init_xskb_dma(struct xdp_buff_xsk *xskb, struct xsk_buff_pool *pool,
				    dma_addr_t *dma_pages, u64 addr)
{
	xskb->frame_dma = (dma_pages[addr >> PAGE_SHIFT] & ~XSK_NEXT_PG_CONTIG_MASK) +
		(addr & ~PAGE_MASK);
	xskb->dma = xskb->frame_dma + pool->headroom + XDP_PACKET_HEADROOM;
}

/* AF_XDP ZC drivers, via xdp_sock_buff.h */
void xp_set_rxq_info(struct xsk_buff_pool *pool, struct xdp_rxq_info *rxq);
void xp_fill_cb(struct xsk_buff_pool *pool, struct xsk_cb_desc *desc);
int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
	       unsigned long attrs, struct page **pages, u32 nr_pages);
void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs);
struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool);
u32 xp_alloc_batch(struct xsk_buff_pool *pool, struct xdp_buff **xdp, u32 max);
bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count);
void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr);
dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr);
static inline dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb)
{
	return xskb->dma;
}

static inline dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb)
{
	return xskb->frame_dma;
}

void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb);
static inline void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb)
{
	xp_dma_sync_for_cpu_slow(xskb);
}

void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
				 size_t size);
static inline void xp_dma_sync_for_device(struct xsk_buff_pool *pool,
					  dma_addr_t dma, size_t size)
{
	if (!pool->dma_need_sync)
		return;

	xp_dma_sync_for_device_slow(pool, dma, size);
}

/* Masks for xdp_umem_page flags.
 * The low 12-bits of the addr will be 0 since this is the page address, so we
 * can use them for flags.
 */
#define XSK_NEXT_PG_CONTIG_SHIFT 0
#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)

static inline bool xp_desc_crosses_non_contig_pg(struct xsk_buff_pool *pool,
						 u64 addr, u32 len)
{
	bool cross_pg = (addr & (PAGE_SIZE - 1)) + len > PAGE_SIZE;

	if (likely(!cross_pg))
		return false;

	return pool->dma_pages &&
	       !(pool->dma_pages[addr >> PAGE_SHIFT] & XSK_NEXT_PG_CONTIG_MASK);
}

static inline bool xp_mb_desc(struct xdp_desc *desc)
{
	return desc->options & XDP_PKT_CONTD;
}

static inline u64 xp_aligned_extract_addr(struct xsk_buff_pool *pool, u64 addr)
{
	return addr & pool->chunk_mask;
}

static inline u64 xp_unaligned_extract_addr(u64 addr)
{
	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
}

static inline u64 xp_unaligned_extract_offset(u64 addr)
{
	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
}

static inline u64 xp_unaligned_add_offset_to_addr(u64 addr)
{
	return xp_unaligned_extract_addr(addr) +
		xp_unaligned_extract_offset(addr);
}

static inline u32 xp_aligned_extract_idx(struct xsk_buff_pool *pool, u64 addr)
{
	return xp_aligned_extract_addr(pool, addr) >> pool->chunk_shift;
}

static inline void xp_release(struct xdp_buff_xsk *xskb)
{
	if (xskb->pool->unaligned)
		xskb->pool->free_heads[xskb->pool->free_heads_cnt++] = xskb;
}

static inline u64 xp_get_handle(struct xdp_buff_xsk *xskb)
{
	u64 offset = xskb->xdp.data - xskb->xdp.data_hard_start;

	offset += xskb->pool->headroom;
	if (!xskb->pool->unaligned)
		return xskb->orig_addr + offset;
	return xskb->orig_addr + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
}

static inline bool xp_tx_metadata_enabled(const struct xsk_buff_pool *pool)
{
	return pool->tx_metadata_len > 0;
}

#endif /* XSK_BUFF_POOL_H_ */
Commit	Line	Data
2b43470a BT	1	/* SPDX-License-Identifier: GPL-2.0 */
	2	/* Copyright(c) 2020 Intel Corporation. */
	3
	4	#ifndef XSK_BUFF_POOL_H_
	5	#define XSK_BUFF_POOL_H_
	6
26062b18	7	#include <linux/if_xdp.h>
2b43470a BT	8	#include <linux/types.h>
2b43470a BT	9	#include <linux/dma-mapping.h>
94033cd8	10	#include <linux/bpf.h>
2b43470a BT	11	#include <net/xdp.h>
	12
	13	struct xsk_buff_pool;
	14	struct xdp_rxq_info;
b4e352ff	15	struct xsk_cb_desc;
2b43470a BT	16	struct xsk_queue;
2b43470a BT	17	struct xdp_desc;
1742b3d5	18	struct xdp_umem;
1c1efc2a	19	struct xdp_sock;
2b43470a BT	20	struct device;
	21	struct page;
	22
94ecc5ca THJ	23	#define XSK_PRIV_MAX 24
94ecc5ca THJ	24
2b43470a BT	25	struct xdp_buff_xsk {
2b43470a BT	26	struct xdp_buff xdp;
94ecc5ca	27	u8 cb[XSK_PRIV_MAX];
2b43470a BT	28	dma_addr_t dma;
	29	dma_addr_t frame_dma;
	30	struct xsk_buff_pool *pool;
2b43470a BT	31	u64 orig_addr;
2b43470a BT	32	struct list_head free_list_node;
24ea5012	33	struct list_head xskb_list_node;
2b43470a BT	34	};
2b43470a BT	35
94ecc5ca	36	#define XSK_CHECK_PRIV_TYPE(t) BUILD_BUG_ON(sizeof(t) > offsetofend(struct xdp_buff_xsk, cb))
48eb03dd	37	#define XSK_TX_COMPL_FITS(t) BUILD_BUG_ON(sizeof(struct xsk_tx_metadata_compl) > sizeof(t))
94ecc5ca	38
921b6869 MK	39	struct xsk_dma_map {
	40	dma_addr_t *dma_pages;
	41	struct device *dev;
	42	struct net_device *netdev;
	43	refcount_t users;
	44	struct list_head list; /* Protected by the RTNL_LOCK */
	45	u32 dma_pages_cnt;
	46	bool dma_need_sync;
	47	};
	48
26062b18	49	struct xsk_buff_pool {
8ef4e27e MK	50	/* Members only used in the control path first. */
	51	struct device *dev;
	52	struct net_device *netdev;
	53	struct list_head xsk_tx_list;
	54	/* Protects modifications to the xsk_tx_list */
	55	spinlock_t xsk_tx_list_lock;
	56	refcount_t users;
	57	struct xdp_umem *umem;
	58	struct work_struct work;
26062b18	59	struct list_head free_list;
24ea5012	60	struct list_head xskb_list;
8ef4e27e MK	61	u32 heads_cnt;
	62	u16 queue_id;
	63
	64	/* Data path members as close to free_heads at the end as possible. */
	65	struct xsk_queue *fq ____cacheline_aligned_in_smp;
	66	struct xsk_queue *cq;
921b6869 MK	67	/* For performance reasons, each buff pool has its own array of dma_pages
	68	* even when they are identical.
	69	*/
26062b18 BT	70	dma_addr_t *dma_pages;
26062b18 BT	71	struct xdp_buff_xsk *heads;
d1bc532e	72	struct xdp_desc *tx_descs;
26062b18 BT	73	u64 chunk_mask;
	74	u64 addrs_cnt;
	75	u32 free_list_cnt;
	76	u32 dma_pages_cnt;
26062b18 BT	77	u32 free_heads_cnt;
	78	u32 headroom;
	79	u32 chunk_size;
94033cd8	80	u32 chunk_shift;
26062b18	81	u32 frame_len;
341ac980	82	u8 tx_metadata_len; /* inherited from umem */
c2d3d6a4 MK	83	u8 cached_need_wakeup;
c2d3d6a4 MK	84	bool uses_need_wakeup;
91d5b702	85	bool dma_need_sync;
26062b18	86	bool unaligned;
11614723	87	bool tx_sw_csum;
26062b18	88	void *addrs;
f09ced40 MK	89	/* Mutual exclusion of the completion ring in the SKB mode. Two cases to protect:
	90	* NAPI TX thread and sendmsg error paths in the SKB destructor callback and when
	91	* sockets share a single cq when the same netdev and queue id is shared.
	92	*/
	93	spinlock_t cq_lock;
26062b18 BT	94	struct xdp_buff_xsk *free_heads[];
	95	};
	96
94033cd8 MK	97	/* Masks for xdp_umem_page flags.
	98	* The low 12-bits of the addr will be 0 since this is the page address, so we
	99	* can use them for flags.
	100	*/
	101	#define XSK_NEXT_PG_CONTIG_SHIFT 0
	102	#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
	103
2b43470a	104	/* AF_XDP core. */
1c1efc2a MK	105	struct xsk_buff_pool xp_create_and_assign_umem(struct xdp_sock xs,
	106	struct xdp_umem *umem);
	107	int xp_assign_dev(struct xsk_buff_pool pool, struct net_device dev,
	108	u16 queue_id, u16 flags);
60240bc2	109	int xp_assign_dev_shared(struct xsk_buff_pool pool, struct xdp_sock umem_xs,
b5aea28d	110	struct net_device *dev, u16 queue_id);
ba3beec2	111	int xp_alloc_tx_descs(struct xsk_buff_pool pool, struct xdp_sock xs);
2b43470a	112	void xp_destroy(struct xsk_buff_pool *pool);
1c1efc2a	113	void xp_get_pool(struct xsk_buff_pool *pool);
e5e1a4bc	114	bool xp_put_pool(struct xsk_buff_pool *pool);
1c1efc2a	115	void xp_clear_dev(struct xsk_buff_pool *pool);
a5aa8e52 MK	116	void xp_add_xsk(struct xsk_buff_pool pool, struct xdp_sock xs);
a5aa8e52 MK	117	void xp_del_xsk(struct xsk_buff_pool pool, struct xdp_sock xs);
2b43470a BT	118
	119	/* AF_XDP, and XDP core. */
	120	void xp_free(struct xdp_buff_xsk *xskb);
	121
94033cd8 MK	122	static inline void xp_init_xskb_addr(struct xdp_buff_xsk xskb, struct xsk_buff_pool pool,
	123	u64 addr)
	124	{
	125	xskb->orig_addr = addr;
	126	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
	127	}
	128
	129	static inline void xp_init_xskb_dma(struct xdp_buff_xsk xskb, struct xsk_buff_pool pool,
	130	dma_addr_t *dma_pages, u64 addr)
	131	{
	132	xskb->frame_dma = (dma_pages[addr >> PAGE_SHIFT] & ~XSK_NEXT_PG_CONTIG_MASK) +
	133	(addr & ~PAGE_MASK);
	134	xskb->dma = xskb->frame_dma + pool->headroom + XDP_PACKET_HEADROOM;
	135	}
	136
2b43470a BT	137	/* AF_XDP ZC drivers, via xdp_sock_buff.h */
2b43470a BT	138	void xp_set_rxq_info(struct xsk_buff_pool pool, struct xdp_rxq_info rxq);
b4e352ff	139	void xp_fill_cb(struct xsk_buff_pool pool, struct xsk_cb_desc desc);
2b43470a BT	140	int xp_dma_map(struct xsk_buff_pool pool, struct device dev,
	141	unsigned long attrs, struct page **pages, u32 nr_pages);
	142	void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs);
	143	struct xdp_buff xp_alloc(struct xsk_buff_pool pool);
47e4075d	144	u32 xp_alloc_batch(struct xsk_buff_pool pool, struct xdp_buff *xdp, u32 max);
2b43470a BT	145	bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count);
	146	void xp_raw_get_data(struct xsk_buff_pool pool, u64 addr);
	147	dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr);
26062b18 BT	148	static inline dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb)
	149	{
	150	return xskb->dma;
	151	}
	152
	153	static inline dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb)
	154	{
	155	return xskb->frame_dma;
	156	}
	157
	158	void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb);
	159	static inline void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb)
	160	{
26062b18 BT	161	xp_dma_sync_for_cpu_slow(xskb);
	162	}
	163
	164	void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
	165	size_t size);
	166	static inline void xp_dma_sync_for_device(struct xsk_buff_pool *pool,
	167	dma_addr_t dma, size_t size)
	168	{
91d5b702	169	if (!pool->dma_need_sync)
26062b18 BT	170	return;
	171
	172	xp_dma_sync_for_device_slow(pool, dma, size);
	173	}
	174
	175	/* Masks for xdp_umem_page flags.
	176	* The low 12-bits of the addr will be 0 since this is the page address, so we
	177	* can use them for flags.
	178	*/
	179	#define XSK_NEXT_PG_CONTIG_SHIFT 0
	180	#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
	181
	182	static inline bool xp_desc_crosses_non_contig_pg(struct xsk_buff_pool *pool,
	183	u64 addr, u32 len)
	184	{
	185	bool cross_pg = (addr & (PAGE_SIZE - 1)) + len > PAGE_SIZE;
	186
2f996198 MK	187	if (likely(!cross_pg))
	188	return false;
	189
6ec7be9a	190	return pool->dma_pages &&
d769ccaf	191	!(pool->dma_pages[addr >> PAGE_SHIFT] & XSK_NEXT_PG_CONTIG_MASK);
26062b18 BT	192	}
26062b18 BT	193
1b725b0c MF	194	static inline bool xp_mb_desc(struct xdp_desc *desc)
	195	{
	196	return desc->options & XDP_PKT_CONTD;
	197	}
	198
26062b18 BT	199	static inline u64 xp_aligned_extract_addr(struct xsk_buff_pool *pool, u64 addr)
	200	{
	201	return addr & pool->chunk_mask;
	202	}
	203
	204	static inline u64 xp_unaligned_extract_addr(u64 addr)
	205	{
	206	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
	207	}
	208
	209	static inline u64 xp_unaligned_extract_offset(u64 addr)
	210	{
	211	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
	212	}
	213
	214	static inline u64 xp_unaligned_add_offset_to_addr(u64 addr)
	215	{
	216	return xp_unaligned_extract_addr(addr) +
	217	xp_unaligned_extract_offset(addr);
	218	}
2b43470a	219
94033cd8 MK	220	static inline u32 xp_aligned_extract_idx(struct xsk_buff_pool *pool, u64 addr)
	221	{
	222	return xp_aligned_extract_addr(pool, addr) >> pool->chunk_shift;
	223	}
	224
	225	static inline void xp_release(struct xdp_buff_xsk *xskb)
	226	{
	227	if (xskb->pool->unaligned)
	228	xskb->pool->free_heads[xskb->pool->free_heads_cnt++] = xskb;
	229	}
	230
	231	static inline u64 xp_get_handle(struct xdp_buff_xsk *xskb)
	232	{
	233	u64 offset = xskb->xdp.data - xskb->xdp.data_hard_start;
	234
	235	offset += xskb->pool->headroom;
	236	if (!xskb->pool->unaligned)
	237	return xskb->orig_addr + offset;
	238	return xskb->orig_addr + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
	239	}
	240
48eb03dd SF	241	static inline bool xp_tx_metadata_enabled(const struct xsk_buff_pool *pool)
	242	{
	243	return pool->tx_metadata_len > 0;
	244	}
	245
2b43470a	246	#endif /* XSK_BUFF_POOL_H_ */