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

/* SPDX-License-Identifier: GPL-2.0
 *
 * page_pool.h
 *	Author:	Jesper Dangaard Brouer <netoptimizer@brouer.com>
 *	Copyright (C) 2016 Red Hat, Inc.
 */

/**
 * DOC: page_pool allocator
 *
 * This page_pool allocator is optimized for the XDP mode that
 * uses one-frame-per-page, but have fallbacks that act like the
 * regular page allocator APIs.
 *
 * Basic use involve replacing alloc_pages() calls with the
 * page_pool_alloc_pages() call.  Drivers should likely use
 * page_pool_dev_alloc_pages() replacing dev_alloc_pages().
 *
 * API keeps track of in-flight pages, in-order to let API user know
 * when it is safe to dealloactor page_pool object.  Thus, API users
 * must make sure to call page_pool_release_page() when a page is
 * "leaving" the page_pool.  Or call page_pool_put_page() where
 * appropiate.  For maintaining correct accounting.
 *
 * API user must only call page_pool_put_page() once on a page, as it
 * will either recycle the page, or in case of elevated refcnt, it
 * will release the DMA mapping and in-flight state accounting.  We
 * hope to lift this requirement in the future.
 */
#ifndef _NET_PAGE_POOL_H
#define _NET_PAGE_POOL_H

#include <linux/mm.h> /* Needed by ptr_ring */
#include <linux/ptr_ring.h>
#include <linux/dma-direction.h>

#define PP_FLAG_DMA_MAP		BIT(0) /* Should page_pool do the DMA
					* map/unmap
					*/
#define PP_FLAG_DMA_SYNC_DEV	BIT(1) /* If set all pages that the driver gets
					* from page_pool will be
					* DMA-synced-for-device according to
					* the length provided by the device
					* driver.
					* Please note DMA-sync-for-CPU is still
					* device driver responsibility
					*/
#define PP_FLAG_PAGE_FRAG	BIT(2) /* for page frag feature */
#define PP_FLAG_ALL		(PP_FLAG_DMA_MAP |\
				 PP_FLAG_DMA_SYNC_DEV |\
				 PP_FLAG_PAGE_FRAG)

/*
 * Fast allocation side cache array/stack
 *
 * The cache size and refill watermark is related to the network
 * use-case.  The NAPI budget is 64 packets.  After a NAPI poll the RX
 * ring is usually refilled and the max consumed elements will be 64,
 * thus a natural max size of objects needed in the cache.
 *
 * Keeping room for more objects, is due to XDP_DROP use-case.  As
 * XDP_DROP allows the opportunity to recycle objects directly into
 * this array, as it shares the same softirq/NAPI protection.  If
 * cache is already full (or partly full) then the XDP_DROP recycles
 * would have to take a slower code path.
 */
#define PP_ALLOC_CACHE_SIZE	128
#define PP_ALLOC_CACHE_REFILL	64
struct pp_alloc_cache {
	u32 count;
	struct page *cache[PP_ALLOC_CACHE_SIZE];
};

struct page_pool_params {
	unsigned int	flags;
	unsigned int	order;
	unsigned int	pool_size;
	int		nid;  /* Numa node id to allocate from pages from */
	struct device	*dev; /* device, for DMA pre-mapping purposes */
	struct napi_struct *napi; /* Sole consumer of pages, otherwise NULL */
	enum dma_data_direction dma_dir; /* DMA mapping direction */
	unsigned int	max_len; /* max DMA sync memory size */
	unsigned int	offset;  /* DMA addr offset */
	void (*init_callback)(struct page *page, void *arg);
	void *init_arg;
};

#ifdef CONFIG_PAGE_POOL_STATS
struct page_pool_alloc_stats {
	u64 fast; /* fast path allocations */
	u64 slow; /* slow-path order 0 allocations */
	u64 slow_high_order; /* slow-path high order allocations */
	u64 empty; /* failed refills due to empty ptr ring, forcing
		    * slow path allocation
		    */
	u64 refill; /* allocations via successful refill */
	u64 waive;  /* failed refills due to numa zone mismatch */
};

struct page_pool_recycle_stats {
	u64 cached;	/* recycling placed page in the cache. */
	u64 cache_full; /* cache was full */
	u64 ring;	/* recycling placed page back into ptr ring */
	u64 ring_full;	/* page was released from page-pool because
			 * PTR ring was full.
			 */
	u64 released_refcnt; /* page released because of elevated
			      * refcnt
			      */
};

/* This struct wraps the above stats structs so users of the
 * page_pool_get_stats API can pass a single argument when requesting the
 * stats for the page pool.
 */
struct page_pool_stats {
	struct page_pool_alloc_stats alloc_stats;
	struct page_pool_recycle_stats recycle_stats;
};

int page_pool_ethtool_stats_get_count(void);
u8 *page_pool_ethtool_stats_get_strings(u8 *data);
u64 *page_pool_ethtool_stats_get(u64 *data, void *stats);

/*
 * Drivers that wish to harvest page pool stats and report them to users
 * (perhaps via ethtool, debugfs, or another mechanism) can allocate a
 * struct page_pool_stats call page_pool_get_stats to get stats for the specified pool.
 */
bool page_pool_get_stats(struct page_pool *pool,
			 struct page_pool_stats *stats);
#else

static inline int page_pool_ethtool_stats_get_count(void)
{
	return 0;
}

static inline u8 *page_pool_ethtool_stats_get_strings(u8 *data)
{
	return data;
}

static inline u64 *page_pool_ethtool_stats_get(u64 *data, void *stats)
{
	return data;
}

#endif

struct page_pool {
	struct page_pool_params p;

	struct delayed_work release_dw;
	void (*disconnect)(void *);
	unsigned long defer_start;
	unsigned long defer_warn;

	u32 pages_state_hold_cnt;
	unsigned int frag_offset;
	struct page *frag_page;
	long frag_users;

#ifdef CONFIG_PAGE_POOL_STATS
	/* these stats are incremented while in softirq context */
	struct page_pool_alloc_stats alloc_stats;
#endif
	u32 xdp_mem_id;

	/*
	 * Data structure for allocation side
	 *
	 * Drivers allocation side usually already perform some kind
	 * of resource protection.  Piggyback on this protection, and
	 * require driver to protect allocation side.
	 *
	 * For NIC drivers this means, allocate a page_pool per
	 * RX-queue. As the RX-queue is already protected by
	 * Softirq/BH scheduling and napi_schedule. NAPI schedule
	 * guarantee that a single napi_struct will only be scheduled
	 * on a single CPU (see napi_schedule).
	 */
	struct pp_alloc_cache alloc ____cacheline_aligned_in_smp;

	/* Data structure for storing recycled pages.
	 *
	 * Returning/freeing pages is more complicated synchronization
	 * wise, because free's can happen on remote CPUs, with no
	 * association with allocation resource.
	 *
	 * Use ptr_ring, as it separates consumer and producer
	 * effeciently, it a way that doesn't bounce cache-lines.
	 *
	 * TODO: Implement bulk return pages into this structure.
	 */
	struct ptr_ring ring;

#ifdef CONFIG_PAGE_POOL_STATS
	/* recycle stats are per-cpu to avoid locking */
	struct page_pool_recycle_stats __percpu *recycle_stats;
#endif
	atomic_t pages_state_release_cnt;

	/* A page_pool is strictly tied to a single RX-queue being
	 * protected by NAPI, due to above pp_alloc_cache. This
	 * refcnt serves purpose is to simplify drivers error handling.
	 */
	refcount_t user_cnt;

	u64 destroy_cnt;
};

struct page *page_pool_alloc_pages(struct page_pool *pool, gfp_t gfp);

static inline struct page *page_pool_dev_alloc_pages(struct page_pool *pool)
{
	gfp_t gfp = (GFP_ATOMIC | __GFP_NOWARN);

	return page_pool_alloc_pages(pool, gfp);
}

struct page *page_pool_alloc_frag(struct page_pool *pool, unsigned int *offset,
				  unsigned int size, gfp_t gfp);

static inline struct page *page_pool_dev_alloc_frag(struct page_pool *pool,
						    unsigned int *offset,
						    unsigned int size)
{
	gfp_t gfp = (GFP_ATOMIC | __GFP_NOWARN);

	return page_pool_alloc_frag(pool, offset, size, gfp);
}

/* get the stored dma direction. A driver might decide to treat this locally and
 * avoid the extra cache line from page_pool to determine the direction
 */
static
inline enum dma_data_direction page_pool_get_dma_dir(struct page_pool *pool)
{
	return pool->p.dma_dir;
}

bool page_pool_return_skb_page(struct page *page, bool napi_safe);

struct page_pool *page_pool_create(const struct page_pool_params *params);

struct xdp_mem_info;

#ifdef CONFIG_PAGE_POOL
void page_pool_unlink_napi(struct page_pool *pool);
void page_pool_destroy(struct page_pool *pool);
void page_pool_use_xdp_mem(struct page_pool *pool, void (*disconnect)(void *),
			   struct xdp_mem_info *mem);
void page_pool_release_page(struct page_pool *pool, struct page *page);
void page_pool_put_page_bulk(struct page_pool *pool, void **data,
			     int count);
#else
static inline void page_pool_unlink_napi(struct page_pool *pool)
{
}

static inline void page_pool_destroy(struct page_pool *pool)
{
}

static inline void page_pool_use_xdp_mem(struct page_pool *pool,
					 void (*disconnect)(void *),
					 struct xdp_mem_info *mem)
{
}
static inline void page_pool_release_page(struct page_pool *pool,
					  struct page *page)
{
}

static inline void page_pool_put_page_bulk(struct page_pool *pool, void **data,
					   int count)
{
}
#endif

void page_pool_put_defragged_page(struct page_pool *pool, struct page *page,
				  unsigned int dma_sync_size,
				  bool allow_direct);

/* pp_frag_count represents the number of writers who can update the page
 * either by updating skb->data or via DMA mappings for the device.
 * We can't rely on the page refcnt for that as we don't know who might be
 * holding page references and we can't reliably destroy or sync DMA mappings
 * of the fragments.
 *
 * When pp_frag_count reaches 0 we can either recycle the page if the page
 * refcnt is 1 or return it back to the memory allocator and destroy any
 * mappings we have.
 */
static inline void page_pool_fragment_page(struct page *page, long nr)
{
	atomic_long_set(&page->pp_frag_count, nr);
}

static inline long page_pool_defrag_page(struct page *page, long nr)
{
	long ret;

	/* If nr == pp_frag_count then we have cleared all remaining
	 * references to the page. No need to actually overwrite it, instead
	 * we can leave this to be overwritten by the calling function.
	 *
	 * The main advantage to doing this is that an atomic_read is
	 * generally a much cheaper operation than an atomic update,
	 * especially when dealing with a page that may be partitioned
	 * into only 2 or 3 pieces.
	 */
	if (atomic_long_read(&page->pp_frag_count) == nr)
		return 0;

	ret = atomic_long_sub_return(nr, &page->pp_frag_count);
	WARN_ON(ret < 0);
	return ret;
}

static inline bool page_pool_is_last_frag(struct page_pool *pool,
					  struct page *page)
{
	/* If fragments aren't enabled or count is 0 we were the last user */
	return !(pool->p.flags & PP_FLAG_PAGE_FRAG) ||
	       (page_pool_defrag_page(page, 1) == 0);
}

static inline void page_pool_put_page(struct page_pool *pool,
				      struct page *page,
				      unsigned int dma_sync_size,
				      bool allow_direct)
{
	/* When page_pool isn't compiled-in, net/core/xdp.c doesn't
	 * allow registering MEM_TYPE_PAGE_POOL, but shield linker.
	 */
#ifdef CONFIG_PAGE_POOL
	if (!page_pool_is_last_frag(pool, page))
		return;

	page_pool_put_defragged_page(pool, page, dma_sync_size, allow_direct);
#endif
}

/* Same as above but will try to sync the entire area pool->max_len */
static inline void page_pool_put_full_page(struct page_pool *pool,
					   struct page *page, bool allow_direct)
{
	page_pool_put_page(pool, page, -1, allow_direct);
}

/* Same as above but the caller must guarantee safe context. e.g NAPI */
static inline void page_pool_recycle_direct(struct page_pool *pool,
					    struct page *page)
{
	page_pool_put_full_page(pool, page, true);
}

#define PAGE_POOL_DMA_USE_PP_FRAG_COUNT	\
		(sizeof(dma_addr_t) > sizeof(unsigned long))

static inline dma_addr_t page_pool_get_dma_addr(struct page *page)
{
	dma_addr_t ret = page->dma_addr;

	if (PAGE_POOL_DMA_USE_PP_FRAG_COUNT)
		ret |= (dma_addr_t)page->dma_addr_upper << 16 << 16;

	return ret;
}

static inline void page_pool_set_dma_addr(struct page *page, dma_addr_t addr)
{
	page->dma_addr = addr;
	if (PAGE_POOL_DMA_USE_PP_FRAG_COUNT)
		page->dma_addr_upper = upper_32_bits(addr);
}

static inline bool is_page_pool_compiled_in(void)
{
#ifdef CONFIG_PAGE_POOL
	return true;
#else
	return false;
#endif
}

static inline bool page_pool_put(struct page_pool *pool)
{
	return refcount_dec_and_test(&pool->user_cnt);
}

/* Caller must provide appropriate safe context, e.g. NAPI. */
void page_pool_update_nid(struct page_pool *pool, int new_nid);
static inline void page_pool_nid_changed(struct page_pool *pool, int new_nid)
{
	if (unlikely(pool->p.nid != new_nid))
		page_pool_update_nid(pool, new_nid);
}

static inline void page_pool_ring_lock(struct page_pool *pool)
	__acquires(&pool->ring.producer_lock)
{
	if (in_softirq())
		spin_lock(&pool->ring.producer_lock);
	else
		spin_lock_bh(&pool->ring.producer_lock);
}

static inline void page_pool_ring_unlock(struct page_pool *pool)
	__releases(&pool->ring.producer_lock)
{
	if (in_softirq())
		spin_unlock(&pool->ring.producer_lock);
	else
		spin_unlock_bh(&pool->ring.producer_lock);
}

#endif /* _NET_PAGE_POOL_H */
Commit	Line	Data
ff7d6b27 JDB	1	/* SPDX-License-Identifier: GPL-2.0
	2	*
	3	* page_pool.h
	4	* Author: Jesper Dangaard Brouer <netoptimizer@brouer.com>
	5	* Copyright (C) 2016 Red Hat, Inc.
	6	*/
	7
	8	/**
	9	* DOC: page_pool allocator
	10	*
	11	* This page_pool allocator is optimized for the XDP mode that
	12	* uses one-frame-per-page, but have fallbacks that act like the
	13	* regular page allocator APIs.
	14	*
	15	* Basic use involve replacing alloc_pages() calls with the
	16	* page_pool_alloc_pages() call. Drivers should likely use
	17	* page_pool_dev_alloc_pages() replacing dev_alloc_pages().
	18	*
99c07c43 JDB	19	* API keeps track of in-flight pages, in-order to let API user know
	20	* when it is safe to dealloactor page_pool object. Thus, API users
	21	* must make sure to call page_pool_release_page() when a page is
	22	* "leaving" the page_pool. Or call page_pool_put_page() where
	23	* appropiate. For maintaining correct accounting.
ff7d6b27	24	*
99c07c43 JDB	25	* API user must only call page_pool_put_page() once on a page, as it
	26	* will either recycle the page, or in case of elevated refcnt, it
	27	* will release the DMA mapping and in-flight state accounting. We
	28	* hope to lift this requirement in the future.
ff7d6b27 JDB	29	*/
	30	#ifndef _NET_PAGE_POOL_H
	31	#define _NET_PAGE_POOL_H
	32
	33	#include <linux/mm.h> /* Needed by ptr_ring */
	34	#include <linux/ptr_ring.h>
	35	#include <linux/dma-direction.h>
	36
e68bc756 LB	37	#define PP_FLAG_DMA_MAP BIT(0) /* Should page_pool do the DMA
	38	* map/unmap
	39	*/
	40	#define PP_FLAG_DMA_SYNC_DEV BIT(1) /* If set all pages that the driver gets
	41	* from page_pool will be
	42	* DMA-synced-for-device according to
	43	* the length provided by the device
	44	* driver.
	45	* Please note DMA-sync-for-CPU is still
	46	* device driver responsibility
	47	*/
0e9d2a0a YL	48	#define PP_FLAG_PAGE_FRAG BIT(2) /* for page frag feature */
	49	#define PP_FLAG_ALL (PP_FLAG_DMA_MAP \|\
	50	PP_FLAG_DMA_SYNC_DEV \|\
	51	PP_FLAG_PAGE_FRAG)
ff7d6b27 JDB	52
	53	/*
	54	* Fast allocation side cache array/stack
	55	*
	56	* The cache size and refill watermark is related to the network
	57	* use-case. The NAPI budget is 64 packets. After a NAPI poll the RX
	58	* ring is usually refilled and the max consumed elements will be 64,
	59	* thus a natural max size of objects needed in the cache.
	60	*
	61	* Keeping room for more objects, is due to XDP_DROP use-case. As
	62	* XDP_DROP allows the opportunity to recycle objects directly into
	63	* this array, as it shares the same softirq/NAPI protection. If
	64	* cache is already full (or partly full) then the XDP_DROP recycles
	65	* would have to take a slower code path.
	66	*/
	67	#define PP_ALLOC_CACHE_SIZE 128
	68	#define PP_ALLOC_CACHE_REFILL 64
	69	struct pp_alloc_cache {
	70	u32 count;
be5dba25	71	struct page *cache[PP_ALLOC_CACHE_SIZE];
ff7d6b27 JDB	72	};
	73
	74	struct page_pool_params {
	75	unsigned int flags;
	76	unsigned int order;
	77	unsigned int pool_size;
	78	int nid; /* Numa node id to allocate from pages from */
	79	struct device dev; / device, for DMA pre-mapping purposes */
8c48eea3	80	struct napi_struct napi; / Sole consumer of pages, otherwise NULL */
ff7d6b27	81	enum dma_data_direction dma_dir; /* DMA mapping direction */
e68bc756 LB	82	unsigned int max_len; /* max DMA sync memory size */
e68bc756 LB	83	unsigned int offset; /* DMA addr offset */
35b2e549 THJ	84	void (init_callback)(struct page page, void *arg);
35b2e549 THJ	85	void *init_arg;
ff7d6b27 JDB	86	};
ff7d6b27 JDB	87
8610037e JD	88	#ifdef CONFIG_PAGE_POOL_STATS
	89	struct page_pool_alloc_stats {
	90	u64 fast; /* fast path allocations */
	91	u64 slow; /* slow-path order 0 allocations */
	92	u64 slow_high_order; /* slow-path high order allocations */
	93	u64 empty; /* failed refills due to empty ptr ring, forcing
	94	* slow path allocation
	95	*/
	96	u64 refill; /* allocations via successful refill */
	97	u64 waive; /* failed refills due to numa zone mismatch */
	98	};
ad6fa1e1 JD	99
	100	struct page_pool_recycle_stats {
	101	u64 cached; /* recycling placed page in the cache. */
	102	u64 cache_full; /* cache was full */
	103	u64 ring; /* recycling placed page back into ptr ring */
	104	u64 ring_full; /* page was released from page-pool because
	105	* PTR ring was full.
	106	*/
	107	u64 released_refcnt; /* page released because of elevated
	108	* refcnt
	109	*/
	110	};
6b95e338 JD	111
	112	/* This struct wraps the above stats structs so users of the
	113	* page_pool_get_stats API can pass a single argument when requesting the
	114	* stats for the page pool.
	115	*/
	116	struct page_pool_stats {
	117	struct page_pool_alloc_stats alloc_stats;
	118	struct page_pool_recycle_stats recycle_stats;
	119	};
	120
f3c5264f LB	121	int page_pool_ethtool_stats_get_count(void);
	122	u8 page_pool_ethtool_stats_get_strings(u8 data);
	123	u64 page_pool_ethtool_stats_get(u64 data, void *stats);
	124
6b95e338 JD	125	/*
	126	* Drivers that wish to harvest page pool stats and report them to users
	127	* (perhaps via ethtool, debugfs, or another mechanism) can allocate a
	128	* struct page_pool_stats call page_pool_get_stats to get stats for the specified pool.
	129	*/
	130	bool page_pool_get_stats(struct page_pool *pool,
	131	struct page_pool_stats *stats);
f3c5264f LB	132	#else
	133
	134	static inline int page_pool_ethtool_stats_get_count(void)
	135	{
	136	return 0;
	137	}
	138
	139	static inline u8 page_pool_ethtool_stats_get_strings(u8 data)
	140	{
	141	return data;
	142	}
	143
	144	static inline u64 page_pool_ethtool_stats_get(u64 data, void *stats)
	145	{
	146	return data;
	147	}
	148
8610037e JD	149	#endif
8610037e JD	150
ff7d6b27	151	struct page_pool {
ff7d6b27 JDB	152	struct page_pool_params p;
ff7d6b27 JDB	153
c3f812ce JL	154	struct delayed_work release_dw;
	155	void (disconnect)(void );
	156	unsigned long defer_start;
	157	unsigned long defer_warn;
	158
	159	u32 pages_state_hold_cnt;
53e0961d YL	160	unsigned int frag_offset;
	161	struct page *frag_page;
	162	long frag_users;
8610037e JD	163
	164	#ifdef CONFIG_PAGE_POOL_STATS
	165	/* these stats are incremented while in softirq context */
	166	struct page_pool_alloc_stats alloc_stats;
	167	#endif
64693ec7	168	u32 xdp_mem_id;
99c07c43	169
ff7d6b27 JDB	170	/*
	171	* Data structure for allocation side
	172	*
	173	* Drivers allocation side usually already perform some kind
	174	* of resource protection. Piggyback on this protection, and
	175	* require driver to protect allocation side.
	176	*
	177	* For NIC drivers this means, allocate a page_pool per
	178	* RX-queue. As the RX-queue is already protected by
	179	* Softirq/BH scheduling and napi_schedule. NAPI schedule
	180	* guarantee that a single napi_struct will only be scheduled
	181	* on a single CPU (see napi_schedule).
	182	*/
	183	struct pp_alloc_cache alloc ____cacheline_aligned_in_smp;
	184
	185	/* Data structure for storing recycled pages.
	186	*
	187	* Returning/freeing pages is more complicated synchronization
	188	* wise, because free's can happen on remote CPUs, with no
	189	* association with allocation resource.
	190	*
	191	* Use ptr_ring, as it separates consumer and producer
	192	* effeciently, it a way that doesn't bounce cache-lines.
	193	*
	194	* TODO: Implement bulk return pages into this structure.
	195	*/
	196	struct ptr_ring ring;
99c07c43	197
ad6fa1e1 JD	198	#ifdef CONFIG_PAGE_POOL_STATS
	199	/* recycle stats are per-cpu to avoid locking */
	200	struct page_pool_recycle_stats __percpu *recycle_stats;
	201	#endif
99c07c43	202	atomic_t pages_state_release_cnt;
1da4bbef IK	203
	204	/* A page_pool is strictly tied to a single RX-queue being
	205	* protected by NAPI, due to above pp_alloc_cache. This
	206	* refcnt serves purpose is to simplify drivers error handling.
	207	*/
	208	refcount_t user_cnt;
7c9e6942 JDB	209
7c9e6942 JDB	210	u64 destroy_cnt;
ff7d6b27 JDB	211	};
	212
	213	struct page page_pool_alloc_pages(struct page_pool pool, gfp_t gfp);
	214
	215	static inline struct page page_pool_dev_alloc_pages(struct page_pool pool)
	216	{
	217	gfp_t gfp = (GFP_ATOMIC \| __GFP_NOWARN);
	218
	219	return page_pool_alloc_pages(pool, gfp);
	220	}
	221
53e0961d YL	222	struct page page_pool_alloc_frag(struct page_pool pool, unsigned int *offset,
	223	unsigned int size, gfp_t gfp);
	224
	225	static inline struct page page_pool_dev_alloc_frag(struct page_pool pool,
	226	unsigned int *offset,
	227	unsigned int size)
	228	{
	229	gfp_t gfp = (GFP_ATOMIC \| __GFP_NOWARN);
	230
	231	return page_pool_alloc_frag(pool, offset, size, gfp);
	232	}
	233
bb005f2a IA	234	/* get the stored dma direction. A driver might decide to treat this locally and
	235	* avoid the extra cache line from page_pool to determine the direction
	236	*/
	237	static
	238	inline enum dma_data_direction page_pool_get_dma_dir(struct page_pool *pool)
	239	{
	240	return pool->p.dma_dir;
	241	}
	242
8c48eea3	243	bool page_pool_return_skb_page(struct page *page, bool napi_safe);
6a5bcd84	244
ff7d6b27 JDB	245	struct page_pool page_pool_create(const struct page_pool_params params);
ff7d6b27 JDB	246
64693ec7 THJ	247	struct xdp_mem_info;
64693ec7 THJ	248
e54cfd7e	249	#ifdef CONFIG_PAGE_POOL
dd64b232	250	void page_pool_unlink_napi(struct page_pool *pool);
c3f812ce	251	void page_pool_destroy(struct page_pool *pool);
64693ec7 THJ	252	void page_pool_use_xdp_mem(struct page_pool pool, void (disconnect)(void *),
64693ec7 THJ	253	struct xdp_mem_info *mem);
458de8a9	254	void page_pool_release_page(struct page_pool pool, struct page page);
78862447 LB	255	void page_pool_put_page_bulk(struct page_pool pool, void *data,
78862447 LB	256	int count);
c3f812ce	257	#else
dd64b232 JK	258	static inline void page_pool_unlink_napi(struct page_pool *pool)
	259	{
	260	}
	261
1da4bbef IK	262	static inline void page_pool_destroy(struct page_pool *pool)
1da4bbef IK	263	{
c3f812ce	264	}
1da4bbef	265
c3f812ce	266	static inline void page_pool_use_xdp_mem(struct page_pool *pool,
64693ec7 THJ	267	void (disconnect)(void ),
64693ec7 THJ	268	struct xdp_mem_info *mem)
c3f812ce	269	{
1da4bbef	270	}
458de8a9 IA	271	static inline void page_pool_release_page(struct page_pool *pool,
	272	struct page *page)
	273	{
	274	}
78862447 LB	275
	276	static inline void page_pool_put_page_bulk(struct page_pool pool, void *data,
	277	int count)
	278	{
	279	}
c3f812ce	280	#endif
1da4bbef	281
52cc6ffc AD	282	void page_pool_put_defragged_page(struct page_pool pool, struct page page,
	283	unsigned int dma_sync_size,
	284	bool allow_direct);
ff7d6b27	285
4d4266e3 IA	286	/* pp_frag_count represents the number of writers who can update the page
	287	* either by updating skb->data or via DMA mappings for the device.
	288	* We can't rely on the page refcnt for that as we don't know who might be
	289	* holding page references and we can't reliably destroy or sync DMA mappings
	290	* of the fragments.
	291	*
	292	* When pp_frag_count reaches 0 we can either recycle the page if the page
	293	* refcnt is 1 or return it back to the memory allocator and destroy any
	294	* mappings we have.
	295	*/
52cc6ffc AD	296	static inline void page_pool_fragment_page(struct page *page, long nr)
	297	{
	298	atomic_long_set(&page->pp_frag_count, nr);
	299	}
	300
	301	static inline long page_pool_defrag_page(struct page *page, long nr)
	302	{
	303	long ret;
	304
	305	/* If nr == pp_frag_count then we have cleared all remaining
	306	* references to the page. No need to actually overwrite it, instead
	307	* we can leave this to be overwritten by the calling function.
	308	*
	309	* The main advantage to doing this is that an atomic_read is
	310	* generally a much cheaper operation than an atomic update,
	311	* especially when dealing with a page that may be partitioned
	312	* into only 2 or 3 pieces.
	313	*/
	314	if (atomic_long_read(&page->pp_frag_count) == nr)
	315	return 0;
	316
	317	ret = atomic_long_sub_return(nr, &page->pp_frag_count);
	318	WARN_ON(ret < 0);
	319	return ret;
	320	}
	321
	322	static inline bool page_pool_is_last_frag(struct page_pool *pool,
	323	struct page *page)
	324	{
	325	/* If fragments aren't enabled or count is 0 we were the last user */
	326	return !(pool->p.flags & PP_FLAG_PAGE_FRAG) \|\|
	327	(page_pool_defrag_page(page, 1) == 0);
	328	}
	329
	330	static inline void page_pool_put_page(struct page_pool *pool,
	331	struct page *page,
	332	unsigned int dma_sync_size,
	333	bool allow_direct)
ff7d6b27	334	{
57d0a1c1 JDB	335	/* When page_pool isn't compiled-in, net/core/xdp.c doesn't
	336	* allow registering MEM_TYPE_PAGE_POOL, but shield linker.
	337	*/
	338	#ifdef CONFIG_PAGE_POOL
52cc6ffc AD	339	if (!page_pool_is_last_frag(pool, page))
	340	return;
	341
	342	page_pool_put_defragged_page(pool, page, dma_sync_size, allow_direct);
57d0a1c1	343	#endif
ff7d6b27	344	}
458de8a9	345
52cc6ffc AD	346	/* Same as above but will try to sync the entire area pool->max_len */
	347	static inline void page_pool_put_full_page(struct page_pool *pool,
	348	struct page *page, bool allow_direct)
	349	{
	350	page_pool_put_page(pool, page, -1, allow_direct);
	351	}
	352
458de8a9	353	/* Same as above but the caller must guarantee safe context. e.g NAPI */
ff7d6b27 JDB	354	static inline void page_pool_recycle_direct(struct page_pool *pool,
	355	struct page *page)
	356	{
458de8a9	357	page_pool_put_full_page(pool, page, true);
6bf071bf JDB	358	}
6bf071bf JDB	359
f915b75b YL	360	#define PAGE_POOL_DMA_USE_PP_FRAG_COUNT \
	361	(sizeof(dma_addr_t) > sizeof(unsigned long))
	362
0afdeeed IA	363	static inline dma_addr_t page_pool_get_dma_addr(struct page *page)
0afdeeed IA	364	{
f915b75b YL	365	dma_addr_t ret = page->dma_addr;
	366
	367	if (PAGE_POOL_DMA_USE_PP_FRAG_COUNT)
	368	ret \|= (dma_addr_t)page->dma_addr_upper << 16 << 16;
	369
	370	return ret;
9ddb3c14 MWO	371	}
	372
	373	static inline void page_pool_set_dma_addr(struct page *page, dma_addr_t addr)
	374	{
0e9d2a0a	375	page->dma_addr = addr;
f915b75b YL	376	if (PAGE_POOL_DMA_USE_PP_FRAG_COUNT)
f915b75b YL	377	page->dma_addr_upper = upper_32_bits(addr);
0e9d2a0a YL	378	}
0e9d2a0a YL	379
57d0a1c1 JDB	380	static inline bool is_page_pool_compiled_in(void)
	381	{
	382	#ifdef CONFIG_PAGE_POOL
	383	return true;
	384	#else
	385	return false;
	386	#endif
	387	}
	388
1da4bbef IK	389	static inline bool page_pool_put(struct page_pool *pool)
	390	{
	391	return refcount_dec_and_test(&pool->user_cnt);
	392	}
	393
bc836748 SM	394	/* Caller must provide appropriate safe context, e.g. NAPI. */
	395	void page_pool_update_nid(struct page_pool *pool, int new_nid);
	396	static inline void page_pool_nid_changed(struct page_pool *pool, int new_nid)
	397	{
	398	if (unlikely(pool->p.nid != new_nid))
	399	page_pool_update_nid(pool, new_nid);
	400	}
78862447 LB	401
	402	static inline void page_pool_ring_lock(struct page_pool *pool)
	403	__acquires(&pool->ring.producer_lock)
	404	{
542bcea4	405	if (in_softirq())
78862447 LB	406	spin_lock(&pool->ring.producer_lock);
	407	else
	408	spin_lock_bh(&pool->ring.producer_lock);
	409	}
	410
	411	static inline void page_pool_ring_unlock(struct page_pool *pool)
	412	__releases(&pool->ring.producer_lock)
	413	{
542bcea4	414	if (in_softirq())
78862447 LB	415	spin_unlock(&pool->ring.producer_lock);
	416	else
	417	spin_unlock_bh(&pool->ring.producer_lock);
	418	}
	419
ff7d6b27	420	#endif /* _NET_PAGE_POOL_H */