[linux-block.git] / include / rdma / ib_umem.h

/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
/*
 * Copyright (c) 2007 Cisco Systems.  All rights reserved.
 * Copyright (c) 2020 Intel Corporation.  All rights reserved.
 */

#ifndef IB_UMEM_H
#define IB_UMEM_H

#include <linux/list.h>
#include <linux/scatterlist.h>
#include <linux/workqueue.h>
#include <rdma/ib_verbs.h>

struct ib_ucontext;
struct ib_umem_odp;
struct dma_buf_attach_ops;

struct ib_umem {
	struct ib_device       *ibdev;
	struct mm_struct       *owning_mm;
	u64 iova;
	size_t			length;
	unsigned long		address;
	u32 writable : 1;
	u32 is_odp : 1;
	u32 is_dmabuf : 1;
	struct work_struct	work;
	struct sg_append_table sgt_append;
};

struct ib_umem_dmabuf {
	struct ib_umem umem;
	struct dma_buf_attachment *attach;
	struct sg_table *sgt;
	struct scatterlist *first_sg;
	struct scatterlist *last_sg;
	unsigned long first_sg_offset;
	unsigned long last_sg_trim;
	void *private;
	u8 pinned : 1;
};

static inline struct ib_umem_dmabuf *to_ib_umem_dmabuf(struct ib_umem *umem)
{
	return container_of(umem, struct ib_umem_dmabuf, umem);
}

/* Returns the offset of the umem start relative to the first page. */
static inline int ib_umem_offset(struct ib_umem *umem)
{
	return umem->address & ~PAGE_MASK;
}

static inline unsigned long ib_umem_dma_offset(struct ib_umem *umem,
					       unsigned long pgsz)
{
	return (sg_dma_address(umem->sgt_append.sgt.sgl) + ib_umem_offset(umem)) &
	       (pgsz - 1);
}

static inline size_t ib_umem_num_dma_blocks(struct ib_umem *umem,
					    unsigned long pgsz)
{
	return (size_t)((ALIGN(umem->iova + umem->length, pgsz) -
			 ALIGN_DOWN(umem->iova, pgsz))) /
	       pgsz;
}

static inline size_t ib_umem_num_pages(struct ib_umem *umem)
{
	return ib_umem_num_dma_blocks(umem, PAGE_SIZE);
}

static inline void __rdma_umem_block_iter_start(struct ib_block_iter *biter,
						struct ib_umem *umem,
						unsigned long pgsz)
{
	__rdma_block_iter_start(biter, umem->sgt_append.sgt.sgl,
				umem->sgt_append.sgt.nents, pgsz);
}

/**
 * rdma_umem_for_each_dma_block - iterate over contiguous DMA blocks of the umem
 * @umem: umem to iterate over
 * @pgsz: Page size to split the list into
 *
 * pgsz must be <= PAGE_SIZE or computed by ib_umem_find_best_pgsz(). The
 * returned DMA blocks will be aligned to pgsz and span the range:
 * ALIGN_DOWN(umem->address, pgsz) to ALIGN(umem->address + umem->length, pgsz)
 *
 * Performs exactly ib_umem_num_dma_blocks() iterations.
 */
#define rdma_umem_for_each_dma_block(umem, biter, pgsz)                        \
	for (__rdma_umem_block_iter_start(biter, umem, pgsz);                  \
	     __rdma_block_iter_next(biter);)

#ifdef CONFIG_INFINIBAND_USER_MEM

struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
			    size_t size, int access);
void ib_umem_release(struct ib_umem *umem);
int ib_umem_copy_from(void *dst, struct ib_umem *umem, size_t offset,
		      size_t length);
unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
				     unsigned long pgsz_bitmap,
				     unsigned long virt);

/**
 * ib_umem_find_best_pgoff - Find best HW page size
 *
 * @umem: umem struct
 * @pgsz_bitmap bitmap of HW supported page sizes
 * @pgoff_bitmask: Mask of bits that can be represented with an offset
 *
 * This is very similar to ib_umem_find_best_pgsz() except instead of accepting
 * an IOVA it accepts a bitmask specifying what address bits can be represented
 * with a page offset.
 *
 * For instance if the HW has multiple page sizes, requires 64 byte alignemnt,
 * and can support aligned offsets up to 4032 then pgoff_bitmask would be
 * "111111000000".
 *
 * If the pgoff_bitmask requires either alignment in the low bit or an
 * unavailable page size for the high bits, this function returns 0.
 */
static inline unsigned long ib_umem_find_best_pgoff(struct ib_umem *umem,
						    unsigned long pgsz_bitmap,
						    u64 pgoff_bitmask)
{
	struct scatterlist *sg = umem->sgt_append.sgt.sgl;
	dma_addr_t dma_addr;

	dma_addr = sg_dma_address(sg) + (umem->address & ~PAGE_MASK);
	return ib_umem_find_best_pgsz(umem, pgsz_bitmap,
				      dma_addr & pgoff_bitmask);
}

struct ib_umem_dmabuf *ib_umem_dmabuf_get(struct ib_device *device,
					  unsigned long offset, size_t size,
					  int fd, int access,
					  const struct dma_buf_attach_ops *ops);
struct ib_umem_dmabuf *ib_umem_dmabuf_get_pinned(struct ib_device *device,
						 unsigned long offset,
						 size_t size, int fd,
						 int access);
int ib_umem_dmabuf_map_pages(struct ib_umem_dmabuf *umem_dmabuf);
void ib_umem_dmabuf_unmap_pages(struct ib_umem_dmabuf *umem_dmabuf);
void ib_umem_dmabuf_release(struct ib_umem_dmabuf *umem_dmabuf);

#else /* CONFIG_INFINIBAND_USER_MEM */

#include <linux/err.h>

static inline struct ib_umem *ib_umem_get(struct ib_device *device,
					  unsigned long addr, size_t size,
					  int access)
{
	return ERR_PTR(-EOPNOTSUPP);
}
static inline void ib_umem_release(struct ib_umem *umem) { }
static inline int ib_umem_copy_from(void *dst, struct ib_umem *umem, size_t offset,
		      		    size_t length) {
	return -EOPNOTSUPP;
}
static inline unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
						   unsigned long pgsz_bitmap,
						   unsigned long virt)
{
	return 0;
}
static inline unsigned long ib_umem_find_best_pgoff(struct ib_umem *umem,
						    unsigned long pgsz_bitmap,
						    u64 pgoff_bitmask)
{
	return 0;
}
static inline
struct ib_umem_dmabuf *ib_umem_dmabuf_get(struct ib_device *device,
					  unsigned long offset,
					  size_t size, int fd,
					  int access,
					  struct dma_buf_attach_ops *ops)
{
	return ERR_PTR(-EOPNOTSUPP);
}
static inline struct ib_umem_dmabuf *
ib_umem_dmabuf_get_pinned(struct ib_device *device, unsigned long offset,
			  size_t size, int fd, int access)
{
	return ERR_PTR(-EOPNOTSUPP);
}
static inline int ib_umem_dmabuf_map_pages(struct ib_umem_dmabuf *umem_dmabuf)
{
	return -EOPNOTSUPP;
}
static inline void ib_umem_dmabuf_unmap_pages(struct ib_umem_dmabuf *umem_dmabuf) { }
static inline void ib_umem_dmabuf_release(struct ib_umem_dmabuf *umem_dmabuf) { }

#endif /* CONFIG_INFINIBAND_USER_MEM */
#endif /* IB_UMEM_H */
Commit	Line	Data
6bf9d8f6	1	/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */
f7c6a7b5 RD	2	/*
f7c6a7b5 RD	3	* Copyright (c) 2007 Cisco Systems. All rights reserved.
368c0159	4	* Copyright (c) 2020 Intel Corporation. All rights reserved.
f7c6a7b5 RD	5	*/
	6
	7	#ifndef IB_UMEM_H
	8	#define IB_UMEM_H
	9
	10	#include <linux/list.h>
	11	#include <linux/scatterlist.h>
e8edc6e0	12	#include <linux/workqueue.h>
b0ea0fa5	13	#include <rdma/ib_verbs.h>
f7c6a7b5 RD	14
f7c6a7b5 RD	15	struct ib_ucontext;
8ada2c1c	16	struct ib_umem_odp;
368c0159	17	struct dma_buf_attach_ops;
f7c6a7b5 RD	18
f7c6a7b5 RD	19	struct ib_umem {
47f725ee	20	struct ib_device *ibdev;
d4b4dd1b	21	struct mm_struct *owning_mm;
a665aca8	22	u64 iova;
f7c6a7b5	23	size_t length;
406f9e5f	24	unsigned long address;
597ecc5a	25	u32 writable : 1;
597ecc5a	26	u32 is_odp : 1;
368c0159	27	u32 is_dmabuf : 1;
1bf66a30	28	struct work_struct work;
79fbd3e1	29	struct sg_append_table sgt_append;
f7c6a7b5 RD	30	};
f7c6a7b5 RD	31
368c0159 JX	32	struct ib_umem_dmabuf {
	33	struct ib_umem umem;
	34	struct dma_buf_attachment *attach;
	35	struct sg_table *sgt;
	36	struct scatterlist *first_sg;
	37	struct scatterlist *last_sg;
	38	unsigned long first_sg_offset;
	39	unsigned long last_sg_trim;
	40	void *private;
1e4df4a2	41	u8 pinned : 1;
368c0159 JX	42	};
	43
	44	static inline struct ib_umem_dmabuf to_ib_umem_dmabuf(struct ib_umem umem)
	45	{
	46	return container_of(umem, struct ib_umem_dmabuf, umem);
	47	}
	48
406f9e5f HE	49	/* Returns the offset of the umem start relative to the first page. */
	50	static inline int ib_umem_offset(struct ib_umem *umem)
	51	{
d2183c6f	52	return umem->address & ~PAGE_MASK;
406f9e5f HE	53	}
406f9e5f HE	54
b045db62 JG	55	static inline unsigned long ib_umem_dma_offset(struct ib_umem *umem,
	56	unsigned long pgsz)
	57	{
79fbd3e1	58	return (sg_dma_address(umem->sgt_append.sgt.sgl) + ib_umem_offset(umem)) &
b045db62 JG	59	(pgsz - 1);
	60	}
	61
a665aca8 JG	62	static inline size_t ib_umem_num_dma_blocks(struct ib_umem *umem,
	63	unsigned long pgsz)
	64	{
	65	return (size_t)((ALIGN(umem->iova + umem->length, pgsz) -
	66	ALIGN_DOWN(umem->iova, pgsz))) /
	67	pgsz;
	68	}
	69
406f9e5f HE	70	static inline size_t ib_umem_num_pages(struct ib_umem *umem)
406f9e5f HE	71	{
a665aca8	72	return ib_umem_num_dma_blocks(umem, PAGE_SIZE);
406f9e5f HE	73	}
406f9e5f HE	74
ebc24096 JG	75	static inline void __rdma_umem_block_iter_start(struct ib_block_iter *biter,
	76	struct ib_umem *umem,
	77	unsigned long pgsz)
	78	{
79fbd3e1 MG	79	__rdma_block_iter_start(biter, umem->sgt_append.sgt.sgl,
79fbd3e1 MG	80	umem->sgt_append.sgt.nents, pgsz);
ebc24096 JG	81	}
	82
	83	/**
	84	* rdma_umem_for_each_dma_block - iterate over contiguous DMA blocks of the umem
	85	* @umem: umem to iterate over
	86	* @pgsz: Page size to split the list into
	87	*
	88	* pgsz must be <= PAGE_SIZE or computed by ib_umem_find_best_pgsz(). The
	89	* returned DMA blocks will be aligned to pgsz and span the range:
	90	* ALIGN_DOWN(umem->address, pgsz) to ALIGN(umem->address + umem->length, pgsz)
a665aca8 JG	91	*
a665aca8 JG	92	* Performs exactly ib_umem_num_dma_blocks() iterations.
ebc24096 JG	93	*/
	94	#define rdma_umem_for_each_dma_block(umem, biter, pgsz) \
	95	for (__rdma_umem_block_iter_start(biter, umem, pgsz); \
	96	__rdma_block_iter_next(biter);)
	97
f7c6a7b5 RD	98	#ifdef CONFIG_INFINIBAND_USER_MEM
f7c6a7b5 RD	99
c320e527	100	struct ib_umem ib_umem_get(struct ib_device device, unsigned long addr,
72b894b0	101	size_t size, int access);
f7c6a7b5	102	void ib_umem_release(struct ib_umem *umem);
c5d76f13 HE	103	int ib_umem_copy_from(void dst, struct ib_umem umem, size_t offset,
c5d76f13 HE	104	size_t length);
4a353399 SS	105	unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
	106	unsigned long pgsz_bitmap,
	107	unsigned long virt);
368c0159	108
b045db62 JG	109	/**
	110	* ib_umem_find_best_pgoff - Find best HW page size
	111	*
	112	* @umem: umem struct
	113	* @pgsz_bitmap bitmap of HW supported page sizes
	114	* @pgoff_bitmask: Mask of bits that can be represented with an offset
	115	*
	116	* This is very similar to ib_umem_find_best_pgsz() except instead of accepting
	117	* an IOVA it accepts a bitmask specifying what address bits can be represented
	118	* with a page offset.
	119	*
	120	* For instance if the HW has multiple page sizes, requires 64 byte alignemnt,
	121	* and can support aligned offsets up to 4032 then pgoff_bitmask would be
	122	* "111111000000".
	123	*
	124	* If the pgoff_bitmask requires either alignment in the low bit or an
	125	* unavailable page size for the high bits, this function returns 0.
	126	*/
	127	static inline unsigned long ib_umem_find_best_pgoff(struct ib_umem *umem,
	128	unsigned long pgsz_bitmap,
	129	u64 pgoff_bitmask)
	130	{
79fbd3e1	131	struct scatterlist *sg = umem->sgt_append.sgt.sgl;
b045db62 JG	132	dma_addr_t dma_addr;
	133
	134	dma_addr = sg_dma_address(sg) + (umem->address & ~PAGE_MASK);
	135	return ib_umem_find_best_pgsz(umem, pgsz_bitmap,
	136	dma_addr & pgoff_bitmask);
	137	}
f7c6a7b5	138
368c0159 JX	139	struct ib_umem_dmabuf ib_umem_dmabuf_get(struct ib_device device,
	140	unsigned long offset, size_t size,
	141	int fd, int access,
	142	const struct dma_buf_attach_ops *ops);
1e4df4a2 GP	143	struct ib_umem_dmabuf ib_umem_dmabuf_get_pinned(struct ib_device device,
	144	unsigned long offset,
	145	size_t size, int fd,
	146	int access);
368c0159 JX	147	int ib_umem_dmabuf_map_pages(struct ib_umem_dmabuf *umem_dmabuf);
	148	void ib_umem_dmabuf_unmap_pages(struct ib_umem_dmabuf *umem_dmabuf);
	149	void ib_umem_dmabuf_release(struct ib_umem_dmabuf *umem_dmabuf);
	150
f7c6a7b5 RD	151	#else /* CONFIG_INFINIBAND_USER_MEM */
	152
	153	#include <linux/err.h>
	154
c320e527	155	static inline struct ib_umem ib_umem_get(struct ib_device device,
f7c6a7b5	156	unsigned long addr, size_t size,
72b894b0	157	int access)
b0ea0fa5	158	{
368c0159	159	return ERR_PTR(-EOPNOTSUPP);
f7c6a7b5 RD	160	}
f7c6a7b5 RD	161	static inline void ib_umem_release(struct ib_umem *umem) { }
c1395a2a HE	162	static inline int ib_umem_copy_from(void dst, struct ib_umem umem, size_t offset,
c1395a2a HE	163	size_t length) {
368c0159	164	return -EOPNOTSUPP;
c1395a2a	165	}
61690d01 JG	166	static inline unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
	167	unsigned long pgsz_bitmap,
	168	unsigned long virt)
	169	{
	170	return 0;
4a353399	171	}
b045db62 JG	172	static inline unsigned long ib_umem_find_best_pgoff(struct ib_umem *umem,
	173	unsigned long pgsz_bitmap,
	174	u64 pgoff_bitmask)
	175	{
	176	return 0;
	177	}
368c0159 JX	178	static inline
	179	struct ib_umem_dmabuf ib_umem_dmabuf_get(struct ib_device device,
	180	unsigned long offset,
	181	size_t size, int fd,
	182	int access,
	183	struct dma_buf_attach_ops *ops)
	184	{
	185	return ERR_PTR(-EOPNOTSUPP);
	186	}
1e4df4a2 GP	187	static inline struct ib_umem_dmabuf *
	188	ib_umem_dmabuf_get_pinned(struct ib_device *device, unsigned long offset,
	189	size_t size, int fd, int access)
	190	{
	191	return ERR_PTR(-EOPNOTSUPP);
	192	}
368c0159 JX	193	static inline int ib_umem_dmabuf_map_pages(struct ib_umem_dmabuf *umem_dmabuf)
	194	{
	195	return -EOPNOTSUPP;
	196	}
	197	static inline void ib_umem_dmabuf_unmap_pages(struct ib_umem_dmabuf *umem_dmabuf) { }
	198	static inline void ib_umem_dmabuf_release(struct ib_umem_dmabuf *umem_dmabuf) { }
4a353399	199
f7c6a7b5	200	#endif /* CONFIG_INFINIBAND_USER_MEM */
f7c6a7b5	201	#endif /* IB_UMEM_H */