1/*
  2 * Copyright (c) 2014 Mellanox Technologies. All rights reserved.
  3 *
  4 * This software is available to you under a choice of one of two
  5 * licenses.  You may choose to be licensed under the terms of the GNU
  6 * General Public License (GPL) Version 2, available from the file
  7 * COPYING in the main directory of this source tree, or the
  8 * OpenIB.org BSD license below:
  9 *
 10 *     Redistribution and use in source and binary forms, with or
 11 *     without modification, are permitted provided that the following
 12 *     conditions are met:
 13 *
 14 *      - Redistributions of source code must retain the above
 15 *        copyright notice, this list of conditions and the following
 16 *        disclaimer.
 17 *
 18 *      - Redistributions in binary form must reproduce the above
 19 *        copyright notice, this list of conditions and the following
 20 *        disclaimer in the documentation and/or other materials
 21 *        provided with the distribution.
 22 *
 23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 30 * SOFTWARE.
 31 */
 32
 33#ifndef IB_UMEM_ODP_H
 34#define IB_UMEM_ODP_H
 35
 36#include <rdma/ib_umem.h>
 37#include <rdma/ib_verbs.h>
 38#include <linux/interval_tree.h>
 39
 40struct ib_umem_odp {
 41	struct ib_umem umem;
 42	struct ib_ucontext_per_mm *per_mm;
 43
 44	/*
 45	 * An array of the pages included in the on-demand paging umem.
 46	 * Indices of pages that are currently not mapped into the device will
 47	 * contain NULL.
 48	 */
 49	struct page		**page_list;
 50	/*
 51	 * An array of the same size as page_list, with DMA addresses mapped
 52	 * for pages the pages in page_list. The lower two bits designate
 53	 * access permissions. See ODP_READ_ALLOWED_BIT and
 54	 * ODP_WRITE_ALLOWED_BIT.
 55	 */
 56	dma_addr_t		*dma_list;
 57	/*
 58	 * The umem_mutex protects the page_list and dma_list fields of an ODP
 59	 * umem, allowing only a single thread to map/unmap pages. The mutex
 60	 * also protects access to the mmu notifier counters.
 61	 */
 62	struct mutex		umem_mutex;
 63	void			*private; /* for the HW driver to use. */
 64
 65	int notifiers_seq;
 66	int notifiers_count;
 67	int npages;
 68
 69	/* Tree tracking */
 70	struct interval_tree_node interval_tree;
 71
 72	/*
 73	 * An implicit odp umem cannot be DMA mapped, has 0 length, and serves
 74	 * only as an anchor for the driver to hold onto the per_mm. FIXME:
 75	 * This should be removed and drivers should work with the per_mm
 76	 * directly.
 77	 */
 78	bool is_implicit_odp;
 79
 80	struct completion	notifier_completion;
 81	int			dying;
 82	unsigned int		page_shift;
 83	struct work_struct	work;
 84};
 85
 86static inline struct ib_umem_odp *to_ib_umem_odp(struct ib_umem *umem)
 87{
 88	return container_of(umem, struct ib_umem_odp, umem);
 89}
 90
 91/* Returns the first page of an ODP umem. */
 92static inline unsigned long ib_umem_start(struct ib_umem_odp *umem_odp)
 93{
 94	return umem_odp->interval_tree.start;
 95}
 96
 97/* Returns the address of the page after the last one of an ODP umem. */
 98static inline unsigned long ib_umem_end(struct ib_umem_odp *umem_odp)
 99{
100	return umem_odp->interval_tree.last + 1;
101}
102
103static inline size_t ib_umem_odp_num_pages(struct ib_umem_odp *umem_odp)
104{
105	return (ib_umem_end(umem_odp) - ib_umem_start(umem_odp)) >>
106	       umem_odp->page_shift;
107}
108
109/*
110 * The lower 2 bits of the DMA address signal the R/W permissions for
111 * the entry. To upgrade the permissions, provide the appropriate
112 * bitmask to the map_dma_pages function.
113 *
114 * Be aware that upgrading a mapped address might result in change of
115 * the DMA address for the page.
116 */
117#define ODP_READ_ALLOWED_BIT  (1<<0ULL)
118#define ODP_WRITE_ALLOWED_BIT (1<<1ULL)
119
120#define ODP_DMA_ADDR_MASK (~(ODP_READ_ALLOWED_BIT | ODP_WRITE_ALLOWED_BIT))
121
122#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
123
124struct ib_ucontext_per_mm {
125	struct mmu_notifier mn;
126	struct pid *tgid;
127
128	struct rb_root_cached umem_tree;
129	/* Protects umem_tree */
130	struct rw_semaphore umem_rwsem;
131};
132
133struct ib_umem_odp *ib_umem_odp_get(struct ib_udata *udata, unsigned long addr,
134				    size_t size, int access);
135struct ib_umem_odp *ib_umem_odp_alloc_implicit(struct ib_udata *udata,
136					       int access);
137struct ib_umem_odp *ib_umem_odp_alloc_child(struct ib_umem_odp *root_umem,
138					    unsigned long addr, size_t size);
139void ib_umem_odp_release(struct ib_umem_odp *umem_odp);
140
141int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 start_offset,
142			      u64 bcnt, u64 access_mask,
143			      unsigned long current_seq);
144
145void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 start_offset,
146				 u64 bound);
147
148typedef int (*umem_call_back)(struct ib_umem_odp *item, u64 start, u64 end,
149			      void *cookie);
150/*
151 * Call the callback on each ib_umem in the range. Returns the logical or of
152 * the return values of the functions called.
153 */
154int rbt_ib_umem_for_each_in_range(struct rb_root_cached *root,
155				  u64 start, u64 end,
156				  umem_call_back cb,
157				  bool blockable, void *cookie);
158
159/*
160 * Find first region intersecting with address range.
161 * Return NULL if not found
162 */
163static inline struct ib_umem_odp *
164rbt_ib_umem_lookup(struct rb_root_cached *root, u64 addr, u64 length)
165{
166	struct interval_tree_node *node;
167
168	node = interval_tree_iter_first(root, addr, addr + length - 1);
169	if (!node)
170		return NULL;
171	return container_of(node, struct ib_umem_odp, interval_tree);
172
173}
174
175static inline int ib_umem_mmu_notifier_retry(struct ib_umem_odp *umem_odp,
176					     unsigned long mmu_seq)
177{
178	/*
179	 * This code is strongly based on the KVM code from
180	 * mmu_notifier_retry. Should be called with
181	 * the relevant locks taken (umem_odp->umem_mutex
182	 * and the ucontext umem_mutex semaphore locked for read).
183	 */
184
185	if (unlikely(umem_odp->notifiers_count))
186		return 1;
187	if (umem_odp->notifiers_seq != mmu_seq)
188		return 1;
189	return 0;
190}
191
192#else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
193
194static inline struct ib_umem_odp *ib_umem_odp_get(struct ib_udata *udata,
195						  unsigned long addr,
196						  size_t size, int access)
197{
198	return ERR_PTR(-EINVAL);
199}
200
201static inline void ib_umem_odp_release(struct ib_umem_odp *umem_odp) {}
202
203#endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
204
205#endif /* IB_UMEM_ODP_H */