ib_umem_odp.h revision 331769 
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 <linux/rbtree.h>
37
38#include <rdma/ib_umem.h>
39#include <rdma/ib_verbs.h>
40
41struct umem_odp_node {
42	u64 __subtree_last;
43	struct rb_node rb;
44};
45
46struct ib_umem_odp {
47	/*
48	 * An array of the pages included in the on-demand paging umem.
49	 * Indices of pages that are currently not mapped into the device will
50	 * contain NULL.
51	 */
52	struct page		**page_list;
53	/*
54	 * An array of the same size as page_list, with DMA addresses mapped
55	 * for pages the pages in page_list. The lower two bits designate
56	 * access permissions. See ODP_READ_ALLOWED_BIT and
57	 * ODP_WRITE_ALLOWED_BIT.
58	 */
59	dma_addr_t		*dma_list;
60	/*
61	 * The umem_mutex protects the page_list and dma_list fields of an ODP
62	 * umem, allowing only a single thread to map/unmap pages. The mutex
63	 * also protects access to the mmu notifier counters.
64	 */
65	struct mutex		umem_mutex;
66	void			*private; /* for the HW driver to use. */
67
68	/* When false, use the notifier counter in the ucontext struct. */
69	bool mn_counters_active;
70	int notifiers_seq;
71	int notifiers_count;
72
73	/* A linked list of umems that don't have private mmu notifier
74	 * counters yet. */
75	struct list_head no_private_counters;
76	struct ib_umem		*umem;
77
78	/* Tree tracking */
79	struct umem_odp_node	interval_tree;
80
81	struct completion	notifier_completion;
82	int			dying;
83};
84
85#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
86
87int ib_umem_odp_get(struct ib_ucontext *context, struct ib_umem *umem);
88
89void ib_umem_odp_release(struct ib_umem *umem);
90
91/*
92 * The lower 2 bits of the DMA address signal the R/W permissions for
93 * the entry. To upgrade the permissions, provide the appropriate
94 * bitmask to the map_dma_pages function.
95 *
96 * Be aware that upgrading a mapped address might result in change of
97 * the DMA address for the page.
98 */
99#define ODP_READ_ALLOWED_BIT  (1<<0ULL)
100#define ODP_WRITE_ALLOWED_BIT (1<<1ULL)
101
102#define ODP_DMA_ADDR_MASK (~(ODP_READ_ALLOWED_BIT | ODP_WRITE_ALLOWED_BIT))
103
104int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 start_offset, u64 bcnt,
105			      u64 access_mask, unsigned long current_seq);
106
107void ib_umem_odp_unmap_dma_pages(struct ib_umem *umem, u64 start_offset,
108				 u64 bound);
109
110void rbt_ib_umem_insert(struct umem_odp_node *node, struct rb_root *root);
111void rbt_ib_umem_remove(struct umem_odp_node *node, struct rb_root *root);
112typedef int (*umem_call_back)(struct ib_umem *item, u64 start, u64 end,
113			      void *cookie);
114/*
115 * Call the callback on each ib_umem in the range. Returns the logical or of
116 * the return values of the functions called.
117 */
118int rbt_ib_umem_for_each_in_range(struct rb_root *root, u64 start, u64 end,
119				  umem_call_back cb, void *cookie);
120
121struct umem_odp_node *rbt_ib_umem_iter_first(struct rb_root *root,
122					     u64 start, u64 last);
123struct umem_odp_node *rbt_ib_umem_iter_next(struct umem_odp_node *node,
124					    u64 start, u64 last);
125
126static inline int ib_umem_mmu_notifier_retry(struct ib_umem *item,
127					     unsigned long mmu_seq)
128{
129	/*
130	 * This code is strongly based on the KVM code from
131	 * mmu_notifier_retry. Should be called with
132	 * the relevant locks taken (item->odp_data->umem_mutex
133	 * and the ucontext umem_mutex semaphore locked for read).
134	 */
135
136	/* Do not allow page faults while the new ib_umem hasn't seen a state
137	 * with zero notifiers yet, and doesn't have its own valid set of
138	 * private counters. */
139	if (!item->odp_data->mn_counters_active)
140		return 1;
141
142	if (unlikely(item->odp_data->notifiers_count))
143		return 1;
144	if (item->odp_data->notifiers_seq != mmu_seq)
145		return 1;
146	return 0;
147}
148
149#else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
150
151static inline int ib_umem_odp_get(struct ib_ucontext *context,
152				  struct ib_umem *umem)
153{
154	return -EINVAL;
155}
156
157static inline void ib_umem_odp_release(struct ib_umem *umem) {}
158
159#endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
160
161#endif /* IB_UMEM_ODP_H */
162