1 /*
2 * Copyright (c) 2004 Topspin Communications. 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 #include <linux/errno.h>
34 #include <linux/slab.h>
35
36 #include "mthca_dev.h"
37
38 /* Trivial bitmap-based allocator */
mthca_alloc(struct mthca_alloc * alloc)39 u32 mthca_alloc(struct mthca_alloc *alloc)
40 {
41 unsigned long flags;
42 u32 obj;
43
44 spin_lock_irqsave(&alloc->lock, flags);
45
46 obj = find_next_zero_bit(alloc->table, alloc->max, alloc->last);
47 if (obj >= alloc->max) {
48 alloc->top = (alloc->top + alloc->max) & alloc->mask;
49 obj = find_first_zero_bit(alloc->table, alloc->max);
50 }
51
52 if (obj < alloc->max) {
53 set_bit(obj, alloc->table);
54 obj |= alloc->top;
55 } else
56 obj = -1;
57
58 spin_unlock_irqrestore(&alloc->lock, flags);
59
60 return obj;
61 }
62
mthca_free(struct mthca_alloc * alloc,u32 obj)63 void mthca_free(struct mthca_alloc *alloc, u32 obj)
64 {
65 unsigned long flags;
66
67 obj &= alloc->max - 1;
68
69 spin_lock_irqsave(&alloc->lock, flags);
70
71 clear_bit(obj, alloc->table);
72 alloc->last = min(alloc->last, obj);
73 alloc->top = (alloc->top + alloc->max) & alloc->mask;
74
75 spin_unlock_irqrestore(&alloc->lock, flags);
76 }
77
mthca_alloc_init(struct mthca_alloc * alloc,u32 num,u32 mask,u32 reserved)78 int mthca_alloc_init(struct mthca_alloc *alloc, u32 num, u32 mask,
79 u32 reserved)
80 {
81 int i;
82
83 /* num must be a power of 2 */
84 if (num != 1 << (ffs(num) - 1))
85 return -EINVAL;
86
87 alloc->last = 0;
88 alloc->top = 0;
89 alloc->max = num;
90 alloc->mask = mask;
91 spin_lock_init(&alloc->lock);
92 alloc->table = kmalloc(BITS_TO_LONGS(num) * sizeof (long),
93 GFP_KERNEL);
94 if (!alloc->table)
95 return -ENOMEM;
96
97 bitmap_zero(alloc->table, num);
98 for (i = 0; i < reserved; ++i)
99 set_bit(i, alloc->table);
100
101 return 0;
102 }
103
mthca_alloc_cleanup(struct mthca_alloc * alloc)104 void mthca_alloc_cleanup(struct mthca_alloc *alloc)
105 {
106 kfree(alloc->table);
107 }
108
109 /*
110 * Array of pointers with lazy allocation of leaf pages. Callers of
111 * _get, _set and _clear methods must use a lock or otherwise
112 * serialize access to the array.
113 */
114
115 #define MTHCA_ARRAY_MASK (PAGE_SIZE / sizeof (void *) - 1)
116
mthca_array_get(struct mthca_array * array,int index)117 void *mthca_array_get(struct mthca_array *array, int index)
118 {
119 int p = (index * sizeof (void *)) >> PAGE_SHIFT;
120
121 if (array->page_list[p].page)
122 return array->page_list[p].page[index & MTHCA_ARRAY_MASK];
123 else
124 return NULL;
125 }
126
mthca_array_set(struct mthca_array * array,int index,void * value)127 int mthca_array_set(struct mthca_array *array, int index, void *value)
128 {
129 int p = (index * sizeof (void *)) >> PAGE_SHIFT;
130
131 /* Allocate with GFP_ATOMIC because we'll be called with locks held. */
132 if (!array->page_list[p].page)
133 array->page_list[p].page = (void **) get_zeroed_page(GFP_ATOMIC);
134
135 if (!array->page_list[p].page)
136 return -ENOMEM;
137
138 array->page_list[p].page[index & MTHCA_ARRAY_MASK] = value;
139 ++array->page_list[p].used;
140
141 return 0;
142 }
143
mthca_array_clear(struct mthca_array * array,int index)144 void mthca_array_clear(struct mthca_array *array, int index)
145 {
146 int p = (index * sizeof (void *)) >> PAGE_SHIFT;
147
148 if (--array->page_list[p].used == 0) {
149 free_page((unsigned long) array->page_list[p].page);
150 array->page_list[p].page = NULL;
151 } else
152 array->page_list[p].page[index & MTHCA_ARRAY_MASK] = NULL;
153
154 if (array->page_list[p].used < 0)
155 pr_debug("Array %p index %d page %d with ref count %d < 0\n",
156 array, index, p, array->page_list[p].used);
157 }
158
mthca_array_init(struct mthca_array * array,int nent)159 int mthca_array_init(struct mthca_array *array, int nent)
160 {
161 int npage = (nent * sizeof (void *) + PAGE_SIZE - 1) / PAGE_SIZE;
162 int i;
163
164 array->page_list = kmalloc(npage * sizeof *array->page_list, GFP_KERNEL);
165 if (!array->page_list)
166 return -ENOMEM;
167
168 for (i = 0; i < npage; ++i) {
169 array->page_list[i].page = NULL;
170 array->page_list[i].used = 0;
171 }
172
173 return 0;
174 }
175
mthca_array_cleanup(struct mthca_array * array,int nent)176 void mthca_array_cleanup(struct mthca_array *array, int nent)
177 {
178 int i;
179
180 for (i = 0; i < (nent * sizeof (void *) + PAGE_SIZE - 1) / PAGE_SIZE; ++i)
181 free_page((unsigned long) array->page_list[i].page);
182
183 kfree(array->page_list);
184 }
185
186 /*
187 * Handling for queue buffers -- we allocate a bunch of memory and
188 * register it in a memory region at HCA virtual address 0. If the
189 * requested size is > max_direct, we split the allocation into
190 * multiple pages, so we don't require too much contiguous memory.
191 */
192
mthca_buf_alloc(struct mthca_dev * dev,int size,int max_direct,union mthca_buf * buf,int * is_direct,struct mthca_pd * pd,int hca_write,struct mthca_mr * mr)193 int mthca_buf_alloc(struct mthca_dev *dev, int size, int max_direct,
194 union mthca_buf *buf, int *is_direct, struct mthca_pd *pd,
195 int hca_write, struct mthca_mr *mr)
196 {
197 int err = -ENOMEM;
198 int npages, shift;
199 u64 *dma_list = NULL;
200 dma_addr_t t;
201 int i;
202
203 if (size <= max_direct) {
204 *is_direct = 1;
205 npages = 1;
206 shift = get_order(size) + PAGE_SHIFT;
207
208 buf->direct.buf = dma_alloc_coherent(&dev->pdev->dev,
209 size, &t, GFP_KERNEL);
210 if (!buf->direct.buf)
211 return -ENOMEM;
212
213 pci_unmap_addr_set(&buf->direct, mapping, t);
214
215 memset(buf->direct.buf, 0, size);
216
217 while (t & ((1 << shift) - 1)) {
218 --shift;
219 npages *= 2;
220 }
221
222 dma_list = kmalloc(npages * sizeof *dma_list, GFP_KERNEL);
223 if (!dma_list)
224 goto err_free;
225
226 for (i = 0; i < npages; ++i)
227 dma_list[i] = t + i * (1 << shift);
228 } else {
229 *is_direct = 0;
230 npages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
231 shift = PAGE_SHIFT;
232
233 dma_list = kmalloc(npages * sizeof *dma_list, GFP_KERNEL);
234 if (!dma_list)
235 return -ENOMEM;
236
237 buf->page_list = kmalloc(npages * sizeof *buf->page_list,
238 GFP_KERNEL);
239 if (!buf->page_list)
240 goto err_out;
241
242 for (i = 0; i < npages; ++i)
243 buf->page_list[i].buf = NULL;
244
245 for (i = 0; i < npages; ++i) {
246 buf->page_list[i].buf =
247 dma_alloc_coherent(&dev->pdev->dev, PAGE_SIZE,
248 &t, GFP_KERNEL);
249 if (!buf->page_list[i].buf)
250 goto err_free;
251
252 dma_list[i] = t;
253 pci_unmap_addr_set(&buf->page_list[i], mapping, t);
254
255 clear_page(buf->page_list[i].buf);
256 }
257 }
258
259 err = mthca_mr_alloc_phys(dev, pd->pd_num,
260 dma_list, shift, npages,
261 0, size,
262 MTHCA_MPT_FLAG_LOCAL_READ |
263 (hca_write ? MTHCA_MPT_FLAG_LOCAL_WRITE : 0),
264 mr);
265 if (err)
266 goto err_free;
267
268 kfree(dma_list);
269
270 return 0;
271
272 err_free:
273 mthca_buf_free(dev, size, buf, *is_direct, NULL);
274
275 err_out:
276 kfree(dma_list);
277
278 return err;
279 }
280
mthca_buf_free(struct mthca_dev * dev,int size,union mthca_buf * buf,int is_direct,struct mthca_mr * mr)281 void mthca_buf_free(struct mthca_dev *dev, int size, union mthca_buf *buf,
282 int is_direct, struct mthca_mr *mr)
283 {
284 int i;
285
286 if (mr)
287 mthca_free_mr(dev, mr);
288
289 if (is_direct)
290 dma_free_coherent(&dev->pdev->dev, size, buf->direct.buf,
291 pci_unmap_addr(&buf->direct, mapping));
292 else {
293 for (i = 0; i < (size + PAGE_SIZE - 1) / PAGE_SIZE; ++i)
294 dma_free_coherent(&dev->pdev->dev, PAGE_SIZE,
295 buf->page_list[i].buf,
296 pci_unmap_addr(&buf->page_list[i],
297 mapping));
298 kfree(buf->page_list);
299 }
300 }
301