2 * Copyright (c) 2004 Topspin Communications. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
33 #include <linux/errno.h>
34 #include <linux/slab.h>
35 #include <linux/bitmap.h>
37 #include "mthca_dev.h"
39 /* Trivial bitmap-based allocator */
40 u32
mthca_alloc(struct mthca_alloc
*alloc
)
45 spin_lock_irqsave(&alloc
->lock
, flags
);
47 obj
= find_next_zero_bit(alloc
->table
, alloc
->max
, alloc
->last
);
48 if (obj
>= alloc
->max
) {
49 alloc
->top
= (alloc
->top
+ alloc
->max
) & alloc
->mask
;
50 obj
= find_first_zero_bit(alloc
->table
, alloc
->max
);
53 if (obj
< alloc
->max
) {
54 set_bit(obj
, alloc
->table
);
59 spin_unlock_irqrestore(&alloc
->lock
, flags
);
64 void mthca_free(struct mthca_alloc
*alloc
, u32 obj
)
68 obj
&= alloc
->max
- 1;
70 spin_lock_irqsave(&alloc
->lock
, flags
);
72 clear_bit(obj
, alloc
->table
);
73 alloc
->last
= min(alloc
->last
, obj
);
74 alloc
->top
= (alloc
->top
+ alloc
->max
) & alloc
->mask
;
76 spin_unlock_irqrestore(&alloc
->lock
, flags
);
79 int mthca_alloc_init(struct mthca_alloc
*alloc
, u32 num
, u32 mask
,
84 /* num must be a power of 2 */
85 if (num
!= 1 << (ffs(num
) - 1))
92 spin_lock_init(&alloc
->lock
);
93 alloc
->table
= kmalloc(BITS_TO_LONGS(num
) * sizeof (long),
98 bitmap_zero(alloc
->table
, num
);
99 for (i
= 0; i
< reserved
; ++i
)
100 set_bit(i
, alloc
->table
);
105 void mthca_alloc_cleanup(struct mthca_alloc
*alloc
)
111 * Array of pointers with lazy allocation of leaf pages. Callers of
112 * _get, _set and _clear methods must use a lock or otherwise
113 * serialize access to the array.
116 #define MTHCA_ARRAY_MASK (PAGE_SIZE / sizeof (void *) - 1)
118 void *mthca_array_get(struct mthca_array
*array
, int index
)
120 int p
= (index
* sizeof (void *)) >> PAGE_SHIFT
;
122 if (array
->page_list
[p
].page
)
123 return array
->page_list
[p
].page
[index
& MTHCA_ARRAY_MASK
];
128 int mthca_array_set(struct mthca_array
*array
, int index
, void *value
)
130 int p
= (index
* sizeof (void *)) >> PAGE_SHIFT
;
132 /* Allocate with GFP_ATOMIC because we'll be called with locks held. */
133 if (!array
->page_list
[p
].page
)
134 array
->page_list
[p
].page
= (void **) get_zeroed_page(GFP_ATOMIC
);
136 if (!array
->page_list
[p
].page
)
139 array
->page_list
[p
].page
[index
& MTHCA_ARRAY_MASK
] = value
;
140 ++array
->page_list
[p
].used
;
145 void mthca_array_clear(struct mthca_array
*array
, int index
)
147 int p
= (index
* sizeof (void *)) >> PAGE_SHIFT
;
149 if (--array
->page_list
[p
].used
== 0) {
150 free_page((unsigned long) array
->page_list
[p
].page
);
151 array
->page_list
[p
].page
= NULL
;
153 array
->page_list
[p
].page
[index
& MTHCA_ARRAY_MASK
] = NULL
;
155 if (array
->page_list
[p
].used
< 0)
156 pr_debug("Array %p index %d page %d with ref count %d < 0\n",
157 array
, index
, p
, array
->page_list
[p
].used
);
160 int mthca_array_init(struct mthca_array
*array
, int nent
)
162 int npage
= (nent
* sizeof (void *) + PAGE_SIZE
- 1) / PAGE_SIZE
;
165 array
->page_list
= kmalloc(npage
* sizeof *array
->page_list
, GFP_KERNEL
);
166 if (!array
->page_list
)
169 for (i
= 0; i
< npage
; ++i
) {
170 array
->page_list
[i
].page
= NULL
;
171 array
->page_list
[i
].used
= 0;
177 void mthca_array_cleanup(struct mthca_array
*array
, int nent
)
181 for (i
= 0; i
< (nent
* sizeof (void *) + PAGE_SIZE
- 1) / PAGE_SIZE
; ++i
)
182 free_page((unsigned long) array
->page_list
[i
].page
);
184 kfree(array
->page_list
);
188 * Handling for queue buffers -- we allocate a bunch of memory and
189 * register it in a memory region at HCA virtual address 0. If the
190 * requested size is > max_direct, we split the allocation into
191 * multiple pages, so we don't require too much contiguous memory.
194 int mthca_buf_alloc(struct mthca_dev
*dev
, int size
, int max_direct
,
195 union mthca_buf
*buf
, int *is_direct
, struct mthca_pd
*pd
,
196 int hca_write
, struct mthca_mr
*mr
)
200 u64
*dma_list
= NULL
;
204 if (size
<= max_direct
) {
207 shift
= get_order(size
) + PAGE_SHIFT
;
209 buf
->direct
.buf
= dma_alloc_coherent(&dev
->pdev
->dev
,
210 size
, &t
, GFP_KERNEL
);
211 if (!buf
->direct
.buf
)
214 pci_unmap_addr_set(&buf
->direct
, mapping
, t
);
216 memset(buf
->direct
.buf
, 0, size
);
218 while (t
& ((1 << shift
) - 1)) {
223 dma_list
= kmalloc(npages
* sizeof *dma_list
, GFP_KERNEL
);
227 for (i
= 0; i
< npages
; ++i
)
228 dma_list
[i
] = t
+ i
* (1 << shift
);
231 npages
= (size
+ PAGE_SIZE
- 1) / PAGE_SIZE
;
234 dma_list
= kmalloc(npages
* sizeof *dma_list
, GFP_KERNEL
);
238 buf
->page_list
= kmalloc(npages
* sizeof *buf
->page_list
,
243 for (i
= 0; i
< npages
; ++i
)
244 buf
->page_list
[i
].buf
= NULL
;
246 for (i
= 0; i
< npages
; ++i
) {
247 buf
->page_list
[i
].buf
=
248 dma_alloc_coherent(&dev
->pdev
->dev
, PAGE_SIZE
,
250 if (!buf
->page_list
[i
].buf
)
254 pci_unmap_addr_set(&buf
->page_list
[i
], mapping
, t
);
256 clear_page(buf
->page_list
[i
].buf
);
260 err
= mthca_mr_alloc_phys(dev
, pd
->pd_num
,
261 dma_list
, shift
, npages
,
263 MTHCA_MPT_FLAG_LOCAL_READ
|
264 (hca_write
? MTHCA_MPT_FLAG_LOCAL_WRITE
: 0),
274 mthca_buf_free(dev
, size
, buf
, *is_direct
, NULL
);
282 void mthca_buf_free(struct mthca_dev
*dev
, int size
, union mthca_buf
*buf
,
283 int is_direct
, struct mthca_mr
*mr
)
288 mthca_free_mr(dev
, mr
);
291 dma_free_coherent(&dev
->pdev
->dev
, size
, buf
->direct
.buf
,
292 pci_unmap_addr(&buf
->direct
, mapping
));
294 for (i
= 0; i
< (size
+ PAGE_SIZE
- 1) / PAGE_SIZE
; ++i
)
295 dma_free_coherent(&dev
->pdev
->dev
, PAGE_SIZE
,
296 buf
->page_list
[i
].buf
,
297 pci_unmap_addr(&buf
->page_list
[i
],
299 kfree(buf
->page_list
);