1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
|
/*
* IOMMU implementation for Cell Broadband Processor Architecture
* We just establish a linear mapping at boot by setting all the
* IOPT cache entries in the CPU.
* The mapping functions should be identical to pci_direct_iommu,
* except for the handling of the high order bit that is required
* by the Spider bridge. These should be split into a separate
* file at the point where we get a different bridge chip.
*
* Copyright (C) 2005 IBM Deutschland Entwicklung GmbH,
* Arnd Bergmann <arndb@de.ibm.com>
*
* Based on linear mapping
* Copyright (C) 2003 Benjamin Herrenschmidt (benh@kernel.crashing.org)
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#undef DEBUG
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/bootmem.h>
#include <linux/mm.h>
#include <linux/dma-mapping.h>
#include <asm/sections.h>
#include <asm/iommu.h>
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/pci-bridge.h>
#include <asm/machdep.h>
#include <asm/pmac_feature.h>
#include <asm/abs_addr.h>
#include <asm/system.h>
#include <asm/ppc-pci.h>
#include "iommu.h"
static inline unsigned long
get_iopt_entry(unsigned long real_address, unsigned long ioid,
unsigned long prot)
{
return (prot & IOPT_PROT_MASK)
| (IOPT_COHERENT)
| (IOPT_ORDER_VC)
| (real_address & IOPT_RPN_MASK)
| (ioid & IOPT_IOID_MASK);
}
typedef struct {
unsigned long val;
} ioste;
static inline ioste
mk_ioste(unsigned long val)
{
ioste ioste = { .val = val, };
return ioste;
}
static inline ioste
get_iost_entry(unsigned long iopt_base, unsigned long io_address, unsigned page_size)
{
unsigned long ps;
unsigned long iostep;
unsigned long nnpt;
unsigned long shift;
switch (page_size) {
case 0x1000000:
ps = IOST_PS_16M;
nnpt = 0; /* one page per segment */
shift = 5; /* segment has 16 iopt entries */
break;
case 0x100000:
ps = IOST_PS_1M;
nnpt = 0; /* one page per segment */
shift = 1; /* segment has 256 iopt entries */
break;
case 0x10000:
ps = IOST_PS_64K;
nnpt = 0x07; /* 8 pages per io page table */
shift = 0; /* all entries are used */
break;
case 0x1000:
ps = IOST_PS_4K;
nnpt = 0x7f; /* 128 pages per io page table */
shift = 0; /* all entries are used */
break;
default: /* not a known compile time constant */
{
/* BUILD_BUG_ON() is not usable here */
extern void __get_iost_entry_bad_page_size(void);
__get_iost_entry_bad_page_size();
}
break;
}
iostep = iopt_base +
/* need 8 bytes per iopte */
(((io_address / page_size * 8)
/* align io page tables on 4k page boundaries */
<< shift)
/* nnpt+1 pages go into each iopt */
& ~(nnpt << 12));
nnpt++; /* this seems to work, but the documentation is not clear
about wether we put nnpt or nnpt-1 into the ioste bits.
In theory, this can't work for 4k pages. */
return mk_ioste(IOST_VALID_MASK
| (iostep & IOST_PT_BASE_MASK)
| ((nnpt << 5) & IOST_NNPT_MASK)
| (ps & IOST_PS_MASK));
}
/* compute the address of an io pte */
static inline unsigned long
get_ioptep(ioste iost_entry, unsigned long io_address)
{
unsigned long iopt_base;
unsigned long page_size;
unsigned long page_number;
unsigned long iopt_offset;
iopt_base = iost_entry.val & IOST_PT_BASE_MASK;
page_size = iost_entry.val & IOST_PS_MASK;
/* decode page size to compute page number */
page_number = (io_address & 0x0fffffff) >> (10 + 2 * page_size);
/* page number is an offset into the io page table */
iopt_offset = (page_number << 3) & 0x7fff8ul;
return iopt_base + iopt_offset;
}
/* compute the tag field of the iopt cache entry */
static inline unsigned long
get_ioc_tag(ioste iost_entry, unsigned long io_address)
{
unsigned long iopte = get_ioptep(iost_entry, io_address);
return IOPT_VALID_MASK
| ((iopte & 0x00000000000000ff8ul) >> 3)
| ((iopte & 0x0000003fffffc0000ul) >> 9);
}
/* compute the hashed 6 bit index for the 4-way associative pte cache */
static inline unsigned long
get_ioc_hash(ioste iost_entry, unsigned long io_address)
{
unsigned long iopte = get_ioptep(iost_entry, io_address);
return ((iopte & 0x000000000000001f8ul) >> 3)
^ ((iopte & 0x00000000000020000ul) >> 17)
^ ((iopte & 0x00000000000010000ul) >> 15)
^ ((iopte & 0x00000000000008000ul) >> 13)
^ ((iopte & 0x00000000000004000ul) >> 11)
^ ((iopte & 0x00000000000002000ul) >> 9)
^ ((iopte & 0x00000000000001000ul) >> 7);
}
/* same as above, but pretend that we have a simpler 1-way associative
pte cache with an 8 bit index */
static inline unsigned long
get_ioc_hash_1way(ioste iost_entry, unsigned long io_address)
{
unsigned long iopte = get_ioptep(iost_entry, io_address);
return ((iopte & 0x000000000000001f8ul) >> 3)
^ ((iopte & 0x00000000000020000ul) >> 17)
^ ((iopte & 0x00000000000010000ul) >> 15)
^ ((iopte & 0x00000000000008000ul) >> 13)
^ ((iopte & 0x00000000000004000ul) >> 11)
^ ((iopte & 0x00000000000002000ul) >> 9)
^ ((iopte & 0x00000000000001000ul) >> 7)
^ ((iopte & 0x0000000000000c000ul) >> 8);
}
static inline ioste
get_iost_cache(void __iomem *base, unsigned long index)
{
unsigned long __iomem *p = (base + IOC_ST_CACHE_DIR);
return mk_ioste(in_be64(&p[index]));
}
static inline void
set_iost_cache(void __iomem *base, unsigned long index, ioste ste)
{
unsigned long __iomem *p = (base + IOC_ST_CACHE_DIR);
pr_debug("ioste %02lx was %016lx, store %016lx", index,
get_iost_cache(base, index).val, ste.val);
out_be64(&p[index], ste.val);
pr_debug(" now %016lx\n", get_iost_cache(base, index).val);
}
static inline unsigned long
get_iopt_cache(void __iomem *base, unsigned long index, unsigned long *tag)
{
unsigned long __iomem *tags = (void *)(base + IOC_PT_CACHE_DIR);
unsigned long __iomem *p = (void *)(base + IOC_PT_CACHE_REG);
*tag = tags[index];
rmb();
return *p;
}
static inline void
set_iopt_cache(void __iomem *base, unsigned long index,
unsigned long tag, unsigned long val)
{
unsigned long __iomem *tags = base + IOC_PT_CACHE_DIR;
unsigned long __iomem *p = base + IOC_PT_CACHE_REG;
pr_debug("iopt %02lx was v%016lx/t%016lx, store v%016lx/t%016lx\n",
index, get_iopt_cache(base, index, &oldtag), oldtag, val, tag);
out_be64(p, val);
out_be64(&tags[index], tag);
}
static inline void
set_iost_origin(void __iomem *base)
{
unsigned long __iomem *p = base + IOC_ST_ORIGIN;
unsigned long origin = IOSTO_ENABLE | IOSTO_SW;
pr_debug("iost_origin %016lx, now %016lx\n", in_be64(p), origin);
out_be64(p, origin);
}
static inline void
set_iocmd_config(void __iomem *base)
{
unsigned long __iomem *p = base + 0xc00;
unsigned long conf;
conf = in_be64(p);
pr_debug("iost_conf %016lx, now %016lx\n", conf, conf | IOCMD_CONF_TE);
out_be64(p, conf | IOCMD_CONF_TE);
}
/* FIXME: get these from the device tree */
#define ioc_base 0x20000511000ull
#define ioc_mmio_base 0x20000510000ull
#define ioid 0x48a
#define iopt_phys_offset (- 0x20000000) /* We have a 512MB offset from the SB */
#define io_page_size 0x1000000
static unsigned long map_iopt_entry(unsigned long address)
{
switch (address >> 20) {
case 0x600:
address = 0x24020000000ull; /* spider i/o */
break;
default:
address += iopt_phys_offset;
break;
}
return get_iopt_entry(address, ioid, IOPT_PROT_RW);
}
static void iommu_bus_setup_null(struct pci_bus *b) { }
static void iommu_dev_setup_null(struct pci_dev *d) { }
/* initialize the iommu to support a simple linear mapping
* for each DMA window used by any device. For now, we
* happen to know that there is only one DMA window in use,
* starting at iopt_phys_offset. */
static void cell_map_iommu(void)
{
unsigned long address;
void __iomem *base;
ioste ioste;
unsigned long index;
base = __ioremap(ioc_base, 0x1000, _PAGE_NO_CACHE);
pr_debug("%lx mapped to %p\n", ioc_base, base);
set_iocmd_config(base);
iounmap(base);
base = __ioremap(ioc_mmio_base, 0x1000, _PAGE_NO_CACHE);
pr_debug("%lx mapped to %p\n", ioc_mmio_base, base);
set_iost_origin(base);
for (address = 0; address < 0x100000000ul; address += io_page_size) {
ioste = get_iost_entry(0x10000000000ul, address, io_page_size);
if ((address & 0xfffffff) == 0) /* segment start */
set_iost_cache(base, address >> 28, ioste);
index = get_ioc_hash_1way(ioste, address);
pr_debug("addr %08lx, index %02lx, ioste %016lx\n",
address, index, ioste.val);
set_iopt_cache(base,
get_ioc_hash_1way(ioste, address),
get_ioc_tag(ioste, address),
map_iopt_entry(address));
}
iounmap(base);
}
static void *cell_alloc_coherent(struct device *hwdev, size_t size,
dma_addr_t *dma_handle, gfp_t flag)
{
void *ret;
ret = (void *)__get_free_pages(flag, get_order(size));
if (ret != NULL) {
memset(ret, 0, size);
*dma_handle = virt_to_abs(ret) | CELL_DMA_VALID;
}
return ret;
}
static void cell_free_coherent(struct device *hwdev, size_t size,
void *vaddr, dma_addr_t dma_handle)
{
free_pages((unsigned long)vaddr, get_order(size));
}
static dma_addr_t cell_map_single(struct device *hwdev, void *ptr,
size_t size, enum dma_data_direction direction)
{
return virt_to_abs(ptr) | CELL_DMA_VALID;
}
static void cell_unmap_single(struct device *hwdev, dma_addr_t dma_addr,
size_t size, enum dma_data_direction direction)
{
}
static int cell_map_sg(struct device *hwdev, struct scatterlist *sg,
int nents, enum dma_data_direction direction)
{
int i;
for (i = 0; i < nents; i++, sg++) {
sg->dma_address = (page_to_phys(sg->page) + sg->offset)
| CELL_DMA_VALID;
sg->dma_length = sg->length;
}
return nents;
}
static void cell_unmap_sg(struct device *hwdev, struct scatterlist *sg,
int nents, enum dma_data_direction direction)
{
}
static int cell_dma_supported(struct device *dev, u64 mask)
{
return mask < 0x100000000ull;
}
void cell_init_iommu(void)
{
cell_map_iommu();
/* Direct I/O, IOMMU off */
ppc_md.iommu_dev_setup = iommu_dev_setup_null;
ppc_md.iommu_bus_setup = iommu_bus_setup_null;
pci_dma_ops.alloc_coherent = cell_alloc_coherent;
pci_dma_ops.free_coherent = cell_free_coherent;
pci_dma_ops.map_single = cell_map_single;
pci_dma_ops.unmap_single = cell_unmap_single;
pci_dma_ops.map_sg = cell_map_sg;
pci_dma_ops.unmap_sg = cell_unmap_sg;
pci_dma_ops.dma_supported = cell_dma_supported;
}
|