Blame view
lib/swiotlb.c
24.1 KB
1da177e4c Linux-2.6.12-rc2 |
1 2 3 |
/* * Dynamic DMA mapping support. * |
563aaf064 [IA64] swiotlb cl... |
4 |
* This implementation is a fallback for platforms that do not support |
1da177e4c Linux-2.6.12-rc2 |
5 6 7 8 9 10 11 12 13 |
* I/O TLBs (aka DMA address translation hardware). * Copyright (C) 2000 Asit Mallick <Asit.K.Mallick@intel.com> * Copyright (C) 2000 Goutham Rao <goutham.rao@intel.com> * Copyright (C) 2000, 2003 Hewlett-Packard Co * David Mosberger-Tang <davidm@hpl.hp.com> * * 03/05/07 davidm Switch from PCI-DMA to generic device DMA API. * 00/12/13 davidm Rename to swiotlb.c and add mark_clean() to avoid * unnecessary i-cache flushing. |
569c8bf5d [PATCH] swiotlb: ... |
14 15 16 |
* 04/07/.. ak Better overflow handling. Assorted fixes. * 05/09/10 linville Add support for syncing ranges, support syncing for * DMA_BIDIRECTIONAL mappings, miscellaneous cleanup. |
1da177e4c Linux-2.6.12-rc2 |
17 18 19 |
*/ #include <linux/cache.h> |
17e5ad6c0 [PATCH] Removed r... |
20 |
#include <linux/dma-mapping.h> |
1da177e4c Linux-2.6.12-rc2 |
21 22 |
#include <linux/mm.h> #include <linux/module.h> |
1da177e4c Linux-2.6.12-rc2 |
23 24 25 26 27 28 |
#include <linux/spinlock.h> #include <linux/string.h> #include <linux/types.h> #include <linux/ctype.h> #include <asm/io.h> |
1da177e4c Linux-2.6.12-rc2 |
29 |
#include <asm/dma.h> |
17e5ad6c0 [PATCH] Removed r... |
30 |
#include <asm/scatterlist.h> |
1da177e4c Linux-2.6.12-rc2 |
31 32 33 |
#include <linux/init.h> #include <linux/bootmem.h> |
a85225092 swiotlb: use iomm... |
34 |
#include <linux/iommu-helper.h> |
1da177e4c Linux-2.6.12-rc2 |
35 36 37 |
#define OFFSET(val,align) ((unsigned long) \ ( (val) & ( (align) - 1))) |
f9527f121 Update swiotlb to... |
38 |
#define SG_ENT_VIRT_ADDRESS(sg) (sg_virt((sg))) |
93fbff63e [IA64] make swiot... |
39 |
#define SG_ENT_PHYS_ADDRESS(sg) virt_to_bus(SG_ENT_VIRT_ADDRESS(sg)) |
1da177e4c Linux-2.6.12-rc2 |
40 41 42 43 44 45 46 47 48 49 50 51 52 |
/* * Maximum allowable number of contiguous slabs to map, * must be a power of 2. What is the appropriate value ? * The complexity of {map,unmap}_single is linearly dependent on this value. */ #define IO_TLB_SEGSIZE 128 /* * log of the size of each IO TLB slab. The number of slabs is command line * controllable. */ #define IO_TLB_SHIFT 11 |
0b9afede3 [IA64] more robus... |
53 54 55 56 57 58 59 60 |
#define SLABS_PER_PAGE (1 << (PAGE_SHIFT - IO_TLB_SHIFT)) /* * Minimum IO TLB size to bother booting with. Systems with mainly * 64bit capable cards will only lightly use the swiotlb. If we can't * allocate a contiguous 1MB, we're probably in trouble anyway. */ #define IO_TLB_MIN_SLABS ((1<<20) >> IO_TLB_SHIFT) |
de69e0f0b [PATCH] swiotlb: ... |
61 62 63 64 65 66 67 |
/* * Enumeration for sync targets */ enum dma_sync_target { SYNC_FOR_CPU = 0, SYNC_FOR_DEVICE = 1, }; |
1da177e4c Linux-2.6.12-rc2 |
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 |
int swiotlb_force; /* * Used to do a quick range check in swiotlb_unmap_single and * swiotlb_sync_single_*, to see if the memory was in fact allocated by this * API. */ static char *io_tlb_start, *io_tlb_end; /* * The number of IO TLB blocks (in groups of 64) betweeen io_tlb_start and * io_tlb_end. This is command line adjustable via setup_io_tlb_npages. */ static unsigned long io_tlb_nslabs; /* * When the IOMMU overflows we return a fallback buffer. This sets the size. */ static unsigned long io_tlb_overflow = 32*1024; void *io_tlb_overflow_buffer; /* * This is a free list describing the number of free entries available from * each index */ static unsigned int *io_tlb_list; static unsigned int io_tlb_index; /* * We need to save away the original address corresponding to a mapped entry * for the sync operations. */ |
25667d675 Revert "[IA64] sw... |
101 |
static unsigned char **io_tlb_orig_addr; |
1da177e4c Linux-2.6.12-rc2 |
102 103 104 105 106 107 108 109 110 111 |
/* * Protect the above data structures in the map and unmap calls */ static DEFINE_SPINLOCK(io_tlb_lock); static int __init setup_io_tlb_npages(char *str) { if (isdigit(*str)) { |
e8579e72c [IA64, X86_64] fi... |
112 |
io_tlb_nslabs = simple_strtoul(str, &str, 0); |
1da177e4c Linux-2.6.12-rc2 |
113 114 115 116 117 118 119 120 121 122 123 124 125 126 |
/* avoid tail segment of size < IO_TLB_SEGSIZE */ io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE); } if (*str == ',') ++str; if (!strcmp(str, "force")) swiotlb_force = 1; return 1; } __setup("swiotlb=", setup_io_tlb_npages); /* make io_tlb_overflow tunable too? */ /* * Statically reserve bounce buffer space and initialize bounce buffer data |
17e5ad6c0 [PATCH] Removed r... |
127 |
* structures for the software IO TLB used to implement the DMA API. |
1da177e4c Linux-2.6.12-rc2 |
128 |
*/ |
563aaf064 [IA64] swiotlb cl... |
129 130 |
void __init swiotlb_init_with_default_size(size_t default_size) |
1da177e4c Linux-2.6.12-rc2 |
131 |
{ |
563aaf064 [IA64] swiotlb cl... |
132 |
unsigned long i, bytes; |
1da177e4c Linux-2.6.12-rc2 |
133 134 |
if (!io_tlb_nslabs) { |
e8579e72c [IA64, X86_64] fi... |
135 |
io_tlb_nslabs = (default_size >> IO_TLB_SHIFT); |
1da177e4c Linux-2.6.12-rc2 |
136 137 |
io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE); } |
563aaf064 [IA64] swiotlb cl... |
138 |
bytes = io_tlb_nslabs << IO_TLB_SHIFT; |
1da177e4c Linux-2.6.12-rc2 |
139 140 141 |
/* * Get IO TLB memory from the low pages */ |
563aaf064 [IA64] swiotlb cl... |
142 |
io_tlb_start = alloc_bootmem_low_pages(bytes); |
1da177e4c Linux-2.6.12-rc2 |
143 144 |
if (!io_tlb_start) panic("Cannot allocate SWIOTLB buffer"); |
563aaf064 [IA64] swiotlb cl... |
145 |
io_tlb_end = io_tlb_start + bytes; |
1da177e4c Linux-2.6.12-rc2 |
146 147 148 149 150 151 152 |
/* * Allocate and initialize the free list array. This array is used * to find contiguous free memory regions of size up to IO_TLB_SEGSIZE * between io_tlb_start and io_tlb_end. */ io_tlb_list = alloc_bootmem(io_tlb_nslabs * sizeof(int)); |
25667d675 Revert "[IA64] sw... |
153 |
for (i = 0; i < io_tlb_nslabs; i++) |
1da177e4c Linux-2.6.12-rc2 |
154 155 |
io_tlb_list[i] = IO_TLB_SEGSIZE - OFFSET(i, IO_TLB_SEGSIZE); io_tlb_index = 0; |
25667d675 Revert "[IA64] sw... |
156 |
io_tlb_orig_addr = alloc_bootmem(io_tlb_nslabs * sizeof(char *)); |
1da177e4c Linux-2.6.12-rc2 |
157 158 159 160 161 |
/* * Get the overflow emergency buffer */ io_tlb_overflow_buffer = alloc_bootmem_low(io_tlb_overflow); |
563aaf064 [IA64] swiotlb cl... |
162 163 164 |
if (!io_tlb_overflow_buffer) panic("Cannot allocate SWIOTLB overflow buffer! "); |
25667d675 Revert "[IA64] sw... |
165 166 167 |
printk(KERN_INFO "Placing software IO TLB between 0x%lx - 0x%lx ", virt_to_bus(io_tlb_start), virt_to_bus(io_tlb_end)); |
1da177e4c Linux-2.6.12-rc2 |
168 |
} |
563aaf064 [IA64] swiotlb cl... |
169 170 |
void __init swiotlb_init(void) |
1da177e4c Linux-2.6.12-rc2 |
171 |
{ |
25667d675 Revert "[IA64] sw... |
172 |
swiotlb_init_with_default_size(64 * (1<<20)); /* default to 64MB */ |
1da177e4c Linux-2.6.12-rc2 |
173 |
} |
0b9afede3 [IA64] more robus... |
174 175 176 177 178 179 |
/* * Systems with larger DMA zones (those that don't support ISA) can * initialize the swiotlb later using the slab allocator if needed. * This should be just like above, but with some error catching. */ int |
563aaf064 [IA64] swiotlb cl... |
180 |
swiotlb_late_init_with_default_size(size_t default_size) |
0b9afede3 [IA64] more robus... |
181 |
{ |
563aaf064 [IA64] swiotlb cl... |
182 |
unsigned long i, bytes, req_nslabs = io_tlb_nslabs; |
0b9afede3 [IA64] more robus... |
183 184 185 186 187 188 189 190 191 192 |
unsigned int order; if (!io_tlb_nslabs) { io_tlb_nslabs = (default_size >> IO_TLB_SHIFT); io_tlb_nslabs = ALIGN(io_tlb_nslabs, IO_TLB_SEGSIZE); } /* * Get IO TLB memory from the low pages */ |
563aaf064 [IA64] swiotlb cl... |
193 |
order = get_order(io_tlb_nslabs << IO_TLB_SHIFT); |
0b9afede3 [IA64] more robus... |
194 |
io_tlb_nslabs = SLABS_PER_PAGE << order; |
563aaf064 [IA64] swiotlb cl... |
195 |
bytes = io_tlb_nslabs << IO_TLB_SHIFT; |
0b9afede3 [IA64] more robus... |
196 197 198 199 200 201 202 203 204 205 206 |
while ((SLABS_PER_PAGE << order) > IO_TLB_MIN_SLABS) { io_tlb_start = (char *)__get_free_pages(GFP_DMA | __GFP_NOWARN, order); if (io_tlb_start) break; order--; } if (!io_tlb_start) goto cleanup1; |
563aaf064 [IA64] swiotlb cl... |
207 |
if (order != get_order(bytes)) { |
0b9afede3 [IA64] more robus... |
208 209 210 211 |
printk(KERN_WARNING "Warning: only able to allocate %ld MB " "for software IO TLB ", (PAGE_SIZE << order) >> 20); io_tlb_nslabs = SLABS_PER_PAGE << order; |
563aaf064 [IA64] swiotlb cl... |
212 |
bytes = io_tlb_nslabs << IO_TLB_SHIFT; |
0b9afede3 [IA64] more robus... |
213 |
} |
563aaf064 [IA64] swiotlb cl... |
214 215 |
io_tlb_end = io_tlb_start + bytes; memset(io_tlb_start, 0, bytes); |
0b9afede3 [IA64] more robus... |
216 217 218 219 220 221 222 223 224 225 226 227 228 229 |
/* * Allocate and initialize the free list array. This array is used * to find contiguous free memory regions of size up to IO_TLB_SEGSIZE * between io_tlb_start and io_tlb_end. */ io_tlb_list = (unsigned int *)__get_free_pages(GFP_KERNEL, get_order(io_tlb_nslabs * sizeof(int))); if (!io_tlb_list) goto cleanup2; for (i = 0; i < io_tlb_nslabs; i++) io_tlb_list[i] = IO_TLB_SEGSIZE - OFFSET(i, IO_TLB_SEGSIZE); io_tlb_index = 0; |
25667d675 Revert "[IA64] sw... |
230 231 |
io_tlb_orig_addr = (unsigned char **)__get_free_pages(GFP_KERNEL, get_order(io_tlb_nslabs * sizeof(char *))); |
0b9afede3 [IA64] more robus... |
232 233 |
if (!io_tlb_orig_addr) goto cleanup3; |
25667d675 Revert "[IA64] sw... |
234 |
memset(io_tlb_orig_addr, 0, io_tlb_nslabs * sizeof(char *)); |
0b9afede3 [IA64] more robus... |
235 236 237 238 239 240 241 242 |
/* * Get the overflow emergency buffer */ io_tlb_overflow_buffer = (void *)__get_free_pages(GFP_DMA, get_order(io_tlb_overflow)); if (!io_tlb_overflow_buffer) goto cleanup4; |
25667d675 Revert "[IA64] sw... |
243 244 245 246 |
printk(KERN_INFO "Placing %luMB software IO TLB between 0x%lx - " "0x%lx ", bytes >> 20, virt_to_bus(io_tlb_start), virt_to_bus(io_tlb_end)); |
0b9afede3 [IA64] more robus... |
247 248 249 250 |
return 0; cleanup4: |
25667d675 Revert "[IA64] sw... |
251 252 |
free_pages((unsigned long)io_tlb_orig_addr, get_order(io_tlb_nslabs * sizeof(char *))); |
0b9afede3 [IA64] more robus... |
253 254 |
io_tlb_orig_addr = NULL; cleanup3: |
25667d675 Revert "[IA64] sw... |
255 256 |
free_pages((unsigned long)io_tlb_list, get_order(io_tlb_nslabs * sizeof(int))); |
0b9afede3 [IA64] more robus... |
257 |
io_tlb_list = NULL; |
0b9afede3 [IA64] more robus... |
258 |
cleanup2: |
563aaf064 [IA64] swiotlb cl... |
259 |
io_tlb_end = NULL; |
0b9afede3 [IA64] more robus... |
260 261 262 263 264 265 |
free_pages((unsigned long)io_tlb_start, order); io_tlb_start = NULL; cleanup1: io_tlb_nslabs = req_nslabs; return -ENOMEM; } |
be6b02678 [PATCH] swiotlb u... |
266 |
static int |
1da177e4c Linux-2.6.12-rc2 |
267 268 269 270 271 272 273 274 275 276 277 278 279 |
address_needs_mapping(struct device *hwdev, dma_addr_t addr) { dma_addr_t mask = 0xffffffff; /* If the device has a mask, use it, otherwise default to 32 bits */ if (hwdev && hwdev->dma_mask) mask = *hwdev->dma_mask; return (addr & ~mask) != 0; } /* * Allocates bounce buffer and returns its kernel virtual address. */ static void * |
25667d675 Revert "[IA64] sw... |
280 |
map_single(struct device *hwdev, char *buffer, size_t size, int dir) |
1da177e4c Linux-2.6.12-rc2 |
281 282 283 284 285 |
{ unsigned long flags; char *dma_addr; unsigned int nslots, stride, index, wrap; int i; |
681cc5cd3 iommu sg merging:... |
286 287 288 289 290 291 292 293 294 |
unsigned long start_dma_addr; unsigned long mask; unsigned long offset_slots; unsigned long max_slots; mask = dma_get_seg_boundary(hwdev); start_dma_addr = virt_to_bus(io_tlb_start) & mask; offset_slots = ALIGN(start_dma_addr, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT; |
b15a3891c avoid endless loo... |
295 296 297 |
max_slots = mask + 1 ? ALIGN(mask + 1, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT : 1UL << (BITS_PER_LONG - IO_TLB_SHIFT); |
1da177e4c Linux-2.6.12-rc2 |
298 299 300 301 302 303 304 305 306 307 |
/* * For mappings greater than a page, we limit the stride (and * hence alignment) to a page size. */ nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT; if (size > PAGE_SIZE) stride = (1 << (PAGE_SHIFT - IO_TLB_SHIFT)); else stride = 1; |
348145458 BUG_ON() Conversi... |
308 |
BUG_ON(!nslots); |
1da177e4c Linux-2.6.12-rc2 |
309 310 311 312 313 314 |
/* * Find suitable number of IO TLB entries size that will fit this * request and allocate a buffer from that IO TLB pool. */ spin_lock_irqsave(&io_tlb_lock, flags); |
a7133a155 lib/swiotlb.c: cl... |
315 316 317 318 319 320 |
index = ALIGN(io_tlb_index, stride); if (index >= io_tlb_nslabs) index = 0; wrap = index; do { |
a85225092 swiotlb: use iomm... |
321 322 |
while (iommu_is_span_boundary(index, nslots, offset_slots, max_slots)) { |
b15a3891c avoid endless loo... |
323 324 325 |
index += stride; if (index >= io_tlb_nslabs) index = 0; |
a7133a155 lib/swiotlb.c: cl... |
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 |
if (index == wrap) goto not_found; } /* * If we find a slot that indicates we have 'nslots' number of * contiguous buffers, we allocate the buffers from that slot * and mark the entries as '0' indicating unavailable. */ if (io_tlb_list[index] >= nslots) { int count = 0; for (i = index; i < (int) (index + nslots); i++) io_tlb_list[i] = 0; for (i = index - 1; (OFFSET(i, IO_TLB_SEGSIZE) != IO_TLB_SEGSIZE - 1) && io_tlb_list[i]; i--) io_tlb_list[i] = ++count; dma_addr = io_tlb_start + (index << IO_TLB_SHIFT); |
1da177e4c Linux-2.6.12-rc2 |
343 |
|
a7133a155 lib/swiotlb.c: cl... |
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 |
/* * Update the indices to avoid searching in the next * round. */ io_tlb_index = ((index + nslots) < io_tlb_nslabs ? (index + nslots) : 0); goto found; } index += stride; if (index >= io_tlb_nslabs) index = 0; } while (index != wrap); not_found: spin_unlock_irqrestore(&io_tlb_lock, flags); return NULL; found: |
1da177e4c Linux-2.6.12-rc2 |
362 363 364 365 366 367 368 |
spin_unlock_irqrestore(&io_tlb_lock, flags); /* * Save away the mapping from the original address to the DMA address. * This is needed when we sync the memory. Then we sync the buffer if * needed. */ |
df336d1c7 Fix swiotlb_sync_... |
369 370 |
for (i = 0; i < nslots; i++) io_tlb_orig_addr[index+i] = buffer + (i << IO_TLB_SHIFT); |
1da177e4c Linux-2.6.12-rc2 |
371 |
if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL) |
25667d675 Revert "[IA64] sw... |
372 |
memcpy(dma_addr, buffer, size); |
1da177e4c Linux-2.6.12-rc2 |
373 374 375 376 377 378 379 380 381 382 383 384 385 |
return dma_addr; } /* * dma_addr is the kernel virtual address of the bounce buffer to unmap. */ static void unmap_single(struct device *hwdev, char *dma_addr, size_t size, int dir) { unsigned long flags; int i, count, nslots = ALIGN(size, 1 << IO_TLB_SHIFT) >> IO_TLB_SHIFT; int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT; |
25667d675 Revert "[IA64] sw... |
386 |
char *buffer = io_tlb_orig_addr[index]; |
1da177e4c Linux-2.6.12-rc2 |
387 388 389 390 |
/* * First, sync the memory before unmapping the entry */ |
25667d675 Revert "[IA64] sw... |
391 |
if (buffer && ((dir == DMA_FROM_DEVICE) || (dir == DMA_BIDIRECTIONAL))) |
1da177e4c Linux-2.6.12-rc2 |
392 393 394 395 |
/* * bounce... copy the data back into the original buffer * and * delete the bounce buffer. */ |
25667d675 Revert "[IA64] sw... |
396 |
memcpy(buffer, dma_addr, size); |
1da177e4c Linux-2.6.12-rc2 |
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 |
/* * Return the buffer to the free list by setting the corresponding * entries to indicate the number of contigous entries available. * While returning the entries to the free list, we merge the entries * with slots below and above the pool being returned. */ spin_lock_irqsave(&io_tlb_lock, flags); { count = ((index + nslots) < ALIGN(index + 1, IO_TLB_SEGSIZE) ? io_tlb_list[index + nslots] : 0); /* * Step 1: return the slots to the free list, merging the * slots with superceeding slots */ for (i = index + nslots - 1; i >= index; i--) io_tlb_list[i] = ++count; /* * Step 2: merge the returned slots with the preceding slots, * if available (non zero) */ for (i = index - 1; (OFFSET(i, IO_TLB_SEGSIZE) != IO_TLB_SEGSIZE -1) && io_tlb_list[i]; i--) io_tlb_list[i] = ++count; } spin_unlock_irqrestore(&io_tlb_lock, flags); } static void |
de69e0f0b [PATCH] swiotlb: ... |
425 426 |
sync_single(struct device *hwdev, char *dma_addr, size_t size, int dir, int target) |
1da177e4c Linux-2.6.12-rc2 |
427 428 |
{ int index = (dma_addr - io_tlb_start) >> IO_TLB_SHIFT; |
25667d675 Revert "[IA64] sw... |
429 |
char *buffer = io_tlb_orig_addr[index]; |
1da177e4c Linux-2.6.12-rc2 |
430 |
|
df336d1c7 Fix swiotlb_sync_... |
431 |
buffer += ((unsigned long)dma_addr & ((1 << IO_TLB_SHIFT) - 1)); |
de69e0f0b [PATCH] swiotlb: ... |
432 433 434 |
switch (target) { case SYNC_FOR_CPU: if (likely(dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)) |
25667d675 Revert "[IA64] sw... |
435 |
memcpy(buffer, dma_addr, size); |
348145458 BUG_ON() Conversi... |
436 437 |
else BUG_ON(dir != DMA_TO_DEVICE); |
de69e0f0b [PATCH] swiotlb: ... |
438 439 440 |
break; case SYNC_FOR_DEVICE: if (likely(dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL)) |
25667d675 Revert "[IA64] sw... |
441 |
memcpy(dma_addr, buffer, size); |
348145458 BUG_ON() Conversi... |
442 443 |
else BUG_ON(dir != DMA_FROM_DEVICE); |
de69e0f0b [PATCH] swiotlb: ... |
444 445 |
break; default: |
1da177e4c Linux-2.6.12-rc2 |
446 |
BUG(); |
de69e0f0b [PATCH] swiotlb: ... |
447 |
} |
1da177e4c Linux-2.6.12-rc2 |
448 449 450 451 |
} void * swiotlb_alloc_coherent(struct device *hwdev, size_t size, |
06a544971 [PATCH] gfp_t: dm... |
452 |
dma_addr_t *dma_handle, gfp_t flags) |
1da177e4c Linux-2.6.12-rc2 |
453 |
{ |
563aaf064 [IA64] swiotlb cl... |
454 |
dma_addr_t dev_addr; |
1da177e4c Linux-2.6.12-rc2 |
455 456 457 458 459 460 461 462 463 |
void *ret; int order = get_order(size); /* * XXX fix me: the DMA API should pass us an explicit DMA mask * instead, or use ZONE_DMA32 (ia64 overloads ZONE_DMA to be a ~32 * bit range instead of a 16MB one). */ flags |= GFP_DMA; |
25667d675 Revert "[IA64] sw... |
464 |
ret = (void *)__get_free_pages(flags, order); |
93fbff63e [IA64] make swiot... |
465 |
if (ret && address_needs_mapping(hwdev, virt_to_bus(ret))) { |
1da177e4c Linux-2.6.12-rc2 |
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 |
/* * The allocated memory isn't reachable by the device. * Fall back on swiotlb_map_single(). */ free_pages((unsigned long) ret, order); ret = NULL; } if (!ret) { /* * We are either out of memory or the device can't DMA * to GFP_DMA memory; fall back on * swiotlb_map_single(), which will grab memory from * the lowest available address range. */ dma_addr_t handle; handle = swiotlb_map_single(NULL, NULL, size, DMA_FROM_DEVICE); |
17a941d85 [PATCH] x86_64: U... |
482 |
if (swiotlb_dma_mapping_error(handle)) |
1da177e4c Linux-2.6.12-rc2 |
483 |
return NULL; |
93fbff63e [IA64] make swiot... |
484 |
ret = bus_to_virt(handle); |
1da177e4c Linux-2.6.12-rc2 |
485 486 487 |
} memset(ret, 0, size); |
93fbff63e [IA64] make swiot... |
488 |
dev_addr = virt_to_bus(ret); |
1da177e4c Linux-2.6.12-rc2 |
489 490 491 |
/* Confirm address can be DMA'd by device */ if (address_needs_mapping(hwdev, dev_addr)) { |
563aaf064 [IA64] swiotlb cl... |
492 493 494 495 |
printk("hwdev DMA mask = 0x%016Lx, dev_addr = 0x%016Lx ", (unsigned long long)*hwdev->dma_mask, (unsigned long long)dev_addr); |
1da177e4c Linux-2.6.12-rc2 |
496 497 498 499 500 501 502 503 504 505 506 |
panic("swiotlb_alloc_coherent: allocated memory is out of " "range for device"); } *dma_handle = dev_addr; return ret; } void swiotlb_free_coherent(struct device *hwdev, size_t size, void *vaddr, dma_addr_t dma_handle) { |
aa24886e3 dma_free_coherent... |
507 |
WARN_ON(irqs_disabled()); |
1da177e4c Linux-2.6.12-rc2 |
508 509 510 511 512 513 514 515 516 517 518 519 520 521 |
if (!(vaddr >= (void *)io_tlb_start && vaddr < (void *)io_tlb_end)) free_pages((unsigned long) vaddr, get_order(size)); else /* DMA_TO_DEVICE to avoid memcpy in unmap_single */ swiotlb_unmap_single (hwdev, dma_handle, size, DMA_TO_DEVICE); } static void swiotlb_full(struct device *dev, size_t size, int dir, int do_panic) { /* * Ran out of IOMMU space for this operation. This is very bad. * Unfortunately the drivers cannot handle this operation properly. |
17e5ad6c0 [PATCH] Removed r... |
522 |
* unless they check for dma_mapping_error (most don't) |
1da177e4c Linux-2.6.12-rc2 |
523 524 525 |
* When the mapping is small enough return a static buffer to limit * the damage, or panic when the transfer is too big. */ |
563aaf064 [IA64] swiotlb cl... |
526 |
printk(KERN_ERR "DMA: Out of SW-IOMMU space for %zu bytes at " |
1da177e4c Linux-2.6.12-rc2 |
527 528 529 530 |
"device %s ", size, dev ? dev->bus_id : "?"); if (size > io_tlb_overflow && do_panic) { |
17e5ad6c0 [PATCH] Removed r... |
531 532 533 534 535 536 |
if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL) panic("DMA: Memory would be corrupted "); if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL) panic("DMA: Random memory would be DMAed "); |
1da177e4c Linux-2.6.12-rc2 |
537 538 539 540 541 |
} } /* * Map a single buffer of the indicated size for DMA in streaming mode. The |
17e5ad6c0 [PATCH] Removed r... |
542 |
* physical address to use is returned. |
1da177e4c Linux-2.6.12-rc2 |
543 544 545 546 547 |
* * Once the device is given the dma address, the device owns this memory until * either swiotlb_unmap_single or swiotlb_dma_sync_single is performed. */ dma_addr_t |
309df0c50 dma/ia64: update ... |
548 549 |
swiotlb_map_single_attrs(struct device *hwdev, void *ptr, size_t size, int dir, struct dma_attrs *attrs) |
1da177e4c Linux-2.6.12-rc2 |
550 |
{ |
563aaf064 [IA64] swiotlb cl... |
551 |
dma_addr_t dev_addr = virt_to_bus(ptr); |
1da177e4c Linux-2.6.12-rc2 |
552 |
void *map; |
348145458 BUG_ON() Conversi... |
553 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
554 555 556 557 558 |
/* * If the pointer passed in happens to be in the device's DMA window, * we can safely return the device addr and not worry about bounce * buffering it. */ |
25667d675 Revert "[IA64] sw... |
559 |
if (!address_needs_mapping(hwdev, dev_addr) && !swiotlb_force) |
1da177e4c Linux-2.6.12-rc2 |
560 561 562 563 564 |
return dev_addr; /* * Oh well, have to allocate and map a bounce buffer. */ |
25667d675 Revert "[IA64] sw... |
565 |
map = map_single(hwdev, ptr, size, dir); |
1da177e4c Linux-2.6.12-rc2 |
566 567 568 569 |
if (!map) { swiotlb_full(hwdev, size, dir, 1); map = io_tlb_overflow_buffer; } |
93fbff63e [IA64] make swiot... |
570 |
dev_addr = virt_to_bus(map); |
1da177e4c Linux-2.6.12-rc2 |
571 572 573 574 575 576 577 578 579 |
/* * Ensure that the address returned is DMA'ble */ if (address_needs_mapping(hwdev, dev_addr)) panic("map_single: bounce buffer is not DMA'ble"); return dev_addr; } |
309df0c50 dma/ia64: update ... |
580 581 582 583 584 585 586 |
EXPORT_SYMBOL(swiotlb_map_single_attrs); dma_addr_t swiotlb_map_single(struct device *hwdev, void *ptr, size_t size, int dir) { return swiotlb_map_single_attrs(hwdev, ptr, size, dir, NULL); } |
1da177e4c Linux-2.6.12-rc2 |
587 588 |
/* |
1da177e4c Linux-2.6.12-rc2 |
589 590 591 592 593 594 595 596 |
* Unmap a single streaming mode DMA translation. The dma_addr and size must * match what was provided for in a previous swiotlb_map_single call. All * other usages are undefined. * * After this call, reads by the cpu to the buffer are guaranteed to see * whatever the device wrote there. */ void |
309df0c50 dma/ia64: update ... |
597 598 |
swiotlb_unmap_single_attrs(struct device *hwdev, dma_addr_t dev_addr, size_t size, int dir, struct dma_attrs *attrs) |
1da177e4c Linux-2.6.12-rc2 |
599 |
{ |
93fbff63e [IA64] make swiot... |
600 |
char *dma_addr = bus_to_virt(dev_addr); |
1da177e4c Linux-2.6.12-rc2 |
601 |
|
348145458 BUG_ON() Conversi... |
602 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
603 604 605 |
if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end) unmap_single(hwdev, dma_addr, size, dir); else if (dir == DMA_FROM_DEVICE) |
cde14bbfb [IA64] swiotlb bu... |
606 |
dma_mark_clean(dma_addr, size); |
1da177e4c Linux-2.6.12-rc2 |
607 |
} |
309df0c50 dma/ia64: update ... |
608 |
EXPORT_SYMBOL(swiotlb_unmap_single_attrs); |
1da177e4c Linux-2.6.12-rc2 |
609 |
|
309df0c50 dma/ia64: update ... |
610 611 612 613 614 615 |
void swiotlb_unmap_single(struct device *hwdev, dma_addr_t dev_addr, size_t size, int dir) { return swiotlb_unmap_single_attrs(hwdev, dev_addr, size, dir, NULL); } |
1da177e4c Linux-2.6.12-rc2 |
616 617 618 619 620 |
/* * Make physical memory consistent for a single streaming mode DMA translation * after a transfer. * * If you perform a swiotlb_map_single() but wish to interrogate the buffer |
17e5ad6c0 [PATCH] Removed r... |
621 622 |
* using the cpu, yet do not wish to teardown the dma mapping, you must * call this function before doing so. At the next point you give the dma |
1da177e4c Linux-2.6.12-rc2 |
623 624 625 |
* address back to the card, you must first perform a * swiotlb_dma_sync_for_device, and then the device again owns the buffer */ |
be6b02678 [PATCH] swiotlb u... |
626 |
static void |
8270f3f1a [PATCH] swiotlb: ... |
627 |
swiotlb_sync_single(struct device *hwdev, dma_addr_t dev_addr, |
de69e0f0b [PATCH] swiotlb: ... |
628 |
size_t size, int dir, int target) |
1da177e4c Linux-2.6.12-rc2 |
629 |
{ |
93fbff63e [IA64] make swiot... |
630 |
char *dma_addr = bus_to_virt(dev_addr); |
1da177e4c Linux-2.6.12-rc2 |
631 |
|
348145458 BUG_ON() Conversi... |
632 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
633 |
if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end) |
de69e0f0b [PATCH] swiotlb: ... |
634 |
sync_single(hwdev, dma_addr, size, dir, target); |
1da177e4c Linux-2.6.12-rc2 |
635 |
else if (dir == DMA_FROM_DEVICE) |
cde14bbfb [IA64] swiotlb bu... |
636 |
dma_mark_clean(dma_addr, size); |
1da177e4c Linux-2.6.12-rc2 |
637 638 639 |
} void |
8270f3f1a [PATCH] swiotlb: ... |
640 641 642 |
swiotlb_sync_single_for_cpu(struct device *hwdev, dma_addr_t dev_addr, size_t size, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
643 |
swiotlb_sync_single(hwdev, dev_addr, size, dir, SYNC_FOR_CPU); |
8270f3f1a [PATCH] swiotlb: ... |
644 645 646 |
} void |
1da177e4c Linux-2.6.12-rc2 |
647 648 649 |
swiotlb_sync_single_for_device(struct device *hwdev, dma_addr_t dev_addr, size_t size, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
650 |
swiotlb_sync_single(hwdev, dev_addr, size, dir, SYNC_FOR_DEVICE); |
1da177e4c Linux-2.6.12-rc2 |
651 652 653 |
} /* |
878a97cfd [PATCH] swiotlb: ... |
654 655 |
* Same as above, but for a sub-range of the mapping. */ |
be6b02678 [PATCH] swiotlb u... |
656 |
static void |
878a97cfd [PATCH] swiotlb: ... |
657 |
swiotlb_sync_single_range(struct device *hwdev, dma_addr_t dev_addr, |
de69e0f0b [PATCH] swiotlb: ... |
658 659 |
unsigned long offset, size_t size, int dir, int target) |
878a97cfd [PATCH] swiotlb: ... |
660 |
{ |
93fbff63e [IA64] make swiot... |
661 |
char *dma_addr = bus_to_virt(dev_addr) + offset; |
878a97cfd [PATCH] swiotlb: ... |
662 |
|
348145458 BUG_ON() Conversi... |
663 |
BUG_ON(dir == DMA_NONE); |
878a97cfd [PATCH] swiotlb: ... |
664 |
if (dma_addr >= io_tlb_start && dma_addr < io_tlb_end) |
de69e0f0b [PATCH] swiotlb: ... |
665 |
sync_single(hwdev, dma_addr, size, dir, target); |
878a97cfd [PATCH] swiotlb: ... |
666 |
else if (dir == DMA_FROM_DEVICE) |
cde14bbfb [IA64] swiotlb bu... |
667 |
dma_mark_clean(dma_addr, size); |
878a97cfd [PATCH] swiotlb: ... |
668 669 670 671 672 673 |
} void swiotlb_sync_single_range_for_cpu(struct device *hwdev, dma_addr_t dev_addr, unsigned long offset, size_t size, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
674 675 |
swiotlb_sync_single_range(hwdev, dev_addr, offset, size, dir, SYNC_FOR_CPU); |
878a97cfd [PATCH] swiotlb: ... |
676 677 678 679 680 681 |
} void swiotlb_sync_single_range_for_device(struct device *hwdev, dma_addr_t dev_addr, unsigned long offset, size_t size, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
682 683 |
swiotlb_sync_single_range(hwdev, dev_addr, offset, size, dir, SYNC_FOR_DEVICE); |
878a97cfd [PATCH] swiotlb: ... |
684 |
} |
309df0c50 dma/ia64: update ... |
685 686 |
void swiotlb_unmap_sg_attrs(struct device *, struct scatterlist *, int, int, struct dma_attrs *); |
878a97cfd [PATCH] swiotlb: ... |
687 |
/* |
1da177e4c Linux-2.6.12-rc2 |
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 |
* Map a set of buffers described by scatterlist in streaming mode for DMA. * This is the scatter-gather version of the above swiotlb_map_single * interface. Here the scatter gather list elements are each tagged with the * appropriate dma address and length. They are obtained via * sg_dma_{address,length}(SG). * * NOTE: An implementation may be able to use a smaller number of * DMA address/length pairs than there are SG table elements. * (for example via virtual mapping capabilities) * The routine returns the number of addr/length pairs actually * used, at most nents. * * Device ownership issues as mentioned above for swiotlb_map_single are the * same here. */ int |
309df0c50 dma/ia64: update ... |
704 705 |
swiotlb_map_sg_attrs(struct device *hwdev, struct scatterlist *sgl, int nelems, int dir, struct dma_attrs *attrs) |
1da177e4c Linux-2.6.12-rc2 |
706 |
{ |
dbfd49fe9 swiotlb: sg chain... |
707 |
struct scatterlist *sg; |
25667d675 Revert "[IA64] sw... |
708 |
void *addr; |
563aaf064 [IA64] swiotlb cl... |
709 |
dma_addr_t dev_addr; |
1da177e4c Linux-2.6.12-rc2 |
710 |
int i; |
348145458 BUG_ON() Conversi... |
711 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
712 |
|
dbfd49fe9 swiotlb: sg chain... |
713 |
for_each_sg(sgl, sg, nelems, i) { |
25667d675 Revert "[IA64] sw... |
714 715 716 717 |
addr = SG_ENT_VIRT_ADDRESS(sg); dev_addr = virt_to_bus(addr); if (swiotlb_force || address_needs_mapping(hwdev, dev_addr)) { void *map = map_single(hwdev, addr, sg->length, dir); |
7e8702334 [PATCH] Fix swiot... |
718 |
if (!map) { |
1da177e4c Linux-2.6.12-rc2 |
719 720 721 |
/* Don't panic here, we expect map_sg users to do proper error handling. */ swiotlb_full(hwdev, sg->length, dir, 0); |
309df0c50 dma/ia64: update ... |
722 723 |
swiotlb_unmap_sg_attrs(hwdev, sgl, i, dir, attrs); |
dbfd49fe9 swiotlb: sg chain... |
724 |
sgl[0].dma_length = 0; |
1da177e4c Linux-2.6.12-rc2 |
725 726 |
return 0; } |
cde14bbfb [IA64] swiotlb bu... |
727 |
sg->dma_address = virt_to_bus(map); |
1da177e4c Linux-2.6.12-rc2 |
728 729 730 731 732 733 |
} else sg->dma_address = dev_addr; sg->dma_length = sg->length; } return nelems; } |
309df0c50 dma/ia64: update ... |
734 735 736 737 738 739 740 741 |
EXPORT_SYMBOL(swiotlb_map_sg_attrs); int swiotlb_map_sg(struct device *hwdev, struct scatterlist *sgl, int nelems, int dir) { return swiotlb_map_sg_attrs(hwdev, sgl, nelems, dir, NULL); } |
1da177e4c Linux-2.6.12-rc2 |
742 743 744 745 746 747 |
/* * Unmap a set of streaming mode DMA translations. Again, cpu read rules * concerning calls here are the same as for swiotlb_unmap_single() above. */ void |
309df0c50 dma/ia64: update ... |
748 749 |
swiotlb_unmap_sg_attrs(struct device *hwdev, struct scatterlist *sgl, int nelems, int dir, struct dma_attrs *attrs) |
1da177e4c Linux-2.6.12-rc2 |
750 |
{ |
dbfd49fe9 swiotlb: sg chain... |
751 |
struct scatterlist *sg; |
1da177e4c Linux-2.6.12-rc2 |
752 |
int i; |
348145458 BUG_ON() Conversi... |
753 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
754 |
|
dbfd49fe9 swiotlb: sg chain... |
755 |
for_each_sg(sgl, sg, nelems, i) { |
1da177e4c Linux-2.6.12-rc2 |
756 |
if (sg->dma_address != SG_ENT_PHYS_ADDRESS(sg)) |
93fbff63e [IA64] make swiot... |
757 758 |
unmap_single(hwdev, bus_to_virt(sg->dma_address), sg->dma_length, dir); |
1da177e4c Linux-2.6.12-rc2 |
759 |
else if (dir == DMA_FROM_DEVICE) |
cde14bbfb [IA64] swiotlb bu... |
760 |
dma_mark_clean(SG_ENT_VIRT_ADDRESS(sg), sg->dma_length); |
dbfd49fe9 swiotlb: sg chain... |
761 |
} |
1da177e4c Linux-2.6.12-rc2 |
762 |
} |
309df0c50 dma/ia64: update ... |
763 764 765 766 767 768 769 770 |
EXPORT_SYMBOL(swiotlb_unmap_sg_attrs); void swiotlb_unmap_sg(struct device *hwdev, struct scatterlist *sgl, int nelems, int dir) { return swiotlb_unmap_sg_attrs(hwdev, sgl, nelems, dir, NULL); } |
1da177e4c Linux-2.6.12-rc2 |
771 772 773 774 775 776 777 778 |
/* * Make physical memory consistent for a set of streaming mode DMA translations * after a transfer. * * The same as swiotlb_sync_single_* but for a scatter-gather list, same rules * and usage. */ |
be6b02678 [PATCH] swiotlb u... |
779 |
static void |
dbfd49fe9 swiotlb: sg chain... |
780 |
swiotlb_sync_sg(struct device *hwdev, struct scatterlist *sgl, |
de69e0f0b [PATCH] swiotlb: ... |
781 |
int nelems, int dir, int target) |
1da177e4c Linux-2.6.12-rc2 |
782 |
{ |
dbfd49fe9 swiotlb: sg chain... |
783 |
struct scatterlist *sg; |
1da177e4c Linux-2.6.12-rc2 |
784 |
int i; |
348145458 BUG_ON() Conversi... |
785 |
BUG_ON(dir == DMA_NONE); |
1da177e4c Linux-2.6.12-rc2 |
786 |
|
dbfd49fe9 swiotlb: sg chain... |
787 |
for_each_sg(sgl, sg, nelems, i) { |
1da177e4c Linux-2.6.12-rc2 |
788 |
if (sg->dma_address != SG_ENT_PHYS_ADDRESS(sg)) |
93fbff63e [IA64] make swiot... |
789 |
sync_single(hwdev, bus_to_virt(sg->dma_address), |
de69e0f0b [PATCH] swiotlb: ... |
790 |
sg->dma_length, dir, target); |
cde14bbfb [IA64] swiotlb bu... |
791 792 |
else if (dir == DMA_FROM_DEVICE) dma_mark_clean(SG_ENT_VIRT_ADDRESS(sg), sg->dma_length); |
dbfd49fe9 swiotlb: sg chain... |
793 |
} |
1da177e4c Linux-2.6.12-rc2 |
794 795 796 |
} void |
8270f3f1a [PATCH] swiotlb: ... |
797 798 799 |
swiotlb_sync_sg_for_cpu(struct device *hwdev, struct scatterlist *sg, int nelems, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
800 |
swiotlb_sync_sg(hwdev, sg, nelems, dir, SYNC_FOR_CPU); |
8270f3f1a [PATCH] swiotlb: ... |
801 802 803 |
} void |
1da177e4c Linux-2.6.12-rc2 |
804 805 806 |
swiotlb_sync_sg_for_device(struct device *hwdev, struct scatterlist *sg, int nelems, int dir) { |
de69e0f0b [PATCH] swiotlb: ... |
807 |
swiotlb_sync_sg(hwdev, sg, nelems, dir, SYNC_FOR_DEVICE); |
1da177e4c Linux-2.6.12-rc2 |
808 809 810 811 812 |
} int swiotlb_dma_mapping_error(dma_addr_t dma_addr) { |
93fbff63e [IA64] make swiot... |
813 |
return (dma_addr == virt_to_bus(io_tlb_overflow_buffer)); |
1da177e4c Linux-2.6.12-rc2 |
814 815 816 |
} /* |
17e5ad6c0 [PATCH] Removed r... |
817 |
* Return whether the given device DMA address mask can be supported |
1da177e4c Linux-2.6.12-rc2 |
818 |
* properly. For example, if your device can only drive the low 24-bits |
17e5ad6c0 [PATCH] Removed r... |
819 |
* during bus mastering, then you would pass 0x00ffffff as the mask to |
1da177e4c Linux-2.6.12-rc2 |
820 821 822 |
* this function. */ int |
563aaf064 [IA64] swiotlb cl... |
823 |
swiotlb_dma_supported(struct device *hwdev, u64 mask) |
1da177e4c Linux-2.6.12-rc2 |
824 |
{ |
25667d675 Revert "[IA64] sw... |
825 |
return virt_to_bus(io_tlb_end - 1) <= mask; |
1da177e4c Linux-2.6.12-rc2 |
826 |
} |
1da177e4c Linux-2.6.12-rc2 |
827 828 829 830 831 832 |
EXPORT_SYMBOL(swiotlb_map_single); EXPORT_SYMBOL(swiotlb_unmap_single); EXPORT_SYMBOL(swiotlb_map_sg); EXPORT_SYMBOL(swiotlb_unmap_sg); EXPORT_SYMBOL(swiotlb_sync_single_for_cpu); EXPORT_SYMBOL(swiotlb_sync_single_for_device); |
878a97cfd [PATCH] swiotlb: ... |
833 834 |
EXPORT_SYMBOL_GPL(swiotlb_sync_single_range_for_cpu); EXPORT_SYMBOL_GPL(swiotlb_sync_single_range_for_device); |
1da177e4c Linux-2.6.12-rc2 |
835 836 837 |
EXPORT_SYMBOL(swiotlb_sync_sg_for_cpu); EXPORT_SYMBOL(swiotlb_sync_sg_for_device); EXPORT_SYMBOL(swiotlb_dma_mapping_error); |
25667d675 Revert "[IA64] sw... |
838 839 |
EXPORT_SYMBOL(swiotlb_alloc_coherent); EXPORT_SYMBOL(swiotlb_free_coherent); |
1da177e4c Linux-2.6.12-rc2 |
840 |
EXPORT_SYMBOL(swiotlb_dma_supported); |