| 1 | #include <filedevice/seadFileDeviceMgr.h> |
| 2 | #include <heap/seadHeap.h> |
| 3 | #include <heap/seadHeapMgr.h> |
| 4 | #include <math/seadMathCalcCommon.h> |
| 5 | #include <prim/seadBitUtil.h> |
| 6 | #include <prim/seadEndian.h> |
| 7 | #include <prim/seadPtrUtil.h> |
| 8 | #include <prim/seadSafeString.h> |
| 9 | #include <resource/seadSZSDecompressor.h> |
| 10 | |
| 11 | namespace |
| 12 | { |
| 13 | #ifdef cafe |
| 14 | __attribute__((aligned(0x20))) s32 decodeSZSCafeAsm_(void* dst, const void* src) |
| 15 | { |
| 16 | asm("lwz r5, 0x4(r4)\n" ); |
| 17 | asm("li r11, 0x20\n" ); |
| 18 | asm("li r6, 0\n" ); |
| 19 | asm("mr r0, r5\n" ); |
| 20 | asm("addi r4, r4, 0xf\n" ); |
| 21 | asm("subi r3, r3, 1\n" ); |
| 22 | asm("cmpwi r5, 0x132\n" ); |
| 23 | asm("ble _final_decloop0\n" ); |
| 24 | asm("subi r5, r5, 0x132\n" ); |
| 25 | asm("nop\n" ); |
| 26 | asm("nop\n" ); |
| 27 | asm("nop\n" ); |
| 28 | asm("nop\n" ); |
| 29 | asm("nop\n" ); |
| 30 | asm("nop\n" ); |
| 31 | asm("nop\n" ); |
| 32 | |
| 33 | asm("_decloop0: rlwinm. r6, r6, 0x1f, 1, 0x1f\n" ); |
| 34 | asm("bne _decloop1\n" ); |
| 35 | asm("lbzu r7, 1(r4)\n" ); |
| 36 | asm("li r6, 0x80\n" ); |
| 37 | |
| 38 | asm("_decloop1: and. r8, r6, r7\n" ); |
| 39 | asm("lbzu r8, 1(r4)\n" ); |
| 40 | asm("beq _decloop2\n" ); |
| 41 | asm("andi. r9, r3, 0x1f\n" ); |
| 42 | asm("bne _decloop1x\n" ); |
| 43 | asm("dcbz r11, r3\n" ); |
| 44 | |
| 45 | asm("_decloop1x: subic. r5, r5, 1\n" ); |
| 46 | asm("stbu r8, 1(r3)\n" ); |
| 47 | asm("bne _decloop0\n" ); |
| 48 | asm("b _decloop8\n" ); |
| 49 | |
| 50 | asm("_decloop2: lbzu r9, 1(r4)\n" ); |
| 51 | asm("rlwinm. r10, r8, 0x1c, 4, 0x1f\n" ); |
| 52 | asm("bne _decloop3\n" ); |
| 53 | asm("lbzu r10, 1(r4)\n" ); |
| 54 | asm("addi r10, r10, 0x10" ); |
| 55 | |
| 56 | asm("_decloop3: addi r10, r10, 2\n" ); |
| 57 | asm("rlwimi r9, r8, 8, 0x14, 0x17\n" ); |
| 58 | asm("subf r5, r10, r5\n" ); |
| 59 | asm("subf r8, r9, r4\n" ); |
| 60 | asm("mtspr CTR, r10\n" ); |
| 61 | asm("addi r8, r8, 1\n" ); |
| 62 | |
| 63 | asm("_decloop4: andi. r9, r3, 0x1f\n" ); |
| 64 | asm("lbz r9, -1(r8)\n" ); |
| 65 | asm("addi r8, r8, 1\n" ); |
| 66 | asm("bne _decloop5\n" ); |
| 67 | asm("dcbz r11, r3\n" ); |
| 68 | |
| 69 | asm("_decloop5: stbu r9, 1(r3)\n" ); |
| 70 | asm("bdnz _decloop4\n" ); |
| 71 | asm("cmpwi r5, 0\n" ); |
| 72 | asm("bgt _decloop0\n" ); |
| 73 | |
| 74 | asm("_decloop8: addi r5, r5, 0x132\n" ); |
| 75 | asm("cmpwi r5, 0\n" ); |
| 76 | asm("ble _final_decloop8\n" ); |
| 77 | |
| 78 | asm("_final_decloop0: rlwinm. r6, r6, 0x1f, 1, 0x1f\n" ); |
| 79 | asm("bne _final_decloop1\n" ); |
| 80 | asm("lbzu r7, 1(r4)\n" ); |
| 81 | asm("li r6, 0x80\n" ); |
| 82 | |
| 83 | asm("_final_decloop1: and. r8, r6, r7\n" ); |
| 84 | asm("lbzu r8, 1(r4)\n" ); |
| 85 | asm("beq _final_decloop2\n" ); |
| 86 | asm("subic. r5, r5, 1\n" ); |
| 87 | asm("stbu r8, 1(r3)\n" ); |
| 88 | asm("bne _final_decloop0\n" ); |
| 89 | asm("b _final_decloop8\n" ); |
| 90 | |
| 91 | asm("_final_decloop2: lbzu r9, 1(r4)\n" ); |
| 92 | asm("rlwinm. r10, r8, 0x1c, 4, 0x1f\n" ); |
| 93 | asm("bne _final_decloop3\n" ); |
| 94 | asm("lbzu r10, 1(r4)\n" ); |
| 95 | asm("addi r10, r10, 0x10\n" ); |
| 96 | |
| 97 | asm("_final_decloop3: addi r10, r10, 2\n" ); |
| 98 | asm("rlwimi r9, r8, 8, 0x14, 0x17\n" ); |
| 99 | asm("subf. r5, r10, r5\n" ); |
| 100 | asm("blt _final_decloop8\n" ); |
| 101 | asm("subf r8, r9, r3\n" ); |
| 102 | asm("mtspr CTR, r10\n" ); |
| 103 | asm("addi r8, r8, 1\n" ); |
| 104 | |
| 105 | asm("_final_decloop4: lbz r9, -1(r8)\n" ); |
| 106 | asm("addi r8, r8, 1\n" ); |
| 107 | asm("stbu r9, 1(r3)\n" ); |
| 108 | asm("bdnz _final_decloop4\n" ); |
| 109 | asm("cmpwi r5, 0\n" ); |
| 110 | asm("bgt _final_decloop0\n" ); |
| 111 | |
| 112 | s32 register error asm("r3" ); |
| 113 | asm("_final_decloop8: mr %0, r0\n" : "=r" (error)); |
| 114 | asm("blr" ); |
| 115 | |
| 116 | return error; |
| 117 | } |
| 118 | #endif // cafe |
| 119 | } // namespace |
| 120 | |
| 121 | #ifdef SWITCH |
| 122 | s32 decodeSZSNxAsm64_(void* dst, const void* src); |
| 123 | #endif |
| 124 | |
| 125 | namespace sead |
| 126 | { |
| 127 | SZSDecompressor::DecompContext::DecompContext() |
| 128 | { |
| 129 | initialize(NULL); |
| 130 | } |
| 131 | |
| 132 | SZSDecompressor::DecompContext::DecompContext(void* dst) |
| 133 | { |
| 134 | initialize(dst); |
| 135 | } |
| 136 | |
| 137 | void SZSDecompressor::DecompContext::initialize(void* dst) |
| 138 | { |
| 139 | destp = static_cast<u8*>(dst); |
| 140 | destCount = 0; |
| 141 | forceDestCount = 0; |
| 142 | flagMask = 0; |
| 143 | flags = 0; |
| 144 | packHigh = 0; |
| 145 | step = SZSDecompressor::cStepNormal; |
| 146 | lzOffset = 0; |
| 147 | headerSize = 0x10; |
| 148 | } |
| 149 | |
| 150 | SZSDecompressor::SZSDecompressor(u32 workSize, u8* workBuffer) : Decompressor("szs" ) |
| 151 | { |
| 152 | if (workBuffer == NULL) |
| 153 | { |
| 154 | mWorkSize = Mathu::roundUpPow2(val: workSize, base: FileDevice::cBufferMinAlignment); |
| 155 | mWorkBuffer = NULL; |
| 156 | } |
| 157 | |
| 158 | else |
| 159 | { |
| 160 | mWorkSize = workSize; |
| 161 | mWorkBuffer = workBuffer; |
| 162 | } |
| 163 | } |
| 164 | |
| 165 | u8* SZSDecompressor::tryDecompFromDevice(const ResourceMgr::LoadArg& loadArg, Resource* resource, |
| 166 | u32* outSize, u32* outAllocSize, bool* outAllocated) |
| 167 | { |
| 168 | Heap* heap = loadArg.load_data_heap; |
| 169 | if (heap == NULL) |
| 170 | heap = HeapMgr::sInstancePtr->getCurrentHeap(); |
| 171 | |
| 172 | FileHandle handle; |
| 173 | FileDevice* device; |
| 174 | u8* src; |
| 175 | |
| 176 | if (loadArg.device != NULL) |
| 177 | device = loadArg.device->tryOpen(handle: &handle, path: loadArg.path, flag: FileDevice::cFileOpenFlag_ReadOnly, |
| 178 | divSize: loadArg.div_size); |
| 179 | else |
| 180 | device = FileDeviceMgr::instance()->tryOpen( |
| 181 | handle: &handle, path: loadArg.path, flag: FileDevice::cFileOpenFlag_ReadOnly, divSize: loadArg.div_size); |
| 182 | |
| 183 | if (device != NULL && |
| 184 | ((src = mWorkBuffer, src != NULL) || |
| 185 | (src = new (heap, -FileDevice::cBufferMinAlignment) u8[mWorkSize], src != NULL))) |
| 186 | { |
| 187 | u32 bytesRead = handle.read(outBuffer: src, bytesToRead: mWorkSize); |
| 188 | if (bytesRead >= 0x10) |
| 189 | { |
| 190 | u32 decompSize = getDecompSize(src); |
| 191 | s32 decompAlignment = getDecompAlignment(src); |
| 192 | |
| 193 | u32 allocSize = loadArg.load_data_buffer_size; |
| 194 | u8* dst = loadArg.load_data_buffer; |
| 195 | |
| 196 | if (decompSize > allocSize && allocSize != 0) |
| 197 | decompSize = allocSize; |
| 198 | |
| 199 | bool allocated = false; |
| 200 | allocSize = Mathu::roundUpPow2(val: decompSize, base: 0x20); |
| 201 | |
| 202 | if (dst == NULL) |
| 203 | { |
| 204 | DirectResource* directResource = DynamicCast<DirectResource, Resource>(obj: resource); |
| 205 | if (directResource != NULL) |
| 206 | { |
| 207 | s32 alignment = loadArg.load_data_alignment; |
| 208 | if (alignment != 0) |
| 209 | decompAlignment = (alignment < 0x20) ? 0x20 : alignment; |
| 210 | |
| 211 | else |
| 212 | { |
| 213 | if (decompAlignment == 0) |
| 214 | decompAlignment = directResource->getLoadDataAlignment(); |
| 215 | |
| 216 | decompAlignment = ((loadArg.instance_alignment < 0) ? -1 : 1) * |
| 217 | ((decompAlignment < 0x20) ? 0x20 : decompAlignment); |
| 218 | } |
| 219 | } |
| 220 | |
| 221 | else |
| 222 | decompAlignment = -(((loadArg.instance_alignment < 0) ? -1 : 1) << 5); |
| 223 | |
| 224 | dst = new (heap, decompAlignment) u8[allocSize]; |
| 225 | |
| 226 | if (dst != NULL) |
| 227 | allocated = true; |
| 228 | } |
| 229 | |
| 230 | if (dst != NULL) |
| 231 | { |
| 232 | s32 error; |
| 233 | if (bytesRead < mWorkSize) |
| 234 | error = decomp(dst, dstSize: allocSize, src, srcSize: mWorkSize); |
| 235 | |
| 236 | else |
| 237 | { |
| 238 | DecompContext context(dst); |
| 239 | context.forceDestCount = decompSize; |
| 240 | |
| 241 | do |
| 242 | { |
| 243 | error = streamDecomp(context: &context, src, srcSize: bytesRead); |
| 244 | if (error <= 0) |
| 245 | break; |
| 246 | } while ((bytesRead = handle.read(outBuffer: src, bytesToRead: mWorkSize), bytesRead != 0)); |
| 247 | } |
| 248 | |
| 249 | if (!(error < 0)) |
| 250 | { |
| 251 | if (mWorkBuffer == NULL) |
| 252 | delete[] src; |
| 253 | |
| 254 | if (outSize != NULL) |
| 255 | *outSize = decompSize; |
| 256 | |
| 257 | if (outAllocSize != NULL) |
| 258 | *outAllocSize = allocSize; |
| 259 | |
| 260 | if (outAllocated != NULL) |
| 261 | *outAllocated = allocated; |
| 262 | |
| 263 | return dst; |
| 264 | } |
| 265 | |
| 266 | if (allocated) |
| 267 | delete[] dst; |
| 268 | } |
| 269 | } |
| 270 | |
| 271 | if (mWorkBuffer == NULL) |
| 272 | delete[] src; |
| 273 | } |
| 274 | |
| 275 | return NULL; |
| 276 | } |
| 277 | |
| 278 | u32 SZSDecompressor::getDecompAlignment(const void* src) |
| 279 | { |
| 280 | return Endian::toHostU32(from: Endian::cBig, v: BitUtil::bitCastPtr<u32>(value: src, offset: 8)); |
| 281 | } |
| 282 | |
| 283 | u32 SZSDecompressor::getDecompSize(const void* src) |
| 284 | { |
| 285 | return Endian::toHostU32(from: Endian::cBig, v: BitUtil::bitCastPtr<u32>(value: src, offset: 4)); |
| 286 | } |
| 287 | |
| 288 | s32 SZSDecompressor::(DecompContext* context, const u8* src, u32 srcSize) |
| 289 | { |
| 290 | s32 len = 0; |
| 291 | |
| 292 | while (context->headerSize != 0) |
| 293 | { |
| 294 | context->headerSize -= 1; |
| 295 | |
| 296 | if (context->headerSize == 0xF) |
| 297 | { |
| 298 | if (*src != 0x59) |
| 299 | return -1; |
| 300 | } |
| 301 | |
| 302 | else if (context->headerSize == 0xE) |
| 303 | { |
| 304 | if (*src != 0x61) |
| 305 | return -1; |
| 306 | } |
| 307 | |
| 308 | else if (context->headerSize == 0xD) |
| 309 | { |
| 310 | if (*src != 0x7A) |
| 311 | return -1; |
| 312 | } |
| 313 | |
| 314 | else if (context->headerSize == 0xC) |
| 315 | { |
| 316 | if (*src != 0x30) |
| 317 | return -1; |
| 318 | } |
| 319 | |
| 320 | else if (7 < context->headerSize) |
| 321 | context->destCount |= static_cast<u32>(*src) << (context->headerSize - 8) * 8; |
| 322 | |
| 323 | src++; |
| 324 | len += 1; |
| 325 | if (--srcSize == 0 && context->headerSize != 0) |
| 326 | return len; |
| 327 | } |
| 328 | |
| 329 | if (context->forceDestCount < 1) |
| 330 | return len; |
| 331 | |
| 332 | if (context->destCount <= context->forceDestCount) |
| 333 | return len; |
| 334 | |
| 335 | context->destCount = context->forceDestCount; |
| 336 | return len; |
| 337 | } |
| 338 | |
| 339 | s32 SZSDecompressor::streamDecomp(DecompContext* context, const void* src, u32 srcSize) |
| 340 | { |
| 341 | const u8* _src = static_cast<const u8*>(src); |
| 342 | u32 n; |
| 343 | |
| 344 | if (context->headerSize != 0) |
| 345 | { |
| 346 | s32 len = readHeader_(context, src: _src, srcSize); |
| 347 | if (len < 0) |
| 348 | return len; |
| 349 | |
| 350 | srcSize -= len; |
| 351 | _src += len; |
| 352 | |
| 353 | if (srcSize == 0) |
| 354 | { |
| 355 | if (context->headerSize == 0) |
| 356 | return context->destCount; |
| 357 | |
| 358 | return -1; |
| 359 | } |
| 360 | } |
| 361 | |
| 362 | while (context->destCount > 0) |
| 363 | { |
| 364 | if (context->step == cStepLong) |
| 365 | { |
| 366 | n = *_src + 0x12; |
| 367 | if (!context->doCopy(n)) |
| 368 | return -2; |
| 369 | } |
| 370 | |
| 371 | else if (context->step == cStepShort) |
| 372 | { |
| 373 | context->lzOffset = (((context->packHigh << 8) & 0xf00) | *_src) + 1; |
| 374 | |
| 375 | n = context->packHigh >> 4; |
| 376 | if (n != 0) |
| 377 | { |
| 378 | n += 2; |
| 379 | if (!context->doCopy(n)) |
| 380 | return -2; |
| 381 | } |
| 382 | |
| 383 | else |
| 384 | context->step = cStepLong; |
| 385 | } |
| 386 | |
| 387 | else |
| 388 | { |
| 389 | if (context->flagMask == 0) |
| 390 | { |
| 391 | context->flags = *_src++; |
| 392 | context->flagMask = 0x80; |
| 393 | if (--srcSize == 0) |
| 394 | break; |
| 395 | } |
| 396 | |
| 397 | if ((context->flags & context->flagMask) == 0) |
| 398 | { |
| 399 | context->packHigh = *_src; |
| 400 | context->step = cStepShort; |
| 401 | } |
| 402 | |
| 403 | else |
| 404 | { |
| 405 | *context->destp++ = *_src; |
| 406 | context->destCount -= 1; |
| 407 | } |
| 408 | |
| 409 | context->flagMask >>= 1; |
| 410 | } |
| 411 | |
| 412 | if (--srcSize == 0) |
| 413 | break; |
| 414 | |
| 415 | _src++; |
| 416 | } |
| 417 | |
| 418 | if (context->destCount == 0 && context->forceDestCount == 0 && 0x20 < srcSize) |
| 419 | return -1; |
| 420 | |
| 421 | else |
| 422 | return context->destCount; |
| 423 | } |
| 424 | |
| 425 | s32 SZSDecompressor::decomp(void* dst, u32 dstSize, const void* src, u32) |
| 426 | { |
| 427 | u32 magic = Endian::toHostU32(from: Endian::cBig, v: BitUtil::bitCastPtr<u32>(value: src)); |
| 428 | if (magic != 0x59617A30) |
| 429 | return -1; |
| 430 | |
| 431 | u32 decompSize = getDecompSize(src); |
| 432 | s32 error = -2; |
| 433 | if (dstSize >= decompSize) |
| 434 | { |
| 435 | #ifdef cafe |
| 436 | error = decodeSZSCafeAsm_(dst, src); |
| 437 | #elif defined(SWITCH) |
| 438 | error = decodeSZSNxAsm64_(dst, src); |
| 439 | #else |
| 440 | SEAD_ASSERT_MSG(false, "SZSDecompressor::decomp not implemented" ); |
| 441 | #endif // cafe |
| 442 | } |
| 443 | |
| 444 | return error; |
| 445 | } |
| 446 | |
| 447 | } // namespace sead |
| 448 | |