github.com/ice-blockchain/go/src@v0.0.0-20240403114104-1564d284e521/runtime/mbitmap.go (about) 1 // Copyright 2009 The Go Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 package runtime 6 7 import ( 8 "internal/goarch" 9 "runtime/internal/atomic" 10 "runtime/internal/sys" 11 "unsafe" 12 ) 13 14 // addb returns the byte pointer p+n. 15 // 16 //go:nowritebarrier 17 //go:nosplit 18 func addb(p *byte, n uintptr) *byte { 19 // Note: wrote out full expression instead of calling add(p, n) 20 // to reduce the number of temporaries generated by the 21 // compiler for this trivial expression during inlining. 22 return (*byte)(unsafe.Pointer(uintptr(unsafe.Pointer(p)) + n)) 23 } 24 25 // subtractb returns the byte pointer p-n. 26 // 27 //go:nowritebarrier 28 //go:nosplit 29 func subtractb(p *byte, n uintptr) *byte { 30 // Note: wrote out full expression instead of calling add(p, -n) 31 // to reduce the number of temporaries generated by the 32 // compiler for this trivial expression during inlining. 33 return (*byte)(unsafe.Pointer(uintptr(unsafe.Pointer(p)) - n)) 34 } 35 36 // add1 returns the byte pointer p+1. 37 // 38 //go:nowritebarrier 39 //go:nosplit 40 func add1(p *byte) *byte { 41 // Note: wrote out full expression instead of calling addb(p, 1) 42 // to reduce the number of temporaries generated by the 43 // compiler for this trivial expression during inlining. 44 return (*byte)(unsafe.Pointer(uintptr(unsafe.Pointer(p)) + 1)) 45 } 46 47 // subtract1 returns the byte pointer p-1. 48 // 49 // nosplit because it is used during write barriers and must not be preempted. 50 // 51 //go:nowritebarrier 52 //go:nosplit 53 func subtract1(p *byte) *byte { 54 // Note: wrote out full expression instead of calling subtractb(p, 1) 55 // to reduce the number of temporaries generated by the 56 // compiler for this trivial expression during inlining. 57 return (*byte)(unsafe.Pointer(uintptr(unsafe.Pointer(p)) - 1)) 58 } 59 60 // markBits provides access to the mark bit for an object in the heap. 61 // bytep points to the byte holding the mark bit. 62 // mask is a byte with a single bit set that can be &ed with *bytep 63 // to see if the bit has been set. 64 // *m.byte&m.mask != 0 indicates the mark bit is set. 65 // index can be used along with span information to generate 66 // the address of the object in the heap. 67 // We maintain one set of mark bits for allocation and one for 68 // marking purposes. 69 type markBits struct { 70 bytep *uint8 71 mask uint8 72 index uintptr 73 } 74 75 //go:nosplit 76 func (s *mspan) allocBitsForIndex(allocBitIndex uintptr) markBits { 77 bytep, mask := s.allocBits.bitp(allocBitIndex) 78 return markBits{bytep, mask, allocBitIndex} 79 } 80 81 // refillAllocCache takes 8 bytes s.allocBits starting at whichByte 82 // and negates them so that ctz (count trailing zeros) instructions 83 // can be used. It then places these 8 bytes into the cached 64 bit 84 // s.allocCache. 85 func (s *mspan) refillAllocCache(whichByte uint16) { 86 bytes := (*[8]uint8)(unsafe.Pointer(s.allocBits.bytep(uintptr(whichByte)))) 87 aCache := uint64(0) 88 aCache |= uint64(bytes[0]) 89 aCache |= uint64(bytes[1]) << (1 * 8) 90 aCache |= uint64(bytes[2]) << (2 * 8) 91 aCache |= uint64(bytes[3]) << (3 * 8) 92 aCache |= uint64(bytes[4]) << (4 * 8) 93 aCache |= uint64(bytes[5]) << (5 * 8) 94 aCache |= uint64(bytes[6]) << (6 * 8) 95 aCache |= uint64(bytes[7]) << (7 * 8) 96 s.allocCache = ^aCache 97 } 98 99 // nextFreeIndex returns the index of the next free object in s at 100 // or after s.freeindex. 101 // There are hardware instructions that can be used to make this 102 // faster if profiling warrants it. 103 func (s *mspan) nextFreeIndex() uint16 { 104 sfreeindex := s.freeindex 105 snelems := s.nelems 106 if sfreeindex == snelems { 107 return sfreeindex 108 } 109 if sfreeindex > snelems { 110 throw("s.freeindex > s.nelems") 111 } 112 113 aCache := s.allocCache 114 115 bitIndex := sys.TrailingZeros64(aCache) 116 for bitIndex == 64 { 117 // Move index to start of next cached bits. 118 sfreeindex = (sfreeindex + 64) &^ (64 - 1) 119 if sfreeindex >= snelems { 120 s.freeindex = snelems 121 return snelems 122 } 123 whichByte := sfreeindex / 8 124 // Refill s.allocCache with the next 64 alloc bits. 125 s.refillAllocCache(whichByte) 126 aCache = s.allocCache 127 bitIndex = sys.TrailingZeros64(aCache) 128 // nothing available in cached bits 129 // grab the next 8 bytes and try again. 130 } 131 result := sfreeindex + uint16(bitIndex) 132 if result >= snelems { 133 s.freeindex = snelems 134 return snelems 135 } 136 137 s.allocCache >>= uint(bitIndex + 1) 138 sfreeindex = result + 1 139 140 if sfreeindex%64 == 0 && sfreeindex != snelems { 141 // We just incremented s.freeindex so it isn't 0. 142 // As each 1 in s.allocCache was encountered and used for allocation 143 // it was shifted away. At this point s.allocCache contains all 0s. 144 // Refill s.allocCache so that it corresponds 145 // to the bits at s.allocBits starting at s.freeindex. 146 whichByte := sfreeindex / 8 147 s.refillAllocCache(whichByte) 148 } 149 s.freeindex = sfreeindex 150 return result 151 } 152 153 // isFree reports whether the index'th object in s is unallocated. 154 // 155 // The caller must ensure s.state is mSpanInUse, and there must have 156 // been no preemption points since ensuring this (which could allow a 157 // GC transition, which would allow the state to change). 158 func (s *mspan) isFree(index uintptr) bool { 159 if index < uintptr(s.freeIndexForScan) { 160 return false 161 } 162 bytep, mask := s.allocBits.bitp(index) 163 return *bytep&mask == 0 164 } 165 166 // divideByElemSize returns n/s.elemsize. 167 // n must be within [0, s.npages*_PageSize), 168 // or may be exactly s.npages*_PageSize 169 // if s.elemsize is from sizeclasses.go. 170 // 171 // nosplit, because it is called by objIndex, which is nosplit 172 // 173 //go:nosplit 174 func (s *mspan) divideByElemSize(n uintptr) uintptr { 175 const doubleCheck = false 176 177 // See explanation in mksizeclasses.go's computeDivMagic. 178 q := uintptr((uint64(n) * uint64(s.divMul)) >> 32) 179 180 if doubleCheck && q != n/s.elemsize { 181 println(n, "/", s.elemsize, "should be", n/s.elemsize, "but got", q) 182 throw("bad magic division") 183 } 184 return q 185 } 186 187 // nosplit, because it is called by other nosplit code like findObject 188 // 189 //go:nosplit 190 func (s *mspan) objIndex(p uintptr) uintptr { 191 return s.divideByElemSize(p - s.base()) 192 } 193 194 func markBitsForAddr(p uintptr) markBits { 195 s := spanOf(p) 196 objIndex := s.objIndex(p) 197 return s.markBitsForIndex(objIndex) 198 } 199 200 func (s *mspan) markBitsForIndex(objIndex uintptr) markBits { 201 bytep, mask := s.gcmarkBits.bitp(objIndex) 202 return markBits{bytep, mask, objIndex} 203 } 204 205 func (s *mspan) markBitsForBase() markBits { 206 return markBits{&s.gcmarkBits.x, uint8(1), 0} 207 } 208 209 // isMarked reports whether mark bit m is set. 210 func (m markBits) isMarked() bool { 211 return *m.bytep&m.mask != 0 212 } 213 214 // setMarked sets the marked bit in the markbits, atomically. 215 func (m markBits) setMarked() { 216 // Might be racing with other updates, so use atomic update always. 217 // We used to be clever here and use a non-atomic update in certain 218 // cases, but it's not worth the risk. 219 atomic.Or8(m.bytep, m.mask) 220 } 221 222 // setMarkedNonAtomic sets the marked bit in the markbits, non-atomically. 223 func (m markBits) setMarkedNonAtomic() { 224 *m.bytep |= m.mask 225 } 226 227 // clearMarked clears the marked bit in the markbits, atomically. 228 func (m markBits) clearMarked() { 229 // Might be racing with other updates, so use atomic update always. 230 // We used to be clever here and use a non-atomic update in certain 231 // cases, but it's not worth the risk. 232 atomic.And8(m.bytep, ^m.mask) 233 } 234 235 // markBitsForSpan returns the markBits for the span base address base. 236 func markBitsForSpan(base uintptr) (mbits markBits) { 237 mbits = markBitsForAddr(base) 238 if mbits.mask != 1 { 239 throw("markBitsForSpan: unaligned start") 240 } 241 return mbits 242 } 243 244 // advance advances the markBits to the next object in the span. 245 func (m *markBits) advance() { 246 if m.mask == 1<<7 { 247 m.bytep = (*uint8)(unsafe.Pointer(uintptr(unsafe.Pointer(m.bytep)) + 1)) 248 m.mask = 1 249 } else { 250 m.mask = m.mask << 1 251 } 252 m.index++ 253 } 254 255 // clobberdeadPtr is a special value that is used by the compiler to 256 // clobber dead stack slots, when -clobberdead flag is set. 257 const clobberdeadPtr = uintptr(0xdeaddead | 0xdeaddead<<((^uintptr(0)>>63)*32)) 258 259 // badPointer throws bad pointer in heap panic. 260 func badPointer(s *mspan, p, refBase, refOff uintptr) { 261 // Typically this indicates an incorrect use 262 // of unsafe or cgo to store a bad pointer in 263 // the Go heap. It may also indicate a runtime 264 // bug. 265 // 266 // TODO(austin): We could be more aggressive 267 // and detect pointers to unallocated objects 268 // in allocated spans. 269 printlock() 270 print("runtime: pointer ", hex(p)) 271 if s != nil { 272 state := s.state.get() 273 if state != mSpanInUse { 274 print(" to unallocated span") 275 } else { 276 print(" to unused region of span") 277 } 278 print(" span.base()=", hex(s.base()), " span.limit=", hex(s.limit), " span.state=", state) 279 } 280 print("\n") 281 if refBase != 0 { 282 print("runtime: found in object at *(", hex(refBase), "+", hex(refOff), ")\n") 283 gcDumpObject("object", refBase, refOff) 284 } 285 getg().m.traceback = 2 286 throw("found bad pointer in Go heap (incorrect use of unsafe or cgo?)") 287 } 288 289 // findObject returns the base address for the heap object containing 290 // the address p, the object's span, and the index of the object in s. 291 // If p does not point into a heap object, it returns base == 0. 292 // 293 // If p points is an invalid heap pointer and debug.invalidptr != 0, 294 // findObject panics. 295 // 296 // refBase and refOff optionally give the base address of the object 297 // in which the pointer p was found and the byte offset at which it 298 // was found. These are used for error reporting. 299 // 300 // It is nosplit so it is safe for p to be a pointer to the current goroutine's stack. 301 // Since p is a uintptr, it would not be adjusted if the stack were to move. 302 // 303 //go:nosplit 304 func findObject(p, refBase, refOff uintptr) (base uintptr, s *mspan, objIndex uintptr) { 305 s = spanOf(p) 306 // If s is nil, the virtual address has never been part of the heap. 307 // This pointer may be to some mmap'd region, so we allow it. 308 if s == nil { 309 if (GOARCH == "amd64" || GOARCH == "arm64") && p == clobberdeadPtr && debug.invalidptr != 0 { 310 // Crash if clobberdeadPtr is seen. Only on AMD64 and ARM64 for now, 311 // as they are the only platform where compiler's clobberdead mode is 312 // implemented. On these platforms clobberdeadPtr cannot be a valid address. 313 badPointer(s, p, refBase, refOff) 314 } 315 return 316 } 317 // If p is a bad pointer, it may not be in s's bounds. 318 // 319 // Check s.state to synchronize with span initialization 320 // before checking other fields. See also spanOfHeap. 321 if state := s.state.get(); state != mSpanInUse || p < s.base() || p >= s.limit { 322 // Pointers into stacks are also ok, the runtime manages these explicitly. 323 if state == mSpanManual { 324 return 325 } 326 // The following ensures that we are rigorous about what data 327 // structures hold valid pointers. 328 if debug.invalidptr != 0 { 329 badPointer(s, p, refBase, refOff) 330 } 331 return 332 } 333 334 objIndex = s.objIndex(p) 335 base = s.base() + objIndex*s.elemsize 336 return 337 } 338 339 // reflect_verifyNotInHeapPtr reports whether converting the not-in-heap pointer into a unsafe.Pointer is ok. 340 // 341 //go:linkname reflect_verifyNotInHeapPtr reflect.verifyNotInHeapPtr 342 func reflect_verifyNotInHeapPtr(p uintptr) bool { 343 // Conversion to a pointer is ok as long as findObject above does not call badPointer. 344 // Since we're already promised that p doesn't point into the heap, just disallow heap 345 // pointers and the special clobbered pointer. 346 return spanOf(p) == nil && p != clobberdeadPtr 347 } 348 349 const ptrBits = 8 * goarch.PtrSize 350 351 // bulkBarrierBitmap executes write barriers for copying from [src, 352 // src+size) to [dst, dst+size) using a 1-bit pointer bitmap. src is 353 // assumed to start maskOffset bytes into the data covered by the 354 // bitmap in bits (which may not be a multiple of 8). 355 // 356 // This is used by bulkBarrierPreWrite for writes to data and BSS. 357 // 358 //go:nosplit 359 func bulkBarrierBitmap(dst, src, size, maskOffset uintptr, bits *uint8) { 360 word := maskOffset / goarch.PtrSize 361 bits = addb(bits, word/8) 362 mask := uint8(1) << (word % 8) 363 364 buf := &getg().m.p.ptr().wbBuf 365 for i := uintptr(0); i < size; i += goarch.PtrSize { 366 if mask == 0 { 367 bits = addb(bits, 1) 368 if *bits == 0 { 369 // Skip 8 words. 370 i += 7 * goarch.PtrSize 371 continue 372 } 373 mask = 1 374 } 375 if *bits&mask != 0 { 376 dstx := (*uintptr)(unsafe.Pointer(dst + i)) 377 if src == 0 { 378 p := buf.get1() 379 p[0] = *dstx 380 } else { 381 srcx := (*uintptr)(unsafe.Pointer(src + i)) 382 p := buf.get2() 383 p[0] = *dstx 384 p[1] = *srcx 385 } 386 } 387 mask <<= 1 388 } 389 } 390 391 // typeBitsBulkBarrier executes a write barrier for every 392 // pointer that would be copied from [src, src+size) to [dst, 393 // dst+size) by a memmove using the type bitmap to locate those 394 // pointer slots. 395 // 396 // The type typ must correspond exactly to [src, src+size) and [dst, dst+size). 397 // dst, src, and size must be pointer-aligned. 398 // The type typ must have a plain bitmap, not a GC program. 399 // The only use of this function is in channel sends, and the 400 // 64 kB channel element limit takes care of this for us. 401 // 402 // Must not be preempted because it typically runs right before memmove, 403 // and the GC must observe them as an atomic action. 404 // 405 // Callers must perform cgo checks if goexperiment.CgoCheck2. 406 // 407 //go:nosplit 408 func typeBitsBulkBarrier(typ *_type, dst, src, size uintptr) { 409 if typ == nil { 410 throw("runtime: typeBitsBulkBarrier without type") 411 } 412 if typ.Size_ != size { 413 println("runtime: typeBitsBulkBarrier with type ", toRType(typ).string(), " of size ", typ.Size_, " but memory size", size) 414 throw("runtime: invalid typeBitsBulkBarrier") 415 } 416 if typ.Kind_&kindGCProg != 0 { 417 println("runtime: typeBitsBulkBarrier with type ", toRType(typ).string(), " with GC prog") 418 throw("runtime: invalid typeBitsBulkBarrier") 419 } 420 if !writeBarrier.enabled { 421 return 422 } 423 ptrmask := typ.GCData 424 buf := &getg().m.p.ptr().wbBuf 425 var bits uint32 426 for i := uintptr(0); i < typ.PtrBytes; i += goarch.PtrSize { 427 if i&(goarch.PtrSize*8-1) == 0 { 428 bits = uint32(*ptrmask) 429 ptrmask = addb(ptrmask, 1) 430 } else { 431 bits = bits >> 1 432 } 433 if bits&1 != 0 { 434 dstx := (*uintptr)(unsafe.Pointer(dst + i)) 435 srcx := (*uintptr)(unsafe.Pointer(src + i)) 436 p := buf.get2() 437 p[0] = *dstx 438 p[1] = *srcx 439 } 440 } 441 } 442 443 // countAlloc returns the number of objects allocated in span s by 444 // scanning the mark bitmap. 445 func (s *mspan) countAlloc() int { 446 count := 0 447 bytes := divRoundUp(uintptr(s.nelems), 8) 448 // Iterate over each 8-byte chunk and count allocations 449 // with an intrinsic. Note that newMarkBits guarantees that 450 // gcmarkBits will be 8-byte aligned, so we don't have to 451 // worry about edge cases, irrelevant bits will simply be zero. 452 for i := uintptr(0); i < bytes; i += 8 { 453 // Extract 64 bits from the byte pointer and get a OnesCount. 454 // Note that the unsafe cast here doesn't preserve endianness, 455 // but that's OK. We only care about how many bits are 1, not 456 // about the order we discover them in. 457 mrkBits := *(*uint64)(unsafe.Pointer(s.gcmarkBits.bytep(i))) 458 count += sys.OnesCount64(mrkBits) 459 } 460 return count 461 } 462 463 // Read the bytes starting at the aligned pointer p into a uintptr. 464 // Read is little-endian. 465 func readUintptr(p *byte) uintptr { 466 x := *(*uintptr)(unsafe.Pointer(p)) 467 if goarch.BigEndian { 468 if goarch.PtrSize == 8 { 469 return uintptr(sys.Bswap64(uint64(x))) 470 } 471 return uintptr(sys.Bswap32(uint32(x))) 472 } 473 return x 474 } 475 476 var debugPtrmask struct { 477 lock mutex 478 data *byte 479 } 480 481 // progToPointerMask returns the 1-bit pointer mask output by the GC program prog. 482 // size the size of the region described by prog, in bytes. 483 // The resulting bitvector will have no more than size/goarch.PtrSize bits. 484 func progToPointerMask(prog *byte, size uintptr) bitvector { 485 n := (size/goarch.PtrSize + 7) / 8 486 x := (*[1 << 30]byte)(persistentalloc(n+1, 1, &memstats.buckhash_sys))[:n+1] 487 x[len(x)-1] = 0xa1 // overflow check sentinel 488 n = runGCProg(prog, &x[0]) 489 if x[len(x)-1] != 0xa1 { 490 throw("progToPointerMask: overflow") 491 } 492 return bitvector{int32(n), &x[0]} 493 } 494 495 // Packed GC pointer bitmaps, aka GC programs. 496 // 497 // For large types containing arrays, the type information has a 498 // natural repetition that can be encoded to save space in the 499 // binary and in the memory representation of the type information. 500 // 501 // The encoding is a simple Lempel-Ziv style bytecode machine 502 // with the following instructions: 503 // 504 // 00000000: stop 505 // 0nnnnnnn: emit n bits copied from the next (n+7)/8 bytes 506 // 10000000 n c: repeat the previous n bits c times; n, c are varints 507 // 1nnnnnnn c: repeat the previous n bits c times; c is a varint 508 509 // runGCProg returns the number of 1-bit entries written to memory. 510 func runGCProg(prog, dst *byte) uintptr { 511 dstStart := dst 512 513 // Bits waiting to be written to memory. 514 var bits uintptr 515 var nbits uintptr 516 517 p := prog 518 Run: 519 for { 520 // Flush accumulated full bytes. 521 // The rest of the loop assumes that nbits <= 7. 522 for ; nbits >= 8; nbits -= 8 { 523 *dst = uint8(bits) 524 dst = add1(dst) 525 bits >>= 8 526 } 527 528 // Process one instruction. 529 inst := uintptr(*p) 530 p = add1(p) 531 n := inst & 0x7F 532 if inst&0x80 == 0 { 533 // Literal bits; n == 0 means end of program. 534 if n == 0 { 535 // Program is over. 536 break Run 537 } 538 nbyte := n / 8 539 for i := uintptr(0); i < nbyte; i++ { 540 bits |= uintptr(*p) << nbits 541 p = add1(p) 542 *dst = uint8(bits) 543 dst = add1(dst) 544 bits >>= 8 545 } 546 if n %= 8; n > 0 { 547 bits |= uintptr(*p) << nbits 548 p = add1(p) 549 nbits += n 550 } 551 continue Run 552 } 553 554 // Repeat. If n == 0, it is encoded in a varint in the next bytes. 555 if n == 0 { 556 for off := uint(0); ; off += 7 { 557 x := uintptr(*p) 558 p = add1(p) 559 n |= (x & 0x7F) << off 560 if x&0x80 == 0 { 561 break 562 } 563 } 564 } 565 566 // Count is encoded in a varint in the next bytes. 567 c := uintptr(0) 568 for off := uint(0); ; off += 7 { 569 x := uintptr(*p) 570 p = add1(p) 571 c |= (x & 0x7F) << off 572 if x&0x80 == 0 { 573 break 574 } 575 } 576 c *= n // now total number of bits to copy 577 578 // If the number of bits being repeated is small, load them 579 // into a register and use that register for the entire loop 580 // instead of repeatedly reading from memory. 581 // Handling fewer than 8 bits here makes the general loop simpler. 582 // The cutoff is goarch.PtrSize*8 - 7 to guarantee that when we add 583 // the pattern to a bit buffer holding at most 7 bits (a partial byte) 584 // it will not overflow. 585 src := dst 586 const maxBits = goarch.PtrSize*8 - 7 587 if n <= maxBits { 588 // Start with bits in output buffer. 589 pattern := bits 590 npattern := nbits 591 592 // If we need more bits, fetch them from memory. 593 src = subtract1(src) 594 for npattern < n { 595 pattern <<= 8 596 pattern |= uintptr(*src) 597 src = subtract1(src) 598 npattern += 8 599 } 600 601 // We started with the whole bit output buffer, 602 // and then we loaded bits from whole bytes. 603 // Either way, we might now have too many instead of too few. 604 // Discard the extra. 605 if npattern > n { 606 pattern >>= npattern - n 607 npattern = n 608 } 609 610 // Replicate pattern to at most maxBits. 611 if npattern == 1 { 612 // One bit being repeated. 613 // If the bit is 1, make the pattern all 1s. 614 // If the bit is 0, the pattern is already all 0s, 615 // but we can claim that the number of bits 616 // in the word is equal to the number we need (c), 617 // because right shift of bits will zero fill. 618 if pattern == 1 { 619 pattern = 1<<maxBits - 1 620 npattern = maxBits 621 } else { 622 npattern = c 623 } 624 } else { 625 b := pattern 626 nb := npattern 627 if nb+nb <= maxBits { 628 // Double pattern until the whole uintptr is filled. 629 for nb <= goarch.PtrSize*8 { 630 b |= b << nb 631 nb += nb 632 } 633 // Trim away incomplete copy of original pattern in high bits. 634 // TODO(rsc): Replace with table lookup or loop on systems without divide? 635 nb = maxBits / npattern * npattern 636 b &= 1<<nb - 1 637 pattern = b 638 npattern = nb 639 } 640 } 641 642 // Add pattern to bit buffer and flush bit buffer, c/npattern times. 643 // Since pattern contains >8 bits, there will be full bytes to flush 644 // on each iteration. 645 for ; c >= npattern; c -= npattern { 646 bits |= pattern << nbits 647 nbits += npattern 648 for nbits >= 8 { 649 *dst = uint8(bits) 650 dst = add1(dst) 651 bits >>= 8 652 nbits -= 8 653 } 654 } 655 656 // Add final fragment to bit buffer. 657 if c > 0 { 658 pattern &= 1<<c - 1 659 bits |= pattern << nbits 660 nbits += c 661 } 662 continue Run 663 } 664 665 // Repeat; n too large to fit in a register. 666 // Since nbits <= 7, we know the first few bytes of repeated data 667 // are already written to memory. 668 off := n - nbits // n > nbits because n > maxBits and nbits <= 7 669 // Leading src fragment. 670 src = subtractb(src, (off+7)/8) 671 if frag := off & 7; frag != 0 { 672 bits |= uintptr(*src) >> (8 - frag) << nbits 673 src = add1(src) 674 nbits += frag 675 c -= frag 676 } 677 // Main loop: load one byte, write another. 678 // The bits are rotating through the bit buffer. 679 for i := c / 8; i > 0; i-- { 680 bits |= uintptr(*src) << nbits 681 src = add1(src) 682 *dst = uint8(bits) 683 dst = add1(dst) 684 bits >>= 8 685 } 686 // Final src fragment. 687 if c %= 8; c > 0 { 688 bits |= (uintptr(*src) & (1<<c - 1)) << nbits 689 nbits += c 690 } 691 } 692 693 // Write any final bits out, using full-byte writes, even for the final byte. 694 totalBits := (uintptr(unsafe.Pointer(dst))-uintptr(unsafe.Pointer(dstStart)))*8 + nbits 695 nbits += -nbits & 7 696 for ; nbits > 0; nbits -= 8 { 697 *dst = uint8(bits) 698 dst = add1(dst) 699 bits >>= 8 700 } 701 return totalBits 702 } 703 704 // materializeGCProg allocates space for the (1-bit) pointer bitmask 705 // for an object of size ptrdata. Then it fills that space with the 706 // pointer bitmask specified by the program prog. 707 // The bitmask starts at s.startAddr. 708 // The result must be deallocated with dematerializeGCProg. 709 func materializeGCProg(ptrdata uintptr, prog *byte) *mspan { 710 // Each word of ptrdata needs one bit in the bitmap. 711 bitmapBytes := divRoundUp(ptrdata, 8*goarch.PtrSize) 712 // Compute the number of pages needed for bitmapBytes. 713 pages := divRoundUp(bitmapBytes, pageSize) 714 s := mheap_.allocManual(pages, spanAllocPtrScalarBits) 715 runGCProg(addb(prog, 4), (*byte)(unsafe.Pointer(s.startAddr))) 716 return s 717 } 718 func dematerializeGCProg(s *mspan) { 719 mheap_.freeManual(s, spanAllocPtrScalarBits) 720 } 721 722 func dumpGCProg(p *byte) { 723 nptr := 0 724 for { 725 x := *p 726 p = add1(p) 727 if x == 0 { 728 print("\t", nptr, " end\n") 729 break 730 } 731 if x&0x80 == 0 { 732 print("\t", nptr, " lit ", x, ":") 733 n := int(x+7) / 8 734 for i := 0; i < n; i++ { 735 print(" ", hex(*p)) 736 p = add1(p) 737 } 738 print("\n") 739 nptr += int(x) 740 } else { 741 nbit := int(x &^ 0x80) 742 if nbit == 0 { 743 for nb := uint(0); ; nb += 7 { 744 x := *p 745 p = add1(p) 746 nbit |= int(x&0x7f) << nb 747 if x&0x80 == 0 { 748 break 749 } 750 } 751 } 752 count := 0 753 for nb := uint(0); ; nb += 7 { 754 x := *p 755 p = add1(p) 756 count |= int(x&0x7f) << nb 757 if x&0x80 == 0 { 758 break 759 } 760 } 761 print("\t", nptr, " repeat ", nbit, " × ", count, "\n") 762 nptr += nbit * count 763 } 764 } 765 } 766 767 // Testing. 768 769 // reflect_gcbits returns the GC type info for x, for testing. 770 // The result is the bitmap entries (0 or 1), one entry per byte. 771 // 772 //go:linkname reflect_gcbits reflect.gcbits 773 func reflect_gcbits(x any) []byte { 774 return getgcmask(x) 775 }