diff options
Diffstat (limited to 'vendor/github.com/bytedance/sonic/loader/funcdata_latest.go')
| -rw-r--r-- | vendor/github.com/bytedance/sonic/loader/funcdata_latest.go | 355 |
1 files changed, 355 insertions, 0 deletions
diff --git a/vendor/github.com/bytedance/sonic/loader/funcdata_latest.go b/vendor/github.com/bytedance/sonic/loader/funcdata_latest.go new file mode 100644 index 000000000..52c6097fb --- /dev/null +++ b/vendor/github.com/bytedance/sonic/loader/funcdata_latest.go @@ -0,0 +1,355 @@ +// go:build go1.18 && !go1.22 +// +build go1.18,!go1.22 + +/* + * Copyright 2021 ByteDance Inc. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package loader + +import ( + `os` + `sort` + `unsafe` + + `github.com/bytedance/sonic/internal/rt` +) + +type funcTab struct { + entry uint32 + funcoff uint32 +} + +type pcHeader struct { + magic uint32 // 0xFFFFFFF0 + pad1, pad2 uint8 // 0,0 + minLC uint8 // min instruction size + ptrSize uint8 // size of a ptr in bytes + nfunc int // number of functions in the module + nfiles uint // number of entries in the file tab + textStart uintptr // base for function entry PC offsets in this module, equal to moduledata.text + funcnameOffset uintptr // offset to the funcnametab variable from pcHeader + cuOffset uintptr // offset to the cutab variable from pcHeader + filetabOffset uintptr // offset to the filetab variable from pcHeader + pctabOffset uintptr // offset to the pctab variable from pcHeader + pclnOffset uintptr // offset to the pclntab variable from pcHeader +} + +type bitVector struct { + n int32 // # of bits + bytedata *uint8 +} + +type ptabEntry struct { + name int32 + typ int32 +} + +type textSection struct { + vaddr uintptr // prelinked section vaddr + end uintptr // vaddr + section length + baseaddr uintptr // relocated section address +} + +type modulehash struct { + modulename string + linktimehash string + runtimehash *string +} + +// findfuncbucket is an array of these structures. +// Each bucket represents 4096 bytes of the text segment. +// Each subbucket represents 256 bytes of the text segment. +// To find a function given a pc, locate the bucket and subbucket for +// that pc. Add together the idx and subbucket value to obtain a +// function index. Then scan the functab array starting at that +// index to find the target function. +// This table uses 20 bytes for every 4096 bytes of code, or ~0.5% overhead. +type findfuncbucket struct { + idx uint32 + _SUBBUCKETS [16]byte +} + +type compilationUnit struct { + fileNames []string +} + +func makeFtab(funcs []_func, maxpc uint32) (ftab []funcTab, pclntabSize int64, startLocations []uint32) { + // Allocate space for the pc->func table. This structure consists of a pc offset + // and an offset to the func structure. After that, we have a single pc + // value that marks the end of the last function in the binary. + pclntabSize = int64(len(funcs)*2*int(_PtrSize) + int(_PtrSize)) + startLocations = make([]uint32, len(funcs)) + for i, f := range funcs { + pclntabSize = rnd(pclntabSize, int64(_PtrSize)) + //writePCToFunc + startLocations[i] = uint32(pclntabSize) + pclntabSize += int64(uint8(_FUNC_SIZE)+f.nfuncdata*4+uint8(f.npcdata)*4) + } + + ftab = make([]funcTab, 0, len(funcs)+1) + + // write a map of pc->func info offsets + for i, f := range funcs { + ftab = append(ftab, funcTab{uint32(f.entryOff), uint32(startLocations[i])}) + } + + // Final entry of table is just end pc offset. + ftab = append(ftab, funcTab{maxpc, 0}) + return +} + +// Pcln table format: [...]funcTab + [...]_Func +func makePclntable(size int64, startLocations []uint32, funcs []_func, maxpc uint32, pcdataOffs [][]uint32, funcdataOffs [][]uint32) (pclntab []byte) { + // Allocate space for the pc->func table. This structure consists of a pc offset + // and an offset to the func structure. After that, we have a single pc + // value that marks the end of the last function in the binary. + pclntab = make([]byte, size, size) + + // write a map of pc->func info offsets + offs := 0 + for i, f := range funcs { + byteOrder.PutUint32(pclntab[offs:offs+4], uint32(f.entryOff)) + byteOrder.PutUint32(pclntab[offs+4:offs+8], uint32(startLocations[i])) + offs += 8 + } + // Final entry of table is just end pc offset. + byteOrder.PutUint32(pclntab[offs:offs+4], maxpc) + + // write func info table + for i := range funcs { + off := startLocations[i] + + // write _func structure to pclntab + fb := rt.BytesFrom(unsafe.Pointer(&funcs[i]), int(_FUNC_SIZE), int(_FUNC_SIZE)) + copy(pclntab[off:off+uint32(_FUNC_SIZE)], fb) + off += uint32(_FUNC_SIZE) + + // NOTICE: _func.pcdata always starts from PcUnsafePoint, which is index 3 + for j := 3; j < len(pcdataOffs[i]); j++ { + byteOrder.PutUint32(pclntab[off:off+4], uint32(pcdataOffs[i][j])) + off += 4 + } + + // funcdata refs as offsets from gofunc + for _, funcdata := range funcdataOffs[i] { + byteOrder.PutUint32(pclntab[off:off+4], uint32(funcdata)) + off += 4 + } + + } + + return +} + +// findfunc table used to map pc to belonging func, +// returns the index in the func table. +// +// All text section are divided into buckets sized _BUCKETSIZE(4K): +// every bucket is divided into _SUBBUCKETS sized _SUB_BUCKETSIZE(64), +// and it has a base idx to plus the offset stored in jth subbucket. +// see findfunc() in runtime/symtab.go +func writeFindfunctab(out *[]byte, ftab []funcTab) (start int) { + start = len(*out) + + max := ftab[len(ftab)-1].entry + min := ftab[0].entry + nbuckets := (max - min + _BUCKETSIZE - 1) / _BUCKETSIZE + n := (max - min + _SUB_BUCKETSIZE - 1) / _SUB_BUCKETSIZE + + tab := make([]findfuncbucket, 0, nbuckets) + var s, e = 0, 0 + for i := 0; i<int(nbuckets); i++ { + // store the start s-th func of the bucket + var fb = findfuncbucket{idx: uint32(s)} + + // find the last e-th func of the bucket + var pc = min + uint32((i+1)*_BUCKETSIZE) + for ; e < len(ftab)-1 && ftab[e+1].entry <= pc; e++ {} + + for j := 0; j<_SUBBUCKETS && (i*_SUBBUCKETS+j)<int(n); j++ { + // store the start func of the subbucket + fb._SUBBUCKETS[j] = byte(uint32(s) - fb.idx) + + // find the s-th end func of the subbucket + pc = min + uint32(i*_BUCKETSIZE) + uint32((j+1)*_SUB_BUCKETSIZE) + for ; s < len(ftab)-1 && ftab[s+1].entry <= pc; s++ {} + } + + s = e + tab = append(tab, fb) + } + + // write findfuncbucket + if len(tab) > 0 { + size := int(unsafe.Sizeof(findfuncbucket{}))*len(tab) + *out = append(*out, rt.BytesFrom(unsafe.Pointer(&tab[0]), size, size)...) + } + return +} + +func makeModuledata(name string, filenames []string, funcsp *[]Func, text []byte) (mod *moduledata) { + mod = new(moduledata) + mod.modulename = name + + // sort funcs by entry + funcs := *funcsp + sort.Slice(funcs, func(i, j int) bool { + return funcs[i].EntryOff < funcs[j].EntryOff + }) + *funcsp = funcs + + // make filename table + cu := make([]string, 0, len(filenames)) + cu = append(cu, filenames...) + cutab, filetab, cuOffs := makeFilenametab([]compilationUnit{{cu}}) + mod.cutab = cutab + mod.filetab = filetab + + // make funcname table + funcnametab, nameOffs := makeFuncnameTab(funcs) + mod.funcnametab = funcnametab + + // mmap() text and funcdata segements + p := os.Getpagesize() + size := int(rnd(int64(len(text)), int64(p))) + addr := mmap(size) + // copy the machine code + s := rt.BytesFrom(unsafe.Pointer(addr), len(text), size) + copy(s, text) + // make it executable + mprotect(addr, size) + + // assign addresses + mod.text = addr + mod.etext = addr + uintptr(size) + mod.minpc = addr + mod.maxpc = addr + uintptr(len(text)) + + // make pcdata table + // NOTICE: _func only use offset to index pcdata, thus no need mmap() pcdata + cuOff := cuOffs[0] + pctab, pcdataOffs, _funcs := makePctab(funcs, cuOff, nameOffs) + mod.pctab = pctab + + // write func data + // NOTICE: _func use mod.gofunc+offset to directly point funcdata, thus need cache funcdata + // TODO: estimate accurate capacity + cache := make([]byte, 0, len(funcs)*int(_PtrSize)) + fstart, funcdataOffs := writeFuncdata(&cache, funcs) + + // make pc->func (binary search) func table + ftab, pclntSize, startLocations := makeFtab(_funcs, uint32(len(text))) + mod.ftab = ftab + + // write pc->func (modmap) findfunc table + ffstart := writeFindfunctab(&cache, ftab) + + // cache funcdata and findfuncbucket + moduleCache.Lock() + moduleCache.m[mod] = cache + moduleCache.Unlock() + mod.gofunc = uintptr(unsafe.Pointer(&cache[fstart])) + mod.findfunctab = uintptr(unsafe.Pointer(&cache[ffstart])) + + // make pclnt table + pclntab := makePclntable(pclntSize, startLocations, _funcs, uint32(len(text)), pcdataOffs, funcdataOffs) + mod.pclntable = pclntab + + // make pc header + mod.pcHeader = &pcHeader { + magic : _Magic, + minLC : _MinLC, + ptrSize : _PtrSize, + nfunc : len(funcs), + nfiles: uint(len(cu)), + textStart: mod.text, + funcnameOffset: getOffsetOf(moduledata{}, "funcnametab"), + cuOffset: getOffsetOf(moduledata{}, "cutab"), + filetabOffset: getOffsetOf(moduledata{}, "filetab"), + pctabOffset: getOffsetOf(moduledata{}, "pctab"), + pclnOffset: getOffsetOf(moduledata{}, "pclntable"), + } + + // sepecial case: gcdata and gcbss must by non-empty + mod.gcdata = uintptr(unsafe.Pointer(&emptyByte)) + mod.gcbss = uintptr(unsafe.Pointer(&emptyByte)) + + return +} + +// makePctab generates pcdelta->valuedelta tables for functions, +// and returns the table and the entry offset of every kind pcdata in the table. +func makePctab(funcs []Func, cuOffset uint32, nameOffset []int32) (pctab []byte, pcdataOffs [][]uint32, _funcs []_func) { + _funcs = make([]_func, len(funcs)) + + // Pctab offsets of 0 are considered invalid in the runtime. We respect + // that by just padding a single byte at the beginning of runtime.pctab, + // that way no real offsets can be zero. + pctab = make([]byte, 1, 12*len(funcs)+1) + pcdataOffs = make([][]uint32, len(funcs)) + + for i, f := range funcs { + _f := &_funcs[i] + + var writer = func(pc *Pcdata) { + var ab []byte + var err error + if pc != nil { + ab, err = pc.MarshalBinary() + if err != nil { + panic(err) + } + pcdataOffs[i] = append(pcdataOffs[i], uint32(len(pctab))) + } else { + ab = []byte{0} + pcdataOffs[i] = append(pcdataOffs[i], _PCDATA_INVALID_OFFSET) + } + pctab = append(pctab, ab...) + } + + if f.Pcsp != nil { + _f.pcsp = uint32(len(pctab)) + } + writer(f.Pcsp) + if f.Pcfile != nil { + _f.pcfile = uint32(len(pctab)) + } + writer(f.Pcfile) + if f.Pcline != nil { + _f.pcln = uint32(len(pctab)) + } + writer(f.Pcline) + writer(f.PcUnsafePoint) + writer(f.PcStackMapIndex) + writer(f.PcInlTreeIndex) + writer(f.PcArgLiveIndex) + + _f.entryOff = f.EntryOff + _f.nameOff = nameOffset[i] + _f.args = f.ArgsSize + _f.deferreturn = f.DeferReturn + // NOTICE: _func.pcdata is always as [PCDATA_UnsafePoint(0) : PCDATA_ArgLiveIndex(3)] + _f.npcdata = uint32(_N_PCDATA) + _f.cuOffset = cuOffset + _f.funcID = f.ID + _f.flag = f.Flag + _f.nfuncdata = uint8(_N_FUNCDATA) + } + + return +} + +func registerFunction(name string, pc uintptr, textSize uintptr, fp int, args int, size uintptr, argptrs uintptr, localptrs uintptr) {} |