golang并发安全-select

前面说了golang的channel，今天我们看看golang select 是怎么实现的。

数据结构

type scase struct {c    *hchan         // chanelem unsafe.Pointer // 数据
}

select 非默认的case 中都是处理channel 的接受和发送，所有scase 结构体中c是用来存储select 的case中使用的channel

处理流程

select case 场景

编译器在中间代码生成期间会根据 select 中 case 的不同对控制语句进行优化，这一过程都发生在cmd/compile/internal/walk/select.go 中，下面会根据不同的场景进行分析代码。

没有case

代码示例

func main() {select {}
}

如果是空的select语句，程序会被阻塞，golang 带有死锁监测机制：如果当前写成无法被唤醒，则会panic

源码解读

在runtime/select.go中可以看到：如果cases为空直接调用gopark函数以waitReasonSelectNoCases的原因挂起当前的协程，并且无法被唤醒，golang监测到直接panic。

同样我们在walk/select.go的walkSelectCases函数中可以看到，如果case为空直接调用runtime.block函数

只有一个case

代码示例

func main() {ch := make(chan int)go func() {ch <- 1}()select {case data := <-ch:fmt.Println("ch data:", data)}
}

如果有输入直接打印ch data : 1 , 没有的话会被检测出all goroutines are asleep - deadlock!（和没有case的一样）

源码解读

如果一个非default case ，将读写转换成 ch <- 或 <- ch，正常的channel读写

func walkSelectCases(cases []*ir.CommClause) []ir.Node {// optimization: one-case select: single op.if ncas == 1 {cas := cases[0] //获取caseir.SetPos(cas)l := cas.Init()if cas.Comm != nil { // 不是默认n := cas.Comm // 获取case的条件语句l = append(l, ir.TakeInit(n)...)switch n.Op() {default:base.Fatalf("select %v", n.Op())case ir.OSEND: // 如果是 send， 无须处理// already okcase ir.OSELRECV2:r := n.(*ir.AssignListStmt)// 如果不是 data, ok := <- ch 类型，处理成<- chif ir.IsBlank(r.Lhs[0]) && ir.IsBlank(r.Lhs[1]) {n = r.Rhs[0]break}// 是的话， op设置成data, ok := <- ch形式r.SetOp(ir.OAS2RECV)}l = append(l, n)}// 将case 条件后要执行的语句加入带执行的列表l = append(l, cas.Body...)// 加入 break类型，跳出select-case l = append(l, ir.NewBranchStmt(base.Pos, ir.OBREAK, nil))return l}// convert case value arguments to addresses.// this rewrite is used by both the general code and the next optimization.var dflt *ir.CommClausefor _, cas := range cases {ir.SetPos(cas)n := cas.Commif n == nil {dflt = cascontinue}switch n.Op() {case ir.OSEND:n := n.(*ir.SendStmt)n.Value = typecheck.NodAddr(n.Value)n.Value = typecheck.Expr(n.Value)case ir.OSELRECV2:n := n.(*ir.AssignListStmt)if !ir.IsBlank(n.Lhs[0]) {n.Lhs[0] = typecheck.NodAddr(n.Lhs[0])n.Lhs[0] = typecheck.Expr(n.Lhs[0])}}}
}

两个case（一个default）

代码示例

func main() {ch := make(chan int)select {case data := <-ch:fmt.Println("ch data:", data)default:fmt.Println("default")}
}

如果写入就走<- 读取，反之走默认

源码解读

如果是两个case，其中一个是default，非default的会根据send还是recv 调用channel的selectnbsend和 selectnbrecv。这两个方法是非阻塞的

func walkSelectCases(cases []*ir.CommClause) []ir.Node){// optimization: two-case select but one is default: single non-blocking op.if ncas == 2 && dflt != nil {cas := cases[0]if cas == dflt { // 如果是default 放在 cases[1]cas = cases[1]}n := cas.Commir.SetPos(n)r := ir.NewIfStmt(base.Pos, nil, nil, nil)r.SetInit(cas.Init())var cond ir.Nodeswitch n.Op() {default:base.Fatalf("select %v", n.Op())case ir.OSEND:// 调用selectnbsend(c, v)// if selectnbsend(c, v) { body } else { default body }n := n.(*ir.SendStmt)ch := n.Chancond = mkcall1(chanfn("selectnbsend", 2, ch.Type()), types.Types[types.TBOOL], r.PtrInit(), ch, n.Value)case ir.OSELRECV2:n := n.(*ir.AssignListStmt)recv := n.Rhs[0].(*ir.UnaryExpr)ch := recv.Xelem := n.Lhs[0]if ir.IsBlank(elem) { //空的话 elem= NodNilelem = typecheck.NodNil()}cond = typecheck.Temp(types.Types[types.TBOOL])// 调用 selectnbrecvfn := chanfn("selectnbrecv", 2, ch.Type())call := mkcall1(fn, fn.Type().Results(), r.PtrInit(), elem, ch)as := ir.NewAssignListStmt(r.Pos(), ir.OAS2, []ir.Node{cond, n.Lhs[1]}, []ir.Node{call})r.PtrInit().Append(typecheck.Stmt(as))}r.Cond = typecheck.Expr(cond)r.Body = cas.Bodyr.Else = append(dflt.Init(), dflt.Body...)return []ir.Node{r, ir.NewBranchStmt(base.Pos, ir.OBREAK, nil)}}    
}

每次尝试从channel读/写值，如果不成功则直接返回，不会阻塞。从selectnbsend和selectnbrecv看出，最后转换成if-else

// compiler implements
//
//	select {
//	case c <- v:
//		... foo
//	default:
//		... bar
//	}
//
// as
//
//	if selectnbsend(c, v) {
//		... foo
//	} else {
//		... bar
//	}
//
func selectnbsend(c *hchan, elem unsafe.Pointer) (selected bool) {// block:false// chan将 select准换if-elsereturn chansend(c, elem, false, getcallerpc())
}// compiler implements
//
//	select {
//	case v, ok = <-c:
//		... foo
//	default:
//		... bar
//	}
//
// as
//
//	if selected, ok = selectnbrecv(&v, c); selected {
//		... foo
//	} else {
//		... bar
//	}
//
func selectnbrecv(elem unsafe.Pointer, c *hchan) (selected, received bool) {// block:false// chan将 select准换if-elsereturn chanrecv(c, elem, false)
}

多个case

代码示例

func main() {ch := make(chan int)go func() {tempArr := []int{1,2,3,4,5,6}for i := range tempArr {ch <- i}}()go func() {for {select {case i := <-ch:println("first: ", i)case i := <-ch:println("second", i)}}}()time.Sleep(3 * time.Second)}

可以看到多个case，会随机选取一个case执行

源码解读

func walkSelectCases(cases []*ir.CommClause) []ir.Node {ncas := len(cases)sellineno := base.Posif dflt != nil {ncas--}// 定义casorder为ncas大小的case语句的数组casorder := make([]*ir.CommClause, ncas)// 分别定义nsends为发送channel的case个数，nrecvs为接收channel的case个数nsends, nrecvs := 0, 0// 多case编译后待执行的语句列表var init []ir.Node// generate sel-structbase.Pos = sellineno// 定义selv为长度为ncas的scase类型的数组// scasetype()函数返回的就是scase结构体，包含c和elem两个字段selv := typecheck.Temp(types.NewArray(scasetype(), int64(ncas)))init = append(init, typecheck.Stmt(ir.NewAssignStmt(base.Pos, selv, nil)))// No initialization for order; runtime.selectgo is responsible for that.// 定义order为2倍的ncas长度的TUINT16类型的数组// 注意：selv和order作为runtime.selectgo()函数的入参，前者存放scase列表内存地址，后者用来做scase排序使用，排序是为了便于挑选出待执行的caseorder := typecheck.Temp(types.NewArray(types.Types[types.TUINT16], 2*int64(ncas)))var pc0, pcs ir.Nodeif base.Flag.Race {pcs = typecheck.Temp(types.NewArray(types.Types[types.TUINTPTR], int64(ncas)))pc0 = typecheck.Expr(typecheck.NodAddr(ir.NewIndexExpr(base.Pos, pcs, ir.NewInt(0))))} else {pc0 = typecheck.NodNil()}// register cases 遍历case生成scase对象放到selv中for _, cas := range cases {ir.SetPos(cas)init = append(init, ir.TakeInit(cas)...)n := cas.Commif n == nil { // default:continue}var i intvar c, elem ir.Nodeswitch n.Op() { // 根据类型获取chan, elem的值default:base.Fatalf("select %v", n.Op())case ir.OSEND: // 发送chan类型，i从0开始递增n := n.(*ir.SendStmt)i = nsendsnsends++c = n.Chanelem = n.Valuecase ir.OSELRECV2: // 接收chan，i从ncas开始递减n := n.(*ir.AssignListStmt)nrecvs++i = ncas - nrecvsrecv := n.Rhs[0].(*ir.UnaryExpr)c = recv.Xelem = n.Lhs[0]}casorder[i] = cas// 定义一个函数，写入c或elem到selv数组setField := func(f string, val ir.Node) {// 放到selv数组r := ir.NewAssignStmt(base.Pos, ir.NewSelectorExpr(base.Pos, ir.ODOT, ir.NewIndexExpr(base.Pos, selv, ir.NewInt(int64(i))), typecheck.Lookup(f)), val)// 添加到带执行列表init = append(init, typecheck.Stmt(r))}c = typecheck.ConvNop(c, types.Types[types.TUNSAFEPTR])setField("c", c)if !ir.IsBlank(elem) {elem = typecheck.ConvNop(elem, types.Types[types.TUNSAFEPTR])setField("elem", elem)}// TODO(mdempsky): There should be a cleaner way to// handle this.if base.Flag.Race {r := mkcallstmt("selectsetpc", typecheck.NodAddr(ir.NewIndexExpr(base.Pos, pcs, ir.NewInt(int64(i)))))init = append(init, r)}}// 如果发送chan和接收chan的个数不等于ncas，直接报错if nsends+nrecvs != ncas {base.Fatalf("walkSelectCases: miscount: %v + %v != %v", nsends, nrecvs, ncas)}// run the select  开始执行select动作base.Pos = sellineno// 定义chosen, recvOK作为selectgo()函数的两个返回值// chosen 表示被选中的case的索引，recvOK表示对于接收操作，是否成功接收chosen := typecheck.Temp(types.Types[types.TINT])recvOK := typecheck.Temp(types.Types[types.TBOOL])r := ir.NewAssignListStmt(base.Pos, ir.OAS2, nil, nil)r.Lhs = []ir.Node{chosen, recvOK}// 调用runtime.selectgo()函数作为运行时实际执行多case的select动作的函数fn := typecheck.LookupRuntime("selectgo")var fnInit ir.Nodesr.Rhs = []ir.Node{mkcall1(fn, fn.Type().Results(), &fnInit, bytePtrToIndex(selv, 0), bytePtrToIndex(order, 0), pc0, ir.NewInt(int64(nsends)), ir.NewInt(int64(nrecvs)), ir.NewBool(dflt == nil))}init = append(init, fnInit...)init = append(init, typecheck.Stmt(r))// selv and order are no longer alive after selectgo.// 执行完selectgo()函数后，销毁selv和order数组.init = append(init, ir.NewUnaryExpr(base.Pos, ir.OVARKILL, selv))init = append(init, ir.NewUnaryExpr(base.Pos, ir.OVARKILL, order))if base.Flag.Race {init = append(init, ir.NewUnaryExpr(base.Pos, ir.OVARKILL, pcs))}// dispatch cases //定义一个函数，根据chosen确定的case分支生成if语句，执行该分支的语句dispatch := func(cond ir.Node, cas *ir.CommClause) {cond = typecheck.Expr(cond)cond = typecheck.DefaultLit(cond, nil)r := ir.NewIfStmt(base.Pos, cond, nil, nil)if n := cas.Comm; n != nil && n.Op() == ir.OSELRECV2 {n := n.(*ir.AssignListStmt)if !ir.IsBlank(n.Lhs[1]) {x := ir.NewAssignStmt(base.Pos, n.Lhs[1], recvOK)r.Body.Append(typecheck.Stmt(x))}}r.Body.Append(cas.Body.Take()...)r.Body.Append(ir.NewBranchStmt(base.Pos, ir.OBREAK, nil))init = append(init, r)}// 如果多case中有default分支，并且chosen小于0，执行该default分支if dflt != nil {ir.SetPos(dflt)dispatch(ir.NewBinaryExpr(base.Pos, ir.OLT, chosen, ir.NewInt(0)), dflt)}// 如果有chosen选中的case分支，即chosen等于i，则执行该分支for i, cas := range casorder {ir.SetPos(cas)dispatch(ir.NewBinaryExpr(base.Pos, ir.OEQ, chosen, ir.NewInt(int64(i))), cas)}return init
}

从上面代码可以看出：

1- 初始化过程: 生成scase数组,定义selv 存放scase数组内存地址，定义order 来给scase排序

2- 遍历所有的case ，将case放到带执行列表（不包括default）

3- 调用runtime。selectgo并将selv和order作为入参传入selectgo

4- 根据selectgo返回的chosen来生成if语句，执行对应的case

解锁加锁

加锁的顺序和解锁的顺序相反。


func sellock(scases []scase, lockorder []uint16) {var c *hchanfor _, o := range lockorder {c0 := scases[o].cif c0 != c {c = c0lock(&c.lock)}}
}func selunlock(scases []scase, lockorder []uint16) {// 我们必须非常小心，在解锁最后一把锁后不要触摸sel，因为sel可以在最后一次解锁后立即释放。//考虑以下情况。第一个M调用runtime·park（）在runtime·selectgo（）中传递sel。//一旦runtime·park（）解锁了最后一个锁，另一个M会使调用select的G再次可运行，//并安排其执行。当G在另一个M上运行时，它锁定所有锁并释放sel。现在，如果第一个M触摸sel，它将访问释放的内存。for i := len(lockorder) - 1; i >= 0; i-- {c := scases[lockorder[i]].cif i > 0 && c == scases[lockorder[i-1]].c {continue // will unlock it on the next iteration}unlock(&c.lock)}
}

selectgo

selectgo 处理逻辑

// cas0指向[ncases]scase类型的数组，order0指向[2*ncases]uint16类型的数组（其中ncases必须<=65536）。
// 返回值有两个, chosen 和 recvOK，分别表示选中的case的序号，和对接收操作是否接收成功的布尔值
func selectgo(cas0 *scase, order0 *uint16, pc0 *uintptr, nsends, nrecvs int, block bool) (int, bool) {if debugSelect {print("select: cas0=", cas0, "\n")}//==== 执行必要的初始化操作，并生成处理case的两种顺序：轮询顺序polIorder和加锁顺序lockorder。// 为了将scase分配到栈上，这里直接给cas1分配了64KB大小的数组，同理， 给order1分配了128KB大小的数组// NOTE: In order to maintain a lean stack size, the number of scases// is capped at 65536.cas1 := (*[1 << 16]scase)(unsafe.Pointer(cas0))order1 := (*[1 << 17]uint16)(unsafe.Pointer(order0))// ncases个数 = 发送chan个数+ 接收chan个数ncases := nsends + nrecvs// scases是cas1数组的前ncases个元素scases := cas1[:ncases:ncases]// 顺序列表pollorder是order1的0- ncases个元素pollorder := order1[:ncases:ncases]// 加锁列表lockorder是order1的ncase到 2 ncases 个元素lockorder := order1[ncases:][:ncases:ncases]// NOTE: 编译器初始化的pollorder/lockorder的基础数组不是零。// Even when raceenabled is true, there might be select// statements in packages compiled without -race (e.g.,// ensureSigM in runtime/signal_unix.go).var pcs []uintptrif raceenabled && pc0 != nil {pc1 := (*[1 << 16]uintptr)(unsafe.Pointer(pc0))pcs = pc1[:ncases:ncases]}casePC := func(casi int) uintptr {if pcs == nil {return 0}return pcs[casi]}var t0 int64if blockprofilerate > 0 {t0 = cputicks()}// 生成排列顺序norder := 0for i := range scases {cas := &scases[i]// Omit cases without channels from the poll and lock orders.// 处理case中channel为空的情况if cas.c == nil {cas.elem = nil // 便于GCcontinue}// 通过fastrandn函数引入随机性，确定pollorder列表中case的随机顺序索引j := fastrandn(uint32(norder + 1))pollorder[norder] = pollorder[j]pollorder[j] = uint16(i)norder++}// 重新生成列表pollorder = pollorder[:norder]lockorder = lockorder[:norder]// 根据chan地址确定lockorder加锁排序列表的顺序// 简单的堆排序，以保证nlogn时间复杂度完成排序for i := range lockorder {j := i// 从轮询顺序开始，在同一channel上排序。c := scases[pollorder[i]].cfor j > 0 && scases[lockorder[(j-1)/2]].c.sortkey() < c.sortkey() {k := (j - 1) / 2lockorder[j] = lockorder[k]j = k}lockorder[j] = pollorder[i]}for i := len(lockorder) - 1; i >= 0; i-- {o := lockorder[i]c := scases[o].clockorder[i] = lockorder[0]j := 0for {k := j*2 + 1if k >= i {break}if k+1 < i && scases[lockorder[k]].c.sortkey() < scases[lockorder[k+1]].c.sortkey() {k++}if c.sortkey() < scases[lockorder[k]].c.sortkey() {lockorder[j] = lockorder[k]j = kcontinue}break}lockorder[j] = o}if debugSelect {for i := 0; i+1 < len(lockorder); i++ {if scases[lockorder[i]].c.sortkey() > scases[lockorder[i+1]].c.sortkey() {print("i=", i, " x=", lockorder[i], " y=", lockorder[i+1], "\n")throw("select: broken sort")}}}// 锁定select中涉及的所有channelsellock(scases, lockorder)var (gp     *gsg     *sudogc      *hchank      *scasesglist *sudogsgnext *sudogqp     unsafe.Pointernextp  **sudog)// === pass 1 - 查找可以等待处理的channelvar casi intvar cas *scasevar caseSuccess boolvar caseReleaseTime int64 = -1var recvOK boolfor _, casei := range pollorder {casi = int(casei) // case的索引cas = &scases[casi]  c = cas.cif casi >= nsends { // 处理接收channel的casesg = c.sendq.dequeue()if sg != nil {// 如果当前channel的sendq上有等待的goroutine，// 跳到recv代码 并从缓冲区读取数据后将等待goroutine中的数据放入到缓冲区中相同的位置goto recv}if c.qcount > 0 {//如果当前channel的缓冲区不为空，就会跳到bufrecv标签处从缓冲区获取数据；goto bufrecv}if c.closed != 0 {//如果当前channel已经被关闭，就会跳到rclose读取末尾数据和收尾工作；goto rclose}} else { // 处理发送channel的caseif raceenabled {racereadpc(c.raceaddr(), casePC(casi), chansendpc)}if c.closed != 0 {// 如果当前channel已经被关闭就会直接跳到sclose标签（panic中止程序）goto sclose}sg = c.recvq.dequeue()if sg != nil {// 如果当前channel的recvq上有等待的goroutine，就会跳到 send标签向channel发送数据；goto send}if c.qcount < c.dataqsiz {// 如果当前channel的缓冲区存在空闲位置，就会将待发送的数据存入缓冲区；goto bufsend}}}if !block { // 如果是非阻塞，即包含default分支，解锁所有channel并返回selunlock(scases, lockorder)casi = -1goto retc}// === pass 2 - 将当前goroutine根据需要挂在chan的sendq或recvq上gp = getg() // 获取当前的groutineif gp.waiting != nil {throw("gp.waiting != nil")}nextp = &gp.waiting // 正在等待的sudog结构；按锁定顺序for _, casei := range lockorder {casi = int(casei)cas = &scases[casi]c = cas.csg := acquireSudog()// 获取sudog，将当前goroutine绑定到sudog上sg.g = gpsg.isSelect = true// 在分配elem和在gp.waiting上排队sg之间没有堆栈分割，copystack可以找到它。sg.elem = cas.elemsg.releasetime = 0if t0 != 0 {sg.releasetime = -1}sg.c = c// 按锁定顺序构建waiting list 。*nextp = sgnextp = &sg.waitlink// 加入相应等待队列if casi < nsends {c.sendq.enqueue(sg)} else {c.recvq.enqueue(sg)}}// 被唤醒后会根据 param 来判断是否是由 close 操作唤醒的，所以先置为 nilgp.param = nil// Signal to anyone trying to shrink our stack that we're about// to park on a channel. The window between when this G's status// changes and when we set gp.activeStackChans is not safe for// stack shrinking.atomic.Store8(&gp.parkingOnChan, 1)// 挂起当前goroutinegopark(selparkcommit, nil, waitReasonSelect, traceEvGoBlockSelect, 1)gp.activeStackChans = false// 加锁所有的channelsellock(scases, lockorder)gp.selectDone = 0sg = (*sudog)(gp.param)// param 存放唤醒 goroutine 的 sudog，如果是关闭操作唤醒的，那么就为 nilgp.param = nil// === pass 3 - 当前 Goroutine 被唤醒之后找到满足条件的 Channel 并进行处理//dequeue from unsuccessful chans// otherwise they stack up on quiet channels// record the successful case, if any.// We singly-linked up the SudoGs in lock order.// 从不成功的通道中退出队列，否则它们会堆积在安静的通道上，记录成功的案例（如果有的话）。我们单独将SudoG按锁定顺序连接起来。casi = -1cas = nilcaseSuccess = false// 当前goroutine 的 waiting 链表按照lockorder顺序存放着case的sudogsglist = gp.waiting// 在从 gp.waiting 取消case的sudog链接之前清除所有元素，便于GCfor sg1 := gp.waiting; sg1 != nil; sg1 = sg1.waitlink {sg1.isSelect = falsesg1.elem = nilsg1.c = nil}// 清楚当前goroutine的waiting链表，因为被sg代表的协程唤醒了gp.waiting = nilfor _, casei := range lockorder {k = &scases[casei]// 如果相等说明，goroutine是被当前case的channel收发操作唤醒的if sg == sglist {// sg唤醒了当前goroutine, 则当前G已经从sg的队列中出队，这里不需要再次出队casi = int(casei)cas = kcaseSuccess = sglist.successif sglist.releasetime > 0 {caseReleaseTime = sglist.releasetime}} else {// 不是此case唤醒当前goroutine, 将goroutine从case对应的队列(发送或接收)出队c = k.cif int(casei) < nsends {c.sendq.dequeueSudoG(sglist)} else {c.recvq.dequeueSudoG(sglist)}}// 释放当前case的sudog，然后处理下一个case的sudogsgnext = sglist.waitlinksglist.waitlink = nilreleaseSudog(sglist)sglist = sgnext}if cas == nil {throw("selectgo: bad wakeup")}c = cas.cif debugSelect {print("wait-return: cas0=", cas0, " c=", c, " cas=", cas, " send=", casi < nsends, "\n")}if casi < nsends {if !caseSuccess {goto sclose}} else {recvOK = caseSuccess}if raceenabled {if casi < nsends {raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)} else if cas.elem != nil {raceWriteObjectPC(c.elemtype, cas.elem, casePC(casi), chanrecvpc)}}if msanenabled {if casi < nsends {msanread(cas.elem, c.elemtype.size)} else if cas.elem != nil {msanwrite(cas.elem, c.elemtype.size)}}if asanenabled {if casi < nsends {asanread(cas.elem, c.elemtype.size)} else if cas.elem != nil {asanwrite(cas.elem, c.elemtype.size)}}selunlock(scases, lockorder)goto retcbufrecv:// 能从buffer获取数据if raceenabled {if cas.elem != nil {raceWriteObjectPC(c.elemtype, cas.elem, casePC(casi), chanrecvpc)}racenotify(c, c.recvx, nil)}if msanenabled && cas.elem != nil {msanwrite(cas.elem, c.elemtype.size)}if asanenabled && cas.elem != nil {asanwrite(cas.elem, c.elemtype.size)}recvOK = trueqp = chanbuf(c, c.recvx)if cas.elem != nil {typedmemmove(c.elemtype, cas.elem, qp)}typedmemclr(c.elemtype, qp)c.recvx++if c.recvx == c.dataqsiz {c.recvx = 0}c.qcount--selunlock(scases, lockorder)goto retcbufsend:// 发送数据到缓存if raceenabled {racenotify(c, c.sendx, nil)raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)}if msanenabled {msanread(cas.elem, c.elemtype.size)}if asanenabled {asanread(cas.elem, c.elemtype.size)}typedmemmove(c.elemtype, chanbuf(c, c.sendx), cas.elem)c.sendx++if c.sendx == c.dataqsiz {c.sendx = 0}c.qcount++selunlock(scases, lockorder)goto retcrecv:// 从休眠sender（sg）接收recv(c, sg, cas.elem, func() { selunlock(scases, lockorder) }, 2)if debugSelect {print("syncrecv: cas0=", cas0, " c=", c, "\n")}recvOK = truegoto retcrclose:// 读取结束的channelselunlock(scases, lockorder)recvOK = falseif cas.elem != nil {typedmemclr(c.elemtype, cas.elem)}if raceenabled {raceacquire(c.raceaddr())}goto retcsend:// 想休眠的接收房发送数据if raceenabled {raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)}if msanenabled {msanread(cas.elem, c.elemtype.size)}if asanenabled {asanread(cas.elem, c.elemtype.size)}send(c, sg, cas.elem, func() { selunlock(scases, lockorder) }, 2)if debugSelect {print("syncsend: cas0=", cas0, " c=", c, "\n")}goto retcretc:if caseReleaseTime > 0 {blockevent(caseReleaseTime-t0, 1)}return casi, recvOKsclose:// 向关闭的channel发送数据selunlock(scases, lockorder)panic(plainError("send on closed channel"))
}