mklockrank.go

     1  // Copyright 2022 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  //go:build ignore
     6  
     7  // mklockrank records the static rank graph of the locks in the
     8  // runtime and generates the rank checking structures in lockrank.go.
     9  package main
    10  
    11  import (
    12  	"bytes"
    13  	"flag"
    14  	"fmt"
    15  	"go/format"
    16  	"internal/dag"
    17  	"io"
    18  	"log"
    19  	"os"
    20  	"strings"
    21  )
    22  
    23  // ranks describes the lock rank graph. See "go doc internal/dag" for
    24  // the syntax.
    25  //
    26  // "a < b" means a must be acquired before b if both are held
    27  // (or, if b is held, a cannot be acquired).
    28  //
    29  // "NONE < a" means no locks may be held when a is acquired.
    30  //
    31  // If a lock is not given a rank, then it is assumed to be a leaf
    32  // lock, which means no other lock can be acquired while it is held.
    33  // Therefore, leaf locks do not need to be given an explicit rank.
    34  //
    35  // Ranks in all caps are pseudo-nodes that help define order, but do
    36  // not actually define a rank.
    37  //
    38  // TODO: It's often hard to correlate rank names to locks. Change
    39  // these to be more consistent with the locks they label.
    40  const ranks = `
    41  # Sysmon
    42  NONE
    43  < sysmon
    44  < scavenge, forcegc, computeMaxProcs, updateMaxProcsG;
    45  
    46  # Defer
    47  NONE < defer;
    48  
    49  # GC
    50  NONE <
    51    sweepWaiters,
    52    assistQueue,
    53    strongFromWeakQueue,
    54    cleanupQueue,
    55    sweep;
    56  
    57  # Test only
    58  NONE < testR, testW;
    59  
    60  # vgetrandom
    61  NONE < vgetrandom;
    62  
    63  NONE < timerSend;
    64  
    65  # Scheduler, timers, netpoll
    66  NONE < allocmW, execW, cpuprof, pollCache, pollDesc, wakeableSleep;
    67  scavenge, sweep, testR, wakeableSleep, timerSend < hchan;
    68  assistQueue,
    69    cleanupQueue,
    70    computeMaxProcs,
    71    cpuprof,
    72    forcegc,
    73    updateMaxProcsG,
    74    hchan,
    75    pollDesc, # pollDesc can interact with timers, which can lock sched.
    76    scavenge,
    77    strongFromWeakQueue,
    78    sweep,
    79    sweepWaiters,
    80    testR,
    81    wakeableSleep
    82  # Above SCHED are things that can call into the scheduler.
    83  < SCHED
    84  # Below SCHED is the scheduler implementation.
    85  < allocmR,
    86    execR;
    87  allocmR, execR, hchan < sched;
    88  sched < allg, allp;
    89  
    90  # Channels
    91  NONE < notifyList;
    92  hchan, notifyList < sudog;
    93  
    94  hchan, pollDesc, wakeableSleep < timers;
    95  timers, timerSend < timer < netpollInit;
    96  
    97  # Semaphores
    98  NONE < root;
    99  
   100  # Itabs
   101  NONE
   102  < itab
   103  < reflectOffs;
   104  
   105  # Synctest
   106  hchan,
   107    notifyList,
   108    reflectOffs,
   109    root,
   110    strongFromWeakQueue,
   111    sweepWaiters,
   112    timer,
   113    timers
   114  < synctest;
   115  
   116  # User arena state
   117  NONE < userArenaState;
   118  
   119  # Tracing without a P uses a global trace buffer.
   120  scavenge
   121  # Above TRACEGLOBAL can emit a trace event without a P.
   122  < TRACEGLOBAL
   123  # Below TRACEGLOBAL manages the global tracing buffer.
   124  # Note that traceBuf eventually chains to MALLOC, but we never get that far
   125  # in the situation where there's no P.
   126  < traceBuf;
   127  # Starting/stopping tracing traces strings.
   128  traceBuf < traceStrings;
   129  
   130  # Malloc
   131  allg,
   132    allocmR,
   133    allp, # procresize
   134    execR, # May grow stack
   135    execW, # May allocate after BeforeFork
   136    hchan,
   137    notifyList,
   138    reflectOffs,
   139    timer,
   140    traceStrings,
   141    userArenaState,
   142    vgetrandom
   143  # Above MALLOC are things that can allocate memory.
   144  < MALLOC
   145  # Below MALLOC is the malloc implementation.
   146  < fin,
   147    spanSetSpine,
   148    mspanSpecial,
   149    traceTypeTab,
   150    MPROF;
   151  
   152  # We can acquire gcBitsArenas for pinner bits, and
   153  # it's guarded by mspanSpecial.
   154  MALLOC, mspanSpecial < gcBitsArenas;
   155  
   156  # Memory profiling
   157  MPROF < profInsert, profBlock, profMemActive;
   158  profMemActive < profMemFuture;
   159  
   160  # Stack allocation and copying
   161  gcBitsArenas,
   162    netpollInit,
   163    profBlock,
   164    profInsert,
   165    profMemFuture,
   166    spanSetSpine,
   167    synctest,
   168    fin,
   169    root
   170  # Anything that can grow the stack can acquire STACKGROW.
   171  # (Most higher layers imply STACKGROW, like MALLOC.)
   172  < STACKGROW
   173  # Below STACKGROW is the stack allocator/copying implementation.
   174  < gscan;
   175  gscan < stackpool;
   176  gscan < stackLarge;
   177  # Generally, hchan must be acquired before gscan. But in one case,
   178  # where we suspend a G and then shrink its stack, syncadjustsudogs
   179  # can acquire hchan locks while holding gscan. To allow this case,
   180  # we use hchanLeaf instead of hchan.
   181  gscan < hchanLeaf;
   182  
   183  # Write barrier
   184  defer,
   185    gscan,
   186    mspanSpecial,
   187    pollCache,
   188    sudog,
   189    timer
   190  # Anything that can have write barriers can acquire WB.
   191  # Above WB, we can have write barriers.
   192  < WB
   193  # Below WB is the write barrier implementation.
   194  < wbufSpans;
   195  
   196  # xRegState allocator
   197  sched < xRegAlloc;
   198  
   199  # Span allocator
   200  stackLarge,
   201    stackpool,
   202    wbufSpans
   203  # Above mheap is anything that can call the span allocator.
   204  < mheap;
   205  # Below mheap is the span allocator implementation.
   206  #
   207  # Specials: we're allowed to allocate a special while holding
   208  # an mspanSpecial lock, and they're part of the malloc implementation.
   209  # Pinner bits might be freed by the span allocator.
   210  mheap, mspanSpecial < mheapSpecial;
   211  # Fixallocs
   212  mheap, mheapSpecial, xRegAlloc < globalAlloc;
   213  
   214  # Execution tracer events (with a P)
   215  hchan,
   216    mheap,
   217    root,
   218    sched,
   219    traceStrings,
   220    notifyList,
   221    fin
   222  # Above TRACE is anything that can create a trace event
   223  < TRACE
   224  < trace
   225  < traceStackTab;
   226  
   227  # panic is handled specially. It is implicitly below all other locks.
   228  NONE < panic;
   229  # deadlock is not acquired while holding panic, but it also needs to be
   230  # below all other locks.
   231  panic < deadlock;
   232  # raceFini is only held while exiting.
   233  panic < raceFini;
   234  
   235  # RWMutex internal read lock
   236  
   237  allocmR,
   238    allocmW
   239  < allocmRInternal;
   240  
   241  execR,
   242    execW
   243  < execRInternal;
   244  
   245  testR,
   246    testW
   247  < testRInternal;
   248  `
   249  
   250  // cyclicRanks lists lock ranks that allow multiple locks of the same
   251  // rank to be acquired simultaneously. The runtime enforces ordering
   252  // within these ranks using a separate mechanism.
   253  var cyclicRanks = map[string]bool{
   254  	// Multiple timers are locked simultaneously in destroy().
   255  	"timers": true,
   256  	// Multiple hchans are acquired in hchan.sortkey() order in
   257  	// select.
   258  	"hchan": true,
   259  	// Multiple hchanLeafs are acquired in hchan.sortkey() order in
   260  	// syncadjustsudogs().
   261  	"hchanLeaf": true,
   262  	// The point of the deadlock lock is to deadlock.
   263  	"deadlock": true,
   264  }
   265  
   266  func main() {
   267  	flagO := flag.String("o", "", "write to `file` instead of stdout")
   268  	flagDot := flag.Bool("dot", false, "emit graphviz output instead of Go")
   269  	flag.Parse()
   270  	if flag.NArg() != 0 {
   271  		fmt.Fprintf(os.Stderr, "too many arguments")
   272  		os.Exit(2)
   273  	}
   274  
   275  	g, err := dag.Parse(ranks)
   276  	if err != nil {
   277  		log.Fatal(err)
   278  	}
   279  
   280  	var out []byte
   281  	if *flagDot {
   282  		var b bytes.Buffer
   283  		g.TransitiveReduction()
   284  		// Add cyclic edges for visualization.
   285  		for k := range cyclicRanks {
   286  			g.AddEdge(k, k)
   287  		}
   288  		// Reverse the graph. It's much easier to read this as
   289  		// a "<" partial order than a ">" partial order. This
   290  		// ways, locks are acquired from the top going down
   291  		// and time moves forward over the edges instead of
   292  		// backward.
   293  		g.Transpose()
   294  		generateDot(&b, g)
   295  		out = b.Bytes()
   296  	} else {
   297  		var b bytes.Buffer
   298  		generateGo(&b, g)
   299  		out, err = format.Source(b.Bytes())
   300  		if err != nil {
   301  			log.Fatal(err)
   302  		}
   303  	}
   304  
   305  	if *flagO != "" {
   306  		err = os.WriteFile(*flagO, out, 0666)
   307  	} else {
   308  		_, err = os.Stdout.Write(out)
   309  	}
   310  	if err != nil {
   311  		log.Fatal(err)
   312  	}
   313  }
   314  
   315  func generateGo(w io.Writer, g *dag.Graph) {
   316  	fmt.Fprintf(w, `// Code generated by mklockrank.go; DO NOT EDIT.
   317  
   318  package runtime
   319  
   320  type lockRank int
   321  
   322  `)
   323  
   324  	// Create numeric ranks.
   325  	topo := g.Topo()
   326  	for i, j := 0, len(topo)-1; i < j; i, j = i+1, j-1 {
   327  		topo[i], topo[j] = topo[j], topo[i]
   328  	}
   329  	fmt.Fprintf(w, `
   330  // Constants representing the ranks of all non-leaf runtime locks, in rank order.
   331  // Locks with lower rank must be taken before locks with higher rank,
   332  // in addition to satisfying the partial order in lockPartialOrder.
   333  // A few ranks allow self-cycles, which are specified in lockPartialOrder.
   334  const (
   335  	lockRankUnknown lockRank = iota
   336  
   337  `)
   338  	for _, rank := range topo {
   339  		if isPseudo(rank) {
   340  			fmt.Fprintf(w, "\t// %s\n", rank)
   341  		} else {
   342  			fmt.Fprintf(w, "\t%s\n", cname(rank))
   343  		}
   344  	}
   345  	fmt.Fprintf(w, `)
   346  
   347  // lockRankLeafRank is the rank of lock that does not have a declared rank,
   348  // and hence is a leaf lock.
   349  const lockRankLeafRank lockRank = 1000
   350  `)
   351  
   352  	// Create string table.
   353  	fmt.Fprintf(w, `
   354  // lockNames gives the names associated with each of the above ranks.
   355  var lockNames = []string{
   356  `)
   357  	for _, rank := range topo {
   358  		if !isPseudo(rank) {
   359  			fmt.Fprintf(w, "\t%s: %q,\n", cname(rank), rank)
   360  		}
   361  	}
   362  	fmt.Fprintf(w, `}
   363  
   364  func (rank lockRank) String() string {
   365  	if rank == 0 {
   366  		return "UNKNOWN"
   367  	}
   368  	if rank == lockRankLeafRank {
   369  		return "LEAF"
   370  	}
   371  	if rank < 0 || int(rank) >= len(lockNames) {
   372  		return "BAD RANK"
   373  	}
   374  	return lockNames[rank]
   375  }
   376  `)
   377  
   378  	// Create partial order structure.
   379  	fmt.Fprintf(w, `
   380  // lockPartialOrder is the transitive closure of the lock rank graph.
   381  // An entry for rank X lists all of the ranks that can already be held
   382  // when rank X is acquired.
   383  //
   384  // Lock ranks that allow self-cycles list themselves.
   385  var lockPartialOrder [][]lockRank = [][]lockRank{
   386  `)
   387  	for _, rank := range topo {
   388  		if isPseudo(rank) {
   389  			continue
   390  		}
   391  		list := []string{}
   392  		for _, before := range g.Edges(rank) {
   393  			if !isPseudo(before) {
   394  				list = append(list, cname(before))
   395  			}
   396  		}
   397  		if cyclicRanks[rank] {
   398  			list = append(list, cname(rank))
   399  		}
   400  
   401  		fmt.Fprintf(w, "\t%s: {%s},\n", cname(rank), strings.Join(list, ", "))
   402  	}
   403  	fmt.Fprintf(w, "}\n")
   404  }
   405  
   406  // cname returns the Go const name for the given lock rank label.
   407  func cname(label string) string {
   408  	return "lockRank" + strings.ToUpper(label[:1]) + label[1:]
   409  }
   410  
   411  func isPseudo(label string) bool {
   412  	return strings.ToUpper(label) == label
   413  }
   414  
   415  // generateDot emits a Graphviz dot representation of g to w.
   416  func generateDot(w io.Writer, g *dag.Graph) {
   417  	fmt.Fprintf(w, "digraph g {\n")
   418  
   419  	// Define all nodes.
   420  	for _, node := range g.Nodes {
   421  		fmt.Fprintf(w, "%q;\n", node)
   422  	}
   423  
   424  	// Create edges.
   425  	for _, node := range g.Nodes {
   426  		for _, to := range g.Edges(node) {
   427  			fmt.Fprintf(w, "%q -> %q;\n", node, to)
   428  		}
   429  	}
   430  
   431  	fmt.Fprintf(w, "}\n")
   432  }
   433
View as plain text