...

Source file src/internal/trace/generation.go

Documentation: internal/trace

     1  // Copyright 2023 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 trace
     6  
     7  import (
     8  	"bufio"
     9  	"bytes"
    10  	"cmp"
    11  	"encoding/binary"
    12  	"fmt"
    13  	"io"
    14  	"slices"
    15  	"strings"
    16  
    17  	"internal/trace/event"
    18  	"internal/trace/event/go122"
    19  )
    20  
    21  // generation contains all the trace data for a single
    22  // trace generation. It is purely data: it does not
    23  // track any parse state nor does it contain a cursor
    24  // into the generation.
    25  type generation struct {
    26  	gen        uint64
    27  	batches    map[ThreadID][]batch
    28  	cpuSamples []cpuSample
    29  	*evTable
    30  }
    31  
    32  // spilledBatch represents a batch that was read out for the next generation,
    33  // while reading the previous one. It's passed on when parsing the next
    34  // generation.
    35  type spilledBatch struct {
    36  	gen uint64
    37  	*batch
    38  }
    39  
    40  // readGeneration buffers and decodes the structural elements of a trace generation
    41  // out of r. spill is the first batch of the new generation (already buffered and
    42  // parsed from reading the last generation). Returns the generation and the first
    43  // batch read of the next generation, if any.
    44  //
    45  // If gen is non-nil, it is valid and must be processed before handling the returned
    46  // error.
    47  func readGeneration(r *bufio.Reader, spill *spilledBatch) (*generation, *spilledBatch, error) {
    48  	g := &generation{
    49  		evTable: &evTable{
    50  			pcs: make(map[uint64]frame),
    51  		},
    52  		batches: make(map[ThreadID][]batch),
    53  	}
    54  	// Process the spilled batch.
    55  	if spill != nil {
    56  		g.gen = spill.gen
    57  		if err := processBatch(g, *spill.batch); err != nil {
    58  			return nil, nil, err
    59  		}
    60  		spill = nil
    61  	}
    62  	// Read batches one at a time until we either hit EOF or
    63  	// the next generation.
    64  	var spillErr error
    65  	for {
    66  		b, gen, err := readBatch(r)
    67  		if err == io.EOF {
    68  			break
    69  		}
    70  		if err != nil {
    71  			if g.gen != 0 {
    72  				// This is an error reading the first batch of the next generation.
    73  				// This is fine. Let's forge ahead assuming that what we've got so
    74  				// far is fine.
    75  				spillErr = err
    76  				break
    77  			}
    78  			return nil, nil, err
    79  		}
    80  		if gen == 0 {
    81  			// 0 is a sentinel used by the runtime, so we'll never see it.
    82  			return nil, nil, fmt.Errorf("invalid generation number %d", gen)
    83  		}
    84  		if g.gen == 0 {
    85  			// Initialize gen.
    86  			g.gen = gen
    87  		}
    88  		if gen == g.gen+1 { // TODO: advance this the same way the runtime does.
    89  			spill = &spilledBatch{gen: gen, batch: &b}
    90  			break
    91  		}
    92  		if gen != g.gen {
    93  			// N.B. Fail as fast as possible if we see this. At first it
    94  			// may seem prudent to be fault-tolerant and assume we have a
    95  			// complete generation, parsing and returning that first. However,
    96  			// if the batches are mixed across generations then it's likely
    97  			// we won't be able to parse this generation correctly at all.
    98  			// Rather than return a cryptic error in that case, indicate the
    99  			// problem as soon as we see it.
   100  			return nil, nil, fmt.Errorf("generations out of order")
   101  		}
   102  		if err := processBatch(g, b); err != nil {
   103  			return nil, nil, err
   104  		}
   105  	}
   106  
   107  	// Check some invariants.
   108  	if g.freq == 0 {
   109  		return nil, nil, fmt.Errorf("no frequency event found")
   110  	}
   111  	// N.B. Trust that the batch order is correct. We can't validate the batch order
   112  	// by timestamp because the timestamps could just be plain wrong. The source of
   113  	// truth is the order things appear in the trace and the partial order sequence
   114  	// numbers on certain events. If it turns out the batch order is actually incorrect
   115  	// we'll very likely fail to advance a partial order from the frontier.
   116  
   117  	// Compactify stacks and strings for better lookup performance later.
   118  	g.stacks.compactify()
   119  	g.strings.compactify()
   120  
   121  	// Validate stacks.
   122  	if err := validateStackStrings(&g.stacks, &g.strings, g.pcs); err != nil {
   123  		return nil, nil, err
   124  	}
   125  
   126  	// Fix up the CPU sample timestamps, now that we have freq.
   127  	for i := range g.cpuSamples {
   128  		s := &g.cpuSamples[i]
   129  		s.time = g.freq.mul(timestamp(s.time))
   130  	}
   131  	// Sort the CPU samples.
   132  	slices.SortFunc(g.cpuSamples, func(a, b cpuSample) int {
   133  		return cmp.Compare(a.time, b.time)
   134  	})
   135  	return g, spill, spillErr
   136  }
   137  
   138  // processBatch adds the batch to the generation.
   139  func processBatch(g *generation, b batch) error {
   140  	switch {
   141  	case b.isStringsBatch():
   142  		if err := addStrings(&g.strings, b); err != nil {
   143  			return err
   144  		}
   145  	case b.isStacksBatch():
   146  		if err := addStacks(&g.stacks, g.pcs, b); err != nil {
   147  			return err
   148  		}
   149  	case b.isCPUSamplesBatch():
   150  		samples, err := addCPUSamples(g.cpuSamples, b)
   151  		if err != nil {
   152  			return err
   153  		}
   154  		g.cpuSamples = samples
   155  	case b.isFreqBatch():
   156  		freq, err := parseFreq(b)
   157  		if err != nil {
   158  			return err
   159  		}
   160  		if g.freq != 0 {
   161  			return fmt.Errorf("found multiple frequency events")
   162  		}
   163  		g.freq = freq
   164  	case b.exp != event.NoExperiment:
   165  		if g.expData == nil {
   166  			g.expData = make(map[event.Experiment]*ExperimentalData)
   167  		}
   168  		if err := addExperimentalData(g.expData, b); err != nil {
   169  			return err
   170  		}
   171  	default:
   172  		g.batches[b.m] = append(g.batches[b.m], b)
   173  	}
   174  	return nil
   175  }
   176  
   177  // validateStackStrings makes sure all the string references in
   178  // the stack table are present in the string table.
   179  func validateStackStrings(
   180  	stacks *dataTable[stackID, stack],
   181  	strings *dataTable[stringID, string],
   182  	frames map[uint64]frame,
   183  ) error {
   184  	var err error
   185  	stacks.forEach(func(id stackID, stk stack) bool {
   186  		for _, pc := range stk.pcs {
   187  			frame, ok := frames[pc]
   188  			if !ok {
   189  				err = fmt.Errorf("found unknown pc %x for stack %d", pc, id)
   190  				return false
   191  			}
   192  			_, ok = strings.get(frame.funcID)
   193  			if !ok {
   194  				err = fmt.Errorf("found invalid func string ID %d for stack %d", frame.funcID, id)
   195  				return false
   196  			}
   197  			_, ok = strings.get(frame.fileID)
   198  			if !ok {
   199  				err = fmt.Errorf("found invalid file string ID %d for stack %d", frame.fileID, id)
   200  				return false
   201  			}
   202  		}
   203  		return true
   204  	})
   205  	return err
   206  }
   207  
   208  // addStrings takes a batch whose first byte is an EvStrings event
   209  // (indicating that the batch contains only strings) and adds each
   210  // string contained therein to the provided strings map.
   211  func addStrings(stringTable *dataTable[stringID, string], b batch) error {
   212  	if !b.isStringsBatch() {
   213  		return fmt.Errorf("internal error: addStrings called on non-string batch")
   214  	}
   215  	r := bytes.NewReader(b.data)
   216  	hdr, err := r.ReadByte() // Consume the EvStrings byte.
   217  	if err != nil || event.Type(hdr) != go122.EvStrings {
   218  		return fmt.Errorf("missing strings batch header")
   219  	}
   220  
   221  	var sb strings.Builder
   222  	for r.Len() != 0 {
   223  		// Read the header.
   224  		ev, err := r.ReadByte()
   225  		if err != nil {
   226  			return err
   227  		}
   228  		if event.Type(ev) != go122.EvString {
   229  			return fmt.Errorf("expected string event, got %d", ev)
   230  		}
   231  
   232  		// Read the string's ID.
   233  		id, err := binary.ReadUvarint(r)
   234  		if err != nil {
   235  			return err
   236  		}
   237  
   238  		// Read the string's length.
   239  		len, err := binary.ReadUvarint(r)
   240  		if err != nil {
   241  			return err
   242  		}
   243  		if len > go122.MaxStringSize {
   244  			return fmt.Errorf("invalid string size %d, maximum is %d", len, go122.MaxStringSize)
   245  		}
   246  
   247  		// Copy out the string.
   248  		n, err := io.CopyN(&sb, r, int64(len))
   249  		if n != int64(len) {
   250  			return fmt.Errorf("failed to read full string: read %d but wanted %d", n, len)
   251  		}
   252  		if err != nil {
   253  			return fmt.Errorf("copying string data: %w", err)
   254  		}
   255  
   256  		// Add the string to the map.
   257  		s := sb.String()
   258  		sb.Reset()
   259  		if err := stringTable.insert(stringID(id), s); err != nil {
   260  			return err
   261  		}
   262  	}
   263  	return nil
   264  }
   265  
   266  // addStacks takes a batch whose first byte is an EvStacks event
   267  // (indicating that the batch contains only stacks) and adds each
   268  // string contained therein to the provided stacks map.
   269  func addStacks(stackTable *dataTable[stackID, stack], pcs map[uint64]frame, b batch) error {
   270  	if !b.isStacksBatch() {
   271  		return fmt.Errorf("internal error: addStacks called on non-stacks batch")
   272  	}
   273  	r := bytes.NewReader(b.data)
   274  	hdr, err := r.ReadByte() // Consume the EvStacks byte.
   275  	if err != nil || event.Type(hdr) != go122.EvStacks {
   276  		return fmt.Errorf("missing stacks batch header")
   277  	}
   278  
   279  	for r.Len() != 0 {
   280  		// Read the header.
   281  		ev, err := r.ReadByte()
   282  		if err != nil {
   283  			return err
   284  		}
   285  		if event.Type(ev) != go122.EvStack {
   286  			return fmt.Errorf("expected stack event, got %d", ev)
   287  		}
   288  
   289  		// Read the stack's ID.
   290  		id, err := binary.ReadUvarint(r)
   291  		if err != nil {
   292  			return err
   293  		}
   294  
   295  		// Read how many frames are in each stack.
   296  		nFrames, err := binary.ReadUvarint(r)
   297  		if err != nil {
   298  			return err
   299  		}
   300  		if nFrames > go122.MaxFramesPerStack {
   301  			return fmt.Errorf("invalid stack size %d, maximum is %d", nFrames, go122.MaxFramesPerStack)
   302  		}
   303  
   304  		// Each frame consists of 4 fields: pc, funcID (string), fileID (string), line.
   305  		frames := make([]uint64, 0, nFrames)
   306  		for i := uint64(0); i < nFrames; i++ {
   307  			// Read the frame data.
   308  			pc, err := binary.ReadUvarint(r)
   309  			if err != nil {
   310  				return fmt.Errorf("reading frame %d's PC for stack %d: %w", i+1, id, err)
   311  			}
   312  			funcID, err := binary.ReadUvarint(r)
   313  			if err != nil {
   314  				return fmt.Errorf("reading frame %d's funcID for stack %d: %w", i+1, id, err)
   315  			}
   316  			fileID, err := binary.ReadUvarint(r)
   317  			if err != nil {
   318  				return fmt.Errorf("reading frame %d's fileID for stack %d: %w", i+1, id, err)
   319  			}
   320  			line, err := binary.ReadUvarint(r)
   321  			if err != nil {
   322  				return fmt.Errorf("reading frame %d's line for stack %d: %w", i+1, id, err)
   323  			}
   324  			frames = append(frames, pc)
   325  
   326  			if _, ok := pcs[pc]; !ok {
   327  				pcs[pc] = frame{
   328  					pc:     pc,
   329  					funcID: stringID(funcID),
   330  					fileID: stringID(fileID),
   331  					line:   line,
   332  				}
   333  			}
   334  		}
   335  
   336  		// Add the stack to the map.
   337  		if err := stackTable.insert(stackID(id), stack{pcs: frames}); err != nil {
   338  			return err
   339  		}
   340  	}
   341  	return nil
   342  }
   343  
   344  // addCPUSamples takes a batch whose first byte is an EvCPUSamples event
   345  // (indicating that the batch contains only CPU samples) and adds each
   346  // sample contained therein to the provided samples list.
   347  func addCPUSamples(samples []cpuSample, b batch) ([]cpuSample, error) {
   348  	if !b.isCPUSamplesBatch() {
   349  		return nil, fmt.Errorf("internal error: addCPUSamples called on non-CPU-sample batch")
   350  	}
   351  	r := bytes.NewReader(b.data)
   352  	hdr, err := r.ReadByte() // Consume the EvCPUSamples byte.
   353  	if err != nil || event.Type(hdr) != go122.EvCPUSamples {
   354  		return nil, fmt.Errorf("missing CPU samples batch header")
   355  	}
   356  
   357  	for r.Len() != 0 {
   358  		// Read the header.
   359  		ev, err := r.ReadByte()
   360  		if err != nil {
   361  			return nil, err
   362  		}
   363  		if event.Type(ev) != go122.EvCPUSample {
   364  			return nil, fmt.Errorf("expected CPU sample event, got %d", ev)
   365  		}
   366  
   367  		// Read the sample's timestamp.
   368  		ts, err := binary.ReadUvarint(r)
   369  		if err != nil {
   370  			return nil, err
   371  		}
   372  
   373  		// Read the sample's M.
   374  		m, err := binary.ReadUvarint(r)
   375  		if err != nil {
   376  			return nil, err
   377  		}
   378  		mid := ThreadID(m)
   379  
   380  		// Read the sample's P.
   381  		p, err := binary.ReadUvarint(r)
   382  		if err != nil {
   383  			return nil, err
   384  		}
   385  		pid := ProcID(p)
   386  
   387  		// Read the sample's G.
   388  		g, err := binary.ReadUvarint(r)
   389  		if err != nil {
   390  			return nil, err
   391  		}
   392  		goid := GoID(g)
   393  		if g == 0 {
   394  			goid = NoGoroutine
   395  		}
   396  
   397  		// Read the sample's stack.
   398  		s, err := binary.ReadUvarint(r)
   399  		if err != nil {
   400  			return nil, err
   401  		}
   402  
   403  		// Add the sample to the slice.
   404  		samples = append(samples, cpuSample{
   405  			schedCtx: schedCtx{
   406  				M: mid,
   407  				P: pid,
   408  				G: goid,
   409  			},
   410  			time:  Time(ts), // N.B. this is really a "timestamp," not a Time.
   411  			stack: stackID(s),
   412  		})
   413  	}
   414  	return samples, nil
   415  }
   416  
   417  // parseFreq parses out a lone EvFrequency from a batch.
   418  func parseFreq(b batch) (frequency, error) {
   419  	if !b.isFreqBatch() {
   420  		return 0, fmt.Errorf("internal error: parseFreq called on non-frequency batch")
   421  	}
   422  	r := bytes.NewReader(b.data)
   423  	r.ReadByte() // Consume the EvFrequency byte.
   424  
   425  	// Read the frequency. It'll come out as timestamp units per second.
   426  	f, err := binary.ReadUvarint(r)
   427  	if err != nil {
   428  		return 0, err
   429  	}
   430  	// Convert to nanoseconds per timestamp unit.
   431  	return frequency(1.0 / (float64(f) / 1e9)), nil
   432  }
   433  
   434  // addExperimentalData takes an experimental batch and adds it to the ExperimentalData
   435  // for the experiment its a part of.
   436  func addExperimentalData(expData map[event.Experiment]*ExperimentalData, b batch) error {
   437  	if b.exp == event.NoExperiment {
   438  		return fmt.Errorf("internal error: addExperimentalData called on non-experimental batch")
   439  	}
   440  	ed, ok := expData[b.exp]
   441  	if !ok {
   442  		ed = new(ExperimentalData)
   443  		expData[b.exp] = ed
   444  	}
   445  	ed.Batches = append(ed.Batches, ExperimentalBatch{
   446  		Thread: b.m,
   447  		Data:   b.data,
   448  	})
   449  	return nil
   450  }
   451  

View as plain text