| 1 | package main |
| 2 | |
| 3 | import ( |
| 4 | "io" |
| 5 | "os" |
| 6 | "path/filepath" |
| 7 | "runtime/debug" |
| 8 | "strconv" |
| 9 | "strings" |
| 10 | "time" |
| 11 | |
| 12 | "reasonix/internal/config" |
| 13 | ) |
| 14 | |
| 15 | const ( |
| 16 | fatalCrashDirName = "crash-fatal" |
| 17 | fatalCrashLogSuffix = ".log" |
| 18 | fatalCrashCoveredSuffix = ".covered" |
| 19 | legacyFatalCrashFile = "crash-fatal.log" |
| 20 | legacyFatalCrashCoveredFile = "crash-fatal-covered" |
| 21 | ) |
| 22 | |
| 23 | var fatalCrashProcessAlive = desktopProcessAlive |
| 24 | |
| 25 | func fatalCrashDir() string { |
| 26 | return filepath.Join(config.MemoryUserDir(), fatalCrashDirName) |
| 27 | } |
| 28 | |
| 29 | func fatalCrashPath() string { |
| 30 | return fatalCrashPathForPID(os.Getpid()) |
| 31 | } |
| 32 | |
| 33 | func fatalCrashCoveredPath() string { |
| 34 | return fatalCrashCoveredPathForPID(os.Getpid()) |
| 35 | } |
| 36 | |
| 37 | func fatalCrashPathForPID(pid int) string { |
| 38 | return filepath.Join(fatalCrashDir(), strconv.Itoa(pid)+fatalCrashLogSuffix) |
| 39 | } |
| 40 | |
| 41 | func fatalCrashCoveredPathForPID(pid int) string { |
| 42 | return filepath.Join(fatalCrashDir(), strconv.Itoa(pid)+fatalCrashCoveredSuffix) |
| 43 | } |
| 44 | |
| 45 | func legacyFatalCrashPath() string { |
| 46 | return filepath.Join(config.MemoryUserDir(), legacyFatalCrashFile) |
| 47 | } |
| 48 | |
| 49 | func legacyFatalCrashCoveredPath() string { |
| 50 | return filepath.Join(config.MemoryUserDir(), legacyFatalCrashCoveredFile) |
| 51 | } |
| 52 | |
| 53 | func markFatalCrashCovered() { |
| 54 | markFatalCrashCoveredForPID(os.Getpid()) |
| 55 | } |
| 56 | |
| 57 | func markFatalCrashCoveredForPID(pid int) { |
| 58 | path := fatalCrashCoveredPathForPID(pid) |
| 59 | if os.MkdirAll(filepath.Dir(path), 0o700) == nil { |
| 60 | _ = os.WriteFile(path, []byte("structured\n"), 0o600) |
| 61 | } |
| 62 | } |
| 63 | |
| 64 | // capturePreviousFatalCrash converts runtime.SetCrashOutput dumps from dead |
| 65 | // processes into the normal scrubbed queue. Per-PID files keep a routine second |
| 66 | // launch from truncating or unlinking the running primary process's dump. |
| 67 | func capturePreviousFatalCrash() { |
| 68 | // Preserve compatibility with the single-file format used by older builds. |
| 69 | // An empty legacy file may still be owned by a running older process, so it |
| 70 | // must be left untouched until it contains a completed crash dump. |
| 71 | captureFatalCrashFile(legacyFatalCrashPath(), legacyFatalCrashCoveredPath(), false) |
| 72 | |
| 73 | entries, err := os.ReadDir(fatalCrashDir()) |
| 74 | if err != nil { |
| 75 | return |
| 76 | } |
| 77 | for _, entry := range entries { |
| 78 | pid, ok := fatalCrashPID(entry.Name()) |
| 79 | if !ok || pid == os.Getpid() || fatalCrashProcessAlive(pid) { |
| 80 | continue |
| 81 | } |
| 82 | captureFatalCrashFile( |
| 83 | filepath.Join(fatalCrashDir(), entry.Name()), |
| 84 | fatalCrashCoveredPathForPID(pid), |
| 85 | true, |
| 86 | ) |
| 87 | } |
| 88 | for _, entry := range entries { |
| 89 | pid, ok := fatalCrashCoveredPID(entry.Name()) |
| 90 | if !ok || pid == os.Getpid() || fatalCrashProcessAlive(pid) { |
| 91 | continue |
| 92 | } |
| 93 | if _, err := os.Stat(fatalCrashPathForPID(pid)); os.IsNotExist(err) { |
| 94 | _ = os.Remove(filepath.Join(fatalCrashDir(), entry.Name())) |
| 95 | } |
| 96 | } |
| 97 | // Best effort: succeeds only when no live/current process artifacts remain. |
| 98 | _ = os.Remove(fatalCrashDir()) |
| 99 | } |
| 100 | |
| 101 | func fatalCrashPID(name string) (int, bool) { |
| 102 | return fatalCrashPIDWithSuffix(name, fatalCrashLogSuffix) |
| 103 | } |
| 104 | |
| 105 | func fatalCrashCoveredPID(name string) (int, bool) { |
| 106 | return fatalCrashPIDWithSuffix(name, fatalCrashCoveredSuffix) |
| 107 | } |
| 108 | |
| 109 | func fatalCrashPIDWithSuffix(name, suffix string) (int, bool) { |
| 110 | if !strings.HasSuffix(name, suffix) { |
| 111 | return 0, false |
| 112 | } |
| 113 | pid, err := strconv.Atoi(strings.TrimSuffix(name, suffix)) |
| 114 | return pid, err == nil && pid > 0 |
| 115 | } |
| 116 | |
| 117 | func captureFatalCrashFile(path, coveredPath string, removeEmpty bool) { |
| 118 | f, err := os.Open(path) |
| 119 | if err != nil { |
| 120 | return |
| 121 | } |
| 122 | occurredAt := time.Now().UTC() |
| 123 | if info, statErr := f.Stat(); statErr == nil { |
| 124 | occurredAt = info.ModTime().UTC() |
| 125 | } |
| 126 | raw, readErr := io.ReadAll(io.LimitReader(f, maxCrashStackBytes+1)) |
| 127 | _ = f.Close() |
| 128 | if readErr != nil || len(strings.TrimSpace(string(raw))) == 0 { |
| 129 | if removeEmpty { |
| 130 | _ = os.Remove(coveredPath) |
| 131 | _ = os.Remove(path) |
| 132 | } |
| 133 | return |
| 134 | } |
| 135 | if _, err := os.Stat(coveredPath); err == nil { |
| 136 | _ = os.Remove(coveredPath) |
| 137 | _ = os.Remove(path) |
| 138 | return |
| 139 | } |
| 140 | stack := sanitizeFatalRuntimeDump(string(raw)) |
| 141 | report := baseCrashReport("crash") |
| 142 | report.SchemaVersion = currentCrashSchema |
| 143 | report.Source = "go.runtime" |
| 144 | report.Label = "go.fatal" |
| 145 | report.ErrorType = "GoRuntimeFatal" |
| 146 | report.ErrorMessage = "Go runtime terminated the desktop process." |
| 147 | report.Stack = stack |
| 148 | report.TopFrame = topFrameFromStack(stack) |
| 149 | report.FingerprintHint = "go.runtime.fatal" |
| 150 | report.OccurredAt = occurredAt.Format(time.RFC3339) |
| 151 | report.Diagnostics = &crashDiagnostics{ |
| 152 | ObserverVersion: version, ObserverBuildCommit: buildCommit(), ProcessRole: "service", |
| 153 | ObservedAt: time.Now().UTC().Format(time.RFC3339Nano), TerminationReason: "unknown", |
| 154 | CleanupOutcome: "interrupted", Evidence: "confirmed", Category: "crash", |
| 155 | } |
| 156 | report.Message = sanitizeCrashText("[go.runtime.fatal]\n\n"+stack, maxCrashDetailBytes) |
| 157 | if writePendingReport(report, true) { |
| 158 | _ = os.Remove(coveredPath) |
| 159 | _ = os.Remove(path) |
| 160 | } |
| 161 | } |
| 162 | |
| 163 | // sanitizeFatalRuntimeDump removes panic values and preamble text that could |
| 164 | // originate in user-controlled errors, while retaining runtime classification |
| 165 | // and symbolized goroutine stacks for diagnosis. |
| 166 | func sanitizeFatalRuntimeDump(raw string) string { |
| 167 | lines := strings.Split(raw, "\n") |
| 168 | classification := "runtime crash output" |
| 169 | stackStart := -1 |
| 170 | for i, line := range lines { |
| 171 | trimmed := strings.TrimSpace(line) |
| 172 | switch { |
| 173 | case strings.HasPrefix(trimmed, "fatal error:"): |
| 174 | classification = sanitizeCrashText(trimmed, 256) |
| 175 | case strings.HasPrefix(trimmed, "panic:"): |
| 176 | classification = "panic: [redacted panic value]" |
| 177 | } |
| 178 | if strings.HasPrefix(trimmed, "goroutine ") { |
| 179 | stackStart = i |
| 180 | break |
| 181 | } |
| 182 | } |
| 183 | stack := "" |
| 184 | if stackStart >= 0 { |
| 185 | stack = strings.Join(lines[stackStart:], "\n") |
| 186 | } |
| 187 | return sanitizeCrashText(classification+"\n\n"+stack, maxCrashStackBytes) |
| 188 | } |
| 189 | |
| 190 | // installFatalCrashOutput asks the Go runtime to mirror unrecovered panics and |
| 191 | // fatal runtime errors to a durable file. The runtime duplicates the descriptor, |
| 192 | // so the file may be closed after SetCrashOutput returns. |
| 193 | func installFatalCrashOutput() { |
| 194 | path := fatalCrashPath() |
| 195 | if err := os.MkdirAll(filepath.Dir(path), 0o700); err != nil { |
| 196 | return |
| 197 | } |
| 198 | f, err := os.OpenFile(path, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0o600) |
| 199 | if err != nil { |
| 200 | return |
| 201 | } |
| 202 | if err := debug.SetCrashOutput(f, debug.CrashOptions{}); err != nil { |
| 203 | _ = f.Close() |
| 204 | _ = os.Remove(path) |
| 205 | return |
| 206 | } |
| 207 | _ = f.Close() |
| 208 | } |
| 209 |