1 // Copyright 2009 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 /* 6 Package runtime contains operations that interact with Go's runtime system, 7 such as functions to control goroutines. It also includes the low-level type information 8 used by the reflect package; see [reflect]'s documentation for the programmable 9 interface to the run-time type system. 10 11 # Environment Variables 12 13 The following environment variables ($name or %name%, depending on the host 14 operating system) control the run-time behavior of Go programs. The meanings 15 and use may change from release to release. 16 17 The GOGC variable sets the initial garbage collection target percentage. 18 A collection is triggered when the ratio of freshly allocated data to live data 19 remaining after the previous collection reaches this percentage. The default 20 is GOGC=100. Setting GOGC=off disables the garbage collector entirely. 21 [runtime/debug.SetGCPercent] allows changing this percentage at run time. 22 23 The GOMEMLIMIT variable sets a soft memory limit for the runtime. This memory limit 24 includes the Go heap and all other memory managed by the runtime, and excludes 25 external memory sources such as mappings of the binary itself, memory managed in 26 other languages, and memory held by the operating system on behalf of the Go 27 program. GOMEMLIMIT is a numeric value in bytes with an optional unit suffix. 28 The supported suffixes include B, KiB, MiB, GiB, and TiB. These suffixes 29 represent quantities of bytes as defined by the IEC 80000-13 standard. That is, 30 they are based on powers of two: KiB means 2^10 bytes, MiB means 2^20 bytes, 31 and so on. The default setting is [math.MaxInt64], which effectively disables the 32 memory limit. [runtime/debug.SetMemoryLimit] allows changing this limit at run 33 time. 34 35 The GODEBUG variable controls debugging variables within the runtime. 36 It is a comma-separated list of name=val pairs setting these named variables: 37 38 allocfreetrace: setting allocfreetrace=1 causes every allocation to be 39 profiled and a stack trace printed on each object's allocation and free. 40 41 clobberfree: setting clobberfree=1 causes the garbage collector to 42 clobber the memory content of an object with bad content when it frees 43 the object. 44 45 cpu.*: cpu.all=off disables the use of all optional instruction set extensions. 46 cpu.extension=off disables use of instructions from the specified instruction set extension. 47 extension is the lower case name for the instruction set extension such as sse41 or avx 48 as listed in internal/cpu package. As an example cpu.avx=off disables runtime detection 49 and thereby use of AVX instructions. 50 51 cgocheck: setting cgocheck=0 disables all checks for packages 52 using cgo to incorrectly pass Go pointers to non-Go code. 53 Setting cgocheck=1 (the default) enables relatively cheap 54 checks that may miss some errors. A more complete, but slow, 55 cgocheck mode can be enabled using GOEXPERIMENT (which 56 requires a rebuild), see https://pkg.go.dev/internal/goexperiment for details. 57 58 disablethp: setting disablethp=1 on Linux disables transparent huge pages for the heap. 59 It has no effect on other platforms. disablethp is meant for compatibility with versions 60 of Go before 1.21, which stopped working around a Linux kernel default that can result 61 in significant memory overuse. See https://go.dev/issue/64332. This setting will be 62 removed in a future release, so operators should tweak their Linux configuration to suit 63 their needs before then. See https://go.dev/doc/gc-guide#Linux_transparent_huge_pages. 64 65 dontfreezetheworld: by default, the start of a fatal panic or throw 66 "freezes the world", preempting all threads to stop all running 67 goroutines, which makes it possible to traceback all goroutines, and 68 keeps their state close to the point of panic. Setting 69 dontfreezetheworld=1 disables this preemption, allowing goroutines to 70 continue executing during panic processing. Note that goroutines that 71 naturally enter the scheduler will still stop. This can be useful when 72 debugging the runtime scheduler, as freezetheworld perturbs scheduler 73 state and thus may hide problems. 74 75 efence: setting efence=1 causes the allocator to run in a mode 76 where each object is allocated on a unique page and addresses are 77 never recycled. 78 79 gccheckmark: setting gccheckmark=1 enables verification of the 80 garbage collector's concurrent mark phase by performing a 81 second mark pass while the world is stopped. If the second 82 pass finds a reachable object that was not found by concurrent 83 mark, the garbage collector will panic. 84 85 gcpacertrace: setting gcpacertrace=1 causes the garbage collector to 86 print information about the internal state of the concurrent pacer. 87 88 gcshrinkstackoff: setting gcshrinkstackoff=1 disables moving goroutines 89 onto smaller stacks. In this mode, a goroutine's stack can only grow. 90 91 gcstoptheworld: setting gcstoptheworld=1 disables concurrent garbage collection, 92 making every garbage collection a stop-the-world event. Setting gcstoptheworld=2 93 also disables concurrent sweeping after the garbage collection finishes. 94 95 gctrace: setting gctrace=1 causes the garbage collector to emit a single line to standard 96 error at each collection, summarizing the amount of memory collected and the 97 length of the pause. The format of this line is subject to change. Included in 98 the explanation below is also the relevant runtime/metrics metric for each field. 99 Currently, it is: 100 gc # @#s #%: #+#+# ms clock, #+#/#/#+# ms cpu, #->#-># MB, # MB goal, # MB stacks, #MB globals, # P 101 where the fields are as follows: 102 gc # the GC number, incremented at each GC 103 @#s time in seconds since program start 104 #% percentage of time spent in GC since program start 105 #+...+# wall-clock/CPU times for the phases of the GC 106 #->#-># MB heap size at GC start, at GC end, and live heap, or /gc/scan/heap:bytes 107 # MB goal goal heap size, or /gc/heap/goal:bytes 108 # MB stacks estimated scannable stack size, or /gc/scan/stack:bytes 109 # MB globals scannable global size, or /gc/scan/globals:bytes 110 # P number of processors used, or /sched/gomaxprocs:threads 111 The phases are stop-the-world (STW) sweep termination, concurrent 112 mark and scan, and STW mark termination. The CPU times 113 for mark/scan are broken down in to assist time (GC performed in 114 line with allocation), background GC time, and idle GC time. 115 If the line ends with "(forced)", this GC was forced by a 116 runtime.GC() call. 117 118 harddecommit: setting harddecommit=1 causes memory that is returned to the OS to 119 also have protections removed on it. This is the only mode of operation on Windows, 120 but is helpful in debugging scavenger-related issues on other platforms. Currently, 121 only supported on Linux. 122 123 inittrace: setting inittrace=1 causes the runtime to emit a single line to standard 124 error for each package with init work, summarizing the execution time and memory 125 allocation. No information is printed for inits executed as part of plugin loading 126 and for packages without both user defined and compiler generated init work. 127 The format of this line is subject to change. Currently, it is: 128 init # @#ms, # ms clock, # bytes, # allocs 129 where the fields are as follows: 130 init # the package name 131 @# ms time in milliseconds when the init started since program start 132 # clock wall-clock time for package initialization work 133 # bytes memory allocated on the heap 134 # allocs number of heap allocations 135 136 madvdontneed: setting madvdontneed=0 will use MADV_FREE 137 instead of MADV_DONTNEED on Linux when returning memory to the 138 kernel. This is more efficient, but means RSS numbers will 139 drop only when the OS is under memory pressure. On the BSDs and 140 Illumos/Solaris, setting madvdontneed=1 will use MADV_DONTNEED instead 141 of MADV_FREE. This is less efficient, but causes RSS numbers to drop 142 more quickly. 143 144 memprofilerate: setting memprofilerate=X will update the value of runtime.MemProfileRate. 145 When set to 0 memory profiling is disabled. Refer to the description of 146 MemProfileRate for the default value. 147 148 pagetrace: setting pagetrace=/path/to/file will write out a trace of page events 149 that can be viewed, analyzed, and visualized using the x/debug/cmd/pagetrace tool. 150 Build your program with GOEXPERIMENT=pagetrace to enable this functionality. Do not 151 enable this functionality if your program is a setuid binary as it introduces a security 152 risk in that scenario. Currently not supported on Windows, plan9 or js/wasm. Setting this 153 option for some applications can produce large traces, so use with care. 154 155 panicnil: setting panicnil=1 disables the runtime error when calling panic with nil 156 interface value or an untyped nil. 157 158 runtimecontentionstacks: setting runtimecontentionstacks=1 enables inclusion of call stacks 159 related to contention on runtime-internal locks in the "mutex" profile, subject to the 160 MutexProfileFraction setting. When runtimecontentionstacks=0, contention on 161 runtime-internal locks will report as "runtime._LostContendedRuntimeLock". When 162 runtimecontentionstacks=1, the call stacks will correspond to the unlock call that released 163 the lock. But instead of the value corresponding to the amount of contention that call 164 stack caused, it corresponds to the amount of time the caller of unlock had to wait in its 165 original call to lock. A future release is expected to align those and remove this setting. 166 167 invalidptr: invalidptr=1 (the default) causes the garbage collector and stack 168 copier to crash the program if an invalid pointer value (for example, 1) 169 is found in a pointer-typed location. Setting invalidptr=0 disables this check. 170 This should only be used as a temporary workaround to diagnose buggy code. 171 The real fix is to not store integers in pointer-typed locations. 172 173 sbrk: setting sbrk=1 replaces the memory allocator and garbage collector 174 with a trivial allocator that obtains memory from the operating system and 175 never reclaims any memory. 176 177 scavtrace: setting scavtrace=1 causes the runtime to emit a single line to standard 178 error, roughly once per GC cycle, summarizing the amount of work done by the 179 scavenger as well as the total amount of memory returned to the operating system 180 and an estimate of physical memory utilization. The format of this line is subject 181 to change, but currently it is: 182 scav # KiB work (bg), # KiB work (eager), # KiB total, #% util 183 where the fields are as follows: 184 # KiB work (bg) the amount of memory returned to the OS in the background since 185 the last line 186 # KiB work (eager) the amount of memory returned to the OS eagerly since the last line 187 # KiB now the amount of address space currently returned to the OS 188 #% util the fraction of all unscavenged heap memory which is in-use 189 If the line ends with "(forced)", then scavenging was forced by a 190 debug.FreeOSMemory() call. 191 192 scheddetail: setting schedtrace=X and scheddetail=1 causes the scheduler to emit 193 detailed multiline info every X milliseconds, describing state of the scheduler, 194 processors, threads and goroutines. 195 196 schedtrace: setting schedtrace=X causes the scheduler to emit a single line to standard 197 error every X milliseconds, summarizing the scheduler state. 198 199 tracebackancestors: setting tracebackancestors=N extends tracebacks with the stacks at 200 which goroutines were created, where N limits the number of ancestor goroutines to 201 report. This also extends the information returned by runtime.Stack. Ancestor's goroutine 202 IDs will refer to the ID of the goroutine at the time of creation; it's possible for this 203 ID to be reused for another goroutine. Setting N to 0 will report no ancestry information. 204 205 tracefpunwindoff: setting tracefpunwindoff=1 forces the execution tracer to 206 use the runtime's default stack unwinder instead of frame pointer unwinding. 207 This increases tracer overhead, but could be helpful as a workaround or for 208 debugging unexpected regressions caused by frame pointer unwinding. 209 210 traceadvanceperiod: the approximate period in nanoseconds between trace generations. Only 211 applies if a program is built with GOEXPERIMENT=exectracer2. Used primarily for testing 212 and debugging the execution tracer. 213 214 asyncpreemptoff: asyncpreemptoff=1 disables signal-based 215 asynchronous goroutine preemption. This makes some loops 216 non-preemptible for long periods, which may delay GC and 217 goroutine scheduling. This is useful for debugging GC issues 218 because it also disables the conservative stack scanning used 219 for asynchronously preempted goroutines. 220 221 The [net] and [net/http] packages also refer to debugging variables in GODEBUG. 222 See the documentation for those packages for details. 223 224 The GOMAXPROCS variable limits the number of operating system threads that 225 can execute user-level Go code simultaneously. There is no limit to the number of threads 226 that can be blocked in system calls on behalf of Go code; those do not count against 227 the GOMAXPROCS limit. This package's [GOMAXPROCS] function queries and changes 228 the limit. 229 230 The GORACE variable configures the race detector, for programs built using -race. 231 See the [Race Detector article] for details. 232 233 The GOTRACEBACK variable controls the amount of output generated when a Go 234 program fails due to an unrecovered panic or an unexpected runtime condition. 235 By default, a failure prints a stack trace for the current goroutine, 236 eliding functions internal to the run-time system, and then exits with exit code 2. 237 The failure prints stack traces for all goroutines if there is no current goroutine 238 or the failure is internal to the run-time. 239 GOTRACEBACK=none omits the goroutine stack traces entirely. 240 GOTRACEBACK=single (the default) behaves as described above. 241 GOTRACEBACK=all adds stack traces for all user-created goroutines. 242 GOTRACEBACK=system is like “all” but adds stack frames for run-time functions 243 and shows goroutines created internally by the run-time. 244 GOTRACEBACK=crash is like “system” but crashes in an operating system-specific 245 manner instead of exiting. For example, on Unix systems, the crash raises 246 SIGABRT to trigger a core dump. 247 GOTRACEBACK=wer is like “crash” but doesn't disable Windows Error Reporting (WER). 248 For historical reasons, the GOTRACEBACK settings 0, 1, and 2 are synonyms for 249 none, all, and system, respectively. 250 The [runtime/debug.SetTraceback] function allows increasing the 251 amount of output at run time, but it cannot reduce the amount below that 252 specified by the environment variable. 253 254 The GOARCH, GOOS, GOPATH, and GOROOT environment variables complete 255 the set of Go environment variables. They influence the building of Go programs 256 (see [cmd/go] and [go/build]). 257 GOARCH, GOOS, and GOROOT are recorded at compile time and made available by 258 constants or functions in this package, but they do not influence the execution 259 of the run-time system. 260 261 # Security 262 263 On Unix platforms, Go's runtime system behaves slightly differently when a 264 binary is setuid/setgid or executed with setuid/setgid-like properties, in order 265 to prevent dangerous behaviors. On Linux this is determined by checking for the 266 AT_SECURE flag in the auxiliary vector, on the BSDs and Solaris/Illumos it is 267 determined by checking the issetugid syscall, and on AIX it is determined by 268 checking if the uid/gid match the effective uid/gid. 269 270 When the runtime determines the binary is setuid/setgid-like, it does three main 271 things: 272 - The standard input/output file descriptors (0, 1, 2) are checked to be open. 273 If any of them are closed, they are opened pointing at /dev/null. 274 - The value of the GOTRACEBACK environment variable is set to 'none'. 275 - When a signal is received that terminates the program, or the program 276 encounters an unrecoverable panic that would otherwise override the value 277 of GOTRACEBACK, the goroutine stack, registers, and other memory related 278 information are omitted. 279 280 [Race Detector article]: https://go.dev/doc/articles/race_detector 281 */ 282 package runtime 283 284 import ( 285 "internal/goarch" 286 "internal/goos" 287 ) 288 289 // Caller reports file and line number information about function invocations on 290 // the calling goroutine's stack. The argument skip is the number of stack frames 291 // to ascend, with 0 identifying the caller of Caller. (For historical reasons the 292 // meaning of skip differs between Caller and [Callers].) The return values report the 293 // program counter, file name, and line number within the file of the corresponding 294 // call. The boolean ok is false if it was not possible to recover the information. 295 func Caller(skip int) (pc uintptr, file string, line int, ok bool) { 296 rpc := make([]uintptr, 1) 297 n := callers(skip+1, rpc[:]) 298 if n < 1 { 299 return 300 } 301 frame, _ := CallersFrames(rpc).Next() 302 return frame.PC, frame.File, frame.Line, frame.PC != 0 303 } 304 305 // Callers fills the slice pc with the return program counters of function invocations 306 // on the calling goroutine's stack. The argument skip is the number of stack frames 307 // to skip before recording in pc, with 0 identifying the frame for Callers itself and 308 // 1 identifying the caller of Callers. 309 // It returns the number of entries written to pc. 310 // 311 // To translate these PCs into symbolic information such as function 312 // names and line numbers, use [CallersFrames]. CallersFrames accounts 313 // for inlined functions and adjusts the return program counters into 314 // call program counters. Iterating over the returned slice of PCs 315 // directly is discouraged, as is using [FuncForPC] on any of the 316 // returned PCs, since these cannot account for inlining or return 317 // program counter adjustment. 318 func Callers(skip int, pc []uintptr) int { 319 // runtime.callers uses pc.array==nil as a signal 320 // to print a stack trace. Pick off 0-length pc here 321 // so that we don't let a nil pc slice get to it. 322 if len(pc) == 0 { 323 return 0 324 } 325 return callers(skip, pc) 326 } 327 328 var defaultGOROOT string // set by cmd/link 329 330 // GOROOT returns the root of the Go tree. It uses the 331 // GOROOT environment variable, if set at process start, 332 // or else the root used during the Go build. 333 func GOROOT() string { 334 s := gogetenv("GOROOT") 335 if s != "" { 336 return s 337 } 338 return defaultGOROOT 339 } 340 341 // buildVersion is the Go tree's version string at build time. 342 // 343 // If any GOEXPERIMENTs are set to non-default values, it will include 344 // "X:<GOEXPERIMENT>". 345 // 346 // This is set by the linker. 347 // 348 // This is accessed by "go version <binary>". 349 var buildVersion string 350 351 // Version returns the Go tree's version string. 352 // It is either the commit hash and date at the time of the build or, 353 // when possible, a release tag like "go1.3". 354 func Version() string { 355 return buildVersion 356 } 357 358 // GOOS is the running program's operating system target: 359 // one of darwin, freebsd, linux, and so on. 360 // To view possible combinations of GOOS and GOARCH, run "go tool dist list". 361 const GOOS string = goos.GOOS 362 363 // GOARCH is the running program's architecture target: 364 // one of 386, amd64, arm, s390x, and so on. 365 const GOARCH string = goarch.GOARCH 366