Source file src/runtime/extern.go

     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  

View as plain text