Support/Unix/Program.inc

//===- llvm/Support/Unix/Program.inc ----------------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements the Unix specific portion of the Program class.
//
//===----------------------------------------------------------------------===//

//===----------------------------------------------------------------------===//
//=== WARNING: Implementation here must contain only generic UNIX
//===          code that is guaranteed to work on *all* UNIX variants.
//===----------------------------------------------------------------------===//

#include "llvm/Support/Program.h"

#include "Unix.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Config/config.h"
#include "llvm/Support/AutoConvert.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Errc.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/Path.h"
#include "llvm/Support/StringSaver.h"
#include "llvm/Support/SystemZ/zOSSupport.h"
#include "llvm/Support/raw_ostream.h"
#include <sys/stat.h>
#include <sys/resource.h>
#include <signal.h>
#include <fcntl.h>
#if HAVE_UNISTD_H
#include <unistd.h>
#endif
#ifdef HAVE_POSIX_SPAWN
#include <spawn.h>

#if defined(__APPLE__)
#include <TargetConditionals.h>
#endif

#if defined(__APPLE__) && !(defined(TARGET_OS_IPHONE) && TARGET_OS_IPHONE)
#define USE_NSGETENVIRON 1
#else
#define USE_NSGETENVIRON 0
#endif

#if !USE_NSGETENVIRON
extern char **environ;
#else
#include <crt_externs.h> // _NSGetEnviron
#endif
#endif

using namespace llvm;
using namespace sys;

ProcessInfo::ProcessInfo() : Pid(0), ReturnCode(0) {}

ErrorOr<std::string> sys::findProgramByName(StringRef Name,
                                            ArrayRef<StringRef> Paths) {
  assert(!Name.empty() && "Must have a name!");
  // Use the given path verbatim if it contains any slashes; this matches
  // the behavior of sh(1) and friends.
  if (Name.contains('/'))
    return std::string(Name);

  SmallVector<StringRef, 16> EnvironmentPaths;
  if (Paths.empty())
    if (const char *PathEnv = std::getenv("PATH")) {
      SplitString(PathEnv, EnvironmentPaths, ":");
      Paths = EnvironmentPaths;
    }

  for (auto Path : Paths) {
    if (Path.empty())
      continue;

    // Check to see if this first directory contains the executable...
    SmallString<128> FilePath(Path);
    sys::path::append(FilePath, Name);
    if (sys::fs::can_execute(FilePath.c_str()))
      return std::string(FilePath); // Found the executable!
  }
  return errc::no_such_file_or_directory;
}

static bool RedirectIO(std::optional<StringRef> Path, int FD, std::string *ErrMsg) {
  if (!Path) // Noop
    return false;
  std::string File;
  if (Path->empty())
    // Redirect empty paths to /dev/null
    File = "/dev/null";
  else
    File = std::string(*Path);

  // Open the file
  int InFD = open(File.c_str(), FD == 0 ? O_RDONLY : O_WRONLY | O_CREAT, 0666);
  if (InFD == -1) {
    MakeErrMsg(ErrMsg, "Cannot open file '" + File + "' for " +
                           (FD == 0 ? "input" : "output"));
    return true;
  }

  // Install it as the requested FD
  if (dup2(InFD, FD) == -1) {
    MakeErrMsg(ErrMsg, "Cannot dup2");
    close(InFD);
    return true;
  }
  close(InFD); // Close the original FD
  return false;
}

#ifdef HAVE_POSIX_SPAWN
static bool RedirectIO_PS(const std::string *Path, int FD, std::string *ErrMsg,
                          posix_spawn_file_actions_t *FileActions) {
  if (!Path) // Noop
    return false;
  const char *File;
  if (Path->empty())
    // Redirect empty paths to /dev/null
    File = "/dev/null";
  else
    File = Path->c_str();

  if (int Err = posix_spawn_file_actions_addopen(
          FileActions, FD, File, FD == 0 ? O_RDONLY : O_WRONLY | O_CREAT, 0666))
    return MakeErrMsg(ErrMsg, "Cannot posix_spawn_file_actions_addopen", Err);
  return false;
}
#endif

static void TimeOutHandler(int Sig) {}

static void SetMemoryLimits(unsigned size) {
  struct rlimit r;
  __typeof__(r.rlim_cur) limit = (__typeof__(r.rlim_cur))(size)*1048576;

  // Heap size
  getrlimit(RLIMIT_DATA, &r);
  r.rlim_cur = limit;
  setrlimit(RLIMIT_DATA, &r);
#ifdef RLIMIT_RSS
  // Resident set size.
  getrlimit(RLIMIT_RSS, &r);
  r.rlim_cur = limit;
  setrlimit(RLIMIT_RSS, &r);
#endif
}

static std::vector<const char *>
toNullTerminatedCStringArray(ArrayRef<StringRef> Strings, StringSaver &Saver) {
  std::vector<const char *> Result;
  for (StringRef S : Strings)
    Result.push_back(Saver.save(S).data());
  Result.push_back(nullptr);
  return Result;
}

static bool Execute(ProcessInfo &PI, StringRef Program,
                    ArrayRef<StringRef> Args,
                    std::optional<ArrayRef<StringRef>> Env,
                    ArrayRef<std::optional<StringRef>> Redirects,
                    unsigned MemoryLimit, std::string *ErrMsg,
                    BitVector *AffinityMask, bool DetachProcess) {
  assert(!AffinityMask && "Starting a process with an affinity mask is "
                          "currently not supported on Unix!");

  BumpPtrAllocator Allocator;
  StringSaver Saver(Allocator);
  std::vector<const char *> ArgVector, EnvVector;
  const char **Argv = nullptr;
  const char **Envp = nullptr;
  ArgVector = toNullTerminatedCStringArray(Args, Saver);
  Argv = ArgVector.data();
  if (Env) {
    EnvVector = toNullTerminatedCStringArray(*Env, Saver);
    Envp = EnvVector.data();
  }

  // If this OS has posix_spawn and there is no memory limit being implied, use
  // posix_spawn.  It is more efficient than fork/exec.
#ifdef HAVE_POSIX_SPAWN
  // Cannot use posix_spawn if you would like to detach the process
  if (MemoryLimit == 0 && !DetachProcess) {
    posix_spawn_file_actions_t FileActionsStore;
    posix_spawn_file_actions_t *FileActions = nullptr;

    // If we call posix_spawn_file_actions_addopen we have to make sure the
    // c strings we pass to it stay alive until the call to posix_spawn,
    // so we copy any StringRefs into this variable.
    std::string RedirectsStorage[3];

    if (!Redirects.empty()) {
      assert(Redirects.size() == 3);
      std::string *RedirectsStr[3] = {nullptr, nullptr, nullptr};
      for (int I = 0; I < 3; ++I) {
        if (Redirects[I]) {
          RedirectsStorage[I] = std::string(*Redirects[I]);
          RedirectsStr[I] = &RedirectsStorage[I];
        }
      }

      FileActions = &FileActionsStore;
      posix_spawn_file_actions_init(FileActions);

      // Redirect stdin/stdout.
      if (RedirectIO_PS(RedirectsStr[0], 0, ErrMsg, FileActions) ||
          RedirectIO_PS(RedirectsStr[1], 1, ErrMsg, FileActions))
        return false;
      if (!Redirects[1] || !Redirects[2] || *Redirects[1] != *Redirects[2]) {
        // Just redirect stderr
        if (RedirectIO_PS(RedirectsStr[2], 2, ErrMsg, FileActions))
          return false;
      } else {
        // If stdout and stderr should go to the same place, redirect stderr
        // to the FD already open for stdout.
        if (int Err = posix_spawn_file_actions_adddup2(FileActions, 1, 2))
          return !MakeErrMsg(ErrMsg, "Can't redirect stderr to stdout", Err);
      }
    }

    if (!Envp)
#if !USE_NSGETENVIRON
      Envp = const_cast<const char **>(environ);
#else
      // environ is missing in dylibs.
      Envp = const_cast<const char **>(*_NSGetEnviron());
#endif

    constexpr int maxRetries = 8;
    int retries = 0;
    pid_t PID;
    int Err;
    do {
      PID = 0; // Make Valgrind happy.
      Err = posix_spawn(&PID, Program.str().c_str(), FileActions,
                        /*attrp*/ nullptr, const_cast<char **>(Argv),
                        const_cast<char **>(Envp));
    } while (Err == EINTR && ++retries < maxRetries);

    if (FileActions)
      posix_spawn_file_actions_destroy(FileActions);

    if (Err)
      return !MakeErrMsg(ErrMsg, "posix_spawn failed", Err);

    PI.Pid = PID;
    PI.Process = PID;

    return true;
  }
#endif // HAVE_POSIX_SPAWN

  // Create a child process.
  int child = fork();
  switch (child) {
  // An error occurred:  Return to the caller.
  case -1:
    MakeErrMsg(ErrMsg, "Couldn't fork");
    return false;

  // Child process: Execute the program.
  case 0: {
    // Redirect file descriptors...
    if (!Redirects.empty()) {
      // Redirect stdin
      if (RedirectIO(Redirects[0], 0, ErrMsg)) {
        return false;
      }
      // Redirect stdout
      if (RedirectIO(Redirects[1], 1, ErrMsg)) {
        return false;
      }
      if (Redirects[1] && Redirects[2] && *Redirects[1] == *Redirects[2]) {
        // If stdout and stderr should go to the same place, redirect stderr
        // to the FD already open for stdout.
        if (-1 == dup2(1, 2)) {
          MakeErrMsg(ErrMsg, "Can't redirect stderr to stdout");
          return false;
        }
      } else {
        // Just redirect stderr
        if (RedirectIO(Redirects[2], 2, ErrMsg)) {
          return false;
        }
      }
    }

    if (DetachProcess) {
      // Detach from controlling terminal
      if (::setsid() == -1) {
        MakeErrMsg(ErrMsg, "Could not detach process, ::setsid failed");
        return false;
      }
    }

    // Set memory limits
    if (MemoryLimit != 0) {
      SetMemoryLimits(MemoryLimit);
    }

    // Execute!
    std::string PathStr = std::string(Program);
    if (Envp != nullptr)
      execve(PathStr.c_str(), const_cast<char **>(Argv),
             const_cast<char **>(Envp));
    else
      execv(PathStr.c_str(), const_cast<char **>(Argv));
    // If the execve() failed, we should exit. Follow Unix protocol and
    // return 127 if the executable was not found, and 126 otherwise.
    // Use _exit rather than exit so that atexit functions and static
    // object destructors cloned from the parent process aren't
    // redundantly run, and so that any data buffered in stdio buffers
    // cloned from the parent aren't redundantly written out.
    _exit(errno == ENOENT ? 127 : 126);
  }

  // Parent process: Break out of the switch to do our processing.
  default:
    break;
  }

  PI.Pid = child;
  PI.Process = child;

  return true;
}

namespace llvm {
namespace sys {

#if defined(_AIX)
static pid_t(wait4)(pid_t pid, int *status, int options, struct rusage *usage);
#elif !defined(__Fuchsia__)
using ::wait4;
#endif

} // namespace sys
} // namespace llvm

#ifdef _AIX
#ifndef _ALL_SOURCE
extern "C" pid_t(wait4)(pid_t pid, int *status, int options,
                        struct rusage *usage);
#endif
pid_t(llvm::sys::wait4)(pid_t pid, int *status, int options,
                        struct rusage *usage) {
  assert(pid > 0 && "Only expecting to handle actual PID values!");
  assert((options & ~WNOHANG) == 0 && "Expecting WNOHANG at most!");
  assert(usage && "Expecting usage collection!");

  // AIX wait4 does not work well with WNOHANG.
  if (!(options & WNOHANG))
    return ::wait4(pid, status, options, usage);

  // For WNOHANG, we use waitid (which supports WNOWAIT) until the child process
  // has terminated.
  siginfo_t WaitIdInfo;
  WaitIdInfo.si_pid = 0;
  int WaitIdRetVal =
      waitid(P_PID, pid, &WaitIdInfo, WNOWAIT | WEXITED | options);

  if (WaitIdRetVal == -1 || WaitIdInfo.si_pid == 0)
    return WaitIdRetVal;

  assert(WaitIdInfo.si_pid == pid);

  // The child has already terminated, so a blocking wait on it is okay in the
  // absence of indiscriminate `wait` calls from the current process (which
  // would cause the call here to fail with ECHILD).
  return ::wait4(pid, status, options & ~WNOHANG, usage);
}
#endif

ProcessInfo llvm::sys::Wait(const ProcessInfo &PI,
                            std::optional<unsigned> SecondsToWait,
                            std::string *ErrMsg,
                            std::optional<ProcessStatistics> *ProcStat,
                            bool Polling) {
  struct sigaction Act, Old;
  assert(PI.Pid && "invalid pid to wait on, process not started?");

  int WaitPidOptions = 0;
  pid_t ChildPid = PI.Pid;
  bool WaitUntilTerminates = false;
  if (!SecondsToWait) {
    WaitUntilTerminates = true;
  } else {
    if (*SecondsToWait == 0)
      WaitPidOptions = WNOHANG;

    // Install a timeout handler.  The handler itself does nothing, but the
    // simple fact of having a handler at all causes the wait below to return
    // with EINTR, unlike if we used SIG_IGN.
    memset(&Act, 0, sizeof(Act));
    Act.sa_handler = TimeOutHandler;
    sigemptyset(&Act.sa_mask);
    sigaction(SIGALRM, &Act, &Old);
    // FIXME The alarm signal may be delivered to another thread.
    alarm(*SecondsToWait);
  }

  // Parent process: Wait for the child process to terminate.
  int status = 0;
  ProcessInfo WaitResult;
#ifndef __Fuchsia__
  rusage Info;
  if (ProcStat)
    ProcStat->reset();

  do {
    WaitResult.Pid = sys::wait4(ChildPid, &status, WaitPidOptions, &Info);
  } while (WaitUntilTerminates && WaitResult.Pid == -1 && errno == EINTR);
#endif

  if (WaitResult.Pid != PI.Pid) {
    if (WaitResult.Pid == 0) {
      // Non-blocking wait.
      return WaitResult;
    } else {
      if (SecondsToWait && errno == EINTR && !Polling) {
        // Kill the child.
        kill(PI.Pid, SIGKILL);

        // Turn off the alarm and restore the signal handler
        alarm(0);
        sigaction(SIGALRM, &Old, nullptr);

        // Wait for child to die
        // FIXME This could grab some other child process out from another
        // waiting thread and then leave a zombie anyway.
        if (wait(&status) != ChildPid)
          MakeErrMsg(ErrMsg, "Child timed out but wouldn't die");
        else
          MakeErrMsg(ErrMsg, "Child timed out", 0);

        WaitResult.ReturnCode = -2; // Timeout detected
        return WaitResult;
      } else if (errno != EINTR) {
        MakeErrMsg(ErrMsg, "Error waiting for child process");
        WaitResult.ReturnCode = -1;
        return WaitResult;
      }
    }
  }

  // We exited normally without timeout, so turn off the timer.
  if (SecondsToWait && !WaitUntilTerminates) {
    alarm(0);
    sigaction(SIGALRM, &Old, nullptr);
  }

#ifndef __Fuchsia__
  if (ProcStat) {
    std::chrono::microseconds UserT = toDuration(Info.ru_utime);
    std::chrono::microseconds KernelT = toDuration(Info.ru_stime);
    uint64_t PeakMemory = 0;
#if !defined(__HAIKU__) && !defined(__MVS__)
    PeakMemory = static_cast<uint64_t>(Info.ru_maxrss);
#endif
    *ProcStat = ProcessStatistics{UserT + KernelT, UserT, PeakMemory};
  }
#endif

  // Return the proper exit status. Detect error conditions
  // so we can return -1 for them and set ErrMsg informatively.
  int result = 0;
  if (WIFEXITED(status)) {
    result = WEXITSTATUS(status);
    WaitResult.ReturnCode = result;

    if (result == 127) {
      if (ErrMsg)
        *ErrMsg = llvm::sys::StrError(ENOENT);
      WaitResult.ReturnCode = -1;
      return WaitResult;
    }
    if (result == 126) {
      if (ErrMsg)
        *ErrMsg = "Program could not be executed";
      WaitResult.ReturnCode = -1;
      return WaitResult;
    }
  } else if (WIFSIGNALED(status)) {
    if (ErrMsg) {
      *ErrMsg = strsignal(WTERMSIG(status));
#ifdef WCOREDUMP
      if (WCOREDUMP(status))
        *ErrMsg += " (core dumped)";
#endif
    }
    // Return a special value to indicate that the process received an unhandled
    // signal during execution as opposed to failing to execute.
    WaitResult.ReturnCode = -2;
  }
  return WaitResult;
}

std::error_code llvm::sys::ChangeStdinMode(fs::OpenFlags Flags) {
  if (!(Flags & fs::OF_Text))
    return ChangeStdinToBinary();
  return std::error_code();
}

std::error_code llvm::sys::ChangeStdoutMode(fs::OpenFlags Flags) {
  if (!(Flags & fs::OF_Text))
    return ChangeStdoutToBinary();
  return std::error_code();
}

std::error_code llvm::sys::ChangeStdinToBinary() {
#ifdef __MVS__
  return disableAutoConversion(STDIN_FILENO);
#else
  // Do nothing, as Unix doesn't differentiate between text and binary.
  return std::error_code();
#endif
}

std::error_code llvm::sys::ChangeStdoutToBinary() {
  // Do nothing, as Unix doesn't differentiate between text and binary.
  return std::error_code();
}

std::error_code
llvm::sys::writeFileWithEncoding(StringRef FileName, StringRef Contents,
                                 WindowsEncodingMethod Encoding /*unused*/) {
  std::error_code EC;
  llvm::raw_fd_ostream OS(FileName, EC,
                          llvm::sys::fs::OpenFlags::OF_TextWithCRLF);

  if (EC)
    return EC;

  OS << Contents;

  if (OS.has_error())
    return make_error_code(errc::io_error);

  return EC;
}

bool llvm::sys::commandLineFitsWithinSystemLimits(StringRef Program,
                                                  ArrayRef<StringRef> Args) {
  static long ArgMax = sysconf(_SC_ARG_MAX);
  // POSIX requires that _POSIX_ARG_MAX is 4096, which is the lowest possible
  // value for ARG_MAX on a POSIX compliant system.
  static long ArgMin = _POSIX_ARG_MAX;

  // This the same baseline used by xargs.
  long EffectiveArgMax = 128 * 1024;

  if (EffectiveArgMax > ArgMax)
    EffectiveArgMax = ArgMax;
  else if (EffectiveArgMax < ArgMin)
    EffectiveArgMax = ArgMin;

  // System says no practical limit.
  if (ArgMax == -1)
    return true;

  // Conservatively account for space required by environment variables.
  long HalfArgMax = EffectiveArgMax / 2;

  size_t ArgLength = Program.size() + 1;
  for (StringRef Arg : Args) {
    // Ensure that we do not exceed the MAX_ARG_STRLEN constant on Linux, which
    // does not have a constant unlike what the man pages would have you
    // believe. Since this limit is pretty high, perform the check
    // unconditionally rather than trying to be aggressive and limiting it to
    // Linux only.
    if (Arg.size() >= (32 * 4096))
      return false;

    ArgLength += Arg.size() + 1;
    if (ArgLength > size_t(HalfArgMax)) {
      return false;
    }
  }

  return true;
}