zig/std/os/index.zig
Andrew Kelley c9ae30d27e delete alloca builtin function
See #225

introduce os.EnvMap
2017-04-03 18:11:57 -04:00

656 lines
22 KiB
Zig

pub const windows = @import("windows.zig");
pub const darwin = @import("darwin.zig");
pub const linux = @import("linux.zig");
pub const posix = switch(@compileVar("os")) {
Os.linux => linux,
Os.darwin, Os.macosx, Os.ios => darwin,
Os.windows => windows,
else => @compileError("Unsupported OS"),
};
pub const max_noalloc_path_len = 1024;
const debug = @import("../debug.zig");
const assert = debug.assert;
const errno = @import("errno.zig");
const linking_libc = @import("../target.zig").linking_libc;
const c = @import("../c/index.zig");
const mem = @import("../mem.zig");
const Allocator = mem.Allocator;
const io = @import("../io.zig");
const HashMap = @import("../hash_map.zig").HashMap;
const cstr = @import("../cstr.zig");
error Unexpected;
error SysResources;
error AccessDenied;
error InvalidExe;
error FileSystem;
error IsDir;
error FileNotFound;
error FileBusy;
/// Fills `buf` with random bytes. If linking against libc, this calls the
/// appropriate OS-specific library call. Otherwise it uses the zig standard
/// library implementation.
pub fn getRandomBytes(buf: []u8) -> %void {
while (true) {
const err = switch (@compileVar("os")) {
Os.linux => {
if (linking_libc) {
if (c.getrandom(buf.ptr, buf.len, 0) == -1) *c._errno() else 0
} else {
posix.getErrno(posix.getrandom(buf.ptr, buf.len, 0))
}
},
Os.darwin, Os.macosx, Os.ios => {
if (linking_libc) {
if (posix.getrandom(buf.ptr, buf.len) == -1) *c._errno() else 0
} else {
posix.getErrno(posix.getrandom(buf.ptr, buf.len))
}
},
Os.windows => {
var hCryptProv: windows.HCRYPTPROV = undefined;
if (!windows.CryptAcquireContext(&hCryptProv, null, null, windows.PROV_RSA_FULL, 0)) {
return error.Unexpected;
}
defer _ = windows.CryptReleaseContext(hCryptProv, 0);
if (!windows.CryptGenRandom(hCryptProv, windows.DWORD(buf.len), buf.ptr)) {
return error.Unexpected;
}
return;
},
else => @compileError("Unsupported OS"),
};
if (err > 0) {
return switch (err) {
errno.EINVAL => unreachable,
errno.EFAULT => unreachable,
errno.EINTR => continue,
else => error.Unexpected,
}
}
return;
}
}
/// Raises a signal in the current kernel thread, ending its execution.
/// If linking against libc, this calls the abort() libc function. Otherwise
/// it uses the zig standard library implementation.
pub coldcc fn abort() -> noreturn {
if (linking_libc) {
c.abort();
}
switch (@compileVar("os")) {
Os.linux, Os.darwin, Os.macosx, Os.ios => {
_ = posix.raise(posix.SIGABRT);
_ = posix.raise(posix.SIGKILL);
while (true) {}
},
else => @compileError("Unsupported OS"),
}
}
fn makePipe() -> %[2]i32 {
var fds: [2]i32 = undefined;
const err = posix.getErrno(posix.pipe(&fds));
if (err > 0) {
return switch (err) {
errno.EMFILE, errno.ENFILE => error.SysResources,
else => error.Unexpected,
}
}
return fds;
}
fn destroyPipe(pipe: &const [2]i32) {
posixClose((*pipe)[0]);
posixClose((*pipe)[1]);
}
/// Calls POSIX close, and keeps trying if it gets interrupted.
pub fn posixClose(fd: i32) {
while (true) {
const err = posix.getErrno(posix.close(fd));
if (err == errno.EINTR) {
continue;
} else {
return;
}
}
}
/// Calls POSIX write, and keeps trying if it gets interrupted.
pub fn posixWrite(fd: i32, bytes: []const u8) -> %void {
while (true) {
const write_ret = posix.write(fd, bytes.ptr, bytes.len);
const write_err = posix.getErrno(write_ret);
if (write_err > 0) {
return switch (write_err) {
errno.EINTR => continue,
errno.EINVAL => unreachable,
errno.EDQUOT => error.DiskQuota,
errno.EFBIG => error.FileTooBig,
errno.EIO => error.Io,
errno.ENOSPC => error.NoSpaceLeft,
errno.EPERM => error.BadPerm,
errno.EPIPE => error.PipeFail,
else => error.Unexpected,
}
}
return;
}
}
/// ::path may need to be copied in memory to add a null terminating byte. In this case
/// a fixed size buffer of size ::max_noalloc_path_len is an attempted solution. If the fixed
/// size buffer is too small, and the provided allocator is null, ::error.NameTooLong is returned.
/// otherwise if the fixed size buffer is too small, allocator is used to obtain the needed memory.
/// Calls POSIX open, keeps trying if it gets interrupted, and translates
/// the return value into zig errors.
pub fn posixOpen(path: []const u8, flags: usize, perm: usize, allocator: ?&Allocator) -> %i32 {
var stack_buf: [max_noalloc_path_len]u8 = undefined;
var path0: []u8 = undefined;
var need_free = false;
if (path.len < stack_buf.len) {
path0 = stack_buf[0...path.len + 1];
} else if (const a ?= allocator) {
path0 = %return a.alloc(u8, path.len + 1);
need_free = true;
} else {
return error.NameTooLong;
}
defer if (need_free) {
(??allocator).free(path0);
};
mem.copy(u8, path0, path);
path0[path.len] = 0;
while (true) {
const result = posix.open(path0.ptr, flags, perm);
const err = posix.getErrno(result);
if (err > 0) {
return switch (err) {
errno.EINTR => continue,
errno.EFAULT => unreachable,
errno.EINVAL => unreachable,
errno.EACCES => error.BadPerm,
errno.EFBIG, errno.EOVERFLOW => error.FileTooBig,
errno.EISDIR => error.IsDir,
errno.ELOOP => error.SymLinkLoop,
errno.EMFILE => error.ProcessFdQuotaExceeded,
errno.ENAMETOOLONG => error.NameTooLong,
errno.ENFILE => error.SystemFdQuotaExceeded,
errno.ENODEV => error.NoDevice,
errno.ENOENT => error.PathNotFound,
errno.ENOMEM => error.NoMem,
errno.ENOSPC => error.NoSpaceLeft,
errno.ENOTDIR => error.NotDir,
errno.EPERM => error.BadPerm,
else => error.Unexpected,
}
}
return i32(result);
}
}
const ErrInt = @intType(false, @sizeOf(error) * 8);
fn writeIntFd(fd: i32, value: ErrInt) -> %void {
var bytes: [@sizeOf(ErrInt)]u8 = undefined;
mem.writeInt(bytes[0...], value, true);
var index: usize = 0;
while (index < bytes.len) {
const amt_written = posix.write(fd, &bytes[index], bytes.len - index);
const err = posix.getErrno(amt_written);
if (err > 0) {
switch (err) {
errno.EINTR => continue,
errno.EINVAL => unreachable,
else => return error.SysResources,
}
}
index += amt_written;
}
}
fn readIntFd(fd: i32) -> %ErrInt {
var bytes: [@sizeOf(ErrInt)]u8 = undefined;
var index: usize = 0;
while (index < bytes.len) {
const amt_written = posix.read(fd, &bytes[index], bytes.len - index);
const err = posix.getErrno(amt_written);
if (err > 0) {
switch (err) {
errno.EINTR => continue,
errno.EINVAL => unreachable,
else => return error.SysResources,
}
}
index += amt_written;
}
return mem.readInt(bytes[0...], ErrInt, true);
}
// Child of fork calls this to report an error to the fork parent.
// Then the child exits.
fn forkChildErrReport(fd: i32, err: error) -> noreturn {
_ = writeIntFd(fd, ErrInt(err));
posix.exit(1);
}
fn dup2NoIntr(old_fd: i32, new_fd: i32) -> %void {
while (true) {
const err = posix.getErrno(posix.dup2(old_fd, new_fd));
if (err > 0) {
return switch (err) {
errno.EBUSY, errno.EINTR => continue,
errno.EMFILE => error.SysResources,
errno.EINVAL => unreachable,
else => error.Unexpected,
};
}
return;
}
}
pub const ChildProcess = struct {
pid: i32,
err_pipe: [2]i32,
stdin: ?io.OutStream,
stdout: ?io.InStream,
stderr: ?io.InStream,
pub const Term = enum {
Clean: i32,
Signal: i32,
Stopped: i32,
Unknown: i32,
};
pub const StdIo = enum {
Inherit,
Ignore,
Pipe,
Close,
};
pub fn spawn(exe_path: []const u8, args: []const []const u8, env_map: &const EnvMap,
stdin: StdIo, stdout: StdIo, stderr: StdIo, allocator: &Allocator) -> %ChildProcess
{
switch (@compileVar("os")) {
Os.linux, Os.macosx, Os.ios, Os.darwin => {
return spawnPosix(exe_path, args, env_map, stdin, stdout, stderr, allocator);
},
else => @compileError("Unsupported OS"),
}
}
pub fn wait(self: &ChildProcess) -> %Term {
defer {
posixClose(self.err_pipe[0]);
posixClose(self.err_pipe[1]);
};
var status: i32 = undefined;
while (true) {
const err = posix.getErrno(posix.waitpid(self.pid, &status, 0));
if (err > 0) {
switch (err) {
errno.EINVAL, errno.ECHILD => unreachable,
errno.EINTR => continue,
else => {
if (const *stdin ?= self.stdin) { stdin.close(); }
if (const *stdout ?= self.stdin) { stdout.close(); }
if (const *stderr ?= self.stdin) { stderr.close(); }
return error.Unexpected;
},
}
}
break;
}
if (const *stdin ?= self.stdin) { stdin.close(); }
if (const *stdout ?= self.stdin) { stdout.close(); }
if (const *stderr ?= self.stdin) { stderr.close(); }
// Write @maxValue(ErrInt) to the write end of the err_pipe. This is after
// waitpid, so this write is guaranteed to be after the child
// pid potentially wrote an error. This way we can do a blocking
// read on the error pipe and either get @maxValue(ErrInt) (no error) or
// an error code.
%return writeIntFd(self.err_pipe[1], @maxValue(ErrInt));
const err_int = %return readIntFd(self.err_pipe[0]);
// Here we potentially return the fork child's error
// from the parent pid.
if (err_int != @maxValue(ErrInt)) {
return error(err_int);
}
return statusToTerm(status);
}
fn statusToTerm(status: i32) -> Term {
return if (posix.WIFEXITED(status)) {
Term.Clean { posix.WEXITSTATUS(status) }
} else if (posix.WIFSIGNALED(status)) {
Term.Signal { posix.WTERMSIG(status) }
} else if (posix.WIFSTOPPED(status)) {
Term.Stopped { posix.WSTOPSIG(status) }
} else {
Term.Unknown { status }
};
}
fn spawnPosix(exe_path: []const u8, args: []const []const u8, env_map: &const EnvMap,
stdin: StdIo, stdout: StdIo, stderr: StdIo, allocator: &Allocator) -> %ChildProcess
{
// TODO issue #295
//const stdin_pipe = if (stdin == StdIo.Pipe) %return makePipe() else undefined;
var stdin_pipe: [2]i32 = undefined;
if (stdin == StdIo.Pipe)
stdin_pipe = %return makePipe();
%defer if (stdin == StdIo.Pipe) { destroyPipe(stdin_pipe); };
// TODO issue #295
//const stdout_pipe = if (stdout == StdIo.Pipe) %return makePipe() else undefined;
var stdout_pipe: [2]i32 = undefined;
if (stdout == StdIo.Pipe)
stdout_pipe = %return makePipe();
%defer if (stdout == StdIo.Pipe) { destroyPipe(stdout_pipe); };
// TODO issue #295
//const stderr_pipe = if (stderr == StdIo.Pipe) %return makePipe() else undefined;
var stderr_pipe: [2]i32 = undefined;
if (stderr == StdIo.Pipe)
stderr_pipe = %return makePipe();
%defer if (stderr == StdIo.Pipe) { destroyPipe(stderr_pipe); };
const any_ignore = (stdin == StdIo.Ignore or stdout == StdIo.Ignore or stderr == StdIo.Ignore);
// TODO issue #295
//const dev_null_fd = if (any_ignore) {
// %return posixOpen("/dev/null", posix.O_RDWR, 0, null)
//} else {
// undefined
//};
var dev_null_fd: i32 = undefined;
if (any_ignore)
dev_null_fd = %return posixOpen("/dev/null", posix.O_RDWR, 0, null);
// This pipe is used to communicate errors between the time of fork
// and execve from the child process to the parent process.
const err_pipe = %return makePipe();
%defer destroyPipe(err_pipe);
const pid = posix.fork();
const pid_err = linux.getErrno(pid);
if (pid_err > 0) {
return switch (pid_err) {
errno.EAGAIN, errno.ENOMEM, errno.ENOSYS => error.SysResources,
else => error.Unexpected,
};
}
if (pid == 0) {
// we are the child
setUpChildIo(stdin, stdin_pipe[0], posix.STDIN_FILENO, dev_null_fd) %%
|err| forkChildErrReport(err_pipe[1], err);
setUpChildIo(stdout, stdout_pipe[1], posix.STDOUT_FILENO, dev_null_fd) %%
|err| forkChildErrReport(err_pipe[1], err);
setUpChildIo(stderr, stderr_pipe[1], posix.STDERR_FILENO, dev_null_fd) %%
|err| forkChildErrReport(err_pipe[1], err);
const err = posix.getErrno(%return execve(exe_path, args, env_map, allocator));
assert(err > 0);
forkChildErrReport(err_pipe[1], switch (err) {
errno.EFAULT => unreachable,
errno.E2BIG, errno.EMFILE, errno.ENAMETOOLONG, errno.ENFILE, errno.ENOMEM => error.SysResources,
errno.EACCES, errno.EPERM => error.AccessDenied,
errno.EINVAL, errno.ENOEXEC => error.InvalidExe,
errno.EIO, errno.ELOOP => error.FileSystem,
errno.EISDIR => error.IsDir,
errno.ENOENT, errno.ENOTDIR => error.FileNotFound,
errno.ETXTBSY => error.FileBusy,
else => error.Unexpected,
});
}
// we are the parent
if (stdin == StdIo.Pipe) { posixClose(stdin_pipe[0]); }
if (stdout == StdIo.Pipe) { posixClose(stdout_pipe[1]); }
if (stderr == StdIo.Pipe) { posixClose(stderr_pipe[1]); }
if (any_ignore) { posixClose(dev_null_fd); }
return ChildProcess {
.pid = i32(pid),
.err_pipe = err_pipe,
.stdin = if (stdin == StdIo.Pipe) {
io.OutStream {
.fd = stdin_pipe[1],
.buffer = undefined,
.index = 0,
}
} else {
null
},
.stdout = if (stdout == StdIo.Pipe) {
io.InStream {
.fd = stdout_pipe[0],
}
} else {
null
},
.stderr = if (stderr == StdIo.Pipe) {
io.InStream {
.fd = stderr_pipe[0],
}
} else {
null
},
};
}
fn setUpChildIo(stdio: StdIo, pipe_fd: i32, std_fileno: i32, dev_null_fd: i32) -> %void {
switch (stdio) {
StdIo.Pipe => %return dup2NoIntr(pipe_fd, std_fileno),
StdIo.Close => posixClose(std_fileno),
StdIo.Inherit => {},
StdIo.Ignore => %return dup2NoIntr(dev_null_fd, std_fileno),
}
}
};
/// This function must allocate memory to add a null terminating bytes on path and each arg.
/// It must also convert to KEY=VALUE\0 format for environment variables, and include null
/// pointers after the args and after the environment variables.
/// Also make the first arg equal to path.
fn execve(path: []const u8, argv: []const []const u8, env_map: &const EnvMap, allocator: &Allocator) -> %usize {
const path_buf = %return allocator.alloc(u8, path.len + 1);
defer allocator.free(path_buf);
@memcpy(&path_buf[0], &path[0], path.len);
path_buf[path.len] = 0;
const argv_buf = %return allocator.alloc(?&const u8, argv.len + 2);
mem.set(?&const u8, argv_buf, null);
defer {
for (argv_buf) |arg, i| {
const arg_buf = if (const ptr ?= arg) ptr[0...argv[i].len + 1] else break;
allocator.free(arg_buf);
}
allocator.free(argv_buf);
}
{
// Add path to the first argument.
const arg_buf = %return allocator.alloc(u8, path.len + 1);
@memcpy(&arg_buf[0], path.ptr, path.len);
arg_buf[path.len] = 0;
argv_buf[0] = arg_buf.ptr;
}
for (argv) |arg, i| {
const arg_buf = %return allocator.alloc(u8, arg.len + 1);
@memcpy(&arg_buf[0], arg.ptr, arg.len);
arg_buf[arg.len] = 0;
argv_buf[i + 1] = arg_buf.ptr;
}
argv_buf[argv.len + 1] = null;
const envp_count = env_map.count();
const envp_buf = %return allocator.alloc(?&const u8, envp_count + 1);
mem.set(?&const u8, envp_buf, null);
defer {
for (envp_buf) |env, i| {
const env_buf = if (const ptr ?= env) ptr[0...cstr.len(ptr)] else break;
allocator.free(env_buf);
}
allocator.free(envp_buf);
}
{
var it = env_map.iterator();
var i: usize = 0;
while (true; i += 1) {
const pair = it.next() ?? break;
const env_buf = %return allocator.alloc(u8, pair.key.len + pair.value.len + 2);
@memcpy(&env_buf[0], pair.key.ptr, pair.key.len);
env_buf[pair.key.len] = '=';
@memcpy(&env_buf[pair.key.len + 1], pair.value.ptr, pair.value.len);
env_buf[env_buf.len - 1] = 0;
envp_buf[i] = env_buf.ptr;
}
assert(i == envp_count);
}
envp_buf[envp_count] = null;
return posix.execve(path_buf.ptr, argv_buf.ptr, envp_buf.ptr);
}
pub var environ_raw: []&u8 = undefined;
pub const EnvMap = struct {
hash_map: EnvHashMap,
const EnvHashMap = HashMap([]const u8, []const u8, hash_slice_u8, eql_slice_u8);
pub fn init(allocator: &Allocator) -> EnvMap {
var self = EnvMap {
.hash_map = undefined,
};
self.hash_map.init(allocator);
return self;
}
pub fn deinit(self: &EnvMap) {
var it = self.hash_map.entryIterator();
while (true) {
const entry = it.next() ?? break;
self.free(entry.key);
self.free(entry.value);
}
self.hash_map.deinit();
}
pub fn set(self: &EnvMap, key: []const u8, value: []const u8) -> %void {
if (const entry ?= self.hash_map.get(key)) {
const value_copy = %return self.copy(value);
%defer self.free(value_copy);
%return self.hash_map.put(key, value_copy);
self.free(entry.value);
} else {
const key_copy = %return self.copy(key);
%defer self.free(key_copy);
const value_copy = %return self.copy(value);
%defer self.free(value_copy);
%return self.hash_map.put(key_copy, value_copy);
}
}
pub fn delete(self: &EnvMap, key: []const u8) {
const entry = self.hash_map.remove(key) ?? return;
self.free(entry.key);
self.free(entry.value);
}
pub fn count(self: &const EnvMap) -> usize {
return self.hash_map.size;
}
pub fn iterator(self: &const EnvMap) -> EnvHashMap.Iterator {
return self.hash_map.entryIterator();
}
fn free(self: &EnvMap, value: []const u8) {
// remove the const
const mut_value = @ptrcast(&u8, value.ptr)[0...value.len];
self.hash_map.allocator.free(mut_value);
}
fn copy(self: &EnvMap, value: []const u8) -> %[]const u8 {
const result = %return self.hash_map.allocator.alloc(u8, value.len);
mem.copy(u8, result, value);
return result;
}
};
pub fn getEnvMap(allocator: &Allocator) -> %EnvMap {
var result = EnvMap.init(allocator);
%defer result.deinit();
for (environ_raw) |ptr| {
var line_i: usize = 0;
while (ptr[line_i] != 0 and ptr[line_i] != '='; line_i += 1) {}
const key = ptr[0...line_i];
var end_i: usize = line_i;
while (ptr[end_i] != 0; end_i += 1) {}
const value = ptr[line_i + 1...end_i];
%return result.set(key, value);
}
return result;
}
pub fn getEnv(key: []const u8) -> ?[]const u8 {
for (environ_raw) |ptr| {
var line_i: usize = 0;
while (ptr[line_i] != 0 and ptr[line_i] != '='; line_i += 1) {}
const this_key = ptr[0...line_i];
if (!mem.eql(u8, key, this_key))
continue;
var end_i: usize = line_i;
while (ptr[end_i] != 0; end_i += 1) {}
const this_value = ptr[line_i + 1...end_i];
return this_value;
}
return null;
}
fn hash_slice_u8(k: []const u8) -> u32 {
// FNV 32-bit hash
var h: u32 = 2166136261;
for (k) |b| {
h = (h ^ b) *% 16777619;
}
return h;
}
fn eql_slice_u8(a: []const u8, b: []const u8) -> bool {
return mem.eql(u8, a, b);
}