std.os: spring-cleanup and specify organization

- Alphabetically sort things, where reasonable.
- Document, that **only** non-portable posix things belong into posix.zig
  * If there is a portable abstraction, do not offer one in posix.zig
  * Reason: Prevent useless abstractions and needless strong coupling.
- Move posix-only functions into posix.zig, which have either incompatible
  or more extensive execution semantics than their counterparts and can be
  grouped into
  * File permission system
  * Process management
  * Memory management
  * IPC
  * Signaling

Work on ziglang#6600.

Author: matu3ba

lib/std/Thread.zig

@@ -946,7 +946,7 @@ const LinuxThreadImpl = struct {
         // map all memory needed without read/write permissions
         // to avoid committing the whole region right away
         // anonymous mapping ensures file descriptor limits are not exceeded
-        const mapped = os.mmap(
+        const mapped = os.posix.mmap(
             null,
             map_bytes,
             os.PROT.NONE,
@@ -962,7 +962,7 @@ const LinuxThreadImpl = struct {
             else => |e| return e,
         };
         assert(mapped.len >= map_bytes);
-        errdefer os.munmap(mapped);
+        errdefer os.posix.munmap(mapped);
 
         // map everything but the guard page as read/write
         os.mprotect(
@@ -1035,7 +1035,7 @@ const LinuxThreadImpl = struct {
     }
 
     fn join(self: Impl) void {
-        defer os.munmap(self.thread.mapped);
+        defer os.posix.munmap(self.thread.mapped);
 
         var spin: u8 = 10;
         while (true) {

lib/std/child_process.zig

@@ -5,6 +5,7 @@ const unicode = std.unicode;
 const io = std.io;
 const fs = std.fs;
 const os = std.os;
+const posix = os.posix;
 const process = std.process;
 const File = std.fs.File;
 const windows = os.windows;
@@ -70,7 +71,7 @@ pub const ChildProcess = struct {
     /// Darwin-only. Start child process in suspended state as if SIGSTOP was sent.
     start_suspended: bool = false,
 
-    pub const Arg0Expand = os.Arg0Expand;
+    pub const Arg0Expand = os.posix.Arg0Expand;
 
     pub const SpawnError = error{
         OutOfMemory,
@@ -86,8 +87,8 @@ pub const ChildProcess = struct {
         /// Windows-only. `cwd` was provided, but the path did not exist when spawning the child process.
         CurrentWorkingDirectoryUnlinked,
     } ||
-        os.ExecveError ||
-        os.SetIdError ||
+        os.posix.ExecveError ||
+        os.posix.SetIdError ||
         os.ChangeCurDirError ||
         windows.CreateProcessError ||
         windows.WaitForSingleObjectError;
@@ -179,7 +180,7 @@ pub const ChildProcess = struct {
             self.cleanupStreams();
             return term;
         }
-        try os.kill(self.id, os.SIG.TERM);
+        try posix.kill(self.pid, os.SIG.TERM);
         try self.waitUnwrapped();
         return self.term.?;
     }
@@ -309,7 +310,7 @@ pub const ChildProcess = struct {
             return term;
         }
 
-        try self.waitUnwrapped();
+        try self.waitUnwrappedPosix();
         return self.term.?;
     }
 
@@ -331,8 +332,8 @@ pub const ChildProcess = struct {
         return result;
     }
 
-    fn waitUnwrapped(self: *ChildProcess) !void {
-        const res: os.WaitPidResult = os.waitpid(self.id, 0);
+    fn waitUnwrappedPosix(self: *ChildProcess) !void {
+        const res: os.posix.WaitPidResult = os.posix.waitpid(self.id, 0);
         const status = res.status;
         self.cleanupStreams();
         self.handleWaitResult(status);
@@ -410,17 +411,17 @@ pub const ChildProcess = struct {
 
     fn spawnPosix(self: *ChildProcess) SpawnError!void {
         const pipe_flags = if (io.is_async) os.O.NONBLOCK else 0;
-        const stdin_pipe = if (self.stdin_behavior == StdIo.Pipe) try os.pipe2(pipe_flags) else undefined;
+        const stdin_pipe = if (self.stdin_behavior == StdIo.Pipe) try os.posix.pipe2(pipe_flags) else undefined;
         errdefer if (self.stdin_behavior == StdIo.Pipe) {
             destroyPipe(stdin_pipe);
         };
 
-        const stdout_pipe = if (self.stdout_behavior == StdIo.Pipe) try os.pipe2(pipe_flags) else undefined;
+        const stdout_pipe = if (self.stdout_behavior == StdIo.Pipe) try os.posix.pipe2(pipe_flags) else undefined;
         errdefer if (self.stdout_behavior == StdIo.Pipe) {
             destroyPipe(stdout_pipe);
         };
 
-        const stderr_pipe = if (self.stderr_behavior == StdIo.Pipe) try os.pipe2(pipe_flags) else undefined;
+        const stderr_pipe = if (self.stderr_behavior == StdIo.Pipe) try os.posix.pipe2(pipe_flags) else undefined;
         errdefer if (self.stderr_behavior == StdIo.Pipe) {
             destroyPipe(stderr_pipe);
         };
@@ -485,12 +486,12 @@ pub const ChildProcess = struct {
                 // end with eventfd
                 break :blk [2]os.fd_t{ fd, fd };
             } else {
-                break :blk try os.pipe2(os.O.CLOEXEC);
+                break :blk try os.posix.pipe2(os.O.CLOEXEC);
             }
         };
         errdefer destroyPipe(err_pipe);
 
-        const pid_result = try os.fork();
+        const pid_result = try os.posix.fork();
         if (pid_result == 0) {
             // we are the child
             setUpChildIo(self.stdin_behavior, stdin_pipe[0], os.STDIN_FILENO, dev_null_fd) catch |err| forkChildErrReport(err_pipe[1], err);
@@ -517,16 +518,16 @@ pub const ChildProcess = struct {
             }
 
             if (self.gid) |gid| {
-                os.setregid(gid, gid) catch |err| forkChildErrReport(err_pipe[1], err);
+                os.posix.setregid(gid, gid) catch |err| forkChildErrReport(err_pipe[1], err);
             }
 
             if (self.uid) |uid| {
-                os.setreuid(uid, uid) catch |err| forkChildErrReport(err_pipe[1], err);
+                os.posix.setreuid(uid, uid) catch |err| forkChildErrReport(err_pipe[1], err);
             }
 
             const err = switch (self.expand_arg0) {
-                .expand => os.execvpeZ_expandArg0(.expand, argv_buf.ptr[0].?, argv_buf.ptr, envp),
-                .no_expand => os.execvpeZ_expandArg0(.no_expand, argv_buf.ptr[0].?, argv_buf.ptr, envp),
+                .expand => os.posix.execvpeZ_expandArg0(.expand, argv_buf.ptr[0].?, argv_buf.ptr, envp),
+                .no_expand => os.posix.execvpeZ_expandArg0(.no_expand, argv_buf.ptr[0].?, argv_buf.ptr, envp),
             };
             forkChildErrReport(err_pipe[1], err);
         }
@@ -832,10 +833,10 @@ pub const ChildProcess = struct {
 
     fn setUpChildIo(stdio: StdIo, pipe_fd: i32, std_fileno: i32, dev_null_fd: i32) !void {
         switch (stdio) {
-            .Pipe => try os.dup2(pipe_fd, std_fileno),
+            .Pipe => try os.posix.dup2(pipe_fd, std_fileno),
             .Close => os.close(std_fileno),
             .Inherit => {},
-            .Ignore => try os.dup2(dev_null_fd, std_fileno),
+            .Ignore => try os.posix.dup2(dev_null_fd, std_fileno),
         }
     }
 };

lib/std/crypto/tlcsprng.zig

@@ -75,7 +75,7 @@ fn tlsCsprngFill(_: *anyopaque, buffer: []u8) void {
         if (want_fork_safety and maybe_have_wipe_on_fork or is_haiku) {
             // Allocate a per-process page, madvise operates with page
             // granularity.
-            wipe_mem = os.mmap(
+            wipe_mem = os.posix.mmap(
                 null,
                 @sizeOf(Context),
                 os.PROT.READ | os.PROT.WRITE,
@@ -111,11 +111,11 @@ fn tlsCsprngFill(_: *anyopaque, buffer: []u8) void {
                 // Qemu user-mode emulation ignores any valid/invalid madvise
                 // hint and returns success. Check if this is the case by
                 // passing bogus parameters, we expect EINVAL as result.
-                if (os.madvise(wipe_mem.ptr, 0, 0xffffffff)) |_| {
+                if (os.posix.madvise(wipe_mem.ptr, 0, 0xffffffff)) |_| {
                     break :wof;
                 } else |_| {}
 
-                if (os.madvise(wipe_mem.ptr, wipe_mem.len, os.MADV.WIPEONFORK)) |_| {
+                if (os.posix.madvise(wipe_mem.ptr, wipe_mem.len, os.MADV.WIPEONFORK)) |_| {
                     return initAndFill(buffer);
                 } else |_| {}
             }

lib/std/debug.zig

@@ -482,9 +482,9 @@ pub const StackIterator = struct {
 
         if (native_os != .windows) {
             if (native_os != .wasi) {
-                os.msync(aligned_memory, os.MSF.ASYNC) catch |err| {
+                os.posix.msync(aligned_memory, os.MSF.ASYNC) catch |err| {
                     switch (err) {
-                        os.MSyncError.UnmappedMemory => {
+                        os.posix.MSyncError.UnmappedMemory => {
                             return false;
                         },
                         else => unreachable,
@@ -1222,15 +1222,15 @@ fn mapWholeFile(file: File) ![]align(mem.page_size) const u8 {
         defer file.close();
 
         const file_len = math.cast(usize, try file.getEndPos()) orelse math.maxInt(usize);
-        const mapped_mem = try os.mmap(
+        const mapped_mem = try os.posix.mmap(
             null,
             file_len,
             os.PROT.READ,
             os.MAP.SHARED,
             file.handle,
             0,
         );
-        errdefer os.munmap(mapped_mem);
+        errdefer os.posix.munmap(mapped_mem);
 
         return mapped_mem;
     }
@@ -1491,7 +1491,7 @@ pub const ModuleDebugInfo = switch (native_os) {
             }
             self.ofiles.deinit();
             allocator.free(self.symbols);
-            os.munmap(self.mapped_memory);
+            os.posix.munmap(self.mapped_memory);
         }
 
         fn loadOFile(self: *@This(), allocator: mem.Allocator, o_file_path: []const u8) !OFileInfo {
@@ -1798,7 +1798,7 @@ pub const ModuleDebugInfo = switch (native_os) {
 
         fn deinit(self: *@This(), allocator: mem.Allocator) void {
             self.dwarf.deinit(allocator);
-            os.munmap(self.mapped_memory);
+            os.posix.munmap(self.mapped_memory);
         }
 
         pub fn getSymbolAtAddress(self: *@This(), allocator: mem.Allocator, address: usize) !SymbolInfo {
@@ -1880,10 +1880,10 @@ pub fn maybeEnableSegfaultHandler() void {
 var windows_segfault_handle: ?windows.HANDLE = null;
 
 pub fn updateSegfaultHandler(act: ?*const os.Sigaction) error{OperationNotSupported}!void {
-    try os.sigaction(os.SIG.SEGV, act, null);
-    try os.sigaction(os.SIG.ILL, act, null);
-    try os.sigaction(os.SIG.BUS, act, null);
-    try os.sigaction(os.SIG.FPE, act, null);
+    try os.posix.sigaction(os.SIG.SEGV, act, null);
+    try os.posix.sigaction(os.SIG.ILL, act, null);
+    try os.posix.sigaction(os.SIG.BUS, act, null);
+    try os.posix.sigaction(os.SIG.FPE, act, null);
 }
 
 /// Attaches a global SIGSEGV handler which calls @panic("segmentation fault");

lib/std/dynamic_library.zig

@@ -123,15 +123,15 @@ pub const ElfDynLib = struct {
 
         // This one is to read the ELF info. We do more mmapping later
         // corresponding to the actual LOAD sections.
-        const file_bytes = try os.mmap(
+        const file_bytes = try os.posix.mmap(
             null,
             mem.alignForward(size, mem.page_size),
             os.PROT.READ,
             os.MAP.PRIVATE,
             fd,
             0,
         );
-        defer os.munmap(file_bytes);
+        defer os.posix.munmap(file_bytes);
 
         const eh = @ptrCast(*elf.Ehdr, file_bytes.ptr);
         if (!mem.eql(u8, eh.e_ident[0..4], elf.MAGIC)) return error.NotElfFile;
@@ -161,15 +161,15 @@ pub const ElfDynLib = struct {
         const dynv = maybe_dynv orelse return error.MissingDynamicLinkingInformation;
 
         // Reserve the entire range (with no permissions) so that we can do MAP.FIXED below.
-        const all_loaded_mem = try os.mmap(
+        const all_loaded_mem = try os.posix.mmap(
             null,
             virt_addr_end,
             os.PROT.NONE,
             os.MAP.PRIVATE | os.MAP.ANONYMOUS,
             -1,
             0,
         );
-        errdefer os.munmap(all_loaded_mem);
+        errdefer os.posix.munmap(all_loaded_mem);
 
         const base = @ptrToInt(all_loaded_mem.ptr);
 
@@ -193,7 +193,7 @@ pub const ElfDynLib = struct {
                         const prot = elfToMmapProt(ph.p_flags);
                         if ((ph.p_flags & elf.PF_W) == 0) {
                             // If it does not need write access, it can be mapped from the fd.
-                            _ = try os.mmap(
+                            _ = try os.posix.mmap(
                                 ptr,
                                 extended_memsz,
                                 prot,
@@ -202,7 +202,7 @@ pub const ElfDynLib = struct {
                                 ph.p_offset - extra_bytes,
                             );
                         } else {
-                            const sect_mem = try os.mmap(
+                            const sect_mem = try os.posix.mmap(
                                 ptr,
                                 extended_memsz,
                                 prot,
@@ -256,7 +256,7 @@ pub const ElfDynLib = struct {
 
     /// Trusts the file
     pub fn close(self: *ElfDynLib) void {
-        os.munmap(self.memory);
+        os.posix.munmap(self.memory);
         self.* = undefined;
     }
 

lib/std/fs.zig

@@ -1194,7 +1194,7 @@ pub const Dir = struct {
         }
 
         if (has_flock_open_flags and flags.lock_nonblocking) {
-            var fl_flags = os.fcntl(fd, os.F.GETFL, 0) catch |err| switch (err) {
+            var fl_flags = os.posix.fcntl(fd, os.F.GETFL, 0) catch |err| switch (err) {
                 error.FileBusy => unreachable,
                 error.Locked => unreachable,
                 error.PermissionDenied => unreachable,
@@ -1203,7 +1203,7 @@ pub const Dir = struct {
                 else => |e| return e,
             };
             fl_flags &= ~@as(usize, os.O.NONBLOCK);
-            _ = os.fcntl(fd, os.F.SETFL, fl_flags) catch |err| switch (err) {
+            _ = os.posix.fcntl(fd, os.F.SETFL, fl_flags) catch |err| switch (err) {
                 error.FileBusy => unreachable,
                 error.Locked => unreachable,
                 error.PermissionDenied => unreachable,
@@ -1350,7 +1350,7 @@ pub const Dir = struct {
         }
 
         if (has_flock_open_flags and flags.lock_nonblocking) {
-            var fl_flags = os.fcntl(fd, os.F.GETFL, 0) catch |err| switch (err) {
+            var fl_flags = os.posix.fcntl(fd, os.F.GETFL, 0) catch |err| switch (err) {
                 error.FileBusy => unreachable,
                 error.Locked => unreachable,
                 error.PermissionDenied => unreachable,
@@ -1359,7 +1359,7 @@ pub const Dir = struct {
                 else => |e| return e,
             };
             fl_flags &= ~@as(usize, os.O.NONBLOCK);
-            _ = os.fcntl(fd, os.F.SETFL, fl_flags) catch |err| switch (err) {
+            _ = os.posix.fcntl(fd, os.F.SETFL, fl_flags) catch |err| switch (err) {
                 error.FileBusy => unreachable,
                 error.Locked => unreachable,
                 error.PermissionDenied => unreachable,

lib/std/fs/file.zig

@@ -399,25 +399,25 @@ pub const File = struct {
         return Stat.fromSystem(st);
     }
 
-    pub const ChmodError = std.os.FChmodError;
+    pub const ChmodError = std.os.posix.FChmodError;
 
     /// Changes the mode of the file.
     /// The process must have the correct privileges in order to do this
     /// successfully, or must have the effective user ID matching the owner
     /// of the file.
     pub fn chmod(self: File, new_mode: Mode) ChmodError!void {
-        try os.fchmod(self.handle, new_mode);
+        try os.posix.fchmod(self.handle, new_mode);
     }
 
-    pub const ChownError = std.os.FChownError;
+    pub const ChownError = std.os.posix.FChownError;
 
     /// Changes the owner and group of the file.
     /// The process must have the correct privileges in order to do this
     /// successfully. The group may be changed by the owner of the file to
     /// any group of which the owner is a member. If the owner or group is
     /// specified as `null`, the ID is not changed.
     pub fn chown(self: File, owner: ?Uid, group: ?Gid) ChownError!void {
-        try os.fchown(self.handle, owner, group);
+        try os.posix.fchown(self.handle, owner, group);
     }
 
     /// Cross-platform representation of permissions on a file.
@@ -899,7 +899,7 @@ pub const File = struct {
         };
     }
 
-    pub const UpdateTimesError = os.FutimensError || windows.SetFileTimeError;
+    pub const UpdateTimesError = os.posix.FutimensError || windows.SetFileTimeError;
 
     /// The underlying file system may have a different granularity than nanoseconds,
     /// and therefore this function cannot guarantee any precision will be stored.
@@ -928,7 +928,7 @@ pub const File = struct {
                 .tv_nsec = math.cast(isize, @mod(mtime, std.time.ns_per_s)) orelse maxInt(isize),
             },
         };
-        try os.futimens(self.handle, &times);
+        try os.posix.futimens(self.handle, &times);
     }
 
     /// Reads all the bytes from the current position to the end of the file.