bootstrap: start herdd instead of systemd

the test will repeated boot the operating system and use the serial log to determine
if the boot was successful. if a problem is detected, a debugger is automatically
started and attached.

CMakeLists.txt

@@ -1,6 +1,9 @@
-cmake_minimum_required(VERSION 3.14)
+cmake_minimum_required(VERSION 4.0)
 project(Rosetta C CXX ASM)
 
+include(CheckPIESupported)
+check_pie_supported()
+
 set(sys_dir ${CMAKE_CURRENT_BINARY_DIR}/sys)
 
 set(kernel_name mango_kernel)
@@ -25,6 +28,7 @@ add_subdirectory(interface)
 add_subdirectory(sys)
 add_subdirectory(lib)
 add_subdirectory(services)
+add_subdirectory(runlevel)
 add_subdirectory(programs)
 
 sysroot_add_program(NAME ${kernel_name} BIN_DIR /boot)

arch/x86_64/Arch.cmake

@@ -1,3 +1,4 @@
 include(System-Disk)
 include(QEMU)
 include(Bochs)
+include(Test)

arch/x86_64/Platform/Rosetta.cmake

@@ -5,17 +5,23 @@ set(CMAKE_TRY_COMPILE_TARGET_TYPE "STATIC_LIBRARY")
 
 find_program(C_COMPILER x86_64-elf-gcc REQUIRED)
 find_program(CXX_COMPILER x86_64-elf-g++ REQUIRED)
-find_program(ASM_COMPILER x86_64-elf-as REQUIRED)
+#find_program(ASM_COMPILER x86_64-elf-as REQUIRED)
 
 add_compile_definitions(__mango__=1)
 
 set(CMAKE_C_COMPILER ${C_COMPILER})
 set(CMAKE_CXX_COMPILER ${CXX_COMPILER})
-set(CMAKE_ASM_COMPILER ${ASM_COMPILER})
+#set(CMAKE_ASM_COMPILER ${ASM_COMPILER})
 
 SET(CMAKE_C_FLAGS "-ffreestanding -nostdlib -z max-page-size=0x1000 -m64 -mcmodel=large -mno-red-zone -mno-mmx -mno-sse -mno-sse2 -D_64BIT -DBYTE_ORDER=1234" CACHE STRING "" FORCE)
 set(CMAKE_SHARED_LINKER_FLAGS "-Wl,-shared" CACHE STRING "" FORCE)
 set(CMAKE_EXE_LINKER_FLAGS "-Wl,--unresolved-symbols=report-all,--dynamic-linker=/lib/ld64.so" CACHE STRING "" FORCE)
+set(CMAKE_C_LINK_OPTIONS_PIE "-pie")
+set(CMAKE_C_LINK_PIE_SUPPORTED TRUE)
+set(CMAKE_C_LINK_NO_PIE_SUPPORTED TRUE)
+SET(CMAKE_ASM_FLAGS "${CFLAGS} -x assembler-with-cpp")
 
 set(CMAKE_C_OUTPUT_EXTENSION .o)
 set(CMAKE_CXX_OUTPUT_EXTENSION .o)
+
+set(CMAKE_POSITION_INDEPENDENT_CODE TRUE)

arch/x86_64/QEMU.cmake

@@ -8,7 +8,8 @@ find_program(LLDB lldb)
 find_program(GDB gdb)
 
 set(patched_kernel ${CMAKE_CURRENT_BINARY_DIR}/kernel/${kernel_name}.elf32)
-set(generic_flags -m 1G)
+set(generic_flags -m 128M -cpu qemu64,+rdrand)
+
 set(no_debug_flags)
 
 if (${CMAKE_HOST_SYSTEM_NAME} STREQUAL "Linux")

arch/x86_64/Test.cmake

@@ -0,0 +1,35 @@
+find_package(Python COMPONENTS Interpreter)
+if (NOT Python_EXECUTABLE)
+	message(STATUS "QEMU: Cannot find python. Direct-kernel boot testing unavailable")
+	return()
+endif ()
+
+find_program(QEMU qemu-system-${TARGET_ARCH} REQUIRED)
+if (NOT QEMU)
+	message(STATUS "QEMU: Cannot find qemu-system-${TARGET_ARCH}. Direct-kernel boot testing unavailable")
+	return()
+endif ()
+
+set(patched_kernel ${CMAKE_CURRENT_BINARY_DIR}/kernel/${kernel_name}.elf32)
+set(generic_flags -m 1G -cpu qemu64,+rdrand)
+set(this_dir ${CMAKE_SOURCE_DIR}/arch/${CMAKE_SYSTEM_PROCESSOR})
+set(no_debug_flags)
+
+if (${CMAKE_HOST_SYSTEM_NAME} STREQUAL "Linux")
+	message(STATUS "QEMU: Enabling KVM acceleration")
+	set(no_debug_flags ${no_debug_flags} -enable-kvm)
+else ()
+	message(STATUS "QEMU: Host system is not Linux. KVM acceleration unavailable")
+endif ()
+
+message(STATUS "Test: Enable direct-kernel boot testing with QEMU")
+add_custom_target(test-successful-boot
+	COMMAND
+		${this_dir}/test/successful-boot
+		${Python_EXECUTABLE}
+		${this_dir}/test/check-results
+		${QEMU}
+		${patched_kernel}
+		${sys_dir}/${bsp_name}
+	USES_TERMINAL
+	DEPENDS ${patched_kernel} bsp)

arch/x86_64/test/check-results

@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+# vim: ft=python
+import sys
+
+def log(f, msg):
+    print(msg)
+    f.write(msg)
+    f.write('\n')
+
+def successful_boot(boot_log, out):
+    nr_panic = boot_log.count("---[ kernel panic")
+    if nr_panic == 1:
+        log(out, "Kernel panic!")
+        return 1
+    if nr_panic > 1:
+        log(out, "Multiple kernel panics!")
+        return 1
+
+    nr_boots = boot_log.count('Mango kernel version')
+    if nr_boots == 0:
+        log(out, "Kernel didn't start!")
+        return 1
+    if nr_boots > 1:
+        log(out, "Kernel rebooted during test!")
+        return 1
+
+    nr_finish = boot_log.count("ld finished")
+    if nr_finish == 0:
+        log(out, "Didn't reach end of boot sequence!")
+        return 1
+    if nr_finish > 1:
+        log(out, "Boot sequence performed multiple times!")
+        return 1
+
+    return 0
+
+
+tests = {
+        'successful-boot': successful_boot,
+}
+
+test_name = sys.argv[1]
+boot_log_path = sys.argv[2]
+out_path = sys.argv[3]
+
+boot_log_file = open(boot_log_path, 'r')
+boot_log = boot_log_file.read()
+boot_log_file.close()
+
+out_file = open(out_path, 'a')
+
+exit(tests[test_name](boot_log, out_file))

arch/x86_64/test/successful-boot

@@ -0,0 +1,69 @@
+#!/bin/bash
+# vim: ft=bash
+
+
+log_dir="test-results/successful-boot"
+rm -rf $log_dir
+mkdir -p $log_dir
+
+logfile="$log_dir/main.log"
+
+log() {
+	if [ -n "$logfile" ]; then
+		printf '%s\n' "$@" >> "$logfile"
+	fi
+
+	printf '%s\n' "$@"
+}
+
+log "Running boot test. Press Ctrl+\\ to stop."
+
+declare -i result
+declare -i count
+declare -i pass
+declare -i fail
+count=0
+pass=0
+fail=0
+python=$1
+validation_script=$2
+qemu=$3
+kernel=$4
+initrd=$5
+
+while true; do
+	log "Test $count"
+	result_file="$log_dir/$count.log"
+	$qemu \
+		-kernel $kernel \
+		-initrd $initrd \
+		-serial file:$result_file \
+		-cpu qemu64,+rdrand \
+		--append kernel.early-console=ttyS0 -s > /dev/null &
+	qemu_id=$!
+
+	sleep 1.2
+
+	$python $validation_script successful-boot $result_file $logfile
+	result=$?
+
+	count=$count+1
+
+	if [ $result -eq 0 ]; then
+		pass=$pass+1
+	else
+		mv $result_file "$result_file.FAIL"
+		fail=$fail+1
+		lldb \
+			-o "file kernel/mango_kernel.debug" \
+			-o "gdb-remote localhost:1234"
+	fi
+
+	kill -INT $qemu_id
+
+	log "---------------"
+	log "Total tests: $count"
+	log "Pass: $pass"
+	log "Fail: $fail"
+	log "---------------"
+done

cmake/BSP.cmake

@@ -9,6 +9,36 @@ function(bsp_reset)
 		COMMAND_ERROR_IS_FATAL ANY)
 endfunction(bsp_reset)
 
+function(bsp_add_file)
+	set(options)
+	set(one_value_args ID SRC_PATH DEST_DIR)
+	set(multi_value_args)
+
+	cmake_parse_arguments(PARSE_ARGV 0 arg
+		"${options}"
+		"${one_value_args}"
+		"${multi_value_args}")
+
+	set(target_name ${arg_ID})
+	set(bsp_target_name _bsp-${target_name})
+
+	get_property(bsp_targets GLOBAL PROPERTY bsp_target_list)
+	list(LENGTH bsp_targets nr_bsp_targets)
+	if (${nr_bsp_targets} GREATER 0)
+		math(EXPR serialiser_index "${nr_bsp_targets}-1")
+		list(GET bsp_targets ${serialiser_index} serialiser)
+	endif ()
+
+	add_custom_target(${bsp_target_name}
+		COMMAND	${Python_EXECUTABLE} ${bsp_tool}
+			add-binary ${bsp_manifest} ${target_name}
+			${arg_DEST_DIR} ${arg_SRC_PATH}
+		COMMENT "Preparing bsp component: ${target_name}"
+		DEPENDS ${arg_SRC_PATH} ${serialiser})
+
+	set_property(GLOBAL PROPERTY bsp_target_list ${bsp_targets} ${bsp_target_name})
+endfunction(bsp_add_file)
+
 function(bsp_add_library)
 	set(options)
 	set(one_value_args NAME HEADER_DIR LIB_DIR)
@@ -70,6 +100,40 @@ function(bsp_add_program)
 	set_property(GLOBAL PROPERTY bsp_target_list ${bsp_targets} ${bsp_target_name})
 endfunction(bsp_add_program)
 
+function(bsp_add_service)
+	set(options)
+	set(one_value_args NAME BIN_DIR SVC_DIR)
+	set(multi_value_args)
+
+	cmake_parse_arguments(PARSE_ARGV 0 arg
+		"${options}"
+		"${one_value_args}"
+		"${multi_value_args}")
+
+	set(target_name ${arg_NAME})
+	set(bsp_target_name _bsp-${target_name})
+
+	get_property(bsp_targets GLOBAL PROPERTY bsp_target_list)
+	get_property(cfg_file TARGET ${arg_NAME} PROPERTY service_cfg_path)
+	list(LENGTH bsp_targets nr_bsp_targets)
+	if (${nr_bsp_targets} GREATER 0)
+		math(EXPR serialiser_index "${nr_bsp_targets}-1")
+		list(GET bsp_targets ${serialiser_index} serialiser)
+	endif ()
+
+	add_custom_target(${bsp_target_name}
+		COMMAND	${Python_EXECUTABLE} ${bsp_tool}
+			add-binary ${bsp_manifest} ${target_name}
+			${arg_BIN_DIR} $<TARGET_FILE:${target_name}>
+		COMMAND	${Python_EXECUTABLE} ${bsp_tool}
+			add-binary ${bsp_manifest} ${target_name}-cfg
+			${arg_SVC_DIR} ${cfg_file}
+		COMMENT "Preparing bsp component: ${target_name}"
+		DEPENDS ${target_name} ${serialiser})
+
+	set_property(GLOBAL PROPERTY bsp_target_list ${bsp_targets} ${bsp_target_name})
+endfunction(bsp_add_service)
+
 function(bsp_finalise)
 	set(options)
 	set(one_value_args BOOTSTRAP_PROGRAM DEST_DIR BSP_NAME)

cmake/Msg-Interface.cmake

@@ -21,6 +21,7 @@ function(add_interface)
 	add_custom_command(
 		OUTPUT ${header_path}
 		COMMAND	${XPCG} ${arg_PATH}
+		DEPENDS ${arg_PATH}
 		WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/${arg_PARENT_DIR}
 		COMMENT "Generating interface: ${arg_NAME}")
 

cmake/Sysroot.cmake

@@ -155,6 +155,40 @@ function(sysroot_add_program)
 	set_property(GLOBAL PROPERTY sysroot_target_list ${sysroot_targets} ${sysroot_target_name})
 endfunction(sysroot_add_program)
 
+function(sysroot_add_service)
+	set(options)
+	set(one_value_args NAME BIN_DIR SVC_DIR)
+	set(multi_value_args)
+
+	cmake_parse_arguments(PARSE_ARGV 0 arg
+		"${options}"
+		"${one_value_args}"
+		"${multi_value_args}")
+
+	set(target_name ${arg_NAME})
+	set(sysroot_target_name _sysroot-${target_name})
+
+	get_property(sysroot_targets GLOBAL PROPERTY sysroot_target_list)
+	get_property(cfg_file TARGET ${arg_NAME} PROPERTY service_cfg_path)
+	list(LENGTH sysroot_targets nr_sysroot_targets)
+	if (${nr_sysroot_targets} GREATER 0)
+		math(EXPR serialiser_index "${nr_sysroot_targets}-1")
+		list(GET sysroot_targets ${serialiser_index} serialiser)
+	endif ()
+
+	add_custom_target(${sysroot_target_name}
+		COMMAND	${Python_EXECUTABLE} ${sysroot_tool}
+			add-binary ${sysroot_manifest} ${target_name}
+			${arg_BIN_DIR} $<TARGET_FILE:${target_name}>
+		COMMAND	${Python_EXECUTABLE} ${sysroot_tool}
+			add-binary ${sysroot_manifest} ${target_name}-cfg
+			${arg_SVC_DIR} ${cfg_file}
+		COMMENT "Preparing sysroot component: ${target_name}"
+		DEPENDS ${target_name} ${serialiser})
+
+	set_property(GLOBAL PROPERTY sysroot_target_list ${sysroot_targets} ${sysroot_target_name})
+endfunction(sysroot_add_service)
+
 function(sysroot_add_file)
 	set(options)
 	set(one_value_args ID SRC_PATH DEST_DIR)

cmake/Templates.cmake

@@ -32,6 +32,29 @@ function(rosetta_add_executable)
 		DESTINATION ${arg_SYSROOT_PATH})
 endfunction(rosetta_add_executable)
 
+function(rosetta_add_service)
+	set(options)
+	set(one_value_args NAME CFG_FILE)
+	set(multi_value_args SOURCES)
+	cmake_parse_arguments(PARSE_ARGV 0 arg
+		"${options}"
+		"${one_value_args}"
+		"${multi_value_args}")
+
+	set(exec_name ${arg_NAME})
+	get_property(programs GLOBAL PROPERTY rosetta_program_list)
+	set_property(GLOBAL PROPERTY rosetta_program_list ${programs} ${exec_name})
+
+	message(STATUS "Building service ${exec_name}")
+	add_executable(${exec_name} ${arg_SOURCES})
+
+	set_target_properties(${exec_name} PROPERTIES
+		POSITION_INDEPENDENT_CODE ON
+		service_cfg_path ${arg_CFG_FILE})
+	install(TARGETS ${exec_name}
+		DESTINATION ${arg_SYSROOT_PATH})
+endfunction(rosetta_add_service)
+
 function(rosetta_add_library)
 	set(options STATIC SHARED)
 	set(one_value_args NAME)

interface/fs.if

@@ -1,4 +0,0 @@
-interface fs {
-	msg open(string path, int flags) -> (int err);
-	msg read(size count) -> (int err, bytebuf data);
-}

interface/fs.xpc

@@ -6,4 +6,6 @@ func close[1]() -> (err: int);
 func read[2](count: size) -> (err: int, nr_read: size, data: buffer);
 func write[3](data: buffer) -> (err: int, nr_written: size);
 
-func map[4](prot: int, flags: int) -> (vmo: handle);
+func seek[4](offset: offset, origin: int) -> (err: int, new_pos: offset);
+
+func map[5](prot: int, flags: int) -> (err: int, vmo: handle);

lib/libc/CMakeLists.txt

@@ -26,5 +26,7 @@ bsp_add_library(
 	NAME libc
 	LIB_DIR /usr/lib)
 
-target_link_libraries(libc libmango libxpc-static interface::fs)
+target_link_libraries(libc PRIVATE libmango librosetta libxpc-static interface::fs)
 target_compile_definitions(libc PRIVATE ENABLE_GLOBAL_HEAP=1)
+
+add_subdirectory(pthread)

lib/libc/core/CMakeLists.txt

@@ -8,13 +8,17 @@ foreach (dir ${source_dirs})
 	set(headers ${headers} ${dir_headers})
 endforeach (dir)
 
-set(component_sources ${sources} PARENT_SCOPE)
+file(GLOB sys_sources
+	${CMAKE_CURRENT_SOURCE_DIR}/sys/${CMAKE_SYSTEM_PROCESSOR}/*.c
+	${CMAKE_CURRENT_SOURCE_DIR}/sys/${CMAKE_SYSTEM_PROCESSOR}/*.S)
+
+set(component_sources ${sources} ${sys_sources} PARENT_SCOPE)
 set(component_headers ${headers} PARENT_SCOPE)
 
 rosetta_add_library(STATIC
 	NAME libc-core
 	PUBLIC_INCLUDE_DIRS ${public_include_dirs}
-	SOURCES ${sources}
+	SOURCES ${sources} ${sys_sources}
 	HEADERS ${headers})
 
 sysroot_add_library(
@@ -22,4 +26,4 @@ sysroot_add_library(
 	HEADER_DIR /usr/include
 	LIB_DIR /usr/lib)
 
-target_link_libraries(libc-core libmango)
+target_link_libraries(libc-core PRIVATE libmango librosetta)

lib/libc/core/errno/errno.c

@@ -1,6 +1,15 @@
 #include <errno.h>
+#include <mango/log.h>
 #include <mango/status.h>
 
+static int __errno = 32;
+
+int __attribute__((weak)) * __errno_location(void)
+{
+	kern_log("using builtin errno");
+	return &__errno;
+}
+
 #if defined(BUILD_STATIC)
 int __set_errno(int err)
 {
@@ -12,6 +21,7 @@ int __set_errno(int err)
 int __set_errno(int err)
 {
 	/* TODO */
+	*__errno_location() = err;
 	return -1;
 }
 #endif

lib/libc/core/string/strcpy.c

@@ -0,0 +1,26 @@
+char *strcpy(char *output, const char *input)
+{
+	unsigned int i;
+	for (i = 0; input[i] != 0; i++) {
+		output[i] = input[i];
+	}
+
+	output[i] = '\0';
+	return output;
+}
+
+char *strncpy(char *output, const char *input, unsigned int count)
+{
+	unsigned int size = count;
+	unsigned int i;
+	for (i = 0; i < size; i++) {
+		output[i] = input[i];
+
+		if (input[i] == 0) {
+			break;
+		}
+	}
+
+	output[i] = '\0';
+	return output;
+}

lib/libc/core/sys/x86_64/init.c

@@ -0,0 +1,18 @@
+#include <mango/log.h>
+#include <rosetta/bootstrap.h>
+#include <stdio.h>
+
+extern int main(int, const char **, const char **);
+
+void *__attribute__((weak)) pthread_self(void)
+{
+	/* Nothing */
+	return NULL;
+}
+
+int __libc_init(const struct rosetta_bootstrap *bsinfo)
+{
+	(volatile void)pthread_self();
+	kern_logf("bsinfo = %p", bsinfo);
+	return 0;
+}

lib/libc/include/errno.h

@@ -1,6 +1,8 @@
 #ifndef ERRNO_H_
 #define ERRNO_H_
 
+#define errno           (*__errno_location())
+
 #define SUCCESS         0  /* Success */
 #define EPERM           1  /* Operation not permitted */
 #define ENOENT          2  /* No such file or directory */
@@ -142,4 +144,6 @@
 extern int __set_errno(int err);
 extern int __errno_from_kern_status(unsigned int err);
 
+extern int __attribute__((weak)) * __errno_location(void);
+
 #endif

lib/libc/include/fcntl.h

@@ -0,0 +1,53 @@
+#ifndef FCNTL_H_
+#define FCNTL_H_
+
+#define O_RDONLY   0x0000 /* open for reading only */
+#define O_WRONLY   0x0001 /* open for writing only */
+#define O_RDWR     0x0002 /* open for reading and writing */
+#define O_ACCMODE  0x0003 /* mask for above modes */
+
+#define O_NONBLOCK 0x00000004 /* no delay */
+#define O_APPEND   0x00000008 /* set append mode */
+
+#define O_SHLOCK   0x00000010 /* open with shared file lock */
+#define O_EXLOCK   0x00000020 /* open with exclusive file lock */
+#define O_ASYNC    0x00000040 /* signal pgrp when data ready */
+#define O_FSYNC    O_SYNC     /* source compatibility: do not use */
+#define O_NOFOLLOW 0x00000100 /* don't follow symlinks */
+#define O_CREAT    0x00000200 /* create if nonexistant */
+#define O_TRUNC    0x00000400 /* truncate to zero length */
+#define O_EXCL     0x00000800 /* error if already exists */
+#define O_RESOLVE_BENEATH                                                      \
+	0x00001000 /* only for open(2), same value as FMARK                    \
+	            */
+
+#define O_EVTONLY                                                              \
+	0x00008000 /* descriptor requested for event notifications only */
+
+#define O_NOCTTY       0x00020000 /* don't assign controlling terminal */
+
+#define O_DIRECTORY    0x00100000
+#define O_SYMLINK      0x00200000 /* allow open of a symlink */
+
+#define O_CLOEXEC      0x01000000 /* implicitly set FD_CLOEXEC */
+
+#define O_NOFOLLOW_ANY 0x20000000 /* no symlinks allowed in path */
+
+#define O_EXEC         0x40000000             /* open file for execute only */
+#define O_SEARCH       (O_EXEC | O_DIRECTORY) /* open directory for search only */
+
+#define AT_FDCWD       -2
+
+#define AT_EACCESS     0x0010 /* Use effective ids in access check */
+#define AT_SYMLINK_NOFOLLOW                                                    \
+	0x0020                   /* Act on the symlink itself not the target */
+#define AT_SYMLINK_FOLLOW 0x0040 /* Act on target of symlink */
+#define AT_REMOVEDIR      0x0080 /* Path refers to directory */
+#define AT_REALDEV                                                             \
+	0x0200 /* Return real device inodes resides on for fstatat(2) */
+#define AT_FDONLY                                                              \
+	0x0400 /* Use only the fd and Ignore the path for fstatat(2) */
+#define AT_SYMLINK_NOFOLLOW_ANY                                                \
+	0x0800 /* Path should not contain any symlinks */
+
+#endif

lib/libc/include/string.h

@@ -11,6 +11,9 @@ extern size_t strlen(const char *s);
 extern int strcmp(const char *s1, const char *s2);
 extern int strncmp(const char *s1, const char *s2, unsigned long n);
 
+extern int strcpy(const char *s1, const char *s2);
+extern int strncpy(const char *s1, const char *s2, unsigned long n);
+
 extern void *memset(void *str, int c, size_t n);
 extern void *memcpy(void *dst, const void *src, size_t len);
 

lib/libc/include/sys/mman.h

@@ -9,8 +9,10 @@
 #define PROT_READ           0x02u
 #define PROT_WRITE          0x04u
 
+#define MAP_FAILED          ((void *)-1)
+
 #define MAP_SHARED          0x01u
-#define MAP_SHARED_VALIDATE 0x03u
+#define MAP_SHARED_VALIDATE 0x02u
 #define MAP_PRIVATE         0x04u
 #define MAP_32BIT           0x08u
 #define MAP_ANON            MAP_ANONYMOUS
@@ -39,5 +41,6 @@ extern void *mmap(
 	int flags,
 	int fd,
 	off_t offset);
+extern int munmap(void *addr, size_t length);
 
 #endif

lib/libc/include/sys/types.h

@@ -0,0 +1,10 @@
+#ifndef SYS_TYPES_H_
+#define SYS_TYPES_H_
+
+#include <mango/types.h>
+
+#define SEEK_SET 0
+#define SEEK_CUR 1
+#define SEEK_END 2
+
+#endif

lib/libc/include/unistd.h

@@ -2,6 +2,7 @@
 #define UNISTD_H_
 
 #include <stddef.h>
+#include <sys/types.h>
 
 extern int open(const char *path, int flags);
 extern int close(int fd);
@@ -9,4 +10,6 @@ extern int close(int fd);
 extern int read(int fd, void *buf, size_t count);
 extern int write(int fd, const void *buf, size_t count);
 
+extern off_t lseek(int fd, off_t offset, int whence);
+
 #endif

lib/libc/io/unistd/close.c

@@ -0,0 +1,11 @@
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/msg.h>
+#include <rosetta/fs.h>
+#include <sys/remote.h>
+
+int close(int fd)
+{
+	kern_handle_close(fd);
+	return 0;
+}

lib/libc/io/unistd/lseek.c

@@ -0,0 +1,22 @@
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/msg.h>
+#include <rosetta/fs.h>
+#include <sys/remote.h>
+
+off_t lseek(int fd, off_t offset, int whence)
+{
+	int err;
+	off_t new_offset;
+
+	kern_status_t status = fs_seek(fd, offset, whence, &err, &new_offset);
+	if (status != KERN_OK) {
+		return __set_errno(__errno_from_kern_status(status));
+	}
+
+	if (err != SUCCESS) {
+		return __set_errno(err);
+	}
+
+	return new_offset;
+}

lib/libc/io/unistd/mmap.c

@@ -1,6 +1,170 @@
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/task.h>
+#include <mango/vm.h>
+#include <rosetta/fs.h>
+#include <stdbool.h>
 #include <sys/mman.h>
 
+static vm_prot_t vm_prot_from_mmap_prot(int prot)
+{
+	vm_prot_t vm_prot = VM_PROT_USER;
+
+	if (prot & PROT_READ) {
+		vm_prot |= VM_PROT_READ;
+	}
+
+	if (prot & PROT_WRITE) {
+		vm_prot |= VM_PROT_WRITE;
+	}
+
+	if (prot & PROT_EXEC) {
+		vm_prot |= VM_PROT_EXEC;
+	}
+
+	return vm_prot;
+}
+
+static int get_vmo_anon(
+	int fd,
+	int prot,
+	int flags,
+	size_t length,
+	kern_handle_t *out)
+{
+	vm_prot_t vm_prot = vm_prot_from_mmap_prot(prot);
+	kern_status_t status = vm_object_create(NULL, 0, length, vm_prot, out);
+	return __errno_from_kern_status(status);
+}
+
+static int get_vmo(
+	int fd,
+	int prot,
+	int flags,
+	size_t length,
+	kern_handle_t *out)
+{
+	if (fd == -1) {
+		return get_vmo_anon(fd, prot, flags, length, out);
+	}
+
+	int err = 0;
+	kern_status_t status = fs_map(fd, prot, flags, &err, out);
+	if (status != KERN_OK) {
+		return __errno_from_kern_status(status);
+	}
+
+	return err;
+}
+
+static void *mreserve(kern_handle_t address_space, void *addr, size_t length)
+{
+	virt_addr_t base = (virt_addr_t)addr;
+	if (!base) {
+		base = MAP_ADDRESS_ANY;
+	}
+
+	kern_status_t status
+		= address_space_reserve(address_space, base, length, &base);
+
+	if (status != KERN_OK) {
+		__set_errno(__errno_from_kern_status(status));
+		return MAP_FAILED;
+	}
+
+	return (void *)base;
+}
+
 void *mmap(void *addr, size_t length, int prot, int flags, int fd, off_t offset)
 {
-	return NULL;
+	int tmp = 0;
+
+	if (flags & MAP_SHARED) {
+		tmp++;
+	}
+
+	if (flags & MAP_SHARED_VALIDATE) {
+		tmp++;
+	}
+
+	if (flags & MAP_PRIVATE) {
+		tmp++;
+	}
+
+	if (tmp != 1) {
+		__set_errno(EINVAL);
+		return MAP_FAILED;
+	}
+
+	if ((flags & MAP_ANONYMOUS) && fd != -1) {
+		__set_errno(EINVAL);
+		return MAP_FAILED;
+	}
+
+	if (!(flags & MAP_ANONYMOUS) && fd == -1) {
+		__set_errno(EINVAL);
+		return MAP_FAILED;
+	}
+
+	kern_status_t status = KERN_OK;
+	kern_handle_t self = KERN_HANDLE_INVALID,
+		      address_space = KERN_HANDLE_INVALID;
+
+	status = task_self(&self);
+	if (status != KERN_OK) {
+		__set_errno(__errno_from_kern_status(status));
+		return MAP_FAILED;
+	}
+
+	status = task_get_address_space(self, &address_space);
+	kern_handle_close(self);
+	if (status != KERN_OK) {
+		__set_errno(__errno_from_kern_status(status));
+		return MAP_FAILED;
+	}
+
+	if ((flags & (MAP_ANONYMOUS | MAP_PRIVATE)) && (prot == PROT_NONE)) {
+		void *ret = mreserve(address_space, addr, length);
+		kern_handle_close(address_space);
+		return ret;
+	}
+
+	kern_handle_t vmo = KERN_HANDLE_INVALID;
+	int err = get_vmo(fd, prot, flags, length, &vmo);
+	if (err != SUCCESS) {
+		kern_handle_close(address_space);
+		__set_errno(err);
+		return MAP_FAILED;
+	}
+
+	virt_addr_t map_address = 0;
+	vm_prot_t vm_prot = vm_prot_from_mmap_prot(prot);
+	if (addr && (flags & MAP_FIXED)) {
+		status = address_space_map(
+			address_space,
+			(virt_addr_t)addr,
+			vmo,
+			offset,
+			length,
+			vm_prot,
+			&map_address);
+	} else {
+		status = address_space_map(
+			address_space,
+			MAP_ADDRESS_ANY,
+			vmo,
+			offset,
+			length,
+			vm_prot,
+			&map_address);
+	}
+
+	kern_handle_close(vmo);
+	kern_handle_close(address_space);
+	if (status != KERN_OK) {
+		__set_errno(__errno_from_kern_status(status));
+		return MAP_FAILED;
+	}
+
+	return (void *)map_address;
 }

lib/libc/io/unistd/munmap.c

@@ -0,0 +1,44 @@
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/task.h>
+#include <mango/vm.h>
+#include <rosetta/fs.h>
+#include <stdbool.h>
+#include <sys/mman.h>
+
+int munmap(void *addr, size_t length)
+{
+	kern_status_t status = KERN_OK;
+	kern_handle_t self = KERN_HANDLE_INVALID,
+		      address_space = KERN_HANDLE_INVALID;
+
+	status = task_self(&self);
+	if (status != KERN_OK) {
+		return __set_errno(EPERM);
+	}
+
+	status = task_get_address_space(self, &address_space);
+	kern_handle_close(self);
+	if (status != KERN_OK) {
+		return __set_errno(EPERM);
+	}
+
+	status = address_space_unmap(address_space, (virt_addr_t)addr, length);
+
+	if (status != KERN_OK) {
+		kern_handle_close(address_space);
+		return __set_errno(__errno_from_kern_status(status));
+	}
+
+	status = address_space_release(
+		address_space,
+		(virt_addr_t)addr,
+		length);
+	kern_handle_close(address_space);
+
+	if (status != KERN_OK) {
+		return __set_errno(__errno_from_kern_status(status));
+	}
+
+	return __set_errno(SUCCESS);
+}

lib/libc/malloc/CMakeLists.txt

@@ -30,4 +30,5 @@ sysroot_add_library(
 	HEADER_DIR /usr/include
 	LIB_DIR /usr/lib)
 
+target_compile_definitions(libc-malloc PRIVATE ENABLE_GLOBAL_HEAP=1)
 target_link_libraries(libc-malloc libc-core libmango)

lib/libc/malloc/heap.c

@@ -38,14 +38,10 @@ static kern_status_t init_heap_region(heap_t *heap)
 	task_get_address_space(self, &address_space);
 	kern_handle_close(self);
 
-	kern_status_t status = vm_region_create(
+	kern_status_t status = address_space_reserve(
 		address_space,
-		"libc-heap",
+		MAP_ADDRESS_ANY,
-		9,
-		VM_REGION_ANY_OFFSET,
 		HEAP_REGION_SIZE,
-		VM_PROT_READ | VM_PROT_WRITE | VM_PROT_USER,
-		&heap->heap_region,
 		&heap->heap_base);
 
 	kern_handle_close(address_space);
@@ -72,9 +68,14 @@ static kern_status_t expand_heap(heap_t *heap)
 
 	virt_addr_t base = 0;
 
-	status = vm_region_map_relative(
+	kern_handle_t self, address_space;
-		heap->heap_region,
+	task_self(&self);
-		heap->heap_sys_alloc,
+	task_get_address_space(self, &address_space);
+	kern_handle_close(self);
+
+	status = address_space_map(
+		address_space,
+		heap->heap_base + heap->heap_sys_alloc,
 		vmo,
 		0,
 		HEAP_EXPAND_INCREMENT,
@@ -90,7 +91,7 @@ static kern_status_t expand_heap(heap_t *heap)
 void *heap_expand(heap_t *heap, size_t size)
 {
 	kern_status_t status = KERN_OK;
-	if (heap->heap_region == KERN_HANDLE_INVALID) {
+	if (!heap->heap_base) {
 		status = init_heap_region(heap);
 	}
 

lib/libc/malloc/include/heap/_liballoc.h

@@ -6,7 +6,6 @@ struct liballoc_minor;
 
 #define HEAP_INIT                                                              \
 	{                                                                      \
-		.heap_region = KERN_HANDLE_INVALID,                            \
 		.heap_liballoc = {                                             \
 			.l_pageSize = 4096,                                    \
 			.l_pageCount = 16,                                     \

lib/libc/malloc/include/heap/heap.h

@@ -12,7 +12,6 @@ typedef enum heap_result {
 } heap_result_t;
 
 typedef struct heap {
-	kern_handle_t heap_region;
 	/* amount of space requested from the heap by the user */
 	size_t heap_req_alloc;
 	/* amount of space requested from the system by the heap */

lib/libc/pthread/CMakeLists.txt

@@ -0,0 +1,46 @@
+set(pthread_source_dirs thread)
+
+foreach (dir ${pthread_source_dirs})
+	file(GLOB dir_sources ${CMAKE_CURRENT_SOURCE_DIR}/${dir}/*.c)
+	file(GLOB dir_headers ${CMAKE_CURRENT_SOURCE_DIR}/${dir}/*.h)
+
+	set(pthread_sources ${pthread_sources} ${dir_sources})
+	set(pthread_headers ${pthread_headers} ${dir_headers})
+endforeach (dir)
+
+file(GLOB pthread_sys_sources
+	${CMAKE_CURRENT_SOURCE_DIR}/sys/${CMAKE_SYSTEM_PROCESSOR}/*.c
+	${CMAKE_CURRENT_SOURCE_DIR}/sys/${CMAKE_SYSTEM_PROCESSOR}/*.S)
+
+set(pthread_sources ${pthread_sources} ${pthread_sys_sources})
+
+set(pthread_headers ${pthread_headers} ${CMAKE_CURRENT_SOURCE_DIR}/include/pthread.h)
+set(pthread_public_include_dir ${CMAKE_CURRENT_SOURCE_DIR}/include)
+
+rosetta_add_library(STATIC
+	NAME libc-pthread
+	PUBLIC_INCLUDE_DIRS ${pthread_public_include_dir}
+	SOURCES ${pthread_sources}
+	HEADERS ${pthread_headers})
+rosetta_add_library(SHARED
+	NAME libpthread
+	PUBLIC_INCLUDE_DIRS ${pthread_public_include_dir}
+	SOURCES ${pthread_sources}
+	HEADERS ${pthread_headers})
+
+sysroot_add_library(
+	NAME libc-pthread
+	HEADER_DIR /usr/include
+	LIB_DIR /usr/lib)
+sysroot_add_library(
+	NAME libpthread
+	HEADER_DIR /usr/include
+	LIB_DIR /usr/lib)
+
+bsp_add_library(
+	NAME libpthread
+	LIB_DIR /usr/lib)
+
+target_link_libraries(libc-pthread PRIVATE libc-io libmango)
+target_link_libraries(libpthread PRIVATE libmango)
+target_link_libraries(libpthread PUBLIC libc)

lib/libc/pthread/include/pthread.h

@@ -0,0 +1,25 @@
+#ifndef PTHREAD_H_
+#define PTHREAD_H_
+
+#define __PTHREAD_ATTR_SIZE__ 32
+
+typedef struct __pthread *pthread_t;
+
+typedef struct __pthread_attr {
+	long __sig;
+	char __opaque[__PTHREAD_ATTR_SIZE__];
+} pthread_attr_t;
+
+extern int __pthread_init(void);
+
+extern int pthread_create(
+	pthread_t *thread,
+	const pthread_attr_t *attr,
+	void *(*start_func)(void *),
+	void *arg);
+extern pthread_t pthread_self(void);
+extern int pthread_join(pthread_t thread, void **retval);
+extern int pthread_detach(pthread_t thread);
+extern void pthread_exit(void *retval);
+
+#endif

lib/libc/pthread/sys/x86_64/pthread_self.S

@@ -0,0 +1,13 @@
+.code64
+
+.global __pthread_self
+.type __pthread_self, @function
+
+__pthread_self:
+	push %rbp
+	mov %rsp, %rbp
+
+	mov %gs:0, %rax
+
+	pop %rbp
+	ret

lib/libc/pthread/sys/x86_64/pthread_unmap_exit.S

@@ -0,0 +1,35 @@
+.code64
+
+#include "mango/syscall.h"
+
+.global __pthread_unmap_exit
+.type __pthread_unmap_exit, @function
+
+/*
+  %rdi = (kern_handle_t)address_space
+  %rsi = (void *)unmap_base
+  %rdx = (size_t)unmap_length
+ */
+__pthread_unmap_exit:
+	push %rbp
+	mov %rsp, %rbp
+
+	/* save the address space handle for later */
+	mov %rdi, %rbx
+
+	/* first, unmap the specified region */
+	mov $SYS_ADDRESS_SPACE_UNMAP, %rax
+	/* args are already in the correct registers */
+	syscall
+
+	/* next, close the handle to the address space */
+	mov $SYS_THREAD_EXIT, %rax
+	mov %rbx, %rdi
+	syscall
+
+	/* finally, stop the current thread */
+	mov $SYS_THREAD_EXIT, %rax
+	syscall
+
+	/* unreachable */
+	ret

lib/libc/pthread/thread/errno.c

@@ -0,0 +1,12 @@
+#include "pthread.h"
+
+#include <mango/log.h>
+#include <pthread.h>
+#include <stdio.h>
+
+int *__errno_location(void)
+{
+	struct __pthread *self = pthread_self();
+	kern_logf("using pthread errno %p", &self->thr_errno);
+	return &self->thr_errno;
+}

lib/libc/pthread/thread/pthread.h

@@ -0,0 +1,26 @@
+#ifndef _THREAD_PTHREAD_H_
+#define _THREAD_PTHREAD_H_
+
+#include <mango/types.h>
+
+enum pthread_flags {
+	THREAD_DETACHED = 0x01u,
+};
+
+struct __pthread {
+	struct __pthread *thr_self;
+	int thr_errno;
+	enum pthread_flags thr_flags;
+	kern_handle_t thr_handle;
+	void *thr_map_base;
+	size_t thr_map_size;
+	void *thr_result;
+};
+
+extern struct __pthread *__pthread_self(void);
+extern void __pthread_unmap_exit(
+	kern_handle_t address_space,
+	void *unmap_base,
+	size_t unmap_length);
+
+#endif

lib/libc/pthread/thread/pthread_create.c

@@ -0,0 +1,89 @@
+#include "pthread.h"
+
+#include <errno.h>
+#include <mango/status.h>
+#include <mango/task.h>
+#include <pthread.h>
+#include <string.h>
+#include <sys/mman.h>
+
+#define DEFAULT_STACK_SIZE 0x40000
+
+static void thread_launchpad(uintptr_t thread, uintptr_t func, uintptr_t arg)
+{
+	struct __pthread *self = (struct __pthread *)thread;
+	void *(*start_func)(void *) = (void *(*)(void *))func;
+
+	void *result = start_func((void *)arg);
+
+	pthread_exit(result);
+}
+
+int pthread_create(
+	pthread_t *thread,
+	const pthread_attr_t *attr,
+	void *(*start_func)(void *),
+	void *arg)
+{
+	kern_status_t status = KERN_OK;
+	kern_handle_t task;
+
+	status = task_self(&task);
+	if (status != KERN_OK) {
+		return __set_errno(EPERM);
+	}
+
+	size_t stack_size = DEFAULT_STACK_SIZE;
+
+	void *base
+		= mmap(NULL,
+	               stack_size,
+	               PROT_READ | PROT_WRITE,
+	               MAP_PRIVATE | MAP_ANONYMOUS,
+	               -1,
+	               0);
+
+	if (!base) {
+		return __set_errno(ENOMEM);
+	}
+
+	virt_addr_t sp = (virt_addr_t)base + DEFAULT_STACK_SIZE;
+
+	sp -= sizeof(struct __pthread);
+	struct __pthread *new = (struct __pthread *)sp;
+	sp &= ~(virt_addr_t)0x10;
+
+	memset(new, 0x0, sizeof *new);
+	new->thr_self = new;
+	new->thr_map_base = base;
+	new->thr_map_size = stack_size;
+
+	uintptr_t args[] = {
+		(uintptr_t)new,
+		(uintptr_t)start_func,
+		(uintptr_t)arg,
+	};
+	size_t nr_args = sizeof args / sizeof args[0];
+
+	status = task_create_thread(
+		task,
+		(virt_addr_t)thread_launchpad,
+		sp,
+		args,
+		nr_args,
+		&new->thr_handle);
+	if (status != KERN_OK) {
+		munmap(base, stack_size);
+		return __set_errno(__errno_from_kern_status(status));
+	}
+
+	thread_config_set(
+		new->thr_handle,
+		THREAD_CFG_GSBASE,
+		&new,
+		sizeof(void *));
+	thread_start(new->thr_handle);
+
+	*thread = new;
+	return __set_errno(SUCCESS);
+}

lib/libc/pthread/thread/pthread_detach.c

@@ -0,0 +1,17 @@
+#include "pthread.h"
+
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/object.h>
+#include <mango/signal.h>
+#include <mango/task.h>
+#include <pthread.h>
+#include <sys/mman.h>
+
+int pthread_detach(struct __pthread *thread)
+{
+	thread->thr_flags |= THREAD_DETACHED;
+	kern_handle_close(thread->thr_handle);
+
+	return 0;
+}

lib/libc/pthread/thread/pthread_exit.c

@@ -0,0 +1,31 @@
+#include "pthread.h"
+
+#include <mango/handle.h>
+#include <mango/task.h>
+#include <pthread.h>
+#include <stdlib.h>
+
+extern void pthread_exit(void *retval)
+{
+	struct __pthread *self = pthread_self();
+	if (!self) {
+		/* TODO: abort(); */
+		return;
+	}
+
+	if (self->thr_flags & THREAD_DETACHED) {
+		kern_handle_t task;
+		kern_handle_t address_space;
+		task_self(&task);
+		task_get_address_space(task, &address_space);
+		kern_handle_close(task);
+
+		__pthread_unmap_exit(
+			address_space,
+			self->thr_map_base,
+			self->thr_map_size);
+	} else {
+		self->thr_result = retval;
+		thread_exit();
+	}
+}

lib/libc/pthread/thread/pthread_join.c

@@ -0,0 +1,31 @@
+#include "pthread.h"
+
+#include <errno.h>
+#include <mango/handle.h>
+#include <mango/object.h>
+#include <mango/signal.h>
+#include <mango/task.h>
+#include <pthread.h>
+#include <sys/mman.h>
+
+int pthread_join(struct __pthread *thread, void **retval)
+{
+	kern_wait_item_t wait;
+	wait.w_handle = thread->thr_handle;
+	wait.w_waitfor = THREAD_SIGNAL_STOPPED;
+	wait.w_observed = 0;
+
+	kern_status_t status = kern_object_wait(&wait, 1);
+	if (status != KERN_OK) {
+		return __set_errno(EINVAL);
+	}
+
+	if (retval) {
+		*retval = thread->thr_result;
+	}
+
+	kern_handle_close(thread->thr_handle);
+	munmap(thread->thr_map_base, thread->thr_map_size);
+
+	return 0;
+}

lib/libc/pthread/thread/pthread_self.c

@@ -0,0 +1,36 @@
+#include "pthread.h"
+
+#include <mango/task.h>
+#include <mango/types.h>
+#include <string.h>
+#include <sys/mman.h>
+
+static struct __pthread main_thread = {0};
+
+struct __pthread *pthread_self(void)
+{
+	static int init = 0;
+	if (!init) {
+		kern_handle_t self_handle = KERN_HANDLE_INVALID;
+		kern_status_t status = thread_self(&self_handle);
+		if (status != KERN_OK) {
+			/* TODO set an errno value in a way that doesn't result
+			 * in a recursive call to pthread_self */
+			return NULL;
+		}
+		struct __pthread *self = &main_thread;
+		self->thr_self = self;
+		self->thr_handle = self_handle;
+		self->thr_map_base = self;
+		self->thr_map_size = sizeof *self;
+		thread_config_set(
+			self_handle,
+			THREAD_CFG_GSBASE,
+			&self,
+			sizeof(void *));
+		init = 1;
+		return self;
+	}
+
+	return __pthread_self();
+}

lib/libc/runtime/CMakeLists.txt

@@ -1,6 +1,5 @@
 file(GLOB runtime_sources
 	${CMAKE_CURRENT_SOURCE_DIR}/${CMAKE_SYSTEM_PROCESSOR}/*.s)
-set_property(SOURCE ${runtime_sources} PROPERTY LANGUAGE C)
 
 rosetta_add_object_library(
 	NAME libc-runtime STATIC

lib/libc/runtime/x86_64/crt0.s

@@ -3,14 +3,22 @@
 .global _start
 .type _start, @function
 
-.extern main
+.extern __libc_init
-.type main, @function
+.type __libc_init, @function
+
+.extern task_exit
+.type task_exit, @function
 
 _start:
-	# Args (as provided by the ABI)
+	# %rdi: (struct rosetta_bootstrap *)bs_info
-	# %rdi: int argc
+	mov %rdi, %rbx
-	# %rsi: const char **argv
+	call __libc_init
-	# %rdx: kern_handle_t task
+
-	# %rcx: kern_handle_t address_space
+	mov 0(%rbx), %rdi # argc
+	mov 8(%rbx), %rsi # argv
+	mov 24(%rbx), %rdx # envp
 	call main
+
+	mov %rax, %rdi
+	call task_exit
 1:	jmp 1b

lib/libfs/context.c

@@ -1,6 +1,7 @@
 #include "btree.h"
 #include "file.h"
 #include "interface.h"
+#include "mapping.h"
 
 #include <fs/allocator.h>
 #include <fs/context.h>
@@ -8,9 +9,16 @@
 #include <fs/inode.h>
 #include <fs/status.h>
 #include <fs/superblock.h>
+#include <mango/handle.h>
 #include <mango/log.h>
+#include <mango/object.h>
+#include <mango/signal.h>
+#include <mango/task.h>
+#include <mango/vm.h>
 #include <stdio.h>
 
+#define TEMP_OBJECT_SIZE 0x10000
+
 BTREE_DEFINE_SIMPLE_GET(struct fs_file, unsigned long, f_node, f_id, get_file);
 BTREE_DEFINE_SIMPLE_INSERT(struct fs_file, f_node, f_id, put_file);
 
@@ -18,12 +26,21 @@ struct fs_context {
 	struct fs_superblock *ctx_sb;
 	struct fs_allocator *ctx_alloc;
 	struct btree ctx_filelist;
+	kern_handle_t ctx_vm_controller;
+	kern_handle_t ctx_channel;
+	kern_handle_t ctx_temp_object;
+	void *ctx_temp_object_buf;
 
 	struct fs_vtable ctx_vtable;
 };
 
 struct fs_context *fs_context_create(struct fs_allocator *alloc)
 {
+	kern_handle_t self, address_space;
+	task_self(&self);
+	task_get_address_space(self, &address_space);
+	kern_handle_close(self);
+
 	struct fs_context *ctx = fs_alloc(alloc, sizeof *ctx);
 	if (!ctx) {
 		return NULL;
@@ -31,10 +48,50 @@ struct fs_context *fs_context_create(struct fs_allocator *alloc)
 
 	memset(ctx, 0x0, sizeof *ctx);
 
+	kern_status_t status = vm_controller_create(&ctx->ctx_vm_controller);
+	if (status != KERN_OK) {
+		fs_free(alloc, ctx);
+		return NULL;
+	}
+
+	status = vm_object_create(
+		NULL,
+		0,
+		TEMP_OBJECT_SIZE,
+		VM_PROT_READ | VM_PROT_WRITE | VM_PROT_USER,
+		&ctx->ctx_temp_object);
+	if (status != KERN_OK) {
+		kern_handle_close(ctx->ctx_vm_controller);
+		fs_free(alloc, ctx);
+		return NULL;
+	}
+
+	virt_addr_t temp_buffer;
+	status = address_space_map(
+		address_space,
+		MAP_ADDRESS_ANY,
+		ctx->ctx_temp_object,
+		0,
+		TEMP_OBJECT_SIZE,
+		VM_PROT_READ | VM_PROT_WRITE | VM_PROT_USER,
+		&temp_buffer);
+	kern_handle_close(address_space);
+	if (status != KERN_OK) {
+		kern_handle_close(ctx->ctx_temp_object);
+		kern_handle_close(ctx->ctx_vm_controller);
+		fs_free(alloc, ctx);
+		return NULL;
+	}
+
+	ctx->ctx_temp_object_buf = (void *)temp_buffer;
 	ctx->ctx_alloc = alloc;
 
 	ctx->ctx_vtable.open = fs_msg_open;
+	ctx->ctx_vtable.close = fs_msg_close;
 	ctx->ctx_vtable.read = fs_msg_read;
+	ctx->ctx_vtable.write = fs_msg_write;
+	ctx->ctx_vtable.seek = fs_msg_seek;
+	ctx->ctx_vtable.map = fs_msg_map;
 
 	return ctx;
 }
@@ -73,6 +130,112 @@ enum fs_status fs_context_unmount_filesystem(struct fs_context *ctx)
 	return FS_ERR_NOT_IMPLEMENTED;
 }
 
+void fs_context_set_channel(struct fs_context *ctx, kern_handle_t channel)
+{
+	ctx->ctx_channel = channel;
+}
+
+kern_handle_t fs_context_get_vm_controller(const struct fs_context *ctx)
+{
+	return ctx->ctx_vm_controller;
+}
+
+static enum fs_status handle_msg(struct fs_context *ctx)
+{
+	xpc_msg_t msg;
+	kern_status_t status = xpc_msg_recv_nowait(ctx->ctx_channel, &msg);
+	if (status == KERN_NO_ENTRY) {
+		return FS_SUCCESS;
+	}
+
+	if (status != KERN_OK) {
+		kern_tracef("message recv error %d", status);
+		return FS_ERR_INTERNAL_FAILURE;
+	}
+
+	switch (msg.msg_header.hdr_interface) {
+	case INTERFACE_FS:
+		status = fs_context_dispatch_msg(ctx, &msg);
+		break;
+	default:
+		kern_tracef(
+			"unknown message protocol %u",
+			msg.msg_header.hdr_interface);
+		xpc_msg_reply_error(&msg, KERN_UNSUPPORTED);
+		break;
+	}
+
+	return FS_SUCCESS;
+}
+
+static enum fs_status handle_page_request(struct fs_context *ctx)
+{
+	equeue_packet_page_request_t packet;
+	vm_controller_recv(ctx->ctx_vm_controller, &packet);
+	struct file_mapping *mapping = (struct file_mapping *)packet.req_vmo;
+	kern_tracef(
+		"received page request [%zx-%zx] for file %s",
+		packet.req_offset,
+		packet.req_offset + packet.req_length,
+		mapping->m_file->f_dent->d_name);
+
+	size_t length = packet.req_length;
+	if (length > TEMP_OBJECT_SIZE) {
+		length = TEMP_OBJECT_SIZE;
+	}
+
+	char *dst = ctx->ctx_temp_object_buf;
+	xpc_buffer_t buf = XPC_LOCAL_BUFFER_OUT(dst, TEMP_OBJECT_SIZE);
+	enum fs_status status = fs_file_read_at(
+		mapping->m_file,
+		&buf,
+		packet.req_offset,
+		length);
+	if (status != FS_SUCCESS) {
+		kern_tracef("map-read failed with code %d", status);
+		return status;
+	}
+
+	vm_controller_supply_pages(
+		ctx->ctx_vm_controller,
+		mapping->m_vmo,
+		packet.req_offset,
+		ctx->ctx_temp_object,
+		0,
+		packet.req_length);
+	return FS_SUCCESS;
+}
+
+enum fs_status fs_context_handle_request(struct fs_context *ctx)
+{
+	kern_wait_item_t waiters[] = {
+		{
+			.w_handle = ctx->ctx_channel,
+			.w_waitfor = CHANNEL_SIGNAL_MSG_RECEIVED,
+		},
+		{
+			.w_handle = ctx->ctx_vm_controller,
+			.w_waitfor = VM_CONTROLLER_SIGNAL_REQUEST_RECEIVED,
+		},
+	};
+	const size_t nr_waiters = sizeof waiters / sizeof waiters[0];
+
+	kern_status_t kstatus = kern_object_wait(waiters, nr_waiters);
+	if (kstatus != KERN_OK) {
+		return FS_ERR_INTERNAL_FAILURE;
+	}
+
+	if (waiters[0].w_observed & CHANNEL_SIGNAL_MSG_RECEIVED) {
+		return handle_msg(ctx);
+	}
+
+	if (waiters[1].w_observed & VM_CONTROLLER_SIGNAL_REQUEST_RECEIVED) {
+		return handle_page_request(ctx);
+	}
+
+	return FS_SUCCESS;
+}
+
 struct fs_file *fs_context_open_file(struct fs_context *ctx, unsigned long id)
 {
 	struct fs_file *f = get_file(&ctx->ctx_filelist, id);

lib/libfs/file.c

@@ -25,9 +25,22 @@ enum fs_status fs_file_read(
 	return status;
 }
 
-enum fs_status fs_file_write(
+enum fs_status fs_file_read_at(
 	struct fs_file *f,
 	struct xpc_buffer *buf,
+	off_t offset,
+	size_t count)
+{
+	if (!f->f_ops || !f->f_ops->f_read) {
+		return FS_ERR_NOT_IMPLEMENTED;
+	}
+
+	return f->f_ops->f_read(f, buf, count, &offset);
+}
+
+enum fs_status fs_file_write(
+	struct fs_file *f,
+	const struct xpc_buffer *buf,
 	size_t count)
 {
 	if (!f->f_ops || !f->f_ops->f_write) {

lib/libfs/file.h

@@ -2,6 +2,7 @@
 #define _FS_FILE_H_
 
 #include "btree.h"
+#include "queue.h"
 
 #include <fs/dentry.h>
 #include <fs/file.h>
@@ -18,6 +19,8 @@ struct fs_file {
 	off_t f_seek;
 	struct fs_inode *f_inode;
 	struct fs_dentry *f_dent;
+
+	struct queue f_mappings;
 };
 
 #endif

lib/libfs/include/fs/context.h

@@ -30,6 +30,12 @@ extern enum fs_status fs_context_mount_filesystem(
 	enum fs_mount_flags flags);
 extern enum fs_status fs_context_unmount_filesystem(struct fs_context *ctx);
 
+extern void fs_context_set_channel(
+	struct fs_context *ctx,
+	kern_handle_t channel);
+extern kern_handle_t fs_context_get_vm_controller(const struct fs_context *ctx);
+extern enum fs_status fs_context_handle_request(struct fs_context *ctx);
+
 extern struct fs_file *fs_context_open_file(
 	struct fs_context *ctx,
 	unsigned long id);

lib/libfs/include/fs/file.h

@@ -27,9 +27,14 @@ extern enum fs_status fs_file_read(
 	struct fs_file *f,
 	struct xpc_buffer *buf,
 	size_t count);
-extern enum fs_status fs_file_write(
+extern enum fs_status fs_file_read_at(
 	struct fs_file *f,
 	struct xpc_buffer *buf,
+	off_t offset,
+	size_t count);
+extern enum fs_status fs_file_write(
+	struct fs_file *f,
+	const struct xpc_buffer *buf,
 	size_t count);
 
 #endif

lib/libfs/include/fs/inode.h

@@ -2,12 +2,14 @@
 #define FS_INODE_H_
 
 #include <fs/status.h>
+#include <mango/types.h>
 #include <stddef.h>
 
 struct fs_inode;
 struct fs_dentry;
 struct fs_superblock;
 struct fs_file_ops;
+struct file_mapping;
 
 enum fs_inode_mode {
 	FS_INODE_REG = 0x01u,
@@ -27,6 +29,7 @@ struct fs_inode {
 	const struct fs_inode_ops *i_ops;
 	const struct fs_file_ops *i_fops;
 	size_t i_size;
+	struct file_mapping *i_shared_mapping;
 };
 
 extern enum fs_status fs_inode_lookup(

lib/libfs/interface.h

@@ -17,9 +17,8 @@ extern kern_status_t fs_msg_open(
 	int *out_err,
 	void *arg);
 extern kern_status_t fs_msg_close(
-	const struct msg_endpoint *sender,
+	xpc_context_t *ctx,
-	const char *path,
+	const xpc_endpoint_t *sender,
-	int flags,
 	int *out_err,
 	void *arg);
 
@@ -31,5 +30,29 @@ extern kern_status_t fs_msg_read(
 	size_t *out_nr_read,
 	xpc_buffer_t *out_data,
 	void *arg);
+extern kern_status_t fs_msg_write(
+	xpc_context_t *ctx,
+	const xpc_endpoint_t *sender,
+	const xpc_buffer_t *data,
+	int *out_err,
+	size_t *out_nr_written,
+	void *arg);
+
+extern kern_status_t fs_msg_seek(
+	xpc_context_t *ctx,
+	const xpc_endpoint_t *sender,
+	off_t offset,
+	int origin,
+	int *out_err,
+	off_t *out_new_pos,
+	void *arg);
+extern kern_status_t fs_msg_map(
+	xpc_context_t *ctx,
+	const xpc_endpoint_t *sender,
+	int prot,
+	int flags,
+	int *out_err,
+	kern_handle_t *out_vmo,
+	void *arg);
 
 #endif

lib/libfs/interface/close.c

@@ -0,0 +1,14 @@
+#include "../interface.h"
+
+#include <errno.h>
+#include <mango/status.h>
+
+extern kern_status_t fs_msg_close(
+	xpc_context_t *ctx,
+	const xpc_endpoint_t *sender,
+	int *out_err,
+	void *arg)
+{
+	*out_err = ENOSYS;
+	return KERN_OK;
+}

lib/libfs/interface/map.c

@@ -0,0 +1,112 @@
+#include "../file.h"
+#include "../mapping.h"
+
+#include <errno.h>
+#include <fs/context.h>
+#include <fs/file.h>
+#include <fs/status.h>
+#include <mango/handle.h>
+#include <mango/log.h>
+#include <mango/vm.h>
+#include <stdio.h>
+#include <sys/mman.h>
+
+static int create_file_mapping(
+	struct fs_context *ctx,
+	int prot,
+	int flags,
+	struct fs_file *f,
+	struct file_mapping **out)
+{
+	if ((flags & MAP_SHARED) && f->f_inode->i_shared_mapping) {
+		*out = f->f_inode->i_shared_mapping;
+		return SUCCESS;
+	}
+
+	struct file_mapping *mapping = fs_context_alloc(ctx, sizeof *mapping);
+	if (!mapping) {
+		return ENOMEM;
+	}
+
+	memset(mapping, 0x0, sizeof *mapping);
+
+	vm_prot_t vm_prot = VM_PROT_USER;
+	if (prot & PROT_READ) {
+		vm_prot |= VM_PROT_READ;
+	}
+
+	if (prot & PROT_WRITE) {
+		vm_prot |= VM_PROT_WRITE;
+	}
+
+	if (prot & PROT_EXEC) {
+		vm_prot |= VM_PROT_EXEC;
+	}
+
+	kern_handle_t vmo = KERN_HANDLE_INVALID;
+	kern_status_t status = vm_controller_create_object(
+		fs_context_get_vm_controller(ctx),
+		f->f_dent->d_name,
+		strlen(f->f_dent->d_name),
+		(equeue_key_t)mapping,
+		f->f_inode->i_size,
+		vm_prot,
+		&vmo);
+
+	if (status != KERN_OK) {
+		fs_context_free(ctx, mapping);
+		return __errno_from_kern_status(status);
+	}
+
+	mapping->m_file = f;
+	mapping->m_vmo = vmo;
+
+	if (flags & MAP_SHARED) {
+		mapping->m_type = FILE_MAPPING_SHARED;
+		f->f_inode->i_shared_mapping = mapping;
+	} else {
+		mapping->m_type = FILE_MAPPING_PRIVATE;
+		queue_push_back(&f->f_mappings, &mapping->m_entry);
+	}
+
+	*out = mapping;
+
+	return SUCCESS;
+}
+
+extern kern_status_t fs_msg_map(
+	xpc_context_t *xpc,
+	const xpc_endpoint_t *sender,
+	int prot,
+	int flags,
+	int *out_err,
+	kern_handle_t *out_vmo,
+	void *arg)
+{
+	struct fs_context *ctx = arg;
+	struct fs_file *f = fs_context_get_file(ctx, sender->e_port);
+	if (!f) {
+		*out_err = EBADF;
+		return KERN_OK;
+	}
+
+	struct file_mapping *mapping = NULL;
+	int err = create_file_mapping(ctx, prot, flags, f, &mapping);
+	if (err != SUCCESS) {
+		*out_err = err;
+		return KERN_OK;
+	}
+
+	kern_tracef(
+		"mapping file %s (%s) using vmo %zx",
+		f->f_dent->d_name,
+		(flags & MAP_SHARED) ? "shared" : "private",
+		mapping->m_vmo);
+
+	kern_handle_t vmo;
+	kern_handle_duplicate(mapping->m_vmo, &vmo);
+
+	*out_err = SUCCESS;
+	*out_vmo = vmo;
+	return KERN_OK;
+}

lib/libfs/interface/open.c

@@ -4,6 +4,8 @@
 #include <fs/context.h>
 #include <fs/file.h>
 #include <fs/status.h>
+#include <mango/log.h>
+#include <stdio.h>
 
 extern kern_status_t fs_msg_open(
 	xpc_context_t *xpc,
@@ -39,7 +41,7 @@ extern kern_status_t fs_msg_open(
 		= fs_context_resolve_path(ctx, path_buf, &dent);
 	if (fs_status != FS_SUCCESS) {
 		fs_context_close_file(ctx, f);
-		*out_err = fs_status_to_errno(status);
+		*out_err = fs_status_to_errno(fs_status);
 		return KERN_OK;
 	}
 

lib/libfs/interface/seek.c

@@ -0,0 +1,52 @@
+#include "../file.h"
+
+#include <errno.h>
+#include <fs/context.h>
+#include <fs/file.h>
+#include <fs/status.h>
+#include <sys/types.h>
+
+extern kern_status_t fs_msg_seek(
+	xpc_context_t *xpc,
+	const xpc_endpoint_t *sender,
+	off_t rel_offset,
+	int origin,
+	int *out_err,
+	off_t *out_new_pos,
+	void *arg)
+{
+	struct fs_context *ctx = arg;
+	struct fs_file *f = fs_context_get_file(ctx, sender->e_port);
+	if (!f) {
+		*out_err = EBADF;
+		return KERN_OK;
+	}
+
+	off_t new_offset = 0;
+
+	switch (origin) {
+	case SEEK_SET:
+		new_offset = rel_offset;
+		break;
+	case SEEK_CUR:
+		new_offset = f->f_seek + rel_offset;
+		break;
+	case SEEK_END:
+		new_offset = f->f_inode->i_size + rel_offset;
+		break;
+	default:
+		*out_err = EINVAL;
+		return KERN_OK;
+	}
+
+	if (new_offset > f->f_inode->i_size) {
+		*out_err = EINVAL;
+		return KERN_OK;
+	}
+
+	f->f_seek = new_offset;
+	*out_err = SUCCESS;
+	*out_new_pos = new_offset;
+
+	return KERN_OK;
+}

lib/libfs/interface/write.c

@@ -0,0 +1,29 @@
+#include <errno.h>
+#include <fs/context.h>
+#include <fs/file.h>
+#include <fs/status.h>
+
+extern kern_status_t fs_msg_write(
+	xpc_context_t *xpc,
+	const xpc_endpoint_t *sender,
+	const xpc_buffer_t *data,
+	int *out_err,
+	size_t *out_nr_written,
+	void *arg)
+{
+	struct fs_context *ctx = arg;
+	struct fs_file *f = fs_context_get_file(ctx, sender->e_port);
+	if (!f) {
+		*out_err = EBADF;
+		return KERN_OK;
+	}
+
+	size_t start = fs_file_get_cursor(f);
+	enum fs_status status = fs_file_write(f, data, data->buf_len);
+	size_t end = fs_file_get_cursor(f);
+
+	*out_err = fs_status_to_errno(status);
+	*out_nr_written = end - start;
+
+	return KERN_OK;
+}

lib/libfs/mapping.h

@@ -0,0 +1,22 @@
+#ifndef MAPPING_H_
+#define MAPPING_H_
+
+#include "queue.h"
+
+#include <mango/types.h>
+
+struct fs_file;
+
+enum file_mapping_type {
+	FILE_MAPPING_PRIVATE,
+	FILE_MAPPING_SHARED,
+};
+
+struct file_mapping {
+	enum file_mapping_type m_type;
+	struct fs_file *m_file;
+	kern_handle_t m_vmo;
+	struct queue_entry m_entry;
+};
+
+#endif

lib/libfs/queue.c

@@ -0,0 +1,138 @@
+#include "queue.h"
+
+size_t queue_length(struct queue *q)
+{
+	size_t i = 0;
+	struct queue_entry *x = q->q_first;
+	while (x) {
+		i++;
+		x = x->qe_next;
+	}
+
+	return i;
+}
+
+void queue_insert_before(
+	struct queue *q,
+	struct queue_entry *entry,
+	struct queue_entry *before)
+{
+	struct queue_entry *x = before->qe_prev;
+	if (x) {
+		x->qe_next = entry;
+	} else {
+		q->q_first = entry;
+	}
+
+	entry->qe_prev = x;
+
+	before->qe_prev = entry;
+	entry->qe_next = before;
+}
+
+void queue_insert_after(
+	struct queue *q,
+	struct queue_entry *entry,
+	struct queue_entry *after)
+{
+	struct queue_entry *x = after->qe_next;
+	if (x) {
+		x->qe_prev = entry;
+	} else {
+		q->q_last = entry;
+	}
+
+	entry->qe_prev = x;
+
+	after->qe_next = entry;
+	entry->qe_prev = after;
+}
+
+void queue_push_front(struct queue *q, struct queue_entry *entry)
+{
+	if (q->q_first) {
+		q->q_first->qe_prev = entry;
+	}
+
+	entry->qe_next = q->q_first;
+	entry->qe_prev = NULL;
+
+	q->q_first = entry;
+
+	if (!q->q_last) {
+		q->q_last = entry;
+	}
+}
+
+void queue_push_back(struct queue *q, struct queue_entry *entry)
+{
+	if (q->q_last) {
+		q->q_last->qe_next = entry;
+	}
+
+	entry->qe_prev = q->q_last;
+	entry->qe_next = NULL;
+
+	q->q_last = entry;
+
+	if (!q->q_first) {
+		q->q_first = entry;
+	}
+}
+
+struct queue_entry *queue_pop_front(struct queue *q)
+{
+	struct queue_entry *x = q->q_first;
+	if (x) {
+		queue_delete(q, x);
+	}
+
+	return x;
+}
+
+struct queue_entry *queue_pop_back(struct queue *q)
+{
+	struct queue_entry *x = q->q_last;
+	if (x) {
+		queue_delete(q, x);
+	}
+
+	return x;
+}
+
+void queue_delete(struct queue *q, struct queue_entry *entry)
+{
+	if (!entry) {
+		return;
+	}
+
+	if (entry == q->q_first) {
+		q->q_first = q->q_first->qe_next;
+	}
+
+	if (entry == q->q_last) {
+		q->q_last = q->q_last->qe_prev;
+	}
+
+	if (entry->qe_next) {
+		entry->qe_next->qe_prev = entry->qe_prev;
+	}
+
+	if (entry->qe_prev) {
+		entry->qe_prev->qe_next = entry->qe_next;
+	}
+
+	entry->qe_next = entry->qe_prev = NULL;
+}
+
+void queue_delete_all(struct queue *q)
+{
+	struct queue_entry *x = q->q_first;
+	while (x) {
+		struct queue_entry *next = x->qe_next;
+		x->qe_next = x->qe_prev = NULL;
+		x = next;
+	}
+
+	q->q_first = q->q_last = NULL;
+}

lib/libfs/queue.h

@@ -0,0 +1,100 @@
+#ifndef QUEUE_H_
+#define QUEUE_H_
+
+#include <stdbool.h>
+#include <string.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define QUEUE_CONTAINER(t, m, v)                                               \
+	((void *)((v) ? (uintptr_t)(v) - (offsetof(t, m)) : 0))
+
+#define QUEUE_INIT ((struct queue) {.q_first = NULL, .q_last = NULL})
+#define QUEUE_ENTRY_INIT                                                       \
+	((struct queue_entry) {.qe_next = NULL, .qe_prev = NULL})
+
+#define queue_foreach(iter_type, iter_name, queue_name, node_member)           \
+	for (iter_type *iter_name = (iter_type *)QUEUE_CONTAINER(              \
+		     iter_type,                                                \
+		     node_member,                                              \
+		     queue_first(queue_name));                                 \
+	     iter_name;                                                        \
+	     iter_name = (iter_type *)QUEUE_CONTAINER(                         \
+		     iter_type,                                                \
+		     node_member,                                              \
+		     queue_next(&((iter_name)->node_member))))
+
+#define queue_foreach_r(iter_type, iter_name, queue_name, node_member)         \
+	for (iter_type *iter_name = (iter_type *)QUEUE_CONTAINER(              \
+		     iter_type,                                                \
+		     node_member,                                              \
+		     queue_last(queue_name));                                  \
+	     iter_name;                                                        \
+	     iter_name = (iter_type *)QUEUE_CONTAINER(                         \
+		     iter_type,                                                \
+		     node_member,                                              \
+		     queue_prev(&((iter_name)->node_member))))
+
+struct queue_entry {
+	struct queue_entry *qe_next;
+	struct queue_entry *qe_prev;
+};
+
+struct queue {
+	struct queue_entry *q_first;
+	struct queue_entry *q_last;
+};
+
+static inline void queue_init(struct queue *q)
+{
+	memset(q, 0x00, sizeof *q);
+}
+static inline bool queue_empty(struct queue *q)
+{
+	return q->q_first == NULL;
+}
+
+static inline struct queue_entry *queue_first(struct queue *q)
+{
+	return q->q_first;
+}
+static inline struct queue_entry *queue_last(struct queue *q)
+{
+	return q->q_last;
+}
+static inline struct queue_entry *queue_next(struct queue_entry *entry)
+{
+	return entry->qe_next;
+}
+static inline struct queue_entry *queue_prev(struct queue_entry *entry)
+{
+	return entry->qe_prev;
+}
+
+extern size_t queue_length(struct queue *q);
+
+extern void queue_insert_before(
+	struct queue *q,
+	struct queue_entry *entry,
+	struct queue_entry *before);
+extern void queue_insert_after(
+	struct queue *q,
+	struct queue_entry *entry,
+	struct queue_entry *after);
+
+extern void queue_push_front(struct queue *q, struct queue_entry *entry);
+extern void queue_push_back(struct queue *q, struct queue_entry *entry);
+
+extern struct queue_entry *queue_pop_front(struct queue *q);
+extern struct queue_entry *queue_pop_back(struct queue *q);
+
+extern void queue_delete(struct queue *q, struct queue_entry *entry);
+extern void queue_delete_all(struct queue *q);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

lib/liblaunch/CMakeLists.txt

@@ -18,4 +18,4 @@ sysroot_add_library(
 	HEADER_DIR /usr/include
 	LIB_DIR /usr/lib)
 
-target_link_libraries(liblaunch librosetta libmango libc-core)
+target_link_libraries(liblaunch PRIVATE librosetta libmango libc-core)

lib/liblaunch/elf.c

@@ -160,15 +160,10 @@ static kern_status_t create_exec_regions(struct elf_image *image)
 		image->e_total_size);
 	kern_status_t status = KERN_OK;
 	if (image->e_local_space != KERN_HANDLE_INVALID) {
-		status = vm_region_create(
+		status = address_space_reserve(
 			image->e_local_space,
-			NULL,
+			MAP_ADDRESS_ANY,
-			0,
-			VM_REGION_ANY_OFFSET,
 			image->e_total_size,
-			VM_PROT_READ | VM_PROT_WRITE | VM_PROT_EXEC
-				| VM_PROT_USER,
-			&image->e_local_exec,
 			&image->e_local_base);
 	}
 
@@ -177,22 +172,18 @@ static kern_status_t create_exec_regions(struct elf_image *image)
 	}
 
 	if (image->e_remote_space != KERN_HANDLE_INVALID) {
-		status = vm_region_create(
+		status = address_space_reserve(
 			image->e_remote_space,
-			NULL,
+			MAP_ADDRESS_ANY,
-			0,
-			VM_REGION_ANY_OFFSET,
 			image->e_total_size,
-			VM_PROT_READ | VM_PROT_WRITE | VM_PROT_EXEC
-				| VM_PROT_USER,
-			&image->e_remote_exec,
 			&image->e_remote_base);
 	}
 
 	if (status != KERN_OK) {
-		vm_region_kill(image->e_local_exec);
+		address_space_release(
-		kern_handle_close(image->e_local_exec);
+			image->e_local_space,
-		image->e_local_exec = KERN_HANDLE_INVALID;
+			image->e_local_base,
+			image->e_total_size);
 
 		return status;
 	}
@@ -200,6 +191,19 @@ static kern_status_t create_exec_regions(struct elf_image *image)
 	return KERN_OK;
 }
 
+static kern_status_t release_exec_regions(struct elf_image *image)
+{
+	address_space_release(
+		image->e_local_space,
+		image->e_local_base,
+		image->e_total_size);
+	address_space_release(
+		image->e_remote_space,
+		image->e_remote_base,
+		image->e_total_size);
+	return KERN_OK;
+}
+
 static enum launch_status map_executable(struct elf_image *image)
 {
 	elf_phdr_t phdr;
@@ -256,10 +260,10 @@ static enum launch_status map_executable(struct elf_image *image)
 			return LAUNCH_ERR_IMAGE_DATA_LOAD_FAILED;
 		}
 
-		if (image->e_local_exec != KERN_HANDLE_INVALID) {
+		if (image->e_local_space != KERN_HANDLE_INVALID) {
-			status = vm_region_map_relative(
+			status = address_space_map(
-				image->e_local_exec,
+				image->e_local_space,
-				phdr.p_vaddr,
+				image->e_local_base + phdr.p_vaddr,
 				vmo,
 				offset,
 				phdr.p_memsz,
@@ -271,10 +275,10 @@ static enum launch_status map_executable(struct elf_image *image)
 			return LAUNCH_ERR_MEMORY_MAP_FAILED;
 		}
 
-		if (image->e_remote_exec != KERN_HANDLE_INVALID) {
+		if (image->e_remote_space != KERN_HANDLE_INVALID) {
-			status = vm_region_map_relative(
+			status = address_space_map(
-				image->e_remote_exec,
+				image->e_remote_space,
-				phdr.p_vaddr,
+				image->e_remote_base + phdr.p_vaddr,
 				vmo,
 				offset,
 				phdr.p_memsz,
@@ -330,6 +334,28 @@ static enum launch_status do_rela(struct elf_image *image, elf_rela_t *rela)
 			ELF64_R_SYM(rela->r_info),
 			rela->r_addend);
 		break;
+	case R_X86_64_GLOB_DAT:
+		sym = get_dynsym(image, ELF64_R_SYM(rela->r_info));
+		if (!sym) {
+			return LAUNCH_ERR_MISSING_SYMBOL;
+		}
+
+		*(uint64_t *)(image->e_local_base + rela->r_offset)
+			= image->e_remote_base + sym->st_value + rela->r_addend;
+		kern_tracef(
+			"GLOB_DAT: offset=%zx, symbol=%zu, addend=%zx",
+			rela->r_offset,
+			ELF64_R_SYM(rela->r_info),
+			rela->r_addend);
+		break;
+	case R_X86_64_RELATIVE:
+		*(uint64_t *)(image->e_local_base + rela->r_offset)
+			= image->e_remote_base + rela->r_addend;
+		kern_tracef(
+			"RELATIVE: offset=%zx, addend=%zx",
+			rela->r_offset,
+			rela->r_addend);
+		break;
 	default:
 		kern_trace("Unknown relocation type");
 		return LAUNCH_ERR_UNSUPPORTED_EXECUTABLE;
@@ -499,8 +525,6 @@ void elf_image_init(struct elf_image *out)
 	out->e_data = KERN_HANDLE_INVALID;
 	out->e_local_space = KERN_HANDLE_INVALID;
 	out->e_remote_space = KERN_HANDLE_INVALID;
-	out->e_local_exec = KERN_HANDLE_INVALID;
-	out->e_remote_exec = KERN_HANDLE_INVALID;
 }
 
 enum launch_status elf_image_load(
@@ -547,6 +571,11 @@ enum launch_status elf_image_load(
 		return status;
 	}
 
+	status = release_exec_regions(image);
+	if (status != KERN_OK) {
+		return status;
+	}
+
 	status = relocate(image);
 	if (status != LAUNCH_OK) {
 		return status;
@@ -557,9 +586,9 @@ enum launch_status elf_image_load(
 
 void elf_image_cleanup(struct elf_image *image)
 {
-	vm_region_unmap_relative(image->e_local_exec, 0, image->e_total_size);
+	address_space_unmap(
+		image->e_local_space,
+		image->e_local_base,
+		image->e_total_size);
 	kern_handle_close(image->e_data);
-	vm_region_kill(image->e_local_exec);
-	kern_handle_close(image->e_local_exec);
-	kern_handle_close(image->e_remote_exec);
 }

lib/liblaunch/elf.h

@@ -292,7 +292,6 @@ struct elf_image {
 	size_t e_page_size;
 	kern_handle_t e_image, e_data;
 	kern_handle_t e_local_space, e_remote_space;
-	kern_handle_t e_local_exec, e_remote_exec;
 	virt_addr_t e_local_base, e_remote_base;
 	elf_ehdr_t e_hdr;
 	elf_phdr_t e_dynamic;

lib/liblaunch/include/launch.h

@@ -45,7 +45,9 @@ struct launch_ctx {
 struct launch_parameters {
 	kern_handle_t p_parent_task;
 	kern_handle_t p_local_address_space;
-	kern_handle_t p_executable;
+
+	const char *p_exec_path;
+	kern_handle_t p_exec_image;
 
 	const char *p_task_name;
 

lib/liblaunch/launch.c

@@ -38,6 +38,7 @@ static kern_handle_t get_library(
 
 static virt_addr_t write_bootstrap_data(
 	struct stack_writer *stack,
+	const char *interpreter,
 	const struct launch_parameters *params)
 {
 	virt_addr_t bs_remote;
@@ -51,9 +52,15 @@ static virt_addr_t write_bootstrap_data(
 	bs->bs_handles_count = params->p_handle_count;
 	bs->bs_channels_count = params->p_channel_count;
 
+	if (interpreter) {
+		/* two extra args: interpreter path and path to executable */
+		bs->bs_argc += 2;
+	}
+
 	const char **argv, **envp;
 
 	if (bs->bs_argc > 0) {
+		int argc = bs->bs_argc;
 		virt_addr_t remote_argv;
 		argv = stack_writer_put(
 			stack,
@@ -63,27 +70,36 @@ static virt_addr_t write_bootstrap_data(
 		bs->bs_argv = (const char **)remote_argv;
 	}
 
-	if (bs->bs_envc > 0) {
+	/* envc+1 so there is space for a null terminator */
 	virt_addr_t remote_envp;
 	envp = stack_writer_put(
 		stack,
 		NULL,
-			bs->bs_envc * sizeof(char *),
+		(bs->bs_envc + 1) * sizeof(char *),
 		&remote_envp);
 	bs->bs_envp = (const char **)remote_envp;
+
+	size_t i = 0, j = 0;
+	if (interpreter) {
+		virt_addr_t arg_ptr;
+		stack_writer_put_string(stack, interpreter, &arg_ptr);
+		argv[i++] = (const char *)arg_ptr;
+		stack_writer_put_string(stack, params->p_exec_path, &arg_ptr);
+		argv[i++] = (const char *)arg_ptr;
 	}
 
-	for (size_t i = 0; i < params->p_argc; i++) {
+	for (; i < bs->bs_argc; i++, j++) {
 		virt_addr_t arg_ptr;
-		stack_writer_put_string(stack, params->p_argv[i], &arg_ptr);
+		stack_writer_put_string(stack, params->p_argv[j], &arg_ptr);
 		argv[i] = (const char *)arg_ptr;
 	}
 
-	for (size_t i = 0; i < params->p_envc; i++) {
 	virt_addr_t env_ptr;
+	for (i = 0; i < bs->bs_envc; i++) {
 		stack_writer_put_string(stack, params->p_envp[i], &env_ptr);
 		envp[i] = (const char *)env_ptr;
 	}
+	envp[i] = NULL;
 
 	return bs_remote;
 }
@@ -118,16 +134,20 @@ enum launch_status launch_ctx_execute(
 		return LAUNCH_ERR_TASK_CREATION_FAILED;
 	}
 
+	char interp_path[4096];
+	interp_path[0] = 0;
+
 	struct elf_image image;
 	elf_image_init(&image);
 
 	enum launch_status status = elf_image_load(
 		&image,
-		params->p_executable,
+		params->p_exec_image,
 		params->p_local_address_space,
 		remote_address_space);
 
 	if (status == LAUNCH_ERR_INTERPRETER_REQUIRED) {
+		snprintf(interp_path, sizeof interp_path, "%s", image.e_interp);
 		kern_handle_t interp = get_library(
 			ctx,
 			image.e_interp,
@@ -157,17 +177,17 @@ enum launch_status launch_ctx_execute(
 	}
 
 	virt_addr_t remote_stack_buf, local_stack_buf;
-	kstatus = vm_region_map_relative(
+	kstatus = address_space_map(
 		remote_address_space,
-		VM_REGION_ANY_OFFSET,
+		MAP_ADDRESS_ANY,
 		stack_vmo,
 		0,
 		STACK_SIZE,
 		VM_PROT_READ | VM_PROT_WRITE | VM_PROT_USER,
 		&remote_stack_buf);
-	kstatus = vm_region_map_relative(
+	kstatus = address_space_map(
 		params->p_local_address_space,
-		VM_REGION_ANY_OFFSET,
+		MAP_ADDRESS_ANY,
 		stack_vmo,
 		0,
 		STACK_SIZE,
@@ -187,7 +207,7 @@ enum launch_status launch_ctx_execute(
 		&stack,
 		local_stack_buf + STACK_SIZE,
 		remote_stack_buf + STACK_SIZE);
-	virt_addr_t bsdata = write_bootstrap_data(&stack, params);
+	virt_addr_t bsdata = write_bootstrap_data(&stack, interp_path, params);
 
 	virt_addr_t ip = image.e_hdr.e_entry + image.e_remote_base;
 
@@ -208,7 +228,10 @@ enum launch_status launch_ctx_execute(
 
 	thread_start(thread);
 
-	kern_handle_close(thread);
+	result->r_task = remote_task;
+	result->r_thread = thread;
+	result->r_address_space = remote_address_space;
+
 	elf_image_cleanup(&image);
 
 	return LAUNCH_OK;

lib/librosetta/CMakeLists.txt

@@ -12,4 +12,4 @@ sysroot_add_library(
 	HEADER_DIR /usr/include
 	LIB_DIR /usr/lib)
 
-target_link_libraries(librosetta libmango)
+target_link_libraries(librosetta PRIVATE libmango)

lib/librosetta/include/rosetta/bootstrap.h

@@ -26,17 +26,17 @@ struct rosetta_bootstrap_channel {
 };
 
 struct rosetta_bootstrap {
-	int bs_argc;
+	uintptr_t bs_argc;
 	const char **bs_argv;
 
-	int bs_envc;
+	uintptr_t bs_envc;
 	const char **bs_envp;
 
 	const struct rosetta_bootstrap_handle *bs_handles;
-	size_t bs_handles_count;
+	uintptr_t bs_handles_count;
 
 	const struct rosetta_bootstrap_channel *bs_channels;
-	size_t bs_channels_count;
+	uintptr_t bs_channels_count;
 };
 
 extern const struct rosetta_bootstrap_channel *rosetta_bootstrap_get_channel(

lib/libxpc/buffer.c

@@ -1,4 +1,5 @@
 #include <mango/status.h>
+#include <string.h>
 #include <xpc/buffer.h>
 #include <xpc/msg.h>
 
@@ -36,8 +37,7 @@ kern_status_t xpc_buffer_write(
 	size_t len,
 	size_t *nr_written)
 {
-	if ((buf->buf_flags & (XPC_BUFFER_F_OUT | XPC_BUFFER_F_REMOTE))
+	if (!(buf->buf_flags & XPC_BUFFER_F_OUT)) {
-	    != (XPC_BUFFER_F_OUT | XPC_BUFFER_F_REMOTE)) {
 		return KERN_BAD_STATE;
 	}
 
@@ -46,14 +46,16 @@ kern_status_t xpc_buffer_write(
 		to_write = buf->buf_max;
 	}
 
-	kern_status_t status
+	if (!(buf->buf_flags & XPC_BUFFER_F_REMOTE)) {
-		= xpc_msg_write(buf->buf_origin, buf->buf_offset, in, to_write);
+		memcpy(buf->buf_ptr, in, to_write);
-	if (status != KERN_OK) {
+		*nr_written = to_write;
-		return status;
+		return KERN_OK;
 	}
 
-	/* TODO */
+	return xpc_msg_write(
-	*nr_written = to_write;
+		buf->buf_origin,
-
+		buf->buf_offset,
-	return KERN_OK;
+		in,
+		to_write,
+		nr_written);
 }

lib/libxpc/include/xpc/buffer.h

@@ -16,7 +16,13 @@
 		.buf_flags = XPC_BUFFER_F_OUT | XPC_BUFFER_F_REMOTE,           \
 		.buf_origin = (msg),                                           \
 		.buf_offset = (offset),                                        \
-		.buf_len = (size),                                             \
+		.buf_max = (size),                                             \
+	}
+#define XPC_LOCAL_BUFFER_OUT(ptr, size)                                        \
+	{                                                                      \
+		.buf_flags = XPC_BUFFER_F_OUT,                                 \
+		.buf_ptr = (ptr),                                              \
+		.buf_max = (size),                                             \
 	}
 
 struct xpc_msg;
@@ -39,9 +45,6 @@ typedef enum xpc_buffer_flags {
 
 typedef struct xpc_buffer {
 	xpc_buffer_flags_t buf_flags;
-	union {
-		/* fields that are only valid if F_OUT is set */
-		struct {
 	/* only valid if F_OUT is set. specifies the maximum
 	 * number of chars that can be written to buf_buf,
 	 * including the null terminator. */
@@ -49,17 +52,11 @@ typedef struct xpc_buffer {
 	/* only valid if F_OUT is set.
 	 * if F_FREE_ON_DISCARD is set, must be either NULL or
 	 * allocated via xpc_context_alloc */
-			const char *buf_ptr;
+	void *buf_ptr;
-		};
+	/* valid for F_IN and F_OUT. offset of the buffer data
-
-		/* fields that are only valid if F_IN is set */
-		struct {
-			/* only valid if F_IN is set. offset of the buffer data
 	 * within the associated message. used when reading
 	 * buffer data from a message. */
 	size_t buf_offset;
-		};
-	};
 
 	/* only valid if F_REMOTE is set.
 	 * used to read/write buffer data from/to the sender's address

lib/libxpc/include/xpc/msg.h

@@ -29,6 +29,7 @@ extern void xpc_msg_header_init(
 	unsigned short func);
 extern bool xpc_msg_header_validate(const xpc_msg_header_t *msg);
 extern kern_status_t xpc_msg_recv(kern_handle_t channel, xpc_msg_t *out);
+extern kern_status_t xpc_msg_recv_nowait(kern_handle_t channel, xpc_msg_t *out);
 extern kern_status_t xpc_msg_read(
 	const xpc_msg_t *msg,
 	size_t offset,
@@ -38,7 +39,8 @@ extern kern_status_t xpc_msg_write(
 	const xpc_msg_t *msg,
 	size_t offset,
 	const void *p,
-	size_t count);
+	size_t count,
+	size_t *nr_written);
 
 extern kern_status_t xpc_msg_reply(
 	const xpc_msg_t *msg,

lib/libxpc/include/xpc/string.h

@@ -18,7 +18,7 @@
 		.s_flags = XPC_STRING_F_OUT | XPC_STRING_F_REMOTE,             \
 		.s_origin = (msg),                                             \
 		.s_offset = (offset),                                          \
-		.s_len = (size),                                               \
+		.s_max = (size),                                               \
 	}
 
 struct xpc_msg;
@@ -41,8 +41,6 @@ typedef enum xpc_string_flags {
 
 typedef struct xpc_string {
 	xpc_string_flags_t s_flags;
-	union {
-		struct {
 	/* only valid if F_OUT is set. specifies the maximum
 	 * number of chars that can be written to s_buf,
 	 * including the null terminator. */
@@ -51,15 +49,11 @@ typedef struct xpc_string {
 	 * if F_FREE_ON_DISCARD is set, must be either NULL or
 	 * allocated via xpc_context_alloc */
 	const char *s_buf;
-		};
 
-		struct {
+	/* valid for F_IN and F_OUT. offset of the string data
-			/* only valid if F_IN is set. offset of the string data
 	 * within the associated message. used when reading
 	 * string data from a message. */
 	size_t s_offset;
-		};
-	};
 
 	/* only valid if F_REMOTE is set.
 	 * used to read/write string data from/to the sender's address space. */

lib/libxpc/msg.c

@@ -1,4 +1,6 @@
 #include <mango/msg.h>
+#include <mango/object.h>
+#include <mango/signal.h>
 #include <string.h>
 #include <xpc/msg.h>
 
@@ -20,8 +22,25 @@ bool xpc_msg_header_validate(const xpc_msg_header_t *msg)
 	return msg->hdr_magic == XPC_MSG_MAGIC;
 }
 
-kern_status_t xpc_msg_recv(kern_handle_t channel, xpc_msg_t *out)
+static kern_status_t __msg_recv(
+	kern_handle_t channel,
+	xpc_msg_t *out,
+	bool nowait)
 {
+	kern_status_t status = KERN_OK;
+
+	if (!nowait) {
+		kern_wait_item_t wait = {
+			.w_handle = channel,
+			.w_waitfor = CHANNEL_SIGNAL_MSG_RECEIVED,
+		};
+
+		status = kern_object_wait(&wait, 1);
+		if (status != KERN_OK) {
+			return status;
+		}
+	}
+
 	kern_iovec_t iov = IOVEC(&out->msg_header, sizeof out->msg_header);
 	kern_msg_t msg = {
 		.msg_data = &iov,
@@ -29,7 +48,8 @@ kern_status_t xpc_msg_recv(kern_handle_t channel, xpc_msg_t *out)
 		.msg_handles = out->msg_handles,
 		.msg_handles_count = KERN_MSG_MAX_HANDLES,
 	};
-	kern_status_t status = msg_recv(channel, &msg);
+
+	status = msg_recv(channel, &msg);
 	if (status != KERN_OK) {
 		return status;
 	}
@@ -46,6 +66,25 @@ kern_status_t xpc_msg_recv(kern_handle_t channel, xpc_msg_t *out)
 	return KERN_OK;
 }
 
+kern_status_t xpc_msg_recv(kern_handle_t channel, xpc_msg_t *out)
+{
+	kern_status_t status = KERN_OK;
+	while (1) {
+		status = __msg_recv(channel, out, false);
+
+		if (status != KERN_NO_ENTRY) {
+			break;
+		}
+	}
+
+	return status;
+}
+
+kern_status_t xpc_msg_recv_nowait(kern_handle_t channel, xpc_msg_t *out)
+{
+	return __msg_recv(channel, out, true);
+}
+
 kern_status_t xpc_msg_read(
 	const xpc_msg_t *msg,
 	size_t offset,
@@ -67,17 +106,17 @@ kern_status_t xpc_msg_write(
 	const xpc_msg_t *msg,
 	size_t offset,
 	const void *p,
-	size_t count)
+	size_t count,
+	size_t *nr_written)
 {
 	kern_iovec_t iov = IOVEC(p, count);
-	size_t w = 0;
 	return msg_write(
 		msg->msg_sender.e_channel,
 		msg->msg_sender.e_msg,
 		offset,
 		&iov,
 		1,
-		&w);
+		nr_written);
 }
 
 kern_status_t xpc_msg_reply(

lib/libxpc/string.c

@@ -47,14 +47,10 @@ xpc_status_t xpc_string_write(
 		to_write = s->s_max - 1;
 	}
 
-	kern_status_t status
+	return xpc_msg_write(
-		= xpc_msg_write(s->s_origin, s->s_offset, in, to_write);
+		s->s_origin,
-	if (status != KERN_OK) {
+		s->s_offset,
-		return status;
+		in,
-	}
+		to_write,
-
+		nr_written);
-	/* TODO */
-	*nr_written = to_write;
-
-	return KERN_OK;
 }

programs/systemd/CMakeLists.txt

@@ -1,10 +0,0 @@
-file(GLOB sources *.c)
-add_executable(systemd ${sources})
-target_link_libraries(systemd libc libc-runtime liblaunch libmango)
-
-sysroot_add_program(
-	NAME systemd
-	BIN_DIR /usr/bin)
-bsp_add_program(
-	NAME systemd
-	BIN_DIR /usr/bin)

programs/test/CMakeLists.txt

@@ -1,9 +0,0 @@
-add_executable(test test.c)
-target_link_libraries(test libc libc-runtime liblaunch)
-
-sysroot_add_program(
-	NAME test
-	BIN_DIR /usr/bin)
-bsp_add_program(
-	NAME test
-	BIN_DIR /usr/bin)

programs/test/test.c

@@ -1,157 +0,0 @@
-#include <string.h>
-
-const char *s
-	= "RHZVNYYFPCGZIKWYLOZAKKLZGKWXDPXSTWNSWDBVDCPUMTAZCRNGNHGQPAGEWVAOCJST"
-	  "PBXVWPKXQAHGETDTDETVYAEQGNBLEQIPHGGWKEDJAULOVMZKNABLUUXOTMHJNAVLSGCZ"
-	  "QIFHIUCBCIRZIOFVHBPSGVBBZSRAQZYRPMYBNJDFTWSUUZVBQZTBUFUDJDYKUXWHDMXD"
-	  "JKBSNWLFPUDGRMQPBCWVJZHOTJBTMGUSXQUZVXLDKWIKINXRFJKSYJEKDAPRQPMQMJPS"
-	  "ICCTMELLKDFFJXOXISIXNVANFHLRPYMNPYCMUOYSPAYCKYJKRBZENHXYRJWRYXZGETMJ"
-	  "XBFTCMSIHZQSCMLESDFKGTTNMCZSXEFMGXZYPWVPODMYTLDUOKGPMTMFBTQQHPBHMNCM"
-	  "LYRGGUAIRSFFBRSKXOLJBWEMEODRQLXZJDRSTXBJOKOMUQKCJVKFHDYXCUUDHDEITHNH"
-	  "VQLJJQMLYWGIDVLYCEJTJFJQLTKSAPZGEEZKVLQPHIVJNJVTXJUGPZODIFXTQNLBSFLG"
-	  "NSPMGLUSEBOFJWXFFRBHIHYGGTILVVOJRPOFOIGDFCHLAZXLSOUCPCLZCBVWGZVKGDON"
-	  "RPYOTSRWUNAGQSPHSGEQHOLUSOZCQQGJBLPKQNGKOPCVKLACDBPAPXDMGKLFPUOFQDWY"
-	  "INIKZPLVFSZZZZHAYKXTETJDPMBHKNRCRLXZHFENZCABDTZULCRHCCVZFETBEBEJFVKJ"
-	  "ADWFSHKKSMMKGTIHPAIAMWXTRJILBMWMBDZGRPZXHMJVCWPKSNPYPPNSAVQDSMANBFZO"
-	  "LJBYKDOZNAPZRWEQDIIZRPNGZGHQWPIONPSAMBNNZERYMIQOVHRGZFXWVUARJMFWNPQP"
-	  "GHDCZABLOYHCBCXAIDSPMDKZVBBOKHAHGTEPRQAIBVWTFBQDGDJPQGMVAGQQVMULVPMG"
-	  "UPGEJUIZZXQRQKRJUCKDDZFTAHAAHMJLSISGFFOXOYNMJCAPPQXAVFSAFFRPRTEQLNCX"
-	  "JVKTHBPZLAEXSIGVJASAERWDGDQDXASXHRSSCNAUMRWSQDZTOJHJGJLLXMJSTXBHOYPH"
-	  "ELYKSXNJSPWPMFAKOOTXXTOEBLYFSOIJDWAOCTHDFOEBEEXVUBXOJDUCJWQOUUDOJNZG"
-	  "PULPYTAOQEXMHSGOOUXHAJOKRHOOMZYMBHQNTQDWZPCXATDEKTYLNBKNUEIINHQTEGTR"
-	  "ZKLMAKIIHHZBQIPXLAGADCFDYWEOJVHFPEMRDIOJYAMWSWEUOPJFEDGRGJOQNTSHRIJY"
-	  "JPTOSWYZXJCGXLYVOWKAFGULLNCKIUZDWUXTHNYSWMMCJGTFVVVPJHEKYQVFRWLIZBBK"
-	  "CNCRITFZYTQZZGZHXELBEYXSVVYRFBGRFPRDROUXKUMAGFYOJRMCLLHJVMQFYOBCXSEL"
-	  "OQAQTRMLSGDAXWMBRSQHCIYYMBQQHMUQOKIANZCBGKHLCPUVUEZVKDTTSOWKKWIUBAIW"
-	  "SOCJAUALJFEQQXJBHRRZBFMJZZMIWTFKQDPOBFIGABJTHFLLSZPWWGGLHLYXKBODKBIV"
-	  "GAYGIKHNMTMJHCPRBQYAACGSFZPJWXUTRZFCGTLFQBVHZBKYBRMYTTCGIDKYWVRPJDTX"
-	  "RKXGOPOQLNSEIGHTSAXGPBROFHQACSIVSLCXTEDUOEPRMGJDYWTKEHCXWINUDCAZWAEY"
-	  "RQDKZRZXKWGHVWJDYHGFLCGKCLYCVHZTWWXPDBKTMBKBASERMURDREKNYVOCPHFSEGBQ"
-	  "LAUHDDCAPPGOAFZYJYXPPAQLUQVKSDEHPPDXMNWAAYLHXBFEFXRLSMQNIYDECYVVPHEE"
-	  "WXFFPEYLSHBXLYJWRABFKJYMJNWAUMVYJKZRDMPUQJKWVNTPRMHTAQGNSDLFTXVNIAIJ"
-	  "HOIISROAJCWEOAIHYMDGWEOABPGIMBTTGWYVJDZOOMUSHYDPCMKWDIXDGJCGTQVXWBKP"
-	  "DBCVJJDYIAPXKINQXACEMTJRCIAENGTTMWXVSQCCXRDVNZZNQDZTTYJHQGXZIJIIKPAU"
-	  "TQJDDQQEPGYZNKCKNMVFCRHUECPVFZYVUWVMSCIQZBVLSTNFAHDDEQRKDOUAQVRVQAVB"
-	  "ZCEMAJRJBVWKVBNEIQWWQSJPUVUKMBJIISCWXGGMWANYYLPCXHCBARMFTFMDXWSMKXPW"
-	  "DZUGFBSGGWXLOFGYJVIDWNTSGODTHQNCKPWRJENDZNSCEYZRLEPNNYVBNUZMXAUWNAJD"
-	  "XGTOLUAAIAHMDJERSESVFSMMHJKHIVIBWZLEAXUKRSWJOTFODRBZIJBRJQTFBVRQHITD"
-	  "TDBAJZCZUKSIZJYDXWTSRPLXULXHGEKMWMICUYVAGNGEICEMMVWLLYWTAKWNGLRYOCIG"
-	  "BEYTLVAZFHBYIJPUAQHITKHPWAQNEBQVAYZEINLRCKUZILAQPAGBJDWWLGCPQZOZVDQP"
-	  "MWTIOAFMEMFKLGVGGHTTKNERTLPPQFALZMCWSOMJQQZMRABNKBCYPFJOWQCXJKXTNOMJ"
-	  "MXAWMPFBJHOYHVOBDNWTKHYTISUQMFSNVBGUHDQFYCSZLZAFABKYQSZRQGKXOXORQPSJ"
-	  "NKRVVAXMMVVPBSWMTHUNXBLSVIOOQPLRPROIBBQGNQVOXQXRNNMSFGUZEIGIYMLMLYYL"
-	  "VINTZYXYXHUFMTQFPDGSFFDVCDMEZXSGQMMGJWMWANFSZNHDIIHVJFOZGMHAOVRUWWVX"
-	  "RCOJJKZLTMAOGSRWNNXPYDCQWSSOFWUKFPKQGYLFSMZBZBKWSBMMZMFPOYYMLVYHQQQF"
-	  "HORVESQYIKEBBKSEUYXUFRNNXZPUYESZWAKQQPAWZUHYJLXRXBFPRFSCHIHHDVAIKYDZ"
-	  "IDLVQBCSGOSGFOUKVMKTODAHQVTACKAONRDYENPGSQFKGKYQROFOEKMJXKFIAEKWNRJH"
-	  "RZCCSMNSSHZNSRTBGFJJDWAPVIOBDCQKMDEMUIWGMETBUTCGMMGXLHWWTXXQYAPMQQGU"
-	  "CHECLYRWNFKHGRWYZISFYKSGTJXKTKIMJKTUWOQFHXUVTOBYUCOZKLCQCUIURHSIGUAU"
-	  "FFSNEHFLHNGBNVTGPKCVCQLHEHLJRWPGSWHKFMMXSSCVTCUOEEZOFOPTDRUPPCYTRSKI"
-	  "NJBPLOHMHQLWKMVBSZLZVDZEKXKSCNWPZTGZUXRYJEPENAYFCEKCAMWQLNYDHCSUUSTJ"
-	  "GFZCTKDQYPEZKOOJQTYMWHZDDTCMYUXCAKFXISWQZDVETOAYIANBRSXLYKOGGHEEEGAX"
-	  "SRIASWZBXTFCHLKMXQDYNLZICZZANCBUVGWFWFHYAWVLQXTQPGPQGRGBEZVIIGZJUVQU"
-	  "PNXEMZDIFXAKSHTCCBQSAXWFBNLRYQKXKXQHAFZTUICFQVYOXQEHASXNSVUTYKSVYTMH"
-	  "UTPBEVILWSKGQXPAEEOPUSMGQVPWFMRZRIMJZRRQRIZBTTRROUWENBHUYVMOMVPFDLZH"
-	  "XMXYRNASODBTLCNRDSLGPTZBZHKTSMHIVNFJOMOFPHDANVSVOYZSTKOQJLDPKLOMAYUU"
-	  "CKWTOTDONTXKCEGVXKZNNFIUXHLJPFFNVMWSFPXZKROEIGQNGXWDRFJWDCUEGJCUJGVG"
-	  "PHYKQEWVOHHPURHBBIEZCKQCFUFFBDXYRNOHAVJISOQYXMRXIEFEISFEZHLNTJXYEAKM"
-	  "EKZPJQVQBKPXPFSOANBBJBAFJQEMPIELYUHYCHUTADTCLFXUQLXWTJEFLJXVBRQXYSIR"
-	  "RPMTNMFMUCHQJEVXUSFYUUYJHLEQBHUQOTVADYQGRXVLUPQXVXEDIGNSCYNMKPWLFLHZ"
-	  "LMQGPTMBDVUJOBSXDFHKSIEXJJQTURPNZVQDLUEJHQZOQSBMPMBEQOCOSEKVXIVSQIQL"
-	  "GSQAUMIAHHLQCBCQWFJWHSYNRFFTBKORISDYNRSMPVERKRJBWGYJRXMKHJDAKRANHFDH"
-	  "WTZOHYRVCTTUXCRAFNOLYPRGDYTXNSOTCDNFVVURJILWVDWOCGPQOZIGGNOEAHCBYGMC"
-	  "XGXAADYMYDAUXPDFADTVEQTHZRGYASPJRDIKJUFIRXGMFSCIURDNFIDUUDEKPFGWECZC"
-	  "OOFZWESHSPSOOBWZKIGODSLXCALULNOLQLCMMLSQDWJDTEOIYXFCLSLKJKGCGURXEEIN"
-	  "SCUQTRDGMYXFZEFBATVSYVAJISCBVDBZAJAPKBBQTQNNRZYBLGWPCIORYJJEKIZXRRCG"
-	  "ZHGQNMGWVNIANJXYJMWRCGDGDFFQISSZOWTQOKWRGXSGRUSJOHABJUEUIHNTLCXJQPNF"
-	  "YUKTQCGRFGZOQWYLJOGOGSRDXESAOTHZVHBEOOYJZYTMOSUUXDQNKTQBVUMRBPJEJIBU"
-	  "TOVXSGYSTADWQKFUEFWJDCAYTEYVZYDCQTHXJWYUESZSLRBKRAMLVVVBMEYSYFNBLKTY"
-	  "UJRQBOKJQTYXTOFPWGWEEANVFYMAVRKMNJARUOKTZTMMJKNVFEVSECABUZGGUEHRJIHO"
-	  "JODXJOOGFZWNURNEXBUCCUHUDYXBZTNJZSQGHAGYLJQSSJERWEGUFAJXGNBXDWVFSCEG"
-	  "SUCLQLHHTRQADZIBKFCBBEXUDLPCFUDONSHUUCREKHDUBQBKMECPPOGFYIUZNUDKTILB"
-	  "IHMYNAMJIDTJEQTVPFZDNONMWFIJEAUOLPWOZVKEFTXRCXNWHWPYDHLWNSWTMUEIBMSK"
-	  "FWROAFBMEOJAVLMNCNWEMGRUDKRPENXCJQQGLPCODNOTGPOQFZTOBEBJIDAMMZARXTMC"
-	  "EAHKYNTYZKWCUFYOSCOPICKIDAUSZHWNTVRRSTIMHFBUALQECIZKYFUYJHDTFDODXXNJ"
-	  "AKZNUMYZSHIGZQQXBTORWFCGQFKFURMZYWBAQSHAJEASIYAFQZDOUHBJXODDUAWNKWVB"
-	  "NABZSUNRULZDXKBRGVCKUIYVRVRMTDFSWCTCDYKBZDELJBDIHLOALYEKHNMECBWRZQBK"
-	  "XLWFYJYECKOJOGXJYBKFSUQZKEUWBHEWNHSZKJPRQRMLRFLJWBDZEJQVYRAFQGEOGUBU"
-	  "SVVWUGKXHSXHRXWCZKASIYPZZDLRUVBNBUQEEPZPHMSNUETUKMYWNJLEZWVOLZBQMLWE"
-	  "YPPVBOTADNFNJWUZKDWRXCJCMDQPGPBIVAVQJVTHEPLXEKPSPJQFNGILKTUETORGMHGH"
-	  "HHXTZIXUXPDLKNYGHNNTKAZFCGCUONFANKRXHGPQLPDJACZDMSSFPJHRPPGGSVEYKQHF"
-	  "JKASAYIFKXXVEYRCLIMLDEUQWIHZXPLKDCHUFYAHLOQUJMJCXTHFAOSOEYWOMFAZHGPE"
-	  "UNYKWLFYQPMRYXDVGWWMOLXHCHQADSYAAQMLBGGNQELFWMYHPWNIDOIFLGHGQUPCVPHS"
-	  "WDGERQMWOZBWFHTOSINKTPXQLFGHLVALHCYSPKFBWSYTUHMZQNZSDAQTAZLPHSYZKROA"
-	  "PSKJEWCRABAGYIIAYAUOVMTYIQOWYWHLLEXOOVZNLXOIPFNRYHDJYVTJLOHODYRBBBSB"
-	  "NPQCNUZHYTWDAQSEBEMDSEDTKORHUCILESZYYYLJNRCFHAWNUMQHDQKXGCJIZOGBDVTU"
-	  "PWNKAHKIRBHINKVRRBOPVAWCSPVMTNQQYDYLCKHPKRFYCAYUPGAOQLJSTHGMCXUUQIVI"
-	  "MNPISKLETFEWNDDTDHLCQXNGRJJZHGJYYWNXKQIWXENKWQAGAXTCTLGWNVXDMMPHYPBQ"
-	  "GWWNCWJWYVBVYOWOEJJPZAZGKJEQHPDBYUQBMOLOIMZXYXFOBNNPMGDCTXLCHBEBHCOS"
-	  "HGKAEBGPLANNUHMOHWHAQEWFJSWPIFTWZNKWHKZYDXDAJODTHPXPIGVFYVDNWEQFKKIC"
-	  "EBMGPSBVTPXODJVAYJAURNSFOCUNJROYEMOELHMIGLFDMJQFVEOSINHIWDUUIPNSBHEC"
-	  "TKUFRERFNYCWSSGCYQWMXOQFCZPCSAVRBMSFZEYDBSWWHYLHIGGIDQJRTLNJOMWQVKES"
-	  "KTFWQIRKJEMAZSMFQQSSTCXKOUZLJJWNYJJKSHPAOTEWEKKABTJDOFRGKVBMJFKFFVSP"
-	  "GPMUCDWAFPHLKKZGEYTQNFJBGJTSATHNVDWRKSMLAJAPHYEJXYEKCTKDFGDILOIRDWLV"
-	  "LAMTOCSMRMXYYHPHYBMKAVDRWYSXVPLZUBPAVUUQDNRCNYPKUSWBTCHJMIHQJNXXXXQX"
-	  "LIUZQDFCTJBHELXALVTAJDFIPIFAKJKCPPPPXAVPTOUTLTIGMBUWOIERHBYOIMWTTXOY"
-	  "KCZKDVRSARRBMSQFZGGSVPVBHKBYXZBITZDBQDBZLQNPVEQTXECOHOJKEXUUIBXSORPF"
-	  "THLTMDDDOYSQZBGMBGZFYAJMHWZOLRUUJAIHOLSCIYGHMRAEIKFLFNLEMHOPKVRTJCMJ"
-	  "DKBKJCMDBPGUGPPZBCXYRLHZUYPMIQOXYOCGKBEHZFHGAAKQINMHUNTSJHPPZGNKFREX"
-	  "HGGFEFDAWMCMIXEDLUPDAXNCTHFDMHJPZOGJILKJXRUQBGXKDXTBZXSPZLZUCNCZZYRU"
-	  "DRUEVXRELACIWMGUIBKEXSYTBIJTJPJJLCIQQBVJGXHPCTSHPASIIAMPATSDCTXZAPCJ"
-	  "ESVMBOTOLKGHVRZSBVOBSAAKRPSAFYNPIDVFUMNMJRGKWOANKHZYCABHWIWUJFLDPSFY"
-	  "SPBXQEFLNEEMIGMWWXYTXNTHRXUZQKXCMBLEHGRBFPSUMGMBJFFWTAEFCLDBOHMNAICE"
-	  "YAZCTBCHKXEIBYUTQEAOVDJVOLTYDJJUCPSXUEPTFZPSJOMQDSSKBAHRIVYHQJFVUQJH"
-	  "HRAQYZLYTOAWWIIPEUPQEBYSKTRETZEDWVALVPISUBTOWJZQLVRWKLLLMWEAZZIGMTRV"
-	  "DXJHBFOTBFYSKQYJSNVKINMYRAMBFMBZUVHEEUWRRCMLAPOKKIODTFIIVTIPTBMVMZIP"
-	  "HKDIQRFOKYTVDLAPPUMYNZBJMZMDDQDTXZWOMAJSJETWLGSAJDNNCODMAMCGADNXJGPP"
-	  "GMPQXTZYICNPVATOCYCCGVSAKGCSGCPVUFGNGJPCRVZQXIDIZYCEBNMLMYHUMJZNHGCZ"
-	  "TIYTNXTCGMGSBGLIDHYABMLEBGAHLOEYVOAMROXQAFDNEIZAOFWDETNEZWJFHTYOEVDH"
-	  "RZDIZNSNBDCERUYZRLFWAANFAETBEWWPNNMUYXVBVDKMWPJUZLEPXOJAZOAYNKZDTBJO"
-	  "MKXEMIAGHIQIHXPZGWDEQJKBNTDIWPLDANSOQJTGVPPSROOXGEBBWKLXMUEKZBKTTQTN"
-	  "HILWSXGGYZZFYPGDVMNGLGGBSZJYWXGVHAACMVKQLPYXWWJMOQJJCXQOUIRCXPYCITBW"
-	  "WCOKSSDXXWHHSOPANMWVIKJFLYNBPQAUHWKZEQBDVWDDULWFVBXUBJVOMHNAELALFGZT"
-	  "FHSVLXJTMSOAKGBXKHKDEKVVFRZPKLTRRKCREVPXQYEVGIHAUNEJDLWERMKWJJBONLPW"
-	  "BNTDWFILGFYOOAPNXVBDGGYDCZWUSJLPLZRVYAVSQFEXRZACHLLGGDEJGURDPYFCQUCX"
-	  "OTNJJUJMMBEYSKJPKBGMQAAWOSPYAWDPJEQRLDJVZMRCYSKBAOQYWTJXSRRRMSFGNFRX"
-	  "VEWFGPXVPZQXGRGOLKIYCCAOAGCZSHGPEGVMWETUWVZXZKCJVGQNEJBICWINFFHIUSUE"
-	  "SVKYPVAOKBQEMQNKAXHFDETVNNFPQAXQXFJSKJCFLAHNRQUQPXYELGDYJDEETNEYGTMQ"
-	  "MYIGQUDGOUMZDQORHNVDXBCJYBFTCDEMATBUCGAEKONTAIRCOHHKJFORZVDQSOWKDLKU"
-	  "UOTTLLKQHVKVYOLJWNRRWXEOJMDZIKPMZFBBJCQSVTMCFFZDCWLRQDKAOKEXFVDZCPHE"
-	  "UGNBVSILDBRPBJKNZXIWTNOZEDDESKOGYFOWPZJBQIQXKQERGWTJCUXGTOHLAOWDLGPF"
-	  "RHVNLXVTKDDCWWZUAXKJZKPMOZCZCYVOIRCJNIJAMFRPAWNKFYAGKHNEPKPFTTRXXGTN"
-	  "XFHFJXQWMKDSODYMNWJIFQZJOCYHUKYNYKXSLJNMBKCUXETKQTSZAXZUEERRGNBFDXJW"
-	  "IKSVAVDNXLHJXVNDGLFZNZYTMJPDDXCOPGEROKJOGHXLDZJTSWOXIVCJKLYYTKANVFIF"
-	  "KRVVEXSZXREAOBPVQRZUQKIWNMPTFUWARTQJCDTPEOVDHTDCWAWTTMNLRTOPYGPBROQA"
-	  "QSMTGDCATMYQDJZQAUBQPOZLXDIJCKEVXDTPDNCHNJXMHNRVLWLKSAUYTHVOIYAAHUCY"
-	  "XUWPHXMFXBDNBSNZIVKZANBYCHTUDOLHVVXGFDRDOIRCCRTZRNPWFTHSQRYTLGETDTGZ"
-	  "KQWQPZYUTFFVFRQCAJAFDJAUMHPRXGUTLMFKOGXQHGOACGELSACTSNJTSOGNJAODFCIB"
-	  "EQGGZETTRWOLAPRYJIGCYSUIMDCRHRGBKJFPDXVGPSPGDMBOLAVARRJSNDMZIOPKTAMA"
-	  "NRZOSWGMPBPDBPCUADFMIEWOEBNNDFHUJCFMWXAFDKSWUTMMSQZSFBOGSSKOITDUGLLA"
-	  "NFCZSGCDFDWUDTBPFKMHROQMIQRLCNGGCSFTNCVXSJJGPUZQREUASWLVOFVHRZPQUYGE"
-	  "TBCJUSBKOJVBDWDEGXWBJPFTBFEALLVFFREQSGIONNYQFBFNKENZKSWLEWGEJJPZLRPV"
-	  "RLDHLELHFXXIEXDJOSAUGITKSWYCPMEBLWOLAEJXXARMXQCBEFKLOOIYRBYVWLNZKNCA"
-	  "ISOZEWVWPUDRGLYYKWWTNCWPZLYYDBKARCATEJHLYYQFNYATGYJIUNMVLEGUYBJVEVVM"
-	  "WBVJYQEGPZUOJYLWNEMODVPHTWTLURFETDHELDHPGEQWLRKZRFCEORRGYSBIJHPVNSNZ"
-	  "CAVXFCESMJIXGJPGLKDREKYGBPQSACESXAAHVQYFBSYWDPHDNTOZTEGSCISTBUMLDOQP"
-	  "WNQCSJUAENKYFCYIYENHTIAARBRTOCYDVMHPCPRGWQVDIFLHWMZWRGCWQTAIKCAJAMLG"
-	  "OKGURQTGNNUSCBKTPUXIGKDUWUGMTLCAOTKMXVTAEYWNWBTIQMLVDIEVBAGJMCWMMGZI"
-	  "USESSUIHYSYOXHFLNZVTCVATIUVWGDETJUEHTVQBJDGHIIDFYHTFDIDPLNGXLIBMKAYG"
-	  "DOZLOFKQXVWSQPRYUKAMGEICLKOMYNLWEMKLWDOPEEZGTXVDQWMULORFGNLKNVVCGQXQ"
-	  "CUIYKIAMJSJVQJKRBNBIEELCZLMPQILCEBLZZTKCPBOQLTMZRGTNWHPIUYWMBCNGIABH"
-	  "WIAILEDQVNWKJGYWMZYWOZGRHQOUFGCAETYUTLYZBCHKANZYPXTLMVREDIWBPAISSWKP"
-	  "IJIBRLWPFXQOLOFIXLUGDVMXNPJWBMMZYJUKUZPPTGVCRIMITUTPYLJOIDQGOMYUFSJB"
-	  "MRAZFVZSFZUSNJYDWBUDMFTFDBRCZCZZERWZSOXVAZISSEOMPKHESJRLUMZBBLBIXPZR"
-	  "UYRFMLZQMMZWMNXULVXZQPOUKPMXISKOPTEDXASVPAIENMUWMBNGIVWVOQQXKGYEAMAT"
-	  "YLDPYAASJSACUYILVJBXLHEMYKRKXEIVCDWKPQUBHACBNBGVTQLFDFPLAGUWUPBNBSIT"
-	  "UEENOTZFWDWNNYZTHJIBDYMCERWQNSDKDDUPPXPJTGDYQTFNSRQZZSZBYGSQHRRGSVAQ"
-	  "QEICLCSLMWQYXGGJEPJWZXKJUCFHJRACHRHLCRQWKXSUFJNBOMGWAIBKNWUDBJTFWVBW"
-	  "UPIUAKBMXXDVVKBEUAEHMOELYCJVEFJEDFBTITDNTEGRAIOACOHFCVERCTRMUZHPRNPQ"
-	  "YCFDKGTRGWFEJAXVTTBMIAYLJZEJAAEOBCTIXHYNDWPXIWOVGXSOLTXIBPBLHYKHIDGX"
-	  "HHXNVRCMUXBZGFIEDZDBZOKSMKRNYTWJGJBIMQIOZQRFROWMLNYPDDKTRESDVHHJNRMN"
-	  "GASH";
-
-int main(void)
-{
-	const char *s = "Hello, world!";
-	return strlen(s);
-}

runlevel/CMakeLists.txt

@@ -0,0 +1,12 @@
+file(GLOB runlevels *.runlevel)
+foreach (f ${runlevels})
+	get_filename_component(name ${f} NAME_WLE)
+	bsp_add_file(
+		ID runlevel-${name}
+		SRC_PATH ${f}
+		DEST_DIR /etc/herdd/runlevels)
+	sysroot_add_file(
+		ID runlevel-${name}
+		SRC_PATH ${f}
+		DEST_DIR /etc/herdd/runlevels)
+endforeach (f)

runlevel/minimal.runlevel

@@ -0,0 +1,3 @@
+[Runlevel]
+Description=Minimal
+Requires=nsd

services/herdd/CMakeLists.txt

@@ -0,0 +1,10 @@
+file(GLOB sources *.c)
+add_executable(herdd ${sources})
+target_link_libraries(herdd libc libc-runtime libmango libpthread)
+
+sysroot_add_program(
+	NAME herdd
+	BIN_DIR /usr/bin)
+bsp_add_program(
+	NAME herdd
+	BIN_DIR /usr/bin)

services/herdd/main.c

@@ -0,0 +1,43 @@
+#include <errno.h>
+#include <mango/log.h>
+#include <pthread.h>
+#include <stdio.h>
+#include <string.h>
+
+static void *thread_func(void *arg)
+{
+	kern_logf("started thread with arg %p", arg);
+	errno = 100;
+	return (void *)0xdeadbeef;
+}
+
+int main(int argc, const char *argv[], const char *envp[])
+{
+	kern_logf("herdd");
+	kern_logf("args:");
+
+	for (int i = 0; i < argc; i++) {
+		kern_logf("[%d]: %s", i, argv[i]);
+	}
+
+	kern_logf("env:");
+
+	for (int i = 0; envp[i]; i++) {
+		kern_logf("[%d]: %s", i, envp[i]);
+	}
+
+	kern_logf("self = %p", pthread_self());
+	errno = 200;
+
+	pthread_t thread;
+	pthread_create(&thread, NULL, thread_func, (void *)0xdeafcafe);
+	kern_logf("started thread %p", thread);
+	void *ret = NULL;
+	pthread_join(thread, &ret);
+	kern_logf("thread returned %p", ret);
+
+	kern_logf("errno...");
+	kern_logf("%u", errno);
+	kern_logf("...errno");
+	return 0;
+}

services/herdd/runlevel.h

@@ -0,0 +1,14 @@
+#ifndef RUNLEVEL_H_
+#define RUNLEVEL_H_
+
+#include <stddef.h>
+
+#define RUNLEVEL_DESCRIPTION_MAX 64
+
+struct runlevel {
+	char rl_description[RUNLEVEL_DESCRIPTION_MAX];
+	char **rl_requires;
+	size_t rl_requires_count;
+};
+
+#endif

services/herdd/service.h

@@ -0,0 +1,17 @@
+#ifndef SERVICE_H_
+#define SERVICE_H_
+
+#define SVC_DESCRIPTION_MAX 64
+
+enum service_role {
+	SVC_ROLE_NONE = 0x00u,
+	SVC_ROLE_NAMESPACE_PROVIDER = 0x01u,
+};
+
+struct service {
+	char s_description[SVC_DESCRIPTION_MAX];
+	enum service_role s_roles;
+	char *s_exec;
+};
+
+#endif

services/ldd/CMakeLists.txt

@@ -1,10 +0,0 @@
-file(GLOB sources *.c)
-add_executable(ldd ${sources})
-target_link_libraries(ldd libc-core libc-runtime libmango)
-
-sysroot_add_program(
-	NAME ldd
-	BIN_DIR /usr/bin)
-bsp_add_program(
-	NAME ldd
-	BIN_DIR /usr/bin)

services/ldd/main.c

@@ -1,4 +0,0 @@
-int main(void)
-{
-	return 0;
-}

services/nsd/CMakeLists.txt

@@ -0,0 +1,15 @@
+file(GLOB sources *.c)
+rosetta_add_service(
+	NAME nsd
+	SOURCES ${sources}
+	CFG_FILE ${CMAKE_CURRENT_SOURCE_DIR}/nsd.service)
+target_link_libraries(nsd libc libc-runtime)
+
+sysroot_add_service(
+	NAME nsd
+	BIN_DIR /usr/bin
+	SVC_DIR /etc/herdd/services)
+bsp_add_service(
+	NAME nsd
+	BIN_DIR /usr/bin
+	SVC_DIR /etc/herdd/services)

services/nsd/nsd.service

@@ -0,0 +1,6 @@
+[Unit]
+Description=Namespace Service
+
+[Service]
+Exec=/usr/bin/nsd
+Role=NamespaceProvider

sys/bootstrap/CMakeLists.txt

@@ -1,16 +1,19 @@
 file(GLOB c_sources *.c *.h)
 file(GLOB arch_sources arch/${CMAKE_SYSTEM_PROCESSOR}/*.S)
 
-set_property(SOURCE ${arch_sources} PROPERTY LANGUAGE C)
-
 add_executable(bootstrap ${c_sources} ${arch_sources})
 
 target_link_libraries(bootstrap
-	libmango libc-core libc-malloc libfs-static liblaunch libxpc-static
+	libmango librosetta
+	libc-core libc-malloc libc-pthread
+	libfs-static
+	liblaunch
+	libxpc-static
 	interface::fs)
 
 target_compile_options(bootstrap PRIVATE
-	 -fno-stack-protector -nostdlib -ffreestanding)
+	-fno-stack-protector -nostdlib -ffreestanding -fno-PIC)
 target_link_options(bootstrap PRIVATE
 	-static -nostdlib -ffreestanding)
-	#-T ${CMAKE_CURRENT_SOURCE_DIR}/arch/${TARGET_ARCH}/layout.ld)
+
+set_target_properties(bootstrap PROPERTIES POSITIION_INDEPENDENT_CODE FALSE)

sys/bootstrap/main.c

@@ -7,16 +7,18 @@
 #include <fs/context.h>
 #include <heap/heap.h>
 #include <launch.h>
+#include <mango/handle.h>
 #include <mango/log.h>
 #include <mango/msg.h>
 #include <mango/task.h>
 #include <mango/types.h>
+#include <pthread.h>
 #include <rosetta/bootstrap.h>
 #include <rosetta/fs.h>
 #include <stdint.h>
 #include <stdio.h>
 
-#define INIT_PATH "/usr/bin/test"
+#define INIT_PATH "/usr/bin/herdd"
 
 static enum launch_status resolve_dependency(
 	struct launch_ctx *ctx,
@@ -29,14 +31,12 @@ static enum launch_status resolve_dependency(
 		name++;
 	}
 
-	snprintf(s, sizeof s, "searching for library %s", name);
+	kern_tracef("searching for library %s", name);
-	kern_log(s);
 
 	struct tar *fs = arg;
 	struct tar_file file = {0};
 	if (tar_open(fs, name, &file) != 0) {
-		snprintf(s, sizeof s, "cannot find library %s", name);
+		kern_tracef("cannot find library %s", name);
-		kern_log(s);
 		return LAUNCH_ERR_CANNOT_RESOLVE_DEPENDENCY;
 	}
 
@@ -142,7 +142,8 @@ int main(
 	};
 
 	struct launch_parameters params = {
-		.p_executable = image,
+		.p_exec_image = image,
+		.p_exec_path = INIT_PATH,
 		.p_parent_task = task,
 		.p_task_name = "init",
 		.p_local_address_space = address_space,
@@ -169,6 +170,10 @@ int main(
 		return -1;
 	}
 
+	kern_handle_close(result.r_task);
+	kern_handle_close(result.r_thread);
+	kern_handle_close(result.r_address_space);
+
 	heap_t heap = HEAP_INIT;
 
 	struct fs_allocator fs_allocator = {
@@ -185,17 +190,20 @@ int main(
 		return -1;
 	}
 
+	fs_context_set_channel(fs, channel);
 	enum fs_status fs_status = fs_context_mount_filesystem(
 		fs,
 		tar_mount,
 		(void *)bsp_base,
 		0);
 	if (fs_status != FS_SUCCESS) {
-		kern_logf("cannot mount filesustem (%d)", fs_status);
+		kern_logf("cannot mount filesystem (%d)", fs_status);
 		return -1;
 	}
 
 	while (1) {
+		fs_context_handle_request(fs);
+#if 0
 		xpc_msg_t msg;
 		kern_status_t status = xpc_msg_recv(channel, &msg);
 		if (status != KERN_OK) {
@@ -219,6 +227,7 @@ int main(
 			kern_logf("message reply error %d", status);
 			continue;
 		}
+#endif
 	}
 
 	return 0;

sys/bootstrap/tar.c

@@ -74,8 +74,10 @@ int tar_open(struct tar *tar, const char *path, struct tar_file *out)
 		}
 
 		s += header.size;
+		if ((uintptr_t)s % (sizeof *bin_header)) {
 			s += ((sizeof *bin_header)
 			      - ((uintptr_t)s % (sizeof *bin_header)));
+		}
 
 		bin_header = (struct tar_bin_header *)s;
 	}
@@ -189,7 +191,7 @@ static enum fs_status file_read(
 	size_t w;
 	xpc_buffer_write(buf, src, count, &w);
 
-	offset += count;
+	offset += w;
 	*seek = offset;
 
 	return FS_SUCCESS;
@@ -382,8 +384,10 @@ static enum fs_status build_dentry_tree(
 		}
 
 		s += header.size;
+		if ((uintptr_t)s % (sizeof *bin_header)) {
 			s += ((sizeof *bin_header)
 			      - ((uintptr_t)s % (sizeof *bin_header)));
+		}
 
 		bin_header = (struct tar_bin_header *)s;
 	}

sys/ld/CMakeLists.txt

@@ -1,8 +1,6 @@
 file(GLOB c_sources *.c *.h)
 file(GLOB arch_sources arch/${CMAKE_SYSTEM_PROCESSOR}/*.S)
 
-set_property(SOURCE ${arch_sources} PROPERTY LANGUAGE C)
-
 add_executable(ld ${c_sources} ${arch_sources})
 set_target_properties(ld PROPERTIES
 	POSITION_INDEPENDENT_CODE ON

sys/ld/arch/x86_64/resolve.S

@@ -0,0 +1,43 @@
+.code64
+
+.global _dl_runtime_resolve
+.type _dl_runtime_resolve, @function
+
+.extern dl_runtime_resolve
+.type dl_runtime_resolve, @function
+
+_dl_runtime_resolve:
+	// pop %rdi
+	// pop %rsi
+	pop %rax
+	pop %r11
+
+	push %rdi
+	push %rsi
+	push %rdx
+	push %rcx
+	push %r8
+	push %r9
+	push %rbx
+	push %r12
+	push %r13
+	push %r14
+	push %r15
+
+	mov %rax, %rdi
+	mov %r11, %rsi
+	call dl_runtime_resolve
+
+	pop %r15
+	pop %r14
+	pop %r13
+	pop %r12
+	pop %rbx
+	pop %r9
+	pop %r8
+	pop %rcx
+	pop %rdx
+	pop %rsi
+	pop %rdi
+
+	jmp *%rax

sys/ld/btree.c

@@ -0,0 +1,691 @@
+/*
+  The Clear BSD License
+
+  Copyright (c) 2023 Max Wash
+  All rights reserved.
+
+  Redistribution and use in source and binary forms, with or without
+  modification, are permitted (subject to the limitations in the disclaimer
+  below) provided that the following conditions are met:
+
+  - Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  - Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in the
+    documentation and/or other materials provided with the distribution.
+
+  - Neither the name of the copyright holder nor the names of its
+    contributors may be used to endorse or promote products derived from this
+    software without specific prior written permission.
+ */
+
+/* templated AVL binary tree implementation
+
+   this file implements an extensible AVL binary tree data structure.
+
+   the primary rule of an AVL binary tree is that for a given node N,
+   the heights of N's left and right subtrees can differ by at most 1.
+
+   the height of a subtree is the length of the longest path between
+   the root of the subtree and a leaf node, including the root node itself.
+
+   the height of a leaf node is 1.
+
+   when a node is inserted into or deleted from the tree, this rule may
+   be broken, in which the tree must be rotated to restore the balance.
+
+   no more than one rotation is required for any insert operations,
+   while multiple rotations may be required for a delete operation.
+
+   there are four types of rotations that can be applied to a tree:
+     - left rotation
+     - right rotation
+     - double left rotations
+     - double right rotations
+
+   by enforcing the balance rule, for a tree with n nodes, the worst-case
+   performance for insert, delete, and search operations is guaranteed
+   to be O(log n).
+
+   this file intentionally excludes any kind of search function implementation.
+   it is up to the programmer to implement their own tree node type
+   using struct btree_node, and their own search function using struct btree.
+   this allows the programmer to define their own node types with complex
+   non-integer key types. btree.h contains a number of macros to help
+   define these functions. the macros do all the work, you just have to
+   provide a comparator function.
+*/
+
+#include "btree.h"
+
+#include <stddef.h>
+
+#define MAX(a, b)            ((a) > (b) ? (a) : (b))
+#define MIN(a, b)            ((a) < (b) ? (a) : (b))
+
+#define IS_LEFT_CHILD(p, c)  ((p) && (c) && ((p)->b_left == (c)))
+#define IS_RIGHT_CHILD(p, c) ((p) && (c) && ((p)->b_right == (c)))
+
+#define HAS_LEFT_CHILD(x)    ((x) && ((x)->b_left))
+#define HAS_RIGHT_CHILD(x)   ((x) && ((x)->b_right))
+
+#define HAS_NO_CHILDREN(x)   ((x) && (!(x)->b_left) && (!(x)->b_right))
+#define HAS_ONE_CHILD(x)                                                       \
+	((HAS_LEFT_CHILD(x) && !HAS_RIGHT_CHILD(x))                            \
+	 || (!HAS_LEFT_CHILD(x) && HAS_RIGHT_CHILD(x)))
+#define HAS_TWO_CHILDREN(x) (HAS_LEFT_CHILD(x) && HAS_RIGHT_CHILD(x))
+
+#define HEIGHT(x)           ((x) ? (x)->b_height : 0)
+
+static inline void update_height(struct btree_node *x)
+{
+	x->b_height = MAX(HEIGHT(x->b_left), HEIGHT((x->b_right))) + 1;
+}
+
+static inline int bf(struct btree_node *x)
+{
+	int bf = 0;
+
+	if (!x) {
+		return bf;
+	}
+
+	if (x->b_right) {
+		bf += x->b_right->b_height;
+	}
+
+	if (x->b_left) {
+		bf -= x->b_left->b_height;
+	}
+
+	return bf;
+}
+
+/* perform a left rotation on a subtree
+
+   if you have a tree like this:
+
+        Z
+       / \
+      X   .
+     / \
+    .   Y
+       / \
+      .   .
+
+   and you perform a left rotation on node X,
+   you will get the following tree:
+
+        Z
+       / \
+      Y   .
+     / \
+    X   .
+   / \
+  .   .
+
+  note that this function does NOT update b_height for the rotated
+  nodes. it is up to you to call update_height_to_root().
+*/
+static void rotate_left(struct btree *tree, struct btree_node *x)
+{
+	struct btree_node *y = x->b_right;
+
+	struct btree_node *p = x->b_parent;
+
+	if (y->b_left) {
+		y->b_left->b_parent = x;
+	}
+
+	x->b_right = y->b_left;
+
+	if (!p) {
+		tree->b_root = y;
+	} else if (x == p->b_left) {
+		p->b_left = y;
+	} else {
+		p->b_right = y;
+	}
+
+	x->b_parent = y;
+	y->b_left = x;
+	y->b_parent = p;
+}
+
+static void update_height_to_root(struct btree_node *x)
+{
+	while (x) {
+		update_height(x);
+		x = x->b_parent;
+	}
+}
+
+/* perform a right rotation on a subtree
+
+   if you have a tree like this:
+
+        Z
+       / \
+      .   X
+         / \
+        Y   .
+       / \
+      .   .
+
+   and you perform a right rotation on node X,
+   you will get the following tree:
+
+        Z
+       / \
+      .   Y
+         / \
+        .   X
+           / \
+          .   .
+
+  note that this function does NOT update b_height for the rotated
+  nodes. it is up to you to call update_height_to_root().
+*/
+static void rotate_right(struct btree *tree, struct btree_node *y)
+{
+	struct btree_node *x = y->b_left;
+
+	struct btree_node *p = y->b_parent;
+
+	if (x->b_right) {
+		x->b_right->b_parent = y;
+	}
+
+	y->b_left = x->b_right;
+
+	if (!p) {
+		tree->b_root = x;
+	} else if (y == p->b_left) {
+		p->b_left = x;
+	} else {
+		p->b_right = x;
+	}
+
+	y->b_parent = x;
+	x->b_right = y;
+	x->b_parent = p;
+}
+
+/* for a given node Z, perform a right rotation on Z's right child,
+   followed by a left rotation on Z itself.
+
+   if you have a tree like this:
+
+        Z
+       / \
+      .   X
+         / \
+        Y   .
+       / \
+      .   .
+
+   and you perform a double-left rotation on node Z,
+   you will get the following tree:
+
+         Y
+        / \
+       /   \
+      Z     X
+     / \   / \
+    .   . .   .
+
+  note that, unlike rotate_left and rotate_right, this function
+  DOES update b_height for the rotated nodes (since it needs to be
+  done in a certain order).
+*/
+static void rotate_double_left(struct btree *tree, struct btree_node *z)
+{
+	struct btree_node *x = z->b_right;
+	struct btree_node *y = x->b_left;
+
+	rotate_right(tree, x);
+	rotate_left(tree, z);
+
+	update_height(z);
+	update_height(x);
+
+	while (y) {
+		update_height(y);
+		y = y->b_parent;
+	}
+}
+
+/* for a given node Z, perform a left rotation on Z's left child,
+   followed by a right rotation on Z itself.
+
+   if you have a tree like this:
+
+        Z
+       / \
+      X   .
+     / \
+    .   Y
+       / \
+      .   .
+
+   and you perform a double-right rotation on node Z,
+   you will get the following tree:
+
+         Y
+        / \
+       /   \
+      X     Z
+     / \   / \
+    .   . .   .
+
+  note that, unlike rotate_left and rotate_right, this function
+  DOES update b_height for the rotated nodes (since it needs to be
+  done in a certain order).
+*/
+static void rotate_double_right(struct btree *tree, struct btree_node *z)
+{
+	struct btree_node *x = z->b_left;
+	struct btree_node *y = x->b_right;
+
+	rotate_left(tree, x);
+	rotate_right(tree, z);
+
+	update_height(z);
+	update_height(x);
+
+	while (y) {
+		update_height(y);
+		y = y->b_parent;
+	}
+}
+
+/* run after an insert operation. checks that the balance factor
+   of the local subtree is within the range -1 <= BF <= 1. if it
+   is not, rotate the subtree to restore balance.
+
+   note that at most one rotation should be required after a node
+   is inserted into the tree.
+
+   this function depends on all nodes in the tree having
+   correct b_height values.
+
+   @param w the node that was just inserted into the tree
+*/
+static void insert_fixup(struct btree *tree, struct btree_node *w)
+{
+	struct btree_node *z = NULL, *y = NULL, *x = NULL;
+
+	z = w;
+	while (z) {
+		if (bf(z) >= -1 && bf(z) <= 1) {
+			goto next_ancestor;
+		}
+
+		if (IS_LEFT_CHILD(z, y)) {
+			if (IS_LEFT_CHILD(y, x)) {
+				rotate_right(tree, z);
+				update_height_to_root(z);
+			} else {
+				rotate_double_right(tree, z);
+			}
+		} else {
+			if (IS_LEFT_CHILD(y, x)) {
+				rotate_double_left(tree, z);
+			} else {
+				rotate_left(tree, z);
+				update_height_to_root(z);
+			}
+		}
+
+	next_ancestor:
+		x = y;
+		y = z;
+		z = z->b_parent;
+	}
+}
+
+/* run after a delete operation. checks that the balance factor
+   of the local subtree is within the range -1 <= BF <= 1. if it
+   is not, rotate the subtree to restore balance.
+
+   note that, unlike insert_fixup, multiple rotations may be required
+   to restore balance after a node is deleted.
+
+   this function depends on all nodes in the tree having
+   correct b_height values.
+
+   @param w one of the following:
+            - the parent of the node that was deleted if the node
+              had no children.
+            - the parent of the node that replaced the deleted node
+              if the deleted node had two children.
+            - the node that replaced the node that was deleted, if
+              the node that was deleted had one child.
+*/
+static void delete_fixup(struct btree *tree, struct btree_node *w)
+{
+	struct btree_node *z = w;
+
+	while (z) {
+		if (bf(z) > 1) {
+			if (bf(z->b_right) >= 0) {
+				rotate_left(tree, z);
+				update_height_to_root(z);
+			} else {
+				rotate_double_left(tree, z);
+			}
+		} else if (bf(z) < -1) {
+			if (bf(z->b_left) <= 0) {
+				rotate_right(tree, z);
+				update_height_to_root(z);
+			} else {
+				rotate_double_right(tree, z);
+			}
+		}
+
+		z = z->b_parent;
+	}
+}
+
+/* updates b_height for all nodes between the inserted node and the root
+   of the tree, and calls insert_fixup.
+
+   @param node the node that was just inserted into the tree.
+*/
+void btree_insert_fixup(struct btree *tree, struct btree_node *node)
+{
+	node->b_height = 0;
+
+	struct btree_node *cur = node;
+	while (cur) {
+		update_height(cur);
+		cur = cur->b_parent;
+	}
+
+	insert_fixup(tree, node);
+}
+
+/* remove a node from a tree.
+
+   this function assumes that `node` has no children, and therefore
+   doesn't need to be replaced.
+
+   updates b_height for all nodes between `node` and the tree root.
+
+   @param node the node to delete.
+*/
+static struct btree_node *remove_node_with_no_children(
+	struct btree *tree,
+	struct btree_node *node)
+{
+	struct btree_node *w = node->b_parent;
+	struct btree_node *p = node->b_parent;
+	node->b_parent = NULL;
+
+	if (!p) {
+		tree->b_root = NULL;
+	} else if (IS_LEFT_CHILD(p, node)) {
+		p->b_left = NULL;
+	} else {
+		p->b_right = NULL;
+	}
+
+	while (p) {
+		update_height(p);
+		p = p->b_parent;
+	}
+
+	return w;
+}
+
+/* remove a node from a tree.
+
+   this function assumes that `node` has one child.
+   the child of `node` is inherited by `node`'s parent, and `node` is removed.
+
+   updates b_height for all nodes between the node that replaced
+   `node` and the tree root.
+
+   @param node the node to delete.
+*/
+static struct btree_node *replace_node_with_one_subtree(
+	struct btree *tree,
+	struct btree_node *node)
+{
+	struct btree_node *p = node->b_parent;
+	struct btree_node *z = NULL;
+
+	if (HAS_LEFT_CHILD(node)) {
+		z = node->b_left;
+	} else {
+		z = node->b_right;
+	}
+
+	struct btree_node *w = z;
+	if (!p) {
+		tree->b_root = z;
+	} else if (IS_LEFT_CHILD(p, node)) {
+		p->b_left = z;
+	} else if (IS_RIGHT_CHILD(p, node)) {
+		p->b_right = z;
+	}
+
+	z->b_parent = p;
+
+	node->b_parent = NULL;
+	node->b_left = node->b_right = NULL;
+
+	while (z) {
+		update_height(z);
+		z = z->b_parent;
+	}
+
+	return w;
+}
+
+/* remove a node from a tree.
+
+   this function assumes that `node` has two children.
+   find the in-order successor Y of `node` (the largest node in `node`'s left
+   sub-tree), removes `node` from the tree and moves Y to where `node` used to
+   be.
+
+   if Y has a child (it will never have more than one), have Y's parent inherit
+   Y's child.
+
+   updates b_height for all nodes between the deepest node that was modified
+   and the tree root.
+
+   @param z the node to delete.
+*/
+static struct btree_node *replace_node_with_two_subtrees(
+	struct btree *tree,
+	struct btree_node *z)
+{
+	/* x will replace z */
+	struct btree_node *x = z->b_left;
+
+	while (x->b_right) {
+		x = x->b_right;
+	}
+
+	/* y is the node that will replace x (if x has a left child) */
+	struct btree_node *y = x->b_left;
+
+	/* w is the starting point for the height update and fixup */
+	struct btree_node *w = x;
+	if (w->b_parent != z) {
+		w = w->b_parent;
+	}
+
+	if (y) {
+		w = y;
+	}
+
+	if (IS_LEFT_CHILD(x->b_parent, x)) {
+		x->b_parent->b_left = y;
+	} else if (IS_RIGHT_CHILD(x->b_parent, x)) {
+		x->b_parent->b_right = y;
+	}
+
+	if (y) {
+		y->b_parent = x->b_parent;
+	}
+
+	if (IS_LEFT_CHILD(z->b_parent, z)) {
+		z->b_parent->b_left = x;
+	} else if (IS_RIGHT_CHILD(z->b_parent, z)) {
+		z->b_parent->b_right = x;
+	}
+
+	x->b_parent = z->b_parent;
+	x->b_left = z->b_left;
+	x->b_right = z->b_right;
+
+	if (x->b_left) {
+		x->b_left->b_parent = x;
+	}
+
+	if (x->b_right) {
+		x->b_right->b_parent = x;
+	}
+
+	if (!x->b_parent) {
+		tree->b_root = x;
+	}
+
+	struct btree_node *cur = w;
+	while (cur) {
+		update_height(cur);
+		cur = cur->b_parent;
+	}
+
+	return w;
+}
+
+/* delete a node from the tree and re-balance it afterwards */
+void btree_delete(struct btree *tree, struct btree_node *node)
+{
+	struct btree_node *w = NULL;
+
+	if (HAS_NO_CHILDREN(node)) {
+		w = remove_node_with_no_children(tree, node);
+	} else if (HAS_ONE_CHILD(node)) {
+		w = replace_node_with_one_subtree(tree, node);
+	} else if (HAS_TWO_CHILDREN(node)) {
+		w = replace_node_with_two_subtrees(tree, node);
+	}
+
+	if (w) {
+		delete_fixup(tree, w);
+	}
+
+	node->b_left = node->b_right = node->b_parent = NULL;
+}
+
+struct btree_node *btree_first(struct btree *tree)
+{
+	/* the first node in the tree is the node with the smallest key.
+	   we keep moving left until we can't go any further */
+	struct btree_node *cur = tree->b_root;
+	if (!cur) {
+		return NULL;
+	}
+
+	while (cur->b_left) {
+		cur = cur->b_left;
+	}
+
+	return cur;
+}
+
+struct btree_node *btree_last(struct btree *tree)
+{
+	/* the first node in the tree is the node with the largest key.
+	   we keep moving right until we can't go any further */
+	struct btree_node *cur = tree->b_root;
+	if (!cur) {
+		return NULL;
+	}
+
+	while (cur->b_right) {
+		cur = cur->b_right;
+	}
+
+	return cur;
+}
+
+struct btree_node *btree_next(struct btree_node *node)
+{
+	if (!node) {
+		return NULL;
+	}
+
+	/* there are two possibilities for the next node:
+
+	   1. if `node` has a right sub-tree, every node in this sub-tree is
+	   bigger than node. the in-order successor of `node` is the smallest
+	   node in this subtree.
+	   2. if `node` has no right sub-tree, we've reached the largest node in
+	      the sub-tree rooted at `node`. we need to go back to our parent
+	      and continue the search elsewhere.
+	*/
+	if (node->b_right) {
+		/* case 1: step into `node`'s right sub-tree and keep going
+		   left to find the smallest node */
+		struct btree_node *cur = node->b_right;
+		while (cur->b_left) {
+			cur = cur->b_left;
+		}
+
+		return cur;
+	}
+
+	/* case 2: keep stepping back up towards the root of the tree.
+	   if we encounter a step where we are our parent's left child,
+	   we've found a parent with a value larger than us. this parent
+	   is the in-order successor of `node` */
+	while (node->b_parent && node->b_parent->b_left != node) {
+		node = node->b_parent;
+	}
+
+	return node->b_parent;
+}
+
+struct btree_node *btree_prev(struct btree_node *node)
+{
+	if (!node) {
+		return NULL;
+	}
+
+	/* there are two possibilities for the previous node:
+
+	   1. if `node` has a left sub-tree, every node in this sub-tree is
+	   smaller than `node`. the in-order predecessor of `node` is the
+	   largest node in this subtree.
+	   2. if `node` has no left sub-tree, we've reached the smallest node in
+	      the sub-tree rooted at `node`. we need to go back to our parent
+	      and continue the search elsewhere.
+	*/
+	if (node->b_left) {
+		/* case 1: step into `node`'s left sub-tree and keep going
+		   right to find the largest node */
+		struct btree_node *cur = node->b_left;
+		while (cur->b_right) {
+			cur = cur->b_right;
+		}
+
+		return cur;
+	}
+
+	/* case 2: keep stepping back up towards the root of the tree.
+	   if we encounter a step where we are our parent's right child,
+	   we've found a parent with a value smaller than us. this parent
+	   is the in-order predecessor of `node`. */
+	while (node->b_parent && node->b_parent->b_right != node) {
+		node = node->b_parent;
+	}
+
+	return node->b_parent;
+}