kernel: add linux compatibility layer

Makefile

@@ -160,6 +160,11 @@ ifeq ($(CONFIG_USE_FAKE_CLOCK), 1)
  libk_c_files += $(kernel_src_dir)/time/fake_clock.c
 endif
 
+ifeq ($(CONFIG_LINUX_COMPAT), 1)
+ KERNEL_CONFIG += -DCONFIG_LINUX_COMPAT
+ LIBC_CONFIG += -DCONFIG_LINUX_COMPAT
+endif
+
 # This file exists in a Docker container because we copy it in `Dockerfile`.
 in_docker = $(wildcard /tmp/install-linux-deps)
 ifneq ($(in_docker),)
@@ -192,7 +197,7 @@ LIBC_CFLAGS += $(LIBC_CONFIG)
 KERNEL_INCLUDES += $(INCLUDES)
 KERNEL_INCLUDES += -I$(include_dir)/kernel/ -I$(arch_src)/
 KERNEL_INCLUDES += $(addprefix -I$(external_dir)/,$(addsuffix /, $(external_deps)))
-KERNEL_ASM_FLAGS +=
+KERNEL_ASM_FLAGS += $(KERNEL_CONFIG)
 # We need to have -fno-omit-frame-pointer or the kernel stack backtrace won't
 # get the stack.
 KERNEL_CFLAGS += -fno-omit-frame-pointer
@@ -434,6 +439,7 @@ what: ## display some information about the current configuration
  echo "OS_NAME : $(OS_NAME)"
  echo "ARCH : $(ARCH)"
  echo ""
+ echo "CONFIG_LINUX_COMPAT = $(CONFIG_LINUX_COMPAT)"
  echo "CONFIG_SEMIHOSTING = $(CONFIG_SEMIHOSTING)"
  echo "CONFIG_USE_DLMALLOC = $(CONFIG_USE_DLMALLOC)"
  echo "CONFIG_USE_FAKE_CLOCK = $(CONFIG_USE_FAKE_CLOCK)"

Makefile-cfg.include

@@ -6,6 +6,8 @@ LLVM_SUFFIX ?=
 
 # When set to 1, enable the Undefined Behavior SANitizer.
 UBSAN ?=
+# When set to 1, enable Linux (binary) compatibility.
+CONFIG_LINUX_COMPAT ?=
 # When set to 1, enable semi-hosting mode (QEMU, mainly).
 CONFIG_SEMIHOSTING ?=
 # When set to 1, use dlmalloc for malloc/free/realloc (instead of liballoc).

data/initrd/etc/hosts

@@ -0,0 +1 @@
+140.82.121.4 github

data/initrd/etc/passwd

@@ -0,0 +1 @@
+root:x:0:0::/:none

include/kernel/proc/descriptor.h

@@ -2,9 +2,11 @@
 #ifndef PROC_DESCRIPTOR_H
 #define PROC_DESCRIPTOR_H
 
+#include <arpa/inet.h>
 #include <fs/vfs.h>
 #include <stdbool.h>
 #include <stdint.h>
+#include <sys/socket.h>
 #include <sys/types.h>
 
 #define STDIN 0
@@ -23,6 +25,8 @@ typedef struct descriptor
  uint32_t type;
  uint32_t protocol;
  uint16_t port;
+ struct sockaddr_in addr;
+ socklen_t addr_len;
 } descriptor_t;
 
 /**
@@ -82,4 +86,8 @@ int descriptor_udp_lookup(uint16_t port);
  */
 bool is_protocol_supported(uint32_t type, uint32_t protocol);
 
+void duplicate_descriptor(int oldfd, int newfd);
+
+int descriptor_raw_lookup(uint32_t protocol, in_addr_t src_addr);
+
 #endif

include/libc/sys/linux_compat.h

@@ -0,0 +1,95 @@
+// https://chromium.googlesource.com/chromiumos/docs/+/HEAD/constants/syscalls.md
+#ifndef SYS_LINUX_COMPAT_H
+#define SYS_LINUX_COMPAT_H
+
+#ifdef __x86_64__
+
+#define SYSCALL_READ 0
+#define SYSCALL_WRITE 1
+#define SYSCALL_OPEN 2
+#define SYSCALL_CLOSE 3
+#define SYSCALL_FSTAT 5
+#define SYSCALL_LSEEK 8
+#define SYSCALL_BRK 12
+#define SYSCALL_IOCTL 16
+#define SYSCALL_WRITEV 20
+#define SYSCALL_DUP2 33
+#define SYSCALL_GETPID 39
+#define SYSCALL_SOCKET 41
+#define SYSCALL_SENDTO 44
+#define SYSCALL_RECVFROM 45
+#define SYSCALL_EXECV 49
+#define SYSCALL_EXIT 60
+#define SYSCALL_GETTIMEOFDAY 96
+#define SYSCALL_GETEUID 107
+#define SYSCALL_ARCH_PRCTL 158
+#define SYSCALL_REBOOT 169
+#define SYSCALL_SET_TID_ADDR 218
+#define SYSCALL_EXIT_GROUP 231
+#define SYSCALL_OPENAT 257
+
+#elif __arm__
+
+#define SYSCALL_EXIT 1
+#define SYSCALL_READ 3
+#define SYSCALL_WRITE 4
+#define SYSCALL_OPEN 5
+#define SYSCALL_CLOSE 6
+#define SYSCALL_EXECV 11
+#define SYSCALL_LSEEK 19
+#define SYSCALL_GETPID 20
+#define SYSCALL_BRK 45
+#define SYSCALL_GETEUID 49
+#define SYSCALL_IOCTL 54
+#define SYSCALL_DUP2 63
+#define SYSCALL_GETTIMEOFDAY 78
+#define SYSCALL_REBOOT 88
+#define SYSCALL_FSTAT 108
+#define SYSCALL_WRITEV 146
+#define SYSCALL_EXIT_GROUP 248
+#define SYSCALL_SET_TID_ADDR 256
+#define SYSCALL_SOCKET 281
+#define SYSCALL_SENDTO 290
+#define SYSCALL_RECVFROM 292
+#define SYSCALL_OPENAT 322
+
+// Not available on AArch32:
+//
+// - SYSCALL_ARCH_PRCTL
+
+#elif __aarch64__
+
+#define SYSCALL_IOCTL 29
+#define SYSCALL_OPENAT 56
+#define SYSCALL_CLOSE 57
+#define SYSCALL_LSEEK 62
+#define SYSCALL_READ 63
+#define SYSCALL_WRITE 64
+#define SYSCALL_WRITEV 66
+#define SYSCALL_FSTAT 80
+#define SYSCALL_EXIT 93
+#define SYSCALL_EXIT_GROUP 94
+#define SYSCALL_SET_TID_ADDR 96
+#define SYSCALL_REBOOT 142
+#define SYSCALL_GETTIMEOFDAY 169
+#define SYSCALL_GETPID 172
+#define SYSCALL_GETEUID 175
+#define SYSCALL_SOCKET 198
+#define SYSCALL_SENDTO 206
+#define SYSCALL_RECVFROM 207
+#define SYSCALL_BRK 214
+#define SYSCALL_EXECV 221
+
+// Not available on AArch64:
+//
+// - SYSCALL_OPEN
+// - SYSCALL_DUP2
+// - SYSCALL_ARCH_PRCTL
+
+#endif
+
+// Not available outside ArvernOS:
+#define SYSCALL_TEST 348
+#define SYSCALL_GETHOSTBYNAME2 349
+
+#endif

include/libc/sys/syscall.h

@@ -11,6 +11,12 @@
 #include <sys/time.h>
 #include <sys/types.h>
 
+#ifdef CONFIG_LINUX_COMPAT
+
+#include <sys/linux_compat.h>
+
+#else
+
 #define SYSCALL_TEST 1
 #define SYSCALL_WRITE 2
 #define SYSCALL_READ 3
@@ -29,6 +35,8 @@
 #define SYSCALL_EXIT 16
 #define SYSCALL_OPENAT 17
 
+#endif // CONFIG_LINUX_COMPAT
+
 #define SYSCALL_SET_ERRNO() \
  if (retval < 0) { \
  errno = -retval; \

src/kernel/arch/x86_64/Makefile.include

@@ -27,7 +27,7 @@ QEMU = qemu-system-x86_64
 # Options for the different tools
 ###############################################################################
 
-QEMU_OPTIONS += -m 512M
+QEMU_OPTIONS += -m 512M -cpu IvyBridge
 QEMU_OPTIONS += -serial file:$(log_file)
 QEMU_OPTIONS += -netdev user,id=u1,ipv6=off,dhcpstart=10.0.2.20
 QEMU_OPTIONS += -device rtl8139,netdev=u1
@@ -51,6 +51,7 @@ libk_c_files += $(wildcard $(arch_src)/drivers/*.c)
 libk_c_files += $(wildcard $(arch_src)/fs/dev/*.c)
 libk_c_files += $(wildcard $(arch_src)/fs/proc/*.c)
 libk_c_files += $(wildcard $(arch_src)/mmu/*.c)
+libk_c_files += $(wildcard $(arch_src)/proc/*.c)
 libk_c_files += $(wildcard $(arch_src)/sys/*.c)
 libk_asm_files += $(wildcard $(arch_src)/asm/*.asm)
 libc_test_files += $(patsubst $(tests_dir)/%.c, %, $(shell find $(tests_dir)/libc -name '*.c'))

src/kernel/arch/x86_64/asm/boot.asm

@@ -28,11 +28,37 @@ start:
  ; load the 64-bit GDT
  lgdt [gdt64.pointer]
 
+%ifdef CONFIG_LINUX_COMPAT
+ call enable_sse
+
+ ; enables the instructions RDFSBASE, RDGSBASE, WRFSBASE, and WRGSBASE
+ mov eax, cr4
+ or eax, 1 << 16
+ mov cr4, eax
+%endif
+
  jmp gdt64.kernel_code:long_mode_start
 
  ; Should not be reached.
  hlt
 
+enable_sse:
+ mov eax, 0x1 ; check for SSE
+ cpuid
+ test edx, 1 << 25
+ jz .no_sse ; after this, SSE can be enabled
+ mov eax, cr0
+ and ax, 0xFFFB ; clear coprocessor emulation CR0.EM
+ or ax, 0x2 ; set coprocessor monitoring CR0.MP
+ mov cr0, eax
+ mov eax, cr4
+ or ax, 3 << 9 ; set CR4.OSFXSR and CR4.OSXMMEXCPT at the same time
+ mov cr4, eax
+ ret
+
+.no_sse:
+ ret
+
 ; -----------------------------------------------------------------------------
 ; make sure the kernel was really loaded by a Multiboot compliant bootloader
 %define MULTIBOOT2_MAGIC_VALUE 0x36d76289

src/kernel/arch/x86_64/asm/k_syscall.asm

@@ -3,6 +3,10 @@ global syscall_handler
 syscall_handler:
  sti
 
+%ifdef CONFIG_LINUX_COMPAT
+ push rsp
+%endif
+
  push rcx ; save the return address
  push r11 ; save the flags
  push rbp
@@ -39,4 +43,21 @@ syscall_handler:
  pop rcx
 
  cli
+
+; TODO: this is needed because usermode has been broken recently (when the
+; multi-tasking code has been introduced). Therefore, we only have kernel mode
+; (ring 0). `sysret` always returns to ring 3 so we cannot use it when a syscall
+; is done from ring 0 (which happens with Linux unmodified binaries).
+%ifdef CONFIG_LINUX_COMPAT
+ pop rbp
+
+ push 0x10
+ push rbp
+ push r11
+ push 0x08
+ push rcx
+
+ iretq
+%else
  o64 sysret
+%endif

src/kernel/arch/x86_64/asm/proc.asm

@@ -30,3 +30,11 @@ ret_from_fork:
  call task_schedule_tail
 
  call r12
+
+%ifdef CONFIG_LINUX_COMPAT
+global do_start
+
+do_start:
+ mov rsp, rsi
+ jmp rdi
+%endif

src/kernel/arch/x86_64/core/elf.c

@@ -92,7 +92,8 @@ void load_segment(uint8_t* data, elf_program_header_t* program_header)
  uint32_t flags =
  PAGING_FLAG_PRESENT | PAGING_FLAG_USER_ACCESSIBLE | PAGING_FLAG_WRITABLE;
 
- if (!(program_header->flags & ELF_PROGRAM_FLAG_EXECUTE)) {
+ if ((program_header->flags & ELF_PROGRAM_FLAG_EXECUTE) !=
+ ELF_PROGRAM_FLAG_EXECUTE) {
  flags |= PAGING_FLAG_NO_EXECUTE;
  }
 

src/kernel/kmain.c

@@ -112,16 +112,12 @@ void kinit()
  argv[_argc] = NULL;
  free(_cmdline);
 
+ INFO("kmain: loading %s...", argv[0]);
+
  if (strcmp(argv[0], "kshell") == 0) {
- INFO("kmain: loading %s...", argv[0]);
  kshell(argc, argv);
  } else {
- // TODO: create task
-
- // INFO("kmain: switching to usermode... (%s)", argv[0]);
- //
- // k_execv(argv[0], argv);
- WARN("cannot execute: %s", saved_cmdline);
+ k_execv(argv[0], argv);
  }
 
  k_exit(EXIT_FAILURE);

src/kernel/net/ipv4.c

@@ -5,9 +5,11 @@
 #include <net/icmpv4.h>
 #include <net/logging.h>
 #include <net/udp.h>
+#include <proc/descriptor.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <sys/syscall.h>
 
 ipv4_header_t ipv4_create_header(uint8_t src_ip[4],
  in_addr_t dst_addr,
@@ -36,6 +38,13 @@ void ipv4_receive_packet(net_interface_t* interface,
  src_ip[3],
  interface->id);
 
+ int sockfd = descriptor_raw_lookup(header.proto, header.src_addr);
+ if (sockfd >= 0) {
+ // Handle SOCK_RAW.
+ write(sockfd, data, len);
+ return;
+ }
+
  switch (header.proto) {
  case IPV4_PROTO_ICMP:
  icmpv4_receive_packet(interface, data, &header);

src/kernel/net/udp.c

@@ -34,7 +34,7 @@ void udp_receive_packet(net_interface_t* interface,
  int sockfd = descriptor_udp_lookup(udp_header.dst_port);
  NET_DEBUG("got sockfd=%d for dst_port=%d", sockfd, udp_header.dst_port);
 
- if (sockfd > 0) {
+ if (sockfd >= 0) {
  write(sockfd, udp_data, udp_header.len - sizeof(udp_header_t));
  return;
  }