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),)
@@ -429,6 +434,7 @@ 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/fcntl.h

@@ -15,6 +15,9 @@
 /// Truncate file.
 #define O_TRUNC 0x0200
 
+/// Mask to determine the file access mode.
+#define O_ACCMODE 0x3
+
 /// Set file descriptor offset to offset.
 #define SEEK_SET 0
 /// Set file descriptor offset to current plus offset.

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/socket.h

@@ -8,8 +8,11 @@
 #define AF_INET 2
 // Types
 #define SOCK_DGRAM 2
+#define SOCK_RAW   3
 // Protocols
-#define IPPROTO_UDP 17
+#define IPPROTO_IP   0
+#define IPPROTO_ICMP 1
+#define IPPROTO_UDP  17
 
 typedef uint16_t sa_family_t;
 

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

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

@@ -28,11 +28,34 @@ start:
   ; load the 64-bit GDT
   lgdt [gdt64.pointer]
 
+  call enable_sse
+
+  mov eax, cr4
+  or eax, 1 << 16
+  mov cr4, eax
+
   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/proc/process.c

@@ -36,11 +36,16 @@ process_t* process_exec(uint8_t* image, const char* name, char* const argv[])
   }
 
   // Load ELF in current process.
+  // TODO: Handle the case where the image isn't a valid/supported ELF.
   elf_header_t* elf = elf_load(image);
   current_process->elf = elf;
 
   // Set current process name.
   current_process->name = strdup(name);
+
+  memset(current_process->user_heap, 0, USER_HEAP_SIZE);
+  current_process->user_brk = (uintptr_t)current_process->user_heap;
+
   // Set up user stack.
   memset(current_process->user_stack, 0, USER_STACK_SIZE);
 
@@ -66,8 +71,35 @@ process_t* process_exec(uint8_t* image, const char* name, char* const argv[])
   void* stack = (void*)&current_process->user_stack[USER_STACK_TOP];
   char* buf = (char*)&current_process->user_stack[USER_STACK_BUF];
 
+  size_t off = 0;
+
   // auxv
+  uint8_t rand_bytes[16] = {
+    0xaa, 0xc0, 0xff, 0xee, 0xc0, 0xff, 0xee, 0xc0,
+    0xff, 0xee, 0xc0, 0xff, 0xee, 0xc0, 0xff, 0xee,
+  };
+  memcpy(&buf[off], rand_bytes, sizeof(rand_bytes));
+  char* rand_ptr = &buf[off];
+  off += sizeof(rand_bytes);
+
+  // aux: AT_NULL
+  PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)NULL);
   PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)NULL);
+  // aux: AT_RANDOM
+  PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)rand_ptr);
+  PUSH_TO_STACK(stack, uint64_t, 25);
+  // aux: AT_PAGESZ
+  PUSH_TO_STACK(stack, uint64_t, 4096);
+  PUSH_TO_STACK(stack, uint64_t, 6);
+  // aux: AT_PHNUM
+  PUSH_TO_STACK(stack, uint64_t, elf->ph_num);
+  PUSH_TO_STACK(stack, uint64_t, 5);
+  // aux: AT_PHENT
+  PUSH_TO_STACK(stack, uint64_t, elf->ph_size);
+  PUSH_TO_STACK(stack, uint64_t, 4);
+  // aux: AT_PHDR
+  PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)image + elf->ph_offset);
+  PUSH_TO_STACK(stack, uint64_t, 3);
 
   // envp
   PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)NULL);
@@ -81,7 +113,6 @@ process_t* process_exec(uint8_t* image, const char* name, char* const argv[])
   // argv
   PUSH_TO_STACK(stack, uintptr_t, (uintptr_t)NULL);
 
-  size_t off = 0;
   for (int i = argc - 1; i >= 0; i--) {
     size_t len = strlen(argv[i]);
     // Copy argv[i] to high address in initial process stack.

src/kernel/arch/x86_64/proc/process.h

@@ -4,17 +4,20 @@
 #include <core/elf.h>
 #include <sys/types.h>
 
-#define USER_STACK_SIZE 0x3000
-#define USER_STACK_TOP  0x1000
-#define USER_STACK_BUF  0x2000
+#define USER_STACK_SIZE 0x30000
+#define USER_STACK_TOP  0x01000
+#define USER_STACK_BUF  0x02000
+#define USER_HEAP_SIZE  0x10000
 
 typedef struct process
 {
   pid_t pid;
   char* name;
   elf_header_t* elf;
   uint64_t user_stack[USER_STACK_SIZE];
-  uint64_t user_rsp;
+  uintptr_t user_rsp;
+  uint64_t user_heap[USER_HEAP_SIZE];
+  uintptr_t user_brk;
   // TODO: We should probably retain the file descriptors that are opened by
   // the process so that we can close them.
 } process_t;

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;
   }