initial commit

.gitignore

@@ -0,0 +1 @@
+builddir/

meson.build

@@ -0,0 +1,12 @@
+project('layer9', 'c')
+
+add_global_arguments('-pie', language : 'c')
+
+executable(
+    'layer9',
+    './src/mach.c',
+    './src/start.s',
+    './src/syscall.c',
+    './src/main.c',
+    dependencies : []
+)

myfile

@@ -0,0 +1 @@
+hello

src/mach.c

@@ -0,0 +1,24 @@
+#include "mach.h"
+
+const char *exec_magic_str(Exec *exec) {
+    switch (exec->magic) {
+        case A_MAGIC: return "68020";
+        case I_MAGIC: return "intel 386";
+        case J_MAGIC: return "intel 960 (retired)";
+        case K_MAGIC: return "sparc";
+        case V_MAGIC: return "mips 3000 BE";
+        case X_MAGIC: return "att dsp 3210 (retired)";
+        case M_MAGIC: return "mips 4000 BE";
+        case D_MAGIC: return "amd 29000 (retired)";
+        case E_MAGIC: return "arm";
+        case Q_MAGIC: return "powerpc";
+        case N_MAGIC: return "mips 4000 LE";
+        case L_MAGIC: return "dec alpha (retired)";
+        case P_MAGIC: return "mips 3000 LE";
+        case U_MAGIC: return "sparc64";
+        case S_MAGIC: return "amd64";
+        case T_MAGIC: return "powerpc64";
+        case R_MAGIC: return "arm64";
+        default: return "[unsupported]";
+    }
+}

src/mach.h

@@ -0,0 +1,44 @@
+#ifndef _MACH_H
+#define _MACH_H
+
+#include <stdint.h>
+
+typedef struct Exec {
+	int32_t       magic;      /* magic number */
+	int32_t       text;       /* size of text segment */
+	int32_t       data;       /* size of initialized data */
+	int32_t       bss;        /* size of uninitialized data */
+	int32_t       syms;       /* size of symbol table */
+	int32_t       entry32;      /* entry point */
+	int32_t       spsz;       /* size of pc/sp offset table */
+	int32_t       pcsz;       /* size of pc/line number table */
+	int64_t       entry;      /* entry point for HDR_MAGIC */
+} __attribute__((__packed__)) Exec;
+
+const static int exec_field_count = sizeof(Exec) / sizeof(int32_t);
+
+#define HDR_MAGIC   0x00008000  /* header expansion */
+
+#define  _MAGIC(f, b)           ((f)|((((4*(b))+0)*(b))+7))
+#define  A_MAGIC    _MAGIC(0, 8)        /* 68020 */
+#define  I_MAGIC    _MAGIC(0, 11)       /* intel 386 */
+#define  J_MAGIC    _MAGIC(0, 12)       /* intel 960 (retired) */
+#define  K_MAGIC    _MAGIC(0, 13)       /* sparc */
+#define  V_MAGIC    _MAGIC(0, 16)       /* mips 3000 BE */
+#define  X_MAGIC    _MAGIC(0, 17)       /* att dsp 3210 (retired) */
+#define  M_MAGIC    _MAGIC(0, 18)       /* mips 4000 BE */
+#define  D_MAGIC    _MAGIC(0, 19)       /* amd 29000 (retired) */
+#define  E_MAGIC    _MAGIC(0, 20)       /* arm */
+#define  Q_MAGIC    _MAGIC(0, 21)       /* powerpc */
+#define  N_MAGIC    _MAGIC(0, 22)       /* mips 4000 LE */
+#define  L_MAGIC    _MAGIC(0, 23)       /* dec alpha (retired) */
+#define  P_MAGIC    _MAGIC(0, 24)       /* mips 3000 LE */
+#define  U_MAGIC    _MAGIC(0, 25)       /* sparc64 */
+#define  S_MAGIC    _MAGIC(HDR_MAGIC, 26)   /* amd64 */
+#define  T_MAGIC    _MAGIC(HDR_MAGIC, 27)   /* powerpc64 */
+#define  R_MAGIC    _MAGIC(HDR_MAGIC, 28)   /* arm64 */
+
+
+const char *exec_magic_str(Exec *exec);
+
+#endif

src/main.c

@@ -0,0 +1,152 @@
+#include "syscall.h"
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdint.h>
+#include <arpa/inet.h>
+#include <sys/mman.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <memory.h>
+#include <inttypes.h>
+#include <stdint.h>
+#include <elf.h>
+#include <link.h>
+#include <sys/prctl.h>
+#include <linux/prctl.h>
+#include <sys/signal.h>
+#include <fcntl.h>
+#include "mach.h"
+#include "tos.h"
+#include "sys.h"
+
+#define ALIGN(v, a) (((v) + (a) - 1) & ~((a) - 1))
+#define DONE_ERR(M_err, M_ret) do { fprintf(stderr, (M_err)); defer_status = (M_ret); goto done; } while(0)
+
+static char sud_intercept = SYSCALL_DISPATCH_FILTER_ALLOW;
+
+
+extern void start(uintptr_t entry, Tos* tos, int argc, char **argv);
+void enter(uintptr_t entry, int argc, char **argv)
+{
+	Tos tos;
+	tos.pid = getpid();
+	printf("jump to entry 0x%"PRIxPTR"\n", entry);
+	sud_intercept = SYSCALL_DISPATCH_FILTER_BLOCK;
+	start(entry, &tos, argc, argv);
+}
+
+void handle_sigsys(int sig, siginfo_t *info, void *ptr)
+{
+	ucontext_t *uc = ptr;
+	mcontext_t *mc = &uc->uc_mcontext;
+	greg_t *gregs = mc->gregs;
+
+	int64_t ret = -1;
+	int64_t *sp = (int64_t *)gregs[REG_RSP];
+	int64_t syscall = gregs[REG_RBP];
+
+	printf("syscall %ld\n", syscall);
+
+	switch (syscall) {
+	case OPEN:
+		ret = sys$open((char *)sp[1], (int)sp[2]);
+		break;
+	}
+
+	gregs[REG_RAX] = ret;
+}
+
+int findlibc(struct dl_phdr_info *info, size_t size, void *ptr)
+{
+	const Elf64_Phdr *p;
+	uintptr_t offset, length, libc;
+
+	libc = (uintptr_t)read;  /* arbitrary function used to identify libc text segment */
+	for (p = info->dlpi_phdr; p < info->dlpi_phdr + info->dlpi_phnum; ++p) {
+		if (p->p_type != PT_LOAD || !(p->p_flags & PF_X))
+			continue;
+		offset = info->dlpi_addr + p->p_vaddr;
+		length = p->p_memsz;
+		if (offset < libc && libc - offset < length) {
+			if (prctl(PR_SET_SYSCALL_USER_DISPATCH, PR_SYS_DISPATCH_ON, offset, length, &sud_intercept) != 0) {
+				perror("prctl PR_SET_SYSCALL_USER_DISPATCH");
+				exit(1);
+			}
+			return 1;
+		}
+	}
+	return 0;
+}
+
+static char *program_argv[] = {
+	"out",
+	"../myfile",
+	NULL
+};
+
+int main()
+{
+	int defer_status = 0;
+	void *file_segments = NULL;
+	FILE *f = NULL;
+	Exec header = {0};
+	size_t text_addr = 0x200000;
+	size_t data_addr;
+
+	f = fopen("../a.out", "rb");
+
+	if (fread(&header, 1, sizeof(header), f) != sizeof(header)) {
+		DONE_ERR("failed to read file header\n", 1);
+	}
+
+	// header fields are big-endian
+	int32_t *buf = (int32_t*)&header;
+	for (int i = 0; i < exec_field_count; i++) {
+		buf[i] = ntohl(buf[i]);
+	}
+
+	printf("load %s binary with entry 0x%"PRIxPTR"\n", exec_magic_str(&header), (uintptr_t)header.entry);
+
+	data_addr = ALIGN(text_addr + header.text, 0x200000);
+
+	size_t segments_size = sizeof(Exec) + header.text + header.data;
+	file_segments = malloc(segments_size);
+	if (!file_segments) {
+		defer_status = 1;
+		goto done;
+	}
+	if (fread(file_segments, 1, segments_size, f) != segments_size) {
+		DONE_ERR("failed to read file header and segments\n", 1);
+	}
+
+	void *text_segment = mmap((void*)text_addr, header.text, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+	if (!text_segment) {
+		DONE_ERR("failed to map text segment into memory\n", 1);
+	}
+	printf("mapped text segment at 0x%"PRIxPTR"\n", (uintptr_t)text_segment);
+
+	void *data_segment = mmap((void*)data_addr, header.data + header.bss, PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+	if (!text_segment) {
+		DONE_ERR("failed to map data segment into memory\n", 1);
+	}
+	printf("mapped data segment at 0x%"PRIxPTR"\n", (uintptr_t)data_segment);
+
+	memcpy(text_segment, file_segments, header.text);
+	memcpy(data_segment, file_segments + sizeof(Exec) + header.text, header.data);
+
+	if (dl_iterate_phdr(findlibc, NULL) != 1) {
+		DONE_ERR("failed to find libc and enable SUD\n", 1);
+	}
+
+	struct sigaction sa = {
+		.sa_sigaction = handle_sigsys,
+		.sa_flags = SA_SIGINFO
+	};
+	sigaction(SIGSYS, &sa, NULL);
+
+	enter(header.entry32, 2, program_argv);
+done:
+	free(file_segments);
+	fclose(f);
+	return defer_status;
+}

src/start.s

@@ -0,0 +1,25 @@
+/* Thanks: https://git.sr.ht/~mcf/nine/tree/main/item/start.s */
+
+/* SPDX-License-Identifier: Unlicense */
+.text
+.globl start
+start:
+	mov %rdi, %rbp /* entry */
+	mov %rsi, %rax /* _tos */
+	mov %rdx, %rbx /* argc */
+	mov %rcx, %rsi /* argv */
+
+	/* push argv onto stack */
+	mov %rbx, %rcx
+	add $1, %rcx
+	sal $3, %rcx
+	sub %rcx, %rsp
+	mov %rsp, %rdi
+	rep movsb
+
+	/* push argc onto stack */
+	push %rbx
+
+	jmp *%rbp
+
+.section .note.GNU-stack,"",@progbits

src/sys.h

@@ -0,0 +1,52 @@
+#define	SYSR1		0
+#define	_ERRSTR		1
+#define	BIND		2
+#define	CHDIR		3
+#define	CLOSE		4
+#define	DUP		5
+#define	ALARM		6
+#define	EXEC		7
+#define	EXITS		8
+#define	_FSESSION	9
+#define	FAUTH		10
+#define	_FSTAT		11
+#define	SEGBRK		12
+#define	_MOUNT		13
+#define	OPEN		14
+#define	_READ		15
+#define	OSEEK		16
+#define	SLEEP		17
+#define	_STAT		18
+#define	RFORK		19
+#define	_WRITE		20
+#define	PIPE		21
+#define	CREATE		22
+#define	FD2PATH		23
+#define	BRK_		24
+#define	REMOVE		25
+#define	_WSTAT		26
+#define	_FWSTAT		27
+#define	NOTIFY		28
+#define	NOTED		29
+#define	SEGATTACH	30
+#define	SEGDETACH	31
+#define	SEGFREE		32
+#define	SEGFLUSH	33
+#define	RENDEZVOUS	34
+#define	UNMOUNT		35
+#define	_WAIT		36
+#define	SEMACQUIRE	37
+#define	SEMRELEASE	38
+#define	SEEK		39
+#define	FVERSION	40
+#define	ERRSTR		41
+#define	STAT		42
+#define	FSTAT		43
+#define	WSTAT		44
+#define	FWSTAT		45
+#define	MOUNT		46
+#define	AWAIT		47
+#define	PREAD		50
+#define	PWRITE		51
+#define	TSEMACQUIRE	52
+#define _NSEC		53

src/syscall.c

@@ -0,0 +1,28 @@
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <stdio.h>
+#include <inttypes.h>
+#include "syscall.h"
+
+static int mode_open(char *name, int mode, int perm, int flag)
+{
+	switch (mode & 3) {
+	case 0: flag |= O_RDONLY; break;
+	case 1: flag |= O_WRONLY; break;
+	case 2: flag |= O_RDWR; break;
+	case 3: flag |= O_PATH; break;
+	}
+	if (mode & 16)
+		flag |= O_TRUNC;
+	if (mode & 32)
+		flag |= O_CLOEXEC;
+	return open(name, flag, perm);
+}
+
+
+int sys$open(char *path, int mode)
+{
+	printf("sys$open(0x%"PRIxPTR", %d)\n", (uintptr_t)path, mode);
+
+    return mode_open(path, mode, 0, 0);
+}

src/syscall.h

@@ -0,0 +1,3 @@
+#pragma once
+
+int sys$open(char *path, int mode);

src/tos.h

@@ -0,0 +1,26 @@
+/* Thanks: https://git.sr.ht/~mcf/nine/tree/main/item/tos.h */
+
+#include <stdint.h>
+
+typedef struct Tos Tos;
+typedef struct Plink Plink;
+
+struct Tos {
+	struct				/* Per process profiling */
+	{
+		Plink		*pp;	/* known to be 0(ptr) */
+		Plink		*next;	/* known to be 4(ptr) */
+		Plink		*last;
+		Plink		*first;
+		uint32_t	pid;
+		uint32_t	what;
+	} prof;
+	uint64_t	cyclefreq;	/* cycle clock frequency if there is one, 0 otherwise */
+	int64_t		kcycles;	/* cycles spent in kernel */
+	int64_t		pcycles;	/* cycles spent in process (kernel + user) */
+	uint32_t	pid;		/* might as well put the pid here */
+	uint32_t	clock;
+	/* top of stack is here */
+};
+
+extern Tos *_tos;