dyngen.c

qemu虚拟机代码

    int offset, min_offset, pc_offset, data_size;
    uint8_t data_allocated[1024];
    unsigned int data_index;
    
    memset(data_allocated, 0, sizeof(data_allocated));
    
    p = p_start;
    min_offset = p_end - p_start;
    while (p < p_start + min_offset) {
        insn = get32((uint32_t *)p);
        if ((insn & 0x0d5f0000) == 0x051f0000) {
            /* ldr reg, [pc, #im] */
            offset = insn & 0xfff;
            if (!(insn & 0x00800000))
                        offset = -offset;
            if ((offset & 3) !=0)
                error("%s:%04x: ldr pc offset must be 32 bit aligned", 
                      name, start_offset + p - p_start);
            pc_offset = p - p_start + offset + 8;
            if (pc_offset <= (p - p_start) || 
                pc_offset >= (p_end - p_start))
                error("%s:%04x: ldr pc offset must point inside the function code", 
                      name, start_offset + p - p_start);
            if (pc_offset < min_offset)
                min_offset = pc_offset;
            if (outfile) {
                /* ldr position */
                fprintf(outfile, "    arm_ldr_ptr->ptr = gen_code_ptr + %d;\n", 
                        p - p_start);
                /* ldr data index */
                data_index = ((p_end - p_start) - pc_offset - 4) >> 2;
                fprintf(outfile, "    arm_ldr_ptr->data_ptr = arm_data_ptr + %d;\n", 
                        data_index);
                fprintf(outfile, "    arm_ldr_ptr++;\n");
                if (data_index >= sizeof(data_allocated))
                    error("%s: too many data", name);
                if (!data_allocated[data_index]) {
                    ELF_RELOC *rel;
                    int i, addend, type;
                    const char *sym_name, *p;
                    char relname[1024];

                    data_allocated[data_index] = 1;

                    /* data value */
                    addend = get32((uint32_t *)(p_start + pc_offset));
                    relname[0] = '\0';
                    for(i = 0, rel = relocs;i < nb_relocs; i++, rel++) {
                        if (rel->r_offset == (pc_offset + start_offset)) {
                            sym_name = get_rel_sym_name(rel);
                            /* the compiler leave some unnecessary references to the code */
                            get_reloc_expr(relname, sizeof(relname), sym_name);
                            type = ELF32_R_TYPE(rel->r_info);
                            if (type != R_ARM_ABS32)
                                error("%s: unsupported data relocation", name);
                            break;
                        }
                    }
                    fprintf(outfile, "    arm_data_ptr[%d] = 0x%x",
                            data_index, addend);
                    if (relname[0] != '\0')
                        fprintf(outfile, " + %s", relname);
                    fprintf(outfile, ";\n");
                }
            }
        }
        p += 4;
    }
    data_size = (p_end - p_start) - min_offset;
    if (data_size > 0 && outfile) {
        fprintf(outfile, "    arm_data_ptr += %d;\n", data_size >> 2);
    }

    /* the last instruction must be a mov pc, lr */
    if (p == p_start)
        goto arm_ret_error;
    p -= 4;
    insn = get32((uint32_t *)p);
    if ((insn & 0xffff0000) != 0xe91b0000) {
    arm_ret_error:
        if (!outfile)
            printf("%s: invalid epilog\n", name);
    }
    return p - p_start;	    
}
#endif


#define MAX_ARGS 3

/* generate op code */
void gen_code(const char *name, host_ulong offset, host_ulong size, 
              FILE *outfile, int gen_switch)
{
    int copy_size = 0;
    uint8_t *p_start, *p_end;
    host_ulong start_offset;
    int nb_args, i, n;
    uint8_t args_present[MAX_ARGS];
    const char *sym_name, *p;
    EXE_RELOC *rel;

    /* Compute exact size excluding prologue and epilogue instructions.
     * Increment start_offset to skip epilogue instructions, then compute
     * copy_size the indicate the size of the remaining instructions (in
     * bytes).
     */
    p_start = text + offset;
    p_end = p_start + size;
    start_offset = offset;
#if defined(HOST_I386) || defined(HOST_X86_64)
#ifdef CONFIG_FORMAT_COFF
    {
        uint8_t *p;
        p = p_end - 1;
        if (p == p_start)
            error("empty code for %s", name);
        while (*p != 0xc3) {
            p--;
            if (p <= p_start)
                error("ret or jmp expected at the end of %s", name);
        }
        copy_size = p - p_start;
    }
#else
    {
        int len;
        len = p_end - p_start;
        if (len == 0)
            error("empty code for %s", name);
        if (p_end[-1] == 0xc3) {
            len--;
        } else {
            error("ret or jmp expected at the end of %s", name);
        }
        copy_size = len;
    }
#endif    
#elif defined(HOST_PPC)
    {
        uint8_t *p;
        p = (void *)(p_end - 4);
        if (p == p_start)
            error("empty code for %s", name);
        if (get32((uint32_t *)p) != 0x4e800020)
            error("blr expected at the end of %s", name);
        copy_size = p - p_start;
    }
#elif defined(HOST_S390)
    {
        uint8_t *p;
        p = (void *)(p_end - 2);
        if (p == p_start)
            error("empty code for %s", name);
        if (get16((uint16_t *)p) != 0x07fe && get16((uint16_t *)p) != 0x07f4)
            error("br %%r14 expected at the end of %s", name);
        copy_size = p - p_start;
    }
#elif defined(HOST_ALPHA)
    {
        uint8_t *p;
        p = p_end - 4;
#if 0
        /* XXX: check why it occurs */
        if (p == p_start)
            error("empty code for %s", name);
#endif
        if (get32((uint32_t *)p) != 0x6bfa8001)
            error("ret expected at the end of %s", name);
        copy_size = p - p_start;	    
    }
#elif defined(HOST_IA64)
    {
        uint8_t *p;
        p = (void *)(p_end - 4);
        if (p == p_start)
            error("empty code for %s", name);
        /* br.ret.sptk.many b0;; */
        /* 08 00 84 00 */
        if (get32((uint32_t *)p) != 0x00840008)
            error("br.ret.sptk.many b0;; expected at the end of %s", name);
	copy_size = p_end - p_start;
    }
#elif defined(HOST_SPARC)
    {
        uint32_t start_insn, end_insn1, end_insn2;
        uint8_t *p;
        p = (void *)(p_end - 8);
        if (p <= p_start)
            error("empty code for %s", name);
        start_insn = get32((uint32_t *)(p_start + 0x0));
        end_insn1 = get32((uint32_t *)(p + 0x0));
        end_insn2 = get32((uint32_t *)(p + 0x4));
        if ((start_insn & ~0x1fff) == 0x9de3a000) {
            p_start += 0x4;
            start_offset += 0x4;
            if ((int)(start_insn | ~0x1fff) < -128)
                error("Found bogus save at the start of %s", name);
            if (end_insn1 != 0x81c7e008 || end_insn2 != 0x81e80000)
                error("ret; restore; not found at end of %s", name);
        } else {
            error("No save at the beginning of %s", name);
        }
#if 0
        /* Skip a preceeding nop, if present.  */
        if (p > p_start) {
            skip_insn = get32((uint32_t *)(p - 0x4));
            if (skip_insn == 0x01000000)
                p -= 4;
        }
#endif
        copy_size = p - p_start;
    }
#elif defined(HOST_SPARC64)
    {
        uint32_t start_insn, end_insn1, end_insn2, skip_insn;
        uint8_t *p;
        p = (void *)(p_end - 8);
        if (p <= p_start)
            error("empty code for %s", name);
        start_insn = get32((uint32_t *)(p_start + 0x0));
        end_insn1 = get32((uint32_t *)(p + 0x0));
        end_insn2 = get32((uint32_t *)(p + 0x4));
        if ((start_insn & ~0x1fff) == 0x9de3a000) {
            p_start += 0x4;
            start_offset += 0x4;
            if ((int)(start_insn | ~0x1fff) < -256)
                error("Found bogus save at the start of %s", name);
            if (end_insn1 != 0x81c7e008 || end_insn2 != 0x81e80000)
                error("ret; restore; not found at end of %s", name);
        } else {
            error("No save at the beginning of %s", name);
        }
        
        /* Skip a preceeding nop, if present.  */
        if (p > p_start) {
            skip_insn = get32((uint32_t *)(p - 0x4));
            if (skip_insn == 0x01000000)
                p -= 4;
        }
        
        copy_size = p - p_start;
    }
#elif defined(HOST_ARM)
    {
        if ((p_end - p_start) <= 16)
            error("%s: function too small", name);
        if (get32((uint32_t *)p_start) != 0xe1a0c00d ||
            (get32((uint32_t *)(p_start + 4)) & 0xffff0000) != 0xe92d0000 ||
            get32((uint32_t *)(p_start + 8)) != 0xe24cb004)
            error("%s: invalid prolog", name);
        p_start += 12;
        start_offset += 12;
        copy_size = arm_emit_ldr_info(name, start_offset, NULL, p_start, p_end, 
                                      relocs, nb_relocs);
    }
#elif defined(HOST_M68K)
    {
        uint8_t *p;
        p = (void *)(p_end - 2);
        if (p == p_start)
            error("empty code for %s", name);
        // remove NOP's, probably added for alignment
        while ((get16((uint16_t *)p) == 0x4e71) &&
               (p>p_start)) 
            p -= 2;
        if (get16((uint16_t *)p) != 0x4e75)
            error("rts expected at the end of %s", name);
        copy_size = p - p_start;
    }
#else
#error unsupported CPU
#endif

    /* compute the number of arguments by looking at the relocations */
    for(i = 0;i < MAX_ARGS; i++)
        args_present[i] = 0;

    for(i = 0, rel = relocs;i < nb_relocs; i++, rel++) {
        host_ulong offset = get_rel_offset(rel);
        if (offset >= start_offset &&
	    offset < start_offset + (p_end - p_start)) {
            sym_name = get_rel_sym_name(rel);
            if(!sym_name)
                continue;
            if (strstart(sym_name, "__op_param", &p) ||
                strstart(sym_name, "__op_gen_label", &p)) {
                n = strtoul(p, NULL, 10);
                if (n > MAX_ARGS)
                    error("too many arguments in %s", name);
                args_present[n - 1] = 1;
            }
        }
    }
    
    nb_args = 0;
    while (nb_args < MAX_ARGS && args_present[nb_args])
        nb_args++;
    for(i = nb_args; i < MAX_ARGS; i++) {
        if (args_present[i])
            error("inconsistent argument numbering in %s", name);
    }

    if (gen_switch == 2) {
        fprintf(outfile, "DEF(%s, %d, %d)\n", name + 3, nb_args, copy_size);
    } else if (gen_switch == 1) {

        /* output C code */
        fprintf(outfile, "case INDEX_%s: {\n", name);
        if (nb_args > 0) {
            fprintf(outfile, "    long ");
            for(i = 0; i < nb_args; i++) {
                if (i != 0)
                    fprintf(outfile, ", ");
                fprintf(outfile, "param%d", i + 1);
            }
            fprintf(outfile, ";\n");
        }
#if defined(HOST_IA64)
        fprintf(outfile, "    extern char %s;\n", name);
#else
        fprintf(outfile, "    extern void %s();\n", name);
#endif

        for(i = 0, rel = relocs;i < nb_relocs; i++, rel++) {
            host_ulong offset = get_rel_offset(rel);
            if (offset >= start_offset &&
                offset < start_offset + (p_end - p_start)) {
                sym_name = get_rel_sym_name(rel);
                if(!sym_name)
                    continue;
                if (*sym_name && 
                    !strstart(sym_name, "__op_param", NULL) &&
                    !strstart(sym_name, "__op_jmp", NULL) &&
                    !strstart(sym_name, "__op_gen_label", NULL)) {
#if defined(HOST_SPARC)
		    if (sym_name[0] == '.') {
			fprintf(outfile,
				"extern char __dot_%s __asm__(\"%s\");\n",
				sym_name+1, sym_name);
			continue;
		    }
#endif
#if defined(__APPLE__)
/* set __attribute((unused)) on darwin because we wan't to avoid warning when we don't use the symbol */
                    fprintf(outfile, "extern char %s __attribute__((unused));\n", sym_name);
#elif defined(HOST_IA64)
			if (ELF64_R_TYPE(rel->r_info) != R_IA64_PCREL21B)
				/*
				 * PCREL21 br.call targets generally
				 * are out of range and need to go
				 * through an "import stub".
				 */
				fprintf(outfile, "    extern char %s;\n",
					sym_name);
#else
                    fprintf(outfile, "extern char %s;\n", sym_name);
#endif
                }
            }
        }

        fprintf(outfile, "    memcpy(gen_code_ptr, (void *)((char *)&%s+%d), %d);\n",
					name, (int)(start_offset - offset), copy_size);

        /* emit code offset information */
        {
            EXE_SYM *sym;
            const char *sym_name, *p;
            unsigned long val;
            int n;

            for(i = 0, sym = symtab; i < nb_syms; i++, sym++) {
                sym_name = get_sym_name(sym);
                if (strstart(sym_name, "__op_label", &p)) {
                    uint8_t *ptr;
                    unsigned long offset;
                    
                    /* test if the variable refers to a label inside
                       the code we are generating */
#ifdef CONFIG_FORMAT_COFF
                    if (sym->st_shndx == text_shndx) {
                        ptr = sdata[coff_text_shndx];
                    } else if (sym->st_shndx == data_shndx) {
                        ptr = sdata[coff_data_shndx];
                    } else {
                        ptr = NULL;
                    }
#elif defined(CONFIG_FORMAT_MACH)
                    if(!sym->n_sect)
                        continue;
                    ptr = sdata[sym->n_sect-1];
#else
                    ptr = sdata[sym->st_shndx];
#endif
                    if (!ptr)
                        error("__op_labelN in invalid section");
                    offset = sym->st_value;
#ifdef CONFIG_FORMAT_MACH
                    offset -= section_hdr[sym->n_sect-1].addr;
#endif
                    val = *(unsigned long *)(ptr + offset);
#ifdef ELF_USES_RELOCA
                    {
                        int reloc_shndx, nb_relocs1, j;

                        /* try to find a matching relocation */
                        reloc_shndx = find_reloc(sym->st_shndx);
                        if (reloc_shndx) {
                            nb_relocs1 = shdr[reloc_shndx].sh_size / 
                                shdr[reloc_shndx].sh_entsize;
                            rel = (ELF_RELOC *)sdata[reloc_shndx];
                            for(j = 0; j < nb_relocs1; j++) {
                                if (rel->r_offset == offset) {
				    val = rel->r_addend;
                                    break;
                                }
				rel++;
                            }