ww/cmd/w6l/out.c

/*
 * out.c — emit a static ELF64 executable.
 *
 * Layout (file order) without .data:
 *   [0..64)       ELF header
 *   [64..120)     one program header (PT_LOAD R+X)
 *   [120..0x1000) zero pad
 *   [0x1000..)    .text bytes
 *
 * With .data (any input .o has writable globals):
 *   [0..64)        ELF header
 *   [64..176)      two program headers (PT_LOAD R+X, PT_LOAD R+W)
 *   [176..0x1000)  zero pad
 *   [0x1000..)     .text bytes
 *   [data_off..)   .data bytes (file offset and vaddr page-aligned)
 *
 * No interpreter, no dynamic, no .bss yet. Entry point is the address
 * of the symbol named "_start" (or whatever main supplies via -e).
 */
#include "l.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#define ET_EXEC		2
#define EM_X86_64	62
#define EV_CURRENT	1
#define ELFCLASS64	2
#define ELFDATA2LSB	1

#define PT_LOAD		1
#define PF_X		1
#define PF_W		2
#define PF_R		4

#pragma pack(push, 1)
typedef struct {
	u8 e_ident[16];
	u16 e_type, e_machine;
	u32 e_version;
	u64 e_entry, e_phoff, e_shoff;
	u32 e_flags;
	u16 e_ehsize, e_phentsize, e_phnum, e_shentsize, e_shnum, e_shstrndx;
} Ehdr;

typedef struct {
	u32 p_type, p_flags;
	u64 p_offset, p_vaddr, p_paddr;
	u64 p_filesz, p_memsz, p_align;
} Phdr;
#pragma pack(pop)

int
l_emit_elf(Lnk *l, FILE *f, u64 base, u64 entry)
{
	/* Dispatch: any loaded shared object plus any dynamic ref means
	 * we owe the loader a real PT_INTERP/PT_DYNAMIC binary. */
	if (l->sos != NULL && l->dyn_n > 0)
		return l_emit_dyn_elf(l, f, base, entry);

	const u64 page     = 0x1000;
	const u64 text_off = 0x1000;
	const u64 rx_end   = text_off + l->textlen;
	const int has_data = (l->datalen > 0);

	/* BSS optimisation: trailing zero bytes in .data can be left out
	 * of the file. The loader zero-fills the gap between p_filesz and
	 * p_memsz, so this shrinks the binary without changing semantics.
	 * Scan after l_relocate has applied DATAR patches — anything still
	 * zero at the tail genuinely is zero-init. */
	u64 bsslen = 0;
	if (has_data) {
		while (bsslen < l->datalen
		    && l->data[l->datalen - 1 - bsslen] == 0)
			bsslen++;
	}
	const u64 data_file_len = l->datalen - bsslen;

	/* data goes at the next page boundary so the loader can grant a
	 * fresh page of R+W permissions without overlapping the R+X mapping. */
	const u64 data_off = has_data ? ((rx_end + page - 1) & ~(page - 1)) : 0;
	const u64 data_va  = has_data ? (base + data_off) : 0;
	const u64 file_end = has_data ? (data_off + data_file_len) : rx_end;
	(void)data_va;

	Ehdr eh = {0};
	memcpy(eh.e_ident, "\x7f""ELF", 4);
	eh.e_ident[4] = ELFCLASS64;
	eh.e_ident[5] = ELFDATA2LSB;
	eh.e_ident[6] = EV_CURRENT;
	eh.e_type = ET_EXEC;
	eh.e_machine = EM_X86_64;
	eh.e_version = EV_CURRENT;
	eh.e_entry = entry;
	eh.e_phoff = sizeof(Ehdr);
	eh.e_ehsize = sizeof(Ehdr);
	eh.e_phentsize = sizeof(Phdr);
	eh.e_phnum = has_data ? 2 : 1;

	/* R+X load covering [0, rx_end). When .data is present we still
	 * round up to a page in memsz so the loader doesn't try to give
	 * the same page both R+X and R+W permissions. */
	Phdr phx = {0};
	phx.p_type = PT_LOAD;
	phx.p_flags = PF_R | PF_X;
	phx.p_offset = 0;
	phx.p_vaddr = base;
	phx.p_paddr = base;
	phx.p_filesz = rx_end;
	phx.p_memsz = rx_end;
	phx.p_align = page;

	Phdr phw = {0};
	if (has_data) {
		phw.p_type = PT_LOAD;
		phw.p_flags = PF_R | PF_W;
		phw.p_offset = data_off;
		phw.p_vaddr = base + data_off;
		phw.p_paddr = base + data_off;
		phw.p_filesz = data_file_len;
		phw.p_memsz = l->datalen;
		phw.p_align = page;
	}

	fwrite(&eh, 1, sizeof eh, f);
	fwrite(&phx, 1, sizeof phx, f);
	if (has_data) fwrite(&phw, 1, sizeof phw, f);

	/* pad to text_off */
	long here = ftell(f);
	for (long i = here; i < (long)text_off; i++) fputc(0, f);

	if (l->textlen) fwrite(l->text, 1, l->textlen, f);

	if (has_data && data_file_len > 0) {
		/* pad to data_off */
		here = ftell(f);
		for (long i = here; i < (long)data_off; i++) fputc(0, f);
		fwrite(l->data, 1, data_file_len, f);
	}
	(void)file_end;

	return 0;
}