added option to run with gdb

1 files changed, 500 insertions, 0 deletions

diff --git a/src/elfo.zig b/src/elfo.zig
new file mode 100644
index 0000000..20f798a
--- /dev/null
+++ b/src/elfo.zig
@@ -0,0 +1,500 @@
+const std = @import("std");
+const cs = @import("capstone");
+const meta_opts = @import("meta");
+
+pub fn main(init: std.process.Init) !void {
+    var args = try init.minimal.args.iterateAllocator(init.gpa);
+    defer args.deinit();
+    _ = args.next(); // skip argv[0]
+
+    var buffer: [64]u8 = undefined;
+    const stderr = try init.io.lockStderr(&buffer, .escape_codes);
+
+    // TODO: finish passing custom alloc operations here
+    // const mem_config: cs.cs_opt_mem = undefined;
+    // std.debug.assert(cs.cs_option(0, cs.CS_OPT_MEM, @intFromPtr(&mem_config)) == cs.CS_ERR_OK);
+
+    if (meta_opts.gdb) @breakpoint();
+
+    try printElf(
+        init.gpa,
+        init.io,
+        args.next() orelse "./study-samples/split",
+        stderr.terminal(),
+        .{
+            .show_unaddressable_sections = true,
+            // .skip_sections_content = true,
+        },
+    );
+}
+
+pub fn printElf(
+    gpa: std.mem.Allocator,
+    io: std.Io,
+    path: []const u8,
+    term: std.Io.Terminal,
+    options: struct {
+        show_unaddressable_sections: bool = false,
+        skip_sections_content: bool = false,
+    },
+) !void {
+    const f = try std.Io.Dir.cwd().openFile(io, path, .{ .mode = .read_only });
+    const bw = term.writer;
+    defer bw.flush() catch {};
+    var buffer = try gpa.alignedAlloc(u8, std.mem.Alignment.of(u64), 1024 * 100);
+    defer gpa.free(buffer);
+
+    var reader = f.reader(io, buffer);
+    const header = try std.elf.Header.read(&reader.interface);
+
+    var handle: usize = undefined;
+    defer _ = cs.cs_close(@ptrCast(&handle));
+    const opts: struct { arch: u64, mode: u64 } = switch (header.machine) {
+        .X86_64 => .{ .arch = cs.CS_ARCH_X86, .mode = cs.CS_MODE_64 },
+        // The arm mode isn't working with symbols on the .text section correctly
+        .ARM => .{ .arch = cs.CS_ARCH_ARM, .mode = cs.CS_MODE_ARM },
+        else => {
+            std.debug.print("found machine: {any}\n", .{header.machine});
+            @panic("unhandled arch");
+        },
+    };
+
+    std.debug.assert(cs.cs_open(@intCast(opts.arch), @intCast(opts.mode), @ptrCast(&handle)) == cs.CS_ERR_OK);
+
+    const shstrtab = blk: {
+        var section_it = header.iterateSectionHeaders(&reader);
+        var section_idx: u32 = 0;
+        while (try section_it.next()) |s| {
+            defer section_idx += 1;
+            if (section_idx == header.shstrndx) {
+                std.debug.assert(s.sh_type == std.elf.SHT_STRTAB);
+                break :blk s;
+            }
+        }
+        break :blk null;
+    };
+
+    const elf_shstrtab_slice = blk: {
+        if (shstrtab == null)
+            break :blk null;
+
+        try reader.seekTo(shstrtab.?.sh_offset);
+        const slice = try reader.interface.readAlloc(gpa, shstrtab.?.sh_size);
+        break :blk slice;
+    };
+    defer {
+        if (elf_shstrtab_slice != null)
+            gpa.free(elf_shstrtab_slice.?);
+    }
+
+    const strtab = blk: {
+        if (elf_shstrtab_slice == null)
+            break :blk null;
+        var section_it = header.iterateSectionHeaders(&reader);
+        while (try section_it.next()) |s| {
+            if (s.sh_type == std.elf.SHT_STRTAB and std.mem.eql(
+                u8,
+                ".strtab",
+                std.mem.sliceTo(elf_shstrtab_slice.?[s.sh_name..], 0),
+            ))
+                // if (s.sh_type == std.elf.SHT_STRTAB and s.sh_name != shstrtab.?.sh_name and s.sh_addr == 0)
+                break :blk s;
+        }
+        break :blk null;
+    };
+
+    const elf_strtab_slice = blk: {
+        if (strtab == null)
+            break :blk null;
+
+        try reader.seekTo(strtab.?.sh_offset);
+        const slice = try reader.interface.readAlloc(gpa, strtab.?.sh_size);
+        break :blk slice;
+    };
+
+    defer {
+        if (elf_strtab_slice != null)
+            gpa.free(elf_strtab_slice.?);
+    }
+
+    var strs: std.ArrayList([]const u8) = try .initCapacity(gpa, 8);
+    {
+        if (elf_shstrtab_slice != null) {
+            var str_it = std.mem.splitScalar(u8, elf_shstrtab_slice.?, 0);
+            while (str_it.next()) |str| {
+                const owned_str = try gpa.alloc(u8, str.len);
+                @memcpy(owned_str, str);
+                try strs.append(gpa, owned_str);
+            }
+        }
+    }
+    defer {
+        for (strs.items) |s| {
+            gpa.free(s);
+        }
+        strs.deinit(gpa);
+    }
+
+    var sections: std.ArrayList(std.elf.Elf64_Shdr) = try .initCapacity(gpa, 8);
+    {
+        var section_it = header.iterateSectionHeaders(&reader);
+        while (try section_it.next()) |section| {
+            try sections.append(gpa, section);
+        }
+        std.mem.sort(std.elf.Elf64_Shdr, sections.items, {}, struct {
+            pub fn inner(_: void, x: std.elf.Elf64_Shdr, y: std.elf.Elf64_Shdr) bool {
+                return x.sh_addr < y.sh_addr;
+            }
+        }.inner);
+    }
+    defer sections.deinit(gpa);
+
+    const symtab = blk: {
+        var section_it = header.iterateSectionHeaders(&reader);
+        while (try section_it.next()) |s| {
+            if (s.sh_type == std.elf.SHT_SYMTAB) {
+
+                // x64: symtab: .{ .sh_name = 1, .sh_type = 2, .sh_flags = 0, .sh_addr = 0, .sh_offset = 4256, .sh_size = 1680, .sh_link = 27, .sh_info = 45, .sh_addralign = 8, .sh_entsize = 24 }
+                // arm32: .{ .sh_name = 1, .sh_type = 2, .sh_flags = 0, .sh_addr = 0, .sh_offset = 4264, .sh_size = 1856, .sh_link = 27, .sh_info = 87, .sh_addralign = 4, .sh_entsize = 16 }
+                try bw.print("symtab: {any}\n", .{s});
+                break :blk s;
+            }
+        }
+        break :blk null;
+    };
+
+    const dynsym = blk: {
+        var section_it = header.iterateSectionHeaders(&reader);
+        while (try section_it.next()) |s| {
+            if (s.sh_type == std.elf.SHT_DYNSYM) {
+                // try bw.print("sym: {any}\n", .{s});
+                break :blk s;
+            }
+        }
+        break :blk null;
+    };
+    // try bw.print("dynsym: {any}\n", .{dynsym});
+
+    var symbols_index = blk: {
+        var syms: std.ArrayList(SymbolRange) = try .initCapacity(gpa, 8);
+        if (symtab != null) {
+            var sym_it = iterateSymbols(header, &reader, symtab.?);
+            while (try sym_it.next()) |s| {
+                const t = s.st_info & 0xf;
+                const name = std.mem.sliceTo(elf_strtab_slice.?[s.st_name..], 0);
+                const owned_name = try gpa.alloc(u8, name.len);
+                @memcpy(owned_name, name);
+                try syms.append(gpa, .{
+                    .start = s.st_value,
+                    .end = s.st_value + s.st_size,
+                    .name = owned_name,
+                    .kind = t,
+                });
+            }
+        }
+
+        // the check on elf_strtab_slice might not be necessary
+        if (dynsym != null and elf_strtab_slice != null) {
+            var sym_it = iterateSymbols(header, &reader, dynsym.?);
+            while (try sym_it.next()) |s| {
+                const t = s.st_info & 0xf;
+                const name = std.mem.sliceTo(elf_strtab_slice.?[s.st_name..], 0);
+                const owned_name = try gpa.alloc(u8, name.len);
+                @memcpy(owned_name, name);
+                try syms.append(gpa, .{
+                    .start = s.st_value,
+                    .end = s.st_value + s.st_size,
+                    .name = owned_name,
+                    .kind = t,
+                });
+            }
+        }
+        std.mem.sort(SymbolRange, syms.items, {}, struct {
+            fn inner(_: void, x: SymbolRange, y: SymbolRange) bool {
+                return x.start < y.start;
+            }
+        }.inner);
+        break :blk syms;
+    };
+    defer {
+        for (symbols_index.items) |sym| {
+            gpa.free(sym.name);
+        }
+        symbols_index.deinit(gpa);
+    }
+
+    for (symbols_index.items) |sym| {
+        if (sym.kind == std.elf.STT_FUNC and sym.name.len > 0)
+            try bw.print("{s} {x}-{x}\n", .{ sym.name, sym.start, sym.end });
+    }
+
+    for (sections.items) |section| {
+        if (section.sh_size > 0 and section.sh_addr > 0) {
+            try term.setColor(.reset);
+            try term.setColor(.dim);
+            try bw.print("\n{x}-{x} (t: {x})  --  ", .{
+                section.sh_addr,
+                section.sh_addr + section.sh_size,
+                section.sh_type,
+            });
+            try term.setColor(.bright_green);
+            if (elf_shstrtab_slice != null)
+                try bw.print("{s}", .{std.mem.sliceTo(elf_shstrtab_slice.?[section.sh_name..], 0)});
+            try bw.print("\n", .{});
+            try term.setColor(.reset);
+
+            // --
+            try reader.seekTo(section.sh_offset);
+
+            if (buffer.len < section.sh_size) {
+                buffer = try gpa.realloc(buffer, section.sh_size);
+                reader = f.reader(io, buffer);
+            }
+            const section_slice = reader.interface.take(section.sh_size) catch |e| blk: {
+                switch (e) {
+                    error.EndOfStream => {
+                        try bw.print("failed\n", .{});
+                        break :blk null;
+                    },
+                    error.ReadFailed => unreachable,
+                }
+            };
+            // TODO: this heuristic is probably wrong
+            if (section_slice != null and !options.skip_sections_content) {
+                if (section.sh_type == std.elf.SHT_PROGBITS and (section.sh_flags & (std.elf.SHF_ALLOC | std.elf.SHF_EXECINSTR)) != 0) {
+                    const instrs: []cs.cs_insn = blk: {
+                        var insn: [*]cs.cs_insn = undefined;
+                        // TODO: use iter API
+                        // https://www.capstone-engine.org/iteration.html
+                        // const count = cs.cs_disasm_iter(handle, section_slice.?.ptr, section_slice.?.len, section.sh_addr, @ptrCast(&insn));
+                        const count = cs.cs_disasm(handle, section_slice.?.ptr, section_slice.?.len, section.sh_addr, 0, @ptrCast(&insn));
+                        break :blk insn[0..count];
+                    };
+
+                    try dumpInstr(gpa, bw, term, instrs, symbols_index.items);
+                } else {
+                    try dumpHexFallible(u64, bw, term, section_slice.?, section.sh_addr);
+                }
+            }
+        }
+    }
+
+    if (options.show_unaddressable_sections) {
+        for (sections.items) |section| {
+            if (section.sh_size > 0 and section.sh_addr == 0) {
+                try term.setColor(.reset);
+                try term.setColor(.dim);
+                try bw.print("{x}-{x} (t: {x})  --  ", .{
+                    section.sh_addr,
+                    section.sh_addr + section.sh_size,
+                    section.sh_type,
+                });
+                try term.setColor(.bright_cyan);
+                if (elf_shstrtab_slice != null)
+                    try bw.print("{s}", .{std.mem.sliceTo(elf_shstrtab_slice.?[section.sh_name..], 0)});
+                try bw.print("\n", .{});
+                try term.setColor(.reset);
+                // --
+
+                try reader.seekTo(section.sh_offset);
+
+                if (buffer.len < section.sh_size) {
+                    buffer = try gpa.realloc(buffer, section.sh_size);
+                    reader = f.reader(io, buffer);
+                }
+                const section_slice = reader.interface.take(section.sh_size) catch |e| blk: {
+                    switch (e) {
+                        error.EndOfStream => {
+                            break :blk null;
+                        },
+                        error.ReadFailed => unreachable,
+                    }
+                };
+                if (section_slice != null and !options.skip_sections_content) {
+                    try dumpHexFallible(u64, bw, term, section_slice.?, section.sh_addr);
+                }
+            }
+        }
+    }
+}
+
+fn allocComment(
+    gpa: std.mem.Allocator,
+    code: []u8,
+    symbols: []SymbolRange,
+) !?[]u8 {
+    var iter = std.mem.splitAny(u8, code, " \t[],+-");
+    while (iter.next()) |s| {
+        if (std.mem.startsWith(u8, s, "0x")) {
+            // todo split at the zero char at the end of string
+            const v = std.fmt.parseInt(u64, std.mem.sliceTo(s[2..], 0), 16) catch |e| blk: {
+                std.debug.print("{any}\n", .{e});
+                std.debug.dumpHex(s);
+                break :blk 0;
+            };
+            // FIXME: this algorithm isn't working to find addresses "inside" symbols
+            if (v > 0) {
+                const idx = std.sort.lowerBound(SymbolRange, symbols, v, struct {
+                    fn inner(a: u64, sym: SymbolRange) std.math.Order {
+                        return std.math.order(a, sym.start);
+                    }
+                }.inner);
+                if (idx < symbols.len and v >= symbols[idx].start and v <= symbols[idx].end) {
+                    // if (idx > 0)
+                    //     idx -= 1;
+                    const d = v - symbols[idx].start;
+                    if (d > 0)
+                        return try std.fmt.allocPrint(gpa, "{s}+0x{x}", .{ symbols[idx].name, d });
+                    return try std.fmt.allocPrint(gpa, "{s}", .{symbols[idx].name});
+                }
+            }
+        }
+    }
+
+    return null;
+}
+
+fn dumpInstr(
+    gpa: std.mem.Allocator,
+    bw: *std.Io.Writer,
+    term: std.Io.Terminal,
+    instrs: []cs.cs_insn,
+    symbols: []SymbolRange,
+) !void {
+    for (instrs) |instr| {
+        const addr = instr.address;
+        const idx = std.sort.lowerBound(SymbolRange, symbols, addr, struct {
+            fn inner(a: u64, sym: SymbolRange) std.math.Order {
+                return std.math.order(a, sym.start);
+            }
+        }.inner);
+
+        if (idx < symbols.len and symbols[idx].start == addr and symbols[idx].name.len > 0) {
+            try term.setColor(.blue);
+            try bw.print("\n{x:0>16}  {s}:\n", .{
+                addr,
+                symbols[idx].name,
+            });
+            try term.setColor(.reset);
+        }
+        try term.setColor(.dim);
+        try bw.print("{x:0>[1]}  ", .{ addr, @sizeOf(usize) * 2 });
+        try term.setColor(.reset);
+        // if(instr.detail.)
+        try term.setColor(.bright_green);
+        try bw.print("{s} ", .{instr.mnemonic});
+        try term.setColor(.reset);
+
+        const mnemonic_strlen: u64 = @intCast(std.mem.find(u8, &instr.mnemonic, &.{0}).?);
+        const mnemonic_pad: u64 = 10;
+        for (0..(mnemonic_pad - mnemonic_strlen)) |_| {
+            try bw.printAsciiChar(' ', .{});
+        }
+        try bw.print("{s}", .{instr.op_str});
+        const asm_comment = try allocComment(gpa, @ptrCast(@constCast(&instr.op_str)), symbols);
+        defer {
+            if (asm_comment != null)
+                gpa.free(asm_comment.?);
+        }
+        if (asm_comment != null and asm_comment.?.len > 0) {
+            try term.setColor(.blue);
+            try bw.print(" <{s}>", .{asm_comment.?});
+        }
+        try bw.print("\n", .{});
+        try term.setColor(.reset);
+    }
+}
+
+/// Prints a hexadecimal view of the bytes, returning any error that occurs.
+pub fn dumpHexFallible(
+    _: type,
+    bw: *std.Io.Writer,
+    term: std.Io.Terminal,
+    bytes: []const u8,
+    offset: u64,
+) !void {
+    // @breakpoint();
+    const nbytes = 16;
+    var chunks = std.mem.window(u8, @ptrCast(@alignCast(bytes)), nbytes, nbytes);
+    while (chunks.next()) |window| {
+        // 1. Print the address.
+        const address = ((0x10 * (std.math.divCeil(usize, chunks.index orelse bytes.len, nbytes) catch unreachable)) - 0x10) + offset;
+        try term.setColor(.dim);
+        // We print the address in lowercase and the bytes in uppercase hexadecimal to distinguish them more.
+        // Also, make sure all lines are aligned by padding the address.
+        try bw.print("{x:0>[1]}  ", .{ address, @sizeOf(usize) * 2 });
+        try term.setColor(.reset);
+
+        // 2. Print the bytes.
+        for (window, 0..) |byte, index| {
+            try bw.print("{X:0>2} ", .{byte});
+            if (index == 7) try bw.writeByte(' ');
+        }
+        try bw.writeByte(' ');
+        if (window.len < 16) {
+            var missing_columns = (16 - window.len) * 3;
+            if (window.len < 8) missing_columns += 1;
+            try bw.splatByteAll(' ', missing_columns);
+        }
+
+        const window_bytes: []const u8 = @ptrCast(@alignCast(window));
+
+        // 3. Print the characters.
+        for (window_bytes) |byte| {
+            if (std.ascii.isPrint(byte)) {
+                try bw.writeByte(byte);
+            } else {
+
+                // Let's print some common control codes as graphical Unicode symbols.
+                // We don't want to do this for all control codes because most control codes apart from
+                // the ones that Zig has escape sequences for are likely not very useful to print as symbols.
+                switch (byte) {
+                    '\n' => try bw.writeAll("␊"),
+                    '\r' => try bw.writeAll("␍"),
+                    '\t' => try bw.writeAll("␉"),
+                    else => try bw.writeByte('.'),
+                }
+            }
+        }
+        try bw.writeByte('\n');
+    }
+}
+
+const SymbolRange = struct {
+    start: u64,
+    end: u64,
+    name: []u8,
+    kind: u8,
+};
+
+fn iterateSymbols(
+    h: std.elf.Header,
+    file_reader: *std.Io.File.Reader,
+    symtab: std.elf.Elf64_Shdr,
+) SymbolIterator {
+    return .{
+        .elf_header = h,
+        .file_reader = file_reader,
+        .symtab = symtab,
+    };
+}
+
+const SymbolIterator = struct {
+    elf_header: std.elf.Header,
+    file_reader: *std.Io.File.Reader,
+    symtab: std.elf.Elf64_Shdr,
+    index: usize = 0,
+
+    pub fn next(it: *SymbolIterator) !?std.elf.Elf64_Sym {
+        defer it.index += 1;
+
+        const size: u64 = if (it.elf_header.is_64) @sizeOf(std.elf.Elf64_Sym) else @sizeOf(std.elf.Elf64_Sym);
+        const offset = it.symtab.sh_offset + size * it.index;
+
+        if (offset >= (it.symtab.sh_size + it.symtab.sh_offset))
+            return null;
+
+        try it.file_reader.seekTo(offset);
+        return try it.file_reader.interface.takeStruct(std.elf.Elf64_Sym, it.elf_header.endian);
+    }
+};