macho: skip GOT for TLVs; handle them separately when lowering

This commit is contained in:
Jakub Konka 2023-04-19 21:00:54 +02:00
parent 9530b95afe
commit f9e9974c8f
3 changed files with 126 additions and 181 deletions

View File

@ -7578,7 +7578,7 @@ fn genSetReg(self: *Self, ty: Type, reg: Register, mcv: MCValue) InnerError!void
const atom_index = try self.getSymbolIndexForDecl(self.mod_fn.owner_decl);
if (self.bin_file.cast(link.File.MachO)) |_| {
_ = try self.addInst(.{
.tag = .mov_linker,
.tag = .lea_linker,
.ops = .tlv_reloc,
.data = .{ .payload = try self.addExtra(Mir.LeaRegisterReloc{
.reg = @enumToInt(Register.rdi),

View File

@ -157,7 +157,6 @@ strtab: StringTable(.strtab) = .{},
got_table: TableSection(SymbolWithLoc) = .{},
stub_table: TableSection(SymbolWithLoc) = .{},
tlv_table: SectionTable = .{},
error_flags: File.ErrorFlags = File.ErrorFlags{},
@ -222,6 +221,10 @@ lazy_syms: LazySymbolTable = .{},
/// Table of tracked Decls.
decls: std.AutoArrayHashMapUnmanaged(Module.Decl.Index, DeclMetadata) = .{},
/// Table of threadlocal variables descriptors.
/// They are emitted in the `__thread_vars` section.
tlv_table: TlvSymbolTable = .{},
/// Hot-code swapping state.
hot_state: if (is_hot_update_compatible) HotUpdateState else struct {} = .{},
@ -238,6 +241,8 @@ const LazySymbolMetadata = struct {
alignment: u32,
};
const TlvSymbolTable = std.AutoArrayHashMapUnmanaged(SymbolWithLoc, Atom.Index);
const DeclMetadata = struct {
atom: Atom.Index,
section: u8,
@ -266,122 +271,6 @@ const DeclMetadata = struct {
}
};
const SectionTable = struct {
entries: std.ArrayListUnmanaged(Entry) = .{},
free_list: std.ArrayListUnmanaged(u32) = .{},
lookup: std.AutoHashMapUnmanaged(SymbolWithLoc, u32) = .{},
pub fn deinit(st: *ST, allocator: Allocator) void {
st.entries.deinit(allocator);
st.free_list.deinit(allocator);
st.lookup.deinit(allocator);
}
pub fn allocateEntry(st: *ST, allocator: Allocator, target: SymbolWithLoc) !u32 {
try st.entries.ensureUnusedCapacity(allocator, 1);
const index = blk: {
if (st.free_list.popOrNull()) |index| {
log.debug(" (reusing entry index {d})", .{index});
break :blk index;
} else {
log.debug(" (allocating entry at index {d})", .{st.entries.items.len});
const index = @intCast(u32, st.entries.items.len);
_ = st.entries.addOneAssumeCapacity();
break :blk index;
}
};
st.entries.items[index] = .{ .target = target, .sym_index = 0 };
try st.lookup.putNoClobber(allocator, target, index);
return index;
}
pub fn freeEntry(st: *ST, allocator: Allocator, target: SymbolWithLoc) void {
const index = st.lookup.get(target) orelse return;
st.free_list.append(allocator, index) catch {};
st.entries.items[index] = .{
.target = .{ .sym_index = 0 },
.sym_index = 0,
};
_ = st.lookup.remove(target);
}
pub fn getAtomIndex(st: *const ST, macho_file: *MachO, target: SymbolWithLoc) ?Atom.Index {
const index = st.lookup.get(target) orelse return null;
return st.entries.items[index].getAtomIndex(macho_file);
}
const FormatContext = struct {
macho_file: *MachO,
st: *const ST,
};
fn fmt(
ctx: FormatContext,
comptime unused_format_string: []const u8,
options: std.fmt.FormatOptions,
writer: anytype,
) @TypeOf(writer).Error!void {
_ = options;
comptime assert(unused_format_string.len == 0);
try writer.writeAll("SectionTable:\n");
for (ctx.st.entries.items, 0..) |entry, i| {
const atom_sym = entry.getSymbol(ctx.macho_file);
const target_sym = ctx.macho_file.getSymbol(entry.target);
try writer.print(" {d}@{x} => ", .{ i, atom_sym.n_value });
if (target_sym.undf()) {
try writer.print("import('{s}')", .{
ctx.macho_file.getSymbolName(entry.target),
});
} else {
try writer.print("local(%{d}) in object({?d})", .{
entry.target.sym_index,
entry.target.file,
});
}
try writer.writeByte('\n');
}
}
fn format(st: *const ST, comptime unused_format_string: []const u8, options: std.fmt.FormatOptions, writer: anytype) !void {
_ = st;
_ = unused_format_string;
_ = options;
_ = writer;
@compileError("do not format SectionTable directly; use st.fmtDebug()");
}
pub fn fmtDebug(st: *const ST, macho_file: *MachO) std.fmt.Formatter(fmt) {
return .{ .data = .{
.macho_file = macho_file,
.st = st,
} };
}
const ST = @This();
const Entry = struct {
target: SymbolWithLoc,
// Index into the synthetic symbol table (i.e., file == null).
sym_index: u32,
pub fn getSymbol(entry: Entry, macho_file: *MachO) macho.nlist_64 {
return macho_file.getSymbol(.{ .sym_index = entry.sym_index });
}
pub fn getSymbolPtr(entry: Entry, macho_file: *MachO) *macho.nlist_64 {
return macho_file.getSymbolPtr(.{ .sym_index = entry.sym_index });
}
pub fn getAtomIndex(entry: Entry, macho_file: *MachO) ?Atom.Index {
return macho_file.getAtomIndexForSymbol(.{ .sym_index = entry.sym_index });
}
pub fn getName(entry: Entry, macho_file: *MachO) []const u8 {
return macho_file.getSymbolName(.{ .sym_index = entry.sym_index });
}
};
};
const BindingTable = std.AutoArrayHashMapUnmanaged(Atom.Index, std.ArrayListUnmanaged(Atom.Binding));
const UnnamedConstTable = std.AutoArrayHashMapUnmanaged(Module.Decl.Index, std.ArrayListUnmanaged(Atom.Index));
const RebaseTable = std.AutoArrayHashMapUnmanaged(Atom.Index, std.ArrayListUnmanaged(u32));
@ -1520,18 +1409,13 @@ fn createDyldPrivateAtom(self: *MachO) !void {
try self.writeAtom(atom_index, &buffer);
}
fn createThreadLocalDescriptorAtom(self: *MachO, target: SymbolWithLoc) !Atom.Index {
fn createThreadLocalDescriptorAtom(self: *MachO, sym_name: []const u8, target: SymbolWithLoc) !Atom.Index {
const gpa = self.base.allocator;
const size = 3 * @sizeOf(u64);
const required_alignment: u32 = 1;
const atom_index = try self.createAtom();
self.getAtomPtr(atom_index).size = size;
const target_sym_name = self.getSymbolName(target);
const name_delimiter = mem.indexOf(u8, target_sym_name, "$").?;
const sym_name = try gpa.dupe(u8, target_sym_name[0..name_delimiter]);
defer gpa.free(sym_name);
const sym = self.getAtom(atom_index).getSymbolPtr(self);
sym.n_type = macho.N_SECT;
sym.n_sect = self.thread_vars_section_index.? + 1;
@ -1706,7 +1590,6 @@ pub fn deinit(self: *MachO) void {
self.got_table.deinit(gpa);
self.stub_table.deinit(gpa);
self.tlv_table.deinit(gpa);
self.strtab.deinit(gpa);
self.locals.deinit(gpa);
@ -1739,14 +1622,12 @@ pub fn deinit(self: *MachO) void {
self.atoms.deinit(gpa);
if (self.base.options.module) |_| {
for (self.decls.values()) |*m| {
m.exports.deinit(gpa);
}
self.decls.deinit(gpa);
} else {
assert(self.decls.count() == 0);
for (self.decls.values()) |*m| {
m.exports.deinit(gpa);
}
self.decls.deinit(gpa);
self.lazy_syms.deinit(gpa);
self.tlv_table.deinit(gpa);
for (self.unnamed_const_atoms.values()) |*atoms| {
atoms.deinit(gpa);
@ -1926,15 +1807,6 @@ fn addStubEntry(self: *MachO, target: SymbolWithLoc) !void {
self.stub_table_count_dirty = true;
}
fn addTlvEntry(self: *MachO, target: SymbolWithLoc) !void {
if (self.tlv_table.lookup.contains(target)) return;
const tlv_index = try self.tlv_table.allocateEntry(self.base.allocator, target);
const tlv_atom_index = try self.createThreadLocalDescriptorAtom(target);
const tlv_atom = self.getAtom(tlv_atom_index);
self.tlv_table.entries.items[tlv_index].sym_index = tlv_atom.getSymbolIndex().?;
self.markRelocsDirtyByTarget(target);
}
pub fn updateFunc(self: *MachO, module: *Module, func: *Module.Fn, air: Air, liveness: Liveness) !void {
if (build_options.skip_non_native and builtin.object_format != .macho) {
@panic("Attempted to compile for object format that was disabled by build configuration");
@ -2081,6 +1953,12 @@ pub fn updateDecl(self: *MachO, module: *Module, decl_index: Module.Decl.Index)
}
}
const is_threadlocal = if (decl.val.castTag(.variable)) |payload|
payload.data.is_threadlocal and !self.base.options.single_threaded
else
false;
if (is_threadlocal) return self.updateThreadlocalVariable(module, decl_index);
const atom_index = try self.getOrCreateAtomForDecl(decl_index);
const sym_index = self.getAtom(atom_index).getSymbolIndex().?;
Atom.freeRelocations(self, atom_index);
@ -2229,6 +2107,101 @@ pub fn getOrCreateAtomForLazySymbol(self: *MachO, sym: File.LazySymbol, alignmen
return atom.*.?;
}
fn updateThreadlocalVariable(self: *MachO, module: *Module, decl_index: Module.Decl.Index) !void {
// Lowering a TLV on macOS involves two stages:
// 1. first we lower the initializer into appopriate section (__thread_data or __thread_bss)
// 2. next, we create a corresponding threadlocal variable descriptor in __thread_vars
// 1. Lower the initializer value.
const init_atom_index = try self.getOrCreateAtomForDecl(decl_index);
const init_atom = self.getAtomPtr(init_atom_index);
const init_sym_index = init_atom.getSymbolIndex().?;
Atom.freeRelocations(self, init_atom_index);
const gpa = self.base.allocator;
var code_buffer = std.ArrayList(u8).init(gpa);
defer code_buffer.deinit();
var decl_state: ?Dwarf.DeclState = if (self.d_sym) |*d_sym|
try d_sym.dwarf.initDeclState(module, decl_index)
else
null;
defer if (decl_state) |*ds| ds.deinit();
const decl = module.declPtr(decl_index);
const decl_metadata = self.decls.get(decl_index).?;
const decl_val = decl.val.castTag(.variable).?.data.init;
const res = if (decl_state) |*ds|
try codegen.generateSymbol(&self.base, decl.srcLoc(), .{
.ty = decl.ty,
.val = decl_val,
}, &code_buffer, .{
.dwarf = ds,
}, .{
.parent_atom_index = init_sym_index,
})
else
try codegen.generateSymbol(&self.base, decl.srcLoc(), .{
.ty = decl.ty,
.val = decl_val,
}, &code_buffer, .none, .{
.parent_atom_index = init_sym_index,
});
var code = switch (res) {
.ok => code_buffer.items,
.fail => |em| {
decl.analysis = .codegen_failure;
try module.failed_decls.put(module.gpa, decl_index, em);
return;
},
};
const required_alignment = decl.getAlignment(self.base.options.target);
const decl_name = try decl.getFullyQualifiedName(module);
defer gpa.free(decl_name);
const init_sym_name = try std.fmt.allocPrint(gpa, "{s}$tlv$init", .{decl_name});
defer gpa.free(init_sym_name);
const sect_id = decl_metadata.section;
const init_sym = init_atom.getSymbolPtr(self);
init_sym.n_strx = try self.strtab.insert(gpa, init_sym_name);
init_sym.n_type = macho.N_SECT;
init_sym.n_sect = sect_id + 1;
init_sym.n_desc = 0;
init_atom.size = code.len;
init_sym.n_value = try self.allocateAtom(init_atom_index, code.len, required_alignment);
errdefer self.freeAtom(init_atom_index);
log.debug("allocated atom for {s} at 0x{x}", .{ init_sym_name, init_sym.n_value });
log.debug(" (required alignment 0x{x})", .{required_alignment});
try self.writeAtom(init_atom_index, code);
if (decl_state) |*ds| {
try self.d_sym.?.dwarf.commitDeclState(
module,
decl_index,
init_sym.n_value,
self.getAtom(init_atom_index).size,
ds,
);
}
try self.updateDeclExports(module, decl_index, module.getDeclExports(decl_index));
// 2. Create a TLV descriptor.
const init_atom_sym_loc = init_atom.getSymbolWithLoc();
const gop = try self.tlv_table.getOrPut(gpa, init_atom_sym_loc);
assert(!gop.found_existing);
gop.value_ptr.* = try self.createThreadLocalDescriptorAtom(decl_name, init_atom_sym_loc);
self.markRelocsDirtyByTarget(init_atom_sym_loc);
}
pub fn getOrCreateAtomForDecl(self: *MachO, decl_index: Module.Decl.Index) !Atom.Index {
const gop = try self.decls.getOrPut(self.base.allocator, decl_index);
if (!gop.found_existing) {
@ -2296,21 +2269,11 @@ fn updateDeclCode(self: *MachO, decl_index: Module.Decl.Index, code: []u8) !u64
const sect_id = decl_metadata.section;
const header = &self.sections.items(.header)[sect_id];
const segment = self.getSegment(sect_id);
const is_threadlocal = if (!self.base.options.single_threaded)
header.flags == macho.S_THREAD_LOCAL_REGULAR or header.flags == macho.S_THREAD_LOCAL_ZEROFILL
else
false;
const code_len = code.len;
const sym_name = if (is_threadlocal)
try std.fmt.allocPrint(gpa, "{s}$tlv$init", .{decl_name})
else
decl_name;
defer if (is_threadlocal) gpa.free(sym_name);
if (atom.size != 0) {
const sym = atom.getSymbolPtr(self);
sym.n_strx = try self.strtab.insert(gpa, sym_name);
sym.n_strx = try self.strtab.insert(gpa, decl_name);
sym.n_type = macho.N_SECT;
sym.n_sect = sect_id + 1;
sym.n_desc = 0;
@ -2320,21 +2283,13 @@ fn updateDeclCode(self: *MachO, decl_index: Module.Decl.Index, code: []u8) !u64
if (need_realloc) {
const vaddr = try self.growAtom(atom_index, code_len, required_alignment);
log.debug("growing {s} and moving from 0x{x} to 0x{x}", .{ sym_name, sym.n_value, vaddr });
log.debug("growing {s} and moving from 0x{x} to 0x{x}", .{ decl_name, sym.n_value, vaddr });
log.debug(" (required alignment 0x{x})", .{required_alignment});
if (vaddr != sym.n_value) {
sym.n_value = vaddr;
// TODO: I think we should update the offset to the initializer here too.
const target: SymbolWithLoc = if (is_threadlocal) blk: {
const tlv_atom_index = self.tlv_table.getAtomIndex(self, .{
.sym_index = sym_index,
}).?;
const tlv_atom = self.getAtom(tlv_atom_index);
break :blk tlv_atom.getSymbolWithLoc();
} else .{ .sym_index = sym_index };
log.debug(" (updating GOT entry)", .{});
const got_atom_index = self.got_table.lookup.get(target).?;
const got_atom_index = self.got_table.lookup.get(.{ .sym_index = sym_index }).?;
try self.writeOffsetTableEntry(got_atom_index);
}
} else if (code_len < atom.size) {
@ -2346,7 +2301,7 @@ fn updateDeclCode(self: *MachO, decl_index: Module.Decl.Index, code: []u8) !u64
self.getAtomPtr(atom_index).size = code_len;
} else {
const sym = atom.getSymbolPtr(self);
sym.n_strx = try self.strtab.insert(gpa, sym_name);
sym.n_strx = try self.strtab.insert(gpa, decl_name);
sym.n_type = macho.N_SECT;
sym.n_sect = sect_id + 1;
sym.n_desc = 0;
@ -2354,21 +2309,13 @@ fn updateDeclCode(self: *MachO, decl_index: Module.Decl.Index, code: []u8) !u64
const vaddr = try self.allocateAtom(atom_index, code_len, required_alignment);
errdefer self.freeAtom(atom_index);
log.debug("allocated atom for {s} at 0x{x}", .{ sym_name, vaddr });
log.debug("allocated atom for {s} at 0x{x}", .{ decl_name, vaddr });
log.debug(" (required alignment 0x{x})", .{required_alignment});
self.getAtomPtr(atom_index).size = code_len;
sym.n_value = vaddr;
if (is_threadlocal) {
try self.addTlvEntry(.{ .sym_index = sym_index });
}
const target: SymbolWithLoc = if (is_threadlocal) blk: {
const tlv_atom_index = self.tlv_table.getAtomIndex(self, .{ .sym_index = sym_index }).?;
const tlv_atom = self.getAtom(tlv_atom_index);
break :blk tlv_atom.getSymbolWithLoc();
} else .{ .sym_index = sym_index };
try self.addGotEntry(target);
try self.addGotEntry(.{ .sym_index = sym_index });
}
try self.writeAtom(atom_index, code);
@ -4169,8 +4116,8 @@ pub fn logSymtab(self: *MachO) void {
log.debug("stubs entries:", .{});
log.debug("{}", .{self.stub_table});
log.debug("threadlocal entries:", .{});
log.debug("{}", .{self.tlv_table.fmtDebug(self)});
// log.debug("threadlocal entries:", .{});
// log.debug("{}", .{self.tlv_table});
}
pub fn logAtoms(self: *MachO) void {

View File

@ -15,7 +15,7 @@ pub const Type = enum {
got,
/// RIP-relative displacement
signed,
/// RIP-relative displacement to GOT pointer to TLV thunk
/// RIP-relative displacement to a TLV thunk
tlv,
// aarch64
@ -52,11 +52,9 @@ pub fn getTargetBaseAddress(self: Relocation, macho_file: *MachO) ?u64 {
return header.addr + got_index * @sizeOf(u64);
},
.tlv => {
const thunk_atom_index = macho_file.tlv_table.getAtomIndex(macho_file, self.target) orelse return null;
const thunk_atom = macho_file.getAtom(thunk_atom_index);
const got_index = macho_file.got_table.lookup.get(thunk_atom.getSymbolWithLoc()) orelse return null;
const header = macho_file.sections.items(.header)[macho_file.got_section_index.?];
return header.addr + got_index * @sizeOf(u64);
const atom_index = macho_file.tlv_table.get(self.target) orelse return null;
const atom = macho_file.getAtom(atom_index);
return atom.getSymbol(macho_file).n_value;
},
.branch => {
if (macho_file.stub_table.lookup.get(self.target)) |index| {