Commit 5dd2bb525d1f19969c450c2b99a71f866a4f01ff | zig

author	Alex Rønne Petersen <alex@alexrp.com> 2024-08-17 11:42:32+02:00
committer	Andrew Kelley <andrew@ziglang.org> 2024-08-23 11:09:20-07:00
log	5dd2bb525d1f19969c450c2b99a71f866a4f01ff
tree	3dac4b3d3d5facfdb2c8998ff50a45bfa0015da2
parent	df6907f6019ec178735f06e1b55cef6a90234201

glibc: Define _IO_stdin_used in start code and reference it in stub asm.

This is necessary to inform the real, non-stub glibc that a program built with
Zig is using a modern `FILE` structure, i.e. glibc 2.1+. This is particularly
important on lesser-used architectures where the legacy code is poorly tested;
for example, glibc 2.40 introduced a regression for the legacy case in the
libio cleanup code, causing all Zig-compiled MIPS binaries to crash on exit.

2 files changed, 59 insertions(+), 1 deletions(-)

lib
- libc
  - glibc
    csu
    init.c
src
- glibc.zig

lib/libc/glibc/csu/init.c created+23

...	@@ -0,0 +1,23 @@
1	/* Special startup support.
2	Copyright (C) 1997-2024 Free Software Foundation, Inc.
3	This file is part of the GNU C Library.
4
5	The GNU C Library is free software; you can redistribute it and/or
6	modify it under the terms of the GNU Lesser General Public
7	License as published by the Free Software Foundation; either
8	version 2.1 of the License, or (at your option) any later version.
9
10	The GNU C Library is distributed in the hope that it will be useful,
11	but WITHOUT ANY WARRANTY; without even the implied warranty of
12	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13	Lesser General Public License for more details.
14
15	You should have received a copy of the GNU Lesser General Public
16	License along with the GNU C Library; if not, see
17	<https://www.gnu.org/licenses/>. */
18
19	/* Vestigial libio version number. Some code in libio checks whether
20	this symbol exists in the executable, but nothing looks at its
21	value anymore; the value it was historically set to has been
22	preserved out of an abundance of caution. */
23	const int _IO_stdin_used = 0x20001;

src/glibc.zig+36-1

...	@@ -286,7 +286,11 @@ pub fn buildCRTFile(comp: *Compilation, crt_file: CRTFile, prog_node: std.Progre	...	@@ -286,7 +286,11 @@ pub fn buildCRTFile(comp: *Compilation, crt_file: CRTFile, prog_node: std.Progre
286	.owner = undefined,	286	.owner = undefined,
287	};	287	};
288	};	288	};
289	var files = [_]Compilation.CSourceFile{ start_o, abi_note_o };	289	const init_o: Compilation.CSourceFile = .{
		290	.src_path = try lib_path(comp, arena, lib_libc_glibc ++ "csu" ++ path.sep_str ++ "init.c"),
		291	.owner = undefined,
		292	};
		293	var files = [_]Compilation.CSourceFile{ start_o, abi_note_o, init_o };
290	return comp.build_crt_file("Scrt1", .Obj, .@"glibc Scrt1.o", prog_node, &files);	294	return comp.build_crt_file("Scrt1", .Obj, .@"glibc Scrt1.o", prog_node, &files);
291	},	295	},
292	.libc_nonshared_a => {	296	.libc_nonshared_a => {
...	@@ -682,6 +686,12 @@ pub const BuiltSharedObjects = struct {	...	@@ -682,6 +686,12 @@ pub const BuiltSharedObjects = struct {
682		686
683	const all_map_basename = "all.map";	687	const all_map_basename = "all.map";
684		688
		689	fn wordDirective(target: std.Target) []const u8 {
		690	// Based on its description in the GNU `as` manual, you might assume that `.word` is sized
		691	// according to the target word size. But no; that would just make too much sense.
		692	return if (target.ptrBitWidth() == 64) ".quad" else ".long";
		693	}
		694
685	pub fn buildSharedObjects(comp: *Compilation, prog_node: std.Progress.Node) !void {	695	pub fn buildSharedObjects(comp: *Compilation, prog_node: std.Progress.Node) !void {
686	const tracy = trace(@src());	696	const tracy = trace(@src());
687	defer tracy.end();	697	defer tracy.end();
...	@@ -923,6 +933,31 @@ pub fn buildSharedObjects(comp: *Compilation, prog_node: std.Progress.Node) !voi	...	@@ -923,6 +933,31 @@ pub fn buildSharedObjects(comp: *Compilation, prog_node: std.Progress.Node) !voi
923		933
924	try stubs_asm.appendSlice(".data\n");	934	try stubs_asm.appendSlice(".data\n");
925		935
		936	// For some targets, the real `libc.so.6` will contain a weak reference to `_IO_stdin_used`,
		937	// making the linker put the symbol in the dynamic symbol table. We likewise need to emit a
		938	// reference to it here for that effect, or it will not show up, which in turn will cause
		939	// the real glibc to think that the program was built against an ancient `FILE` structure
		940	// (pre-glibc 2.1).
		941	//
		942	// Note that glibc only compiles in the legacy compatibility code for some targets; it
		943	// depends on what is defined in the `shlib-versions` file for the particular architecture
		944	// and ABI. Those files are preprocessed by 2 separate tools during the glibc build to get
		945	// the final `abi-versions.h`, so it would be quite brittle to try to condition our emission
		946	// of the `_IO_stdin_used` reference in the exact same way. The only downside of emitting
		947	// the reference unconditionally is that it ends up being unused for newer targets; it
		948	// otherwise has no negative effect.
		949	//
		950	// glibc uses a weak reference because it has to work with programs compiled against pre-2.1
		951	// versions where the symbol didn't exist. We only care about modern glibc versions, so use
		952	// a strong reference.
		953	if (std.mem.eql(u8, lib.name, "c")) {
		954	try stubs_asm.writer().print(
		955	\\.globl _IO_stdin_used
		956	\\{s} _IO_stdin_used
		957	\\
		958	, .{wordDirective(target)});
		959	}
		960
926	const obj_inclusions_len = mem.readInt(u16, metadata.inclusions[inc_i..][0..2], .little);	961	const obj_inclusions_len = mem.readInt(u16, metadata.inclusions[inc_i..][0..2], .little);
927	inc_i += 2;	962	inc_i += 2;
928		963