[Tarantool-patches] [PATCH luajit 2/2] OSX/iOS: Always generate 64 bit non-FAT Mach-O object files.
Sergey Bronnikov
sergeyb at tarantool.org
Wed Jul 10 15:43:54 MSK 2024
Sergey,
thanks for review. Fixes applied and force-pushed to the same branch.
Sergey
On 09.07.2024 16:03, Sergey Kaplun via Tarantool-patches wrote:
> Hi, Sergey!
> Thanks for the patch!
> Please consider my comments below.
>
> May we add the test [1] to verify that there will be no regression in the
> future?
This "test" was for a FAT binary generated by LuaJIT.
With backported patches, FAT (aka Universal Binary) support is gone.
mach-O header is checked by tests in proposed patch.
>
> On 05.07.24, Sergey Bronnikov wrote:
>> Reported by Sergey Bronnikov.
>>
>> (cherry picked from commit 7110b935672489afd6ba3eef3e5139d2f3bd05b6)
>>
>> Previously, LuaJIT generated Mach-O FAT object files for ARM and
>> ARM64 on macOS. The patch removes support of 32-bit ARM and
>> FAT object files and now LuaJIT generate Mach-O object files for
>> ARM64.
> I suppose we should mention that no x86/x86_64 objects are generated
> now.
x86_64 is still there:
$ ./build/gc64/src/luajit -b -o osx -a arm64 empty.lua empty.o
$ file empty.o
empty.o: Mach-O 64-bit arm64 object
$ ./build/gc64/src/luajit -b -o osx -a x64 empty.lua empty.o
$ file empty.o
empty.o: Mach-O 64-bit x86_64 object
$
Anyway, commit message has been updated.
>> Sergey Bronnikov:
>> * added the description and the trimmed the test for the problem
>>
>> Part of tarantool/tarantool#10199
>> ---
>> src/jit/bcsave.lua | 155 ++-------
>> ...-865-cross-generation-mach-o-file.test.lua | 294 +++---------------
>> 2 files changed, 70 insertions(+), 379 deletions(-)
>>
>> diff --git a/src/jit/bcsave.lua b/src/jit/bcsave.lua
>> index 26ec29c6..61953c2d 100644
>> --- a/src/jit/bcsave.lua
>> +++ b/src/jit/bcsave.lua
> <snipped>
>
>> diff --git a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
>> index f008f3bd..6a58de95 100644
>> --- a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
>> +++ b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
>> @@ -3,109 +3,11 @@ local test = tap.test('lj-865-cross-generation-mach-o-file')
>> local utils = require('utils')
>> local ffi = require('ffi')
>>
>> -test:plan(2)
>> +test:plan(1)
>>
>> -- The test creates an object file in Mach-O format with LuaJIT
>> -- bytecode and checks the validity of the object file fields.
>> ---
>> --- The original problem is reproduced with LuaJIT, which is built
>> --- with enabled AVX512F instructions. The support for AVX512F
>> --- could be checked in `/proc/cpuinfo` on Linux and
>> --- `sysctl hw.optional.avx512f` on Mac. AVX512F must be
>> --- implicitly enabled in a C compiler by passing a CPU codename.
>> --- Please take a look at the GCC Online Documentation [1] for
>> --- available CPU codenames. Also, see the Wikipedia for CPUs with
>> --- AVX-512 support [2].
>> --- Execute command below to detect the CPU codename:
>> --- `gcc -march=native -Q --help=target | grep march`.
>> ---
>> --- 1.https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html
>> --- 2.https://en.wikipedia.org/wiki/AVX-512#CPUs_with_AVX-512
>> ---
>> --- Manual steps for reproducing are the following:
>> ---
>> --- $ CC=gcc TARGET_CFLAGS='skylake-avx512' cmake -S . -B build
>> --- $ cmake --build build --parallel
>> --- $ echo > test.lua
>> --- $ LUA_PATH="src/?.lua;;" luajit -b -o osx -a arm test.lua test.o
>> --- $ file test.o
>> --- empty.o: DOS executable (block device driver)
>>
>> --- LuaJIT can generate so called Universal Binary with Lua
>> --- bytecode. The Universal Binary format is a format for
>> --- executable files that run natively on hardware platforms with
>> --- different hardware architectures. This concept is more
>> --- generally known as a fat binary.
>> ---
>> --- The format of the Mach-O is described in the document
>> --- "OS X ABI Mach-O File Format Reference", published by Apple
>> --- company. The copy of the (now removed) official documentation
>> --- can be found here [1]. Yet another source of truth is
>> --- XNU headers, see the definition of C-structures in:
>> --- [2] (`nlist_64`), [3] (`fat_arch` and `fat_header`).
>> ---
>> --- There is a good visual representation of Universal Binary
>> --- in "Mac OS X Internals" book (pages 67-68) [5] and in the [6].
>> --- Below is the schematic structure of Universal Binary, which
>> --- includes two executables for PowerPC and Intel i386 (omitted):
>> ---
>> --- 0x0000000 ---------------------------------------
>> --- |
>> --- struct | 0xcafebabe FAT_MAGIC magic
>> --- fat_header | -------------------------------------
>> --- | 0x00000003 nfat_arch
>> --- ---------------------------------------
>> --- | 0x00000012 CPU_TYPE_POWERPC cputype
>> --- | -------------------------------------
>> --- | 0x00000000 CPU_SUBTYPE_POWERPC_ALL cpusubtype
>> --- struct | -------------------------------------
>> --- fat_arch | 0x00001000 4096 bytes offset
>> --- | -------------------------------------
>> --- | 0x00004224 16932 bytes size
>> --- | -------------------------------------
>> --- | 0x0000000c 2^12 = 4096 bytes align
>> --- ---------------------------------------
>> --- ---------------------------------------
>> --- | 0x00000007 CPU_TYPE_I386 cputype
>> --- | -------------------------------------
>> --- | 0x00000003 CPU_SUBTYPE_I386_ALL cpusubtype
>> --- struct | -------------------------------------
>> --- fat_arch | 0x00006000 24576 bytes offset
>> --- | -------------------------------------
>> --- | 0x0000292c 10540 bytes size
>> --- | -------------------------------------
>> --- | 0x0000000c 2^12 = 4096 bytes align
>> --- ---------------------------------------
>> --- Unused
>> --- 0x00001000 ---------------------------------------
>> --- | 0xfeedface MH_MAGIC magic
>> --- | ------------------------------------
>> --- | 0x00000012 CPU_TYPE_POWERPC cputype
>> --- | ------------------------------------
>> --- struct | 0x00000000 CPU_SUBTYPE_POWERPC_ALL cpusubtype
>> --- mach_header | ------------------------------------
>> --- | 0x00000002 MH_EXECUTE filetype
>> --- | ------------------------------------
>> --- | 0x0000000b 10 load commands ncmds
>> --- | ------------------------------------
>> --- | 0x00000574 1396 bytes sizeofcmds
>> --- | ------------------------------------
>> --- | 0x00000085 DYLDLINK TWOLEVEL flags
>> --- --------------------------------------
>> --- Load commands
>> --- ---------------------------------------
>> --- Data
>> --- ---------------------------------------
>> ---
>> --- < x86 executable >
>> ---
>> --- 1.https://github.com/aidansteele/osx-abi-macho-file-format-reference
>> --- 2.https://github.com/apple-oss-distributions/xnu/blob/xnu-10002.1.13/EXTERNAL_HEADERS/mach-o/nlist.h
>> --- 3.https://github.com/apple-oss-distributions/xnu/blob/xnu-10002.1.13/EXTERNAL_HEADERS/mach-o/fat.h
>> --- 4.https://developer.apple.com/documentation/apple-silicon/addressing-architectural-differences-in-your-macos-code
>> --- 5.https://reverseengineering.stackexchange.com/a/6357/46029
>> --- 6.http://formats.kaitai.io/mach_o/index.html
> I prefer to keep the non-FAT part of this comment since it is very
> useful.
reverted:
diff --git a/build/src/luajit b/build/src/luajit
deleted file mode 100755
index bd10e09a..00000000
Binary files a/build/src/luajit and /dev/null differ
diff --git
a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
index 6a58de95..f9ca6e38 100644
--- a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
+++ b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
@@ -68,6 +68,62 @@ end
-- Parses a buffer in the Mach-O format and returns its fields
-- in a table.
+-- The test creates an object file in Mach-O format with LuaJIT
+-- bytecode and checks the validity of the object file fields.
+---
+--- The original problem is reproduced with LuaJIT, which is built
+--- with enabled AVX512F instructions. The support for AVX512F
+--- could be checked in `/proc/cpuinfo` on Linux and
+--- `sysctl hw.optional.avx512f` on Mac. AVX512F must be
+--- implicitly enabled in a C compiler by passing a CPU codename.
+--- Please take a look at the GCC Online Documentation [1] for
+--- available CPU codenames. Also, see the Wikipedia for CPUs with
+--- AVX-512 support [2].
+--- Execute command below to detect the CPU codename:
+--- `gcc -march=native -Q --help=target | grep march`.
+---
+--- 1. https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html
+--- 2. https://en.wikipedia.org/wiki/AVX-512#CPUs_with_AVX-512
+---
+--- Manual steps for reproducing are the following:
+---
+--- $ CC=gcc TARGET_CFLAGS='skylake-avx512' cmake -S . -B build
+--- $ cmake --build build --parallel
+--- $ echo > test.lua
+--- $ LUA_PATH="src/?.lua;;" luajit -b -o osx -a arm test.lua test.o
+--- $ file test.o
+--- empty.o: DOS executable (block device driver)
+
+--- The format of the Mach-O is described in the document
+--- "OS X ABI Mach-O File Format Reference", published by Apple
+--- company. The copy of the (now removed) official documentation
+--- can be found here [1]. There is a good visual representation
+--- of Mach-O format in "Mac OS X Internals" book (pages 67-68)
+--- [2] and in the [3].
+--
+--- 0x0000000 ---------------------------------------
+--- | 0xfeedface MH_MAGIC magic
+--- | ------------------------------------
+--- | 0x00000012 CPU_TYPE_POWERPC cputype
+--- | ------------------------------------
+--- struct | 0x00000000 CPU_SUBTYPE_POWERPC_ALL cpusubtype
+--- mach_header | ------------------------------------
+--- | 0x00000002 MH_EXECUTE filetype
+--- | ------------------------------------
+--- | 0x0000000b 10 load commands ncmds
+--- | ------------------------------------
+--- | 0x00000574 1396 bytes sizeofcmds
+--- | ------------------------------------
+--- | 0x00000085 DYLDLINK TWOLEVEL flags
+--- --------------------------------------
+--- Load commands
+--- ---------------------------------------
+--- Data
+--- ---------------------------------------
+---
+--- 1. https://github.com/aidansteele/osx-abi-macho-file-format-reference
+--- 2. https://reverseengineering.stackexchange.com/a/6357/46029
+--- 3. http://formats.kaitai.io/mach_o/index.html
local function read_mach_o(buf, hw_arch)
local is64 = hw_arch == 'arm64'
>
>> ---
>> -- Using the same declarations as defined in <src/jit/bcsave.lua>.
>> ffi.cdef[[
> <snipped>
>
>> local function create_obj_file(name, arch)
>> @@ -212,108 +66,37 @@ local function create_obj_file(name, arch)
>> return mach_o_path
>> end
>>
>> --- Parses a buffer in the Mach-O format and returns the FAT magic
>> --- number and `nfat_arch`.
>> +-- Parses a buffer in the Mach-O format and returns its fields
>> +-- in a table.
>> local function read_mach_o(buf, hw_arch)
> I suggest renaming it to `read_mach_o_hdr()` and returning only the
> header without any additional wrapping in the table.
Updated:
diff --git a/build/src/luajit b/build/src/luajit
deleted file mode 100755
index bd10e09a..00000000
Binary files a/build/src/luajit and /dev/null differ
diff --git
a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
index 6a58de95..0f269a8c 100644
--- a/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
+++ b/test/tarantool-tests/lj-865-cross-generation-mach-o-file.test.lua
@@ -78,15 +134,7 @@ local function read_mach_o(buf, hw_arch)
-- Mach-O object header.
local mach_header = obj.hdr
- return {
- header = {
- magic = mach_header.magic,
- cputype = mach_header.cputype,
- cpusubtype = mach_header.cpusubtype,
- filetype = mach_header.filetype,
- ncmds = mach_header.ncmds,
- },
- }
+ return mach_header
end
-- The function builds Mach-O object file and retrieves
@@ -104,22 +152,22 @@ local function build_and_check_mach_o(subtest)
local mach_o_buf = utils.tools.read_file(mach_o_obj_path)
assert(mach_o_buf ~= nil and #mach_o_buf ~= 0, 'cannot read an
object file')
- local mach_o = read_mach_o(mach_o_buf, hw_arch)
+ local mach_o = read_mach_o_hdr(mach_o_buf, hw_arch)
-- Teardown.
assert(os.remove(mach_o_obj_path), 'remove an object file')
- local magic_str = string.format('%#x', mach_o.header.magic)
+ local magic_str = string.format('%#x', mach_o.magic)
subtest:is(magic_str, '0xfeedfacf',
'magic is correct in Mach-O')
- local cputype_str = string.format('%#x', mach_o.header.cputype)
+ local cputype_str = string.format('%#x', mach_o.cputype)
subtest:is(cputype_str, '0x100000c',
'cputype is correct in Mach-O')
- subtest:is(mach_o.header.cpusubtype, 0,
+ subtest:is(mach_o.cpusubtype, 0,
'cpusubtype is correct in Mach-O')
- subtest:is(mach_o.header.filetype, 1,
+ subtest:is(mach_o.filetype, 1,
'filetype is correct in Mach-O')
- subtest:is(mach_o.header.ncmds, 2,
+ subtest:is(mach_o.ncmds, 2,
'ncmds is correct in Mach-O')
end
>> - local res = {
>> - header = {
>> - magic = 0,
>> - nfat_arch = 0,
>> - },
>> - fat_arch = {},
>> - }
>> -
>> local is64 = hw_arch == 'arm64'
> Maybe it is better to use assert here, like we do for the
> `build_and_check_mach_o()` routine?
Added, but read_mach_o_hdr is called by build_and_check_mach_o where
assert already added.
>
>>
>> - -- Mach-O FAT object.
>> - local mach_fat_obj_type = ffi.typeof(is64 and
>> - 'mach_fat_obj_64 *' or
>> - 'mach_fat_obj *')
>> - local obj = ffi.cast(mach_fat_obj_type, buf)
>> + -- Mach-O object.
>> + local mach_obj_type = ffi.typeof(is64 and 'mach_obj_64 *')
> Maybe just use mach_obj_64 since there is no alternative?
ok
>> + local obj = ffi.cast(mach_obj_type, buf)
>>
>> - -- Mach-O FAT object header.
>> - local mach_fat_header = obj.fat
>> - -- Mach-O FAT is BE, target arch is LE.
>> - local be32 = bit.bswap
>> - res.header.magic = be32(mach_fat_header.magic)
>> - res.header.nfat_arch = be32(mach_fat_header.nfat_arch)
>> + -- Mach-O object header.
>> + local mach_header = obj.hdr
>>
>> - -- Mach-O FAT object arches.
>> - for i = 0, res.header.nfat_arch - 1 do
>> - local fat_arch = obj.fat_arch[i]
>> - local arch = {
>> - cputype = be32(fat_arch.cputype),
>> - cpusubtype = be32(fat_arch.cpusubtype),
>> - }
>> - table.insert(res.fat_arch, arch)
>> - end
>> -
>> - return res
>> + return {
>> + header = {
>> + magic = mach_header.magic,
>> + cputype = mach_header.cputype,
>> + cpusubtype = mach_header.cpusubtype,
>> + filetype = mach_header.filetype,
>> + ncmds = mach_header.ncmds,
>> + },
>> + }
>> end
>>
>> --- Universal Binary can contain executables for more than one
>> --- CPU architecture. For simplicity, the test compares the *sum*
>> --- of CPU types and CPU subtypes.
>> ---
>> --- <src/jit/bcsave.lua:bcsave_machobj> has the definitions of the
>> --- numbers below. The original XNU source code may be found in
>> --- <osfmk/mach/machine.h> [1].
>> ---
>> --- 1.https://opensource.apple.com/source/xnu/xnu-4570.41.2/osfmk/mach/machine.h.auto.html
>> ---
>> -local SUM_CPUTYPE = {
>> - -- x86 + arm.
>> - arm = 7 + 12,
>> - -- x64 + arm64.
>> - arm64 = 0x01000007 + 0x0100000c,
>> -}
>> -local SUM_CPUSUBTYPE = {
>> - -- x86 + arm.
>> - arm = 3 + 9,
>> - -- x64 + arm64.
>> - arm64 = 3 + 0,
>> -}
>> -
>> --- The function builds Mach-O FAT object file and retrieves
>> --- its header fields (magic and nfat_arch) and fields of each arch
>> --- (cputype, cpusubtype).
>> ---
>> --- The Mach-O FAT object header can be retrieved with `otool` on
>> --- macOS:
>> ---
>> --- $ otool -f empty.o
>> --- Fat headers
>> --- fat_magic 0xcafebabe
>> --- nfat_arch 2
>> --- <snipped>
>> ---
>> --- CPU type and subtype can be retrieved with `lipo` on macOS:
>> ---
>> --- $ luajit -b -o osx -a arm empty.lua empty.o
>> --- $ lipo -archs empty.o
>> --- i386 armv7
>> --- $ luajit -b -o osx -a arm64 empty.lua empty.o
>> --- $ lipo -archs empty.o
>> --- x86_64 arm64
>> +-- The function builds Mach-O object file and retrieves
>> +-- its header fields.
>> local function build_and_check_mach_o(subtest)
>> local hw_arch = subtest.name
>> - assert(hw_arch == 'arm' or hw_arch == 'arm64')
>> + -- LuaJIT always generate 64-bit non-FAT Mach-O object files.
> The fact that these files are generated only for M1 CPUs looks worth
> mentioning.
>
>> + assert(hw_arch == 'arm64')
>>
>> - subtest:plan(4)
>> - -- FAT_MAGIC is an integer containing the value 0xCAFEBABE in
>> - -- big-endian byte order format. On a big-endian host CPU,
>> - -- this can be validated using the constant FAT_MAGIC;
>> - -- on a little-endian host CPU, it can be validated using
>> - -- the constant FAT_CIGAM.
>> - --
>> - -- FAT_NARCH is an integer specifying the number of fat_arch
>> - -- data structures that follow. This is the number of
>> - -- architectures contained in this binary.
>> - --
>> - -- See the aforementioned "OS X ABI Mach-O File Format
>> - -- Reference".
> The similar comment about Mach-O format will be appretiated.
mach-o header struct described in a comment above
>
>> - local FAT_MAGIC = '0xffffffffcafebabe'
>> - local FAT_NARCH = 2
>> + subtest:plan(5)
>>
>> local MODULE_NAME = 'lango_team'
>>
>> @@ -327,24 +110,19 @@ local function build_and_check_mach_o(subtest)
>> assert(os.remove(mach_o_obj_path), 'remove an object file')
>>
>> local magic_str = string.format('%#x', mach_o.header.magic)
>> - subtest:is(magic_str, FAT_MAGIC,
>> - 'fat_magic is correct in Mach-O')
>> - subtest:is(mach_o.header.nfat_arch, FAT_NARCH,
>> - 'nfat_arch is correct in Mach-O')
>> -
>> - local total_cputype = 0
>> - local total_cpusubtype = 0
>> - for i = 1, FAT_NARCH do
>> - total_cputype = total_cputype + mach_o.fat_arch[i].cputype
>> - total_cpusubtype = total_cpusubtype + mach_o.fat_arch[i].cpusubtype
>> - end
>> - subtest:is(total_cputype, SUM_CPUTYPE[hw_arch],
>> + subtest:is(magic_str, '0xfeedfacf',
> Please use MH_MAGIC_64 named constant for this magic string.
Added
>
>> + 'magic is correct in Mach-O')
> Looks like this line may be joined with the previous one.
not enough free space for this
>> + local cputype_str = string.format('%#x', mach_o.header.cputype)
>> + subtest:is(cputype_str, '0x100000c',
> Please use the named constant CPU_TYPE_ARM64 for this magic string.
Added
>
>> 'cputype is correct in Mach-O')
> Looks like this line may be joined with the previous one.
not enough free space for this
>
>> - subtest:is(total_cpusubtype, SUM_CPUSUBTYPE[hw_arch],
>> + subtest:is(mach_o.header.cpusubtype, 0,
> Please use the named constant CPU_SUBTYPE_ARM64 for this magic constant.
added
>
>> 'cpusubtype is correct in Mach-O')
> Looks like this line may be joined with the previous one.
not enough free space for this
>
>> + subtest:is(mach_o.header.filetype, 1,
> What does the 1 filetype mean?
> Please use the named constant.
MH_OBJECT, added * The file type MH_OBJECT is a compact format intended
as output of the * assembler and input (and possibly output) of the link
editor (the .o * format). All sections are in one unnamed segment with
no segment padding. * This format is used as an executable format when
the file is so small the * segment padding greatly increases it's size.
https://opensource.apple.com/source/xnu/xnu-344/EXTERNAL_HEADERS/mach-o/loader.h.auto.html
>
>> + 'filetype is correct in Mach-O')
> Looks like this line may be joined with the previous one.
no free space for this
>
>> + subtest:is(mach_o.header.ncmds, 2,
> Why there are 2 commands for Mach-O format?
> Please use the named constant.
is constant name 'ncmds' self-explained?
>
>> + 'ncmds is correct in Mach-O')
> Looks like this line may be joined with the previous one.
joined
>
>> end
>>
>> -test:test('arm', build_and_check_mach_o)
>> test:test('arm64', build_and_check_mach_o)
>>
>> test:done(true)
>> --
>> 2.34.1
>>
> [1]:https://github.com/LuaJIT/LuaJIT/issues/1181#issue-2202788411
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.tarantool.org/pipermail/tarantool-patches/attachments/20240710/c1e62f18/attachment.htm>
More information about the Tarantool-patches
mailing list