diff options
author | fanquake <fanquake@gmail.com> | 2020-04-10 07:10:01 +0800 |
---|---|---|
committer | fanquake <fanquake@gmail.com> | 2020-04-10 07:42:20 +0800 |
commit | d486991aa59d807b1cd62169dc094ac0d53647cb (patch) | |
tree | 793eb6cec94efc977310788706f6699a37b8124d /test/functional/mining_basic.py | |
parent | 405713d00fb414808f3eb61d02899b53cc3f1cc1 (diff) | |
parent | 5ca90f8b598978437340bb8467f527b9edfb2bbf (diff) |
Merge #18295: scripts: add MACHO lazy bindings check to security-check.py
5ca90f8b598978437340bb8467f527b9edfb2bbf scripts: add MACHO lazy bindings check to security-check.py (fanquake)
Pull request description:
This is a slightly belated follow up to #17686 and some discussion with Cory. It's not entirely clear if we should make this change due to the way the macOS dynamic loader appears to work. However I'm opening this for some discussion. Also related to #17768.
#### Issue:
[`LD64`](https://opensource.apple.com/source/ld64/) doesn't set the [MH_BINDATLOAD](https://opensource.apple.com/source/xnu/xnu-6153.11.26/EXTERNAL_HEADERS/mach-o/loader.h.auto.html) bit in the header of MACHO executables, when building with `-bind_at_load`. This is in contradiction to the [documentation](https://opensource.apple.com/source/ld64/ld64-450.3/doc/man/man1/ld.1.auto.html):
```bash
-bind_at_load
Sets a bit in the mach header of the resulting binary which tells dyld to
bind all symbols when the binary is loaded, rather than lazily.
```
The [`ld` in Apples cctools](https://opensource.apple.com/source/cctools/cctools-927.0.2/ld/layout.c.auto.html) does set the bit, however the [cctools-port](https://github.com/tpoechtrager/cctools-port/) that we use for release builds, bundles `LD64`.
However; even if the linker hasn't set that bit, the dynamic loader ([`dyld`](https://opensource.apple.com/source/dyld/)) doesn't seem to ever check for it, and from what I understand, it looks at a different part of the header when determining whether to lazily load symbols.
Note that our release binaries are currently working as expected, and no lazy loading occurs.
#### Example:
Using a small program, we can observe the behaviour of the dynamic loader.
Conducted using:
```bash
clang++ --version
Apple clang version 11.0.0 (clang-1100.0.33.17)
Target: x86_64-apple-darwin18.7.0
ld -v
@(#)PROGRAM:ld PROJECT:ld64-530
BUILD 18:57:17 Dec 13 2019
LTO support using: LLVM version 11.0.0, (clang-1100.0.33.17) (static support for 23, runtime is 23)
TAPI support using: Apple TAPI version 11.0.0 (tapi-1100.0.11)
```
```cpp
#include <iostream>
int main() {
std::cout << "Hello World!\n";
return 0;
}
```
Compile and check the MACHO header:
```bash
clang++ test.cpp -o test
otool -vh test
...
Mach header
magic cputype cpusubtype caps filetype ncmds sizeofcmds flags
MH_MAGIC_64 X86_64 ALL LIB64 EXECUTE 16 1424 NOUNDEFS DYLDLINK TWOLEVEL WEAK_DEFINES BINDS_TO_WEAK PIE
# Run and dump dynamic loader bindings:
DYLD_PRINT_BINDINGS=1 DYLD_PRINT_TO_FILE=no_bind.txt ./test
Hello World!
```
Recompile with `-bind_at_load`. Note still no `BINDATLOAD` flag:
```bash
clang++ test.cpp -o test -Wl,-bind_at_load
otool -vh test
Mach header
magic cputype cpusubtype caps filetype ncmds sizeofcmds flags
MH_MAGIC_64 X86_64 ALL LIB64 EXECUTE 16 1424 NOUNDEFS DYLDLINK TWOLEVEL WEAK_DEFINES BINDS_TO_WEAK PIE
...
DYLD_PRINT_BINDINGS=1 DYLD_PRINT_TO_FILE=bind.txt ./test
Hello World!
```
If we diff the outputs, you can see that `dyld` doesn't perform any lazy bindings when the binary is compiled with `-bind_at_load`, even if the `BINDATLOAD` flag is not set:
```diff
@@ -1,11 +1,27 @@
+dyld: bind: test:0x103EDF030 = libc++.1.dylib:__ZNKSt3__16locale9use_facetERNS0_2idE, *0x103EDF030 = 0x7FFF70C9FA58
+dyld: bind: test:0x103EDF038 = libc++.1.dylib:__ZNKSt3__18ios_base6getlocEv, *0x103EDF038 = 0x7FFF70CA12C2
+dyld: bind: test:0x103EDF068 = libc++.1.dylib:__ZNSt3__113basic_ostreamIcNS_11char_traitsIcEEE6sentryC1ERS3_, *0x103EDF068 = 0x7FFF70CA12B6
+dyld: bind: test:0x103EDF070 = libc++.1.dylib:__ZNSt3__113basic_ostreamIcNS_11char_traitsIcEEE6sentryD1Ev, *0x103EDF070 = 0x7FFF70CA1528
+dyld: bind: test:0x103EDF080 = libc++.1.dylib:__ZNSt3__16localeD1Ev, *0x103EDF080 = 0x7FFF70C9FAE6
<trim>
-dyld: lazy bind: test:0x10D4AC0C8 = libsystem_platform.dylib:_strlen, *0x10D4AC0C8 = 0x7FFF73C5C6E0
-dyld: lazy bind: test:0x10D4AC068 = libc++.1.dylib:__ZNSt3__113basic_ostreamIcNS_11char_traitsIcEEE6sentryC1ERS3_, *0x10D4AC068 = 0x7FFF70CA12B6
-dyld: lazy bind: test:0x10D4AC038 = libc++.1.dylib:__ZNKSt3__18ios_base6getlocEv, *0x10D4AC038 = 0x7FFF70CA12C2
-dyld: lazy bind: test:0x10D4AC030 = libc++.1.dylib:__ZNKSt3__16locale9use_facetERNS0_2idE, *0x10D4AC030 = 0x7FFF70C9FA58
-dyld: lazy bind: test:0x10D4AC080 = libc++.1.dylib:__ZNSt3__16localeD1Ev, *0x10D4AC080 = 0x7FFF70C9FAE6
-dyld: lazy bind: test:0x10D4AC070 = libc++.1.dylib:__ZNSt3__113basic_ostreamIcNS_11char_traitsIcEEE6sentryD1Ev, *0x10D4AC070 = 0x7FFF70CA1528
```
Note: `dyld` also has a `DYLD_BIND_AT_LAUNCH=1` environment variable, that when set, will force any lazy bindings to be non-lazy:
```bash
dyld: forced lazy bind: test:0x10BEC8068 = libc++.1.dylib:__ZNSt3__113basic_ostream
```
#### Thoughts:
After looking at the dyld source, I can't find any checks for `MH_BINDATLOAD`. You can see the flags it does check for, such as MH_PIE or MH_BIND_TO_WEAK [here](https://opensource.apple.com/source/dyld/dyld-732.8/src/ImageLoaderMachO.cpp.auto.html).
It seems that the lazy binding of any symbols depends on whether or not [lazy_bind_size](https://opensource.apple.com/source/xnu/xnu-6153.11.26/EXTERNAL_HEADERS/mach-o/loader.h.auto.html) from the `LC_DYLD_INFO_ONLY` load command is > 0. Which was mentioned in [#17686](https://github.com/bitcoin/bitcoin/pull/17686#issue-350216254).
#### Changes:
This PR is one of [Corys commits](https://github.com/theuni/bitcoin/commit/7b6ba26178d2754568a1308d3d44e038e9ebf450), that I've rebased and modified to make build. I've also included an addition to the `security-check.py` script to check for the flag.
However, given the above, I'm not entirely sure this patch is the correct approach. If the linker no-longer inserts it, and the dynamic loader doesn't look for it, there might be little benefit to setting it. Or, maybe this is an oversight from Apple and needs some upstream discussion. Looking for some thoughts / Concept ACK/NACK.
One alternate approach we could take is to drop the patch and modify security-check.py to look for `lazy_bind_size` == 0 in the `LC_DYLD_INFO_ONLY` load command, using `otool -l`.
ACKs for top commit:
theuni:
ACK 5ca90f8b598978437340bb8467f527b9edfb2bbf
Tree-SHA512: 444022ea9d19ed74dd06dc2ab3857a9c23fbc2f6475364e8552d761b712d684b3a7114d144f20de42328d1a99403b48667ba96885121392affb2e05b834b6e1c
Diffstat (limited to 'test/functional/mining_basic.py')
0 files changed, 0 insertions, 0 deletions