Initial Linux support #97

Dadoum · 2024-07-07T16:11:23Z

I just adapted part of XNU headers in the folder. Hopefully I didn't break anything on macOS.

Addresses part of #11

p-x9 · 2024-07-08T01:06:08Z

Thanks.

As for MachOImage, I don't think all the features are available on Linux.
Therefore, I was thinking that everything could be wrapped up in #if canImport(Darwin).

Similar problems also occurred when supporting Linux with ELFKit, but it seems that for some header files it is necessary to use #if !__has_include to avoid duplicate definitions.
https://github.com/p-x9/ELFKit/blob/ca073217cfc55921a33f01aa2360b56bcb78db66/Sources/ELFKitC/include/elf32.h#L37

Dadoum · 2024-08-19T22:46:49Z

As for MachOImage, I don't think all the features are available on Linux. Therefore, I was thinking that everything could be wrapped up in #if canImport(Darwin).

Since they compile, they can be useful to make a mach-o loader on Linux so I think it's okay to leave them enabled.

Similar problems also occurred when supporting Linux with ELFKit, but it seems that for some header files it is necessary to use #if !__has_include to avoid duplicate definitions. https://github.com/p-x9/ELFKit/blob/ca073217cfc55921a33f01aa2360b56bcb78db66/Sources/ELFKitC/include/elf32.h#L37

Now working on that (I also noticed the use of Apple's Crypto library bumps the requirements from macOS 10.13 to 10.15, is that okay for you?).

p-x9 · 2024-08-22T19:00:06Z

Thanks @Dadoum .

Since they compile, they can be useful to make a mach-o loader on Linux so I think it's okay to leave them enabled.

That's true.

Now working on that (I also noticed the use of Apple's Crypto library bumps the requirements from macOS 10.13 to 10.15, is that okay for you?).

I thought the way it was written when using Apple's Crypto library was simpler and better than the current implementation, so the required OS version bump was not a problem for me.

p-x9 · 2024-08-22T19:00:39Z

Sources/MachOKit/Model/Codesign/CodeDirectory/CodeSignCodeDirectory.swift

-            guard let baseAddress = $0.baseAddress else {
-                return nil
-            }
-            switch hashType {
+        switch hashType {
            case .sha1:
-                var digest = [UInt8](repeating: 0, count: Int(CC_SHA1_DIGEST_LENGTH))
-                CC_SHA1(baseAddress, length, &digest)
-                return Data(digest)
+                return Data(Insecure.SHA1.hash(data: data))
            case .sha256, .sha256_truncated:
-                var digest = [UInt8](repeating: 0, count: Int(CC_SHA256_DIGEST_LENGTH))
-                CC_SHA256(baseAddress, length, &digest)
-                return Data(digest)
+                return Data(SHA256.hash(data: data))
            case .sha384:
-                var digest = [UInt8](repeating: 0, count: Int(CC_SHA384_DIGEST_LENGTH))
-                CC_SHA384(baseAddress, length, &digest)
-                return Data(digest)
+                return Data(SHA384.hash(data: data))


p-x9 · 2024-09-12T09:20:29Z

Hello, @Dadoum
I tried to build on macOS and I got some errors related to include in the header file.
I have made some fixes, can I push some additional commits to this pull request?
(I am using the virtual environment Ubuntu 22.04 for testing in a Linux environment.)

Dadoum · 2024-09-12T09:28:21Z

Yes you can. I tried to work with the updated branch but my swift compiler now crashes when trying to compile it (I need to find a fix, or to wait for a swift update for fedora)

p-x9 · 2024-09-12T13:49:30Z

It was not possible to push additional commits to this pull request.
(This may be due to the fact that it is the main branch.)

In my environment, I was able to build on both MacOS and Ubuntu with the following changes.

0001-Move-c-headers-for-linux-platforms.patch
0001-Update-supported-version-of-platforms.patch
0001-Do-not-run-MachOPrintTests-in-a-Linux-environment.patch

Dadoum · 2024-09-12T16:31:29Z

I can't test it unfortunately because I can't build PrebuiltLoader.swift unfortunately. I tried on Swift 5 and 6 on my main Fedora machine. Will commit your changes though.

For the record, here is a snippet of what outputs Swift 6 experimental dev version from Ubuntu on my computer:

2.      Compiling with effective version 5.10
3.      While walking into 'prebuilt_loader' (in module 'MachOKitC')
4.      While verifying VarDecl 'isPrebuilt' (in module 'MachOKitC')
 #0 0x000055c6ec6f3f67 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/dadoum/KDLs/swift-DEVELOPMENT-SNAPSHOT-2024-09-11-a-ubuntu20.04/usr/bin/swift-frontend+0x886ef67)
 #1 0x000055c6ec6f1c6e llvm::sys::RunSignalHandlers() (/home/dadoum/KDLs/swift-DEVELOPMENT-SNAPSHOT-2024-09-11-a-ubuntu20.04/usr/bin/swift-frontend+0x886cc6e)
 #2 0x000055c6ec6f45df SignalHandler(int) Signals.cpp:0:0
 #3 0x00007f315f64fd00 __restore_rt (/lib64/libc.so.6+0x40d00)
 #4 0x00007f315f6a84a4 __pthread_kill_implementation (/lib64/libc.so.6+0x994a4)
 #5 0x00007f315f64fc4e gsignal (/lib64/libc.so.6+0x40c4e)
 #6 0x00007f315f637902 abort (/lib64/libc.so.6+0x28902)
 #7 0x000055c6e74802ec (anonymous namespace)::Verifier::verifyChecked(swift::AbstractStorageDecl*) ASTVerifier.cpp:0:0
 #8 0x000055c6e747fc41 (anonymous namespace)::Verifier::verifyChecked(swift::VarDecl*) ASTVerifier.cpp:0:0
 #9 0x000055c6e746ad38 (anonymous namespace)::Verifier::walkToDeclPost(swift::Decl*) ASTVerifier.cpp:0:0
#10 0x000055c6e74819ba (anonymous namespace)::Traversal::doIt(swift::Decl*) ASTWalker.cpp:0:0
#11 0x000055c6e7485daf (anonymous namespace)::Traversal::visitNominalTypeDecl(swift::NominalTypeDecl*) ASTWalker.cpp:0:0
#12 0x000055c6e7484bd9 (anonymous namespace)::Traversal::visit(swift::Decl*) ASTWalker.cpp:0:0
#13 0x000055c6e74819aa (anonymous namespace)::Traversal::doIt(swift::Decl*) ASTWalker.cpp:0:0

Dadoum · 2024-09-16T19:10:11Z

It's due to the bitfields used in the structure!

Sources/MachOKitC/include/runetype.h

p-x9 · 2024-09-17T11:40:02Z

I also used bit field in dyld_cache_format.h.
I am therefore wondering why the PrebuiltLoader part is an error.
(Is it a bug in swift 6.0 or a platform-specific issue?)

I am using the following environment and it works.

- Ubuntu 22.04
- swift-5.10.1-RELEASE-ubuntu22.04-aarch64

If you go to the MachOKit directory and then run the following command, the MachOKit-enabled interpreter will start up.

swift run -repl

The interpreter was then tested by typing:

import Foundation
import MachOKit

let machO = MachOFile(url: URL(fileURLWithPath: "Path To mach-o file"))

Dadoum · 2024-09-17T12:03:15Z

Apple just released Swift 6, with a fedora build that does no longer have the issue.

Co-authored-by: p-x9 <chi.iosdev@gmail.com>

p-x9 · 2024-10-01T11:41:21Z

Hello @Dadoum
I am sorry for the delay, but I have been able to confirm that it works correctly on both macOS/Linux.
I will merge it if it is not a problem for you.

Dadoum · 2024-10-01T11:46:13Z

It is not a problem for me, you can merge it.

p-x9 · 2024-10-01T11:54:34Z

OK, Thanks for your much work.

Initial Linux support

cdc252b

Merge branch 'p-x9:main' into main

b66b9e1

p-x9 reviewed Aug 22, 2024

View reviewed changes

Add p-x9 patches

4a79c62

p-x9 reviewed Sep 17, 2024

View reviewed changes

Sources/MachOKitC/include/runetype.h Outdated Show resolved Hide resolved

Fix Sources/MachOKitC/include/runetype.h

d1e515e

Co-authored-by: p-x9 <chi.iosdev@gmail.com>

p-x9 merged commit c3ef0ac into p-x9:main Oct 1, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Initial Linux support #97

Initial Linux support #97

Dadoum commented Jul 7, 2024 •

edited

Loading

p-x9 commented Jul 8, 2024

Dadoum commented Aug 19, 2024

p-x9 commented Aug 22, 2024

p-x9 Aug 22, 2024

p-x9 commented Sep 12, 2024

Dadoum commented Sep 12, 2024

p-x9 commented Sep 12, 2024

Dadoum commented Sep 12, 2024

Dadoum commented Sep 16, 2024

p-x9 commented Sep 17, 2024 •

edited

Loading

Dadoum commented Sep 17, 2024

p-x9 commented Oct 1, 2024

Dadoum commented Oct 1, 2024

p-x9 commented Oct 1, 2024

Initial Linux support #97

Initial Linux support #97

Conversation

Dadoum commented Jul 7, 2024 • edited Loading

p-x9 commented Jul 8, 2024

Dadoum commented Aug 19, 2024

p-x9 commented Aug 22, 2024

p-x9 Aug 22, 2024

Choose a reason for hiding this comment

p-x9 commented Sep 12, 2024

Dadoum commented Sep 12, 2024

p-x9 commented Sep 12, 2024

Dadoum commented Sep 12, 2024

Dadoum commented Sep 16, 2024

p-x9 commented Sep 17, 2024 • edited Loading

Dadoum commented Sep 17, 2024

p-x9 commented Oct 1, 2024

Dadoum commented Oct 1, 2024

p-x9 commented Oct 1, 2024

Dadoum commented Jul 7, 2024 •

edited

Loading

p-x9 commented Sep 17, 2024 •

edited

Loading