aboutsummaryrefslogtreecommitdiff
path: root/lib (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Revert <stdatomic.h> dependency.comex2015-07-161-10/+8
| | | | | | | | it was added in the otherwise cosmetic dfd16cc1f759a103707ebd27743abc621486e131 because I didn't change the generated code at all - just switched to the standard aliases of the functions I was already using. But I forgot that <stdatomic.h> isn't present in relatively recent Xcode, so I'm switching back for now.
* Add options field to struct substitute_{function,import}_hook.comex2015-07-161-2/+12
| | | | | | | | | | | | I will probably need to add an option to deal with GCC 5.2's new optimization - to not save caller-saved registers if the called function is known not to use them, which would break with hooked functions. Don't want to emit the required stub by default because it will likely be slow. I need to verify that Clang doesn't currently have such an optimization... (ABI break - need to avoid this once it gets widely distributed Soon, obviously)
* style policecomex2015-07-1612-45/+94
|
* work around (incorrect, since it's not code) alignment warningcomex2015-07-161-1/+1
|
* fix accidentally exported symbolscomex2015-07-162-2/+2
|
* typocomex2015-07-161-1/+1
|
* orderingcomex2015-07-161-8/+8
|
* Import some code to handle <redacted> syms.comex2015-07-162-11/+199
|
* hook up the list of dylibscomex2015-07-132-0/+9
|
* don't be n^2 for removal, lol. i should probably switch to rust's robin ↵comex2015-07-121-21/+12
| | | | hood hashing - it isn't that much code and supposedly performs much better when the table gets full. *however*, it requires checking the hash of every entry in the chain during insert, which basically means storing it, which means more memory usage ... but by allowing fuller tables it could decrease memory usage. but if you have a big table anyway to avoid copying, you don't want *extra*... and storing the hash twice in the simple case is so dumb feeling. dunno.
* fix my hash table algorithm - arghcomex2015-07-121-14/+35
|
* fixescomex2015-07-121-0/+4
|
* redo crash reporting - untested (but it compiles)comex2015-07-122-21/+41
|
* how did that style violation get in there? must have been tiredcomex2015-07-121-4/+8
|
* working on safetycomex2015-07-111-5/+18
|
* fix anti-noisecomex2015-07-111-1/+4
|
* stuff!comex2015-07-103-5/+6
|
* redo all the things with xpc - not tested yetcomex2015-07-071-0/+63
|
* testscomex2015-07-071-0/+4
|
* fix for compiling objc-asm.S on ARM with newer Apple compilercomex2015-04-151-1/+1
|
* fixes for compiling on older OS Xcomex2015-03-032-1/+6
|
* A number of critical fixes painstakingly discovered in the slowest way possible.comex2015-03-0112-23/+39
|
* fix stack alignmentcomex2015-03-011-3/+6
|
* Let's get hacky!comex2015-03-011-3/+16
|
* make jump-dis use a vec as a stack, rather than a hacky queuecomex2015-03-012-26/+18
|
* add veccomex2015-03-014-9/+156
|
* Add extra argument to substitute_hook_functions and interpose_imports for ↵comex2015-02-284-12/+68
| | | | | | | | | | use with unhooking. I think we'll need to explicitly record trampoline locations to avoid the possibility of stomping on someone else's stuff if the function was re-patched... Also, document substitute_hook_functions.
* fix accidental usage of panic() from mach/mach.h instead of substitute_paniccomex2015-02-282-7/+9
|
* Add substituted and fix things up to use it. Still untested.comex2015-02-281-0/+5
|
* Rename ios-bootstrap to darwin-bootstrap; cleanup posixspawn-hook and ↵comex2015-02-281-145/+0
| | | | | | unrestrict. Not tested yet.
* fix (and make more robust) strerror.ccomex2015-02-243-14/+20
|
* Add unaligned read/write functions.comex2015-02-247-8/+34
| | | | | | | I thought I could get away without since I wasn't (presently) targeting systems without hardware support for unaligned accesses, but on armv7 clang insists on optimizing into the one ARM instruction that requires alignment anyway - LDM/STM. Oops. Damnit, clang.
* fix armv7 syscall registers :ocomex2015-02-241-1/+6
|
* fix "panic" in substrate-compat.ccomex2015-02-241-4/+4
|
* fix arm64/armv7 ishcomex2015-02-245-6/+13
|
* Add transform-dis-cases-arm64.S and fix a serious bug thus discovered.comex2015-02-231-3/+3
| | | | Commit transform-dis-cases-x86_64.S, which I forgot to earlier.
* Ban calls within transform regions in threadsafe mode.comex2015-02-2314-37/+104
|
* fix some i386 stuffcomex2015-02-233-43/+56
|
* fix i386 manual syscall, mmap return checkcomex2015-02-232-5/+9
|
* Fix vm_remap, of trouble with which the previous diagnosis was completely ↵comex2015-02-231-10/+12
| | | | | | | | | | | incorrect. It's actually the fact that no vm_map_server actually exists on the kernel side - even though the otherwise useless function vm_remap exists in the kernel source, and mig generates server code for it, apparently that server code isn't built or linked into the kobject table. Bizarre. Anyway, switch to mach_vm_remap, which is what vm_remap in userspace calls into.
* remove old commentcomex2015-02-231-5/+0
|
* and avoid call to pthreadscomex2015-02-231-1/+5
|
* Redo manual syscalls, and use them for hooking.. And fix mmap, which now ↵comex2015-02-234-84/+114
| | | | makes the whole thing slightly broken, because vm_remap into the middle of the shared region apparently silently does nothing.
* various fixescomex2015-02-186-55/+61
|
* er, don't unnecessarily spam thread start/resume.comex2015-02-181-27/+34
|
* Fix hook-function:comex2015-02-189-311/+466
| | | | | | | | | | - Thread stoppage is now complemented by sigaction to catch injected threads (sigaction is not used exclusively because the rest of the program could be trying to use sigaction itself in the meantime - this is a real thing, ask Dolphin) - mprotect is no longer used due to max_protection possibly getting in the way; instead, a copy is created and mapped onto the original.
* more fixescomex2015-02-153-20/+48
|
* fix test casescomex2015-02-142-2/+2
|
* Trampoline fixes.comex2015-02-1410-48/+86
| | | | | | | | | | | | | The transformed code was incorrect because it assumed the pointer it was writing to was where the code would execute, but it was actually 'rewritten_temp'. Changed transform_dis_main to take a pc_trampoline pointer, which also helps the test harness. However, this means that it has to be called after the trampoline has been allocated, while before the trampoline allocation depended on the generated size; this change doesn't bother to use two passes or anything, but just allocates a new code buffer if the maximum possible size isn't available - not the end of the world, since trampoline_ptr will still only be increased by the actual size before the next hook in the series (if any).
* theoretically support x86 PICcomex2015-02-091-14/+46
|
Copyright (C) 2019-2021 Reiko Asakura. All Rights Reserved.