forums.ps2dev.org

[NoEffex]

How would one go about hooking the net modules? It's easy with the kernel modules that are loaded at startup, but with the net modules I'm at a loss because they're loaded during runtime.

Basically, what my code currently does is wait until the inet module is loaded and then hooks various syscalls, but it seems that it doesn't even regard those as used. Then, it directs those syscalls to send data over a usbhostfs-based app (Tested and works just fine), but that's not the problem.

What I don't understand is how the syscalls are used during runtime modules like the net modules, since they obviously don't use the syscalls that the M33 systemctrl can hook.
_________________
Programming with:
Geany + Latest PSPSDK from svn

[Torch]

Because they don't use syscalls at all. They are user mode modules. You have to patch the functions in the net modules memory space to jump into your own code.

Depending on the function, it might be possible to hook it from kernel mode code. Look at Coldbird's sample. If not, you need a user mode hooking module which won't play nice with games.

[NoEffex]

Thanks, I'm fairly certain that's exactly what I needed.

They are exports and as such I can get their addresses via systemctrl, which is what his sample appears to do.

Coldbird, if you're reading this, thanks.

Alright, it appears to work, except for one thing.

I'm assuming when it attempts to call to the kernel stuff it goes KABOOM!

[Coldbird]

Might not be the best spot to write that but~

My hooking sample, while still being the best sample out so far, isn't up 2 date anymore...

1. it relies on the m33 sdk... which isn't really a good idea, especially with gen firmwares taking over the place right now... it does work linking it in but i really think this shouldnt be done anymore.

2. the jump hooking method is outdated aswell, or better said - bug infested.
I figured out what was the reason several functions didn't hook properly and fixed it in a new version of my sample which I didn't release so far.

The problem is that the old sample doesn't check whether a delay slot has been destroyed in the hooking process or not...

The new one does check for this case, and avoids it properly.

I intend to update my old hook sample topic with the new sample once I finished my newest project (the first tool out making use of the hooking code).

If you need a more relyable way for hooking immediately though, feel free to contact me via MSN.

As I've said before though, no problem, I like to help even though I'm not yet on par with people like Torch.
_________________
Been gone for some time. Now I'm back. Someone mind getting me up-2-date?

[Torch]

The syscall related functions from the M33 SDK are basically Tyranid's. You can get the source & headers from RemoteJoy and just statically link them into your modules.

And for the record I'm no expert in any of this.

[Coldbird]

Quickly misusing this thread, but - Torch every time I try to read / write to usermemory from a kernel module the psp crashes.

I suppose it's a memory access privileg problem - got a idea what I could do to circumvent that?
_________________
Been gone for some time. Now I'm back. Someone mind getting me up-2-date?

[Torch]

I have no idea. Post some code. Depending on what you're doing you might need to disable interrupts or something. Actually you should always disable interrupts to be 100% thread safe. Don't see why reading is a problem when you can just set a pointer to any address and read away.

[Coldbird]

Again problem solved myself - was just a typo in the address calculation. ^^
Guess it often is the most simple thing you can think of isnt it?
_________________
Been gone for some time. Now I'm back. Someone mind getting me up-2-date?