| View previous topic :: View next topic |
| Author |
Message |
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
 Posted: Thu Nov 27, 2008 6:43 Post subject: TMI fix (new thread from game obj array) |
 |
|
From- http://www.nwnx.org/phpBB2/viewtopic.php?t=1167
I think that TMI is happening here:
| Code: |
.text:005BD1A1 inc ecx
.text:005BD1A2 mov eax, ecx
.text:005BD1A4 cmp eax, 20000h
.text:005BD1A9 mov [esi+8], ecx
.text:005BD1AC jge loc_5BF535
|
ecx, the counter, is incremented (it's counting instructions). It's moved into the accumulator (eax) and compared with 20,000 hex, or 131,072. This number is apparently the long sought after TMI limit. It records this value in the stack somewhere (mov [esi+8], ecx), and if the counter is greater than 131,072 it jumps 5BF535. I'm thinking the function at 5BF535 (not shown in post) is just setting an error number, and then jumping to some I/O function which will flush out the dreaded TOO MANY INSTRUCTIONS message.
I'm really excited by my findings. Now I just got to write the plugin. The plugin templates we've been given so far are handlers from SetLocalString function hooks.
Can the maintainer of the NWNX source tell me how to proceed from here? I could use the SetLocalString call to form a function for resetting TMI, but that would mean tracing where ecx is stored (obviously its a register, but I'm assuming there is a pointer equivalent lying around somewhere) and punching a hole in it. Since memory allocation can be a fairly dynamic thing, I'm not sure if I feel comfortable promenading thorugh memory with my NWNX plugin to find it. At least today I'm not  Who knows about tomorrow? Any thoughts on the *best* way to go about writing a TMI fix now? |
|
| Back to top |
|
 |
virusman
Joined: 30 Jan 2005
Posts: 1020
Location: Russia
|
| Posted: Thu Nov 27, 2008 7:19 Post subject: |
|
|
That's right, this is the TMI limit.
You can just change this value in the code memory. It's what the original plugin does.
By the way, esi is "this", which in this case is a CVirtualMachine object. |
|
| Back to top |
|
|
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
| Posted: Thu Nov 27, 2008 9:57 Post subject: |
|
|
Thank you so much virusman. I had done some local tests and had modified my nwserver.exe already, and even sent this information to some friends at Arelith, but its nice hearing it from an experienced debugger.
I'll set about writing something for it right away. Unless you beat me there, but either way, its a victory for the community =). TMI has been a thorn in the side of men and women all over the NWN world for FAR too long. It will be really nice to have this killed off.
I'll post here again when it is done. Hopefully I'll do a nice enough job of it to get it included in the modules selection at nwnx.org. |
|
| Back to top |
|
|
acaos
Joined: 08 May 2007
Posts: 153
|
| Posted: Thu Nov 27, 2008 10:46 Post subject: |
|
|
It already exists for Linux:
nwnx_system
There's also an older nwnx_tmi plugin which only modifies the TMI limit. You could probably use one of those two as a base for coding yours.
Acaos |
|
| Back to top |
|
|
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
| Posted: Fri Nov 28, 2008 9:20 Post subject: |
|
|
I think I blew those plugins away with this rather elegant implementation I came up with today. This one is a real fire and forget, I love it!
I'm really excited, its a remarkably elegant. Just one call to WriteProcessMemory =-)
I'll be posting it soon (lol I sitll have my VS open minimized and my NWNServer running as just successfully completing my first test of it). I just had to dash to the forums to share my excitement, hahah.
On a more serious note, what steps would I have to take to get a download of it hosted at nwnx.org? Any particular packaging instructions for the source and DLL or peer review needed? I'm ready to jump through any hoops - I feel it wouldn't serve the community well if it were lost in that gargantuan NWVault; it really needs the official backing and stamping of nwnx.org |
|
| Back to top |
|
|
virusman
Joined: 30 Jan 2005
Posts: 1020
Location: Russia
|
| Posted: Fri Nov 28, 2008 10:48 Post subject: |
|
|
| addicted2rpg wrote: | | Just one call to WriteProcessMemory |
You can use memcpy instead. NWNX plugins are running in nwserver process space. |
|
| Back to top |
|
|
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
| Posted: Fri Nov 28, 2008 11:27 Post subject: |
|
|
I think I'll shoot myself now, especially since its all workingn and ready to ship
But...I'll go back and change it for memcpy. If a thing should be done, it should be done well. |
|
| Back to top |
|
|
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
| Posted: Fri Nov 28, 2008 11:36 Post subject: |
|
|
Ok, memcpy segfaulted. hmmm.
WriteProcessMemory was working... hmmm.
maybe I just phucked the pointers, but I really don't think so. I'm usually very careful about that. I think WriteProcessMemory is affording me some extra priviledges with the windows kernel as far as to what address space I can write to. Even within a process, there are limitations with the ANSI C functions.
argh...
edit: You do know I'm writing in the instruction image area, right?
| Code: |
void CNWNXTMI::ModifyInstructionImage() {
SIZE_T numBytes;
unsigned char countoverwrite;
HANDLE procHandle;
procHandle = GetCurrentProcess();
if(unlimited == 1) {
tmiPatternStart = tmiPatternStart + 3;
countoverwrite = 0xC3;
}
else {
tmiPatternStart = tmiPatternStart + 7;
countoverwrite = 0xFF & (unsigned char)instruction_cap;
}
if(TRUE != WriteProcessMemory(procHandle, tmiPatternStart, &countoverwrite, 1, &numBytes)) {
Log("o Couldn't write to the nwnserver process. Your TMI settings have not been applied.\n");
return;
}
if(*tmiPatternStart == countoverwrite) {
Log("o New instruction settings successfully applied.\n");
}
CloseHandle(procHandle);
}
|
Here is my final version, as far as 1.0 goes
http://metafocus.net/~cs60/addicteds_TMI.zip
Runs like a dream, but maybe we can shorten it sometime later if I can get that memcpy to work. |
|
| Back to top |
|
|
addicted2rpg
Joined: 01 Aug 2008
Posts: 106
|
| Posted: Sun Nov 30, 2008 7:37 Post subject: |
|
|
I've simplified the code some more and have polished up the documentation (which means its very raw by normal standards, but there IS documentation). I believe the PW world of Arelith may start using this in the future, it is still going through their quality assurance processes which of course is very understandable considering the number of users they have and the sensitivity and care that comes with supporting such a massive PW.
In any case, I've put it on the vault.
http://nwvault.ign.com/View.php?view=other.detail&id=1364 |
|
| Back to top |
|
|
Fireboar
Joined: 17 Feb 2008
Posts: 323
|
| Posted: Sun Nov 30, 2008 11:58 Post subject: |
|
|
| addicted2rpg wrote: | | I believe the PW world of Arelith may start using this in the future |
That's an understatement. The impression I got from Mithreas, the lead scripter there, is that he's positively dancing at the prospect of finally losing that awful arbitrary limit on number of instructions. Seems to work perfectly every time in testing too, so I guess they're going to go live with it pretty soon and see if anything breaks (2x75-player servers is very different to a single 2 player test server). |
|
| Back to top |
|
|
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
