Slashdot Mirror
.ANI Vulnerability Patch Breaks Applications
Jud writes "Microsoft's fix for the .ANI vulnerability was part of Patch Tuesday yesterday. However, all is not well with the update. Reportedly, installing the patch will break applications such as Realtek HD Audio Control Panel and CD-Tag, which mentions they are affected by the problem on their main page. A hotfix is currently available from Microsoft, however their current position is this is an isolated problem and the fix is not planned to be pushed out through Microsoft Update. "
"their current position is this is an isolated problem"
Weird, 'cause I hear about one of these stories almost every week. Isolated in what sense?
My turnips listen for the soft cry of your love
My archiving application of choice, Tugzip is also affected by this update and the mentioned fix took care of the problem.
Microsoft breaks something when patching something else? I'm surprised. Really. I am. No, really. I am.
Patch Tuesday is the second Tuesday of each month. This was an out of cycle patch released.
I just wanted to make a quick post before I see all the standard lame M$ bashing gets out of hands from a ton of idiots that are most likely using Windows while posting.
This is exactly why it takes Microsoft so long to put out patches sometimes. Unlikely all these free and open source packages, Microsoft Windows is actually used by tons of users at home and in the business world. People need their machines to do their daily activities and jobs. This is why so much testing is needed before something can just be shoved out there. This is why you tend to see this sort of thing from patches released out of cycle. It obviously has not and could not have been tested as much (and yes sometimes problems occur with patch Tuesday patches).
You might not see as many issues with *nix based systems. Why? Well, there just are as many users. This might sound like a cliche but it is a fact. Look at when official Redhat patches and other updated packages actually come out. They come out days, weeks, and months later. Sure there is some patch that some random guy hatched together -- the power of open source!! However, if you were to apply that untested P.O.S. across the world in tons of real environments, you'd probably have a shitton of problems.
This does not excuse problems with patches, but at least it came quicker. Remember, M$ has to release stuff that fortune 1000, government, home users, and everyone else can live with. Pushing some patch 30 minutes later for an OSS package that 2000 rag tag home users use.. just isn't the same.
If it does affect calc.exe, it rather seems like you have some DLL injection (keylogger/spyware, or something legit) that then causes this. If they messed up the base address, or just increased the size over a previously valid boundary, all kinds of DLLs with preferred addresses in the same region could start causing interference.
You simply have to be careful with the address space if you are a library that will be dynamically loaded in plenty of images, especially if you are loaded very early on.
(Heh, last summer, I got the genius idea that the base addresses were probably not optimal after all hotfixes and 3rd party software, so I started a gigantic rebase on the complete system32. That's a baaaad idea. I should at least have had enough sense to exclude NTOSKRNL, but I obviously didn't. Repair was fun...)
A lot of machines have the Realtek HD Audio thing in them to provide audio - notably most of the Core/Core2 based ones (HD Audio is a standard by Intel, Realtek being one of the first to offer it).
Seems like this isn't really an "isolated" problem, but a fairly common one if you own a desktop made in the last year or a recent laptop...
... and all I had to do to solve it was go to Realtek's site and download the latest version of their driver. Problem solved (knock on wood).
So.. If the fix is that simple, is this issue really an issue or is this issue blown out of proportion?
This is my opinion. To make sure you don't steal it, it's covered by the DMCA.
"There's been a little complication with my complication"
...unless you're in NZ or Australia, when Patch Tuesday is on Wednesday.
"It doesn't cost enough, and it makes too much sense."
But of course it's available if you do want it.
Naturally Linux Genuine Advantage is open source, and not to be outdone by Microsoft platform hackers a hack is available to auto-certify LGA without actually contacting the LGA server.
Help stamp out iliturcy.
Well, for starters, changing the mouse cursor is a part of the official W3C CSS specs...
http://www.w3.org/TR/CSS21/ui.html#propdef-cursor
In other words, *something* has to be able to load and process the mouse cursor. And if the thing that loads and processes the mouse cursor falls prey to a buffer overflow, then you've got yourself a vulnerability. Since it's the OS that handles and draws the mouse (so it's not an IE thing; FF will fall prey to this too), it's the OS that handles the mouse cursor, so a buffer overflow there means that it's the OS that gets compromised--the very same thing could happen in any other OS if there was a similar mistake. So they forgot a length check. Shit happens.
I'm a developer for a software package that lots of automotive engineers use to do bus analysis. The patch broke our software, and we've gotten calls from lots of people at our smaller companies wondering what was going on. The bigger (think Big 3) customers have huge turn around times on Windows Update patches, but as of now we have lots of angry people wondering why our software won't work. Nothing like MS giving us bad rep for essentially us doing nothing.
You don't need WGA to receive Windows patches either. Automatic updates will work perfectly fine even if your serial is blatantly stolen; but 'upgrades' won't (IE7, WMP, and all that good fun).
Yeah, WGA sucks, but software updates will come either way.
Also, to say Linux is completely independently modular isn't entirely accurate either (although, in fairness it's not like I completely disagree). Upgrading kernels (due to 'serious' security vulns) will break more things than I'd like for instance - my (yes I know, proprietary) NVidia drivers will be the first victims.
Finally, you think MS don't feel your patch-pain? Just by working for them in one form or other you've subscribed your machine to every patch, patch-for-the-patch, and patch-for-the-hacked-patched-service-pack that gets tested; all before releasing to a unsuspecting public, almost every day, just so you get as little bother staying secure with as little hassle as possible. It's not a pretty sight sometimes.
throw new NoSignatureException();
The screw up is in Microsoft's release system allowing hhctrl.ocx and user32.dll to be shipped with the exact same "DLL Base Address". They both share address 0x7E410000. I'm sure Microsoft has a system to prevent this, but either someone didn't follow it, or the system has flaws.
Normally clashes of base addresses happen all the time. For most DLLs the base address is sort of a suggested location, the OS load the DLL to this area if it can, but will "relocate" DLLs to free memory area if that space is taken. User32.dll isn't allowed to be relocated for some reason (a very good reason, I'm sure). If it's space is already taken (by HHctrl) the program using it cannot load. HHCtrl.ocx has no problem being relocated, but this will only happen if it's loaded after user32.
I'm surprised that anyone could manage to make an application that would load these DLLs in this bad order - but that's not the point I guess. Usually you'd HAVE to call a function in User32.dll long before loading anything COM - esp an HTML help control (which is what hhctrl is).
A big HA-HA! goes out to the vendors who insist on using every imaginable gimmick and gee-wiz animation / transparency effect / irregular shaped window trick to try to make their product appeal to their target audience of 8 year olds. Stick with the basics, please! There's no reason for an audio control panel to require an animated cursor, for christsakes!!!
Reminds me of when I bought a little FM radio controlled by a serial link. The crapplet they sent on the CD-ROM was so annoying, the first thing I did was sniff the serial protocol and write my own little non-obtrusive applet. I asked the manufacturer for the proto specs first, but they delined, even after I pointed out how easy it was going to be to reverse engineer them... idiots!
Never thought I'd write something like this, but kudos to MS for saying we're not going to work around your crappy little app.
</rant>
What bothers me is that it makes me feel like this "fix" may not even patch the real problem.
You see, moving where a DLL is stored in memory might break the proof of concept, but it might not actually fix the vulnerability. Sure, the code it hooked into before in order to hack the machine won't be in the same place, but it might well be possible to fix the exploit to point to the code's new location.
In short, I wonder if they're playing tricks to make it more difficult to exploit without actually fixing the underlying problem?
Blame Microsoft all you want for security holes in user32 and GDI, but don't blame them for these programs breaking.
The change that broke these applications was changing the base address of user32.dll (from 77D40000 to 7ED10000 I believe). The programs that broke were using the 7ED10000 range. When user32.dll can't load at its desired address, it will fail to start the process. DLLs hard-linked by an EXE will be loaded before user32 or kernel32 if they are the first in the import table. Once loaded, user32.dll can't load at its desired address and will get relocated. user32.dll doesn't like this and aborts starting the process. This happens even if the bad DLL is relocatable, because preferred addresses are first-come, first-serve.
The entire 60000000-7FFFFFFF address range is reserved for Microsoft DLLs and special memory. Don't stick your own DLLs in there! (This address range is true of Win64 as well: just add 8 zeros to the front.)
As for why user32.dll and kernel32.dll don't relocate like other DLLs: so many badly-coded programs that do DLL injection assume that the addresses of LoadLibraryW, etc. in other processes are the same as the ones in their own processes. Almost all DLL injection code snippets assume this, and it's just wrong. The proper way is to use either EnumProcessModules or CreateToolhelp32Snapshot to list all the DLLs in the target to find the one you want. The correct address to use is then remote_function = local_function - local_dll + remote_dll. The types HMODULE, HINSTANCE, and IMAGE_DOS_HEADER * are equivalent, so just cast to a DWORD_PTR.
"Screw Sun, cross-platform will never work. Let's move on and steal the Java language." - Visual J++ Product Manager