NVIDIA Releases Fix For Dangerous Display Driver Exploit

wiredmikey writes "NVIDIA on Saturday quietly released a driver update (version 310.90) that fixes a recently-uncovered security vulnerability in the NVIDIA Display Driver service (nvvsvc.exe). The vulnerability was disclosed on Christmas day by Peter Winter-Smith, a researcher from the U.K. According to Rapid7's HD Moore, the vulnerability allows a remote attacker with a valid domain account to gain super-user access to any desktop or laptop running the vulnerable service, and allows an attacker (or rogue user) with a low-privileged account to gain super-access to their own system. In addition to the security fix, driver version 310.90 addresses other bugs and brings performance increases for several games and applications for a number of GPUs including the GeForce 400/500/600 Series."

Comments

No 7-series support?

[qwertphobia]

Looks like they're now dropping support for the Geforce 7-series cards. Bummer, I have a 7800GT and it's still pretty quick.

Re:No 7-series support?

[Anonymous Coward]

Really? Try this page.
http://www.geforce.com/drivers/results/49740
Still plenty of support for the 7 series.

Re:No 7-series support?

[qwertphobia]

Not far off.. it's a '96 f-150 and I drive it like I stole it some days. And I get what you're trying to say... but my 7800 runs just fine and these days 90% of my time on my PC is spent in email and web browsers.

Re:No 7-series support?

[Osgeld]

then you would be fine on any card, whats the problem, use the windows default drivers

Re:No 7-series support?

[Anonymous Coward]

here you go a nice affordable replacement for your aging card. yeah yeah no need to thank me for the hot find. it's a real steal!

Re:No 7-series support?

[Elbart]

They've announced to stop support 6- and 7-series a few months back. You should be reading the release notes once in a while.

Dangerous ? Nope.

[lemur3]

Not like a CRT catching fire...

I remember hooking up an old CRT to the wrong video card.. one with way too a high resolution for that screen..

A while later, hooked up to the correct video card, I noticed a bit of smoke coming out from where the dials were.. removed the case.. plugged it in again to see if it was OK .. it burst into 3 foot high flames.

thankfully a fire extinguisher was about 3 feet away... mom would have been awfully mad if i had burned down the house.... scared the bejeezes out of me ... the burnt electrical smell was horrendous..

(bonus: it was a fancy no mess extinguisher)

lesson learned.

Re:Dangerous ? Nope.

[earlzdotnet]

The more I learn about the past of computing the more I'm convinced they only ever considered one failure mode: catastrophic.

Re:Dangerous ? Nope.

[Anonymous Coward]

It appears as though you have found a use for is_computer_on_fire().

http://www.tycomsystems.com/beos/BeBook/The%20Kernel%20Kit/System.html

Good old BeOS. Man I miss that operating system (though Haiku fills that gap nicely), but moreso the radical hardware that came with it (BeBox).

Re:Dangerous ? Nope.

[Anonymous Coward]

http://en.wikipedia.org/wiki/Lp0_on_fire

Re:Dangerous ? Nope.

[jamesh]

Not like a CRT catching fire...

I remember hooking up an old CRT to the wrong video card.. one with way too a high resolution for that screen..

A while later, hooked up to the correct video card, I noticed a bit of smoke coming out from where the dials were.. removed the case.. plugged it in again to see if it was OK .. it burst into 3 foot high flames.

thankfully a fire extinguisher was about 3 feet away... mom would have been awfully mad if i had burned down the house.... scared the bejeezes out of me ... the burnt electrical smell was horrendous..

(bonus: it was a fancy no mess extinguisher)

lesson learned.

Bullshit...

Possibly, but not necessarily. Older hardware wasn't always manufactured to "must not burn" specifications, and scanning at too high resolution could definitely cause damage to the monitor (and there were warnings in the manuals to that effect). My guess would be that the damage would be through overheating so fire is definitely a possibility, especially if the thing is full of flammable crud (fluff, paper dust, etc). 3 foot high flames might be an exaggerated memory of a distant past but I'm willing to believe the essence of the story.

Re:Dangerous ? Nope.

[Black LED]

I have never seen flames come out of a CRT display, but I have witnessed at least two which have died after emitting a decent amount of black smoke. One was an old 14" Hyundai SVGA monitor and the other was a Compaq of some sort. The Hyundai had been in use for years at that point and had already suffered some kind of circuit board failure that prevented the the blue gun from firing, resulting in a golden image tone.

Re:Dangerous ? Nope.

[Gaygirlie]

Bullshit...

I doubt that. I've actually seen myself an old CRT bursting into flames shortly after I noticed the plastic on its side turning brownish and starting to melt. The thing is, CRTs are a very much different kind of a beast than our LCDs and a CRT can indeed be permanently damaged just by sending a wrong kind of a signal. Sending a wrong signal enough could cause the capacitors to blow and this could result in a fire. Have you ever opened a CRT-display? Those things have huge voltages going on there. I once opened this 21" high-end CRT and the warning labels on the rails there read 17,000 volts.

Re:Dangerous ? Nope.

[X0563511]

I raise you a CPU opcode, HCF

Re:Dangerous ? Nope.

[mrmeval]

NEC had one glorious little monitor, pretty thing and expensive but if were plugged into the wrong card it would most of the time sheer the picture tube off, it would at best damage the phosphor. This would happen right at the part of the yoke closest to the face. Couple an odd reaction by the vertical circuit to a higher vertical frequency to an increase in horizontal frequency resulting in a very high boost in high voltage unchecked by an inadequate x-ray protect and you had in effect an electron cutting beam till air entered and arcing caused it to finally draw enough current to trip the safety. NEC just replaced them with more expensive units.

Re:Dangerous ? Nope.

[ATMD]

Nice screen saver.

Re:Dangerous ? Nope.

[Shark]

Made in Britain?

Re:Dangerous ? Nope.

[Zero__Kelvin]

"mom would have been awfully mad if i had burned down the house"

... until she realized that it finally got you out of her basement?

No it didn't

[ArchieBunker]

I've *NEVER* heard of a single instance of a refresh rate or too high of a scanning frequency causing monitor failure. Seems like a trivial thing to fix for a monitor manufacturer. Would you sell a product that shot out fire if someone clicked a slider setting too high?

Re:No it didn't

[lemur3]

it was one of these..

GATEWAY 2000 CRYSTALSCAN 1024NI

  http://meghan.schnooze.com/Photos/ToSell/monitor.jpg

think it was 14 inch..

and to the naysayers! yes! it really did happen.

Re:No it didn't

[Black LED]

But dude! Crystalscan! CRYSTALSCAN!

Re:No it didn't

[Kjella]

I think I experienced that. It was in the 80s, old IBM PC and somehow the machine froze - not just the screen but it didn't respond to input or anything, but I left it running in the hope it'd recover. Suddenly there was a rather loud bang as a capacitor blew and released its magic smoke. Of course it could be coincidence, but I suspect it was the computer crashing and sending a very bad signal to the screen. This was the age where you'd check machine and screen compatibility before plugging it in, probably for a good reason.

Re:No it didn't

[drinkypoo]

I've *NEVER* heard of a single instance of a refresh rate or too high of a scanning frequency causing monitor failure.

It's not the kind of thing that is likely to crop up, is it? The only monitors which will even try to sync to a bad frequency were ancient multisyncs, modern ones are smart enough to detect a signal out of range. And you'd pretty much have to have some bad hardware for it to happen.

Re:No it didn't

[ais523]

If you overclock a monitor, you can expect similar results to if you try to overclock a CPU. (In general, it's going to overheat.) Because it's rather easier to do by mistake than it is for a CPU, monitors tend to check for it nowadays and cut out intentionally.

Re:No it didn't

[ATMD]

Would you sell a product that shot out fire if someone clicked a slider setting too high?

I have one of those, I use it for sticking bits of metal together.

Turn your flipping auto-updater on

[GodfatherofSoul]

The days of trying to manually screen each update your system needs are over. Too many components are vulnerable and the turnaround time for an exploit is too short.

Re:Turn your flipping auto-updater on

[DrXym]

Laptop driver support can be *horrible* because manufacturers twiddle with the chipsets params so that means their drivers are machine specific and certified. I'm writing this on an old netbook with Intel IGP. The OpenGL implementation is bugged so I want to install a later driver which is up on Intel's site. Can I install this driver? No because it decides "the driver being installed is not validated for this computer". And HP don't give a fuck about providing a certified version.

Re:Turn your flipping auto-updater on

[Black LED]

I have been using notebook drivers direct from Nvidia for quite some time; at least since I bought the laptop prior to my current one, which would have been in late 2008. Maybe it was different for the Quadro, but they have had mobile GeForce drivers available for download for years.

I don't know if this will work for you, but here are the Quadro Notebook Drivers v310.90 dated yesterday.

Re:Turn your flipping auto-updater on

[jones_supa]

Which netbook?

Re:Turn your flipping auto-updater on

[DrXym]

An HP Mini 210. OpenGL ES 2.0 driver is totally broken - I'm on holiday and using it for development and the driver just crashes whenever glLinkProgram is called. Seems to be a widespread issue so I assume it's a driver fault.

Re:Turn your flipping auto-updater on

[jittles]

I have been using notebook drivers direct from Nvidia for quite some time; at least since I bought the laptop prior to my current one, which would have been in late 2008. Maybe it was different for the Quadro, but they have had mobile GeForce drivers available for download for years. I don't know if this will work for you, but here are the Quadro Notebook Drivers v310.90 dated yesterday.

Whether you can install the latest and greatest drivers from NVidia, or have to download from your OEM depends on the OEM. They usually tweak some settings (PCI Device ID) specifically so that the default NVidia drivers will not install. They do this for support purposes. You can modify the INF for the Nvidia drivers and force them to install, but it can be a pain to get them working right sometimes. For instance, when Vista came out, I bought a new laptop. Dell restricted the drivers so I couldn't use the Windows XP drivers for my video card. I had to find this solution to be able to run XP on the device.

Re:Turn your flipping auto-updater on

[Black LED]

That used to be true, but it hasn't been like that for a long time. The Nvidia notebook drivers seem to be pretty universal. While I can't say with certainty that they will work with all laptops, I haven't run across that particular problem in years. I've installed the unified notebook drivers across a dozen different models among perhaps a half dozen brands, none of them any older than seven years, and they have worked without any fuss.

All I can recommend is for you to give them a shot. They will probably work.

It's also pretty old

[Sycraft-fu]

About 7.5 years old. It is reasonable that they cease supporting it with new drivers. You can still get drivers for it, they have drivers for OSes up to and including Windows 8, they just aren't keeping support in newer unified drivers.

Sounds pretty reasonable to me. They gave you over 7 years of driver updates. It is fairly unrealistic to assume that they'll continue with new support forever, particularly given that there is little reason. The 7 series can't do WDDM 1.1 or 1.2, it can't handle DirectX 10, 10.1, 11 or 11.1, it can't do CUDA, DirectCompute or OpenCL. There is just little in the way of things to implement for it.

If you wish to continue using the card, no problem (though be aware that an Intel 4000 series GPU found in Ivy Bridge processors is likely to be faster, and certainly has far more features) just use the 306 series drivers. It will continue to operate with those no problem.

If the security issues is what you are worried about, it looks like it only affected the 310 drivers, so no issues there.

Re:It's also pretty old

[Billly Gates]

Nvidia and ATI have great cheap $49 cards if you want aero. That can cream the gaming 7800 series easily. No meed to get a new system.

If it is on XP you have a lot more security issues than this card though.

Re:It's also pretty old

[qwertphobia]

Yes, it's reasonable, that doesn't mean I like it. I won't gracefully give up my right to complain on the Internet.

Frankly it's linux kernel compatibility I'm most concerned about. If Fedora 18 comes out next week with an updated kernel which breaks compatibility with the current 7-series driver, what are the chances it's going to get fixed?

In the other hand, things are moving along in the Nouveau open source driver so there are alternatives.

Re:It's also pretty old

[davydagger]

" it can't do CUDA, DirectCompute or OpenCL"

nouveau is feature complete for 2D and 3D rendering, in addition, without OpenCL, CUDA, or the bells and whistles of new hardware, it just might work better for this guy's application.

nouveau renders the world wide web just fine, it also works all the way back to NV04.

Your also not going to be playing the latest games on your old video card anyway.

Re:It's also pretty old

[drinkypoo]

I got a 1GB 610GT for $12 on Black Friday. It's in my system right now because the new fan for my 240GT (still better than the budget cards, anyway) has only just arrived. It's a gimped 520 IIRC and still better than your 7800. Its specs are almost as good as my 240GT (which was a spectacularly good core when the card was new, in spite of it being based on a core which was already old when it was released) but it has 1/4 the fill rate so 1920x1200 is murder.

I do miss the days when the nVidia driver went all the way back to geforce, but not very much. My oldest card is a 9600 and I don't even use it, it's just my backup for my backup.

Re:It's also pretty old

[Anonymous Coward]

NVC0 family [400 series]
All sorts of fun. Feature-wise it isn't too different but the architecture has changed a lot.
These cards are generally working with the latest kernel and Mesa but may still have power management issues. It is recommended to use the Linux 3.1 kernel or newer (or a backported driver from this kernel).

Oh yea, it sounds perfect!

Re:It's also pretty old

[hairyfeet]

Ya know, I never understood why folks have a fit when Nvidia and AMD drop support, as you pointed out they are several years old now and simply can't do WDDM 1.2 which is required for all the features of Win 8. MSFT has always had legacy drivers built in so you probably won't need a driver at all for an Nvidia 7 or 8, and on the AMD side 2, 3, and 4, so what is there to complain about?

Hell my HD4850 has had support dropped for nearly a year but Win 7 and all my games run fine so why should I care? Its not like these cards are gonna magically have more performance squeezed out of them via software, all the bugs are pretty much worked out by now, so what good would new drivers do? Considering the fact you can buy an Nvidia 210 for like $20 that will run rings around the old 7 series if he really wants to run win 8 he'd be better off just getting a new card and if he is on XP-7 it should run just fine with the drivers he has.

The only place i could see it being a problem is Linux but until Torvalds joins the rest of the world and has a stable ABI so older drivers can work on the latest kernel its either the hacked together FOSS drivers or you're SOL. Neither Nvidia nor AMD will ever be able to fully open their drivers thanks to HDCP, AMD were able to give around 65% but that's it and with Nvidia you run a supported card or give it up. But you can't blame the GPU companies for that, when its no longer supported that's it and if the old drivers won't run tough luck, you can't expect them to pay a team of devs to support cards they aren't even selling anymore.

Re:It's also pretty old

[Lost Race]

If it is on XP you have a lot more security issues than this card though.

Such as?

I have a few XP systems that are still getting regular automatic patches and updates from Microsoft. You seem pretty confident that all XP systems are vulnerable, so you must be aware of something specific. Care to share?

ATI had an exploit too

[Billly Gates]

Do we as geeks and IT professionals need to worry about this?

First it was the OS that got you owned. Then when Linux, Macosx, and NT/XP came it was about IE. IE 5.5 and 6 were instant targets. Then as that died off it was flash, java, and ODF addons.

Are video drivers next? Which never gets updated? The video drivers. Which has its own cpu, ram, and is never checked by AV? The video card. A reflash would be a nightmate.

Re:ATI had an exploit too

[DrXym]

Do we as geeks and IT professionals need to worry about this?

Absolutely. WebGL allows any random website to tap your hardware through the browser. WebGL is essentially OpenGL ES 2.0 give or take a few APIs and is supported by just about every modern browser except IE. Some enable WebGL by default on suitable hardware, some have it disabled by default. When it is enabled a page has carte blanche to abuse the chipset six ways to sunday. The only protection afforded by browsers is the driver has to implement a GL extension called GL_EXT_robustness which says the driver promises, fingers crossed to be really good about checking and recovering from errors.

ActiveX had something similar called the "safe for scripting" bit. IE wouldn't load a page unless the control said it was safe and look what happened there. While there are less graphics drivers than activeX controls, it's easy to imagine a driver version claiming it's robust when in fact it isn't. It's easy to imagine a malicious site using that fact to break a lot of machines. I assume browsers could implement a whitelist of "good" drivers and update the list in addition to checking for the extension but it's obviously imperfect and offers additional browser exploits where none existed before.

Re:Let's see how long it takes to download.

[Osgeld]

what sucks is a 170 meg driver download, thats just fucking stupid

sadly enough I think they got a little smaller

Re:nvvsvc.exe

[X0563511]

Probably detecting driver freezes and restarting it.

It's something I've only seen happen rarely - but I've had a game "lock up" for a few seconds, only to be greeted by a notification that the driver froze and was restarted. The game died, but the whole system did not (without this functionality it would have been a power cycle)

Re:nvvsvc.exe

[jones_supa]

Isn't that functionality built in Windows instead of nvvsvc.exe?

Re:Let's see how long it takes to download.

[jittles]

Part of the reason the driver is so big is because they now package all cards into one driver. Well, at least all of their GeForce cards. You literally have generations worth of drivers in one file. Sure they added the PhysX and the HD AUdio driver, 3D crud, and a few other things. However, I think most of that size comes from different driver files. I don't think all of them get installed.

Re:nvvsvc.exe

[X0563511]

I don't believe so, but I could be wrong. If it is, it doesn't seem to trigger correctly for ATI/AMD cards when I've had some limited time using them.

Re:nvvsvc.exe

[jones_supa]

It works for Intel cards.

NVidia forgot to test their drivers again ...

[Cammi]

A NVidia, how about fixing your drivers so that it will stop quit providing a signal on windows 8 machines after an hour or so? Should have tested your drivers before release.

Re:NVidia forgot to test their drivers again ...

[Zero__Kelvin]

" how about fixing your drivers so that it will stop quit providing a signal on windows 8 machines after an hour or so"

You make that plus sound like a drawback.

Re:NVidia forgot to test their drivers again ...

[Cammi]

If a driver does not work, it is not a plus. For instance ... you would not be on a computer ...

Re:NVidia forgot to test their drivers again ...

[Zero__Kelvin]

"If a driver does not work, it is not a plus. For instance ... you would not be on a computer ..."

... and by logical inference, if a driver for Windows 8 does not work, I would not be using a Windows 8 computer. Ergo, it is not a drawback; it is a major plus. Please try to follow along with the rest of the class.

Re:NVidia forgot to test their drivers again ...

[Cammi]

Yeah, that made no sense. Please try to follow along with the rest of the class.

Re:NVidia forgot to test their drivers again ...

[Zero__Kelvin]

I know you are but what am I ...

Drivers and the network

[obeythefist]

Excuse me if this is a dumb question, but why is the display driver exposed to the network at all?

Re:Drivers and the network

[hobarrera]

No network access as far as I can see:

the vulnerability allows a remote attacker with a valid domain account to gain super-user access

You need an account on the machine to log into it first.