Skip to content

Changing the cursor shape in Windows proven difficult by NVIDIA (and AMD)

Ten post nie jest dostępny w języku polskim!

{ 14 } Comments

  1. Honorary_BoT | 01-lip-13 at 05:34:50 | Permalink

    Cool stuff!

    I’ve had an experience of debugging with all of the types of physical media and I have to say that since net debugging support in W8 I’ll never get back to any other type of media! You should definitely switch to net debugging ;)

  2. j00ru | 01-lip-13 at 07:04:45 | Permalink

    @Honorary BoT: what makes net debugging so great? :)

  3. pakt | 01-lip-13 at 07:19:11 | Permalink

    These are useful for reclaiming screenspace taken by gui elements:

    set guioptions-=m “remove menu bar
    set guioptions-=T “remove toolbar
    set guioptions+=c “console dialogs

  4. j00ru | 01-lip-13 at 07:22:46 | Permalink

    Nice, thanks pakt!

  5. tobi | 01-lip-13 at 13:30:40 | Permalink

    a) The vendor graphics drivers must contain an endless amount of severe security problems. It seems that the exploit community hasn’t really tried to analyze them yet (which is surprising – is it impossible to make money from driver exploits?).

    b) Exposing OpenGL via WebGL does not seem like a smart idea. Now you can root the box from a (Chrome-sandboxed) browser frame. And Google does not have the power to fix OpenGL exploits.

  6. Yuhong Bao | 01-lip-13 at 15:03:25 | Permalink

    I still remember the inflated cursor bug in ATI’s drivers that was covered on Tech Report.

  7. Honorary_BoT | 02-lip-13 at 00:01:31 | Permalink

    @j00ru: its the speed and better stability! I’ve watched some video on Channel9 about new kernel debugging transports (net and usb3). The guy mentioned that pre8 transports were artificially slowed down. As for USB3 – it is still difficult to find an XHCI with debugging port support.

  8. jusme | 02-lip-13 at 09:55:29 | Permalink

    are you saying that debugging via hardware-cables is faster than going via VMs and VKD?

  9. j00ru | 03-lip-13 at 03:50:03 | Permalink

    @tobi: right.
    a) I guess they may contain an endless amount of severe security problems, but it’s not certain if this is really the case (and also – how many of them are really reachable from a low-integrity process). There might also be several reasons for lack of public interest in graphics card drivers such as “they’re already widely exploited, but no one has realized yet” or “the market share of each individual driver is too limited to allow for usage in generic attacks” or “they’re more difficult to find and exploit compared to Windows itself” and so forth. There’s been some research done on it anyways, e.g. see Ben Hawkes’ “Browser GFX Security” presentation from Ruxcon 2011.

    b) Well yeah, new features are usually added for improved usability/performance and not security. :) Theoretically, you could do that provided a very fortunate vulnerability in OpenGL, but I still think it’s much more realistic to be attacked through the regular use-after-free to execution within sandbox + OS kernel bug combo than with a sophisticated one-shot GL exploit. Let’s see how things unfold, though.

    @Yuhong Bao: Are you referring to this one: http://techreport.com/news/17819/radeon-hd-5000-cards-inflate-cursors? I haven’t seen that one before, nice find! :-D

    @Honorary BoT: Interesting! I must try it one day then, but honestly, I can’t imagine it being even better than 1394 (which is already 100% responsive and reliable for me thus far). Anyway, thanks for the recommendation :)

    @jusme: yep, that’s what I’m saying. Going via VMs is usually implemented via COM ports which makes things slow even if we’re operating within the same physical machine. VKD seems to help (sometimes), but I’ve had multiple issues using it in the past which was quite annoying.

  10. mohammad Modarres | 07-wrz-13 at 03:21:24 | Permalink

    hi,i have some of the mentioned issues with my graphic card exactly the same shapes appears every time and system jammed in every 15-20 sec ,the issues started when i installed the new windows 7 64bit with new Nvidia Driver,the card is 560 ti and since my job is doing heavy graphic related works(3d modeling)this is very painful issue.
    any idea how to fix this problem?

  11. j00ru | 10-wrz-13 at 05:33:09 | Permalink

    @mohammad Modarres: I think you should be best off writing to Nvidia directly. Unfortunately, I am not aware of any workaround for the problem.

  12. ender | 31-gru-13 at 04:32:22 | Permalink

    I remember seeing the same problem on my computer, most commonly when moving the cursor to/from virtual machines in VMWare Workstation (so I blamed it at first, until I noticed it happening in other windows) – but at least to me, it was immediately obvious that the cursor shape was just rendered at the wrong colour depth (maybe it helped that I use a very distinct pointer arrow shape).

  13. kek | 03-maj-14 at 04:39:36 | Permalink

    hehe, original PoC software made such a nice ‘demoscene’ effect, so I now take this bug as an awesome feature. Thank’s NVIDIA! ;D

  14. redbone | 17-sie-16 at 13:02:22 | Permalink

    Is there a way to hide the global cursor through WDDM?

{ 1 } Trackback

  1. […] you interested in reverse engineering? You might be interested in this walkthrough of a reverse engineering / debugging session. The problem starts when the intrepid blogger noted […]

Post a Comment

Your email is never published nor shared. Required fields are marked *