Three New Security Advisories Hit X.Org's X11 Library

[phoronix]

Phoronix: Three New Security Advisories Hit X.Org's X11 Library

It's been a while since last having any big security bulletins for the X.Org Server even though some of the code-base dates back decades and security researchers have said the security is even worse than it looks and numerous advisories have come up in recent years. But it's not because X11 is bug-free as today three more security bulletins were made public affecting libX11...

http://www.phoronix.com/scan.php?page=news_item&px=Three-New-LibX11-Security-Holes

[uid313]

There is libxcb which is more modern than libx11.
But many widget toolkits, window managers, applications, libraries, etc never migrated over to libxcb and still use libx11.

[treba]

Wayland compositors won't be bug free either, but x11 exposes so many functions that I'd expect them to be much cleaner. Hopefully we will soon be able to run wayland session without xwayland (do I get this right that this can also get exploited in an wayland session with xwayland?)

[R41N3R]

It would be great to see broader support for Wayland from distributions and application developers. At the moment it still feels strange if you report bugs that are related to xcb dependencies or to missing Wayland support. And in several applications they are either slowly addressed or not at all. It is like everybody is waiting for somebody else to fix something for them. Even in XWayland 1.20.1 are quite a few bugs that create issues in Vulkan or Wine games. I understand that new features take time to develop and that it is more fun to develop them, but I would expect at least that "real" bugs get addressed faster. I do use Wayland with Plasma 99% of my time, but there are still several glitches that need to be addressed and then there are applications like Firefox, Chromium, Steam, Wine, SDDM, ... that still have no Wayland support.

[NotMine999]

And if it was all rewritten in RUST it would not have any bugs, right? /sarcasm

[jpg44]

Wayland compositors won't be bug free either, but x11 exposes so many functions that I'd expect them to be much cleaner. Hopefully we will soon be able to run wayland session without xwayland (do I get this right that this can also get exploited in an wayland session with xwayland?)

This is a myth, the API used by a Wayland app is huge, and most of it is actually an exposed driver API from a driver that is mapped into application address space. We are talking a large OpenGL API. This also means apps are communicating directly with video hardware, whereas X.org had the X server exclusively handle the hardware. This could expose hardware bugs to apps and flaws that are harder to fix than software bugs. Wayland's not cleaner. Ive used all the APis, and its not, its actually quite a bit more dangerous than the X way of doing things.

The idea that Wayland is intrinsically safer is so opposite of reality that its jaw dropping and frustrating that people think this. The X model has a cleaner seperation between applications and the video hardware and is actually a much safer model.

[jpg44]

It would be great to see broader support for Wayland from distributions and application developers. At the moment it still feels strange if you report bugs that are related to xcb dependencies or to missing Wayland support. And in several applications they are either slowly addressed or not at all. It is like everybody is waiting for somebody else to fix something for them. Even in XWayland 1.20.1 are quite a few bugs that create issues in Vulkan or Wine games. I understand that new features take time to develop and that it is more fun to develop them, but I would expect at least that "real" bugs get addressed faster. I do use Wayland with Plasma 99% of my time, but there are still several glitches that need to be addressed and then there are applications like Firefox, Chromium, Steam, Wine, SDDM, ... that still have no Wayland support.

Porting stuff to Wayland is overrated, for low bandwidth apps a waste of time. if you are an app maintainer, please do not target Wayland directly and instead go through Qt or Gtk/Cairo, so people who want to use X11 will not end up not being able to use your app. The model Wayland is actually worse security wise than X because you are exposing the video hardware directly to applications and loading drivers as a library into the App. X has a clean separation between the app and the video hardware, only the X server is allowed to touch the video hardware. All software has bugs and I fear wayland will have them too, and it could be pretty ugly, also hardware bugs end up being exposed to the apps because the app has direct contact with the video hardware. This could get pretty ugly if a hardware bug is found.

[dkasak]

This is a myth, the API used by a Wayland app is huge, and most of it is actually an exposed driver API from a driver that is mapped into application address space. We are talking a large OpenGL API. This also means apps are communicating directly with video hardware, whereas X.org had the X server exclusively handle the hardware. This could expose hardware bugs to apps and flaws that are harder to fix than software bugs. Wayland's not cleaner. Ive used all the APis, and its not, its actually quite a bit more dangerous than the X way of doing things.

The idea that Wayland is intrinsically safer is so opposite of reality that its jaw dropping and frustrating that people think this. The X model has a cleaner seperation between applications and the video hardware and is actually a much safer model.

Heh. This is pretty confused. What gives you the idea that wayland apps "are communicating directly with video hardware"? How do you explain the fact that you can run a wayland compositor without hardware acceleration? Wayland cuts out the libraries that no-one was using any more ( because modern apps do all their rendering themselves ), and has clients pass buffers of rendered content to the compositor. The compositor then renders all clients as it sees fit. Nothing in wayland forces either clients or compositor to "use graphics hardware directly" - other than the obvious requirement of a method for the compositor to actually render full-screen frames, whether via a simple frame buffer, or OpenGL or Vulkan, or some other method.

[Zan Lynx]

Wayland compositors won't be bug free either, but x11 exposes so many functions that I'd expect them to be much cleaner. Hopefully we will soon be able to run wayland session without xwayland (do I get this right that this can also get exploited in an wayland session with xwayland?)

Wayland and XCB avoid some issues by using machine generated interface code. The protocol ABI is defined in XML.

And you can already run Weston for Wayland without X11. It's just that there's a surprising amount of stuff that won't work.