Systemd 230 Opens Up A New Graphics Vulnerability & FBDEV Still Should Die

[trek]

Completely untrue, but keep the bullshit coming!

CVE-2015-3164, CVE-2012-3524 and this phoronix article

Not sure what your beef with NT is either. You do realise the Windows NT kernel is pretty advanced as far as security is concerned, right?

yes, the kernel is really good, thanks to the DEC developers, but the final product is... nothing like Unix
or may be would you like Linux become more closer to the Windows experiece?

[SystemCrasher]

That's funny, every time this is raised with actual OSS GPU driver devs present, the conclusion is KMS drivers are really simple if you don't need things like HW accel. I don't know which level of HW accel said devices currently have

There is one problem: "desktop GPU" devs have very specific views on simplicity. Somehow mentioned HW is so simple it only could receive few commands over I2C or SPI. It got no notion of usual DRM things like CRTC, encoder or connector. So setting up fake abstractions which do not even exist in this HW could take more than rest of code. At which point helpers aren't so helpful, etc. Somehow Intel devs and somesuch never looked into SPI or I2C LCD drivers to get idea their notion of simplicity isn't really simple by such hardware standards. There were "SimpleDRM" patches trying to address this but they've been met with "desktop" GPU devs failing to get the point due to really different notion of simplicity.

[nanonyme]

Somehow, that's how software development works everywhere. Systemd isnt't unique in it. Root cause is the fact devs can't foresee everything, neither all use cases, nor all shortcomings, not to mention bugs. So if someone stumbles on bug or shortcoming, it is up to them to deal with it by default. That's how opensource works. OTOH nobody could force you to use systemd, Linux or whatever. You're perfectly free to use whatever. With some little caveat: it could turn out you have to maintain your favorite solution yourself if nobody else steps in. That's how it works.


Nothing wrong with it. Someone have to introduce changes, and it happens all the time. I do not get why systemd devs should be unable to do something like this. Are they second-class citizens, or what? Though in this particular case they did quite strange thing. Still, fbdev appears to be poorly maintained these days. Yet it is eventually used in embedded devices, so it would be nice if there is reasonable replacement. This is especially true for small simplest i2c/spi/... LCDs where full fledged KMS driver looks like an overkill. There were patches floating around to provide simplified layer on top of KMS but it seems they haven't made it so far.

That's funny, every time this is raised with actual OSS GPU driver devs present, the conclusion is KMS drivers are really simple if you don't need things like HW accel. I don't know which level of HW accel said devices currently have

[unixfan2001]

Just kill systemd please.

Marketing version: hey! X11 is old like dinosaurs and a security mess, you can even grab the screen if you are connected to the X server.
Thanks to god, we developed a cutting edge architecture composed by systemd to restrict who can run X11, wayland to restrict what the apps can do and dbus to let's them talk together. We end up to a system that works only in the original author environment and with worst security bugs: thanks to systemd any logged user can grab the screen, not only the ones connected to the X server, then you can execute arbitrary commands through dbus and anyone who can connect to the xwayland socket can do what they want without authentication!
So, what are you waiting for before to install our novel view of Unix NT?

Completely untrue, but keep the bullshit coming!
Not sure what your beef with NT is either. You do realise the Windows NT kernel is pretty advanced as far as security is concerned, right?

[unixfan2001]

This is exactly why I don't want systemd. They introduce a bug and expect others to fix it. Because they want systemd to do or don't do something the whole world has to change. Why not let systemd die and keep fbdev?

1. It's not a bug.
2. It's not systemd's fault.
3. The article is clickbait.

[pal666]

But I have a sneaky suspicion that it doesn't have "linux correct" support

it has no maintainer and it is broken

[cjcox]

not unless you provide patches

So I guess what everybody is effectively saying is that fbdev has no support. Then I guess it has to die. But I have a sneaky suspicion that it doesn't have "linux correct" support and therefore it must die (the special kind of support that is deemed "good" by the few that make those types of decisions). Sort of like Reiser 4.

[pal666]

Why not let systemd die and keep fbdev?

because all smart people are developing systemd and all idiots can only post shit on forums and can't fix their fbdev

[pal666]

Before we kill anymore functionality, can we make sure that there are viable pathways that actually work for everyone?

not unless you provide patches

[trek]

Just kill fbdev please.

Just kill systemd please.

Marketing version: hey! X11 is old like dinosaurs and a security mess, you can even grab the screen if you are connected to the X server.
Thanks to god, we developed a cutting edge architecture composed by systemd to restrict who can run X11, wayland to restrict what the apps can do and dbus to let's them talk together. We end up to a system that works only in the original author environment and with worst security bugs: thanks to systemd any logged user can grab the screen, not only the ones connected to the X server, then you can execute arbitrary commands through dbus and anyone who can connect to the xwayland socket can do what they want without authentication!
So, what are you waiting for before to install our novel view of Unix NT?