8 ms·

I'm really starting to understand why jwz left software and bought a nightclub. swaylocker, the wayland screen locker I'm currently using has the exact same ar

by dmm 6y ago

I'm really starting to understand why jwz left software and bought a nightclub. swaylocker, the wayland screen locker I'm currently using has the exact same architectural flaw: crash the dialog and get access to the desktop. https://github.com/swaywm/swaylock/issues/162 https://github.com/swaywm/swaylock/issues/162 https://github.com/swaywm/swaylock/issues/158 https://github.com/swaywm/swaylock/issues/158 https://github.com/swaywm/swaylock/issues/10 https://github.com/swaywm/swaylock/issues/10

hedora 6y ago
Heh. I’ve seen Wayland proponents cite the “impossibility” of screen locking in X11 as a reason to rewrite the entire thing from scratch. I’ve also seen this argument from the systemd crowd (spoiler alert: the logind ecosystem is also unreliable).
- dmm 6y ago
  > Heh. I’ve seen Wayland proponents cite the “impossibility” of screen locking in X11 as a reason to rewrite the entire thing from scratch. Supposedly Gnome is better but I have no idea. EDIT: I also found this looking through github issues. It's not only a know issue, it's a WONTFIX: >> This also brings up a good point. If sway detects that swaylock (or whatever locking process is used) didn't exit cleanly, it should try and recreate it, so we don't just end up giving the a malicious user access to the session. > I don't think this is worth it https://github.com/swaywm/sway/issues/2031#issuecomment-392038949 https://github.com/swaywm/sway/issues/2031#issuecomment-3920...
  - mscarborough 6y ago
    Directly below the supposedly juicy "I don't think this is worth it" comment is a merged PR linked to this issue. In that PR they changed the render loop rather than implement the "not worth it" suggested solution. At no point was it marked as WONTFIX.
    - hpfr 6y ago
      Yes, they fixed the bug, which is good, but if you read jwz’s post, the issue is that these bugs don’t fail safe. A screen locker shouldn’t crash to an unlocked session. The design of a screen locker should take into account that bugs and crashes will exist and strongly prioritize failing safely.
      - shawnz 6y ago
        Actually, Xscreensaver doesn't fail safe either. If it crashes, the same issue will occur. I don't know why the post implies otherwise. Jwz says it himself in the linked article: > But if xscreensaver crashes, the screen is unlocked, and our attacker is now logged in as the person who locked their screen. EDIT: To clarify I meant that jwz' blog post is the one which implies otherwise.
        dmm 6y ago
        I just said swaylock, a wayland screenlocker, had the same architectural flaw as all X11 screenlockers: when it crashes the screen unlocks. Wayland was supposed to be better but it isn't, apparently. Xscreensaver attempts to mitigate this by being as simple as possible to prevent crashes. swaylock doesn't do this either, apparently. Several crashes have been reported in the past year alone. I was just expressing disappointment in the lack of interest in fixing #1.
        josefx 6y ago
        His approach was basically to leave out complex features that could crash when possible and document why doing that was necessary. As far as he thinks a better solution is not possible with how Linux interacts with screen lockers. Modern implementations seem to do the inverse, include as much complexity as possible and fix every security critical exploit after the fact.
- marbu 6y ago
  Based on such arguments, I got an impression that screen saver and locking functionality is handled in wayland protocol, but when I tried to look it up, I haven't found it. It seems that wayland architecture can solve this only when a compositor directly contains screen locking code. And it also means that every wayland compositor have to do this separately (btw is there any reusable library covering this functionality?) and that independent wayland screen lockers (such as swaylocker) have the same problem as X server implementations.
jasonjayr 6y ago
My kids(all under 11yo) figured out how to buffer overflow the lock screen on our linux family games machine. It took a few times of trying to figure out if I had forgotten to lock it to figure out how they did it. They were bored and upset it was locked, and just kept typing characters into the dialog box for 5 minutes. After they were gently scolded for breaking the spirit of the rules, I praised them for being clever hackers, and explained what exactly they had done.
- jacquesm 6y ago
  This is one of the bad things about rust. It will take away these intermediate steps of learning how to hack stuff ;)
  - monocasa 6y ago
    Would the buffer index panic still crash the process and let them in? Albeit, it'd probably be a Vec instead of a slice.
maxerickson 6y ago
My impression is that he would make more money doing nothing than running the night club.

by dmm 6y ago

hedora 6y ago

Heh. I’ve seen Wayland proponents cite the “impossibility” of screen locking in X11 as a reason to rewrite the entire thing from scratch. I’ve also seen this argument from the systemd crowd (spoiler alert: the logind ecosystem is also unreliable).

dmm 6y ago
> Heh. I’ve seen Wayland proponents cite the “impossibility” of screen locking in X11 as a reason to rewrite the entire thing from scratch. Supposedly Gnome is better but I have no idea. EDIT: I also found this looking through github issues. It's not only a know issue, it's a WONTFIX: >> This also brings up a good point. If sway detects that swaylock (or whatever locking process is used) didn't exit cleanly, it should try and recreate it, so we don't just end up giving the a malicious user access to the session. > I don't think this is worth it https://github.com/swaywm/sway/issues/2031#issuecomment-392038949 https://github.com/swaywm/sway/issues/2031#issuecomment-3920...
- mscarborough 6y ago
  Directly below the supposedly juicy "I don't think this is worth it" comment is a merged PR linked to this issue. In that PR they changed the render loop rather than implement the "not worth it" suggested solution. At no point was it marked as WONTFIX.
  - hpfr 6y ago
    Yes, they fixed the bug, which is good, but if you read jwz’s post, the issue is that these bugs don’t fail safe. A screen locker shouldn’t crash to an unlocked session. The design of a screen locker should take into account that bugs and crashes will exist and strongly prioritize failing safely.
    - shawnz 6y ago
      Actually, Xscreensaver doesn't fail safe either. If it crashes, the same issue will occur. I don't know why the post implies otherwise. Jwz says it himself in the linked article: > But if xscreensaver crashes, the screen is unlocked, and our attacker is now logged in as the person who locked their screen. EDIT: To clarify I meant that jwz' blog post is the one which implies otherwise.
      - dmm 6y ago
        I just said swaylock, a wayland screenlocker, had the same architectural flaw as all X11 screenlockers: when it crashes the screen unlocks. Wayland was supposed to be better but it isn't, apparently. Xscreensaver attempts to mitigate this by being as simple as possible to prevent crashes. swaylock doesn't do this either, apparently. Several crashes have been reported in the past year alone. I was just expressing disappointment in the lack of interest in fixing #1.
      - josefx 6y ago
        His approach was basically to leave out complex features that could crash when possible and document why doing that was necessary. As far as he thinks a better solution is not possible with how Linux interacts with screen lockers. Modern implementations seem to do the inverse, include as much complexity as possible and fix every security critical exploit after the fact.
marbu 6y ago
Based on such arguments, I got an impression that screen saver and locking functionality is handled in wayland protocol, but when I tried to look it up, I haven't found it. It seems that wayland architecture can solve this only when a compositor directly contains screen locking code. And it also means that every wayland compositor have to do this separately (btw is there any reusable library covering this functionality?) and that independent wayland screen lockers (such as swaylocker) have the same problem as X server implementations.

jasonjayr 6y ago

My kids(all under 11yo) figured out how to buffer overflow the lock screen on our linux family games machine. It took a few times of trying to figure out if I had forgotten to lock it to figure out how they did it. They were bored and upset it was locked, and just kept typing characters into the dialog box for 5 minutes. After they were gently scolded for breaking the spirit of the rules, I praised them for being clever hackers, and explained what exactly they had done.

jacquesm 6y ago
This is one of the bad things about rust. It will take away these intermediate steps of learning how to hack stuff ;)
- monocasa 6y ago
  Would the buffer index panic still crash the process and let them in? Albeit, it'd probably be a Vec instead of a slice.

maxerickson 6y ago

My impression is that he would make more money doing nothing than running the night club.