| Summary: | Part 1 of the MASTERCRASH created by ToBeFree | ||
|---|---|---|---|
| Product: | [Unmaintained] kscreensaver | Reporter: | Unknown <null> |
| Component: | locker | Assignee: | kscreensaver bugs tracking <kscreensaver-bugs-null> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | crash | CC: | mpyne |
| Priority: | NOR | ||
| Version: | 2.0 | ||
| Target Milestone: | --- | ||
| Platform: | Ubuntu | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
|
Description
Unknown
2011-10-23 21:41:27 UTC
I found out which option causes the bug. It is: bleachbit -d kde.cache I found out which option causes the bug. It is: bleachbit -d kde.cache it would seem that KSharedDataCache shares a bit too much ... like, with other users. :} alternatively, the problem could be also in a higher layer, like KIconLoader. though that would be bizarre. "where full" instead of "bt" in the debugger could help with that. or you found a kernel bug. are you *sure* the files of the two users weren't (sym-)linked in any way? did you isolate which option made apt-get crash? as ths is a security-sensitive matter, i would expect a bit more discretion next time. please inform yourself about responsible disclosure. "as ths is a security-sensitive matter, i would expect a bit more discretion next time. please inform yourself about responsible disclosure." <-- I'm sorry, but I created the report using the KDE bug reporting tool - and there was no option to make this report secret or something like this. :/ I am not sure if there were symlinks. When I found the bug, I used another session (Ctrl+Alt+F1) and entered the command. When I came back to the F7 KDE session, everything crashed. Also, if I run the bleachbit-gui in that session, lock the screen and return after some time, everything including the screen lock crashes. I was absolutely sure that it also works with every other user running the command because I tested it in another session and on other computers multiple times. Now I created another user and did the same thing. Whatever caused the bug/crash before - I was pretty surprised that it did not work again. O.O I am no longer sure that it works using another user. Maybe this makes the bug security-irrelevant, but it exists and it is at least VERY annoying. I hope that I did not waste your time with this supposedly security-sensitive bug. Nevertheless, this bug _exists_, even if it does not work as another user. A bug that crashes _every_ program I currently use is always bad. By the way, should I mark the other parts of this bug as duplicates of this one? (In reply to comment #4) > there was no option to make this report secret or something like this. :/ > well, yeah. but maybe that means that bugzilla simply isn't the right address to report security problems? we have security@kde.org for that. maybe the bug reporting page should make that clearer. > I am no longer sure that it works using another user. > ok, please triple-check. > it exists and it is at least VERY annoying. > well, one can argue that using bleachbit on running applications is the bug here. you are essentially destroying their data structures under their feet. mmap is known to be fragile, so this is no big surprise. we could catch sigbus and handle errors resulting from accesses to our maps, but this is going to be ugly, in particular because signal handling in library code is not particularly popular (qprocess has the same problem). > By the way, should I mark the other parts of this bug as duplicates of this > one? > yes, it's all the same problem. "invalid - multiple submissions" would be even better in this case. FTR: find . -perm /022 in ~/.kde/cache-localhost or ~/.kde/tmp-localhost returns two files with write permissions on the user group (ie, owner group, NOT "users") - non for others and the dirs are drwx------ - no sticky bit "solution". It's likely either a (local/distro?) config issue or "i can 'Clear the memory and swap on Linux'" snake... "bleachbit" raises privileges - try "stat `which bleachbit`" and check whether the first acces line has sort of a (4755/-rwsr-xr-x) part (important are the leading 4 and the "s" in the first rwx block) bleachbit is a python script and thus so totally not setuid. BleachBit after version 0.8.7 whitelists /var/tmp/kdecache-* to avoid overwriting/"shredding" the file data before deleting it. A better "MASTERCRASH" is to send SIGABRT to a given process, since we're assuming we can already run arbitrary commands... *** This bug has been marked as a duplicate of bug 271889 *** As an aside be sure to unlink (*not* shred) the *.kcache files under /var/tmp/kdecache-$USER before starting KDE next time to clear out your corrupted caches. |