Description of problem: Anaconda runs into numerous selinux problems with a custom install. You perform a NFS install, with a custom layout and you can see many selinux errors from one of the background consoles. Version-Release number of selected component (if applicable): Fedora Core 5 Test 2 How reproducible: Always Steps to Reproduce: Choose a NFS install, custom layout, select packages, then wait about 5 minutes into the install, and watch the tty console fill with selinux errors, then crash and reboot Actual results: here's an example: AVC denied {transition} for pid=610 comm="anaconda" name="libgcc_post_upgrade" dev=hda2 ino=1358789 scontext=system_u: system_r: anaconda_t:s0 tcontext=system_u: system_r: rpm_script_t:s0 tclass=process and many many more with name= write, connectto, search, execmem, use, depmod, etc. crashed installer, reboot, unusable system Expected results: sucessful install Additional info: I can't understand why selinux has to be used for the linux install
We have to have policy loaded so that we can set xattrs properly, but we're running in permissive mode, so none of the messages should be fatal. Can you provide the actual messages you got when it crashed?
I see, The problem is there is no error message at all, and even if there was there's hardly any time to see because it just reboots itself. I have had this problem with FC4 also, and to some degree FC3; all tested on different computers too. However in text mode it does sow an error: Install exited abnormally ... I can try and get the whole message, but it's didn;t seem to indicate anything that I could tell. Do you know if this could have something to do with an already known problem. The amount of packages selected to install seems like it might be a factor.
Since Core 5 Test 3 came out, the problem seems to have improved. I wonder what changes were made, it could provide some clue to where the source of the problem was. Also, with T3, it was interesting that my test installs would have better success with more ram being added, adding more than 256 MB of ram seemed to help the install go farther along before it rebooted. Here's some details: Text mode install, custom layout, 256MB or more, no swap space P-III 866, 20 GB HD All it used to say before the crash was: Sending termination signals... Done. ... Hope this helps...
We've fixed cases where ram usage was spiralling out of control post-test3, so this should be better for the release