Description of problem: This happens on Rawhide-2008-04-09, which has error in installing selinux-policy-targeted. Version-Release number of selected component (if applicable): selinux-policy-targeted-3.3.1-31.fc9.noarch policycoreutils-2.0.46-2.fc9.i386 libsemanage-2.0.24-1.fc9.i386 How reproducible: I've seen it each time I looked (twice) Steps to Reproduce: 1. Install Rawhide-2008-04-09 (that might be important) 2. user login on X11, logout. (which sometimes leaves Xorg :0 tty7 running) 3. ps -ef | grep [u]serlogin; secon --pid 4. tty1# semanage user -m -R unconfined_r user_u 5. ps -ef | grep [u]serlogin Actual results: 3. bonobo-activation-server had context user_u user_r user_t 4. semanage got error message libsemanage_dbase_llist_query: could not query record value (no such file or directory), after a long delay (maybe a minute). 5. the process context was system_u object_r unlabeled_t. Expected results: Not this. I imagine the process might have elevated privileges now. Additional info: install.log: Installing selinux-policy-targeted-3.3.1-31.fc9.noarch /usr/sbin/semanage: You must specify a prefix /usr/sbin/semanage: You must specify a prefix
My notes also say on first attempt to change user_r to unconfined_r, system crashed hard, no keyboard response even to shift-lock, no ping response, power-off button which usually triggers shutdown was ignored, reset required. On reboot, the role was changed. This happened on both systems I had installed 04-09 on. I am repeating install to see if the crash is repeatable. Presumably the fixes to targeted policy might mask the symptoms but I don't think it should be possible to crash the system just because s-p-t-3.3.1-31 is installed, and this probably reflects a different bug. Any pointers to instructions for setting up serial port kernel debugger to get the crash state?
Repeated install on two systems. The crucial factor which seems necessary for a crash is that the user is logged in. If no user process is active, there is no crash. If user is logged in, the crash occurs a short (but not zero) time after the dbase_llist_query no such file or directory. It is definitely a hard crash: shift-lock, num-lock, scroll-lock lights do not respond, mouse tracking stops, ping response stops, all at the same time.
dan, although clearly not impossible that this was an actual kernel crash more likely is that system_u:object_r:kernel_t became and invalid type and all of the sudden darn near everything went boom. To collect a kernel crash though attach a serial console and add console=ttyS0,9600 console=tty0 to the boot line. All kernel meesage will them go out the first serial port and to the console...
Did we get any messages from the console? Anything relevant you could find in /var/log/messages dmesg or /var/log/audit/audit.log
Fixed in latest policy Fixed in selinux-policy-3.3.1-40.fc9.noarch