Bug 89748

Created attachment 114590 [details]
Add missing xselinux hook to shm PIXMAP resource creation

I spent some time trying to track down an X segfault and believe I have found a Bug in the file "Xext/shm.c" where it is missing an Xace hook initialization call which leads to a NULL pointer in the xselinux code later when something tries to access the resource which was never initialized. 

A patch is attached which makes the segfault go away and I wanted to share it to see if others believe it made sense.The segfault I was seeing was in the function "SELinuxDoCheck" in the file "Xext/selinux_hooks.c" where "obj->sid" would be "NULL".

It would probably make sense to try to add better error handling to the xselinux code to catch the case where something attempts to access a resource which has not received xselinux initialization.

Below are the reproduction steps to reproduce my initial problem without the patch. 


Steps: (Any computer which has hardware that supports the fglrx driver)
-------------------------------
1. Ensure that 2 monitors are connected to video card
2. Perform a standard "Desktop" CentOS 6.6 i386 install using the standard ISO: "CentOS-6.6-i386-bin-DVD1.iso"
3. Once fully installed - login, "su" to root and run "yum update"
4. Once the update finishes, then run "yum install kernel-devel gcc"
5. Download the the fgrlx driver from http://www2.ati.com/drivers/linux/amd-catalyst-omega-14.12-linux-run-installers.zip
	NOTE: older fglrx drivers exhibit the same issue
6. Unzip the driver to expose the "amd-driver-installer-14.501.1003-x86.x86_64.run" script.
7. Reboot the system with "shutdown -r now"
8. At the bootloader press "a" twice and then append "3" to end of the kernel parameters to boot the system into runlevel 3.
9. Login as root and run the fglrx installer "sh amd-driver-installer-14.501.1003-x86.x86_64.run" script. 
10. After the fglrx driver installer successfully completes then reboot the system with "shutdown -r now".
	NOTE: fglrx installer should ask for a reboot after completion anyways
11. At the bootloader press "a" twice and then append "3" to end of the kernel parameters to boot the system into runlevel 3.
12. Login as root and run the following command "aticonfig --initial=dual-head --xinerama=on"
        NOTE: this sets up a dual head xinerama configuration, the use of Xinerama is critical to reproduce the issue
13. Run "setsebool -P xserver_object_manager on" as root
        NOTE: This is critical to reproduce the issue
        NOTE: The setsebool command will take awhile to run as it needs to make the boolean permanent and rebuild the selinux policy, so that the change will be in place after reboot
14. Reboot the system with "shutdown -r now".
15. Let the system boot normally. The X Server will not start properly. If you checkout the Xorg and/or gdm logs then you will see a backtrace of the segfault.