I'm not sure this is an anaconda bug, but need to file it somewhere so it doesn't get lost, feel free to move as appropriate. When installing rawhide onto ext4dev, the root inode does not get an selinux security context label. This ultimately causes the system boot to fail post-installation. Simply running restorecon on / fixes it up. I haven't yet sorted out *what* should be labeling / during install; perhaps the recent "fixfiles" fix will fix it. I'll test that today.
Yes I think the fixed fixfiles should fix this problem
Putting in MODIFIED on the basis of comment #1.
Something else is still wrong... at the end of the install, ls -Zd /mnt/sysimage shows root_t... but after the box reboots, it's file_t again. It appears it's not getting flushed. ls -Zd /mnt/sysimage/boot (separate filesystem...) shows boot_t before & after. I notice that at the end of the install, /mnt/sysimage is still held busy via files in /mnt/sysimage/var/lib/rpm... perhaps it's not getting unmounted cleanly? This may be a kernel problem unique to ext4 I suppose; maybe journaling and/or flushing problems....
Funky, the root attr is there but it's not reported/returned: [root@magnesium tmp]# mount | grep sda9 /dev/sda9 on /mnt type ext4dev (rw) [root@magnesium tmp]# ls -Zd /mnt drwxr-xr-x root root system_u:object_r:file_t:s0 /mnt (unlabeled) [root@magnesium tmp]# getfattr -d -m "^security\\." /mnt [root@magnesium tmp]# <nada> [root@magnesium tmp]# debugfs /dev/sda9 debugfs 1.40.4 (31-Dec-2007) debugfs: stat . Inode: 2 Type: directory Mode: 0755 Flags: 0x0 Generation: 0 User: 0 Group: 0 Size: 4096 File ACL: 0 Directory ACL: 0 Links: 23 Blockcount: 8 Fragment: Address: 0 Number: 0 Size: 0 ctime: 0x479ea01f -- Mon Jan 28 21:40:15 2008 atime: 0x479ea020 -- Mon Jan 28 21:40:16 2008 mtime: 0x479ea01f -- Mon Jan 28 21:40:15 2008 Size of extra inode fields: 28 Extended attributes stored in inode body: selinux = "73 79 73 74 65 6d 5f 75 3a 6f 62 6a 65 63 74 5f 72 3a 72 6f 6f 74 5f 74 3a 73 30 00 " (28) BLOCKS: (0):816 TOTAL: 1 (that nice hex string should be printed (that's another bug...) - it's "system_u:object_r:root_t:s0" so looks like the attr is there, but we're not fetching/reporting it. Grr. Well, will fix this up tomorrow I hope.
Found it. Can fix it a couple ways; will go to the ext4 list & see which people like best. -Eric
Patch is in the upstream ext4 git tree; will be heading to linus shortly. Applied to rawhide kernel.