Description of problem: There are some UEFI and maybe kernel bugs that cause the efibootmgr command to write NVRAM entries to fail. When that fails, anaconda fails to complete installation tasks. Because grub and shim packages have a fallback contingency for no NVRAM entry, we should move the anaconda issued efibootmgr command later than present, pretty much the last possible step. That way the root and user configurations are written, /etc/default/grub can be written, grub.cfg produced, and so on. So even if the NVRAM write fails, chances are the system can still boot. Present behavior, it will not boot without a lot of manual intervention because so many post install things failed to complete. Version-Release number of selected component (if applicable): anaconda-20-25-1 How reproducible: Always, *if* there is failure to write NVRAM entry via efibootmgr. Steps to Reproduce: 1. 2. 3. Actual results: If NVRAM write fails, anaconda stops, no post install completes, system not bootable. Expected results: Do all the other stuff first, including post install, and cause efibootmgr to run after all of that. Then if it fails, we can probably still boot (and if not, there's still no harm done). Additional info: Bug 1006304 Bug 1006304
Add Bug 1047993 where I report the efibootmgr fails due to kernel problems.
This bug can probably be closed, see bug 1006304 comment 41 (and 11).
Closing per comment 2