On an NVIDIA Jetson TK1 (and likely any other embedded/... board with a directly attached recent eMMC device), Anaconda's storage selection menu (at least in text mode) offers /dev/mmcblk0rpmb as a potential installation location. IIUC, the "rpmb" HW partition on an eMMC device is something special that wouldn't make a good installation location. Besides, it's tiny (~4MB on my ~16GB eMMC device). I believe that /dev/mmcblk0boot0 and /dev/mmcblk0boot1 are culled out for similar reasons; they certainly don't show up in Anaconda's storage list even though boot0/boot1 are the same size as rpmb. Here's part of a log from the installer run: 3) [!] Software selection 4) [!] Installation Destination (Processing...) (No disks selected) 5) [x] Kdump 6) [x] Network configuration (Kdump is disabled) (Wired (eth0) connected) 7) [!] Root password 8) [!] User creation (Password is not set.) (No user will be created) Please make your choice from above ['q' to quit | 'b' to begin installation | 'r' to refresh]: 4 ================================================================================ ================================================================================ Probing storage... Installation Destination [ ] 1) MMC SEM16G: 14.68 GiB (mmcblk0) [ ] 2) Generic SD/MMC Storage Card: 4096 KiB (mmcblk0rpmb) [ ] 3) Select all No disks selected; please select at least one disk to install to. Please make your choice from above ['q' to quit | 'c' to continue | 'r' to refresh]:
Aha, /dev/mmcblk0boot* are ignored by Blivet's devicetree.py:isIgnored() because they are read-only by default. I suppose the correct fix might be to enhance that function to ignore mmcblk\d+boot\d.
I've sent a github pull request to the upstream blivet repo to fix this: https://github.com/dwlehman/blivet/pull/5 Also changing this bug's component to python-blivet.