[pve-devel] applied: [PATCH installer v2 0/6] add 4kn support and small improvements

[Thomas Lamprecht]

On 11/27/19 5:06 PM, Stoiko Ivanov wrote:
> This patchset adds support for installing on 4kn disks.
> 
> v1->v2:
> * Dominik suggested to do it correctly - so i tried, which resulted in a
>   rather different structure of the patch :)
> * Instead of checking the logical blocksize before partitioning the info now
>   gets collected into the hd list, which the installer uses as source at
>   various stages
> * This makes it possible to do the sanity checks during the disk-selection,
>   instead of while starting the install, which results in a better experience
> * 2 further unrelated patches were added:
>   * the creating of a fresh partition table now happens after all sanity checks
>     which don't need it, meaning we don't wipe an existing partition table,
>     just to error out directly afterwards
>   * all vdevs are now created with the by-id links (until now only the
>     bootable first vdev got created with by-id links, the rest remained as
>     sdX
> * Did some more testing (also in legacy mode) - learned that grub refuses to
>   install, if any 4kn disk is part of the zpool (even if it's not in the first
>   vdev/bootable)
> 
> Thanks for the feedback!
> 
> original cover-letter:
> The catchy part for me was the neccessary parameter -s1 for mkfs.vfat
> (the installer ran through, but ovmf (and I guess many UEFI implementations
> out there) did not boot from the esp.
> 
> Points I'm not too sure about:
> * should we still create an empty 1M partition on 4k disks (we cannot use it
>   to install grub with target i386 anyways) for more consistency in `lsblk`
>   outputs and the like
> * should we still install grub with target i386 on 512b sector disks in
>   setups that mix 4kn and 512e disks?
> 
> Feedback welcome!
> 
> Stoiko Ivanov (6):
>   expand comment on /sys/block/$dev/size
>   add logical_blocksize to hd_list infos
>   linewrap bootdevinfo generation
>   fix #1211: allow install on 4kn disks
>   wipe partitiontable after early exits
>   use by-id paths for all vdevs on pool creation
> 
>  proxinstall | 116 +++++++++++++++++++++++++++++++++++++++++-----------
>  1 file changed, 92 insertions(+), 24 deletions(-)
> 

applied, added Fabians requested comment in a followup and tried to massage
the commit-message info about the dosfstools bug into your 4/6 commit.

Thanks!