Richacl and ext4 vs. xfs

[Steve French]

On Fri, Nov 4, 2016 at 4:15 PM, Andreas Gruenbacher <agruenba at redhat.com> wrote:
> On Fri, Nov 4, 2016 at 11:48 PM, Andreas Dilger <adilger at dilger.ca> wrote:
>> [added linux-ext4 to CC list]
>>
>> On Nov 4, 2016, at 1:19 PM, Steve French <smfrench at gmail.com> wrote:
>>>
>>> Andreas,
>>> I had been working on getting some VMs together to experiment with a
>>> few different Richacl configurations and it looks like the e2fsprogs
>>> git hub repo with richacl enhanced ext4 utilities may need to be
>>> updated or have an incompatibility with current kernels (See e2fsprogs
>>> github link at http://www.bestbits.at/richacl/ ).  XFS worked ok.
>>>
>>> Fedora 24 Workstation, created two extra disks
>>> /dev/sdb for xfs
>>> /dev/sdc for ext4
>>>
>>> Built and installed Andreas's 4.9.0-rc1 based kernel richacl branch
>>> and the other repos (xfsprogs, richacl etc.).    Had no problem
>>> formatting and using /dev/sdb1 as xfs, but once I installed e2fsprogs
>>> from Andreas's tree I couldn't boot the system (removed e2fsprogs and
>>> the ext4 partition and booted fine).  With Andreas's e2fsprogs I
>>> couldn't boot 4.9.0-rc1
>>
>> What version of e2fsprogs is (the other) Andreas' tree based on?  The
>> newer e2fsprogs-1.43 has support for metadata checksums, but this should
>> be off by default in the released version.  The testing version had it
>> enabled by default.
>
> The richacl patches were based on v1.42.12+. They only define a new
> "richacl" feature flag, so I don't think they are to blame for the
> failure Steve is seeing. I've still rebased the patches onto v1.43.3+
> now.
>
> Steve, I assume that /dev/sdc is not the root filesystem or listed in
> /etc/fstab; if so, then the device will have no influence on booting.
/dev/sdc should have no impact on booting (it was empty)

/dev/sda1 was /boot and ext4
/dev/sda2 was /
/dev/sdb1 was xfs formatted with richacl and his xfsprogs (mounted a /xfs)
installing e2fsprogs and formatting /dev/sdc1 (mounted as /ext4) was
ok until the reboot when it


> Does the richacl kernel use an initrd? If so, then please check for
> mkinitrd errors: I've seen versions of mkinitrd (dracut) that use a
> blkid option that isn't part of e2fsprogs v1.42. Not sure if v1.43 has
> this fixed or if this is a distro specific problem.

I will check


-- 
Thanks,

Steve