Please apply
Upstream commit ea23994edc4169bd90d7a9b5908c6ccefd82fa40
to kernel versions 4.14, 4.19, 5.4 and above.
Reason:
Commits c84a1372df929033cb1a0441fb57bd3932f39ac9 "md/raid0: avoid RAID0
data corruption due to layout confusion." and
33f2c35a54dfd75ad0e7e86918dcbe4de799a56c "md: add feature flag
MD_FEATURE_RAID0_LAYOUT" added handling of original and alternate
layouts of RAID0 arrays with members of different sizes. However they
introduced a regression: assembly of such RAID0 array fails if the
per-array or default layout is not defined even when the layout is
irrelevant and can be safely ignored. One common case is when the RAID0
array is composed of two members of different sizes because the disk or
partition sizes are slightly different. This patch aims to fix this
regression.
Newer versions of mdadm can set a per-array RAID0 layout but some stable
distributions such as Debian 10 ship an older version of mdadm which
does not handle RAID0 layouts and a kernel series (4.19.y) which now
includes the backported commits. As a result, assembly fails after the
kernel upgrade unless the default layout is defined with a kernel parameter.
Related Debian bug reports :
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=944676
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=954816