[andreas.reichel.ext@siemens.com: Re: qemu test with start_vm fails]

[andreas.reichel.ext@siemens.com: Re: qemu test with start_vm fails]

[Andreas Reichel]

Hi,

short interim result:

multiconfig:qemuarm-stretch:isar-image-base:

ROOTFS_EXTRA ?= "64"

is set correctly.

/meta/image.bbclass calculates the size of ROOTFS,
giving:

720332 KiB for the rootfs without the 64 extra and
785868 KiB with the ROOTFS_EXTRA.

diff 65536 KiB (correct.)

This number is identical between inside the docker and outside the
docker with manual 'du -s --block=1k'.

[By the way, man-page does not encourage use of small letter 'k'.
It states that 'K=1024' and 'kB=1000', whereas it seems that 'k'
equals 1024 luckily.]

The generated .ext4.img file has exactly 804728832 bytes,

giving $((804728832 / 1024)) == 785868, identical to the requested size.

However, mounting this file and using df gives:

/dev/loop0	757116 1k-blocks,  721844 1k-blocks used, 0 1k-blocks
available, Use 100%.

So if everything is built correctly and all calculations are correct,
how can it be that in the CI, this problem did not arise?

Again, when I use this file and start qemu, systemd hangs with infinite
timeout. When I enable journald output to console, i run out of disk
space. I could only boot with root ro, which is not set in the start_vm
script so CI test should run with root rw. This is illogical.

vG
Andreas

-- 
Andreas Reichel
Dipl.-Phys. (Univ.)
Software Consultant

Andreas.Reichel@tngtech.com, +49-174-3180074
TNG Technology Consulting GmbH, Betastr. 13a, 85774 Unterfoehring
Geschaeftsfuehrer: Henrik Klagges, Dr. Robert Dahlke, Gerhard Mueller
Sitz: Unterfoehring * Amtsgericht Muenchen * HRB 135082