Re: [PATCH 0/8] Assorted patches

[Alexander Smirnov <asmirnov@ilbers.de>]

On 02/19/2018 05:08 PM, Jan Kiszka wrote:
> On 2018-02-19 14:56, Alexander Smirnov wrote:
>> On 02/16/2018 11:52 AM, Jan Kiszka wrote:
>>> Flush of my remaining queue:
>>>    - improvements for mount and umount
>>>    - arm64 target
>>>    - copyright and credit enhancements
>>>
>>
>> CI failed on wheezy, log is attached. I saw the same problem when mount
>> sys/dev via '--bind'.
>>
> 
> Is that the error?
> 
>> Setting up initscripts (2.88dsf-41+deb7u1) ...
>> /bin/rm: cannot remove `/dev/shm': Device or resource busy
> 
> Hmm, seems someone would like to wipe /dev content here. That's a
> warning sign that we definitely should not expose a host /dev to the chroot.
> 
> What else can we do here? Copy /dev over?
> 

The issue here is in following:

  - initstripts.postinst performs check whether we are under chroot
  - then it performs lots of checks:

8<--
		# Note trailing slash for mountpoint dir needed to
		# resolve mountpoints where the dir is a symlink.
		DEV=d
		DEVSHM=d
		RUN=d
		RUNSHM=d
		if mountpoint -q /dev/; then
		    DEV=m
		fi
		if mountpoint -q /dev/shm/; then
		    DEVSHM=m
		fi
		if [ -L /dev/shm ]; then
		    DEVSHM=l
		fi
		if mountpoint -q /run/; then
		    RUN=m
		fi
		if mountpoint -q /run/shm/; then
		    RUNSHM=m
		fi
		if [ -L /run/shm ]; then
		    RUNSHM=l
		fi

		RUNSTATE="${DEV}${DEVSHM}${RUN}${RUNSHM}"
		RUNACTION="NONE"
8<--

  - then according to this is performs complex case:

8<--
		case "$RUNSTATE" in
8<--

So changing of mount way in Isar, somehow changes the flow of this 
script, so it tries to re-create /dev/shm. That's all I know about this 
issue.

Alex