[CRIU] Dumping a FUSE mount

Pavel Emelyanov xemul at parallels.com
Mon Oct 12 10:02:38 PDT 2015


On 10/12/2015 06:54 PM, Jonathan 'Wolf' Rentzsch wrote:
>> On Oct 12, 2015, at 3:00 AM, Pavel Emelyanov <xemul at parallels.com> wrote:
>> OK, now this lines in log:
>>
>>> (00.322206)      1: Reading mountpoint images
>> ...
>>> (00.322533)      1: 		Getting root for 626
>>> (00.322536)      1: 		Getting mpt for 626 ./app/available_modules
>>> (00.322540)      1: 		Getting source for 626
>>> (00.322543)      1: 		Getting opts for 626
>>> (00.322546)      1: 	Read 626 mp @ ./app/available_modules
>>
>> mean that the mountpoint entry in the image is not marked as external.
> 
> Sharp eye. I missed that.
> 
>> Can you show the dump log as well?
> 
> Attached, thanks for taking a look at this. Really helps.

This log is strange... and that's why:

> -Wolf
> 
> root at ip-172-31-46-108:/tmp/checkpoints/6533f087751536cd506bc1c98a8a5984/3d74f66b2398232ad415d99fc8bc7be195affd3e at 2/checkpoint-work# cat dump.log 
...
> (00.004068) 	type fuse.dupe source dupe mnt_id 0x272 s_dev 0x21 root / mountpoint ./app/available_modules flags 0x200001 external false options user_id=0,group_id=0

Here's the log-line for collected mountpoint. But I cannot find the place in 1.5.2
code where it is printed. The message is 90% alike the one from parse_mountinfo(), 
but the latter doesn't print the "external" word. Did you patch the message? May I
see the overall patch you use?

But let's go next.


> (00.004295) Found /etc/hosts mapping for ./etc/hosts mountpoint
> (00.004297) Found /etc/hostname mapping for ./etc/hostname mountpoint
> (00.004299) Found /etc/resolv.conf mapping for ./etc/resolv.conf mountpoint

Above is the external mounts resolving. No message for our one. Which means
that the mountpoint is

a) fsroot_mounted()
b) fstype points to unsupported entry

so in validate_mount() that prints the above messages it should report -1.

Would you patch the validate_mount() to see which mountpoints it validates, which
branches it travels and which results gets?

-- Pavel



More information about the CRIU mailing list