Artem,

We moved the tmp mic dir from /var/tmp/mic/cache to /var/run/mic to be
faster: this greatly reduces the image creation time, specially the rpm
install phase.

But... /var/run is mounted on tmpfs... And it's probably why mic fails
to generate the bmap file (FIEMAP fails). If we change the temp dir to
ext4, it works (but is slower :-))

Thanks for pointing the problem.

-- 
Stéphane Desneux
Intel OTC - Vannes/FR
gpg:1CA35726/DFA9B0232EF80493AF2891FA24E3A2841CA35726

On 15/01/2014 08:32, Artem Bityutskiy wrote:
> On Tue, 2014-01-14 at 19:21 +0100, Stéphane Desneux wrote:
>> AFAIK, mic doesn't start bmaptool but "is compatible with" the bmaptool
>> principle. If we run 'bmaptool create xxx.raw', this doesn't work as mic
>> copies the temp sparse file to the output file (which is a regular one).
> 
> Yeah, MIC would need to do smart copy. Should be fixable.
> 
>> Some ideas ?
> 
> FIEMAP is a relatively new IOCTL in some file-systems. But not supported
> everywhere. What is your host FS? I only tested this with ext4 and
> btrfs.
> 
> There is an older and slower FIBMAP ioctl, which is more widely
> supported. I used it in the past, but then switched to FIEMAP and dumped
> FIMBAP altogether. I can easilly bring it back and fall-back to FIBMAP
> if FIEMAP is not supported.
> 
> So please, let me know what is your file-system, and I'll check v3.7 WRT
> what is supported there.
> 
> But there FSes where this won't ever work, unfortunately. E.g., FAT-FS.
> May be some networking FSes.
> 
_______________________________________________
Dev mailing list
[email protected]
https://lists.tizen.org/listinfo/dev

Reply via email to