Re: make-bcache bug?

[Kai Krakow <hurikhan77@gmail.com>]

Am Mon, 15 May 2017 17:58:48 -0700
schrieb Marc MERLIN <marc@merlins.org>:

> On Mon, May 15, 2017 at 08:37:45PM +0200, Kai Krakow wrote:
> > Am Mon, 15 May 2017 13:52:09 +0100
> > schrieb Nix <nix@esperi.org.uk>:
> >   
> > > On 14 May 2017, Kai Krakow said:  
>  [...]  
> > > 
> > > udev does a blkid to see how the block device needs to be
> > > activated: this relies on precisely the information wipefs
> > > removes. Avoiding this problem is why wipefs *exists*. :)  
> > 
> > Yes, but something triggered udev when it shouldn't...
> > 
> > This ends up in unregistering/stopping the bcache, then wipefs the
> > cdev, then look if udev was triggered and eventually stop it again.
> > 
> > I guess everytime I ran "fdisk -l" to double-check the devices, or
> > "lsblk", or "blkid", udev was triggered and re-enabled the bcache.
> > This is a race you cannot win. ;-)  
> 
> Note that this is orthogonal to the problem I reported.
> It tells me 
> Already a bcache device on /dev/sde2, overwrite with --wipe-bcache
> when it fact this does not work, apparently ever.
> 
> So either the message gets changed, or 
> make-bcache  --wipe-bcache
> -C /dev/sde2 Device /dev/sde2 already has a non-bcache superblock,
> remove it using wipefs and wipefs -a does not happen. there was no
> non-bcache superblock.

I look at both source codes of wipefs and make-bcache should give a
clue.

Since you made me curious, I'll probably take a look this weekend.


-- 
Regards,
Kai

Replies to list-only preferred.