Freespacenotifier KDED module in kdereview

[Ingo Klöcker]

On Monday 17 May 2010, Lubos Lunak wrote:
> On Monday 17 of May 2010, Jacopo De Simoi wrote:
> > I've a question about this module, as far as I can understand it
> > polls  the disk on which $HOME is mounted every minute to check
> > the available disk space. This is something we carefully avoided
> > to do in the plasma device notifier because by doing so we keep
> > the disks awake instead of letting them sleep if unused if the
> > user desires so. This module would probably upset laptop users, am
> > I right? (ervin?)
> 
>  Do you have some more details on this? I tried to keep my HDD spun
> down after I had bought SSD and there was always something that kept
> spinning it up, so eventually I've given up and I keep it unmounted
> until needed. As far as I know laptops have a special mode for
> allowing them to spin down the drive that does more aggressive
> caching and I don't understand why free space, which is something
> very easy to cache, should keep the actual drive active.
> 
> > I would personally  like to have the service disabled _at the very
> > least_ if running on batteries (you can get an idea of how to do
> > it by looking at how it's done for the strigi indexer),
> 
>  If we really need to do this by hand, could this be done somewhere
> in the framework? I can't imagine this would be the only place in
> KDE that'd need adjusting.
> 
> > and  to check available space less
> > often than every minute in the general case.
> 
>  No, and actually even a minute is not that short. The point of the
> service is to avoid running out of space in $HOME, which can lead to
> loss of data as a consequence of that. And running out of space is
> quite easy, a download running in background, compilation, whatever.
> Which means the service should ideally check all the time,
> everything (and that means doing it in KIO is not enough). And
> having it disabled by default makes including it pointless, how many
> users will turn it on? About the same number as those that turn on
> quotas.

IMO this needs to be handled on a much lower level probably at the same 
level as file/directory change notifications. We don't do polling to 
watch for file/directory changes anymore, right? Why should we do such 
nonsense for something even more simple?

The file system (or some kernel service or whatever) should allow to 
register a callback or whatever that is called/notified if the amount of 
free space falls below a certain threshold. I mean, the file system will 
surely know best at any time how much space is available so it is the 
natural place for such functionality. This solution would make polling 
completely unnecessary and at the same time prevent any problems with 
energy management of hard disks.

I propose to contact the Linux kernel people about this.


Regards,
Ingo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20100517/9a3e63d3/attachment.sig>