Re: [PATCHv4 00/10] add on-demand device creation

[Minchan Kim <minchan@kernel.org>]

On Wed, May 06, 2015 at 02:25:57PM +0900, Sergey Senozhatsky wrote:
> 
> Hi,
> 
> On (05/06/15 14:01), Minchan Kim wrote:
> > Hello Sergey,
> > 
> > On Mon, May 04, 2015 at 09:38:52PM +0900, Sergey Senozhatsky wrote:
> > > We currently don't support zram on-demand device creation.  The only way
> > > to have N zram devices is to specify num_devices module parameter (default
> > > value 1).  That means that if, for some reason, at some point, user wants
> > > to have N + 1 devies he/she must umount all the existing devices, unload
> > > the module, load the module passing num_devices equals to N + 1.
> > > 
> > > This patchset introduces zram-control sysfs class, which has two sysfs
> > > attrs:
> > > 
> > >  - zram_add     -- add a new zram device
> > >  - zram_remove  -- remove a specific (device_id) zram device
> > > 
> > >     Usage example:
> > >         # add a new specific zram device
> > >         cat /sys/class/zram-control/zram_add
> > >         1
> > > 
> > >         # remove a specific zram device
> > >         echo 4 > /sys/class/zram-control/zram_remove
> > 
> > I just reported bug. Please handle it.
> 
> a-ha... found it:
> http://lkml.iu.edu/hypermail/linux/kernel/1505.0/04389.html
> 
> will take a look. thanks!
> 
> > Other nits:
> > 
> > 1) How about inserting a step to reset before zram_remove?
> > IOW, user should echo "1" > /sys/block/zram[0-9]*/reset before
> > echo zram_id > /sys/class/zram-control/zram_remove.
> > 
> > Actually, I can't think any benefit other than consistency of
> > zram interface but you might have.
> 
> well, I did this the way it is because there is no requirement to reset any
> devices before `rmmod zram' (which eventually removes all zram devices from the
> system), a set of umount-s is enough. so requiring both umount and reset before
> hot_remove seems to be a bit different.

Okay.

> 
> zram_remove() is called from both hot_remove and zram_exit()->destroy_devices()
> (which requires reset step anyway). so I'm not sure about this change. do you
> have any strong objections?

No. I just wanted to know you have any idea.

> 
> 
> > 2) How about using hot_add/hot_remove?
> > 
> > /class/zram-control includes prefix zram meaning so I think
> > we don't need zram prefix of the knobs. Instead, let's add
> > *hot* which is more straightforward for representing *dynamic*.
> > 
> > What do you think about it?
> 
> ok. I can change it. I'll ask Andrew to drop the entire patch series and
> will resubmit once we settle it down.
> 

Thanks!

-- 
Kind regards,
Minchan Kim