Re: [PATCH v3] ALSA: control: Add memory consumption limit to user controls

[Takashi Sakamoto <o-takashi@sakamocchi.jp>]

Hi,

Sorry to be late.

On Sat, Apr 10, 2021 at 10:47:22AM +0200, Takashi Iwai wrote:
> On Sat, 10 Apr 2021 10:20:16 +0200,
> Takashi Sakamoto wrote:
> > 
> > On Fri, Apr 09, 2021 at 06:09:02PM +0200, Takashi Iwai wrote:
> > > On Fri, 09 Apr 2021 12:59:10 +0200,
> > > Takashi Iwai wrote:
> > > > 
> > > > On Fri, 09 Apr 2021 04:27:35 +0200,
> > > > Takashi Sakamoto wrote:
> > > > > 
> > > > > 4. Add `max_user_ctl_alloc_size` kobject attribute to the control device
> > > > > 
> > > > > In the patch, a new module parameter 'max_user_ctl_alloc_size' is added.
> > > > > In the item, I use the value of this parameter as initial value per
> > > > > control device. The value per control device can be changed via sysfs
> > > > > node.
> > > > > 
> > > > > The `max_user_ctl_alloc_size` is really the attribute of control device,
> > > > > so I think it acceptable. Additionally, 'curr_user_ctl_alloc_size' is
> > > > > also added so that userspace applications get current status.
> > > > 
> > > > So that's the primary purpose?  Then it makes sense, yeah.
> > > 
> > > You meant something like below, right?
> > 
> > If you were carefully reading my items in the order, you would have
> > realized that the patch includes problem to share attribute group
> > table between several modules...
> 
> Which several modules...?  The control API is mandatory, hence it
> can't be separated from each card core stuff.  So splitting to another
> module makes no sense, as I already replied earlier.
> 
> > Device attribute is one of userspace interface expected to be stable. I'd
> > like to avoid careless changes which our known developer tends to do.
> 
> Sure, it has to be set on stone once after put in the tree.
> However, currently it's just a brain storming phase, and no need to
> grumble for a dreadful future vision.

Oops, I realised to have overlooked your previous post:
https://lore.kernel.org/alsa-devel/s5h5z0v67wh.wl-tiwai@suse.de/

Let me back to it so that we share the premises for the discussion.


Thanks

Takashi Sakamoto