Re: [PATCH v2] zram: introduce comp algorithm fallback functionality

[Minchan Kim <minchan@kernel.org>]

Hello Andrew,

On Tue, Sep 15, 2015 at 04:07:00PM -0700, Andrew Morton wrote:
> On Thu, 10 Sep 2015 14:03:51 +0900 Minchan Kim <minchan@kernel.org> wrote:
> 
> > On Tue, Sep 08, 2015 at 07:42:56PM +0100, Luis Henriques wrote:
> > > When the user supplies an unsupported compression algorithm, keep the
> > > previously selected one (knowingly supported) or the default one (if the
> > > compression algorithm hasn't been changed yet).
> > > 
> > > Note that previously this operation (i.e. setting an invalid algorithm)
> > > would result in no algorithm being selected, which means that this
> > > represents a small change in the default behaviour.
> > 
> > It seems it is hard for Andrew to parse so I will add more.
> 
> Thanks ;)
> 
> What's missing here is an understandable-by-andrew *reason* for the
> patch.  What's wrong with the old behaviour and why is the new
> behaviour better?

Oops, I said it in detail but it seems I got failed.

For initializing zram, we need to set up 3 optional parameters in advance.

1. the number of compression streams
2. memory limitation
3. compression alrogithm

Although user pass completely wrong value to set up for 1 and 2 parameters,
it's okay because they have default value so zram will be initialized
with the default value(Of course, when user pass wrong value via *echo*,
sysfs returns -EINVAL so user can notice it).

But 3 is not consistent with other optional parameters.
IOW, If user pass wrong value to set up 3 parameter, zram's initialization
would be failed unlike other optional parameters.

So, this patch make them consistent.