Re: [PATCH 1/2] staging: zram: add Crypto API support

[Piotr Sarna <p.sarna@partner.samsung.com>]

On 07/31/2013 02:49 PM, Greg KH wrote:
> On Wed, Jul 31, 2013 at 12:16:36PM +0200, Piotr Sarna wrote:
>> Hi,
>>
>> On 07/30/2013 03:53 PM, Greg KH wrote:
>>> On Tue, Jul 30, 2013 at 02:30:48PM +0200, Piotr Sarna wrote:
>>>> Current version of zram does not allow any substitution of a default
>>>> compression algorithm. Therefore, I decided to change the existing
>>>> implementation of page compression by adding Crypto API compability.
>>>
>>> As I have stated before, I want this code to get merged out of the
>>> drivers/staging/ area _before_ any new features get added to it.  People
>>> are taking too much time adding new stuff, like this, and no one seems
>>> to be working to get it merged to the proper place anymore.
>>>
>>
>> OK, but we actually need those features in order to test zram
>> against other, competitive solutions - and then decide whether
>> and how to merge it out of /drivers/staging.
> 
> Where is another "competitive solution" in the kernel?
> 
> And you are implying that as-is, zram isn't acceptable, right?  If so, I
> should just delete it now as I was originally told otherwise, that's why
> I merged it :(
> 
>>> So again, I'm going to have to say no to a new feature here, sorry.
>>> zcache, zram, and zsmalloc need to get cleaned up and merged out of
>>> staging before anything new can be added to them.
>>>
>>> Or, if no one is working on that, I guess I can just delete them,
>>> right?...
>>>
>>
>> Is there any official TODO list of cleaning up and merging out zram?
> 
> As it came from the "zswap" code, there doesn't seem to be one :(
> 
> The code is over 2 years old now, with no percieved movement out of
> staging.  If it doesn't get fixed up in the next kernel version or so, I
> will have to remove it entirely.
> 
> sorry,
> 
> greg k-h
> 

"Another competitive solutions" I thought of were zswap and zcache.

On one hand, during my tests (of compressed swap feature), zram
turned out to be no faster than zswap/zcache (but no slower either).

On the other, zram happens to have some more, perhaps reasonable
use cases (described in zram.txt), e.g. mounting it as /tmp.

So far, I haven't tested those other features of zram,
so I can't possibly say whether it should be considered for a merge-out
or kept in staging for now.

Regards,
Piotr Sarna