Re: [PATCHv4 3/7] zswap: add to mm/

[Seth Jennings]

On 02/01/2013 09:31 AM, Seth Jennings wrote:
> On 01/31/2013 08:38 PM, Minchan Kim wrote:
>> On Thu, Jan 31, 2013 at 01:06:46PM -0600, Seth Jennings wrote:
>>> On 01/31/2013 01:07 AM, Minchan Kim wrote:
>>>> On Tue, Jan 29, 2013 at 03:40:23PM -0600, Seth Jennings wrote:
>>>>> zswap is a thin compression backend for frontswap. It receives
>>>>> pages from frontswap and attempts to store them in a compressed
>>>>> memory pool, resulting in an effective partial memory reclaim and
>>>>> dramatically reduced swap device I/O.
>>>>>
>>>>> Additionally, in most cases, pages can be retrieved from this
>>>>> compressed store much more quickly than reading from tradition
>>>>> swap devices resulting in faster performance for many workloads.
>>>>>
>>>>> This patch adds the zswap driver to mm/
>>>>>
>>>>> Signed-off-by: Seth Jennings <sjenning@xxxxxxxxxxxxxxxxxx>
>>>>> ---
>>>>>  mm/Kconfig  |  15 ++
>>>>>  mm/Makefile |   1 +
>>>>>  mm/zswap.c  | 656 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>>  3 files changed, 672 insertions(+)
>>>>>  create mode 100644 mm/zswap.c
>>>>>
>>>>> diff --git a/mm/Kconfig b/mm/Kconfig
>>>>> index 278e3ab..14b9acb 100644
>>>>> --- a/mm/Kconfig
>>>>> +++ b/mm/Kconfig
>>>>> @@ -446,3 +446,18 @@ config FRONTSWAP
>>>>>  	  and swap data is stored as normal on the matching swap device.
>>>>>  
>>>>>  	  If unsure, say Y to enable frontswap.
>>>>> +
>>>>> +config ZSWAP
>>>>> +	bool "In-kernel swap page compression"
>>>>> +	depends on FRONTSWAP && CRYPTO
>>>>> +	select CRYPTO_LZO
>>>>> +	select ZSMALLOC
>>>>
>>>> Again, I'm asking why zswap should have a dependent on CRPYTO?
>>>> Couldn't we support it as a option? I'd like to use zswap without CRYPTO
>>>> like zram.
>>>
>>> The reason we need CRYPTO is that zswap uses it to support a pluggable
>>> compression model.  zswap can use any compressor that has a crypto API
>>> driver.  zswap has _symbol dependencies_ on CRYPTO.  If it isn't
>>> selected, the build breaks.
>>
>> I think we can factor out compressoin part and remove dependency
>> at compile time by Kconfig. No?
> 
> I'm still not following.  How would one "factor out" the crypto API
> dependency when we use it to access the compressor modules.
> 
> The only thing I can think you're saying is to hack up the code with
> ifdefs to call the lzo code directly based on a Kconfig option.  I
> really hope you aren't saying that though :-/

Looking into this more, I found out that INET also selects CRYPTO, so
unless the kernel is not doing networking, CRYPTO will be enabled
anyway.  EXT4 also selects it.

> 
>> Of course, if we disable CRYPTO in Kconfig,
>> we lost pluggable model but not a problem for embedded system.
> 
> The pluggable model is _very_ necessary for us because we use it to
> access our hardware compression accelerator.  We do not use lzo in
> that case.  We use 842 (crypto/842.c and drivers/crypto/nx/nx-842.c).
> 
> I'm not sure why we are misunderstanding on this.  Is there a specific
> objection to depending the crypto API here? I understand that you are
> thinking about embedded systems.  Does the enabling CRYPTO and
> CRYPTO_LZO add significant size to the kernel or something?

If size is the concern, I did a quick size diff between a vmlinux with
and without CRYPTO:
			
		text	data	bss	dec
CRYPTO=n	4747622	602704	7774208	13124534
CRYPTO=y	4755437	602960	7774208	13132605
diff		7815	256	0	8071

Bottom line is CRYPTO adds about 8k to vmlinux.

Thanks,
Seth

>Just trying to understand why this is a problem.
> 
> Thanks,
> Seth
> 
>>
>> Anyway, If it's a burden for you at a moment, I'm not going to insist on it.
>> Will do it for myself.
> 
> 
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/