lkml.org 
[lkml]   [2012]   [May]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 4/4] zsmalloc: zsmalloc: align cache line size
On 05/11/2012 09:03 AM, Dan Magenheimer wrote:

>> From: Minchan Kim [mailto:minchan@kernel.org]
>> Subject: Re: [PATCH 4/4] zsmalloc: zsmalloc: align cache line size
>>
>> On 05/08/2012 11:00 PM, Dan Magenheimer wrote:
>>
>>>> From: Minchan Kim [mailto:minchan@kernel.org]
>>>>> zcache can potentially create a lot of pools, so the latter will save
>>>>> some memory.
>>>>
>>>>
>>>> Dumb question.
>>>> Why should we create pool per user?
>>>> What's the problem if there is only one pool in system?
>>>
>>> zcache doesn't use zsmalloc for cleancache pages today, but
>>> that's Seth's plan for the future. Then if there is a
>>> separate pool for each cleancache pool, when a filesystem
>>> is umount'ed, it isn't necessary to walk through and delete
>>> all pages one-by-one, which could take quite awhile.
>>
>>>
>>
>>> ramster needs one pool for each client (i.e. machine in the
>>> cluster) for frontswap pages for the same reason, and
>>> later, for cleancache pages, one per mounted filesystem
>>> per client
>>
>>
>> Fair enough.
>> But some subsystems can't want a own pool for not waste unnecessary memory.
>>
>> Then, how about this interfaces like slab?
>>
>> 1. zs_handle zs_malloc(size_t size, gfp_t flags) - share a pool by many subsystem(like kmalloc)
>> 2. zs_handle zs_malloc_pool(struct zs_pool *pool, size_t size) - use own pool(like kmem_cache_alloc)
>>
>> Any thoughts?
>
> I don't have any objections to adding this kind of
> capability to zsmalloc. But since we are just speculating
> that this capability would be used by some future
> kernel subsystem, isn't it normal kernel protocol for
> this new capability NOT to be added until that future
> kernel subsystem creates a need for it.


Now zram makes pool per block device and a embedded system may use zram
for several block device, ex) swap device, several compressed tmpfs
In such case, share pool is better than private pool because embedded system
don't mount/umount frequently on such directories since booting.

>
> As I said in reply to the other thread, there is missing
> functionality in zsmalloc that is making it difficult for
> it to be used by zcache. It would be good if Seth
> and Nitin (and any other kernel developers) would work


So, if you guys post TODO list, it helps fix the direction.

> on those issues before adding capabilities for non-existent
> future users of zsmalloc.


I think it's not urgent than zs_handle mess.

>
> Again, that's just my opinion.

> Dan
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
> Don't email: <a href=ilto:"dont@kvack.org"> email@kvack.org </a>
>



--
Kind regards,
Minchan Kim


\
 
 \ /
  Last update: 2012-05-11 03:01    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site