lkml.org 
[lkml]   [2011]   [Jun]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v3 0/2] fadvise: support POSIX_FADV_NOREUSE
On 27/06/11 08:11, Andrea Righi wrote:
> On Mon, Jun 27, 2011 at 12:04:41PM +0900, KOSAKI Motohiro wrote:
>> (2011/06/24 22:49), Andrea Righi wrote:
>>> There were some reported problems in the past about trashing page cache
>>> when a backup software (i.e., rsync) touches a huge amount of pages (see
>>> for example [1]).
>>>
>>> This problem has been almost fixed by the Minchan Kim's patch [2] and a
>>> proper use of fadvise() in the backup software. For example this patch
>>> set [3] has been proposed for inclusion in rsync.
>>>
>>> However, there can be still other similar trashing problems: when the
>>> backup software reads all the source files, some of them may be part of
>>> the actual working set of the system. When a POSIX_FADV_DONTNEED is
>>> performed _all_ pages are evicted from pagecache, both the working set
>>> and the use-once pages touched only by the backup software.
>>>
>>> A previous proposal [4] tried to resolve this problem being less
>>> agressive in invalidating active pages, moving them to the inactive list
>>> intead of just evict them from the page cache.
>>>
>>> However, this approach changed completely the old behavior of
>>> invalidate_mapping_pages(), that is not only used by fadvise.
>>>
>>> The new solution maps POSIX_FADV_NOREUSE to the less-agressive page
>>> invalidation policy.
>>>
>>> With POSIX_FADV_NOREUSE active pages are moved to the tail of the
>>> inactive list, and pages in the inactive list are just removed from page
>>> cache. Pages mapped by other processes or unevictable pages are not
>>> touched at all.
>>>
>>> In this way if the backup was the only user of a page, that page will be
>>> immediately removed from the page cache by calling POSIX_FADV_NOREUSE.
>>> If the page was also touched by other tasks it'll be moved to the
>>> inactive list, having another chance of being re-added to the working
>>> set, or simply reclaimed when memory is needed.
>>>
>>> In conclusion, now userspace applications that want to drop some page
>>> cache pages can choose between the following advices:
>>>
>>> POSIX_FADV_DONTNEED = drop page cache if possible
>>> POSIX_FADV_NOREUSE = reduce page cache eligibility
>>
>> Eeek.
>>
>> Your POSIX_FADV_NOREUSE is very different from POSIX definition.
>> POSIX says,
>>
>> POSIX_FADV_NOREUSE
>> Specifies that the application expects to access the specified data once and then
>> not reuse it thereafter.
>>
>> IfI understand correctly, it designed for calling _before_ data access
>> and to be expected may prevent lru activation. But your NORESE is designed
>> for calling _after_ data access. Big difference might makes a chance of
>> portability issue.
>
> You're right. NOREUSE is designed to implement drop behind policy.

Hmm fair enough.
NOREUSE is meant for specifying you _will_ need the data _once_

Isn't this what rsync actually wants though?
I.E. to specify NOREUSE for the file up front
so it would drop from cache automatically as processed,
(if not already in cache).

I realize that would be a more invasive patch.

> I'll post a new patch that will plug this logic in DONTNEED (like the
> presious version), but without breaking the old /proc/sys/vm/drop_caches
> behavior.

But will that break existing apps (running as root) that expect DONTNEED
to drop cache for a _file_. Perhaps posix_fadvise() is meant to have
process rather than system scope, but that has not been the case until now.

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

\
 
 \ /
  Last update: 2011-06-27 12:39    [W:0.114 / U:0.224 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site