[lkml]   [2013]   [Dec]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] [RFC] mm: slab: separate slab_page from 'struct page'
On 12/10/2013 02:00 PM, Christoph Lameter wrote:
>> > We _need_ to share fields when the structure is handed between different
>> > subsystems and it needs to be consistent in both places. For slab page
>> > at least, the only data that actually gets used consistently is
>> > page->flags. It seems silly to bend over backwards just to share a
>> > single bitfield.
> If you get corruption in one field then you need to figure out which other
> subsystem could have accessed that field. Its not a single bitfield. There
> are numerous relationships between the fields in struct page.

I'm not saying that every 'struct page' user should get their own
complete structure. I'm just saying that the *slabs* should get their
own structure. Let's go through it field by field for the "normal"
'struct page' without debugging options:

page->flags: shared by everybody, needs to be consistent for things
like memory error handling
mapping: unioned over by s_mem for slab
index: unioned over by freelist for sl[oua]b
_count: unioned over by lots of stuff by sl[oua]b
lru: unioned over by lots of stuff by sl[oua]b, including another
list_head called 'list' which blk-mq.c is now using.
private: opaque storage anyway, but unioned over by sl[au]b

See? *EVERYTHING* is overridden by at least one of the sl?b allocators
except ->flags. In other words, there *ARE* no relationships when it
comes to the sl?bs, except for page->flags.

 \ /
  Last update: 2013-12-10 23:41    [W:0.054 / U:1.620 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site