• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1
2** ZCACHE PLAN FOR PROMOTION FROM STAGING **
3
4Last updated: Feb 13, 2013
5
6PLAN STEPS
7
81. merge zcache and ramster to eliminate horrible code duplication
92. converge on a predictable, writeback-capable allocator
103. use debugfs instead of sysfs (per akpm feedback in 2011)
114. zcache side of cleancache/mm WasActive patch
125. zcache side of frontswap exclusive gets
136. zcache must be able to writeback to physical swap disk
14    (per Andrea Arcangeli feedback in 2011)
157. implement adequate policy for writeback
168. frontswap/cleancache work to allow zcache to be loaded
17    as a module
189. get core mm developer to review
1910. incorporate feedback from review
2011. get review/acks from 1-2 additional mm developers
2112. incorporate any feedback from additional mm reviews
2213. propose location/file-naming in mm tree
2314. repeat 9-13 as necessary until akpm is happy and merges
24
25STATUS/OWNERSHIP
26
271. DONE as part of "new" zcache; in staging/zcache for 3.9
282. DONE as part of "new" zcache (cf zbud.[ch]); in staging/zcache for 3.9
29    (this was the core of the zcache1 vs zcache2 flail)
303. DONE as part of "new" zcache; in staging/zcache for 3.9
314. DONE (w/caveats) as part of "new" zcache; per cleancache performance
32    feedback see https://lkml.org/lkml/2011/8/17/351, in
33    staging/zcache for 3.9; dependent on proposed mm patch, see
34    https://lkml.org/lkml/2012/1/25/300
355. DONE as part of "new" zcache; performance tuning only,
36    in staging/zcache for 3.9; dependent on frontswap patch
37    merged in 3.7 (33c2a174)
386. DONE (w/caveats), prototyped as part of "new" zcache, had
39    bad memory leak; reimplemented to use sjennings clever tricks
40    and proposed mm patches with new version in staging/zcache
41    for 3.9, see https://lkml.org/lkml/2013/2/6/437;
427. PROTOTYPED as part of "new" zcache; in staging/zcache for 3.9;
43    needs more review (plan to discuss at LSF/MM 2013)
449. IN PROGRESS; owned by Konrad Wilk; Mel Gorman provided
45   great feedback in August 2012 (unfortunately of "old"
46   zcache)
4711. NOT DONE; owned by Konrad Wilk and Bob Liu
4812. TBD (depends on quantity of feedback)
4913. PROPOSED; one suggestion proposed by Dan; needs more ideas/feedback
5014. TBD (depends on feedback)
51
52WHO NEEDS TO AGREE
53
54Not sure.  Seth Jennings is now pursuing a separate but semi-parallel
55track.  Akpm clearly has to approve for any mm merge to happen.  Minchan
56Kim has interest but may be happy if/when zram is merged into mm.  Konrad
57Wilk may be maintainer if akpm decides compression is maintainable
58separately from the rest of mm.  (More LSF/MM 2013 discussion.)
59
60ZCACHE FUTURE NEW FUNCTIONALITY
61
62A. Support zsmalloc as an alternative high-density allocator
63    (See https://lkml.org/lkml/2013/1/23/511)
64B. Possibly support three zbuds per pageframe when space allows
65