blob: d0c18fa957435aee85349c37d766f8be043ca11b (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
|
** ZCACHE PLAN FOR PROMOTION FROM STAGING **
Last updated: Feb 13, 2013
PLAN STEPS
1. merge zcache and ramster to eliminate horrible code duplication
2. converge on a predictable, writeback-capable allocator
3. use debugfs instead of sysfs (per akpm feedback in 2011)
4. zcache side of cleancache/mm WasActive patch
5. zcache side of frontswap exclusive gets
6. zcache must be able to writeback to physical swap disk
(per Andrea Arcangeli feedback in 2011)
7. implement adequate policy for writeback
8. frontswap/cleancache work to allow zcache to be loaded
as a module
9. get core mm developer to review
10. incorporate feedback from review
11. get review/acks from 1-2 additional mm developers
12. incorporate any feedback from additional mm reviews
13. propose location/file-naming in mm tree
14. repeat 9-13 as necessary until akpm is happy and merges
STATUS/OWNERSHIP
1. DONE as part of "new" zcache; in staging/zcache for 3.9
2. DONE as part of "new" zcache (cf zbud.[ch]); in staging/zcache for 3.9
(this was the core of the zcache1 vs zcache2 flail)
3. DONE as part of "new" zcache; in staging/zcache for 3.9
4. DONE (w/caveats) as part of "new" zcache; per cleancache performance
feedback see https://lkml.org/lkml/2011/8/17/351, in
staging/zcache for 3.9; dependent on proposed mm patch, see
https://lkml.org/lkml/2012/1/25/300
5. DONE as part of "new" zcache; performance tuning only,
in staging/zcache for 3.9; dependent on frontswap patch
merged in 3.7 (33c2a174)
6. DONE (w/caveats), prototyped as part of "new" zcache, had
bad memory leak; reimplemented to use sjennings clever tricks
and proposed mm patches with new version in staging/zcache
for 3.9, see https://lkml.org/lkml/2013/2/6/437;
7. PROTOTYPED as part of "new" zcache; in staging/zcache for 3.9;
needs more review (plan to discuss at LSF/MM 2013)
9. IN PROGRESS; owned by Konrad Wilk; Mel Gorman provided
great feedback in August 2012 (unfortunately of "old"
zcache)
11. NOT DONE; owned by Konrad Wilk and Bob Liu
12. TBD (depends on quantity of feedback)
13. PROPOSED; one suggestion proposed by Dan; needs more ideas/feedback
14. TBD (depends on feedback)
WHO NEEDS TO AGREE
Not sure. Seth Jennings is now pursuing a separate but semi-parallel
track. Akpm clearly has to approve for any mm merge to happen. Minchan
Kim has interest but may be happy if/when zram is merged into mm. Konrad
Wilk may be maintainer if akpm decides compression is maintainable
separately from the rest of mm. (More LSF/MM 2013 discussion.)
ZCACHE FUTURE NEW FUNCTIONALITY
A. Support zsmalloc as an alternative high-density allocator
(See https://lkml.org/lkml/2013/1/23/511)
B. Possibly support three zbuds per pageframe when space allows
|
