blob: 79f007e9f716cfb293d78d90902f72eb0c45128d (
plain) (
blame)
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
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
|
$KTH: TODO,v 1.140 2001/01/07 18:36:47 lha Exp $
Release 0.36:
Check Solaris
SMP
xfs_map/xfs_getpage (check why patch doesn't work)
Linux
webserver (symlink loose contents)
Release 0.35:
MacOS X
run-tests
ran (find fd/vnode leek)
Check with E people is bugs are fixed (rename of volume, ...)
TODO-list, in no particular order.
Please mail arla-drinkers@stacken.kth.se if you want to work on
any of these things so we don't do double work.
- fix weirdness with fake mount points
installattr (token w/o data) (ls -l ; ls foo ; ls bar)
arlad: add regular pinging of file servers
fs flushall:
-volume ?
is this really good ?
xfs/linux:
check that we DTRT wrt Documentation/filesystem/Locking.txt
xfs/*: teach xfs to read AFS directories for readdir
xfs/linux: check if we can hook into vm_operations/address_space to
figure out when cache pages are dirty and need to flushed.
*/Makefile.in: allow creation of shared libraries
*/Makefile.in: homogenize
arlad/xfs: return from open before fetching the whole file
arlad/xfs: cache in blocks (iff ?) file too large to fit in cache ?
arlad/xfs: implement advisory locking
arlad: more improvements on writing in disconnected mode
arlad: fallback to disconnected mode if networking is not working (possibly per cell)
arlad: more consistency checks before recovering state
arlad: remove unused stuff from FCacheEntry
xfs/aix: figure out where the PAG should be stored
xfs/irix: fix syscall and implement setgroups wrapper
xfs/irix: port to Irix 6.5
xfs/solaris: verify multithreadness, merge xnode and vnode mutex locks
xfs/sunos: implement setgroups wrapper
xfs: support building as part of the kernel
OPTIMIZATIONS
This is a list of posible optimizations that could be done to different parts
of arla.
- Delay reads and writes to cache and AFS-servers. Make the more operations
in one batch. (possible with rx?) Especially directories don't have to be
synced for every operation.
- send more messages to/from arlad in one sweep. (lookup on whole path in one
message, getattr+getdata, etc..)
MILKO
Filbunke, the nnp-class-server generation:
- callbacks
- logging
- fsck/salvage, stability, more tests
- SPEED, real i*op
- volser (?)
- pts,vldb
- security
|
