[go: up one dir, main page]

File: TODO

package info (click to toggle)
arla 0.36.2-11
  • links: PTS
  • area: main
  • in suites: sarge
  • size: 13,236 kB
  • ctags: 12,096
  • sloc: ansic: 125,016; sh: 10,590; makefile: 5,533; asm: 1,617; objc: 1,236; yacc: 372; perl: 367; lex: 309; awk: 35; lisp: 14
file content (104 lines) | stat: -rw-r--r-- 2,432 bytes parent folder | download
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
91
92
93
94
95
96
97
98
99
100
101
102
103
104

  $Id: TODO,v 1.155 2004/05/04 11:37:26 lha Exp $


	Release 0.36
		Redhat kernel + syscall

	Release 0.37:
		Darwin 10.2
			ubc
			find fd/vnode leak (fhopen)
		OSF/1 5.1a
			ubc
		FreeBSD 4.5
			ubc ?
		OpenBSD 3.1 and NetBSD 1.5
			dubble-buffer (X ?)
		Linux 2.4
			ubc
			(sparc64)syscall table 64 vs 32 bits syscalls
		Solaris 9
			ubc

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.

Store fcache entries on a list on the appropriate volcache entry to
make volume callbacks faster.

split cred for !bsd to rd_cred and wr_cred 

fix weirdness with fake mount points
  installattr (token w/o data) (ls -l ; ls foo ; ls bar)

write2: bsd

fix usage of token between nnpfs and arla

fix usage of open-token between nnpfs and arla

fs flushall:
-volume ?
is this really good ?

nnpfs/linux: 
check that we DTRT wrt Documentation/filesystem/Locking.txt

nnpfs/*: teach nnpfs to read AFS directories for readdir

   merge in tol's code from winnt

nnpfs/linux: check if we can hook into vm_operations/address_space to
figure out when cache pages are dirty and need to flushed.

   dubble buffer instead, look how its done on netbsd nnpfs

*/Makefile.in: allow creation of shared libraries

*/Makefile.in: homogenize

arlad/nnpfs: 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

nnpfs/aix: figure out where the PAG should be stored

nnpfs/irix: fix syscall and implement setgroups wrapper

nnpfs/irix: port to Irix 6.5

nnpfs/solaris: verify multithreadness, merge xnode and vnode mutex locks

nnpfs/sunos: implement setgroups wrapper

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