-
Notifications
You must be signed in to change notification settings - Fork 54
/
README
19 lines (18 loc) · 1015 Bytes
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
[This Document is Under Construction]
TODO
- free fids (+users) when connection fails and they are not properly clunked
- rework auth to use afid and Tauth?
- POSIX issue: newgrp + creat = created file will have old gid?
- flock advisory file lock
- fcntl(lock) advisory record lock
- one stuck fs can hang all npfs worker threads
- seekdir (offset != 0), telldir incompat with 9P
- limit aread/awrite size (breaks atomicity, but better than malloc==NULL /oom)
- handle simultaneous aread/awrite on same fid
- mount generates attach + stat / - can we cache stat / result to avoid
touching all the file system at mount time (e.g. when doing it at the
beginning of every cluster job)? If just one fs is down, everything hangs.
- diodmount should umount control file system when it aborts for other reasons
- init scripts 'restart' fails to start server if it is not already running
- init scripts should create rundir if it doesn't exist
- diodctl dynamic allocation of ports needs to use all server interfaces