freebsd-nq/contrib/amd/tasks
David E. O'Brien ad8bcc147f Virgin import of AMD (am-utils) v6.1.5
Sponsored by: Juniper Networks
2007-12-05 15:48:03 +00:00

112 lines
3.3 KiB
Plaintext

AM-UTILS-6.1 TASKS TODO
Please volunteer to do any of the following:
- autofs support: see README.autofs for details.
- deal with everything that has XXX on it in the sources
- documentation update ("XXX: FILL IN" sections).
- compatibility with Sun's automount maps?
should be easier via the amd.conf file to specify type of map
- convert to using my own rpcgen .x files for amq/amq/nfs (v2 and v3)
- use packaging info for various OSs (such as RPM, Redhat Package Manager)
- $mindelay and $maxdelay in milliseconds
- multiple nfsl should be matched if one matched and nfs mount failed. fall
through.
- support multiple "fail-over read-only" NFS mounts in Solaris 2.6+.
- random nfs rhost:={srv1, srv2, srv3}
- nfslx, same as nfsl, but using linkx
- loadable info_*, amfs_*, and ops_* modules (lazy evaluation).
- hlfsd should daemonize even if -DDEBUG, then use -D nofork
- fixmount should use generic code from transp/transp_{tli,sockets}.c
- reverse notion of -F and other cmd-line options, so they override the
amd.conf file (right now amd.conf overrides cmd-line options).
- add am-utils URL and COPYRIGHT to all distributed scripts and sources and
programs etc. also to amq -v output.
- ion: browsable_dirs only works for nis if in [global] broken if it is file
maps.
- browsable 'auto' style maps.
- mention signals in amd man page
- ALLOWED_MOUNT_TIME of 40 seconds is way too long!
- detecting down'ed hosts faster: use default portmap TTL variables?
- push get_nfs_version() down into a child process
- type:=program should not require umount command. can use default
umount(). it works for smbmount/umount.
- document var[0-7] variables/selectors
- need a way to export a nfs-mount point after it is being mounted (jukebox
mounts). can do this with type:=program.
- support port=n, for machines that use NFS on a different port
- nfsl may be buggy when fs:= is specified explicitly (lockup?)
- y2k: amq /home (mounted time uses YY instead of YYYY)
- new amd.conf option to disable amq listener altogether (security)
Should be available even if DEBUG is off.
- new amd.conf option dynamic_system_params to recompute osver, karch,
etc. from uname() dynamically.
- solaris 8 doesn't pick up hsfs/pcfs/etc *_args structures and mount flags
- avoid having to compile ops_nfs3.o
- linux support for ext3
- be able to pass generic mount options to systems w/ a mount(2) that
supports it (Linux). Perhaps genopt="foo,bar" syntax.
- support several different nfs servers inside one nfsx entry
- reimplement the "parallel mounting" for multiple matching sub-entries
- fix amd.conf parser so it can parse key=value without spaces around the
"=" sign. Note this isn't so trivial because 'value' may include "="
signs.
- NFSv4 support (such a short "tasks" entry given how much work it is :-)
* documentation
- document what can and what can't be a top-level mount type (i.e. a
map-backed or map-holding type). Currently that's toplvl, auto, and
direct.
- document how to add a direct map to amd.conf.
* autofs stuff
Linux:
- host mounts broken with autofs v3
- nfsx status unknown
- local filesystems (ufs, etc) status unknown
Solaris:
- host mounts status unknown
- nfsx status unknown
- local filesystems (ufs, etc) status unknown
FreeBSD:
- just do it
All:
- mntfs->mf_fo->opt_fs can contain garbage sometimes??