From Fedora Project Wiki
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:
= Init scripts =
= Init scripts =
The interesting init scripts for power management are those which are '''switch on''' by default. New services had been probably added after this review (February 2009). The simplest way how to check most of them was installed all packages which have init script in rawhide (future F-11) and then review them. Our attempt lead to new daemon which should started services on demand. But because of small number of services, which could be handled that way was [[https://bugzilla.redhat.com/show_bug.cgi?id=487665 | soud]] refused. Services of appropriate devices should be catching events from upstart/udev/DeviceKit instead of new daemon.
The interesting init scripts for power management are those which are '''switch on''' by default. New services had been probably added after this review (February 2009). The simplest way how to check most of them was installed all packages which have init script in rawhide (future F-11) and then review them. Our attempt lead to new daemon which should started services on demand. But because of small number of services, which could be handled that way was <strike>[[https://bugzilla.redhat.com/show_bug.cgi?id=487665 soud]]</strike> refused. Services of appropriate devices should be catching events from upstart/udev/DeviceKit instead of new daemon.


Part of the review was going through comps, whether there exist some barely used services in @base or so on, but nothing had been found.
Part of the review was going through comps, whether there exist some barely used services in @base or so on, but nothing had been found.
Line 9: Line 9:
|'''package''' || '''device''' || '''/etc/rc.d/init.d/''' || '''comment'''
|'''package''' || '''device''' || '''/etc/rc.d/init.d/''' || '''comment'''
|-
|-
|bluez || bluetooth || bluetooth || patch in [[https://bugzilla.redhat.com/show_bug.cgi?id=484345 bugzilla]]
|bluez || bluetooth || bluetooth || <strike>[[https://bugzilla.redhat.com/show_bug.cgi?id=484345 bugzilla]]</strike> FIXED in F-12
|-
|-
|fnfx || only for Toshiba laptops || fnfxd || Verify manufacture: "dmidecode -s system-manufacturer" and output will be f.e. "Dell Inc."
|fnfx || only for Toshiba laptops || fnfxd || Verify manufacture: "dmidecode -s system-manufacturer" and output will be f.e. "Dell Inc."

Latest revision as of 12:25, 19 May 2009

Init scripts

The interesting init scripts for power management are those which are switch on by default. New services had been probably added after this review (February 2009). The simplest way how to check most of them was installed all packages which have init script in rawhide (future F-11) and then review them. Our attempt lead to new daemon which should started services on demand. But because of small number of services, which could be handled that way was [soud] refused. Services of appropriate devices should be catching events from upstart/udev/DeviceKit instead of new daemon.

Part of the review was going through comps, whether there exist some barely used services in @base or so on, but nothing had been found.

The different types of hardware services

package device /etc/rc.d/init.d/ comment
bluez bluetooth bluetooth [bugzilla] FIXED in F-12
fnfx only for Toshiba laptops fnfxd Verify manufacture: "dmidecode -s system-manufacturer" and output will be f.e. "Dell Inc."
isdn4k-utils configuration of isdn isdn /etc/rc.d/init.d/isdn
microcode_ctl updates the microcode on Intel x86/x86-64 CPU's microcode_ctl Verify CPU: "dmidecode -s processor-manufacturer" output will be f.e. "Intel" and command uname give us architecture name f.e. "x86_64"
mt-st tape drive management Mt (for magnetic tape drives) and st (for SCSI tape devices) stinit

devices are connected through scsi (the better one) and usb (low-end devices)

openct PC/SC-Lite, as CT-API driver, have primitive mechanism to export smart card readers to remote machines via TCP/IP openct udev wasn't sending information about our smart card reader inclued in keyboard
pcsc-lite PC/SC Lite is to provide a Windows(R) SCard interface pcscd probably same problem as the previous smart card reader

Notes: bluez is widely installed, so it should be fixed. Other services are not installed in any base group and the hardware like magnetic tape is not widely used. So probably there is no point of fixing these other services. In this list is not irda which has not service on by default. If there were fixes for interacting services with hardware on the event then it should be also fixed.

The service for printer

package device /etc/rc.d/init.d/ comment
oki4linux printers OL400w oki4linux
cups printers with various input: lpt, usb, lan cups

For possible switching off printing service:

  • must be checked lpt, usb, lan (we can hardly check whether here is printer or not)
  • some programmes could be dependent on running cups even for printing into file

The services related with network

package device comment
ices streaming server maybe it should be up only for streaming?
ifplugd automatically configure your ethernet device when a cable is plugged in this is now done by NetworkManager?
nas audio equivalent of an X display server could be off without network?
nfs-utils NFS utilities and supporting clients and daemons for the kernel NFS server only with network?
qpidd An AMQP message broker daemon only on network?
rpcbind Universal Addresses to RPC Program Number Mapper only on network?
samba Samba Suite of programs only on network?
ser sip server only on network?
munin-node Network-wide graphing framework only on network?

Probably many of these services could be switch off, if the network is down. But probably these services for advanced users which are not of top of our list.

Various services

package device comment
libvirt virtual machine
mdadm for sw raid
qemu needs libvirtd on
wine-core wine emulator
wine-desktop
xen virtual machine
xen-runtime
zfs-fuse filesystem daemon?
ocfs2 only for cluster

Probably for libvirt, qemu, xen or wine could be daemon up only when some application need it. But this will have definitely problem with permission. This could save boot time.

Progress

With new packages should be done better testing of power saving with/out services. In case there will be considerable difference with services on/off, then it might be worth trying find power save in init scripts.

  • Measurement with services up/down [0%]