我发现firewalld在 Fedora 24 中这需要很长时间：
# systemd-analyze --no-pager blame
12.051s firewalld.service
7.613s dev-sda8.device
6.347s akmods.service
5.832s dev-sdb3.device
3.956s systemd-udev-settle.service
3.561s ldconfig.service
3.435s lvm2-monitor.service
3.037s upower.service
2.983s lm_sensors.service
2.895s systemd-logind.service
2.850s akmods-shutdown.service
2.847s gssproxy.service
2.847s sysstat.service
2.841s rsyslog.service
2.801s rtkit-daemon.service
2.695s systemd-hwdb-update.service
1.988s dnf-makecache.service
1.783s systemd-udevd.service
1.777s systemd-fsck@dev-disk-by\x2duuid-623D\x2d6AE0.service
1.705s systemd-modules-load.service
1.584s home.mount
1.472s fedora-readonly.service
1.457s systemd-tmpfiles-setup-dev.service
1.456s proc-fs-nfsd.mount
1.448s primage.mount
1.209s plymouth-start.service
1.144s systemd-fsck-root.service
1.132s abrtd.service
1.050s systemd-vconsole-setup.service
1.040s systemd-sysctl.service
关键链如下：
The time after the unit is active or started is printed after the "@" character.
The time the unit takes to start is printed after the "+" character.
graphical.target @29.297s
└─multi-user.target @29.296s
└─crond.service @29.186s
└─systemd-user-sessions.service @29.060s +100ms
└─remote-fs.target @29.048s
└─remote-fs-pre.target @29.048s
└─iscsi-shutdown.service @29.047s
└─network.target @29.046s
└─NetworkManager.service @28.444s +601ms
└─network-pre.target @28.434s
└─firewalld.service @16.383s +12.051s
└─dbus.service @13.557s
└─basic.target @13.531s
└─sockets.target @13.531s
└─dbus.socket @13.531s
我可以做些什么来减少firewalld的启动时间？