Intro
 
This benchmark should demonstrate the speed of Hotplug2, compared to Udev,
when performing trivial tasks common in early Linux userspace: loading modules
and trivial creating of devices.

It should be noted that unlike udev, hotplug2 only provides information that it
has read from uevent socket. It doesn't grab information from sysfs etc. However,
in this test, none of those informations were requested in any rule - this test
only measured performance for basic tasks.

The benchmark was performed in QEMU 0.8.1, with kqemu module, with 128MB of
virtual RAM.

The host is:
AMD Athlon XP 2000+ (1667MHz)
512MB DDR RAM

The kernel:
2.6.18-rc4-git1 (with several performance or uevent unrelated patches, such as
                 squashfs, unionfs, reiser4 or vesafb-tng)

The test took place after bootup which already loaded all modules, therefore
no overhead was created by one having to wait until modules are loaded and the
other having modules already preloaded.

If you think the measurement is unjoust, please let me know how to alter the
environment to even it out.


 Udev, version: 098

Measurement method:
time { udevtrigger && udevsettle; }

Results:
real    0m13.164s
user    0m0.128s
sys     0m0.612s

Rules for udev:
---[/etc/udev/rules.d/hotplug.rules]---
ENV{MODALIAS}=="?*",    RUN+="/sbin/modprobe $env{MODALIAS}"
---[cut]---


  Hotplug 2, version: 0.3

Measurement method:
time hotplug2

Results:
real    0m6.498s
user    0m1.424s
sys     0m4.436s

Rules for hotplug2:
---[/etc/hotplug2.rules]---
MODALIAS is set {
    exec modprobe %MODALIAS%;
}

DEVPATH is set {
    makedev /tmp/hotplug2/%DEVICENAME% 0644
}
---[cut]---