* Udevstart & udevtrigger
@ 2008-09-01 14:44 Francesco RUNDO
2008-09-01 15:26 ` Kay Sievers
0 siblings, 1 reply; 2+ messages in thread
From: Francesco RUNDO @ 2008-09-01 14:44 UTC (permalink / raw)
To: linux-hotplug
Hi,
My name is Francesco Rundo and I work porting Linux (kernel, device
drivers, etc..) for embedded system.
I've a question about udev, in particular way, I need to get some info
about udevstart and udevtrigger.
In order to improve the boot speed, I've removed in my udev init scripts
the tool udevstart. Basically, in my udev init scripts (which start
firstly in the user space boot sequence) there are in order:
1) some operation on /dev with manually creation of some devices;
2) udevstart call;
3) starting of udevd;
4) udevtrigger call;
5) udevsettle call;
After that, the common services will be initialized: ssh, D-bus, HAL,
Lirc, etc...;
I've noted that this sequence is used in different Linux distribution
for embedded system.
As far as I know, the tool "udevstart" has the target to populate
initial device directoty (/dev). It walks trough the sysfs device tree
and calls udev (daemon) to create the nodes for every valid device
found. It is used to fill the initial empty device directory with nodes
for all devices currently available. Is it right ?
Now, I know that "udevtrigger" replies the uevents generated by system
coldplug of the devices (passing the info to the udev daemon) while
udevsettle waits watching the udev event queue and exits if all current
events are handled. Is it right ?
By taking into account what above mentioned, I've removed from udev
starting script the tool "udevstart" which needs a lot of time to
populate the device (In my kernel config, I've a lot of device included
in the kernel ) and I've managed udev only with udevtrigger and
udevsettle. In this way, the device creation under /dev will be made by
udevtrigger. I've saved 3/4 sec of boot time!!
Finally, I've to say that in some web site I've noted that some Linux
distribution were using this approach. Also the initNG project suggests
to remove udevstart using only udevtrigger & udevsettle. Please, see
http://www.initng.org/wiki/udevstart for more detail.
Now the question is:
Is it right to remove udevstart using only udevtrigger ?
Are there some funcionality-restrictions in the udev if I don't use
udevstart ?
Is it true that starting from udev-117 udevstart will not be delivered ?
In my distribution, we are aligned to udev-116
Thanks in advance for your support.
Best Regards,
--
Francesco Rundo
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: Udevstart & udevtrigger
2008-09-01 14:44 Udevstart & udevtrigger Francesco RUNDO
@ 2008-09-01 15:26 ` Kay Sievers
0 siblings, 0 replies; 2+ messages in thread
From: Kay Sievers @ 2008-09-01 15:26 UTC (permalink / raw)
To: linux-hotplug
On Mon, Sep 1, 2008 at 16:44, Francesco RUNDO <francesco.rundo@st.com> wrote:
> My name is Francesco Rundo and I work porting Linux (kernel, device drivers,
> etc..) for embedded system.
>
> I've a question about udev, in particular way, I need to get some info about
> udevstart and udevtrigger.
>
> In order to improve the boot speed, I've removed in my udev init scripts the
> tool udevstart. Basically, in my udev init scripts (which start firstly in
> the user space boot sequence) there are in order:
> 1) some operation on /dev with manually creation of some devices;
> 2) udevstart call;
> 3) starting of udevd;
> 4) udevtrigger call;
> 5) udevsettle call;
> After that, the common services will be initialized: ssh, D-bus, HAL, Lirc,
> etc...;
>
> I've noted that this sequence is used in different Linux distribution for
> embedded system.
>
> As far as I know, the tool "udevstart" has the target to populate initial
> device directoty (/dev). It walks trough the sysfs device tree and calls
> udev (daemon) to create the nodes for every valid device found. It is used
> to fill the initial empty device directory with nodes for all devices
> currently available. Is it right ?
>
> Now, I know that "udevtrigger" replies the uevents generated by system
> coldplug of the devices (passing the info to the udev daemon) while
> udevsettle waits watching the udev event queue and exits if all current
> events are handled. Is it right ?
>
> By taking into account what above mentioned, I've removed from udev starting
> script the tool "udevstart" which needs a lot of time to populate the device
> (In my kernel config, I've a lot of device included in the kernel ) and I've
> managed udev only with udevtrigger and udevsettle. In this way, the device
> creation under /dev will be made by udevtrigger. I've saved 3/4 sec of boot
> time!!
>
> Finally, I've to say that in some web site I've noted that some Linux
> distribution were using this approach. Also the initNG project suggests to
> remove udevstart using only udevtrigger & udevsettle. Please, see
> http://www.initng.org/wiki/udevstart for more detail.
>
> Now the question is:
> Is it right to remove udevstart using only udevtrigger ?
> Are there some funcionality-restrictions in the udev if I don't use
> udevstart ?
> Is it true that starting from udev-117 udevstart will not be delivered ?
> In my distribution, we are aligned to udev-116
Udevstart is obsolete and does not exist anymore. It was replaced, and
is not installed since udev 082:
http://git.kernel.org/?p=linux/hotplug/udev.git;a=commit;hÂ33b59baaabe55a8f99dc37c4ee9011b85f8917
Kay
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2008-09-01 15:26 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-09-01 14:44 Udevstart & udevtrigger Francesco RUNDO
2008-09-01 15:26 ` Kay Sievers
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).