Skip to contentSkip to conversation

connman 1.33 takes very long time to start

DISCUSSION Flavio Suligoi 2 participants 3 messages Jan 19, 2017

lore
Share & tools
lore
#0 lore raw

Hi all,

I'm using Yocto "Morty" for my custom i.MX6 DualLite embedded system board, with the Freescale kernel 4.1.35.

The new "connman 1.33" takes a lot of time to start; sometime it hangs for 120 seconds and more

The cause is the poor entropy of my kernel:

[    4.659894] random: udevd urandom read with 12 bits of entropy available

If, for test, I connect an USB keyboard and, during the boot, press the keys quickly, the entropy level goes to 45 bits or more and the boot speed is normal.

Using the previous "Krogoth" version, with "connman 1.31", the boot speed was always fast, so the problem is related to the new connman version.

As workaround for this problem, now I use the "rng-tools", starting the "rngd" daemon before starting the "connmand" daemon .

Do you have any idea to solve this problem or other suggestions?

Thanks,

Flavio

#1 lore raw
Quoted 4 lines · Flavio Suligoi

Using the previous "Krogoth" version, with "connman 1.31", the boot speed was always fast, so the problem is related to the new connman version.

Unless you can demonstrate that running connman 1.31 on morty with

everything else the same is also slow then this is unlikely: a change in kernel behaviour for random numbers is more likely.

If you have low entropy then rng-tools is one solution, or persist the value of /dev/urandom when shutting down to carry some entropy between boots (
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/tree/drivers/char/random.c
).

Ross

#2 lore raw

Thank you Ross!

Flavio

________________________________
Da: Burton, Ross <[email protected]>
Inviato: giovedì 19 gennaio 2017 11:49
A: Flavio Suligoi
Cc: [email protected]
Oggetto: Re: [poky] connman 1.33 takes very long time to start

On 19 January 2017 at 10:01, Flavio Suligoi <[email protected]<mailto:[email protected]>> wrote:

Using the previous "Krogoth" version, with "connman 1.31", the boot speed was always fast, so the problem is related to the new connman version.

Unless you can demonstrate that running connman 1.31 on morty with everything else the same is also slow then this is unlikely: a change in kernel behaviour for random numbers is more likely.

If you have low entropy then rng-tools is one solution, or persist the value of /dev/urandom when shutting down to carry some entropy between boots (https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/tree/drivers/char/random.c).

Ross