[systemd-devel] systemd-udev causes slow boot and high load

[Marco Steinacher]

Am 19.05.2015 um 18:32 schrieb Marco Steinacher:
> Am 19.05.2015 um 12:37 schrieb Lennart Poettering:
>> On Tue, 19.05.15 12:02, Marco Steinacher (mailinglists at websource.ch) wrote:
>>
>>> Hi,
>>>
>>> After upgrading my system to Debian jessie and switching to systemd I
>>> have the following problem:
>>>
>>> 1. During boot there is the message
>>>
>>> [    **] A start job is running for udev Wait for Complete De...on (11s
>>> / 3min)
>>>
>>> after "systemd-fsck[452]: /dev/sda5: clean" and it takes about 2 minutes
>>> until the boot process continues.
>>> systemd-analyze blame shows:
>>>
>>>   2min 2.945s systemd-udev-settle.service
>>
>> systemd-udev-settle.service just waits for devices to be probed. It
>> will wait for the kernel drivers to initialize and the udev rules to
>> be pliugged in.
>>
>> In fact, there's really no need to have this in the boot process at
>> all, unless you are using some broken software that assumes that
>> there's a point in time where all hardware has been plugged in, but
>> such a point in time does not actually exist, and hasn't existed in a
>> long time (simply because USB and other busses can take any time they
>> want before initialization is complete and the devices have shown up).
>>
>>> 2. dmesg also shows:
>>>
>>> [  254.305831] INFO: task systemd-udevd:296 blocked for more than 120
>>> seconds.
>>> [  254.305962]       Not tainted 3.16.0-4-686-pae #1
>>> [  254.306069] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
>>> disables this message.
>>> [  254.306165] systemd-udevd   D 00000000     0   296    234 0x00000004
>>>
>>> and similar messages for the tasks systemd-udevd:308 and
>>> systemd-udevd:307. The messages repeat a couple of times (see full log
>>> below).
>>>
>>
>> There you go: your kernel is hung, some driver or your xen irq
>> handling is borked and hangs during initialization. This is a kernel
>> problem, please report this to the kernel bugzilla.
> 
> Thank you for the swift reply and for clarifying the situation.  I
> understand now that this is a kernel issue and not a problem with
> systemd.  I'll try to track down the kernel issue now.

Just for the record - in case somebody faces a similar issue and
stumbles upon this thread:

I solved the problem. It was embarrassingly simple: Loading the I2C
driver JC42 with kernel 3.16.0 caused the "irq 16: nobody cared" issue
and the subsequent timeouts. Removing the module jc42 from /etc/modules
solved it!

Thanks again and sorry for the noise on the wrong mailing list,
Marco

-- 
OpenPGP Key ID: 0x62937F7F