Arch Linux

kokoko3k

Member

Registered: 2008-11-14

Posts: 2,448

udev 146-2 much slower than udev 145-1

Hi all, yesterday i upgraded udev from 145 to 146, now my arch box keeps 15 seconds to "process udev events".
I rolled back to 145 and time went down to 5,5 seconds.

My system is not fully upgraded,
I have to keep kernel to 2.6.27 because of some suspend/hibernate issues;
because of this, some packages which requires kernel >2.6.27 are in mi ignorepkgs (here the full list)

Does anyone observed the same thing or has a clue in how to investigate the problem?

Thanks in advance!

---

Help me to improve ssh-rdp !
Retroarch User? Try my koko-aio shader !

lawmaker

Member

Registered: 2007-03-08

Posts: 8

Re: udev 146-2 much slower than udev 145-1

I have exactly the same problem (and I'm glad that I'm not alone) - udev event processing takes exactly 15 seconds instead of roughly 5 (even the numbers match...). I've searched this forum for udev related problems and I've tried the following to no avail:

- removed all blacklists from /etc/modprobe.d/
- removed all modules from MODULES variable in /etc/rc.conf
- tried latest stable kernel (originaly I didn't - just thought I did (IgnorePkg) - upgrading to the latest kernel DOES solve this issue, but newer (custom) kernels still have this problem)
- refreshed initrd

I'll look into it some more and if I find anything, I'll post it here. Unfortunately, I don't have the older 145 in package cache, so I cannot downgrade.

If it's relevant in any way, I use 32-bit Arch on this machine.

Last edited by lawmaker (2009-09-17 17:54:48)

hw-tph

Member

Registered: 2006-11-01

Posts: 149

Re: udev 146-2 much slower than udev 145-1

Load the modules you need with the MODULES array in rc.conf and set MOD_AUTOLOAD to "no". That way it will take about 1/10 second.

bangkok_manouel

Member

From: indicates a starting point

Registered: 2005-02-07

Posts: 1,556

Re: udev 146-2 much slower than udev 145-1

try to revert this commit to see if it helps:
http://projects.archlinux.org/?p=initsc … 13b79c69c5

lawmaker

Member

Registered: 2007-03-08

Posts: 8

Re: udev 146-2 much slower than udev 145-1

Setting MOD_AUTOLOAD to "no" decreased the udev event processing time from 15 seconds to 14 seconds.

Using specified initscripts commit didn't change anything.

Downgrading the "udev" package to previous revision (145-1) did it - less than 2 seconds.

Last edited by lawmaker (2009-09-17 15:05:55)

lawmaker

Member

Registered: 2007-03-08

Posts: 8

Re: udev 146-2 much slower than udev 145-1

Upgrading kernel to the latest version seems to solve this problem (again, sorry for misleading you guys, I forgot I had kernel in IgnorePkg, so the latest kernel I tried was actually an older one).

If you're holding to an older kernel, you shouldn't upgrade udev beyond the 145-1 version (extend your IgnorePkg to udev as well). If you don't have the older (udev 145-1) package in cache, you can find it here: http://schlunix.org/archlinux/

Last edited by lawmaker (2009-09-17 15:58:27)

lawmaker

Member

Registered: 2007-03-08

Posts: 8

Re: udev 146-2 much slower than udev 145-1

Actually, this is odd, Arch latest kernel (2.6.30) seems to work fine, but booting with the stable vanilla (2.6.31) still exhibits this problem. So older kernel is not the problem - other kernels than 2.6.30 produce this behaviour. Quite strange...

Last edited by lawmaker (2009-09-17 18:06:38)

toofishes

Developer

From: Chicago, IL

Registered: 2006-06-06

Posts: 602

Website

Re: udev 146-2 much slower than udev 145-1

try to revert this commit to see if it helps:
http://projects.archlinux.org/?p=initsc … 13b79c69c5

This is a horrible suggestion...

bangkok_manouel

Member

From: indicates a starting point

Registered: 2005-02-07

Posts: 1,556

Re: udev 146-2 much slower than udev 145-1

try to revert this commit to see if it helps:
http://projects.archlinux.org/?p=initsc … 13b79c69c5

This is a horrible suggestion...

well, this has been the default on all arch machines for at least 1 month and i didn't see many people suffer from that (i know you did). i would have been surprised that OP would have gained back his 10 seconds with this change but why not trying at least. looks like there's a lot of black magic surrounding udev, we still do not know what was OP _real_ issue.