http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c1 Philippe De Muyter changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |phdm@macqel.be --- Comment #1 from Philippe De Muyter 2010-02-21 01:11:56 UTC --- after reading the previous log, I tried to change slightly /etc/init.d/boot.clock, by adding : start=`expr $start +1` before /sbin/udevadm settle --quiet --seq-start=$start --seq-end=$end This makes my boot fast. boot.clock is now finished in 2-3 seconds. But, I know nothing about udev or /sys/kernel/uevent_seqnum, so I cannot determine if the bug is in udev, in /etc/init.d/boot.clock or if there's no bug at all -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c2 Philippe De Muyter changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #343674|0 |1 is obsolete| | --- Comment #2 from Philippe De Muyter 2010-02-21 01:20:46 UTC --- Created an attachment (id=343675) --> (http://bugzilla.novell.com/attachment.cgi?id=343675) boot.msg with added debugging (but udevadm settle debug truncated) -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c3 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED --- Comment #3 from Kay Sievers 2010-05-18 22:43:12 UTC --- Sorry for getting to this that late. It looks like a bug in udev. I have no idea how the logic was ever supposed to work. The queue file contains only unfinished events, not necessarily finished ones, so it looks like, this can go wrong at any time. This needs to be fixed by replacing the entire logic. The call in boot.clock should probably just use: --exit-if-exists=/dev/rtc0 instead of the sequence number magic. Note: For 11.3 we switched to devtmpfs, where we can be sure, that the node is created before modprobe returns and has loaded the module, then all the settle calls are not needed anymore. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c5 --- Comment #5 from Philippe De Muyter 2010-05-24 20:30:33 UTC --- Created an attachment (id=364316) --> (http://bugzilla.novell.com/attachment.cgi?id=364316) /var/log/boot.msg excerpt for /etc/init.d/boot.clock -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c7 --- Comment #7 from Kay Sievers 2010-05-25 05:25:42 UTC --- Awesome work. Thanks a lot. I think I've found it, and found out why I was never able to reproduce it. Seems it can only happen, with the very first event after udev has started. We export the current sequence number to the /dev/.udev/queue.bin file. This sequence has already finished, because it is is the current value of the kernel counter. Now boot.clock is the first user without any event in-between, and it will wait for exactly this event to finish. All this should never happen anytime later, or if we did load any module before runnign settle with a sequence argument, because the queue file would also contain the finished entry, which we look for. I'll fix that later today. Thanks again for all your help so far. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c9 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO InfoProvider| |phdm@macqel.be --- Comment #9 from Kay Sievers 2010-05-25 13:32:11 UTC --- Fix released upstream and a new udev package is in Factory. How did you run into this? There is no other report like this for openSUSE 11.2. You compiled your own kernel with a modular rtc_cmos? We usually have rtc_cmos compiled-in, so that the kernel itself can sync the system time at bootup before any userspace runs, and hwclock does not necessarily need to be called: $ dmesg rtc_cmos 00:07: setting system clock to 2010-05-25 11:11:02 UTC -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c10 --- Comment #10 from Philippe De Muyter 2010-05-26 21:43:36 UTC --- Created an attachment (id=365025) --> (http://bugzilla.novell.com/attachment.cgi?id=365025) boot.clock related excerpt of /var/log/boot.msg Thanks for your patch. I have merged your patch into my verbose version and udevadm settle now finish before the timeout as it should. I cannot tell if it really works as it should, because I notice that the loop is still entered many times. Is that normal ? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c12 Philippe De Muyter changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED InfoProvider|phdm@macqel.be | --- Comment #12 from Philippe De Muyter 2010-05-26 21:51:36 UTC --- (In reply to comment #9)

Fix released upstream and a new udev package is in Factory.

How did you run into this? There is no other report like this for openSUSE 11.2. You compiled your own kernel with a modular rtc_cmos?

I did not compile my kernel myself. It comes from SuSE : diabolo:~ # cat /proc/version Linux version 2.6.31.12-0.1-default (geeko@buildhost) (gcc version 4.4.1 [gcc-4_4-branch revision 150839] (SUSE Linux) ) #1 SMP 2010-01-27 08:20:11 +0100

We usually have rtc_cmos compiled-in, so that the kernel itself can sync the system time at bootup before any userspace runs, and hwclock does not necessarily need to be called: $ dmesg rtc_cmos 00:07: setting system clock to 2010-05-25 11:11:02 UTC

But you're right : rtc_cmos is a module (see attached /proc/config.gz (unzipped)) -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c14 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO InfoProvider| |phdm@macqel.be --- Comment #14 from Kay Sievers 2010-05-27 05:11:23 UTC --- (In reply to comment #12)

I did not compile my kernel myself. It comes from SuSE. But you're right : rtc_cmos is a module (see attached /proc/config.gz

Interesting. This bug should trigger for all users then, but I did not see any other report. I hope they don't wait all for 3 minutes. :) Weird, I don't see rtc_cmos as a module in the 11.2 kernel. The kernel you have installed is just from the standard update repo? I don't have a recent 11.2 installation anymore. In the Factory kernel, the one that will be 11.3, it is also compiled in. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

http://bugzilla.novell.com/show_bug.cgi?id=581612 http://bugzilla.novell.com/show_bug.cgi?id=581612#c16 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #16 from Kay Sievers 2010-05-27 09:56:07 UTC --- I see. Thanks for your excellent debugging and testing. And sorry for that taking so long, I just had no idea what was going wrong here. As it seems that the released 11.2 does not have this problem, I'll queue this fix for a possible udev update in 11.2, but release it only if there will be an update because of other issues which need to be fixed. Thanks again! -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.