adjtimex calls failing on 3.18.5

[rmenessec]

The new code in 3.18.4 for the bcm2708 pps_gpio_pin parameter was extremely welcome, but now I'm stuck on that kernel.

Immediately after upgrading to 3.18.5 yesterday, I started getting these failures from chronyd:

Jan 31 13:03:50 localhost chronyd[3496]: chronyd version 1.31 starting
Jan 31 13:03:50 localhost chronyd[3496]: Frequency 38.867 +/- 0.011 ppm read from /var/lib/chrony/chrony.drift
Jan 31 13:03:50 localhost chronyd[3496]: Fatal error : adjtimex failed for set_frequency, freq_ppm=3.8867e+01 required_freq=-3.8867e+01 required_tick=10000
Jan 31 13:10:19 localhost chronyd[3564]: chronyd version 1.31 starting
Jan 31 13:10:19 localhost chronyd[3564]: linux_hz directive is no longer supported
Jan 31 13:10:19 localhost chronyd[3564]: Frequency 38.867 +/- 0.011 ppm read from /var/lib/chrony/chrony.drift
Jan 31 13:10:19 localhost chronyd[3564]: Fatal error : adjtimex failed for set_frequency, freq_ppm=3.8867e+01 required_freq=-3.8867e+01 required_tick=10000

I found a few references to chronyd failing on the adjtimex call when presented with odd kernel HZ values or the NOHZ option. I tested booting with 'nohz=off' and I tested specifying the kernel HZ value in chrony.conf ('linux_hz', seen above), but got nowhere. Help?

[chron0]

Hi, I've stumbled over a very similar issue with 3.10.67 on an Odroid C1 though. I don't know it it's related but since I've found my problem here, maybe my solution will help too. I needed to tweak kernel/time/ntp.c to get chrony to work again, see https://apollo.open-resource.org/mission:log:2015:02:06:patching-kernel-ntp-to-fix-odroid-armv7a-chrony-adjtimex-bug

[kukabu]

It's upstream bug. https://bugzilla.kernel.org/show_bug.cgi?id=92481

[pelwell]

Thanks for that.

[chron0]

Patched, compiled, tested: Works (for Odroid C1 at least). Thanks for the update kukabu.

[vk4tux]

On 08/02/15 19:16, chrono wrote:

Patched, compiled, tested: Works (for Odroid C1 at least). Thanks for
the update kukabu.

—
Reply to this email directly or view it on GitHub
#782 (comment).

chrono pls contact vk4tux at bigpond dot com re: C1?

[chron0]

Should have mail now :)

[popcornmix]

Please run rpi-update and test.
That includes the patch referenced by @kukabu.

[clivem]

Is this patch going to be committed to rpi-3.18.y now it has "passed" testing? ;)

[pelwell]

It's already in there: 772fa10

[popcornmix]

@rmenessec Can you test and close if okay?

[rmenessec]

Testing with today's 3.18.y (.7). chrony starts and runs again, but my PPS signal is unstable. I'm unclear whether this might be a problem with the patch, changes from .6 to .7, or none of the above. Any thoughts?

[Ruffio]

@rmenessec has this issue been resolved? If yes, please close this issue. Thanks.

[rmenessec]

@Ruffio I'm not aware of any action that's been taken to resolve this, as you can see from the thread.

[Ruffio]

@rmenessec I see multiple references on 25 Feb 2015 plus a request to test it. Have you tried with the latest firmware? Current kernel version 4.4 and you refer to 3.18. Could you test with latest version?

[rmenessec]

I don't even have the system in question any more. The voltage converter failed a year ago. What happened to asking me to test back in the beginning of 2015?

[Ruffio]

@rmenessec, how should I know? I'm an independent C# developer that just happens to think that RPi is great product, and I want to contribute to make it better. I can't fix anything on RPi, but I can help with administrative tasks, such as takng a look on all old issues and ask people if they are still relevant. IMHO if the real issue has been resolved or doesn't show up anymore/no more relevant, then the issue should be closed as there is no real issue. Alot of the old and untouched issues (for a year) has been closed now after I asked people to retest. I think that I have posted/asked people to verify their issue 350 times now, which you would have noticed if you have followed/watched Fully and not just when you are referenced. So why haven't you been asked before to test? I guess that nobody before me has taken this task on themself to ping people/old issues.