[gentoo-user] [OT] Recent git kernels break rtc (real-time-clock)

[gentoo-user] [OT] Recent git kernels break rtc (real-time-clock)

[walt]

After the recent 3.9 --> 3.10 kernel merge window, udev no longer creates
/dev/rtc (or /dev/rtc0) during bootup on my ~amd64 machines. (The only
machines I have now.)

So, I have two idle (very non-urgent) questions for you git-kernel nerds
out there (I know you're there ;)

First, anyone else seeing the same breakage?

Second, anyone understand how udev knows if I do, or don't, have a real
time clock on my machine?

(BTW, I didn't even notice that /dev/rtc is missing for a week or two
because the error messages zip by so fast, and nothing seems broken
because of it, other than /sbin/hwclock whining quietly.)

Re: [gentoo-user] [OT] Recent git kernels break rtc (real-time-clock)

[Paul Hartman]

On Mon, Jun 10, 2013 at 7:31 PM, walt <w41ter@gmail.com> wrote:
> After the recent 3.9 --> 3.10 kernel merge window, udev no longer creates
> /dev/rtc (or /dev/rtc0) during bootup on my ~amd64 machines. (The only
> machines I have now.)

Not a git-kernel nerd, but just an ordinary kernel nerd. :)

The kernel RTC driver creates /dev/rtc0 and then I assume udev creates
/dev/rtc from there, so if you are missing rtc0 that's a possible
source of the problem. Which would coincide with your kernel upgrade.

I know 3.9 introduced a couple new RTC-related option so maybe it
changed around some more to 3.10 series. I would run menuconfig and
see what it thinks you have enabled in that section, just in case
something got lost in transition from one kernel to the next.

In dmesg on my non-git 3.9.4 kernel it looks like:

[    1.237994] rtc_cmos 00:04: RTC can wake from S4
[    1.238158] rtc_cmos 00:04: rtc core: registered rtc_cmos as rtc0
[    1.238177] rtc_cmos 00:04: alarms up to one month, y3k, 114 bytes
nvram, hpet irqs
[    1.241101] rtc_cmos 00:04: setting system clock to 2013-06-04
04:28:34 UTC (1370320114)

I'm using the "PC-Style CMOS" RTC driver, and I have all of the
RTC-related options enabled except for the debugging options.

Re: [gentoo-user] [OT] Recent git kernels break rtc (real-time-clock)

[Francisco Ares]

[-- Attachment #1: Type: text/plain, Size: 1590 bytes --]

2013/6/11 Paul Hartman <paul.hartman+gentoo@gmail.com>

> On Mon, Jun 10, 2013 at 7:31 PM, walt <w41ter@gmail.com> wrote:
> > After the recent 3.9 --> 3.10 kernel merge window, udev no longer creates
> > /dev/rtc (or /dev/rtc0) during bootup on my ~amd64 machines. (The only
> > machines I have now.)
>
> Not a git-kernel nerd, but just an ordinary kernel nerd. :)
>
> The kernel RTC driver creates /dev/rtc0 and then I assume udev creates
> /dev/rtc from there, so if you are missing rtc0 that's a possible
> source of the problem. Which would coincide with your kernel upgrade.
>
> I know 3.9 introduced a couple new RTC-related option so maybe it
> changed around some more to 3.10 series. I would run menuconfig and
> see what it thinks you have enabled in that section, just in case
> something got lost in transition from one kernel to the next.
>
> In dmesg on my non-git 3.9.4 kernel it looks like:
>
> [    1.237994] rtc_cmos 00:04: RTC can wake from S4
> [    1.238158] rtc_cmos 00:04: rtc core: registered rtc_cmos as rtc0
> [    1.238177] rtc_cmos 00:04: alarms up to one month, y3k, 114 bytes
> nvram, hpet irqs
> [    1.241101] rtc_cmos 00:04: setting system clock to 2013-06-04
> 04:28:34 UTC (1370320114)
>
> I'm using the "PC-Style CMOS" RTC driver, and I have all of the
> RTC-related options enabled except for the debugging options.
>
>

I'm using gentoo-sources-3.8.13 - out of your scope, though, but I have
recently faced the same issue.

I rebuilt the kernel using "--menuconfig", to make sure that all RTC
options were enabled.  It works, now.

Good luck
Francisco

[-- Attachment #2: Type: text/html, Size: 2067 bytes --]

Re: [gentoo-user] [OT] Recent git kernels break rtc (real-time-clock)

[Walter Dnes]

On Tue, Jun 11, 2013 at 02:50:46PM -0300, Francisco Ares wrote

> I'm using gentoo-sources-3.8.13 - out of your scope, though, but I have
> recently faced the same issue.
> 
> I rebuilt the kernel using "--menuconfig", to make sure that all RTC
> options were enabled.  It works, now.

  Me too.  It seems that there were some changes recently in the .config
file.  I recently built a 3.7.10-gentoo-r1 kernel on a new machine and
was on the verge of sending an email to the list asking what the bleep I
was doing wrong.  After some experimentation, I found that I need either
CONFIG_RTC=y or CONFIG_HPET_EMULATE_RTC=y.

make menuconfig
    Device Drivers  --->
    Character devices  --->
    <*> Enhanced Real Time Clock Support (legacy PC RTC driver)
    [*] HPET - High Precision Event Timer

  You might be able to get away without the the legacy RTC support if
you enable HPET.  I don't know.

-- 
Walter Dnes <waltdnes@waltdnes.org>
I don't run "desktop environments"; I run useful applications

[gentoo-user] Re: [OT] Recent git kernels break rtc (real-time-clock)

[walt]

On 06/11/2013 04:00 PM, Walter Dnes wrote:
> It seems that there were some changes recently in the .config
> file.  I recently built a 3.7.10-gentoo-r1 kernel on a new machine and
> was on the verge of sending an email to the list asking what the bleep I
> was doing wrong.  After some experimentation, I found that I need either
> CONFIG_RTC=y ...

That fixed the problem for me, thanks.

OTOH, I've never had CONFIG_RTC enabled before, including 3.9.5, yet I've
always had /dev/rtc until 3.10.  Now I have /dev/rtc && /dev/rtc0, FWIW.

I have no idea why you guys are having the same problem with older kernels,
except maybe some useflag has changed recently.  Are all of you running
~amd64, maybe?

BTW, I pull directly from kernel.org, I don't use sys-kernel/git-sources,
so my git kernels are not affected by any useflags, FWIW.

Thanks guys.

Re: [gentoo-user] Re: [OT] Recent git kernels break rtc (real-time-clock)

[Walter Dnes]

On Tue, Jun 11, 2013 at 05:33:37PM -0700, walt wrote

> I have no idea why you guys are having the same problem with older
> kernels, except maybe some useflag has changed recently.  Are all
> of you running ~amd64, maybe?
> 
> BTW, I pull directly from kernel.org, I don't use
> sys-kernel/git-sources, so my git kernels are not affected by any
> useflags, FWIW.

  I'm running stable amd64.  I don't think a USE flag is involved.
sys-kernel/gentoo-sources only has...

-build -deblob -symlink

...flags available.  I think it's a change in .config defaults.  Note
that .config calls it legacy RTC.  I think that the "legacy" RTC support
has been made a non-default option, and HPET's emulation of RTC is now
the norm.

-- 
Walter Dnes <waltdnes@waltdnes.org>
I don't run "desktop environments"; I run useful applications