Re: [Regression] d96ac6f2: time: Revert ALWAYS_USE_PERSISTENT_CLOCK compile time optimizaitons

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, May 31, 2013 at 03:02:00AM +0200, Jens Taprogge wrote:
> On Wed, May 29, 2013 at 06:08:18PM -0700, John Stultz wrote:
> > On 05/29/2013 01:41 PM, Jens Taprogge wrote:
> > >On Tue, May 28, 2013 at 12:51:19PM -0700, John Stultz wrote:
> > >>On 05/26/2013 12:25 PM, Jens Taprogge wrote:
> > >>>Hi all.
> > >>>
> > >>>Linux 3.9.3 hangs early during boot on a Lenovo Thinkpad X1 Carbon (Ivy
> > >>>Bridge).  v3.9.2 works.  I have bisected the hang to
> > >>>   d96ac6f2: time: Revert ALWAYS_USE_PERSISTENT_CLOCK compile time optimizaitons.
> > >>>
> > >>>Reverting d96ac6f2 on top of v3.9.4 makes the kernel boot.  I do not
> > >>>have any additional information as the kernel hangs before anything is
> > >>>displayed.  The config of v3.9.4 with d96ac6f2 reverted is attached.
> > >>So with 3.9.4 (nothing reverted), what was the config you were using
> > >>to boot?
> > >>
> > >>Does changing CONFIG_RTC_HCTOSYS change the behavior?
> > >>
> > >>thanks
> > >>-john
> > >Hello John,
> > >
> > >turns out things behave quite differently with the patch reverted and
> > >not reverted.  Since there is no output, I can not tell whether I am
> > >seeing one or two problems.
> > >
> > >With vanilla 3.9.4:
> > >
> > >   CONFIG_RTC_HCTOSYS=y works
> > >   CONFIG_RTC_HCTOSYS=n does not work
> > 
> > Huh. Strange, I actually was expecting the opposite.
> > 
> > The odd thing is, commit d96ac6f2 *allows* HCTOSYS=y to be set. In
> > other words, without that patch, HCTOSYS is always N.
> > 
> > So its weird HCTOSYS=y works, but also reverting the patch (which
> > forces HCTOSYS=N) works.
> > 
> > Still stranger, option really shouldn't change much in early boot.
> > HCTOSYS driver triggers as a late_initcall, so I'd expect you'd get
> > some output before hitting trouble.
> > 
> > 
> > >   CONFIG_CPU_FREQ_STAT has no effect
> > >
> > >With d96ac6f2 reverted:
> > >
> > >   CONFIG_CPU_FREQ_STAT=n works
> > >   CONFIG_CPU_FREQ_STAT=y does not work
> > >
> > >Please see several configurations attached.
> > 
> > Hrm.. I do wonder if we're seeing something else going on here.
> 
> It turns out 3.9.4 with d96ac6f2 reverted only fails with
> CONFIG_CPU_FREQ_STAT=y if I manually enable the Intel PState driver by
> adding the Thinkpad's IvyBridge processor to the list of supported
> devices.  So this is not part of the regression per se.

Please forget what I said here.  I switched a config option.  And that
made the system boot.  Sorry for the confusion.

So, in the case of the reverted v3.9.4 with CONFIG_CPU_FREQ_STAT=y,
CONFIG_X86_VERBOSE_BOOTUP makes a difference.  To update the little
table from above:

With d96ac6f2 reverted:

  CONFIG_CPU_FREQ_STAT=n works
  CONFIG_CPU_FREQ_STAT=y and CONFIG_X86_VERBOSE_BOOTUP=y does not work
  CONFIG_CPU_FREQ_STAT=y and CONFIG_X86_VERBOSE_BOOTUP=n works

Please find the new working configuration attached.

Best Regards,
Jens

Attachment: config3.9.4_withrevert_working2.gz
Description: Binary data


[Index of Archives]     [Linux Kernel]     [Kernel Development Newbies]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Hiking]     [Linux Kernel]     [Linux SCSI]