Re: Hang while booting 4.15.7

[Brian Rak]

On 3/8/2018 1:02 PM, Brian Rak wrote:
> On 3/8/2018 12:49 PM, Randy Dunlap wrote:
> > On 03/08/2018 08:21 AM, Brian Rak wrote:
> > > We have some Dell servers running Intel Gold 6126 processors. Some 
> > > of them hang on boot under 4.15.7, but work fine on 4.14.14. When 
> > > they hang, we see the following on console:
> > >
> > > Error parsing PCC subspaces from PCCT
> > > watchdog: BUG: soft lockup - CPU #16 stuck for 23s! [swapper/0:1]
> > >
> > > We see that PCC subspaces error under 4.14 as well, but it doesn't 
> > > cause the machine to hang.
> > >
> > > So far we haven't been able to correlate these hangs with anything 
> > > in particular. Some machines will hang, some machines will boot. 
> > > They're otherwise identical as far as hardware and firmware goes.
> > >
> > > I've tried pcie_aspm=off, since that seems to be the next bit of 
> > > code that's being executed. This resulted in the machine booting a 
> > > little further, but then oopsing somewhere in acpi_os_purge_cache. 
> > > I'm not able to get a full trace there, as I don't have serial 
> > > access easily available.
> > >
> > > Any suggestions?
> > Hi,
> >
> > The first thing that I would do is boot with:
> > ÂÂ ignore_loglevel initcall_debug
> > on the kernel boot command line.
> >
> > That will add lots of messages and maybe give us a stronger hint 
> > about where
> > the hang is actually happening.
> >
> > And then worst case (without a boot log via serial console or 
> > netconsole) is
> > to take a photo of the screen with the oops messages.
> >
> > And if you are fairly certain that it's an ACPI issue, also write to the
> > linux-acpi@xxxxxxxxxxxxxxx mailing list.
> Thanks!
>
> I booted with those parameters, and this certainly seems like an ACPI 
> issue. During bootup, the machine paused here for about 20s:
> https://www.dropbox.com/s/39us0tlhbzuay7t/2018-03-08%2012_52_35.png?dl=0
>
> then it started printing this trace:
> https://www.dropbox.com/s/nxdhm19wcitrgm0/2018-03-08%2012_53_04.png?dl=0
>
> (I can't see to figure out a decent way to capture the beginning of 
> the trace here, I'll have to see if I can get the serial console working)
I got the serial console working, that's just the end of the CPU stuck 
message:

https://gist.githubusercontent.com/devicenull/abe9022877d0a7354fa2ffc8b8a8f042/raw/e497624f90037eb272760f7c5c3d2a0f21e5ea83/gistfile1.txt