[Skiboot] trace settings

Oliver oohall at gmail.com
Fri Sep 14 15:07:32 AEST 2018

On Fri, Sep 14, 2018 at 2:47 PM, Andrew Donnellan
<andrew.donnellan at au1.ibm.com> wrote:
> On 14/09/18 08:45, Raymond Higgs wrote:
>> Hi,
>> Is it possible to change the trace settings without recompiling the
>> firmware?  I'm interested in messages like this:
>>     [   90.072348476,5] OPAL skiboot-v6.0-p1da203b starting...
>>     [   90.072352660,7] initial console log level: memory 7, driver 5
>>     [   90.072354749,6] CPU: P9 generation processor (max 4 threads/core)
>>     [   90.072356576,7] CPU: Boot CPU PIR is 0x0014 PVR is 0x004e1202
>> I wouldn't normally post to a development mailing list, but I have looked
>> through a lot of documentation, and cannot find a procedure for this.
>> We have an MT 9006-22P machine running RHEL7.5.  It has this firmware:
>>     [root at power9 ibm,firmware-versions]# for xxx in `ls`; do echo $xxx:
>>     `cat $xxx`; done
>>     bmc-firmware-version: 1.23
>>     hcode: hw051018a.op920
>>     hostboot: f911e5c-pda8239f
>>     linux: 4.16.7-openpower2-pbc45895
>>     machine-xml: 218a77a
>>     name: ibm,firmware-versions
>>     occ: 77bb5e6-p623d1cd
>>     petitboot: v1.7.1-pf773c0d
>>     phandle: a
>>     sbe: 8e0105e
>>     skiboot: v6.0-p1da203b
>>     version: SUPERMICRO-P9DSU-V1.16-20180531-prod
>> Please cc me with any replies because I'm not subscribed to the mailing
>> list.
> As far as I'm aware, there's no way to do this at present.
> The most obvious place to put a setting like this would be NVRAM, but that
> poses an obvious issue - on the machine I'm currently logged into, NVRAM
> itself isn't initialised until 752 lines into the log.

We already support setting it from NVRAM. See here:

To set it you need the nvram utility which is available from the
petitboot shell. You can also install the opal-utils package to get it
on RHEL. The syntax is something like this:

nvram -p ibm,skiboot --update-config log-level-driver=7
nvram -p ibm,skiboot --update-config log-level-memory=8

That will cause PR_DEBUG (lvl 7) messages to appear on the console and
PR_TRACE (lvl 8) messages to be logged to the in-memory console. You
can access the console from the OS by `cat /sys/firmware/opal/msglog`

> Now we could do that anyway, and just accept that for the first few seconds
> while we don't have NVRAM, we'd rely on the platform defaults. That would
> still let you turn up the log level in time to see PCI, OpenCAPI, etc etc
> messages. (I'm guessing you probably want to turn the log level up to see
> those messages on the console rather than in the memory log buffer?)

Yeah, we can't apply nvram settings until we've discovered what
platform we are running on since the method used to access NVRAM
depends on the platform. I don't think it's a big deal though, since
the only people who should care what's going on that early are
probably skiboot developers. At a guess anyone posting from an
@broadcom.com address is probably more interested what's happening
during PCI probing.

> Andrew
> --
> Andrew Donnellan              OzLabs, ADL Canberra
> andrew.donnellan at au1.ibm.com  IBM Australia Limited
> _______________________________________________
> Skiboot mailing list
> Skiboot at lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/skiboot

More information about the Skiboot mailing list