[Skiboot] [PATCH] plat/qemu: use the common OpenPOWER routines to initialize

[Cédric Le Goater]

On 12/12/18 1:40 AM, Joel Stanley wrote:
> On Tue, 11 Dec 2018 at 16:57, Stewart Smith <stewart at linux.vnet.ibm.com> wrote:
>> I've pulled it into master as of
>> 8340a9642bba297a481d3d07ee06d84bff3f9aa8 even though we see some test
>> failures with it. The failures seem to be "op-test could do better"
>> rather than anything else, so I'll submit patches there.
> 
> Thanks! Looks like we resolved the xscom one with Cedric's addition to
> the pnv model.
> 
> We have regressed behaviour when a user tries to boot without the bmc
> simulator on the command line. We see 30 seconds of this each time
> skiboot tries to send an IPMI BT message:
> 
> [   32.069661976,3] BT: seq 0x9e netfn 0x06 cmd 0x31: Maximum queue
> length exceeded
> [   32.071144699,3] BT: seq 0x9d netfn 0x06 cmd 0x31: Removed from queue
> [   32.177366645,3] BT: seq 0x9f netfn 0x06 cmd 0x31: Maximum queue
> length exceeded
> [   32.179578975,3] BT: seq 0x9e netfn 0x06 cmd 0x31: Removed from queue
> [   32.286597167,3] BT: seq 0xa0 netfn 0x06 cmd 0x31: Maximum queue
> length exceeded
> [   32.288307209,3] BT: seq 0x9f netfn 0x06 cmd 0x31: Removed from queue
> 
> This is because we used to have this check:
> 
>        /* check if the BT device was defined by qemu */
>        dt_for_each_child(primary_lpc, n) {
>                if (dt_node_is_compatible(n, "bt"))
>                        bt_device_present = true;
>        }

Indeed. I forgot that QEMU work around ... 

> Cedric, can we have the pnv machine enable the bmc simulator by
> default if there's no bmc options on the command line?

I tried in the past, without success. The BT device is tied to the 
backend and it is difficult to untangle them without the command line. 
I will send a skiboot fix instead. 

But yes, I think we should have a default set of devices instantiated 
by the QEMU PowerNV machine and the BMC is part of it. 

Thanks, 

C.