Re: [RFC 0/6] ARC: merge HAPS-HS with nSIM-HS configs

[Vineet Gupta]

Ok but this if off topic and not directly related to patches - or am I missing something.

Indeed if we run UP only for nsim we should add pseudo SMP and also a true SMP.

For uClibc-ng regression, it doesn't matter what they use and not for us to decide anyways - we just need to tell them (or point to wiki etc) that starting with 5.x kernel,any nsim uart config needs to change of they are using the defconfigs.

-Vineet

________________________________________
From: Eugeniy Paltsev <paltsev@xxxxxxxxxxxx>
Sent: Tuesday, October 22, 2019 11:18 AM
To: Vineet Gupta; Eugeniy Paltsev; linux-snps-arc@xxxxxxxxxxxxxxxxxxx; Vineet Gupta
Cc: Alexey Brodkin; linux-kernel@xxxxxxxxxxxxxxx
Subject: Re: [RFC 0/6] ARC: merge HAPS-HS with nSIM-HS configs

Btw, about nSIM regression runs:
We are using UP configuration for all nSIM runs and I propose to use pseudo-SMP  configuration (SMP configuration with only one CPU online) instead as our main target are SMP systems.

So -
run our verification jobs on both UP and pseudo-SMP nSIM
run uClibc-ng regression on pseudo-SMP only

 Any objections / thoughts?
---
 Eugeniy Paltsev


________________________________________
From: Vineet Gupta <vineetg76@xxxxxxxxx>
Sent: Friday, October 18, 2019 21:48
To: Eugeniy Paltsev; linux-snps-arc@xxxxxxxxxxxxxxxxxxx; Vineet Gupta
Cc: Alexey Brodkin; linux-kernel@xxxxxxxxxxxxxxx
Subject: Re: [RFC 0/6] ARC: merge HAPS-HS with nSIM-HS configs

On 10/18/19 5:15 AM, Eugeniy Paltsev wrote:
> Starting from nSIM 2019.06 is possible to use DW UART
> instead of ARC UART. That allows us to merge
> "nsim_hs" with "haps_hs" and "nsim_hs_smp" with "haps_hs_smp"
> mith some minor changes.
>
> We eliminate nsim_hs_defconfig and nsim_hs_smp_defconfig
> and leave haps_hs_defconfig and haps_hs_smp_defconfig
> which can be used on HAPS / nSIM / ZEBU / QEMU platforms
> without additionall changes in Linux kernel.

Thx for doing this, I was planning to do this myself.

But remember that doing this will disturb existing nsim setups
- Our internal linux/gnu regression jobs one of which tracks my for-curr and will
  fail immediately
- uClibc-ng maintainer who also seems to use nsim AFAIK for regression runs

So we need to notify parties involves (and it would be best that we align this to
a kernel release which anyways will be next one.

Also go thru the github wiki pages and wherever applicable please add the config
info for nsim (keep the old settings there for reference as well)

Thx,
-Vineet