Re: port-sgimips/57595: Indigo2 cannot boot

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

The following reply was made to PR port-sgimips/57595; it has been noted by GNATS.

From: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: tsutsui%ceres.dti.ne.jp@localhost
Subject: Re: port-sgimips/57595: Indigo2 cannot boot
Date: Wed, 30 Aug 2023 00:20:50 +0900

 > [   1.0000000] int0 at mainbus0 addr 0x1fbd9880
 > 
 > The address of device int0 shown in the last line is for Indy, not for Indigo2.
 > The address of device int0 is determined by the variable "mach_subtype", which is set on device ioc0.
 > I think the device ioc0 that sets the variable "mach_subtype" should be probed before device int0 and some other devices that reference the variable "mach_subtype".
 > 
 > The variable "mach_subtype" is referenced below for IP22/24.
 
  :
 
 > --- NetBSD-9.3-original/usr/src/sys/arch/sgimips/conf/files.sgimips	2019-03-24 02:44:49.000000000 +0900
 > +++ NetBSD-9.3-sgimips-indigo2/usr/src/sys/arch/sgimips/conf/files.sgimips	2023-08-28 20:16:09.474100759 +0900
 > @@ -38,8 +38,9 @@
 >  include	"external/bsd/sljit/conf/files.sljit"
 >  
 >  # Machine-dependent drivers
 > +include "arch/sgimips/ioc/files.ioc"
 >  include "arch/sgimips/dev/files.dev"
 > -include "arch/sgimips/ioc/files.ioc"		# depends on int0 (files.dev)
 > +include "arch/sgimips/ioc/files.oioc"		# depends on int0 (files.dev)
 >  defflag opt_sgimace.h MACE_NEEDS_DELAYS
 >  include "arch/sgimips/mace/files.mace"
 
 Your analysis looks correct, and it looks the Indigo2 support has been
 broken since the following Personal IRIS 4D/20 and 4D/25 support was added:
  https://mail-index.netbsd.org/source-changes/2009/02/12/msg217033.html
  http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/arch/sgimips/conf/files.sgimips#rev1.48
  https://github.com/NetBSD/src/commit/85716c44
 
 I.e. oioc(4) on IP6 and IP10 requires int(4) attached, but
 int(4) on IP22 (especially Indigo2) requires ioc(4) attached.
 
 In the perfect world[TM], we should reconsider device config tree
 (or handle it via config_defer(9) etc.), but for now it's safer
 and easier to split oioc(4) config definitions from files.ioc to
 independent files.oioc, as your patch does.
 
 ---
 Izumi Tsutsui