NetBSD Problem Report #54252

From gson@gson.org  Fri May 31 10:20:16 2019
Return-Path: <gson@gson.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits))
	(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 6A6367A184
	for <gnats-bugs@gnats.NetBSD.org>; Fri, 31 May 2019 10:20:16 +0000 (UTC)
Message-Id: <20190531102012.2A689989DFB@guava.gson.org>
Date: Fri, 31 May 2019 13:20:12 +0300 (EEST)
From: gson@gson.org (Andreas Gustafsson)
Reply-To: gson@gson.org (Andreas Gustafsson)
To: gnats-bugs@NetBSD.org
Subject: NetBSD/hpcmips install fails under GXemul 0.6.1
X-Send-Pr-Version: 3.95

>Number:         54252
>Category:       pkg
>Synopsis:       NetBSD/hpcmips install fails under GXemul 0.6.1
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    pkg-manager
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri May 31 10:25:01 +0000 2019
>Closed-Date:    Mon Jun 03 09:25:42 +0000 2019
>Last-Modified:  Mon Jun 03 09:25:42 +0000 2019
>Originator:     Andreas Gustafsson
>Release:        NetBSD-current
>Organization:

>Environment:
System: NetBSD
Architecture: mips
Machine: hpcmips
>Description:

The gxemul-based NetBSD/hpcmips testbed has not successfully completed
an install since gxemul was upgraded from version 0.6.0.1 to 0.6.1.
The currently installed version is gxemul-0.6.1nb2 from pkgsrc.

The other gxemul-based MIPS testbed, for NetBSD/pmax, is working.

Log output from a recent install attempt is at:

  http://releng.netbsd.org/b5reports/hpcmips/2019/2019.05.29.11.27.34/install.log

which shows a timeout after the error message

  warning: LOW reference: vaddr=0x00000000000002f0, exception TLBL, pc=0xffffffff802b7400 <(no symbol)> ]

Other runs appear to time out during set extraction without any error
message, for example:

  http://releng.netbsd.org/b5reports/hpcmips/2019/2019.05.30.11.15.29/install.log

>How-To-Repeat:

>Fix:

>Release-Note:

>Audit-Trail:
From: David Holland <dholland-pbugs@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: pkg/54252: NetBSD/hpcmips install fails under GXemul 0.6.1
Date: Sat, 1 Jun 2019 06:46:44 +0000

 On Fri, May 31, 2019 at 10:25:01AM +0000, Andreas Gustafsson wrote:
  > The gxemul-based NetBSD/hpcmips testbed has not successfully completed
  > an install since gxemul was upgraded from version 0.6.0.1 to 0.6.1.
  > The currently installed version is gxemul-0.6.1nb2 from pkgsrc.
  > 
  > The other gxemul-based MIPS testbed, for NetBSD/pmax, is working.
  > 
  > Log output from a recent install attempt is at:
  > 
  >   http://releng.netbsd.org/b5reports/hpcmips/2019/2019.05.29.11.27.34/install.log
  > 
  > which shows a timeout after the error message
  > 
  >   warning: LOW reference: vaddr=0x00000000000002f0, exception TLBL, pc=0xffffffff802b7400 <(no symbol)> ]

 One of the less endearing properties of mips is that all-bits-0 is a
 nop, so if you jump off the deep end you're likely to keep executing
 until you run into something unrelated.

 Can you look at the kernel from this (e.g. with nm -n) and check where
 the address 0x(80)2b7400 is relative to its .etext, .end, etc.?

 It's suspicious that it's round like that, and yet it's not a page
 boundary.

 -- 
 David A. Holland
 dholland@netbsd.org

From: Andreas Gustafsson <gson@gson.org>
To: David Holland <dholland-pbugs@netbsd.org>
Cc: gnats-bugs@netbsd.org
Subject: Re: pkg/54252: NetBSD/hpcmips install fails under GXemul 0.6.1
Date: Sat, 1 Jun 2019 11:27:28 +0300

 David Holland wrote:
 >  Can you look at the kernel from this (e.g. with nm -n) and check where
 >  the address 0x(80)2b7400 is relative to its .etext, .end, etc.?

 This is the installation RAMDISK kernel which is stripped, but
 /bracket/hpcmips/test/2019.05.29.11.27.34/release/hpcmips/installation/netbsd.symbols.gz
 on lyta.netbsd.org contains:

   ...
   ffffffff80368864 R spec_nfsv2nodeop_entries
   ffffffff8036885c R spec_nfsv2nodeop_opv_desc
   ffffffff808ae094 B spec_nfsv2nodeop_p
   ffffffff802b74ac T spec_node_destroy
   ffffffff802b73f4 T spec_node_getmountedfs
   ffffffff802b705c T spec_node_init
   ...

 -- 
 Andreas Gustafsson, gson@gson.org

From: Andreas Gustafsson <gson@gson.org>
To: gnats-bugs@netbsd.org
Cc: David Holland <dholland-pbugs@netbsd.org>
Subject: Re: pkg/54252: NetBSD/hpcmips install fails under GXemul 0.6.1
Date: Sat, 1 Jun 2019 12:40:05 +0300

 I reported the issue upstream, and Anders Gavare is
 investigating a possible cause and fix:

   https://sourceforge.net/p/gxemul/mailman/message/36682877/

 -- 
 Andreas Gustafsson, gson@gson.org

From: David Holland <dholland-pbugs@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: pkg/54252: NetBSD/hpcmips install fails under GXemul 0.6.1
Date: Sun, 2 Jun 2019 20:50:55 +0000

 On Sat, Jun 01, 2019 at 09:45:01AM +0000, Andreas Gustafsson wrote:
  >  I reported the issue upstream, and Anders Gavare is
  >  investigating a possible cause and fix:
  >  
  >    https://sourceforge.net/p/gxemul/mailman/message/36682877/

 Translation caching is evil in general :-)  If there's a reasonable
 way to do so, check if the oddities occur immediately after exception
 return. Also, just for fun, check if 0x802b73fc contains a jump or
 branch instruction. The interactions between branch delay slots, page
 crossings, exception returns, and mmu lookups have broken more than
 one mips implementation...

 (pass this on if you feel so inclined, not in a position to send it
 directly)

 -- 
 David A. Holland
 dholland@netbsd.org

State-Changed-From-To: open->closed
State-Changed-By: gson@NetBSD.org
State-Changed-When: Mon, 03 Jun 2019 09:25:42 +0000
State-Changed-Why:
gxemul-0.6.1nb3 has a provisional fix, and Anders Gavare is working
on a cleaner fix for the next release.


>Unformatted:

Home	PR Database Search