NetBSD Problem Report #54560

From gson@gson.org  Sun Sep 22 14:38:40 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 74BC07A158
	for <gnats-bugs@gnats.NetBSD.org>; Sun, 22 Sep 2019 14:38:40 +0000 (UTC)
Message-Id: <20190922143834.A7CD1989652@guava.gson.org>
Date: Sun, 22 Sep 2019 17:38:34 +0300 (EEST)
From: gson@gson.org (Andreas Gustafsson)
Reply-To: gson@gson.org (Andreas Gustafsson)
To: gnats-bugs@NetBSD.org
Subject: PXE netboot regression
X-Send-Pr-Version: 3.95

>Number:         54560
>Category:       kern
>Synopsis:       PXE netboot regression
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    manu
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Sep 22 14:40:00 +0000 2019
>Closed-Date:    Fri Sep 27 13:00:03 +0000 2019
>Last-Modified:  Fri Oct 04 11:35:01 +0000 2019
>Originator:     Andreas Gustafsson
>Release:        NetBSD-current, source date >= 2019.09.13.05.13.54
>Organization:

>Environment:
System: NetBSD
Architecture: x86_64
Machine: amd64
>Description:

I have an automated testing setup that netboots an INSTALL kernel and
performs a scripted installation of NetBSD/amd64 on physical hardware
using sysinst.

This recently stopped working.  As pxeboot starts to load the
kernel over TFTP, it restarts, leading to repeating output on the
console like this:

  >> NetBSD/x86 PXE boot, Revision 5.1 (Fri Sep 13 05:13:54 UTC 2019) (from NetBSD 9.99.11)
  >> Memory: 552/3668992 k
  Press return to boot now, any other key for boot menu
  booting netbsd - starting in 0 seconds.
  PXE BIOS Version 2.1
  Using PCI device at bus 3 device 0 function 0
  Ethernet address 98:4b:e1:67:68:98
  |
  >> NetBSD/x86 PXE boot, Revision 5.1 (Fri Sep 13 05:13:54 UTC 2019) (from NetBSD 9.99.11)
  >> Memory: 552/3668992 k
  Press return to boot now, any other key for boot menu
  booting netbsd - starting in 0 seconds.
  PXE BIOS Version 2.1
  Using PCI device at bus 3 device 0 function 0
  Ethernet address 98:4b:e1:67:68:98
  |
  >> NetBSD/x86 PXE boot, Revision 5.1 (Fri Sep 13 05:13:54 UTC 2019) (from NetBSD 9.99.11)
  >> Memory: 552/3668992 k
  Press return to boot now, any other key for boot menu
  booting netbsd - starting in 0 seconds.
  PXE BIOS Version 2.1
  Using PCI device at bus 3 device 0 function 0
  Ethernet address 98:4b:e1:67:68:98
  |

After this loop has repeated several more times, the system gives up
on netbooting and boots from its second priority boot device (in this
case, the hard disk).

A full console log is at:

  http://www.gson.org/netbsd/bugs/build/amd64-baremetal/2019/2019.09.13.05.13.54/install.log

A packet capture of the network traffic is at:

  http://www.gson.org/netbsd/bugs/pxeboot/2019.09.13.05.13.54.pcap

The problem started around the time of manu's changes to add multiboot 2
support on September 13.

>How-To-Repeat:

Attempt to netboot an amd64 system using pxeboot.

>Fix:

>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: kern-bug-people->manu
Responsible-Changed-By: gson@NetBSD.org
Responsible-Changed-When: Sun, 22 Sep 2019 14:41:19 +0000
Responsible-Changed-Why:
Over to committer.


From: manu@netbsd.org (Emmanuel Dreyfus)
To: gnats-bugs@netbsd.org, kern-bug-people@netbsd.org, netbsd-bugs@netbsd.org,
        gnats-admin@netbsd.org, gson@NetBSD.org,
        gson@gson.org (Andreas Gustafsson)
Cc: 
Subject: Re: kern/54560 (PXE netboot regression)
Date: Sun, 22 Sep 2019 16:57:43 +0200

 Hello

 Could you try this patch?

 Index: sys/arch/i386/stand/pxeboot/Makefile
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/i386/stand/pxeboot/Makefile,v
 retrieving revision 1.26
 diff -U4 -r1.26 Makefile
 --- sys/arch/i386/stand/pxeboot/Makefile        13 Sep 2019 02:19:46 -0000
 1.26
 +++ sys/arch/i386/stand/pxeboot/Makefile        22 Sep 2019 14:56:48 -0000
 @@ -28,8 +28,9 @@
  KERNMISCMAKEFLAGS="LIBKERN_ARCH=i386"
  .endif

  CPPFLAGS+= -DSLOW      # for libz
 +CPPFLAGS+= -DNO_MULTIBOOT2     # kern/54560

  .if (${BASE} == "pxeboot_ia32")
  # Take config values from patchable header
  CPPFLAGS+= -DSUPPORT_SERIAL=boot_params.bp_consdev



 -- 
 Emmanuel Dreyfus
 http://hcpnet.free.fr/pubz
 manu@netbsd.org

From: Andreas Gustafsson <gson@NetBSD.org>
To: manu@netbsd.org (Emmanuel Dreyfus)
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/54560 (PXE netboot regression)
Date: Mon, 23 Sep 2019 18:51:54 +0300

 Emmanuel Dreyfus wrote:
 > Could you try this patch?

 With the patch, the system netboots successfully (my first try failed,
 but that was probably caused by an unrelated issue).
 -- 
 Andreas Gustafsson, gson@NetBSD.org

From: manu@netbsd.org (Emmanuel Dreyfus)
To: gnats-bugs@netbsd.org, gnats-admin@netbsd.org, netbsd-bugs@netbsd.org,
        gson@gson.org (Andreas Gustafsson)
Cc: 
Subject: Re: kern/54560 (PXE netboot regression)
Date: Tue, 24 Sep 2019 01:28:25 +0200

 Andreas Gustafsson <gson@NetBSD.org> wrote:

 >  With the patch, the system netboots successfully (my first try failed,
 >  but that was probably caused by an unrelated issue).

 I am not sure there are real multiboot + PXE usage. Perhaps we can just
 disable multiboot 2 for PXE without looking further into what the
 problem actually is.

 In case you ar ecurious and want to look into it further, my first hunch
 is that the multiboot 2 code increases the pxeboot_ia32.bin file beyond
 64kB. You can easily test if this is the problem, remove CPPFLAGS+=
 -DNO_MULTIBOOT2 and instead add

 CPPFLAGS+= -DNO_GPT 
 CPPFLAGS+= -DNO_RAIDFRAME

 You will get a pxeboot_ia32.bin below 64 kB with multiboot 2 enabled. If
 that boots, then we have our answer.

 -- 
 Emmanuel Dreyfus
 http://hcpnet.free.fr/pubz
 manu@netbsd.org

From: Andreas Gustafsson <gson@gson.org>
To: manu@netbsd.org (Emmanuel Dreyfus)
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/54560 (PXE netboot regression)
Date: Tue, 24 Sep 2019 12:25:02 +0300

 Emmanuel Dreyfus wrote:
 > I am not sure there are real multiboot + PXE usage. Perhaps we can just
 > disable multiboot 2 for PXE without looking further into what the
 > problem actually is.
 > 
 > In case you ar ecurious and want to look into it further, my first hunch
 > is that the multiboot 2 code increases the pxeboot_ia32.bin file beyond
 > 64kB. You can easily test if this is the problem, remove CPPFLAGS+=
 > -DNO_MULTIBOOT2 and instead add
 > 
 > CPPFLAGS+= -DNO_GPT 
 > CPPFLAGS+= -DNO_RAIDFRAME
 > 
 > You will get a pxeboot_ia32.bin below 64 kB with multiboot 2 enabled. If
 > that boots, then we have our answer.

 That did not boot, but your hunch may be correct anyway.  Here's a
 summary of my test results, with the size of pxeboot_ia32.bin in each:

 Source date          CPPFLAGS                     Size      Result
 ==================================================================
 2019.09.13.01.34.19  none                         61664     pass
 2019.09.13.05.13.54  none                         73952     fail
 2019.09.22.18.31.59  -DNO_MULTIBOOT2              61664     pass
 2019.09.21.15.56.09  -DNO_GPT -DNO_RAIDFRAME      65760     fail

 That is, the test failed when and only when pxeboot_ia32.bin was
 larger than 65536 bytes.

 I guess the next question is, why does multiboot 2 use more space
 than GPT and raidframe combined?
 -- 
 Andreas Gustafsson, gson@gson.org

From: Andreas Gustafsson <gson@gson.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: kern/54560: PXE netboot regression
Date: Fri, 27 Sep 2019 11:41:09 +0300

 As discussed on tech-kern under the subject "x86 bootstrap features",
 it may be possible to increase the 64k size limit.  With the following
 patch, the boot succeeds for me:

 Index: Makefile
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/i386/stand/pxeboot/Makefile,v
 retrieving revision 1.27
 diff -u -r1.27 Makefile
 --- Makefile	23 Sep 2019 13:42:37 -0000	1.27
 +++ Makefile	27 Sep 2019 05:27:41 -0000
 @@ -66,7 +66,7 @@
  #CFLAGS= -O2 -fomit-frame-pointer -fno-defer-pop
  CFLAGS+= -Wall -Wmissing-prototypes -Wstrict-prototypes -Wno-main

 -SAMISCCPPFLAGS+= -DHEAP_START=0x10000 -DHEAP_LIMIT=0x30000
 +SAMISCCPPFLAGS+= -DHEAP_START=0x20000 -DHEAP_LIMIT=0x40000
  SAMISCMAKEFLAGS+= SA_USE_CREAD=yes	# Read compressed kernels

  CPPFLAGS+= -DPASS_BIOSGEOM
 Index: start_pxe.S
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/i386/stand/pxeboot/start_pxe.S,v
 retrieving revision 1.6
 diff -u -r1.6 start_pxe.S
 --- start_pxe.S	18 Mar 2011 17:46:26 -0000	1.6
 +++ start_pxe.S	27 Sep 2019 05:27:41 -0000
 @@ -69,7 +69,7 @@
  	# set up %ss and %sp
  	movl	$_end, %eax		/* top of bss */
  	shrl	$4, %eax		/* as a segment */
 -	addw	$0x1001, %ax		/* and + 64k */
 +	addw	$0x2001, %ax		/* and + 128k */
  	movw	%ax, %ss		/* for stack */
  	movw	$0xfffc, %sp		/* %sp at top of it */

 -- 
 Andreas Gustafsson, gson@gson.org

From: "Andreas Gustafsson" <gson@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/54560 CVS commit: src/sys/arch/i386/stand/pxeboot
Date: Fri, 27 Sep 2019 08:57:10 +0000

 Module Name:	src
 Committed By:	gson
 Date:		Fri Sep 27 08:57:10 UTC 2019

 Modified Files:
 	src/sys/arch/i386/stand/pxeboot: Makefile start_pxe.S

 Log Message:
 Incrase pxeboot code size limit from 64k to 128k.  Fixes PR kern/54560.
 The start_pxe.S part was suggested by mlelstv.


 To generate a diff of this commit:
 cvs rdiff -u -r1.27 -r1.28 src/sys/arch/i386/stand/pxeboot/Makefile
 cvs rdiff -u -r1.6 -r1.7 src/sys/arch/i386/stand/pxeboot/start_pxe.S

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

State-Changed-From-To: open->closed
State-Changed-By: gson@NetBSD.org
State-Changed-When: Fri, 27 Sep 2019 13:00:03 +0000
State-Changed-Why:
Fix committed.


From: "Martin Husemann" <martin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/54560 CVS commit: [netbsd-9] src/sys/arch/i386/stand/pxeboot
Date: Thu, 3 Oct 2019 17:26:00 +0000

 Module Name:	src
 Committed By:	martin
 Date:		Thu Oct  3 17:26:00 UTC 2019

 Modified Files:
 	src/sys/arch/i386/stand/pxeboot [netbsd-9]: Makefile start_pxe.S

 Log Message:
 Pull up following revision(s) (requested by manu in ticket #277):

 	sys/arch/i386/stand/pxeboot/start_pxe.S: revision 1.7
 	sys/arch/i386/stand/pxeboot/Makefile: revision 1.28

 Incrase pxeboot code size limit from 64k to 128k.  Fixes PR kern/54560.
 The start_pxe.S part was suggested by mlelstv.


 To generate a diff of this commit:
 cvs rdiff -u -r1.25.6.1 -r1.25.6.2 src/sys/arch/i386/stand/pxeboot/Makefile
 cvs rdiff -u -r1.6 -r1.6.60.1 src/sys/arch/i386/stand/pxeboot/start_pxe.S

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: "Martin Husemann" <martin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/54560 CVS commit: [netbsd-8] src/sys/arch/i386/stand/pxeboot
Date: Fri, 4 Oct 2019 11:34:18 +0000

 Module Name:	src
 Committed By:	martin
 Date:		Fri Oct  4 11:34:18 UTC 2019

 Modified Files:
 	src/sys/arch/i386/stand/pxeboot [netbsd-8]: Makefile start_pxe.S

 Log Message:
 Pull up following revision(s) (requested by manu in ticket #1400):

 	sys/arch/i386/stand/pxeboot/start_pxe.S: revision 1.7
 	sys/arch/i386/stand/pxeboot/Makefile: revision 1.28

 Incrase pxeboot code size limit from 64k to 128k.  Fixes PR kern/54560.
 The start_pxe.S part was suggested by mlelstv.


 To generate a diff of this commit:
 cvs rdiff -u -r1.24.10.1 -r1.24.10.2 src/sys/arch/i386/stand/pxeboot/Makefile
 cvs rdiff -u -r1.6 -r1.6.48.1 src/sys/arch/i386/stand/pxeboot/start_pxe.S

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

>Unformatted:

Home	PR Database Search