https://bugzilla.novell.com/show_bug.cgi?id=214568
Summary: Fully Virtualized openSUSE 10.2 guest will not install Product: openSUSE 10.2 Version: Alpha 5 plus Platform: Other OS/Version: Other Status: NEW Severity: Critical Priority: P5 - None Component: Xen AssignedTo: cgriffin@novell.com ReportedBy: lbendixs@novell.com QAContact: jdouglas@novell.com CC: cgriffin@novell.com
Using a Os 10.2 installation source I cannot install on either a oS10.2 dom0 or a SLES10 dom0. The last line on the VM as it is trying to begin the install says something about loading isolinux.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #1 from torsten.hallmann@novell.com 2006-10-25 06:08 MST ------- I assume you mean domU installations with vmx. This does also not work here with IBM T60 including a T2400 cpu.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #2 from torsten.hallmann@novell.com 2006-10-27 11:02 MST ------- Here the problem was solved after I enabled the VT functionality in BIOS. But to take effect, I had to disconnect powercable and remove battery.
https://bugzilla.novell.com/show_bug.cgi?id=214568
cgriffin@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|cgriffin@novell.com |carnold@novell.com
https://bugzilla.novell.com/show_bug.cgi?id=214568
cgriffin@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Critical |Blocker
https://bugzilla.novell.com/show_bug.cgi?id=214568
carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Blocker |Critical Status|NEW |ASSIGNED Summary|Fully Virtualized openSUSE |HVM Xen: Fully Virtualized openSUSE 10.2 guest |10.2 guest will not install |will not install
------- Comment #4 from carnold@novell.com 2006-11-06 10:07 MST ------- The following information is obtained from xen-dbg on the serial line when the guest freezes. It seems to indicate a real mode problem perhaps in isolinux.
(XEN) vmx_do_launch(): GUEST_CR3<=76761000, HOST_CR3<=76762000 (XEN) (GUEST: 1) HVM Loader (XEN) (GUEST: 1) Detected Xen v3.0.3_11774-0.01_x (XEN) (GUEST: 1) Writing SMBIOS tables ... (XEN) (GUEST: 1) Loading ROMBIOS ... (XEN) (GUEST: 1) Loading Cirrus VGABIOS ... (XEN) (GUEST: 1) Loading VMXAssist ... (XEN) (GUEST: 1) VMX go ... (XEN) (GUEST: 1) VMXAssist (Nov 2 2006) (XEN) (GUEST: 1) Memory size 256 MB (XEN) (GUEST: 1) E820 map: (XEN) (GUEST: 1) 0000000000000000 - 000000000009F000 (RAM) (XEN) (GUEST: 1) 000000000009F000 - 00000000000A0000 (Reserved) (XEN) (GUEST: 1) 00000000000A0000 - 00000000000C0000 (Type 16) (XEN) (GUEST: 1) 00000000000F0000 - 0000000000100000 (Reserved) (XEN) (GUEST: 1) 0000000000100000 - 000000000FFF0000 (RAM) (XEN) (GUEST: 1) 000000000FFF0000 - 000000000FFFA000 (ACPI Data) (XEN) (GUEST: 1) 000000000FFFA000 - 000000000FFFD000 (ACPI NVS) (XEN) (GUEST: 1) 000000000FFFD000 - 000000000FFFE000 (Type 19) (XEN) (GUEST: 1) 000000000FFFE000 - 000000000FFFF000 (Type 18) (XEN) (GUEST: 1) 000000000FFFF000 - 0000000010000000 (Type 17) (XEN) (GUEST: 1) 00000000FEC00000 - 0000000100000000 (Type 16) (XEN) (GUEST: 1) (XEN) (GUEST: 1) Start BIOS ... (XEN) (GUEST: 1) Starting emulated 16-bit real-mode: ip=F000:FFF0 (XEN) (GUEST: 1) 0x000FFFF0: 0xF000:0xFFF0 (0) <VM86_REAL> (XEN) (GUEST: 1) Trap (0x6) while in real mode (XEN) (GUEST: 1) eax 50060 ecx DFFFF edx 1 ebx 0 (XEN) (GUEST: 1) esp D71F4 ebp D0000 esi 0 edi D0031 (XEN) (GUEST: 1) eip E209 eflags 33006 cs F000 ds 0 (XEN) (GUEST: 1) es 0 fs 0 uss 0 uesp E (XEN) (GUEST: 1) ves F000 vds F000 vfs F000 vgs F000 (XEN) (GUEST: 1) trapno 6 errno 0 (XEN) (GUEST: 1) cr0 50032 cr2 0 cr3 0 cr4 651 (XEN) (GUEST: 1) Here is the eip dump: (XEN) (GUEST: 1) (XEN) (GUEST: 1) Dumping FE1B9..FE259 eip=FE209 (XEN) (GUEST: 1) 000D70D0: 66 61 69 6C 65 64 0A 00-42 6F 6F 74 20 66 72 6F failed..Boot fro (XEN) (GUEST: 1) 000D70E0: 6D 20 25 73 20 66 61 69-6C 65 64 0A 00 42 6F 6F m %s failed..Boo (XEN) (GUEST: 1) 000D70F0: 74 69 6E 67 20 66 72 6F-6D 20 25 73 2E 2E 2E 0A ting from %s.... (XEN) (GUEST: 1) 000D7100: 00 0A 00 25 73 20 25 73-0A 00 00 73 00 48 56 4D ..%s %s...s.HVM (XEN) (GUEST: 1) 000D7110: 41 73 73 69 73 74 20 42-49 4F 53 2C 20 25 64 20 Assist BIOS, %d (XEN) (GUEST: 1) 000D7120: 63 70 75 25 73 2C 20 00-55 6E 69 6D 70 6C 65 6D cpu%s, .Unimplem (XEN) (GUEST: 1) 000D7130: 65 6E 74 65 64 20 73 68-75 74 64 6F 77 6E 20 73 ented shutdown s (XEN) (GUEST: 1) 000D7140: 74 61 74 75 73 3A 20 25-30 32 78 0A 00 4B 65 79 tatus: %02x..Key (XEN) (GUEST: 1) 000D7150: 62 6F 61 72 64 20 65 72-72 6F 72 3A 25 75 0A 00 board error:%u.. (XEN) (GUEST: 1) 000D7160: 62 69 6F 73 5F 70 72 69-6E 74 66 3A 20 75 6E 6B bios_printf: unk (XEN) (GUEST: 1) Halt called from %eip 0xD0FA2
This is not a considered blocker at this time.
https://bugzilla.novell.com/show_bug.cgi?id=214568
carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |snwint@novell.com
------- Comment #5 from carnold@novell.com 2006-11-09 10:46 MST ------- I have taken the isolinux.bin and isolinux.cfg from sles10 and dropped them on the openSUSE 10.2 Beta 2 build and after creating a CD I was then able to start a fully virtualized guest install without the trap. Debugging into the 3.31 version of isolinux (the community version, not our openSUSE modified one in 10.2) I noticed in the following bit of code that if I reduced the 512 to 128 I was able to boot the guest.
cli ; Abandon hope, ye who enter here mov si,trackbuf mov di,07C00h mov cx,512 ; Probably overkill, but should be safe rep movsd lss sp,[cs:InitStack] jmp 0:07C00h ; Jump to new boot sector
This may only be a symptom of some other problem but I would like some help debugging this further as I suspect the problem to be in isolinux (at least when booting under xen).
https://bugzilla.novell.com/show_bug.cgi?id=214568
carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Critical |Blocker
------- Comment #6 from carnold@novell.com 2006-11-09 11:06 MST ------- We need to get this resolved for RC1. Upgrading to a blocker.
https://bugzilla.novell.com/show_bug.cgi?id=214568
snwint@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |snwint@novell.com Status|NEEDINFO |ASSIGNED Info Provider|snwint@novell.com |
------- Comment #7 from snwint@novell.com 2006-11-10 03:56 MST ------- The cited code is only used for the 'localboot' option, that is to boot from disk. But the report is about not being able to install, where that code is never run. Or not?
Could you please state more exactly what goes on?
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #8 from lbendixs@novell.com 2006-11-10 09:59 MST ------- Hi Steffen, Sorry for the succinctness in my original report. Heres what I did:
1. Put in a 10.2 install DVD. (fails for CD also) 2. Open yast2 Virtual Machine Manager. 3. Click Add to add a VM 4. Select full virtualization and CD/DVD and click next. 5. At this point a VM is created which attempts to boot off of the DVD so that installation can occur. 6. The screen displays a message about isolinux starting but the isolinux menu never appears.
Thanks for looking into this for us, Lynn
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #9 from snwint@novell.com 2006-11-13 03:25 MST ------- Ok, so what happens is: syslinux 3.11 works, 3.31 (both the unmodified and the SuSE version) does not.
It starts working if you copy only 512 bytes in the code fragment in comment 5. But that code is not executed (it's the localboot path, but you are going for installation). So a good question would be: why?
Can you put a hardware break point (or whatever equivalent of that in xen) on that 512 and see whether it is read accidentally by some other part of the code? Otherwise I don't see why it could make a difference.
BTW, does the bug occur only with intel-vt or with amd-svm, too?
BTW2, though I can't make much sense out of the log in comment 4, but it crashed at f000:e209 which would be inside the (emulated) BIOS, not the boot loader. Or not?
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #10 from aj@novell.com 2006-11-17 06:48 MST ------- What is the status here? What needs to be done to get this fixed?
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #11 from carnold@novell.com 2006-11-17 09:02 MST ------- This also fails on amd-svm in a slightly different way. The guest actually crashes instead of hanging with the following message.
(XEN) (GUEST: 1) Booting from CD-Rom... (XEN) 0, This opcode isn't handled yet! (XEN) handle_mmio: failed to decode instruction (XEN) mmio opcode: va 0xa0000, gpa 0xa0000, len 2: 00 00 (XEN) domain_crash_sync called from platform.c:915 (XEN) Domain 1 (vcpu#0) crashed on cpu#1: (XEN) ----[ Xen-3.0.3_11774-0.02_xtp x86_64 debug=y Not tainted ]---- (XEN) CPU: 1 (XEN) RIP: 0018:[<0000000000005d94>] (XEN) RFLAGS: 0000000000000246 CONTEXT: hvm (XEN) rax: 0000000000000093 rbx: 0000000000000640 rcx: 0000000000000258 (XEN) rdx: 0000000000000640 rsi: 0000000000801017 rdi: 0000000000000000 (XEN) rbp: 0000000000000000 rsp: 00000000000551f4 r8: 0000000000000000 (XEN) r9: 0000000000000000 r10: 0000000000000000 r11: 0000000000000000 (XEN) r12: 0000000000000000 r13: 0000000000000000 r14: 0000000000000000 (XEN) r15: 0000000000000000 cr0: 0000000000000011 cr4: 0000000000000000 (XEN) cr3: 0000000000000000 cr2: 0000000000000000
I also tried installing a 10.2 guest on a sles10 GA dom0 and saw the same results. The unhandled opcode of 0 may indicate a memory overwrite.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #12 from snwint@novell.com 2006-11-17 09:47 MST ------- Using rip of 0x5d94 above, the relevant code fragment would be:
13721 00005D8A 0FAF15[6B010000] imul edx,[pixel_bytes] 13722 13723 restore_bg_20: 13724 00005D91 52 push edx 13725 13726 restore_bg_30: 13727 00005D92 26AC es lodsb 13728 00005D94 658807 mov [gs:edi],al ^^^^^^^^ 13729 00005D97 6647 inc di 13730 00005D99 7505 jnz restore_bg_50 13731 00005D9B E8AFF0FFFF call inc_winseg 13732 restore_bg_50: 13733 00005DA0 4A dec edx 13734 00005DA1 75EF jnz restore_bg_30
indicating it crashes the first time it tries to write to the screen memory at 0xa0000. The only difference to 10.1 that comes to mind is that the code now runs in 32bit protected mode instead of real mode.
Could it be the vga emulation has problems with that? I don't think the code was overwritten with 0 (and there isn't anything special about opcode 0 anyway on x86).
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #15 from carnold@novell.com 2006-11-20 16:50 MST ------- For a sles10ga hvm guest, realmode is always set (at least during boot) and the code segment equals 0xC000 and later switches to 0x4004. For a 10.2 hvm guest, it starts out the same as sles10 with realmode set to 1 and the code seg equal to 0xC000. It fails when suddenly realmode changes to zero and the code seg changes to 0x18.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #16 from snwint@novell.com 2006-11-21 03:51 MST ------- ad 14: Yes, in PM the segment base is not always 0. CS points to the current code segment (to avoid relocation), DS normally is the matching data segment for that (to have write access). And, selectors with base 0xa0000 are used for display access.
https://bugzilla.novell.com/show_bug.cgi?id=214568
jbeulich@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |snwint@novell.com
------- Comment #17 from jbeulich@novell.com 2006-11-21 08:53 MST ------- We have this fixed now. However, it now hangs in what appears to be an endless loop. To analyse this, we really need to understand what isolinux and the gfxboot stuff are doing - dumping the VMCB I get a consistent value of 4004:6c1f for CS:IP that the boot process is sitting at (but that might just be the place where the last intercept happened, e.g. the write to CR0 to re-enable protected mode or the iret/sti that re-enabled interrupts). Sources and knowledge how these map to specific addresses at run time would be rather helpful in debugging this. (I have a suspicion that the overhead of handling interrupts by reflecting them to real mode may be so high that the process is live locking, but that would be more likely on Intel than on AMD.)
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #19 from snwint@novell.com 2006-11-21 09:12 MST ------- Created an attachment (id=106417) --> (https://bugzilla.novell.com/attachment.cgi?id=106417&action=view) gfxboot code listing
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #20 from snwint@novell.com 2006-11-21 09:18 MST ------- Created an attachment (id=106419) --> (https://bugzilla.novell.com/attachment.cgi?id=106419&action=view) same as above, but compressed
Hm, at least I can't read the attachment above, maybe too large. Trying compressed.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #21 from carnold@novell.com 2006-11-21 11:23 MST ------- With the latest patch from Jan the Conroe outputs the following error. This is different than the trap generated in comment #4. The guest does remain up in a hung state after this message.
(XEN) (GUEST: 1) Start BIOS ... (XEN) (GUEST: 1) Starting emulated 16-bit real-mode: ip=F000:FFF0 (XEN) (GUEST: 1) rombios.c,v 1.138 2005/05/07 15:55:26 vruppert Exp $ (XEN) (GUEST: 1) Remapping master: ICW2 0x8 -> 0x20 (XEN) (GUEST: 1) Remapping slave: ICW2 0x70 -> 0x28 (XEN) (GUEST: 1) VGABios $Id: vgabios.c,v 1.61 2005/05/24 16:50:50 vruppert Exp $ (XEN) (GUEST: 1) HVMAssist BIOS, 1 cpu, $Revision: 1.138 $ $Date: 2005/05/07 15:55:26 $ (XEN) (GUEST: 1) (XEN) (GUEST: 1) ata0-0: PCHS=8322/16/63 translation=lba LCHS=522/255/63 (XEN) (GUEST: 1) ata0 master: QEMU HARDDISK ATA-7 Hard-Disk (4096 MBytes) (XEN) (GUEST: 1) ata0 slave: Unknown device (XEN) (GUEST: 1) ata1 master: QEMU CD-ROM ATAPI-4 CD-Rom/DVD-Rom (XEN) (GUEST: 1) ata1 slave: Unknown device (XEN) (GUEST: 1) (XEN) (GUEST: 1) Booting from CD-Rom... (XEN) (GUEST: 1) Unknown opcode at 4004:0001=0x40041 (XEN) (GUEST: 1) Halt called from %eip 0xD3F5A
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #22 from jbeulich@novell.com 2006-11-22 03:23 MST ------- With an extended patch install is now coming up properly on AMD-V.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #23 from jdouglas@novell.com 2006-11-22 16:36 MST ------- I have confirmed that this is working for AMD with Jan's patch, but it still does not work on Intel. I have tried to workaround the issue using tricks that worked for SLES 10, but those efforts have failed.
https://bugzilla.novell.com/show_bug.cgi?id=214568
aj@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Blocker |Critical
------- Comment #24 from aj@novell.com 2006-11-29 05:09 MST ------- Let's solve this via Online Update, not a shipment blocker for 10.2.
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #25 from jan.albrecht@bertelsmann.de 2006-12-14 23:50 MST ------- Hi,
I've the same problem on a DL380G5 (Intel VT). Is there any solution in development for SLES10?
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #26 from carnold@novell.com 2006-12-15 08:35 MST ------- The problem was solved for AMD-VT and we are still looking to resolve it for Intel VT.
https://bugzilla.novell.com/show_bug.cgi?id=214568
jan.albrecht@bertelsmann.de changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jan.albrecht@bertelsmann.de
------- Comment #27 from jan.albrecht@bertelsmann.de 2007-01-15 03:11 MST ------- Is here any update in place?
Thanks Jan
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #28 from jan.albrecht@bertelsmann.de 2007-02-16 00:48 MST ------- Bug is still present in boot.iso of 10.3 from 14th Feb.
Do I have to open a new bug report for this or will manage that in this one?
Jan
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #29 from carnold@novell.com 2007-02-16 09:07 MST ------- Continue to use this bug. Intel was working on a fix but may have stopped in favor of new work being done by xensource in the 3.0.5 xen version. We will grab xen 3.0.5 for 10.3 when it becomes available in about a month (the release date for 3.0.5 has not been set).
https://bugzilla.novell.com/show_bug.cgi?id=214568
lbendixs@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |nine@detonation.org
------- Comment #30 from lbendixs@novell.com 2007-02-28 14:53 MST ------- *** Bug 248669 has been marked as a duplicate of this bug. ***
https://bugzilla.novell.com/show_bug.cgi?id=214568
fkohler@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Critical |Normal
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #31 from jan.albrecht@bertelsmann.de 2007-03-19 01:35 MST ------- That an Intel fix is needed to fix this issue seem to be a bit odd to me.
The latest RedHat AS 5 does boot and install with no problem at all so why isn't this possible for openSuSE 10.2 or 10.3?
Waiting for XEN 3.0.5 can take some time which most users don't have...
https://bugzilla.novell.com/show_bug.cgi?id=214568
------- Comment #32 from snwint@novell.com 2007-03-19 04:00 MST ------- There is nothing odd about it. Virtualization via Xen on Intel is not yet perfect and can't deal with certain situations. Including, unfortunately, our bootloader graphics code. AFAIUI Intel is trying to get this right.
https://bugzilla.novell.com/show_bug.cgi?id=214568#c33
--- Comment #33 from Jan Albrecht jan.albrecht@bertelsmann.de 2007-06-26 02:43:00 MST --- Hi, are there any news in this case?
Thanks Jan
https://bugzilla.novell.com/show_bug.cgi?id=214568#c34
Charles Arnold carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Component|Xen |Xen Product|openSUSE 10.2 |openSUSE 10.3 Summary|HVM Xen: Fully Virtualized openSUSE 10.2 guest |HVM Xen: Fully Virtualized openSUSE 10.x guest |will not install |will not install Target Milestone|--- |unspecified Version|Alpha 5 plus |unspecified
--- Comment #34 from Charles Arnold carnold@novell.com 2007-08-30 15:57:31 MST --- Retesting this on openSuse 10.3. There are no plans to backport a newer xen to 10.2.
https://bugzilla.novell.com/show_bug.cgi?id=214568#c35
Stephen Shaw stshaw@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |stshaw@novell.com Severity|Normal |Critical Version|unspecified |Beta 2
--- Comment #35 from Stephen Shaw stshaw@novell.com 2007-09-04 11:35:50 MST --- This appears to still be an issue with opensuse 10.3. It stops when trying to load ISOLINUX 3.31 0x46c601a6
https://bugzilla.novell.com/show_bug.cgi?id=214568
Stephen Shaw stshaw@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|Beta 2 |Beta 3
https://bugzilla.novell.com/show_bug.cgi?id=214568
Stephan Kulow coolo@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |coolo@novell.com Flag| |SHIP_STOPPER-
https://bugzilla.novell.com/show_bug.cgi?id=214568
Frank Kohler fkohler@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P3 - Medium
https://bugzilla.novell.com/show_bug.cgi?id=214568
Charles Arnold carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |lbendixs@novell.com
https://bugzilla.novell.com/show_bug.cgi?id=214568
Frank Kohler fkohler@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Severity|Critical |Normal Summary|HVM Xen: Fully Virtualized openSUSE 10.x guest |HVM: openSUSE 10.x guest will not install |will not install |
https://bugzilla.novell.com/show_bug.cgi?id=214568
User carnold@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=214568#c37
Charles Arnold carnold@novell.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |WONTFIX
--- Comment #37 from Charles Arnold carnold@novell.com 2009-03-11 11:20:24 MST --- Upgrading to newer versions of openSUSE provides better HVM support in Xen.
-
bugzilla_noreply＠novell.com