Bug 104271 - Boot crashes on Lenovo W520 with KVM and Discrete Graphic enabled in BIOS
Summary: Boot crashes on Lenovo W520 with KVM and Discrete Graphic enabled in BIOS
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: PCI (show other bugs)
Hardware: Intel Linux
: P1 blocking
Assignee: virtualization_kvm
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-09-09 09:05 UTC by Hans Streibel
Modified: 2016-02-15 15:00 UTC (History)
4 users (show)

See Also:
Kernel Version: 4.1.6
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Output of lspci -vvv (39.56 KB, text/plain)
2015-09-11 09:23 UTC, Hans Streibel
Details

Description Hans Streibel 2015-09-09 09:05:03 UTC
Booting stops/crashes on a Lenovo W520 Notebook with these BIOS Settings:
- Virtualization
  Intel Virtualization Technology: Enabled
  Intel VT-d Feature: Enabled
- Display
  Boot Display Device: ThinkPad LCD
  Graphics Device: Discrete Graphics
  OS Detection for NVIDIA Optimus: Disabled

Booting starts and the last lines show that the kernel cannot access the hard disks (lines starting with ata2 and ata1). The last lines are:

[ TIME ] Timed out waiting for device dev-disk-by\x2duuid-c30...4394b6b.device.
[ DEPEND ] Dependency failed for /dev/disk/by-uuid/c305bd9f-9d6...7-359b94394b6b.
[ DEPEND ] Dependency failed for Swap.

After that booting stops.

Additional Information:
- This is no hardware fault. The error can be reproduced on a different W520.

- Booting also crashes with these kernels:
  -  3.16.0 (kernel from Debian Jessie)
  -  4.1.0 (backported kernel to Debian Jessie)

- Booting succeeds if:
  - both virtualization flags are set to Disabled in the BIOS
  or
  - Graphics Device is set to "Optimus" in the BIOS
Comment 1 Huaitong Han 2015-09-09 12:41:57 UTC
Does it work well on windows? I guess it's a BIOS problem. finding a avaliable disk device is the job of BIOS, and the problem just right occurs when you change the BIOS settings.
Comment 2 Hans Streibel 2015-09-09 16:36:48 UTC
Yes, it works on Windows 7. I did not recognize any problems there.
Ok, I did not try any virtualization like Virtual Box under Windows,
but everything else seems to work.

Even if it is a BIOS problem, then Windows obviously found a way to handle it.
Comment 3 Alex Williamson 2015-09-09 19:25:14 UTC
You might get more attention moving this bug to Drivers/PCI and including the full dmesg or console log of the failure and lspci -vvv so we know something about your system.  If it doesn't boot, there are always screen shots or netconsole if you're serial port challenged.

KVM is a hypervisor, it has nothing to do with your system not booting.  More likely it's something with the IOMMU.  Does disabling VT-d help?  What about booting with iommu=pt?  BTW, reproducing on another W520 doesn't mean the hardware isn't broken, it just means this isn't a point defect.  Hardware is broken a disturbing amount of the time.
Comment 4 Hans Streibel 2015-09-11 09:23:22 UTC
Created attachment 187321 [details]
Output of lspci -vvv
Comment 5 Hans Streibel 2015-09-11 09:38:59 UTC
Ok, will move it to Drivers/PCI.

Output of lspci -vvv now is in the appendix.
Output of dmesg is not included because that only shows output of a successfully booted kernel. I even installed bootlogd but its log file does not show error messages.

I made some pictures with my camera but I do not dare to attach those somewhat big jpg files here. But if you really need them I can attach them.

The trouble starts (most of the time) right behind the line:
shpchp: Standard Hot Plug PCI Controller Driver Version: 0.4

Yes, disabling VT-d did help. Booting succeeded again.
Using "iommu=pt" however did not help. Same symptoms.

BTW, I wanted to express that my special notebook is not broken.
Not more broken that all (many/most) of the others W520s around.
Comment 6 Hans Streibel 2015-09-29 09:31:07 UTC
As Alex suggested I moved this bug to Drivers/PCI to get more attention.
However now I have the impression that this bug does not get any attention at all any more.

Did I miss any necessary action beside moving this to Drivers/PCI?

Note You need to log in before you can comment on or make changes to this bug.