FS#46589 - [linux] [linux-lts] Kernel failure 4.2 on AMD core.

Attached to Project: Arch Linux
Opened by Arnaud (arnaud) - Tuesday, 06 October 2015, 20:44 GMT
Last edited by Andreas Radke (AndyRTR) - Saturday, 27 August 2016, 12:07 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Tobias Powalowski (tpowa)
Andreas Radke (AndyRTR)
Architecture x86_64
Severity Critical
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 4
Private No

Details

Description: Can not start your computer with this kernel > 4.2
Not found port SATA, but the USB ports are functional.

No problem with the LTS kernel.

See attached file dmesg.

https://forums.archlinux.fr/viewtopic.php?f=15&t=17372
https://lkml.org/lkml/2015/7/28/902

Additional info:
Desktop computer with:
AMD A8-5500 APU with Radeon (tm) HD Graphics

No problem with Intel core.

Steps to reproduce: at every startup.



   dmesg.log (49.3 KiB)
This task depends upon

Closed by  Andreas Radke (AndyRTR)
Saturday, 27 August 2016, 12:07 GMT
Reason for closing:  Upstream
Comment by stoffl (stoffl) - Sunday, 11 October 2015, 22:08 GMT
Confirmed here on Supermicro Server AMD Board, broken on all 4.2.X Kernels no matter if used with grsec, zen or other kernel patch sets/flavours.

Issue seams to be AMD-Vi:

[ 1.721997] AMD-Vi: Event logged [IO_PAGE_FAULT device=00:11.0 domain=0x0005 address=0x0000000000000000 flags=0x0020]

Disabling iommu via kernel commandline fixes the issue:

iommu=soft

But leaves the user/operator without memory mapped io for advanced virtualization support, AMD equivalent to Intel vt-d. So no hardware pass through into virtual machines.
May be an upstream issue?
Comment by Jacob Litewski (HACKhalo2) - Monday, 26 October 2015, 03:06 GMT
I have the same issue.

Desktop with A10-5800K would not boot unless I disabled IOMMU in BIOS. It worked before the 4.2 kernel, so I'm assuming it's a regression either upstream or here.
Comment by Jacob Litewski (HACKhalo2) - Monday, 26 October 2015, 15:04 GMT
lspci -vvks output for the device that the IO_PAGE_FAULT issues on my machine
   output (0.9 KiB)
Comment by Britt Yazel (brittyazel) - Sunday, 29 November 2015, 08:09 GMT
I believe I also have this issue with an AMD fx 8120 board on 990fx chipset. Using kernel 4.2.x 3 out of every 5 boots or so results in my boot locking up and become unresponsive. Kernel 4.1.x lts seems to not have this issue. I do not know how to help debug this
Comment by Arnaud (arnaud) - Monday, 14 December 2015, 02:25 GMT
With the new 4.3.2 kernel available in the repository [Testing],
this bug is still not resolved.
Comment by Arnaud (arnaud) - Saturday, 26 December 2015, 22:16 GMT
It's the same with version 4.3.3
Comment by Arnaud (arnaud) - Sunday, 17 January 2016, 01:44 GMT
It's the same with version 4.4.0-3
Comment by Arnaud (arnaud) - Wednesday, 20 January 2016, 23:23 GMT
It's the same with version 4.4.0-4
Comment by Arnaud (arnaud) - Monday, 01 February 2016, 16:52 GMT
Here is the copy of the dmesg Xubuntu 15.10 distribution.
There are the same mistakes but that does not prevent the system from starting.
Comment by alex (kabolt) - Monday, 22 February 2016, 19:43 GMT
had maybe the same problem:
bugfix:
https://paulphilippov.com/articles/how-to-fix-slow-boot-with-ata-errors
(in /usr/lib/udev...)
and then regenerating the linux kernel
Comment by Arnaud (arnaud) - Tuesday, 23 February 2016, 00:23 GMT
It's the same with version 4.4.2-1
Comment by Claire Farron (clfarron4) - Monday, 07 March 2016, 13:46 GMT
With my AMD FX 6300 and Gigabyte 990XA-UD3, I had to enable IOMMU in BIOS and use the iommu=soft kernel parameter to get it to work properly.

But the first comment point about the disadvantage of this should be remembered.
Comment by Arnaud (arnaud) - Tuesday, 22 March 2016, 01:12 GMT
With the new 4.5 kernel available in the repository [Testing],
this bug is still not resolved.
Comment by Arnaud (arnaud) - Sunday, 17 April 2016, 23:33 GMT
Since the update to version 4.4 of the package linux-lts April 16 , the same bug appeared.
Comment by Andreas Radke (AndyRTR) - Thursday, 28 July 2016, 07:54 GMT
Is this still happening in linux 4.7 release? Is there any proper upstream bug to follow?

Loading...