Bug 1154225

Summary: AMD-Vi page faults on AMD FX-8350 when IOMMU or HPET enabled
Product: [Fedora] Fedora Reporter: bob <redzilla.coralnut>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED WONTFIX QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: urgent Docs Contact:
Priority: unspecified    
Version: 20CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda, mchehab
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2014-10-18 13:13:55 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
lspci -vvv none

Description bob 2014-10-17 23:05:03 UTC 
Description of problem:

AMD-Vi: Event logged [IO_PAGE_FAULT device=01:00.0 domain=0x0012 address=0x0000000000001000 flags=0x0000]

This error message is continuously flooded to /var/log/messages, causing the following:

a. /var/log/messages grows to several GB in size
b. syslog saturates a core at 100% utilization

Version-Release number of selected component (if applicable):

kernel: 3.16.4-200.fc20.x86_64 
nvidia: kmod-nvidia-3.16.4-200.fc20.x86_64.x86_64 1:331.89-2.fc20 

How reproducible:

Occurs if either IOMMU or HPET is enabled in BIOS.  Either one alone is sufficient condition for the error to occur.  Disabling both is required for the error not to occur.


Steps to Reproduce:
1. Enable either IOMMU or HPET on an AMD-FX system.  (In my case, the ASUS M5A97LE R2.0 motherboard with a 970 AMD chipset) 
2. check CPU utilization and system logs
3.

Actual results:

AMD-Vi page faults flood the system logs and monopolize a CPU core at 100% utilization.  Log files grow to enormous size.

Expected results:

No page faults.

Additional info:

Graphics card is an nVidia Quadro 400 workstation GPU, using the proprietary driver.  (Let the finger pointing begin!)  The verbose output of lspci is attached.  It is not possible to post the output of dmesg, as dmesg also becomes flooded with the same error message and all other pertinent information is scrolled away and lost.
Comment 1 bob 2014-10-17 23:10:00 UTC 
Created attachment 948029 [details]
lspci -vvv
Comment 2 Josh Boyer 2014-10-18 13:13:55 UTC 
The device called out by the IOMMU is indeed the nVidia card.  I have no idea if the driver is capable of doing the proper setup to run under an IOMMU and we can't figure that out.  Sorry.