Hi All

I am trying NVME pass through in VM and able to do it with host OS 4.6
and guest OS SL 2.6.696. i am using below pci pass through rule added in my vm.xml and starting the VM with virsh start command

<hostdev mode='subsystem' type='pci' managed='yes'>
<source>
<address domain='0x0000' bus='0xd8' slot='0x00' function='0x0'/>
</source>
</hostdev>

it is not working in 4.5 host OS and i am getting below panic as soon as starting VM with virsh start command in 4.5 with the pci rule in xml

please suggest if some specific patch went into 4.6.0 kernel from 4.5.7 for the same

br0: port 1(vnet0) entered listening state
virbr0: port 1(vnet0) entered listening state
vfio_ecap_init: 0000:d8:00.0 hiding ecap 0x19@0x2a0
mce: [Hardware Error]: CPU 11: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8104c95f> {acpi_processor_ffh_cstate_enter+0x8f/0xc0}
mce: [Hardware Error]: TSC 38f053f78f6
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 22 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: CPU 14: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8104c95f> {acpi_processor_ffh_cstate_enter+0x8f/0xc0}
mce: [Hardware Error]: TSC 38f053f74a0
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 28 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: CPU 13: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff815bb09f> {pci_conf1_write+0xdf/0xf0}
mce: [Hardware Error]: TSC 38f053f7e60
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 26 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: CPU 16: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8104c95f> {acpi_processor_ffh_cstate_enter+0x8f/0xc0}
mce: [Hardware Error]: TSC 38f053f7bc4
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 32 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: CPU 19: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8104c95f> {acpi_processor_ffh_cstate_enter+0x8f/0xc0}
mce: [Hardware Error]: TSC 38f053f880c
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 38 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: CPU 12: Machine Check Exception: 5 Bank 6: f380000000000e0b
mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff8104c95f> {acpi_processor_ffh_cstate_enter+0x8f/0xc0}
mce: [Hardware Error]: TSC 38f053f7d3a
mce: [Hardware Error]: PROCESSOR 0:50654 TIME 1515579774 SOCKET 1 APIC 24 microcode 200001e
mce: [Hardware Error]: Run the above through 'mcelog --ascii'

This looks like a machine problem to me. Do you have a kernel version that works fine? I'd run a memtest on the system to make sure it's not a broken memory chip.

Hi Mara

Thanks for your reply . if i remove below rule machine boot fine with same kernel 4.5 .
for my understanding this is something to related to PCI pass through of NVME devcie which is not working up to 4.5.7 kernel
but 4.6.0 its working ..

<hostdev mode='subsystem' type='pci' managed='yes'>
<source>
<address domain='0x0000' bus='0xd8' slot='0x00' function='0x0'/>
</source>
</hostdev>

Does 4.6 work for you? 4.5 isn't an LTS release so if this bug has been fixed after 4.5.7, it wasn't backported.

yes Mara 4.6 is working fine for me .. i am just wondering which patch went into 4.6 regarding NVME passthrough in 4.6 so that i can try backporting into 4.5 .

May be this one (but not 100% sure about this):
https://git.kernel.org/pub/scm/linux...47e3359a51a8b5

Hi Mara

Thanks for your reply . is there any method where i can get the all the patch for 4.6.0 kernel and one by one i can apply on 4.5.7 kernel and test .

Regards
Nitin

Hmm... You could do it using git format-patch between the two revisions. However, there will be many and application isn't obvious because of the merges.

For the patch in question you can download the single one from the web interface here https://git.kernel.org/pub/scm/linux...47e3359a51a8b5

Then you just to git am to apply it.