It seems that something with interrupt remapping is the problem...
And I don´t know how to solve the problem.

But I tested the intel NVMe in passthrouged Windows 2012. There it
works perfectly. Second test - intel P4600 and Samsung simultaneously:
Samsung works, intel has the error.
Further tests without success:
- esxcli system settings kernel set --setting=iovDisableIR -v TRUE
- Remapping disabled in BIOS


My hardware configuration:
- SC826BAC4-R920LPB - 2U
- X10DRC-T4+ (BIOS-FW 3.58)
- 2x intel Xeon E5-2630 v4
- BPN-SAS3-826A-N4
- AOC-SLG3-4E4T
- 3x intel DC P4600
- AOC-SLG3-2M2 with 2x Samsung SM/PM961 added for tests
- ESXi 6.5U1

Sunday, December 10, 2017, 7:18:24 PM, you wrote:

FM> Anybody can help?
FM> Removed NVMe´s for now and found this:
FM>         root@vsan-01:~# fmadm faulty
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         TIME            EVENT-ID                              MSG-ID        
 SEVERITY
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         Dec 09 20:06:26 2c3507dc-6936-6a0a-a36d-a831a64939d4  PCIEX-8000-0A 
 Critical
FM>         Host        : vsan-01
FM>         Platform    : VMware-Virtual-Platform   Chassis_id  :
FM> VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23
FM>         Product_sn  :
FM>         Fault class : fault.io.pciex.device-interr
FM>         Affects     : dev:////pci@0,0/pci15ad,7a0@18/pci8086,4702@0
FM>                                           faulted and taken out of service
FM>         FRU         : "MB"
FM> 
(hc://:product-id=VMware-Virtual-Platform:server-id=vsan-01:chassis-id=VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23/motherboard=0)
FM>                                           faulty
FM>         Description : A problem was detected for a PCIEX device.
FM>                                   Refer to
FM> http://illumos.org/msg/PCIEX-8000-0A for more
FM>                                   information.
FM>         Response    : One or more device instances may be disabled
FM>         Impact      : Loss of services provided by the device instances 
associated with
FM>                                   this fault
FM>         Action      : Schedule a repair procedure to replace the affected 
device.  Use
FM>                                   fmadm faulty to identify the device or 
contact your illumos
FM>                                   distribution team for support.
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         TIME            EVENT-ID                              MSG-ID        
 SEVERITY
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         Dec 09 20:06:26 c5accbb9-26cd-4aba-dde0-ff15b11c6823  PCIEX-8000-0A 
 Critical
FM>         Host        : vsan-01
FM>         Platform    : VMware-Virtual-Platform   Chassis_id  :
FM> VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23
FM>         Product_sn  :
FM>         Fault class : fault.io.pciex.device-interr
FM>         Affects     : dev:////pci@0,0/pci15ad,7a0@17/pci15ad,7f0@0
FM>                                           faulted and taken out of service
FM>         FRU         : "MB"
FM> 
(hc://:product-id=VMware-Virtual-Platform:server-id=vsan-01:chassis-id=VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23/motherboard=0)
FM>                                           faulty
FM>         Description : A problem was detected for a PCIEX device.
FM>                                   Refer to
FM> http://illumos.org/msg/PCIEX-8000-0A for more
FM>                                   information.
FM>         Response    : One or more device instances may be disabled
FM>         Impact      : Loss of services provided by the device instances 
associated with
FM>                                   this fault
FM>         Action      : Schedule a repair procedure to replace the affected 
device.  Use
FM>                                   fmadm faulty to identify the device or 
contact your illumos
FM>                                   distribution team for support.
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         TIME            EVENT-ID                              MSG-ID        
 SEVERITY
FM>         --------------- ------------------------------------  
-------------- ---------
FM>         Dec 09 20:06:26 4ed4fad4-9c61-cbe2-8c71-9f7dbea60d14  PCIEX-8000-0A 
 Critical
FM>         Host        : vsan-01
FM>         Platform    : VMware-Virtual-Platform   Chassis_id  :
FM> VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23
FM>         Product_sn  :
FM>         Fault class : fault.io.pciex.device-interr
FM>         Affects     : dev:////pci@0,0/pci15ad,7a0@15,1/pci8086,4702@0
FM>                                           faulted and taken out of service
FM>         FRU         : "MB"
FM> 
(hc://:product-id=VMware-Virtual-Platform:server-id=vsan-01:chassis-id=VMware-56-4d-08-4b-a7-e9-d9-7a-1d-a1-5f-95-55-89-d7-23/motherboard=0)
FM>                                           faulty
FM>         Description : A problem was detected for a PCIEX device.
FM>                                   Refer to
FM> http://illumos.org/msg/PCIEX-8000-0A for more
FM>                                   information.
FM>         Response    : One or more device instances may be disabled
FM>         Impact      : Loss of services provided by the device instances 
associated with
FM>                                   this fault
FM>         Action      : Schedule a repair procedure to replace the affected 
device.  Use
FM>                                   fmadm faulty to identify the device or 
contact your illumos
FM>                                   distribution team for support.
FM> I cleared this with:
FM> root@vsan-01:~# fmadm repaired
FM> dev:////pci@0,0/pci15ad,7a0@17/pci15ad,7f0@0
FM> fmadm: recorded repair to of
FM> dev:////pci@0,0/pci15ad,7a0@17/pci15ad,7f0@0
FM> root@vsan-01:~# fmadm repaired
FM> dev:////pci@0,0/pci15ad,7a0@15,1/pci8086,4702@0
FM> fmadm: recorded repair to of
FM> dev:////pci@0,0/pci15ad,7a0@15,1/pci8086,4702@0
FM> root@vsan-01:~# fmadm repaired 
FM> dev:////pci@0,0/pci15ad,7a0@18/pci8086,4702@0
FM> fmadm: recorded repair to of
FM> dev:////pci@0,0/pci15ad,7a0@18/pci8086,4702@0



FM> Sunday, December 10, 2017, 4:29:08 AM, you wrote:
FM>> I just remembered that in another system works a Samsung SM961 with
FM>> NVMe 1.2 perfectly.
FM>> Anybody knows why the intel P4600 doesn´t work? Or better, anybody has
FM>> a solution that the intel works?



FM>> Sunday, December 10, 2017, 1:54:35 AM, you wrote:
FM>>> Hi,
FM>>> are there any news regarding NVMe 1.2 in ESXi-passthrough systems or
FM>>> is there any trick?
FM>>> I can´t see the intel P4600 as a blkdev-device (updated omnios-r151024
FM>>> on ESXi 6.5).
FM>>> prtconv -v produces this:
FM>>>             pci8086,4702, instance #0 (retired)
FM>>>                 Hardware properties:
FM>>>                     name='pci-msix-capid-pointer' type=int items=1
FM>>>                         value=00000050
FM>>>                     name='pci-msi-capid-pointer' type=int items=1
FM>>>                         value=000000a0
FM>>>                     name='pcie-serial' type=byte items=8
FM>>>                         value=fa.1a.23.4e.41.2e.cd.55
FM>>>                     name='acpi-namespace' type=string items=1
FM>>>                         value='\_SB_.PCI0.PE41.S1F0'
FM>>>                     name='assigned-addresses' type=int items=5
FM>>>                        
FM>>> value=83040010.00000000.fcffc000.00000000.00004000
FM>>>                     name='reg' type=int items=10
FM>>>                        
FM>>> 
value=00040000.00000000.00000000.00000000.00000000.03040010.00000000.00000000.00000000.00004000
FM>>>                     name='compatible' type=string items=13
FM>>>                         value='pciex8086,a54.8086.4702.0' +
FM>>> 'pciex8086,a54.8086.4702' + 'pciex8086,a54.0' + 'pciex8086,a54' +
FM>>> 'pciexclass,010802' + 'pciexclass,0108' +
FM>>> 'pci8086,a54.8086.4702.0' + 'pci8086,a54.8086.4702' +
FM>>> 'pci8086,4702' + 'pci8086,a54.0' + 'pci8086,a54' +
FM>>> 'pciclass,010802' + 'pciclass,0108'
FM>>>                     name='model' type=string items=1
FM>>>                         value='Unknown class of pci/pnpbios device'
FM>>>                     name='power-consumption' type=int items=2
FM>>>                         value=00000001.00000001
FM>>>                     name='devsel-speed' type=int items=1
FM>>>                         value=00000000
FM>>>                     name='interrupts' type=int items=1
FM>>>                         value=00000001
FM>>>                     name='subsystem-vendor-id' type=int items=1
FM>>>                         value=00008086
FM>>>                     name='subsystem-id' type=int items=1
FM>>>                         value=00004702
FM>>>                     name='unit-address' type=string items=1
FM>>>                         value='0'
FM>>>                     name='class-code' type=int items=1
FM>>>                         value=00010802
FM>>>                     name='revision-id' type=int items=1
FM>>>                         value=00000000
FM>>>                     name='vendor-id' type=int items=1
FM>>>                         value=00008086
FM>>>                     name='device-id' type=int items=1
FM>>>                         value=00000a54
FM>>>                     name='vendor-name' type=string items=1
FM>>>                         value='Intel Corporation'
FM>>>                     name='device-name' type=string items=1
FM>>>                         value='unknown device'
FM>>>                     name='subsystem-name' type=string items=1
FM>>>                         value='unknown subsystem'


FM>>> my last try was with /kernel/drv/nvme.conf:
FM>>>                 #
FM>>>                 # The driver was tested only against devices supporting 
v1.0 of the
FM>>>                 # NVMe specification. Uncomment this to be able to use 
devices conforming
FM>>>                 # to newer specifications.
FM>>>                 strict-version=0;
FM>>>                 #
FM>>>                 # The driver does currently not support any vendor 
specific extension to the
FM>>>                 # specification. By default it will fault the device if 
it receives a vendor-
FM>>>                 # specific command status. Uncomment this to disable this 
behaviour.
FM>>>                 ignore-unknown-vendor-status=1;
FM>>>                 #
FM>>>                 # The maximum length of the admin queue can be overridden 
here (16-4096).
FM>>>                 #admin-queue-len=256;
FM>>>                 #
FM>>>                 # The maximum length of the individual I/O queues can be 
overriden here
FM>>>                 # (16-65536).
FM>>>                 #io-queue-len=1024;
FM>>>                 #
FM>>>                 # The maximum number of outstanding asynchronous event 
requests can
FM>>>                 # overridden here.
FM>>>                 #async-event-limit=10;
FM>>>                 #
FM>>>                 # Enable (1) or Disable (0) the volatile write cache, if 
present.
FM>>>                 #
FM>>>                 #volatile-write-cache-enable=1;
FM>>>                 #
FM>>>                 # The minimum physical block size to be reported to 
blkdev. This value must
FM>>>                 # be a power of 2 greater than or equal to 512.
FM>>>                 #
FM>>>                 #min-phys-block-size=512;



FM>>> Greets
FM>>> Frank




FM> ------------------------------------------
FM> illumos-discuss
FM> Archives:
FM> 
https://illumos.topicbox.com/groups/discuss/discussions/Tbd539f25eed62d6a-M2ca046c4e5bf2678193f154f
FM> Powered by Topicbox: https://topicbox.com



-- 
Best regards,
 Frank                            mailto:[email protected]


------------------------------------------
illumos-discuss
Archives: 
https://illumos.topicbox.com/groups/discuss/discussions/Tbd539f25eed62d6a-Madb6a780b4e35b0ad6c21e44
Powered by Topicbox: https://topicbox.com

Reply via email to