Ubuntu16.04.5 failed to create namespace

Bug #1812765 reported by shangsong
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
intel
Incomplete
Undecided
Unassigned

Bug Description

1.Fresh install Ubuntu16.04.5 on the lenovo ThinkSystem SR630MLK;
2.Install ipmctl and libsafec packages from below link:
https://launchpad.net/~jhli/+archive/ubuntu/ipmctl
https://launchpad.net/~jhli/+archive/ubuntu/libsafec
2.Create regions and namespaces via 128G QS AEP,
the regions can be created normally ,
but namespaces creation is failed:
root@ubuntu:~# ipmctl show -region
SocketID ISetID PersistentMemoryType Capacity FreeCapacity HealthState
       0 0xa354da902f248a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
       0 0x5190da905a2a8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
       1 0x3fa4da90a52c8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
       1 0xd590da90b42a8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
root@ubuntu:~#
root@ubuntu:~# ndctl create-namespace --type=pmem --mode=sector --size=50G --name=pmem1
failed to create namespace: Resource temporarily unavailable
root@ubuntu:~#
root@ubuntu:~# dmesg | grep -i region
[ 3.184547] PM: Registering ACPI NVS region [mem 0xa5ebf000-0xa6b05fff] (12873728 bytes)
[ 16.168259] nd_region region0: online cpus (112) < concurrent i/o lanes (256) < possible cpus (512)
[ 16.168262] nd_region region0: setting nr_cpus=256 may yield better libnvdimm device performance
[ 16.168511] nd_region region2: invalid BLK region
[ 16.168597] nd_region region4: invalid BLK region
[ 16.168609] nd_region region3: invalid BLK region
[ 16.168685] nd_region region5: invalid BLK region
root@ubuntu:~#

Revision history for this message
Jeff Lane  (bladernr) wrote :

Please confirm the following:

1: The failure occurs on 16.05.5 running the 4.15 kernel
2: The failure DOES NOT occur on 18.04 running the 4.15 kernel

That was my understanding from the email thread that lead to this bug being filed. Is this true?

Changed in intel:
status: New → Incomplete
Revision history for this message
shangsong (shangsong) wrote :

Hi Jeff,
  1.I am still not test on the 16.05.5 with kernel 4.15;
  2.18.04.1 with kernel 4.15 can not reproduce the failure.

shangsong1

Revision history for this message
Jeff Lane  (bladernr) wrote :

Another interesting question, why does each socket show two different IDs?

SocketID ISetID PersistentMemoryType Capacity FreeCapacity HealthState
0 0xa354da902f248a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
0 0x5190da905a2a8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
1 0x3fa4da90a52c8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy
1 0xd590da90b42a8a22 AppDirectNotInterleaved 126.0 GiB 126.0 GiB Healthy

From your description, you're using 128Gb NVDIMMs so why does socket 0 show two different IDs?

I'm only asking because these are entirely new to me. Is Socket 0 actually a pair of DIMM slots so Socket 0 = 2x 128GB NVDIMMs??

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.