[CentOS] mlx4_0 Initializing and... (infiniband)

... Initializing..

hi all,

I've a a very basic setup, directly two boxes via two 
MHEH28-XTC and I cannot activate them.
One peculiar thing is I get (randomly & !often):

[85947.090496] AMD-Vi: Event logged [
[85947.090539] IO_PAGE_FAULT device=09:00.7 domain=0x0000 
address=0x00000000f6ffb000 flags=0x0050]
[85947.298509] AMD-Vi: Event logged [
[85947.298550] IO_PAGE_FAULT device=09:00.7 domain=0x0000 
address=0x00000000f6ffb000 flags=0x0050]

which is the card itself, judging by the device id
Would you have and share some thoughts please?

$ ./flint/mstflint -d 09:00.0 q # for both cards

-W- Running quick query - Skipping full image integrity checks.

Image type:      FS2
FW Version:      2.9.1000
Device ID:       25408
Description:     Node             Port1            Port2 Sys 
image
GUIDs:           0008f104039a62a0 0008f104039a62a1 
0008f104039a62a2 0008f104039a62a3
MACs:                                 000000000000     
000000000001
VSD:
PSID:            MT_04A0110001

$ ibstat
CA 'mlx4_0'
     CA type: MT25408
     Number of ports: 2
     Firmware version: 2.9.1000
     Hardware version: a0
     Node GUID: 0x0008f104039a08dc
     System image GUID: 0x0008f104039a08df
     Port 1:
         State: Initializing
         Physical state: LinkUp
         Rate: 10
         Base lid: 1
         LMC: 0
         SM lid: 1
         Capability mask: 0x0259086a
         Port GUID: 0x0008f104039a08dd
         Link layer: InfiniBand
     Port 2:
         State: Down
         Physical state: Polling
         Rate: 10
         Base lid: 0
         LMC: 0
         SM lid: 0
         Capability mask: 0x0259086a
         Port GUID: 0x0008f104039a08de
         Link layer: InfiniBand

in opensm log:

Jan 06 17:00:28 817185 [F6D5A700] 0x01 -> 
sm_mad_ctrl_send_err_cb: ERR 3113: MAD completed in error 
(IB_TIMEOUT): SubnGet(NodeInfo), attr_mod 0x0, TID 0x1cd1
Jan 06 17:00:28 817200 [F6D5A700] 0x01 -> 
sm_mad_ctrl_send_err_cb: ERR 3120 Timeout while getting 
attribute 0x11 (NodeInfo); Possible mis-set mkey?