[openib-general] ibv_devinfo
Pradeep Satyanarayana
pradeep at us.ibm.com
Thu Oct 5 11:11:48 PDT 2006
Roland,
Here is what I mean (partial output of "ibv_devinfo -v"):
port: 1
state: PORT_ACTIVE (4)
max_mtu: 2048 (4)
active_mtu: 2048 (4)
sm_lid: 4
port_lid: 128
port_lmc: 0x02
max_msg_sz: 0x0
port_cap_flags: 0x00000000
max_vl_num: 0
bad_pkey_cntr: 0x0
qkey_viol_cntr: 0x0
sm_sl: 0
pkey_tbl_len: 1
gid_tbl_len: 1
subnet_timeout: 8
init_type_reply: 0
active_width: 12X (8)
active_speed: 2.5 Gbps (1)
phys_state: invalid physical state (0)
GID[ 0]: fe80:0000:0000:0003:0002:5500:1002:1b3d
port: 2
state: PORT_ACTIVE (4)
max_mtu: 2048 (4)
active_mtu: 2048 (4)
sm_lid: 4
port_lid: 136
port_lmc: 0x02
max_msg_sz: 0x0
port_cap_flags: 0x00000000
max_vl_num: 0
bad_pkey_cntr: 0x0
qkey_viol_cntr: 0x0
sm_sl: 0
pkey_tbl_len: 1
gid_tbl_len: 1
subnet_timeout: 8
init_type_reply: 0
active_width: 12X (8)
active_speed: 2.5 Gbps (1)
phys_state: invalid physical state (0)
GID[ 0]: fe80:0000:0000:0002:0002:5500:1002:1b7d
The phys_state above does not correspond to what is in
/sys/class/infiniband/ehca0/ports/1
However, the "phys state" that ibstatus displays is indeed correct and as
expected (see below).
ibstatus ehca0
Infiniband device 'ehca0' port 1 status:
default gid: fe80:0000:0000:0000:0002:5500:0000:933d
base lid: 0xb
sm lid: 0x2
state: 4: ACTIVE
phys state: 0: <unknown>
rate: 30 Gb/sec (12X)
Infiniband device 'ehca0' port 2 status:
default gid: fe80:0000:0000:0000:0002:5500:0000:937d
base lid: 0x7
sm lid: 0x2
state: 4: ACTIVE
phys state: 0: <unknown>
rate: 30 Gb/sec (12X)
The fix in dev_info.c for this problem is elementary (replace "invalid
physical state" with "unknown"). Since the sources for this
program occurred in the examples directory I presumed that this was just a
sample program written as an example. That
was why I asked if it was "supported" or not.
Pradeep
pradeep at us.ibm.com
Roland Dreier <rdreier at cisco.com> wrote on 10/05/2006 10:53:41 AM:
> Pradeep> On eHCA we find that there are some discrepencies (for
> Pradeep> example the phys_state) between the outputs of
> Pradeep> "ibv_devinfo -v" and the corresponding output of
> Pradeep> ibstatus. Indeed ibstatus does display what is in the
> Pradeep> /sys/class/infiniband/* files, and as expected.
>
> Could you give some details on the differences? Is ibv_devinfo or
> ibstatus giving the correct output?
>
> Pradeep> I would like to understand the status of ibv_devinfo. Is
> Pradeep> this a supported program -given that the sources for this
> Pradeep> is in the example directory? Also, I find the source
> Pradeep> (devinfo.c) does appear in the OFED 1.0 tar ball.
>
> I'm not sure what "supported" would mean exactly, but certainly I
> would like ibv_devinfo to work as well as possible. Unfortunately I
> can't fix anything without a more detailed report...
>
> - R.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20061005/932c6e0d/attachment.html>
More information about the general
mailing list