[ofa-general] ibv_devinfo -v conventions for RNIC in 1.4.1-X
Mike Aho
ahomike at us.ibm.com
Thu May 21 05:42:16 PDT 2009
Steve, Below is the verbose version I got as an example for a Chelsio
card on a Dell machine. I REALLY like that an RNIC can show its settings
via ibv_devinfo. But an RNIC is different from an IB card and perhaps
ibv_devinfo should have an indicator [by port?] that differentiates an
RNIC from an IB HCA and do the output differently. I doubt 1.4.1 can
change this short term but 1.5 could incorporate a meaningful change.
Developing a "readme" to cover it would recover separate maintenance and
be challenging to keep updated and RNIC meaningful output would be more
useful.
It appears that the node_guid and sys_image_guid incorporate the MAC
addresses into the format. Are these fields really useful as GUIDs? This
also assumes a single port RNIC and I could see a two-port RNIC coming
along or an IO card with an IB port and Ethernet (RNIC) port.
I could see MAC address(es) being shown in the port subsection based on a
port indicator enum of IB HCA, Ethernet, etc.
The max_raw_ipv6_qp and max_raw_eth_qp seem superfluous on a plain RNIC
but perhaps make sense on a combined RNIC and HCA adapter. So I think it
can stay.
Under the port subsection, the use of the enum values for mtu size are
good for an IB HCA but seem not applicable to an RNIC. Perhaps these
should move from an enum to a range of values for RNIC.
I can live with some of the other IB artifacts under the port subsection
such as pkey and the table lengths but these could be suppressed on an
RNIC port indicator in the future. The active width, speed, and physical
state need to change to address an RNIC port.
Mike Aho
hca_id: cxgb3_0
fw_ver: 7.0.0
node_guid: 0007:4305:6009:0000
sys_image_guid: 0007:4305:6009:0000
vendor_id: 0x1425
vendor_part_id: 48
hw_ver: 0x1
board_id: 1425.30
phys_port_cnt: 1
max_mr_size: 0x100000000
page_size_cap: 0xffff000
max_qp: 32736
max_qp_wr: 1023
device_cap_flags: 0x00228000
max_sge: 4
max_sge_rd: 1
max_cq: 32767
max_cqe: 8192
max_mr: 32768
max_pd: 32767
max_qp_rd_atom: 8
max_ee_rd_atom: 0
max_res_rd_atom: 0
max_qp_init_rd_atom: 8
max_ee_init_rd_atom: 0
atomic_cap: ATOMIC_NONE (0)
max_ee: 0
max_rdd: 0
max_mw: 0
max_raw_ipv6_qp: 0
max_raw_ethy_qp: 0
max_mcast_grp: 0
max_mcast_qp_attach: 0
max_total_mcast_qp_attach: 0
max_ah: 0
max_fmr: 0
max_srq: 0
max_pkeys: 0
local_ca_ack_delay: 0
port: 1
state: PORT_ACTIVE (4)
max_mtu: 4096 (5)
active_mtu: 2048 (4)
sm_lid: 0
port_lid: 0
port_lmc: 0x00
max_msg_sz: 0xffffffff
port_cap_flags: 0x009f0000
max_vl_num: invalid value (0)
bad_pkey_cntr: 0x0
qkey_viol_cntr: 0x0
sm_sl: 0
pkey_tbl_len: 1
gid_tbl_len: 1
subnet_timeout: 0
init_type_reply: 0
active_width: 4X (2)
active_speed: 5.0 Gbps (2)
phys_state: invalid physical state (0)
Mike Aho
From:
Steve Wise <swise at opengridcomputing.com>
To:
Mike Aho/Rochester/IBM at IBMUS
Cc:
general at lists.openfabrics.org
Date:
05/20/2009 08:47 PM
Subject:
Re: [ofa-general] ibv_devinfo -v conventions for RNIC in 1.4.1-X
Mike Aho wrote:
> I ran ibv_devinfo -v for 1.4.1-rc4 and got all the information dumped
> for an RNIC (Ethernet RDMA card). Is there doc somewhere to explain
> how the fields are addressed differently than an IB HCA? Is there a
> set of conventions?
>
> Mike Aho
Hey Mike,
Unfortunately, no doc. What fields are not clear?
Steve.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20090521/b5daa376/attachment.html>
More information about the general
mailing list