[ewg] iWARP uDAPL RDMA_READ(RR) test hangs on Chelsio T320
John Jolly
jjolly at suse.com
Tue Jun 11 07:01:26 PDT 2013
All,
We're trying to track down a problem where dapltest hangs using more
than one segment on the RR (RDMA read) performance test.
Tried uDAPL RDMA_READ test with "number of segments" parameter set to 1, test
worked fine.
RDMA_WRITE and SEND/RECV tests were executed without any issue.
---Steps to Reproduce---
1. Configure iWARP on server and client machines connected via Chelsio
adapter.
Server configuration:
x3755b:/ # ifconfig eth0
eth0 Link encap:Ethernet HWaddr 00:07:43:07:1B:B8
inet addr:10.0.0.1 Bcast:10.255.255.255 Mask:255.0.0.0
inet6 addr: fe80::207:43ff:fe07:1bb8/64 Scope:Link
UP BROADCAST MULTICAST MTU:1500 Metric:1
RX packets:10011 errors:0 dropped:0 overruns:0 frame:0
TX packets:9969 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:12629636 (12.0 Mb) TX bytes:2510976 (2.3 Mb)
Interrupt:11 Memory:e7880000-e7880fff
x3755b:/ # cat /etc/dat.conf
ofa-v2-iwarp u2.0 nonthreadsafe default libdaplofa.so.2 dapl.2.0 "eth0 0" ""
Client configuration:
3650m2:~ # ifconfig eth4
eth4 Link encap:Ethernet HWaddr 00:07:43:07:1B:62
inet addr:10.0.0.2 Bcast:10.255.255.255 Mask:255.0.0.0
inet6 addr: fe80::207:43ff:fe07:1b62/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:8730 errors:0 dropped:0 overruns:0 frame:0
TX packets:8780 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:2407318 (2.2 Mb) TX bytes:10847636 (10.3 Mb)
Interrupt:26 Memory:9c800000-9c800fff
3650m2:~ # cat /etc/dat.conf
ofa-v2-iwarp u2.0 nonthreadsafe default libdaplofa.so.2 dapl.2.0 "eth4 0" ""
2. Run the below command on the server,
x3755b:~ # dapltest -T S -D ofa-v2-iwarp
Dapltest: Service Point Ready - ofa-v2-iwarp
Dapltest: Service Point Ready - ofa-v2-iwarp
3. On client, run RR dapltest as below: Note the 2 parameter for number of segments.
3650m2:~ # dapltest -T P -s 10.0.0.1 -D ofa-v2-iwarp -d -i 100 RR 4096 2
-------------------------------------
PerfCmd.server_name : 10.0.0.1
PerfCmd.dapl_name : ofa-v2-iwarp
PerfCmd.mode : BLOCKING
PerfCmd.num_iterations : 100
PerfCmd.pipeline_len : -1
PerfCmd.op.transfer_type : RDMA_READ
PerfCmd.op.num_segs : 2
PerfCmd.op.seg_size : 4096
Server Name: 10.0.0.1
Server Net Address: 10.0.0.1
DT_cs_Client: Starting Test ...
DT_cs_Client: IA ofa-v2-iwarp opened
DT_cs_Client: EP created
***** DAPL Characteristics *****
Provider: ofa-v2-iwarp Version 2.0 DAPL 2.0
Adapter: by Version 2.0
Supporting:
32736 EPs with 1023 DTOs and 8 RDMA/RDs each
32767 EVDs of up to 65536 entries (default S/R size is 16/16)
IOVs of up to 4 elements
32768 LMRs (and 0 RMRs) of up to 0xffffffff bytes
Maximum MTU 0xffffffff bytes, RDMA 0xffffffff bytes
Maximum Private data size 256 bytes
Local IP address 10.0.0.2
***** ***** ***** ***** ***** *****
DT_cs_Client: Posting 1 recv buffer
DT_cs_Client: Connect Endpoint
DT_cs_Client: Await connection ...
DT_cs_Client: Connected!
DAT_STATE: DAT_EP_STATE_CONNECTED
DAT_STATE: Inbound DTO Status: Idle
DAT_STATE: Outbound DTO Status: Active
DT_cs_Client: Sending Client_Info
DT_cs_Client: Sent Client_Info - awaiting completion
DT_cs_Client: Sending Command
DT_cs_Client: Sent Command - awaiting completion
DT_cs_Client: Waiting for Server_Info
DT_cs_Client: Server_Info Received
DT_cs_Client: Version OK!
-------------------------------------
Server_Info.dapltest_version : 6
Server_Info.is_little_endian : 1
-------------------------------------
Client_Info.dapltest_version : 6
Client_Info.is_little_endian : 1
Client_Info.test_type : 2
-------------------------------------
PerfCmd.server_name : 10.0.0.1
PerfCmd.dapl_name : ofa-v2-iwarp
PerfCmd.mode : BLOCKING
PerfCmd.num_iterations : 100
PerfCmd.pipeline_len : -1
PerfCmd.op.transfer_type : RDMA_READ
PerfCmd.op.num_segs : 2
PerfCmd.op.seg_size : 4096
Client: Starting performance test
Client[b0e0]: Connect on port 0xb0e0
Client[b0e0]: Got Connection
Test[b0e0]: Waiting for Sync Msg
Test[b0e0]: Sync Msg Received
Test[b0e0]: Begin...
<...test hangs...>
Let me know if I can provide more information.
Regards,
John, SUSE
More information about the ewg
mailing list