[openib-general] Re: ib_sdp ERR: IOCB dmesg output

[Grant Grundler]

On Thu, Jan 12, 2006 at 09:52:45AM +0200, Michael S. Tsirkin wrote:
> > As noted earlier, netperf TCP_RR over SDP ran to completion
> > with no problems. netperf TCP_STREAM over SDP started spewing
> > the same errors despite the patches. :(
>
> OK, but can you unload the module now?

Sorry - I just realized I checked the "netserver" machine
and not the "netperf" ("client").  client side still fails. :(

iota:~# reload_ib
+ IPoIB=30
+ ifconfig ib0 down
+ ifconfig ib1 down
+ rmmod ib_umad ib_ipoib ib_uverbs ib_sdp ib_cm ib_sa ib_mthca ib_mad ib_core
ERROR: Module ib_sdp is in use
ERROR: Module ib_cm is in use by ib_sdp
ERROR: Module ib_sa is in use by ib_sdp
ACPI: PCI interrupt for device 0000:81:00.0 disabled
GSI 60 (level, low) -> CPU 1 (0x0100) vector 60 unregistered
ERROR: Module ib_mad is in use by ib_cm,ib_sa
ERROR: Module ib_core is in use by ib_sdp,ib_cm,ib_sa,ib_mad
+ modprobe ib_mthca msi_x=1
ib_mthca: Mellanox InfiniBand HCA driver v0.06 (June 23, 2005)
ib_mthca: Initializing 0000:81:00.0
GSI 60 (level, low) -> CPU 0 (0x0000) vector 60
ACPI: PCI Interrupt 0000:81:00.0[A] -> GSI 60 (level, low) -> IRQ 60
ib_mthca 0000:81:00.0: HCA FW version 3.3.2 is old (3.3.3 is current).
ib_mthca 0000:81:00.0: If you have problems, try updating your HCA FW.
+ modprobe ib_ipoib
+ modprobe ib_sdp
+ modprobe ib_uverbs
+ modprobe ib_umad
+ ifconfig ib0 10.0.0.30 netmask 255.255.255.0 broadcast 10.0.0.255
+ ifconfig ib1 10.0.1.30 netmask 255.255.255.0 broadcast 10.0.1.255
iota:~# lsmod
Module                  Size  Used by
ib_umad                33648  0
ib_uverbs              93096  0
ib_ipoib               96128  0
ib_mthca              274728  0
ib_sdp                230480  3
ib_cm                  93964  1 ib_sdp
ib_sa                  25324  2 ib_ipoib,ib_sdp
ib_mad                 85952  4 ib_umad,ib_mthca,ib_cm,ib_sa
ib_core                93096  8 ib_umad,ib_uverbs,ib_ipoib,ib_mthca,ib_sdp,ib_cm,ib_sa,ib_mad
tulip                 118064  0
e1000                 233420  0
tg3                   227280  0
e100                   83592  0
iota:~#

Looks like the error messages and sdp refcnt might be related.
(IIRC, 4 error msgs and SDP ref cnt is 3)
Since netperf is terminated by a timer signal, it's possible
traffic is still outstanding when it exits. Could that be a
cause of the "ERR: IOCB <-1> cancel" error messages?

thanks,
grant