[ofw] Re: Completion with bad status: IBV_WC_EXC_RETRY_EXC_ERROR

[Smith, Stan]

Diego Guella wrote:
> Hi,
> 
> Has anyone read this message? I got a failure delivery, maybe my
> message has been dropped.
> I can't still get 2 QP connected between a Linux and a Windows machine

You never will until the Windows IB CM (Communications/Connection
Manager) is compatible with the Linux IB/RDMA CM...
Currently the only working approach is to exchange QP information via a
TCP/IP socket connection and then transition the QPs manually; rather
advanced IB programming.

Stick with IPoIB/SDP, far easier thus allowing you to concentrate on the
_real_ problem you are trying to solve.

Stan.
> 
> 
> 
> 
> ----- Original Message -----
>> From: Dotan Barak
>> 
>> I don't have an available Linux <--> windows setup, can you please
>> send me the following data: * LID values (from ibv_devinfo/vstat)
> LIDs are 1 for Windows side, and 2 for Linux side. (in host order, as
> seen by Linux)
>> * the values that you set to the AV.LID (the value that you set,
>> after fixing/handling endianess)
> 
> Using the printfs in the code I can see this:
> 
> ### Linux ###
> commandline:
> ./verbs_test --ip=193.168.200.5
> 
> ----------
> QP with number 0x406 was created
> MR was registered, vaddr=0x0000000000507570, rkey = 0x00003900
> <<now reads the LID and htons()' it>>
> local LID: 512
> sizeof cm data: 18
> <<now exchanges data with other side, and htons()' on the received
> LID>> Remote data info: lid=1, QP num=0x05040000,
> vaddr=0x00000000003d7290, rkey=0x00230000
> Posting receive request
> ----------
> And this side blocks here indefinitely, since the other completes with
> error.
> 
> ### ###
> 
> 
> ### Windows ###
> commandline:
> verbs_test.exe -d
> 
> ----------
> QP with number 0x5040000 was created
> MR was registered, vaddr=0x00000000003d7290, rkey = 0x00230000
> Local LID: 256
> sizeof cm data: 18
> Remote data info: lid=512, QP num=0x00000406,
> vaddr=0x0000000000507570, rkey=0x00003900
> Sending the string 'Hello world' to the other side
> Posting send request
> Polling for Completion
> Error, got completion with bad status: IB_WCS_TIMEOUT_RETRY_ERR
> Starting destruction of resources
> ----------
> 
> ### ###
> 
>> * the QP number in each size
> See above. Note that the QPN changes at every test, this results are
> for one particular test
> 
>> * the remote QP number that you set in each side
> Same as above. I do not change endianness of QPN, vaddr, rkey.
> 
> 
> Thanks,
> Diego
> 
> _______________________________________________
> ofw mailing list
> ofw at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw