[openib-general] IB kernel on x86_64 architecture

Sun Apr 18 18:02:55 PDT 2004

    Andrey> The same here with kernel 2.6.5. Oops in the same place. A
    Andrey> quick look shows that in ib_mad_send_no_copy() in the line

    Andrey> struct ib_mad_private *priv = device->mad;

    Andrey> priv becomes a zero pointer which later dereferenced
    Andrey> during (inlined) call of ib_mad_queue_work()

I'm not sure why this would happen.  With my latest tree, I don't see
this happen (tested on RHEL 3 with 2.4.21-9.0.1 SMP on dual Opteron).
As soon as Eitan is done checking in his code and the Subversion
server is stablized, we should be able to release this.

    Andrey> test26 kernel: THH(1): page shift calculated :32 , due to
    Andrey> MPT restrictions page shift wil be 32

    Andrey> Rather error message or TAVOR_IF_MAX_MPT_PAGE_SIZE should
    Andrey> be corrected, because TAVOR_IF_MAX_MPT_PAGE_SIZE actual
    Andrey> value is 31 not 32...

31 is the correct value, so this is just a cosmetic problem.

 - Roland

-- 
To unsubscribe send an email with subject unsubscribe to openib-general at openib.org.
Please contact moderator at openib.org for questions.