[ofw] wrong assert - patch

Uri Habusha urih at mellanox.co.il
Thu Aug 18 06:11:59 PDT 2011


Thanks Hal,

I'm looking in ib_types and don't find any status related to IB_SM. Did you add it in the last period. Do you suggest to add only IB_SM_STATUS_SUCCESS for ib_types.h

For now I will make the checking with checking that the status is

Uri

-----Original Message-----
From: Hal Rosenstock [mailto:hal.rosenstock at gmail.com] 
Sent: Monday, August 15, 2011 8:58 PM
To: Uri Habusha
Cc: ofw at lists.openfabrics.org; Firas Mahameed; Alex Netes
Subject: Re: [ofw] wrong assert - patch

On Mon, Aug 15, 2011 at 1:22 AM, Uri Habusha <urih at mellanox.co.il> wrote:
> Description:
>
> If one starts opensm with –P flag. (opensm –P config.conf).
>
> He will get the following ASSERT in spl_qp_svc_update_cache function:
>
> CL_ASSERT(!p_mad->status) , while p_mad->status == 0x80
>
>
>
> Fix:
>
> Use ib_smp_get_status instead to refine the assert.
>
>
>
>
>
> Index: al_smi.c
>
> ===================================================================
>
> --- al_smi.c          (revision 3262)
>
> +++ al_smi.c       (working copy)
>
> @@ -1661,7 +1661,7 @@
>
>                CL_ASSERT( p_mad );
>
>                CL_ASSERT( p_mad->mgmt_class == IB_MCLASS_SUBN_DIR ||
>
>
>  p_mad->mgmt_class == IB_MCLASS_SUBN_LID);
>
> -              CL_ASSERT(!p_mad->status);
>
> +             CL_ASSERT(ib_smp_get_status(p_mad) ==
> IB_SA_MAD_STATUS_SUCCESS);

This may work but looks a little like apples and oranges to me. SMPs are SM not SA class. Rather than IB_SA_MAD_STATUS, should this use 0 or alternatively define IB_MAD_STATUS_OK as 0 in ib_types.h and then use that define (so adding Alex N) ?

-- Hal

>
>                 cl_spinlock_acquire(&p_spl_qp_svc->cache_lock);
>
>
>
>
>
> Uri Habusha
>
> Windows SW Development Lead
>
>
>
> Mellanox Technologies
> P.OBox 586, Yokneam 20692
>
> Israel
>
>
>
>
>
>
>
> _______________________________________________
> ofw mailing list
> ofw at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
>


More information about the ofw mailing list