[ofw] [PATCH] ib_types.h - be 'more' compatible with Linux OpenSM 3.3.9

[Hal Rosenstock]

On Tue, Apr 26, 2011 at 11:15 AM, Alex Naslednikov <xalex at mellanox.co.il> wrote:
> Hi Stan,
> I already asked the community about this ASSERT.
> Why can't we change the function to be like this:
>        if( attr_size & 0x07 )
>                return( cl_hton16( (uint16_t)(attr_size >> 3) + 1 ) );
>        else
>                return( cl_hton16( (uint16_t)(attr_size >> 3) ) );
> instead of removing the assert ?

That shouldn't be needed. InformInfo is the only one which is not a
multiple of 8 bytes but it is never used in an RMPP operation so this
should never be called for this attribute.

Anyhow, this should mirror what's done in the main/Linux OpenSM rather
than be different for Windows unless this is indeed something
different about Windows.

-- Hal

>
> -----Original Message-----
> From: ofw-bounces at lists.openfabrics.org [mailto:ofw-bounces at lists.openfabrics.org] On Behalf Of Smith, Stan
> Sent: Tuesday, April 26, 2011 3:08 AM
> To: ofw at lists.openfabrics.org
> Subject: [ofw] [PATCH] ib_types.h - be 'more' compatible with Linux OpenSM 3.3.9
>
> Patch to update ib_types.h to be 'more' compatible with Linux opensm/ib_types.h such that Windows Opensm 3.3.9 (checked) will run.
>
> Whitespace and spelling updates.
>
> Removal of CL_ASSERT() is required by the 3.3.9 opensm code base (Linux or Windows).
> In the OpenSM 3.3.6 timeframe the CL_ASSERT() was introduced into the Linux OpenSM ib_types.h.
> Unfortunately the CL_ASSERT() produced an assert() halt when running a Linux OpenSM debug version.
> The non-debug Linux version was deemed to be working correctly without the CL_ASSERT().
> The Linux OpenSM maintainer at that time decided to remove the CL_ASSERT() claiming it was unnecessary.
> Curious in that the windows OpenSM at the 3.3.6 code base time did not fire the assert().
>
> Now that the Windows code base has been updated to OpenSM 3.3.9, the CL_ASSERT() now fires.
> Remove the CL_ASSERT() to be compatible with Linux OpenSM 3.3.9 ib_types.h.
>
> signed-off-by: stan smith <stan.smith at intel.com>
>
> Index: ib_types.h
> ===================================================================
> --- ib_types.h  (revision 3111)
> +++ ib_types.h  (working copy)
> @@ -2013,6 +2013,7 @@
>  *
>  * DESCRIPTION
>  *      Returns a string for the specified node type.
> +*      14.2.5.3 NodeInfo
>  *
>  * SYNOPSIS
>  */
> @@ -2246,7 +2247,7 @@
>  *
>  * SYNOPSIS
>  */
> -static inline boolean_t
> +static inline boolean_t
>  ib_pkey_is_invalid(
>        IN      const   ib_net16_t              pkey )
>  {
> @@ -4549,7 +4550,6 @@
>  ib_get_attr_offset(
>        IN              const   uint32_t                                        attr_size )
>  {
> -       CL_ASSERT((attr_size & 0x07) == 0);
>        return( cl_hton16( (uint16_t)(attr_size >> 3) ) );  }
>
> @@ -5869,7 +5869,7 @@
>  *      ib_port_info_set_vl_stall_count
>  *
>  * DESCRIPTION
> -*      Sets the VL Stall Count which define the number of contiguous
> +*      Sets the VL Stall Count which define the number of contiguous
>  *  HLL (hoq) drops that will put the VL into stalled mode.
>  *
>  * SYNOPSIS
> @@ -10268,7 +10268,7 @@
>  *              Indicates the requested maximum number of work requests that may be
>  *              outstanding on the queue pair's receive queue.  This value must be less
>  *              than or equal to the maximum reported by the channel adapter associated
> -*              with the queue pair.
> +*              with the queue pair; Must be zero if h_srq is non-NULL.
>  *
>  *      sq_sge
>  *              Indicates the maximum number scatter-gather elements that may be
>
> _______________________________________________
> ofw mailing list
> ofw at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ofw
>