[openib-general] ib_modify_qp() too many arguments

[Hal Rosenstock]

On Tue, 2004-11-02 at 19:10, Grant Grundler wrote:
> I am trying to build roland-merge #1119 on top of 2.6.10-rc1 for ia64.
> And yes, the usage noted below doesn't match the declaration:
> 
>   CC [M]  drivers/infiniband/core/cm_main.o
>   In file included from drivers/infiniband/core/cm_main.c:24:
>   drivers/infiniband/core/cm_priv.h: In function `ib_cm_qp_modify':
>   drivers/infiniband/core/cm_priv.h:183: error: too many arguments to
>   function `ib_modify_qp'
> 
> Should I not (yet) be enabling CONFIG_INFINIBAND_CM?
> Trivial patch appended to "fix". 

There was recently a verbs change which caused this and it looks like we
missed this as CM is not "used" yet.

> Though I don't know if it's "right"

Looks right to me.

> since it seems ib_cm_qp_modify() could just go away.

QP modify is needed by CM (as it walks a connection through it's states
it modifies the QP state) and UD QPs as well. It can't go away.

> Index: src/linux-kernel/infiniband/core/cm_priv.h
> ===================================================================
> --- src/linux-kernel/infiniband/core/cm_priv.h	(revision 1116)
> +++ src/linux-kernel/infiniband/core/cm_priv.h	(working copy)
> @@ -178,9 +178,7 @@
>  				  struct ib_qp_attr *attr,
>  				  int                attr_mask)
>  {
> -	struct ib_qp_cap qp_cap;
> -
> -	return qp ? ib_modify_qp(qp, attr, attr_mask, &qp_cap) : 0;
> +	return qp ? ib_modify_qp(qp, attr, attr_mask) : 0;
>  }
>  
>  int ib_cm_timeout_to_jiffies(int timeout);

-- Hal