Re: [PATCH] Add const qualifier in rdma_reg*, rdma_post_send*|write and ibv_reg_mr (was: Re: non-const pointer void * addr in rdma_reg_* and rdma_post_[send|write])

[Yann Droneaud <yann-WYdEh2jsUj5Wj0EZb7rXcA@public.gmane.org>]

Hi,

Le jeudi 28 novembre 2013 à 20:47 +0100, Hannes Weisbach a écrit :

> > But adding "const" should not break compilation of existing userspace program.

> That was my thought too.  However, because those functions are defined
> in as inline in a header, they actually do break user code.  For
> example, rdma_reg_msgs passes a const void * to ibv_reg_mr, which
> takes a void *, thus resulting in a warning.  If you (like me) compile
> with the -Werror flag, this results in broken code.  The only
> offending libibverbs function I found so far is ibv_reg_mr().

> > 
> > PS: this issue could also be raised against libibverbs, so once you've
> > fixed librdmacm, why not fix libibverbs after ?
> This was my plan.  But first I wanted to see if there even is interest
> in fixing this.  As a user I deal with librdmacm directly, so I
> started to fix it first.
> 
> Anyway, since patching librdmacm alone would break user code, I also
> included a patch for libibverbs, which adds the const qualifier to
> 'addr' parameter ib ibv_reg_mr().
> 

To avoid adding cast to non-const in librdmacm that no one will remove
once added, you're correct in patching libibverbs first.

> lists.openfabrics.org says, this is also the mailing list for
> libibverbs (README is outdated ;)), I hope this is ok.
> 

It's OK. Roland Dreier, libibverbs maintainer, is here (but sadly not as
often as we would like).

Regards.

-- 
Yann Droneaud
OPTEYA


--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html