From mboxrd@z Thu Jan  1 00:00:00 1970
From: Vincent JARDIN <vincent.jardin@6wind.com>
Subject: Re: [PATCH v4 02/16] net/mrvl: add mrvl net pmd driver
 skeleton
Date: Thu, 12 Oct 2017 09:59:49 +0200
Message-ID: <6bfe47de-c5b4-2b01-7991-2ac48913b2c3@6wind.com>
References: <1507031500-11473-1-git-send-email-tdu@semihalf.com>
 <1507561244-20115-1-git-send-email-tdu@semihalf.com>
 <1507561244-20115-3-git-send-email-tdu@semihalf.com> <9041127.34t6OW5FrT@xps>
 <20171012065104.GC19106@tdu>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Cc: dev@dpdk.org, mw@semihalf.com, dima@marvell.com, nsamsono@marvell.com,
 Jianbo.liu@linaro.org, Jacek Siuda <jck@semihalf.com>
To: Tomasz Duszynski <tdu@semihalf.com>, Thomas Monjalon <thomas@monjalon.net>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-wm0-f47.google.com (mail-wm0-f47.google.com [74.125.82.47])
 by dpdk.org (Postfix) with ESMTP id 725081D90
 for <dev@dpdk.org>; Thu, 12 Oct 2017 09:59:55 +0200 (CEST)
Received: by mail-wm0-f47.google.com with SMTP id u138so10887195wmu.4
 for <dev@dpdk.org>; Thu, 12 Oct 2017 00:59:55 -0700 (PDT)
In-Reply-To: <20171012065104.GC19106@tdu>
Content-Language: en-US
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

+1 with Thomas, see below,

Le 12/10/2017 à 08:51, Tomasz Duszynski a écrit :
>> What is MUSDK_DMA_MEMSIZE?
>> If the value cannot change, it must be a constant in the code.
>> If it can change, it should be a run-time driver option.
> It's up to the user what MUSDK_DMA_MEMSIZE is going to be. Currently it's
> set to value that should work it all cases.
> 
> Except that, MUSDK_DMA_MEMSIZE is used as synchronization point for net
> and crypto (on condition they are used together i.e ipsec-secgw).
> 
> Suppose we have two different MUSDK_DMA_MEMSIZE defined for net/crypto then
> dma memsize allocated will depend on driver probing sequence which might
> confuse user.
It does not make sense,
+	/*
+	 * ret == -EEXIST is correct, it means DMA
+	 * has been already initialized (by another PMD).
+	 */
+	ret = mv_sys_dma_mem_init(RTE_MRVL_MUSDK_DMA_MEMSIZE

int mv_sys_dma_mem_init(u64 size)
{
         struct sys_dma  *i_sys_dma;
         int err;

         if (sys_dma) {
                 pr_err("Dma object already exits.\n");
                 return -EEXIST;
         }

So, I do not understand why you cannot add some checks into the drivers 
to assert that users must have set the same value for both when calling:
   ret = mv_sys_dma_mem_init(my_best_size);
maybe, you need to fix and improve musdk first to avoid DPDK from 
getting such compilation issues.

best regards,
   Vincent