From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sebastian Ott Subject: Re: [PATCH 06/10] s390/cio: add basic protected virtualization support Date: Thu, 9 May 2019 10:52:45 +0200 (CEST) Message-ID: References: <20190426183245.37939-1-pasic@linux.ibm.com> <20190426183245.37939-7-pasic@linux.ibm.com> <20190508230809.6b0faaf7.pasic@linux.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20190508230809.6b0faaf7.pasic@linux.ibm.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: virtualization-bounces@lists.linux-foundation.org Errors-To: virtualization-bounces@lists.linux-foundation.org To: Halil Pasic Cc: Farhan Ali , linux-s390@vger.kernel.org, Thomas Huth , Claudio Imbrenda , Vasily Gorbik , kvm@vger.kernel.org, "Michael S. Tsirkin" , Cornelia Huck , Eric Farman , virtualization@lists.linux-foundation.org, Christoph Hellwig , Martin Schwidefsky , Viktor Mihajlovski , Janosch Frank List-Id: virtualization@lists.linuxfoundation.org On Wed, 8 May 2019, Halil Pasic wrote: > On Wed, 8 May 2019 15:46:42 +0200 (CEST) > Sebastian Ott wrote: > > On Fri, 26 Apr 2019, Halil Pasic wrote: > > > static struct ccw_device * io_subchannel_allocate_dev(struct subchannel *sch) > > > { > > [..] > > > + cdev->private = kzalloc(sizeof(struct ccw_device_private), > > > + GFP_KERNEL | GFP_DMA); > > > > Do we still need GFP_DMA here (since we now have cdev->private->dma_area)? > > > > We probably do not. I kept it GFP_DMA to keep changes to the > minimum. Should changing this in your opinion be a part of this patch? This can be changed on top. > > > @@ -1062,6 +1082,14 @@ static int io_subchannel_probe(struct subchannel *sch) > > > if (!io_priv) > > > goto out_schedule; > > > > > > + io_priv->dma_area = dma_alloc_coherent(&sch->dev, > > > + sizeof(*io_priv->dma_area), > > > + &io_priv->dma_area_dma, GFP_KERNEL); > > > > This needs GFP_DMA. > > Christoph already answered this one. Thanks Christoph! Yes, I'm still struggling to grasp the whole channel IO + DMA API + protected virtualization thing.. > > > You use a genpool for ccw_private->dma and not for iopriv->dma - looks > > kinda inconsistent. > > > > Please have a look at patch #9. A virtio-ccw device uses the genpool of > it's ccw device (the pool from which ccw_private->dma is allicated) for > the ccw stuff it needs to do. AFAICT for a subchannel device and its API > all the DMA memory we need is iopriv->dma. So my thought was > constructing a genpool for that would be an overkill. > > Are you comfortable with this answer, or should we change something? Nope, I'm good with that one - ccw_private->dma has multiple users that all fit into one page.