Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Ofir Bitton <obitton@habana.ai>
To:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Daniel Vetter <daniel.vetter@ffwll.ch>
CC:     Jason Gunthorpe <jgg@ziepe.ca>,
        Oded Gabbay <oded.gabbay@gmail.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        SW_Drivers <SW_Drivers@habana.ai>
Subject: RE: [PATCH 1/3] habanalabs: implement dma-fence mechanism
Thread-Topic: [PATCH 1/3] habanalabs: implement dma-fence mechanism
Thread-Index: AQHWWSShqFCRK4JZYk2Bp6Q6RakrCqkFqnEAgAArrQCAAAhQgIAAAWOAgADAMgCAAFu6MA==
Date:   Tue, 14 Jul 2020 12:07:29 +0000
Message-ID: <VI1PR0202MB32774769723B6CE154865086BC610@VI1PR0202MB3277.eurprd02.prod.outlook.com>
References: <20200713155424.24721-1-oded.gabbay@gmail.com>
 <20200713155752.GC267581@kroah.com>
 <CAKMK7uH=Ch4ce-9D5e-RvVwq_oK6Doqtq5QbvpmQ8uPWkMCi2w@mail.gmail.com>
 <20200713190357.GC25301@ziepe.ca>
 <CAKMK7uEvehX2CV3Q5FJrF49-_Xe9gXJ11wDo7xyVsipyuZm23Q@mail.gmail.com>
 <20200714063648.GC662760@kroah.com>
In-Reply-To: <20200714063648.GC662760@kroah.com>
Accept-Language: he-IL, en-US
Content-Language: en-US
dlp-product: dlpe-windows
dlp-version: 11.2.0.6
dlp-reaction: no-action
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-MS-Exchange-CrossTenant-Network-Message-Id: 35a2d9a4-ce65-4ada-c2bc-08d827ee7b66
X-MS-Exchange-CrossTenant-originalarrivaltime: 14 Jul 2020 12:07:29.6888
 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 0d4d4539-213c-4ed8-a251-dc9766ba127a
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-CrossTenant-userprincipalname: UWwxtJ+WjpxGEx1b03apWCLDy9fTG4XTTI9HKXGXDPiWmFFEhic2tYUo9HXs2yX4RmR1nqaRCLGB0/89zYvU4g==
X-MS-Exchange-Transport-CrossTenantHeadersStamped: VI1PR02MB4926
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

Sure,
I will send a new patch using completion instead of dma-fence

Thanks,
Ofir

-----Original Message-----
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>=20
Sent: Tuesday, July 14, 2020 09:37
To: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jason Gunthorpe <jgg@ziepe.ca>; Oded Gabbay <oded.gabbay@gmail.com>; Li=
nux Kernel Mailing List <linux-kernel@vger.kernel.org>; SW_Drivers <SW_Driv=
ers@habana.ai>; Ofir Bitton <obitton@habana.ai>
Subject: Re: [PATCH 1/3] habanalabs: implement dma-fence mechanism

On Mon, Jul 13, 2020 at 09:08:55PM +0200, Daniel Vetter wrote:
> On Mon, Jul 13, 2020 at 9:03 PM Jason Gunthorpe <jgg@ziepe.ca> wrote:
> >
> > On Mon, Jul 13, 2020 at 08:34:12PM +0200, Daniel Vetter wrote:
> > > On Mon, Jul 13, 2020 at 5:57 PM Greg Kroah-Hartman=20
> > > <gregkh@linuxfoundation.org> wrote:
> > > >
> > > > On Mon, Jul 13, 2020 at 06:54:22PM +0300, Oded Gabbay wrote:
> > > > > From: Ofir Bitton <obitton@habana.ai>
> > > > >
> > > > > Instead of using standard dma-fence mechanism designed for=20
> > > > > GPU's, we introduce our own implementation based on the former=20
> > > > > one. This implementation is much more sparse than the=20
> > > > > original, contains only mandatory functionality required by the d=
river.
> > > >
> > > > Sad you can't use the in-kernel code for this, I really don't=20
> > > > understand what's wrong with using it as-is.
> > > >
> > > > Daniel, why do we need/want duplicate code floating around in=20
> > > > the tree like this?
> > >
> > > The rules around dma-fence are ridiculously strict, and it only=20
> > > makes sense to inflict that upon you if you actually want to=20
> > > participate in the cross driver uapi built up around dma-buf and dma-=
fence.
> > >
> > > I've recently started some lockdep annotations to better enforce=20
> > > these rules (and document them), and it's finding tons of subtle=20
> > > bugs even in drivers/gpu (and I only just started with annotating dri=
vers:
> > >
> > > https://lore.kernel.org/dri-devel/20200707201229.472834-1-daniel.v
> > > etter@ffwll.ch/
> > >
> > > You really don't want to deal with this if you don't have to. If=20
> > > drivers/gpu folks (who created this) aren't good enough to=20
> > > understand it, maybe it's not a good idea to sprinkle this all=20
> > > over the tree. And fundamentally all this is is a slightly fancier=20
> > > struct completion. Use that one instead, or a wait_queue.
> > >
> > > I discussed this a bit with Oded, and he thinks it's easier to=20
> > > copypaste and simplify, but given that all other drivers seem to=20
> > > get by perfectly well with completion or wait_queue, I'm not sure=20
> > > that's a solid case.
> > >
> > > Also adding Jason Gunthorpe, who very much suggested this should=20
> > > be limited to dma-buf/gpu related usage only.
> >
> > Without all the cross-driver stuff dma_fence is just a completion.=20
> > Using dma_fence to get a completion is big abuse of what it is=20
> > intended for.
> >
> > I think the only problem with this patch is that it keeps too much=20
> > of the dma_fence stuff around. From what I could tell it really just=20
> > wants to add a kref and completion to struct hl_cs_compl and delete=20
> > everything to do with dma_fence.
>=20
> Yeah, that's what I recommended doing too. error flag might be needed=20
> too I think, but that's it.

Ok, so this should be made much simpler and not use this copy/paste code at=
 all.  I can accept that :)

Ofir, care to redo this?

thanks,

greg k-h