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:   Sanket Parmar <sparmar@cadence.com>
To:     Peter Chen <peter.chen@kernel.org>
CC:     Pawel Laszczak <pawell@cadence.com>,
        "a-govindraju@ti.com" <a-govindraju@ti.com>,
        "linux-usb@vger.kernel.org" <linux-usb@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Rahul Kumar <kurahul@cadence.com>,
        "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
        "kishon@ti.com" <kishon@ti.com>
Subject: RE: [PATCH 2/2] usb: cdns3: Optimize DMA request buffer allocation
Thread-Topic: [PATCH 2/2] usb: cdns3: Optimize DMA request buffer allocation
Thread-Index: AQHXFKPlwE5egmmpm02b66W+ylYqvKqC94IAgAJBL9A=
Date:   Mon, 15 Mar 2021 15:51:04 +0000
Message-ID: <BY5PR07MB81194FB5A32CE9D6B793FF30B06C9@BY5PR07MB8119.namprd07.prod.outlook.com>
References: <1615267180-9289-1-git-send-email-sparmar@cadence.com>
 <1615267180-9289-2-git-send-email-sparmar@cadence.com>
 <20210314051048.GA30122@b29397-desktop>
In-Reply-To: <20210314051048.GA30122@b29397-desktop>
Accept-Language: en-US
Content-Language: en-US
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-MS-Exchange-CrossTenant-Network-Message-Id: 32548bfa-e3bf-44c2-b364-08d8e7ca23e2
X-MS-Exchange-CrossTenant-originalarrivaltime: 15 Mar 2021 15:51:04.3439
 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: d36035c5-6ce6-4662-a3dc-e762e61ae4c9
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-CrossTenant-userprincipalname: 7Ld2ZufdJsT5aAypJWfc86RpGBMsSVEPPIiemYDuS/IRKsakqJiUMLfXb/3cfgdVl7SlJsMI9Wd0R4ePyFTaMlGqCPIs4YK4jvyz9Q2Vp+M=
X-MS-Exchange-Transport-CrossTenantHeadersStamped: SJ0PR07MB8568
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.369,18.0.761
 definitions=2021-03-15_08:2021-03-15,2021-03-15 signatures=0
X-Proofpoint-Spam-Details: rule=outbound_check_notspam policy=outbound_check score=0 phishscore=0
 mlxlogscore=729 priorityscore=1501 bulkscore=0 impostorscore=0 spamscore=0
 adultscore=0 malwarescore=0 suspectscore=0 mlxscore=0 lowpriorityscore=0
 clxscore=1015 classifier=spam adjust=0 reason=mlx scancount=1
 engine=8.12.0-2009150000 definitions=main-2103150111
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

>=20
> On 21-03-09 06:19:40, Sanket Parmar wrote:
> > dma_alloc_coherent() might fail on the platform with a small DMA region=
.
> >
> > To avoid such failure in cdns3_prepare_aligned_request_buf(),
> > dma_alloc_coherent() is replaced with kmalloc and dma_map API to
> > allocate aligned request buffer of dynamic length.
> >
> > Fixes: 7733f6c32e36 ("usb: cdns3: Add Cadence USB3 DRD Driver")
>=20
> The comment with the 1st patch, it is not a bug-fix.

I will remove this.=20

>=20
> > Reported-by: Aswath Govindraju <a-govindraju@ti.com>
> > Signed-off-by: Sanket Parmar <sparmar@cadence.com>
> > ---
> >  drivers/usb/cdns3/cdns3-gadget.c |   73 +++++++++++++++++++++++++--
> ----------
> >  drivers/usb/cdns3/cdns3-gadget.h |    2 +
> >  2 files changed, 51 insertions(+), 24 deletions(-)
> >
> > diff --git a/drivers/usb/cdns3/cdns3-gadget.c b/drivers/usb/cdns3/cdns3=
-
> gadget.c
> > index 5f51215..b4955ce 100644
> > --- a/drivers/usb/cdns3/cdns3-gadget.c
> > +++ b/drivers/usb/cdns3/cdns3-gadget.c
> > @@ -818,10 +818,26 @@ void cdns3_gadget_giveback(struct
> cdns3_endpoint *priv_ep,
> >  	usb_gadget_unmap_request_by_dev(priv_dev->sysdev, request,
> >  					priv_ep->dir);
> >
> > -	if ((priv_req->flags & REQUEST_UNALIGNED) &&
> > -	    priv_ep->dir =3D=3D USB_DIR_OUT && !request->status)
> > -		memcpy(request->buf, priv_req->aligned_buf->buf,
> > -		       request->length);
> > +	if ((priv_req->flags & REQUEST_UNALIGNED) && priv_req-
> >aligned_buf) {
> > +		struct cdns3_aligned_buf *buf;
> > +
> > +		buf =3D priv_req->aligned_buf;
> > +		dma_unmap_single(priv_dev->sysdev, buf->dma, buf->size,
> > +			buf->dir);
> > +		priv_req->flags &=3D ~REQUEST_UNALIGNED;
> > +
> > +		if (priv_ep->dir =3D=3D USB_DIR_OUT && !request->status) {
> > +			memcpy(request->buf, priv_req->aligned_buf->buf,
> > +			       request->length);
> > +		}
> > +
> > +		trace_cdns3_free_aligned_request(priv_req);
> > +		priv_req->aligned_buf->in_use =3D 0;
> > +		queue_work(system_freezable_wq,
> > +			   &priv_dev->aligned_buf_wq);
> > +		priv_req->aligned_buf =3D NULL;
> > +
> > +	}
> >
> >  	priv_req->flags &=3D ~(REQUEST_PENDING | REQUEST_UNALIGNED);
> >  	/* All TRBs have finished, clear the counter */
> > @@ -883,8 +899,7 @@ static void cdns3_free_aligned_request_buf(struct
> work_struct *work)
> >  			 * interrupts.
> >  			 */
> >  			spin_unlock_irqrestore(&priv_dev->lock, flags);
> > -			dma_free_coherent(priv_dev->sysdev, buf->size,
> > -					  buf->buf, buf->dma);
> > +			kfree(buf->buf);
> >  			kfree(buf);
> >  			spin_lock_irqsave(&priv_dev->lock, flags);
> >  		}
> > @@ -910,27 +925,16 @@ static int
> cdns3_prepare_aligned_request_buf(struct cdns3_request *priv_req)
> >  		if (!buf)
> >  			return -ENOMEM;
> >
> > -		buf->size =3D priv_req->request.length;
> > +		buf->size =3D usb_endpoint_dir_out(priv_ep->endpoint.desc)
> ?
> > +				usb_ep_align(&(priv_ep->endpoint),
> priv_req->request.length)
> > +				: priv_req->request.length;
> >
> > -		buf->buf =3D dma_alloc_coherent(priv_dev->sysdev,
> > -					      buf->size,
> > -					      &buf->dma,
> > -					      GFP_ATOMIC);
> > +		buf->buf =3D kmalloc(buf->size, GFP_ATOMIC);
> >  		if (!buf->buf) {
> >  			kfree(buf);
> >  			return -ENOMEM;
> >  		}
> >
> > -		if (priv_req->aligned_buf) {
> > -			trace_cdns3_free_aligned_request(priv_req);
> > -			priv_req->aligned_buf->in_use =3D 0;
> > -			queue_work(system_freezable_wq,
> > -				   &priv_dev->aligned_buf_wq);
> > -		}
> > -
> > -		buf->in_use =3D 1;
> > -		priv_req->aligned_buf =3D buf;
> > -
> >  		list_add_tail(&buf->list,
> >  			      &priv_dev->aligned_buf_list);
> >  	}
> > @@ -940,6 +944,27 @@ static int
> cdns3_prepare_aligned_request_buf(struct cdns3_request *priv_req)
> >  		       priv_req->request.length);
> >  	}
> >
> > +	if (priv_req->aligned_buf) {
> > +		trace_cdns3_free_aligned_request(priv_req);
> > +		priv_req->aligned_buf->in_use =3D 0;
> > +		queue_work(system_freezable_wq,
> > +			   &priv_dev->aligned_buf_wq);
>=20
> @Pawel, do you remember when this condition is met?
>=20
> > +	}
> > +
> > +	buf->dir =3D  priv_ep->dir ? DMA_TO_DEVICE : DMA_FROM_DEVICE;
> > +	buf->in_use =3D 1;
> > +	priv_req->aligned_buf =3D buf;
> > +
> > +	buf->dma =3D dma_map_single(priv_dev->sysdev, buf->buf, buf-
> >size,
> > +				buf->dir);
> > +
> > +	if (dma_mapping_error(priv_dev->sysdev, buf->dma)) {
> > +		dev_err(priv_dev->dev, "Failed to map buffer\n");
> > +		kfree(buf->buf);
> > +		kfree(buf);
> > +		return -EFAULT;
> > +	}
> > +
> >  	priv_req->flags |=3D REQUEST_UNALIGNED;
> >  	trace_cdns3_prepare_aligned_request(priv_req);
> >
> > @@ -3088,11 +3113,11 @@ static void cdns3_gadget_exit(struct cdns
> *cdns)
> >  		struct cdns3_aligned_buf *buf;
> >
> >  		buf =3D cdns3_next_align_buf(&priv_dev->aligned_buf_list);
> > -		dma_free_coherent(priv_dev->sysdev, buf->size,
> > -				  buf->buf,
> > -				  buf->dma);
> > +		dma_unmap_single(priv_dev->sysdev, buf->dma, buf->size,
> > +			buf->dir);
>=20
> It only needs to DMA unmap after DMA has completed, this buf will not be
> used, otherwise, the kfree below will cause issue.

This part is not clear.  Aligned DMA buffer is allocated and mapped in cdns=
3_prepare_aligned_request_buf()
and put into aligned_buf_list. While unloading the gadget, We need to undo =
the same if aligned_buf_list is not
empty.  Am I missing something here?=20

Also, I will post v2 of this patch which uses dma_*_noncoherent APIs sugges=
ted by Christoph Hellwig.
=20
>=20
> >
> >  		list_del(&buf->list);
> > +		kfree(buf->buf);
> >  		kfree(buf);
> >  	}
> >
> > diff --git a/drivers/usb/cdns3/cdns3-gadget.h b/drivers/usb/cdns3/cdns3=
-
> gadget.h
> > index ecf9b91..c5660f2 100644
> > --- a/drivers/usb/cdns3/cdns3-gadget.h
> > +++ b/drivers/usb/cdns3/cdns3-gadget.h
> > @@ -12,6 +12,7 @@
> >  #ifndef __LINUX_CDNS3_GADGET
> >  #define __LINUX_CDNS3_GADGET
> >  #include <linux/usb/gadget.h>
> > +#include <linux/dma-direction.h>
> >
> >  /*
> >   * USBSS-DEV register interface.
> > @@ -1205,6 +1206,7 @@ struct cdns3_aligned_buf {
> >  	void			*buf;
> >  	dma_addr_t		dma;
> >  	u32			size;
> > +	enum dma_data_direction dir;
> >  	unsigned		in_use:1;
> >  	struct list_head	list;
> >  };
> > --
> > 1.7.1
> >
>=20
> --
>=20
> Thanks,
> Peter Chen


Thanks,
Sanket