linux-media.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Phil Elwell <phil@raspberrypi.com>
To: Nicolas Saenz Julienne <nsaenzjulienne@suse.de>
Cc: "Laurent Pinchart" <laurent.pinchart@ideasonboard.com>,
	linux-media@vger.kernel.org,
	"Kieran Bingham" <kieran.bingham@ideasonboard.com>,
	"Jacopo Mondi" <jacopo@jmondi.org>,
	"Niklas Söderlund" <niklas.soderlund@ragnatech.se>,
	"Naushir Patuck" <naush@raspberrypi.com>,
	"Dave Stevenson" <dave.stevenson@raspberrypi.com>,
	"Phil Elwell" <phil@raspberrypi.org>
Subject: Re: [PATCH v2 29/34] staging: vchiq: Add 36-bit address support
Date: Tue, 5 May 2020 11:57:37 +0100	[thread overview]
Message-ID: <CAMEGJJ1qT1Kup4BAg_48Pk=zYKQTXV9A-2ZThDd5YJyQvSRzjw@mail.gmail.com> (raw)
In-Reply-To: <09a896b78b42d2c78ed3ad468de7e90163b9d6c4.camel@suse.de>

Hi Nicolas,

On Tue, 5 May 2020 at 11:13, Nicolas Saenz Julienne
<nsaenzjulienne@suse.de> wrote:
>
> On Mon, 2020-05-04 at 21:46 +0100, Phil Elwell wrote:
> > Hi Nicolas,
> >
> > On 04/05/2020 18:40, Nicolas Saenz Julienne wrote:
> > > Hi Phil, Laurent,
> > >
> > > On Mon, 2020-05-04 at 12:26 +0300, Laurent Pinchart wrote:
> > > > From: Phil Elwell <phil@raspberrypi.org>
> > > >
> > > > Conditional on a new compatible string, change the pagelist encoding
> > > > such that the top 24 bits are the pfn, leaving 8 bits for run length
> > > > (-1).
> > > >
> > > > Signed-off-by: Phil Elwell <phil@raspberrypi.org>
> > > > Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
> > > > ---
> > > >   .../interface/vchiq_arm/vchiq_2835_arm.c      | 88 ++++++++++++++-----
> > > >   .../interface/vchiq_arm/vchiq_arm.c           |  6 ++
> > > >   .../interface/vchiq_arm/vchiq_arm.h           |  1 +
> > > >   3 files changed, 74 insertions(+), 21 deletions(-)
> > > >
> > > > diff --git
> > > > a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
> > > > b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
> > > > index 3e422a7eb3f1..ecec84ad4345 100644
> > > > --- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
> > > > +++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c
> > > > @@ -16,6 +16,8 @@
> > > >   #include <soc/bcm2835/raspberrypi-firmware.h>
> > > >     #define TOTAL_SLOTS (VCHIQ_SLOT_ZERO_SLOTS + 2 * 32)
> > > > +#define VC_SAFE(x) (g_use_36bit_addrs ? ((u32)(x) | 0xc0000000) :
> > > > (u32)(x))
> > > > +#define IS_VC_SAFE(x) (g_use_36bit_addrs ? !((x) & ~0x3fffffffull) : 1)
> > >
> > > As I commented earlier, this is a sign your dma-ranges are wrong, most of
> > > the
> > > code below reimplements what is already done by dma-direct (see
> > > kernel/dma/direct.c). Once properly setup, you should be able to use
> > > whatever
> > > phys address dmam_alloc_coherent() provides and drop g_use_36bit_addrs.
> > >
> > > Note that on arm32+LPAE, dma-direct/swiotlb are the default dma_ops, so this
> > > also applies there.
> >
> > As I explained in an offline email, the problem is that VCHIQ needs access to
>
> We discussed this privately, but I wanted to start from scratch, specially as I
> hope I won't be the only one reviewing this :).
>
> > two
> > kinds of "DMA" addresses - those suitable for the 40-bit DMA channel, and the
> > 30-bit addresses that the VPU can use. Since each DT node only has access to a
> > single set of DMA ranges, I can't see how to use dma-direct to calculate
> > addreses
> > for everything, but feel free to submit an alternative implementation showing
> > how
> > it should be done.
>
> How about this):
>  - Move vchiq to /soc
>  - Get a handle to the relevant dma controller device pointer (so 30-bit DMA
>    controller on old RPis, 40-bit on RPi4)
>  - Allocate slotmem/pagelist with dma_alloc_coherent(vpu_dev, ...)
>  - Map pages with dma_map_sg(dma_dev, ...)
>
> I hope I'm not missing anything, but short of implementing it and seeing the
> end result, I think you'd be free of any rpi[1-3]/rpi4 distinction in the code.

Thanks for the suggestion - I hadn't considered using a device pointer
for the controller rather than the current (client) device. If that
works then I propose
to make the 40-bit dma property optional such that dma_dev defaults to
vpu_dev.

Phil

  reply	other threads:[~2020-05-05 10:57 UTC|newest]

Thread overview: 104+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-04  9:25 [PATCH v2 00/34] Drivers for the BCM283x CSI-2/CCP2 receiver and ISP Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 01/34] media: uapi: v4l2-core: Add sensor ancillary data V4L2 fourcc type Laurent Pinchart
2020-05-04 13:48   ` Hans Verkuil
2020-05-04 14:39     ` Dave Stevenson
2020-05-04 15:32       ` Hans Verkuil
2020-05-04 16:08         ` Laurent Pinchart
2020-05-05 11:20           ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 02/34] media: uapi: Add MEDIA_BUS_FMT_SENSOR_DATA media bus format Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 03/34] dt-bindings: media: Document BCM283x CSI2/CCP2 receiver Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 04/34] media: bcm2835-unicam: Driver for CCP2/CSI2 camera interface Laurent Pinchart
2020-05-04 15:21   ` Hans Verkuil
2020-05-05  1:26   ` kbuild test robot
2020-05-06 18:01   ` Nicolas Saenz Julienne
2020-08-29 11:20   ` Jacopo Mondi
2020-08-29 18:32     ` Laurent Pinchart
2020-08-31  7:38       ` Jacopo Mondi
2020-08-31 14:17         ` Laurent Pinchart
2020-08-31 14:46           ` Jacopo Mondi
2020-08-31 14:56             ` Laurent Pinchart
2020-09-01  8:41               ` Dave Stevenson
2020-09-01 10:22                 ` Jacopo Mondi
2020-09-01 16:37                   ` Dave Stevenson
2020-09-01 17:11                     ` Laurent Pinchart
2020-09-15  7:03   ` Sakari Ailus
2020-09-15  9:32     ` Laurent Pinchart
2020-09-15 13:28       ` Dave Stevenson
2020-10-30 17:53         ` Jacopo Mondi
2020-09-15 17:30     ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 05/34] ARM: dts: bcm2711: Add Unicam DT nodes Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 06/34] staging: vc04_services: Add new vc-sm-cma driver Laurent Pinchart
2020-05-05  2:38   ` kbuild test robot
2020-05-05 12:17   ` kbuild test robot
2020-05-06  3:05   ` kbuild test robot
2020-05-06 18:04   ` Nicolas Saenz Julienne
2020-05-06 19:24     ` Dave Stevenson
2020-05-08  0:11       ` Laurent Pinchart
2020-05-18 12:06         ` Hans Verkuil
2020-08-24 16:39       ` Jacopo Mondi
2020-08-25 17:52         ` Dave Stevenson
2020-08-27 10:38           ` Jacopo Mondi
2020-08-27 12:51             ` Dave Stevenson
2020-08-27 16:46               ` Jacopo Mondi
2020-08-27 17:19                 ` Dave Stevenson
2020-05-11 18:42   ` Nicolas Saenz Julienne
2020-05-18 15:48     ` Dave Stevenson
2020-05-20 14:41       ` Nicolas Saenz Julienne
2020-05-21 11:01         ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 07/34] staging: bcm2835: Break MMAL support out from camera Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 08/34] staging: mmal-vchiq: Allocate and free components as required Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 09/34] staging: mmal-vchiq: Avoid use of bool in structures Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 10/34] staging: mmal-vchiq: Make timeout a defined parameter Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 11/34] staging: mmal-vchiq: Make a mmal_buf struct for passing parameters Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 12/34] staging: mmal-vchiq: Add support for event callbacks Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 13/34] staging: mmal-vchiq: Support sending data to MMAL ports Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 14/34] staging: mmal-vchiq: Fixup vchiq-mmal include ordering Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 15/34] staging: mmal-vchiq: Use vc-sm-cma to support zero copy Laurent Pinchart
2020-05-04 16:30   ` Nicolas Saenz Julienne
2020-05-05 16:07   ` kbuild test robot
2020-05-11 19:15   ` Nicolas Saenz Julienne
2020-05-04  9:25 ` [PATCH v2 16/34] staging: mmal-vchiq: Fix client_component for 64 bit kernel Laurent Pinchart
2020-05-18 10:19   ` Hans Verkuil
2020-05-04  9:25 ` [PATCH v2 17/34] staging: mmal_vchiq: Add in the Bayer encoding formats Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 18/34] staging: mmal-vchiq: Always return the param size from param_get Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 19/34] staging: mmal-vchiq: If the VPU returns an error, don't negate it Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 20/34] staging: mmal-vchiq: Fix handling of VB2_MEMORY_DMABUF buffers Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 21/34] staging: mmal-vchiq: Update mmal_parameters.h with recently defined params Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 22/34] staging: mmal-vchiq: Free the event context for control ports Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 23/34] staging: mmal-vchiq: Fix memory leak in error path Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 24/34] staging: mmal-vchiq: Fix formatting errors in mmal_parameters.h Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 25/34] staging: vchiq_arm: Register vcsm-cma as a platform driver Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 26/34] staging: vchiq_arm: Set up dma ranges on child devices Laurent Pinchart
2020-05-04 16:54   ` Nicolas Saenz Julienne
2020-08-25 16:57     ` Jacopo Mondi
2020-05-04  9:26 ` [PATCH v2 27/34] staging: vchiq: Use the old dma controller for OF config on platform devices Laurent Pinchart
2020-05-04 15:44   ` Nicolas Saenz Julienne
2020-05-04  9:26 ` [PATCH v2 28/34] staging: vchiq_2835_arm: Implement a DMA pool for small bulk transfers Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 29/34] staging: vchiq: Add 36-bit address support Laurent Pinchart
2020-05-04 17:40   ` Nicolas Saenz Julienne
2020-05-04 20:46     ` Phil Elwell
2020-05-05 10:13       ` Nicolas Saenz Julienne
2020-05-05 10:57         ` Phil Elwell [this message]
2020-05-05 13:22   ` kbuild test robot
2020-05-04  9:26 ` [PATCH v2 30/34] staging: vchiq_arm: Give vchiq children DT nodes Laurent Pinchart
2020-05-04 17:12   ` Nicolas Saenz Julienne
2020-05-04 19:42     ` Phil Elwell
2020-05-05 10:37       ` Nicolas Saenz Julienne
2020-05-05 10:50         ` Phil Elwell
2020-05-04  9:26 ` [PATCH v2 31/34] staging: vchiq_arm: Add a matching unregister call Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 32/34] media: videobuf2: Allow exporting of a struct dmabuf Laurent Pinchart
2020-05-04 13:36   ` Hans Verkuil
2020-05-04  9:26 ` [PATCH v2 33/34] staging: bcm2835-isp: Add support for BC2835 ISP Laurent Pinchart
2020-05-11 19:19   ` Nicolas Saenz Julienne
2020-05-18 13:38     ` Dave Stevenson
2020-05-20 13:46       ` Nicolas Saenz Julienne
2020-05-18 12:02   ` Hans Verkuil
2020-05-18 14:36     ` Dave Stevenson
2020-05-18 15:07       ` Hans Verkuil
2020-05-19 12:47     ` Naushir Patuck
2020-05-19 13:03       ` Jacopo Mondi
2020-06-24 11:28       ` Hans Verkuil
2020-05-04  9:26 ` [PATCH v2 34/34] staging: vchiq: Load bcm2835_isp driver from vchiq Laurent Pinchart
2020-05-04 15:15 ` [PATCH v2 00/34] Drivers for the BCM283x CSI-2/CCP2 receiver and ISP Nicolas Saenz Julienne
2020-05-04 15:38   ` Laurent Pinchart
2020-05-04 16:15     ` Nicolas Saenz Julienne

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAMEGJJ1qT1Kup4BAg_48Pk=zYKQTXV9A-2ZThDd5YJyQvSRzjw@mail.gmail.com' \
    --to=phil@raspberrypi.com \
    --cc=dave.stevenson@raspberrypi.com \
    --cc=jacopo@jmondi.org \
    --cc=kieran.bingham@ideasonboard.com \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=linux-media@vger.kernel.org \
    --cc=naush@raspberrypi.com \
    --cc=niklas.soderlund@ragnatech.se \
    --cc=nsaenzjulienne@suse.de \
    --cc=phil@raspberrypi.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).