| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20250812134222.GN30054@pendragon.ideasonboard.com>
Date: Tue, 12 Aug 2025 16:42:22 +0300
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: "Du, Bin" <bin.du@....com>
Cc: mchehab@...nel.org, hverkuil@...all.nl, bryan.odonoghue@...aro.org,
sakari.ailus@...ux.intel.com,
prabhakar.mahadev-lad.rj@...renesas.com,
linux-media@...r.kernel.org, linux-kernel@...r.kernel.org,
pratap.nirujogi@....com, benjamin.chan@....com, king.li@....com,
gjorgji.rosikopulos@....com, Phil.Jawich@....com,
Dominic.Antony@....com, Svetoslav.Stoilov@....com
Subject: Re: [PATCH v2 8/8] Documentation: add documentation of AMD isp 4
driver
On Tue, Aug 12, 2025 at 09:36:04AM +0800, Du, Bin wrote:
> On 8/5/2025 7:37 PM, Laurent Pinchart wrote:
> > On Wed, Jun 18, 2025 at 05:19:59PM +0800, Bin Du wrote:
> >> Add documentation for AMD isp 4 and describe the main components
> >>
> >> Signed-off-by: Bin Du <Bin.Du@....com>
> >> Signed-off-by: Svetoslav Stoilov <Svetoslav.Stoilov@....com>
> >> ---
> >> Documentation/admin-guide/media/amdisp4-1.rst | 64 +++++++++++++++++++
> >> Documentation/admin-guide/media/amdisp4.dot | 8 +++
> >> MAINTAINERS | 2 +
> >> 3 files changed, 74 insertions(+)
> >> create mode 100644 Documentation/admin-guide/media/amdisp4-1.rst
> >> create mode 100644 Documentation/admin-guide/media/amdisp4.dot
> >>
> >> diff --git a/Documentation/admin-guide/media/amdisp4-1.rst b/Documentation/admin-guide/media/amdisp4-1.rst
> >> new file mode 100644
> >> index 000000000000..417b15af689a
> >> --- /dev/null
> >> +++ b/Documentation/admin-guide/media/amdisp4-1.rst
> >> @@ -0,0 +1,64 @@
> >> +.. SPDX-License-Identifier: GPL-2.0
> >> +
> >> +.. include:: <isonum.txt>
> >> +
> >> +====================================
> >> +AMD Image Signal Processor (amdisp4)
> >> +====================================
> >> +
> >> +Introduction
> >> +============
> >> +
> >> +This file documents the driver for the AMD ISP4 that is part of
> >> +AMD Ryzen AI Max 385 SoC.
> >> +
> >> +The driver is located under drivers/media/platform/amd/isp4 and uses
> >> +the Media-Controller API.
> >> +
> >> +Topology
> >> +========
> >> +
> >> +.. _amdisp4_topology_graph:
> >> +
> >> +.. kernel-figure:: amdisp4.dot
> >> + :alt: Diagram of the media pipeline topology
> >> + :align: center
> >> +
> >> +
> >> +
> >> +The driver has 1 sub-device:
> >> +
> >> +- isp: used to resize and process bayer raw frames in to yuv.
> >> +
> >> +The driver has 1 video device:
> >> +
> >> +- <capture video device: capture device for retrieving images.
> >> +
> >> +
> >> + - ISP4 Image Signal Processing Subdevice Node
> >> +-----------------------------------------------
> >> +
> >> +The isp4 is represented as a single V4L2 subdev, the sub-device does not
> >> +provide interface to the user space.
> >
> > Doesn't it ? The driver sets the V4L2_SUBDEV_FL_HAS_DEVNODE flag for the
> > subdev, and calls v4l2_device_register_subdev_nodes().
>
> We have exported subdev device to user space during the testing with
> libcamera sample pipeline.
But it's not needed anymore ? If not, you could stop exposing the subdev
to userspace for the time being.
> > As far as I understand, the camera is exposed by the firmware with a
> > webcam-like interface. We need to better understand your plans with this
> > driver. If everything is handled by the firmware, why are the sensor and
> > subdev exposed to userspace ? Why can't you expose a single video
> > capture device, with a media device, and handle everything behind the
> > scene ? I assume there may be more features coming later. Please
> > document the plan, we can't provide feedback on the architecture
> > otherwise.
>
> Currently, isp fw is controlling the sensor to update just the exposure
> and gain, since the 3A algorithms run on ISP HW rather than on x86.
This design decision makes my hair stand on end :-( Exposing the camera
sensor to both the firmware and the host concurrently is asking for
trouble. If you really want to abstract the camera behind a firmware and
only expose a webcam-like API (or not even that in this version, as the
driver exposes no control as far as I can see), then you should push the
whole sensor handling to the firmware too. In my opinion that would not
be a good solution compared to exposing the ISP to the host, but it
would be better than this hybrid model.
> In a
> future version, we plan to introduce raw output support in the ISP
> driver, allowing users to choose between AMD’s 3A running on ISP
> hardware or a custom 3A running on x86. If the user opts for the
> x86-based 3A, the firmware will relinquish control of the sensor, and
> hands over full control to the x86 system.
Will the firmware at that point expose to Linux all the ISP statistics
needed to implement auto-exposure ? What if I want to also set digital
gain ? Or have manual white balance (requiring statistics), and manual
CCM ?
> >> The sub-device is connected to one video node
> >> +(isp4_capture) with immutable active link. The isp entity is connected
> >> +to sensor pad 0 and receives the frames using CSI-2 protocol. The sub-device is
> >> +also responsible to configure CSI2-2 receiver.
> >> +The sub-device processes bayer raw data from the connected sensor and output
> >> +them to different YUV formats. The isp also has scaling capabilities.
> >> +
> >> + - isp4_capture - Frames Capture Video Node
> >> +--------------------------------------------
> >> +
> >> +Isp4_capture is a capture device to capture frames to memory.
> >> +This entity is the DMA engine that write the frames to memory.
> >> +The entity is connected to isp4 sub-device.
> >> +
> >> +Capturing Video Frames Example
> >> +==============================
> >> +
> >> +.. code-block:: bash
> >> +
> >> + # set the links
> >
> > This seems very under-documented.
>
> Yes, documentation needs to be updated.
>
> >> +
> >> + # start streaming:
> >> + v4l2-ctl "-d" "/dev/video0" "--set-fmt-video=width=1920,height=1080,pixelformat=NV12" "--stream-mmap" "--stream-count=10"
> >> diff --git a/Documentation/admin-guide/media/amdisp4.dot b/Documentation/admin-guide/media/amdisp4.dot
> >> new file mode 100644
> >> index 000000000000..a4c2f0cceb30
> >> --- /dev/null
> >> +++ b/Documentation/admin-guide/media/amdisp4.dot
> >> @@ -0,0 +1,8 @@
> >> +digraph board {
> >> + rankdir=TB
> >> + n00000001 [label="{{<port0> 0} | amd isp4\n | {<port1> 1}}", shape=Mrecord, style=filled, fillcolor=green]
> >> + n00000001:port1 -> n00000004 [style=bold]
> >> + n00000004 [label="Preview\n/dev/video0", shape=box, style=filled, fillcolor=yellow]
> >> + n0000000a [label="{{} | ov05c10 22-0010\n/dev/v4l-subdev0 | {<port0> 0}}", shape=Mrecord, style=filled, fillcolor=green]
> >> + n0000000a:port0 -> n00000001:port0 [style=bold]
> >> +}
> >> diff --git a/MAINTAINERS b/MAINTAINERS
> >> index 15070afb14b5..e4455bde376f 100644
> >> --- a/MAINTAINERS
> >> +++ b/MAINTAINERS
> >> @@ -1113,6 +1113,8 @@ M: Nirujogi Pratap <pratap.nirujogi@....com>
> >> L: linux-media@...r.kernel.org
> >> S: Maintained
> >> T: git git://linuxtv.org/media.git
> >> +F: Documentation/admin-guide/media/amdisp4-1.rst
> >> +F: Documentation/admin-guide/media/amdisp4.dot
> >> F: drivers/media/platform/amd/Kconfig
> >> F: drivers/media/platform/amd/Makefile
> >> F: drivers/media/platform/amd/isp4/*
--
Regards,
Laurent Pinchart
Powered by blists - more mailing lists