linux-kernel - [PATCH v3 0/4] Add support of mt8183 APU

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20210819151340.741565-1-abailon@baylibre.com>
Date:   Thu, 19 Aug 2021 17:13:36 +0200
From:   Alexandre Bailon <abailon@...libre.com>
To:     ohad@...ery.com, bjorn.andersson@...aro.org,
        mathieu.poirier@...aro.org, robh+dt@...nel.org
Cc:     matthias.bgg@...il.com, linux-remoteproc@...r.kernel.org,
        devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-mediatek@...ts.infradead.org, linux-kernel@...r.kernel.org,
        stephane.leprovost@...iatek.com, gpain@...libre.com,
        khilman@...libre.com, Alexandre Bailon <abailon@...libre.com>
Subject: [PATCH v3 0/4]  Add support of mt8183 APU

Some Mediatek's SoC have an Accelerated Processing Unit.
This adds support of the one available in the mt8183
(aswell some derivative SoC).

This v3 attempt to handle memory differently.
In the v2, the memory was reserved in dts and all dma allocation made by
remoteproc was using it. Since we may load many firmwares with different size,
this could lead to a memory waste or out of memory issues.
In addition, the APU itself expect the memory to mapped at a specific address.
The firmware, because of many hardware / software limitations use hardcoded
address.
To simplify memory management of userspace application, kernel and firmware,
and to rmeove a couple of hacks, we changed the way to manage the memory.
With the v3, only the virtio memory (that won't never change) is reserved
in the dts. The firmware declare resource similar to the CARVOUT_RSC 
that used by this driver to allocate and map the memory using the IOMMU.
I am not sure that the best solution but the simplest one I found so far.

Changes in v3:
- Remove IOMMU hack. Instead, manage the IOMMU directly from the driver
- Update the device tree bindings: only use reserved memory for virtio.
  All the other memory allocation will be done using DMA API.
  This sould simplify the memory management.
- Add more comments
Changes in v2:
- Drop the workarounds needed to load bad firmwares
- There are many name for the APU (most common one is VPU).
  Rename many functions and dts nodes to be more consistent.
- Use the bulk clock API, and enable / disable clock at a better place
- add few comments explaining how to start the APU
- update the way to use pinctl for JTAG
- fix some minors issues
- fix device tree bindings

Alexandre Bailon (4):
  dt bindings: remoteproc: Add bindings for MT8183 APU
  remoteproc: Add a remoteproc driver for the MT8183's APU
  remoteproc: mtk_vpu_rproc: Add support of JTAG
  ARM64: mt8183: Add support of APU to mt8183

 .../bindings/remoteproc/mtk,apu.yaml          | 118 ++++
 .../boot/dts/mediatek/mt8183-pumpkin.dts      |  48 ++
 arch/arm64/boot/dts/mediatek/mt8183.dtsi      |  40 ++
 drivers/remoteproc/Kconfig                    |  19 +
 drivers/remoteproc/Makefile                   |   1 +
 drivers/remoteproc/mtk_apu.c                  | 590 ++++++++++++++++++
 6 files changed, 816 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/remoteproc/mtk,apu.yaml
 create mode 100644 drivers/remoteproc/mtk_apu.c

-- 
2.31.1