mbox series

[v6,0/3] nvmem: stm32: add OP-TEE support for STM32MP13x

Message ID 20230118172940.841094-1-patrick.delaunay@foss.st.com
Headers
Series nvmem: stm32: add OP-TEE support for STM32MP13x |

Message

Patrick Delaunay Jan. 18, 2023, 5:29 p.m. UTC 
  The v5 patchset is rebased on next-20221226.

This serie update the NVMEM BSEC driver to be compatible with STM32MP13x
SoC and the trusted application STM32MP BSEC in OP-TEE

This serie solve issue in initial support of STM32MP131
(using BSEC STM32MP15 compatible) and so it break the STM32MP13x DTS
compatible.

I create this serie for more efficient review, including support for
STM32MP15x.

The first patches of the V1 series is already merged:
"dt-bindings: nvmem: add new stm32mp13 compatible for stm32-romem"

This STM32MP13x DTS break is acceptable as
- the STM32MP13x SoC is not yet available outside STMicroelectronics
  (not official)
- the same patch is already integrated or modifications are in progress in
  the other users (arm-trusted-firmware/TF-A, OP-TEE and U-Boot) of
  stm32mp131 device tree.

It is the good time to correct this issue before the real availability of
the SoC and before full support of STM32MP13x SoC in Linux kernel.

Regards

Patrick

Changes in v6:
- Add reviewed by  Etienne Carierre review
- added reviewed by Etienne Carriere

Changes in v5:
- minor changes after Etienne Carierre review (comments,
  change %x to %#x, remove goto to out_tee_session)
- update the BSEC SMC detection logic in stm32_romem_probe()
  after Etienne Carierre review to support NVMEM probe after OP-TEE probe

Changes in v4:
- fixe warning reported by kernel test robot for 64 bits support in
  drivers/nvmem/stm32-bsec-optee-ta.c:260:18:
  warning: format '%d' expects argument of type 'int',
  but argument 4 has type 'size_t'

Changes in v3:
- add a separate file stm32-bsec-optee-ta.c with STM32MP BSEC TA
  communication functions to avoid #if in romem driver.
- use of_find_compatible_node in optee_presence_check function
  instead of of_find_node_by_path("/firmware/optee")

Changes in v2:
- rebase series on linux-next/master
- minor update after V1 revue

Changes in v1:
- update commit message to indicate DTS break reason.

Patrick Delaunay (3):
  ARM: dts: stm32mp13: fix compatible for BSEC
  nvmem: stm32: add OP-TEE support for STM32MP13x
  nvmem: stm32: detect bsec pta presence for STM32MP15x

 arch/arm/boot/dts/stm32mp131.dtsi   |   2 +-
 drivers/nvmem/Kconfig               |  11 +
 drivers/nvmem/Makefile              |   1 +
 drivers/nvmem/stm32-bsec-optee-ta.c | 298 ++++++++++++++++++++++++++++
 drivers/nvmem/stm32-bsec-optee-ta.h |  80 ++++++++
 drivers/nvmem/stm32-romem.c         |  84 +++++++-
 6 files changed, 472 insertions(+), 4 deletions(-)
 create mode 100644 drivers/nvmem/stm32-bsec-optee-ta.c
 create mode 100644 drivers/nvmem/stm32-bsec-optee-ta.h

base-commit: c76083fac3bae1a87ae3d005b5cb1cbc761e31d5
prerequisite-patch-id: 5aaa8fffbdd16871143808180b3932d80f4045d0
prerequisite-patch-id: ae711dc528e191e4751cbb7402041fc5f185d6b3

Comments

Arnd Bergmann Jan. 26, 2023, 9:02 p.m. UTC | #1 
On Wed, Jan 18, 2023, at 18:29, Patrick Delaunay wrote:
> 
> +config NVMEM_STM32_BSEC_OPTEE_TA
> +	bool "STM32MP BSEC OP-TEE TA support for nvmem-stm32-romem driver"
> +	depends on OPTEE
> +	help
> +	  Say y here to enable the accesses to STM32MP SoC OTPs by the OP-TEE
> +	  trusted application STM32MP BSEC.
> +
> +	  This library is a used by stm32-romem driver or included in the 
> module
> +	  called nvmem-stm32-romem.
> +
>  config NVMEM_STM32_ROMEM
>  	tristate "STMicroelectronics STM32 factory-programmed memory support"
>  	depends on ARCH_STM32 || COMPILE_TEST
> +	imply NVMEM_STM32_BSEC_OPTEE_TA
>  	help
>  	  Say y here to enable read-only access for STMicroelectronics STM32
>  	  factory-programmed memory area.

This is now causing a link failure with CONFIG_OPTEE=m if
NVMEM_STM32_ROMEM is built-in. My guess is that you saw something
like that earlier and someone recommended using the 'imply' keyword
without understanding what it does (no worries, nobody understands it).

I've prepared a patch now based on the most likely interpretation
of what you actually meant here:

diff --git a/drivers/nvmem/Kconfig b/drivers/nvmem/Kconfig
index ed8ef7460be2..ae2c5257ed97 100644
--- a/drivers/nvmem/Kconfig
+++ b/drivers/nvmem/Kconfig
@@ -295,8 +295,7 @@ config NVMEM_SPRD_EFUSE
          will be called nvmem-sprd-efuse.
 
 config NVMEM_STM32_BSEC_OPTEE_TA
-       bool "STM32MP BSEC OP-TEE TA support for nvmem-stm32-romem driver"
-       depends on OPTEE
+       def_bool NVMEM_STM32_ROMEM && OPTEE
        help
          Say y here to enable the accesses to STM32MP SoC OTPs by the OP-TEE
          trusted application STM32MP BSEC.
@@ -307,7 +306,7 @@ config NVMEM_STM32_BSEC_OPTEE_TA
 config NVMEM_STM32_ROMEM
        tristate "STMicroelectronics STM32 factory-programmed memory support"
        depends on ARCH_STM32 || COMPILE_TEST
-       imply NVMEM_STM32_BSEC_OPTEE_TA
+       depends on OPTEE || !OPTEE
        help
          Say y here to enable read-only access for STMicroelectronics STM32
          factory-programmed memory area.

This enables NVMEM_STM32_BSEC_OPTEE_TA whenever OPTEE is
available, but prevents the link error by forcing NVMEM_STM32_ROMEM
to also be a loadable module if that is how OPTEE is built.

I'll send that if it passes the randconfig builds over night.

      Arnd