Date: Thu, 20 Jun 2024 00:16:24 +0530
From: Mukesh Ojha <quic_mojha@...cinc.com>
To: Bjorn Andersson <quic_bjorande@...cinc.com>
CC: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>,
        Krzysztof Kozlowski
	<krzysztof.kozlowski@...aro.org>,
        Komal Bajaj <quic_kbajaj@...cinc.com>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>,
        Konrad Dybcio <konrad.dybcio@...aro.org>,
        Srinivas Kandagatla
	<srinivas.kandagatla@...aro.org>,
        <linux-arm-kernel@...ts.infradead.org>, <linux-kernel@...r.kernel.org>,
        <linux-arm-msm@...r.kernel.org>
Subject: Re: [PATCH] arm64: defconfig: Enable secure QFPROM driver

On Wed, Jun 19, 2024 at 09:51:48AM -0700, Bjorn Andersson wrote:
> On Wed, Jun 19, 2024 at 10:08:10PM +0530, Mukesh Ojha wrote:
> > On Wed, Jun 19, 2024 at 04:14:50PM +0300, Dmitry Baryshkov wrote:
> > > On Wed, Jun 19, 2024 at 05:40:42PM GMT, Mukesh Ojha wrote:
> > > > On Wed, Jun 19, 2024 at 01:08:48PM +0200, Krzysztof Kozlowski wrote:
> > > > > On 19/06/2024 12:56, Komal Bajaj wrote:
> > > > > > Enable the secure QFPROM driver which is used by QDU1000
> > > > > 
> > > > > Qualcomm QDU1000. You are changing kernel-wide defconfig, not some
> > > > > Qualcomm downstream stuff.
> > > > > 
> > > > > > platform for reading the secure qfprom region to get the
> > > > > > DDR channel configuration.
> > > > > > 
> > > > > > Signed-off-by: Komal Bajaj <quic_kbajaj@...cinc.com>
> > > > > > ---
> > > > > >  arch/arm64/configs/defconfig | 1 +
> > > > > >  1 file changed, 1 insertion(+)
> > > > > > 
> > > > > > diff --git a/arch/arm64/configs/defconfig b/arch/arm64/configs/defconfig
> > > > > > index 838b4466d6f6..c940437ae1b3 100644
> > > > > > --- a/arch/arm64/configs/defconfig
> > > > > > +++ b/arch/arm64/configs/defconfig
> > > > > > @@ -1575,6 +1575,7 @@ CONFIG_NVMEM_LAYERSCAPE_SFP=m
> > > > > >  CONFIG_NVMEM_MESON_EFUSE=m
> > > > > >  CONFIG_NVMEM_MTK_EFUSE=y
> > > > > >  CONFIG_NVMEM_QCOM_QFPROM=y
> > > > > > +CONFIG_NVMEM_QCOM_SEC_QFPROM=y
> > > > > 
> > > > > Module
> > > > 
> > > > Should not this be inline with what CONFIG_NVMEM_QCOM_QFPROM is having ?
> > > > Either both CONFIG_NVMEM_QCOM_QFPROM and CONFIG_NVMEM_QCOM_SEC_QFPROM
> > > > should be m or both y
> > > 
> > > Looking back in time, CONFIG_NVMEM_QCOM_QFPROM was enabled as built-in
> > > to get TSENS to work (which makes sense, we don't want the CPUs to
> > > burn). What the actual users for NVMEM_QCOM_SEC_QFPROM?
> > 
> > 
> > CONFIG_NVMEM_QCOM_QFPROM and CONFIG_NVMEM_QCOM_SEC_QFPROM are similar driver
> > for same device and only difference is register region lies in secure space
> > for the latter;
> > 
> > Currently, LLCC is the only client for CONFIG_NVMEM_QCOM_SEC_QFPROM, however
> > if someday if the region lies in non-secure space in that case, client
> > started depending on CONFIG_NVMEM_QCOM_QFPROM.
> > 
> > It reminds me, we have not yet put depends on for LLCC on NVMEM_QCOM_SEC_QFPROM
> > doing which is resulting in recursive dependency[1].
> > 
> 
> Kconfig "depends on" describes build time dependencies and LLCC doesn't
> depend on NVMEM_QCOM_SEC_QFPROM.
> 
> It does depend on nvmem_cell_read_u8() being defined, but as
> CONFIG_NVMEM is bool there will either be a proper implementation (=y)
> or a static inline stub (=n) of this function, so we don't need to
> describe this dependency.

I remember nvmem_cell_read_u8() patch., however, got misleaded by some of the
in kernel example like.,

config ARM_QCOM_CPUFREQ_NVMEM
        tristate "Qualcomm nvmem based CPUFreq"
        depends on ARCH_QCOM
        depends on NVMEM_QCOM_QFPROM
        depends on QCOM_SMEM
        select PM_OPP
        help
          This adds the CPUFreq driver for Qualcomm Kryo SoC based boards.

          If in doubt, say N.

Here, ARM_QCOM_CPUFREQ_NVMEM should not depend on NVMEM_QCOM_QFPROM
for compilation, as NVMEM_QCOM_QFPROM is the provider driver while
all the stubs are available and ARM_QCOM_CPUFREQ_NVMEM is consumer here.

-Mukesh
> 
> > It looks we need to select NVMEM_QCOM_SEC_QFPROM from QCOM_LLCC config and that
> > may not need config enablement here.
> > 
> 
> This is a runtime dependency, created through the nvmem framework by the
> relationship defined in DeviceTree. We don't describe those.
> 
> Regards,
> Bjorn
> 
> > [1]
> > fs/sysfs/Kconfig:2:error: recursive dependency detected!
> > fs/sysfs/Kconfig:2:     symbol SYSFS is selected by CONFIGFS_FS
> > fs/configfs/Kconfig:2:  symbol CONFIGFS_FS is selected by GPIO_SIM
> > drivers/gpio/Kconfig:1884:      symbol GPIO_SIM depends on GPIOLIB
> > drivers/gpio/Kconfig:6: symbol GPIOLIB is selected by I2C_MUX_LTC4306
> > drivers/i2c/muxes/Kconfig:47:   symbol I2C_MUX_LTC4306 depends on I2C_MUX
> > drivers/i2c/Kconfig:62: symbol I2C_MUX is selected by DRM_SII902X
> > drivers/gpu/drm/bridge/Kconfig:270:     symbol DRM_SII902X depends on DRM_BRIDGE
> > drivers/gpu/drm/bridge/Kconfig:2:       symbol DRM_BRIDGE is selected by DRM_MSM
> > drivers/gpu/drm/msm/Kconfig:3:  symbol DRM_MSM depends on QCOM_LLCC
> > drivers/soc/qcom/Kconfig:47:    symbol QCOM_LLCC depends on NVMEM_QCOM_SEC_QFPROM
> > drivers/nvmem/Kconfig:230:      symbol NVMEM_QCOM_SEC_QFPROM depends on NVMEM
> > drivers/nvmem/Kconfig:2:        symbol NVMEM is selected by EEPROM_AT24
> > drivers/misc/eeprom/Kconfig:4:  symbol EEPROM_AT24 depends on SYSFS
> > For a resolution refer to Documentation/kbuild/kconfig-language.rst
> > 
> > -Mukesh
> > 
> > > 
> > > > 
> > > > -Mukesh
> > > > > 
> > > > > >  CONFIG_NVMEM_RMEM=m
> > > > > >  CONFIG_NVMEM_ROCKCHIP_EFUSE=y
> > > > > >  CONFIG_NVMEM_ROCKCHIP_OTP=y
> > > > > > --
> > > > > > 2.42.0
> > > > > > 
> > > > > 
> > > > > Best regards,
> > > > > Krzysztof
> > > > > 
> > > 
> > > -- 
> > > With best wishes
> > > Dmitry

Powered by blists - more mailing lists