Re: [PATCH v2 2/3] nvmem: Add read-only SpacemiT K3 eFuse support

From: Jiayu Du

Date: Wed Nov 18 2026 - 01:15:53 EST


On Tue, Nov 17, 2026 at 10:09:45PM +0100, Aurelien Jarno wrote:
> Hi,
>
> On 2026-11-16 15:54, Jiayu Du wrote:
> > Add read-only NVMEM access to bank 0 and bank 7 of the SpacemiT K3 eFuse
> > controller. These banks contain temperature calibration parameters and
> > other configuration data.
> >
> > Keep the root key and SSK registers outside the exposed bank ranges.
> >
> > FUSE_READY reports completion of a software-triggered reload. Trigger a
> > reload through GEU_CONFIG bit 22 with a 1 us reset pulse, then poll
> > FUSE_READY at a 1 us interval with a 10 ms timeout before taking the
> > snapshot. Fail probe if the reload times out.
> >
> > Snapshot the shadow registers with aligned 32-bit reads and expose the
> > data in little-endian order at controller-relative offsets. Serve reads
> > from the snapshot and use keepouts for the remaining register ranges.
> >
> > Enable the AES_WTM clock only while accessing the hardware. Leave its
> > shared reset deasserted on teardown because WTM_RST resets the entire GEU.
> >
> > Signed-off-by: Jiayu Du <duge1256@xxxxxxxxx>
> > ---
> > MAINTAINERS | 1 +
> > drivers/nvmem/Kconfig | 11 ++
> > drivers/nvmem/Makefile | 1 +
> > drivers/nvmem/spacemit-k3-efuse.c | 199 ++++++++++++++++++++++++++++++
> > 4 files changed, 212 insertions(+)
> > create mode 100644 drivers/nvmem/spacemit-k3-efuse.c
>
> I have been looking at the code with more details. Is there any reason
> to "expose" the whole 1024-byte efuse address space in the nvmem layout
> (of course with some masking) as opposed to the 16 banks x 32 bytes =
> 4096 bits only (still with masking)?
>
> I find strange that user space has to know the address of bank 0 or bank
> 7 instead of just looking respectively at addresses 0x000 and 0x0e0.
>
> As the layout exposed to user space also includes status or control
> registers (although masked), this might makes difficult to support a
> newer version of the efuse controller which have a similar data layout,
> but slightly different registers colliding with the bank data.
>

Hi Aurelien,

I will use a 512-byte NVMEM space, with bank 0 starting at 0x000
and bank 7 at 0x0e0. The other banks will remain covered by keepouts.

I will also update the temperature calibration cell offsets in the
device tree to 0x1c and 0xea. The hardware accesses will stay unchanged.

Best regards,
Jiayu Du