Re: [PATCH v2 3/3] scsi: ufs: core: Add OP-TEE based RPMB driver for UFS devices

[Bean Huo <huobean@gmail.com>]

Jens, 

I incorporated your suggestions in my v3 excpet these two:


On Wed, 2025-10-01 at 09:50 +0200, Jens Wiklander wrote:
> > diff --git a/drivers/ufs/core/Makefile b/drivers/ufs/core/Makefile
> > index cf820fa09a04..51e1867e524e 100644
> > --- a/drivers/ufs/core/Makefile
> > +++ b/drivers/ufs/core/Makefile
> > @@ -2,6 +2,7 @@
> > 
> >   obj-$(CONFIG_SCSI_UFSHCD)              += ufshcd-core.o
> >   ufshcd-core-y                          += ufshcd.o ufs-sysfs.o ufs-mcq.o
> > +ufshcd-core-$(CONFIG_RPMB)             += ufs-rpmb.o
> 
> SCSI_UFSHCD might need the same trick ("depends on RPMB || !RPMB") in
> Kconfig as we have for MMC_BLOCK.
> 
> > 
When RPMB=m and SCSI_UFSHCD=y, the ufs-rpmb.o is compiled into the built-in
ufshcd-core, ufs-rpmb.c calls functions from the OP-TEE RPMB subsystem module,
The kernel allows built-in code to reference module symbols (they become runtime
dependencies, not link-time), please check, I tested.

> > 
> > 
> 
> > +
> > +       struct rpmb_descr descr = {
> > +               .type = RPMB_TYPE_UFS,
> 
> We'll need another type if the device uses the extended RPMB frame
> format. How about you clarify this, where RPMB_TYPE_UFS is defined to
> avoid confusion?

As ufs-bsg.c, we could use ARPMB_TYPE_UFS for UFS advanced RPMB frame, if it is
RPMB, we take it as normal RPMB, the frame should be the same as MMC RPMB.


Kind regards,
Bean