Re: [PATCH v2 6/6] ufs: core: Add advanced RPMB support in ufs_bsg

[Bean Huo <beanhuo@iokpp.de>]

On Tue, 2022-11-22 at 11:55 +0000, Avri Altman wrote:
> > +static int ufs_bsg_exec_advanced_rpmb_req(struct ufs_hba *hba,
> > struct
> > +bsg_job *job) {
> > +       struct ufs_rpmb_request *rpmb_request = job->request;
> > +       struct ufs_rpmb_reply *rpmb_reply = job->reply;
> > +       struct bsg_buffer *payload = NULL;
> > +       enum dma_data_direction dir;
> > +       struct scatterlist *sg_list;
> > +       int rpmb_req_type;
> > +       int sg_cnt;
> > +       int ret;
> > +       int data_len;
> > +
> > +       if (hba->ufs_version < ufshci_version(4, 0) || !hba-
> > > dev_info.b_advanced_rpmb_en ||
> > +           !(hba->capabilities & MASK_EHSLUTRD_SUPPORTED))
> > +               return -EINVAL;
> > +
> > +       if (rpmb_request->ehs_req.length != 2 || rpmb_request-
> > > ehs_req.ehs_type != 1)
> > +               return -EINVAL;
> Maybe you could also check:
> In case of rpmb write, the request payload 4096 × Advanced RPMB Block
> Count,
> And same goes for response payload for rpmb read.
> 
> Thanks,
> Avri
> 

Hi Avri, 

in Spec:

"If the Block Count indicates a value greater than bRPMB_ReadWriteSize,
then the authenticated data write/read operation fails and the Result
is set to “General failure” (0001h)."


I think this should be checked in the application in userspace because
if the application passes a wrong/incorrect payload length, it will
error out and have no effect on UFS. In order to add this check in a
driver in the kernel, we need to maintain bRPMB_ReadWriteSize in kernel
space. Sometimes this is a waste of resources because we don't know if
the client will access the RPMB.

I have enabled Advanced RPMB feature in the ufs-utils as an example,
will be refered in cover-letter in the next version.

Kind regards,
Bean