Re: [PATCH v6 4/4] x86/sev: register tpm-svsm platform device

[Stefano Garzarella]

On Tue, Apr 08, 2025 at 01:28:20PM +0200, Borislav Petkov wrote:
> On Tue, Apr 08, 2025 at 01:08:36PM +0200, Stefano Garzarella wrote:
> > We discussed a bit on v3, but I'm open to change it:
> > https://lore.kernel.org/linux-integrity/nrn4ur66lz2ocbkkjl2bgiex3xbp552szerfhalsaefunqxf7p@ki7xf66zrf6u/
> >
> >  I tried to keep the logic of whether or not the driver is needed all in
> > the tpm_svsm_probe()/snp_svsm_vtpm_probe() (where I check for SVSM).
> >  If you prefer to move some pieces here, though, I'm open.
>
> Yes please.
>
> It doesn't make a whole lotta sense right now to register a TPM platform
> driver at one place without even knowing you're running with an SVSM inside
> the guest blob or not.
>
> The usual approach is to register upon a successful detection.

I see, so IIUC I can just apply the following change to this patch and 
avoid to export snp_svsm_vtpm_probe() at all, right?

diff --git a/arch/x86/coco/sev/core.c b/arch/x86/coco/sev/core.c
index acbd9bc526b1..fa83e6c7f990 100644
--- a/arch/x86/coco/sev/core.c
+++ b/arch/x86/coco/sev/core.c
@@ -2702,8 +2702,10 @@ static int __init snp_init_platform_device(void)
        if (platform_device_register(&sev_guest_device))
                return -ENODEV;
 
-       if (platform_device_register(&tpm_svsm_device))
-               return -ENODEV;
+       if (snp_svsm_vtpm_probe()) {
+               if (platform_device_register(&tpm_svsm_device))
+                       return -ENODEV;
+       }
 
        pr_info("SNP guest platform device initialized.\n");
        return 0;

Thanks,
Stefano