Re: [PATCH v3 3/7] tpm: Validate the eventlog access before tpm_bios_log_setup

[Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>]

On Fri, Sep 09, 2016 at 10:54:25PM +0530, Nayna wrote:
> 
> 
> On 08/30/2016 11:22 PM, Jason Gunthorpe wrote:
> >On Tue, Aug 30, 2016 at 12:50:15AM -0400, Nayna Jain wrote:
> >>@@ -382,6 +370,8 @@ int tpm_chip_register(struct tpm_chip *chip)
> >>  		return rc;
> >>  	}
> >>
> >>+	tpm_bios_log_setup(chip);
> >
> >Surely this can fail, right? At least if the security fs setup fails
> >this should propogate that error.
> 
> What action we want to take if it fails to do bios_log_setup ?
> I have done all other fixes, just am not sure that if we propogate this
> error, then will it mean that tpm_chip_register (where this function is
> called) should fail ? or it is just an error logging on failure of
> bios_log_setup.

Typically we'd want to fail probe, so keep flowing it up.

I'm expecting this should only happen in fatal cases like security_fs
functions failing.

Failures like no optional ACPI/DT description should not log or return
an error up to probe..

Jason

------------------------------------------------------------------------------