Re: [PATCH v2 2/2] tpm: tpm_crb: enhance resource mapping mechanism for supporting AMD's fTPM
From: Seunghun Han
Date: Tue Sep 10 2019 - 11:28:35 EST
> On Tue, Sep 10, 2019 at 03:42:15PM +0100, Jarkko Sakkinen wrote:
> > On Mon, Sep 09, 2019 at 06:09:06PM +0900, Seunghun Han wrote:
> > > I got an AMD system which had a Ryzen Threadripper 1950X and MSI
> > > mainboard, and I had a problem with AMD's fTPM. My machine showed an error
> > > message below, and the fTPM didn't work because of it.
> > >
> > > [ 5.732084] tpm_crb MSFT0101:00: can't request region for resource
> > > [mem 0x79b4f000-0x79b4ffff]
> > > [ 5.732089] tpm_crb: probe of MSFT0101:00 failed with error -16
> > >
> > > When I saw the iomem, I found two fTPM regions were in the ACPI NVS area.
> > > The regions are below.
> > >
> > > 79a39000-79b6afff : ACPI Non-volatile Storage
> > > 79b4b000-79b4bfff : MSFT0101:00
> > > 79b4f000-79b4ffff : MSFT0101:00
> > >
> > > After analyzing this issue, I found that crb_map_io() function called
> > > devm_ioremap_resource() and it failed. The ACPI NVS didn't allow the TPM
> > > CRB driver to assign a resource in it because a busy bit was set to
> > > the ACPI NVS area.
> > >
> > > To support AMD's fTPM, I added a function to check intersects between
> > > the TPM region and ACPI NVS before it mapped the region. If some
> > > intersects are detected, the function just calls devm_ioremap() for
> > > a workaround. If there is no intersect, it calls devm_ioremap_resource().
> > >
> > > Signed-off-by: Seunghun Han <kkamagui@xxxxxxxxx>
> > This problem is still valid and not addressed by Vanya's patch (and
> > should not be as it is a disjoint issue). However, calling
> > devm_ioremap() is somewhat racy as the NVS driver is not aware of that.
> > My take is that this should be fixed in the code that assigns regions to
> > the NVS driver e.g. it could look up the regions assigned to the
> > MSFT0101 and ignore those regions. In the end linux-acpi maintainers
> > have the say on this but this would be the angle that I'd take to
> > implement such patch probably.
> Matthew pointed out that having a hook in NVS driver is better solution
> because it is nil functionality if the TPM driver is loaded. We need
> functions to:
> 1. Request a region from the NVS driver (when tpm_crb loads)
> 2. Release a region back to the NVS Driver (when tpm_crb unloads).
> My proposal would unnecessarily duplicate code and also leave a
> side-effect when TPM is not used in the first place.
> I see this as the overally best solution. If you can come up with a
> patch for the NVS side and changes to CRB drivers to utilize the new
> hooks, then combined with Vanya's changes we have a sustainable solution
> for AMD fTPM.
It's a great solution. I will update this patch on your advice and
send it to you soon.
By the way, I have a question about your advice.
If we handle the NVS region with NVS driver, calling devm_ioremap()
function is fine like crb_ioremap_resource() function in this patch?