Re: [PATCH] mei: improve Denverton HSM & IFSI support

[Alex Williamson <alex.williamson@redhat.com>]

On Thu, 19 Aug 2021 10:07:03 -0500
Bjorn Helgaas <helgaas@kernel.org> wrote:

> [+cc Alex]
> 
> On Thu, Aug 19, 2021 at 04:51:14PM +0200, Lukas Bulwahn wrote:
> > The Intel Denverton chip provides HSM & IFSI. In order to access
> > HSM & IFSI at the same time, provide two HECI hardware IDs for accessing.
> > 
> > Suggested-by: Ionel-Catalin Mititelu <ionel-catalin.mititelu@intel.com>
> > Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
> > ---
> > Tomas, please pick this quick helpful extension for the hardware.
> > 
> >  drivers/misc/mei/hw-me-regs.h | 3 ++-
> >  drivers/misc/mei/pci-me.c     | 1 +
> >  drivers/pci/quirks.c          | 3 +++
> >  3 files changed, 6 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/misc/mei/hw-me-regs.h b/drivers/misc/mei/hw-me-regs.h
> > index cb34925e10f1..c1c41912bb72 100644
> > --- a/drivers/misc/mei/hw-me-regs.h
> > +++ b/drivers/misc/mei/hw-me-regs.h
> > @@ -68,7 +68,8 @@
> >  #define MEI_DEV_ID_BXT_M      0x1A9A  /* Broxton M */
> >  #define MEI_DEV_ID_APL_I      0x5A9A  /* Apollo Lake I */
> >  
> > -#define MEI_DEV_ID_DNV_IE     0x19E5  /* Denverton IE */
> > +#define MEI_DEV_ID_DNV_IE	0x19E5  /* Denverton for HECI1 - IFSI */
> > +#define MEI_DEV_ID_DNV_IE_2	0x19E6  /* Denverton 2 for HECI2 - HSM */
> >  
> >  #define MEI_DEV_ID_GLK        0x319A  /* Gemini Lake */
> >  
> > diff --git a/drivers/misc/mei/pci-me.c b/drivers/misc/mei/pci-me.c
> > index c3393b383e59..30827cd2a1c2 100644
> > --- a/drivers/misc/mei/pci-me.c
> > +++ b/drivers/misc/mei/pci-me.c
> > @@ -77,6 +77,7 @@ static const struct pci_device_id mei_me_pci_tbl[] = {
> >  	{MEI_PCI_DEVICE(MEI_DEV_ID_APL_I, MEI_ME_PCH8_CFG)},
> >  
> >  	{MEI_PCI_DEVICE(MEI_DEV_ID_DNV_IE, MEI_ME_PCH8_CFG)},
> > +	{MEI_PCI_DEVICE(MEI_DEV_ID_DNV_IE_2, MEI_ME_PCH8_SPS_CFG)},
> >  
> >  	{MEI_PCI_DEVICE(MEI_DEV_ID_GLK, MEI_ME_PCH8_CFG)},
> >  
> > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> > index 6899d6b198af..2ab767ef8469 100644
> > --- a/drivers/pci/quirks.c
> > +++ b/drivers/pci/quirks.c
> > @@ -4842,6 +4842,9 @@ static const struct pci_dev_acs_enabled {
> >  	{ PCI_VENDOR_ID_INTEL, 0x15b7, pci_quirk_mf_endpoint_acs },
> >  	{ PCI_VENDOR_ID_INTEL, 0x15b8, pci_quirk_mf_endpoint_acs },
> >  	{ PCI_VENDOR_ID_INTEL, PCI_ANY_ID, pci_quirk_rciep_acs },
> > +	/* Denverton */
> > +	{ PCI_VENDOR_ID_INTEL, 0x19e5, pci_quirk_mf_endpoint_acs },
> > +	{ PCI_VENDOR_ID_INTEL, 0x19e6, pci_quirk_mf_endpoint_acs },  
> 
> This looks like it should be a separate patch with a commit log that
> explains it.  For example, see these:
> 
>   db2f77e2bd99 ("PCI: Add ACS quirk for Broadcom BCM57414 NIC")
>   3247bd10a450 ("PCI: Add ACS quirk for Intel Root Complex Integrated Endpoints")
>   299bd044a6f3 ("PCI: Add ACS quirk for Zhaoxin Root/Downstream Ports")
>   0325837c51cb ("PCI: Add ACS quirk for Zhaoxin multi-function devices")
>   76e67e9e0f0f ("PCI: Add ACS quirk for Amazon Annapurna Labs root ports")
>   46b2c32df7a4 ("PCI: Add ACS quirk for iProc PAXB")
>   01926f6b321b ("PCI: Add ACS quirk for HXT SD4800")
> 
> It should be acked by somebody at Intel since this quirk relies on
> behavior of the device for VM security.

+1 Thanks Bjorn.  I got curious and AFAICT these functions are the
interface for the host system to communicate with "Innovation Engine"
processors within the SoC, which seem to be available for system
builders to innovate and differentiate system firmware features.  I'm
not sure then how we can assume a specific interface ("HSM" or "IFSI",
whatever those are) for each function, nor of course how we can assume
isolation between them.  Thanks,

Alex