On Mon, Nov 25, 2019 at 08:45:52AM +0100, Jan Kiszka wrote:
> On 28.10.19 11:55, Michael S. Tsirkin wrote:
> > Vendors might want to add their own capability
> > in the PCI capability list. However, Virtio already
> > uses the vendor specific capability ID (0x09)
> > for its own purposes.
>
> Did some vendor express that need, or do we only assume it so far? IOW: Do
> we know at least once concrete use case?
Good point, I should have added this in the log.
I know of a device that implements virtio and has a bug.
Device is a transitional one so can not have vendor specific
subsystem IDs (that violates the SHOULD below. Do we want to qualify
that this recommendation is for non-transitional devices?). While that
specific device can't be fixed like this on bare metal, it creates a consistent way
for future devices to trigger vendor/device specific quirks,
and it creates an option for the hypervisor to add the capability.
> >
> > Provide a structure for vendor specific extensions.
> >
> > Signed-off-by: Michael S. Tsirkin <>
> > ---
> > content.tex | 80 +++++++++++++++++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 80 insertions(+)
> >
> > diff --git a/content.tex b/content.tex
> > index b1ea9b9..8a79b03 100644
> > --- a/content.tex
> > +++ b/content.tex
> > @@ -691,6 +691,8 @@ \subsection{Virtio Structure PCI Capabilities}\label{sec:Virtio Transport Option
> > #define VIRTIO_PCI_CAP_PCI_CFG 5
> > /* Shared memory region */
> > #define VIRTIO_PCI_CAP_SHARED_MEMORY_CFG 8
> > +/* Vendor-specific data */
> > +#define VIRTIO_PCI_CAP_VENDOR_CFG 9
> > \end{lstlisting}
> > Any other value is reserved for future use.
> > @@ -1099,6 +1101,84 @@ \subsubsection{Shared memory capability}\label{sec:Virtio Transport Options / Vi
> > The \field{cap.id} MUST be unique for any one device instance.
> > +\devicenormative{\paragraph}{Device-specific configuration}{Virtio Transport Options / Virtio Over PCI Bus / PCI Device Layout / Shared memory capability}
> > +
> > +The region defined by the combination of the \field {cap.offset},
> > +\field {cap.offset_hi}, and \field {cap.length}, \field
> > +{cap.length_hi} fields MUST be contained within the declared bar.
> > +
> > +The \field{cap.id} MUST be unique for any one device instance.
> > +
> > +\subsubsection{Vendor data capability}\label{sec:Virtio
> > +Transport Options / Virtio Over PCI Bus / PCI Device Layout /
> > +Vendor data capability}
> > +
> > +The optional Vendor data capability allows the device to present
> > +vendor-specific data to the driver, without
> > +conflicts, for debugging and/or reporting purposes,
> > +and without conflicting with standard functionality.
> > +
> > +This capability augments but does not replace the standard
> > +subsystem ID and subsystem vendor ID fields
> > +(offsets 0x2C and 0x2E in the PCI configuration space header)
> > +as specified by \hyperref[intro:PCI]{[PCI]}.
> > +
> > +Vendor data capability is enumerated on the PCI transport
> > +as a VIRTIO_PCI_CAP_VENDOR_CFG capability.
> > +
> > +The capability has the following structure:
> > +\begin{lstlisting}
> > +struct virtio_pci_vndr_data {
> > + u8 cap_vndr; /* Generic PCI field: PCI_CAP_ID_VNDR */
> > + u8 cap_next; /* Generic PCI field: next ptr. */
> > + u8 cap_len; /* Generic PCI field: capability length */
> > + u8 cfg_type; /* Identifies the structure. */
> > + u16 vendor_id; /* Identifies the vendor-specific format. */
> > + /* For Vendor Definition */
> > + /* Pads structure to a multiple of 4 bytes */
> > + /* Reads must not have side effects */
> > +};
> > +\end{lstlisting}
> > +
> > +Where \field{vendor_id} identifies the PCI-SIG assigned Vendor ID
> > +as specified by \hyperref[intro:PCI]{[PCI]}.
> > +
> > +Note that the capability size is required to be a multiple of 4.
> > +
> > +To make it safe for a generic driver to access the capability,
> > +reads from this capability MUST NOT have any side effects.
> > +
> > +\devicenormative{\subsection}{Vendor data capability}{Virtio
> > +Transport Options / Virtio Over PCI Bus / PCI Device Layout /
> > +Vendor data capability}
> > +
> > +The device SHOULD present the PCI subsystem vendor ID matching
> > +the device vendor, at offset 0x2C in its PCI configuration space
> > +header.
> > +
> > +Devices CAN present \field{vendor_id} that does not match
> > +either the PCI Vendor ID or the PCI Subsystem Vendor ID.
> > +
> > +Devices CAN present multiple Vendor data capabilities with
> > +either different or identical \field{vendor_id} values.
> > +
> > +The value \field{vendor_id} MUST NOT equal 0x1AF4.
> > +
> > +The size of the Vendor data capability MUST be a multiple of 4 bytes.
> > +
> > +Reads of the Vendor data capability by the driver MUST NOT have any
> > +side effects.
> > +
> > +\drivernormative{\subsection}{Vendor data capability}{Virtio
> > +Transport Options / Virtio Over PCI Bus / PCI Device Layout /
> > +Vendor data capability}
> > +
> > +The driver SHOULD NOT use the Vendor data capability except
> > +for debugging and reporting purposes.
>
> Every vendor is free to mess up the device behavior by adding more to this
> than the spec defines. Still, I'm undecided if this shouldn't be stronger.
>
> Jan
>
> > +
> > +The driver MUST qualify the \field{vendor_id} before
> > +interpreting or writing into the Vendor data capability.
> > +
> > \subsubsection{PCI configuration access capability}\label{sec:Virtio Transport Options / Virtio Over PCI Bus / PCI Device Layout / PCI configuration access capability}
> > The VIRTIO_PCI_CAP_PCI_CFG capability
> >
>
> --
> Siemens AG, Corporate Technology, CT RDA IOT SES-DE
> Corporate Competence Center Embedded Linux
