On Thu, Jun 11, 2026 at 3:35 AM Christian König <[email protected]> wrote: > > > > On 6/10/26 21:31, Zhiping Zhang wrote: > > Add an optional dma_buf_ops.get_tph callback and a dma_buf_get_tph() > > wrapper for importers. > > > > 8-bit ST and 16-bit Extended ST are distinct PCIe TPH namespaces, so > > the importer requests the namespace it can emit and the exporter > > returns the matching ST/PH tuple or -EOPNOTSUPP. > > > > dma_buf_get_tph() is the importer entry point. It returns -EOPNOTSUPP > > when the exporter lacks the callback and requires dmabuf->resv to be > > held while the callback runs. > > > > The first user is VFIO_DEVICE_FEATURE_DMA_BUF_TPH in vfio-pci, with > > mlx5 as the first importer. > > > > Signed-off-by: Zhiping Zhang <[email protected]> > > --- > > drivers/dma-buf/dma-buf.c | 25 +++++++++++++++++++++++++ > > include/linux/dma-buf.h | 21 +++++++++++++++++++++ > > 2 files changed, 46 insertions(+) > > > > diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c > > index d504c636dc29..aff79ea12e43 100644 > > --- a/drivers/dma-buf/dma-buf.c > > +++ b/drivers/dma-buf/dma-buf.c > > @@ -1144,6 +1144,31 @@ void dma_buf_unpin(struct dma_buf_attachment *attach) > > } > > EXPORT_SYMBOL_NS_GPL(dma_buf_unpin, "DMA_BUF"); > > > > +/** > > + * dma_buf_get_tph - Retrieve TPH metadata from an exporter > > + * @dmabuf: DMA buffer to query > > + * @extended: false for 8-bit ST, true for 16-bit Extended ST > > + * @steering_tag: returns the raw steering tag for the requested namespace > > + * @ph: returns the TPH processing hint > > + * > > + * Wrapper for the optional &dma_buf_ops.get_tph callback. > > + * > > + * Must be called with &dma_buf.resv held. Returns -EOPNOTSUPP if the > > + * exporter does not implement the callback or has no metadata for the > > + * requested namespace. > > + */ > > +int dma_buf_get_tph(struct dma_buf *dmabuf, bool extended, > > + u16 *steering_tag, u8 *ph) > > That name needs improvement, maybe something like dma_buf_get_pci_tph(). > > It also needs some brief explanation what TPH is, maybe a reference to the > PCIe spec name etc... > > And document in the list of functions that this one should be called with the > lock held. > > > +{ > > + dma_resv_assert_held(dmabuf->resv); > > + > > + if (!dmabuf->ops->get_tph) > > + return -EOPNOTSUPP; > > + > > + return dmabuf->ops->get_tph(dmabuf, extended, steering_tag, ph); > > +} > > +EXPORT_SYMBOL_NS_GPL(dma_buf_get_tph, "DMA_BUF"); > > + > > /** > > * dma_buf_map_attachment - Returns the scatterlist table of the > > attachment; > > * mapped into _device_ address space. Is a wrapper for map_dma_buf() of > > the > > diff --git a/include/linux/dma-buf.h b/include/linux/dma-buf.h > > index d1203da56fc5..6a54e0f251a2 100644 > > --- a/include/linux/dma-buf.h > > +++ b/include/linux/dma-buf.h > > @@ -113,6 +113,25 @@ struct dma_buf_ops { > > */ > > void (*unpin)(struct dma_buf_attachment *attach); > > > > + /** > > + * @get_tph: > > + * @dmabuf: DMA buffer for which to retrieve TPH metadata > > + * @extended: false for 8-bit ST, true for 16-bit Extended ST > > + * @steering_tag: Returns the raw TPH steering tag for the requested > > + * namespace > > + * @ph: Returns the TPH processing hint (2-bit value) > > + * > > + * Return TPH metadata for the namespace selected by @extended. Return > > + * 0 on success, or -EOPNOTSUPP if no metadata is available. > > + * > > + * This callback is optional. Importers must not call it directly; > > + * the dma_buf_get_tph() wrapper is the only entry point and handles > > + * the NULL-callback case. The callback is invoked with > > + * &dma_buf.resv held. > > That most of that should be obvious, we only need that it's optional and that > the lock should be held. Everything else can be dropped. > > And most of the description/documentation should be on the wrapper function, > exporters who implement the callback should know what they are doing. > > Regards, > Christian. >
sure will do Thanks, Zhiping
