Re: [PATCH v9 7/7] PCI: endpoint: pci-ep-msi: Add embedded doorbell fallback

[Koichiro Den]

On Fri, Feb 27, 2026 at 04:34:02PM +0100, Niklas Cassel wrote:
> On Sat, Feb 21, 2026 at 02:42:35AM +0900, Koichiro Den wrote:
> > On second thought, I'm wondering whether it makes sense to handle the case where
> > the embedded doorbell target resides behind an IOMMU in this series.
> > 
> > In v9, we simply expose the raw physical address without establishing an IOMMU
> > mapping. When the EPC parent device is attached to an IOMMU domain, a Host->EP
> > MMIO write through the BAR window may result in an IOMMU fault.
> > 
> > Initially, I planned to submit IOMMU support separately as a follow-up series
> > once this series is accepted, to avoid making this series too large [1].
> >
> > Niklas, any comments would be appreciated.
> 
> These patches are for NTB/vNTB:
> 
> > [1] Supporting such an IOMMU-backed case would likely require additional
> >     patches for vNTB + ntb_transport to demonstrate usability, such as:
> >     https://lore.kernel.org/all/20260118135440.1958279-12-den@xxxxxxxxxxxxx/
> >     https://lore.kernel.org/all/20260118135440.1958279-16-den@xxxxxxxxxxxxx/
> >     https://lore.kernel.org/all/20260118135440.1958279-19-den@xxxxxxxxxxxxx/
> 
> If the diff you pasted above is all you need for pci-epf-test to pass the
> doorbell test case when running against a PCI endpoint with IOMMU enabled,
> then I suggest you just send a v10 with those changes included, without
> any additional dependencies.

Thanks for the comment, Niklas.

The three patches mentioned above are unrelated to whether pci-epf-test passes
when running against a PCI endpoint with IOMMU enabled. From that perspective,
the code snippet I pasted in the same comment (which you trimmed here) is the
only relevant change.

I'll prepare a v10 with that snippet (polished up as needed) squashed into Patch
7. Before doing so, I'll also take a closer look at your v2 series to see if I
can help move it forward.

> 
> I understand that you would need additional dependencies for vNTB to be able
> to use eDMA doorbell on a PCI endpoint with IOMMU enabled.
> 
> Perhaps, drop the vNTB changes from this series, focus only on pci-epf-test,
> get this series merged, and then submit an follow up series that adds eDMA
> doorbells with IOMMU support for vNTB?

The vNTB changes already included in this series (i.e. [PATCH v9 5/7]) are
necessary. Without them, vNTB can break because the doorbell fallback can occur
behind pci_epf_alloc_doorbell().

Regarding the three patches mentioned above, since I included the vNTB ping
performance improvement in the cover letter, I thought it made sense to consider
those three patches as prerequisites for this series. However, as you suggest,
it makes sense to treat them as a follow-up series.

Best regards,
Koichiro

> 
> 
> Kind regards,
> Niklas