On Tue, Jan 21, 2025 at 01:16:05PM +0000, Saladi.Rakeshbabu@xxxxxxxxxxxxx wrote: > On Mon, 2025-01-20 at 16:56 -0600, Bjorn Helgaas wrote: > > [Some people who received this message don't often get email from > > helgaas@xxxxxxxxxx. Learn why this is important at > > https://aka.ms/LearnAboutSenderIdentification ;] > > > > EXTERNAL EMAIL: Do not click links or open attachments unless you > > know the content is safe > > > > On Mon, Jan 20, 2025 at 03:25:24PM +0530, Rakesh Babu Saladi wrote: > > > Add the Microchip Parts to the existing device ID > > > table so that the driver supports PCI100x devices too. > > > > > > Add a new macro to quirk the Microchip switchtec PCI100x parts > > > to allow DMA access via NTB to work when the IOMMU is turned on. > > > > > > PCI100x family has 6 variants, each variant is designed for > > > different > > > application usages, different port counts and lane counts. > > > > > > PCI1001 has 1 x4 upstream port and 3 x4 downstream ports. > > > PCI1002 has 1 x4 upstream port and 4 x2 downstream ports. > > > PCI1003 has 2 x4 upstream ports, 2 x2 upstream ports and 2 x2 > > > downstream ports. > > > PCI1004 has 4 x4 upstream ports. > > > PCI1005 has 1 x4 upstream port and 6 x2 downstream ports. > > > PCI1006 has 6 x2 upstream ports and 2 x2 downstream ports. > > > > > > Signed-off-by: Rakesh Babu Saladi <Saladi.Rakeshbabu@xxxxxxxxxxxxx> > > > --- > > > drivers/pci/quirks.c | 11 +++++++++++ > > > drivers/pci/switch/switchtec.c | 26 ++++++++++++++++++++++++++ > > > 2 files changed, 37 insertions(+) > > > > > > diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c > > > index eeec1d6f9023..266ab5f8c6e1 100644 > > > --- a/drivers/pci/quirks.c > > > +++ b/drivers/pci/quirks.c > > > @@ -5906,6 +5906,17 @@ SWITCHTEC_QUIRK(0x5552); /* PAXA 52XG5 */ > > > SWITCHTEC_QUIRK(0x5536); /* PAXA 36XG5 */ > > > SWITCHTEC_QUIRK(0x5528); /* PAXA 28XG5 */ > > > > > > +#define SWITCHTEC_PCI100X_QUIRK(vid) \ > > > + DECLARE_PCI_FIXUP_CLASS_FINAL(PCI_VENDOR_ID_EFAR, vid, \ > > > + PCI_CLASS_BRIDGE_OTHER, 8, > > > quirk_switchtec_ntb_dma_alias) > > > +SWITCHTEC_PCI100X_QUIRK(0x1001); /* PCI1001XG4 */ > > > +SWITCHTEC_PCI100X_QUIRK(0x1002); /* PCI1002XG4 */ > > > +SWITCHTEC_PCI100X_QUIRK(0x1003); /* PCI1003XG4 */ > > > +SWITCHTEC_PCI100X_QUIRK(0x1004); /* PCI1004XG4 */ > > > +SWITCHTEC_PCI100X_QUIRK(0x1005); /* PCI1005XG4 */ > > > +SWITCHTEC_PCI100X_QUIRK(0x1006); /* PCI1006XG4 */ > > > + > > > + > > > /* > > > * The PLX NTB uses devfn proxy IDs to move TLPs between NT > > > endpoints. > > > * These IDs are used to forward responses to the originator on > > > the other > > > diff --git a/drivers/pci/switch/switchtec.c > > > b/drivers/pci/switch/switchtec.c > > > index 5b921387eca6..faaca76407c8 100644 > > > --- a/drivers/pci/switch/switchtec.c > > > +++ b/drivers/pci/switch/switchtec.c > > > @@ -1726,6 +1726,26 @@ static void switchtec_pci_remove(struct > > > pci_dev *pdev) > > > .driver_data = gen, \ > > > } > > > > > > +#define SWITCHTEC_PCI100X_DEVICE(device_id, gen) \ > > > + { \ > > > + .vendor = PCI_VENDOR_ID_EFAR, \ > > > + .device = device_id, \ > > > + .subvendor = PCI_ANY_ID, \ > > > + .subdevice = PCI_ANY_ID, \ > > > + .class = (PCI_CLASS_MEMORY_OTHER << 8), \ > > > + .class_mask = 0xFFFFFFFF, \ > > > + .driver_data = gen, \ > > > + }, \ > > > + { \ > > > + .vendor = PCI_VENDOR_ID_EFAR, \ > > > + .device = device_id, \ > > > + .subvendor = PCI_ANY_ID, \ > > > + .subdevice = PCI_ANY_ID, \ > > > + .class = (PCI_CLASS_BRIDGE_OTHER << 8), \ > > > + .class_mask = 0xFFFFFFFF, \ > > > + .driver_data = gen, \ > > > > We have this: > > > > #define PCI_VENDOR_ID_EFAR 0x1055 > > > > but searching the PCI-SIG Vendor ID database for 0x1055 doesn't find > > anything: > > > > https://pcisig.com/membership/member-companies > > > > You mention "Microchip", and it looks like Microchip is assigned > > Vendor ID 0x11f8: > > > > https://pcisig.com/membership/member-companies?combine=microchip > > > > which we also have defined: > > > > #define PCI_VENDOR_ID_PMC_Sierra 0x11f8 > > #define PCI_VENDOR_ID_MICROSEMI 0x11f8 > > > > Can you clarify what's going on here? I assume these parts actually > > do have Vendor ID 0x1055. But the PCI-SIG really should know about > > the usage of this ID. There's an email contact address on that web > > page; can you straighten this out? > > PMC, Microsemi are now part of Microchip. The switchtec products are > from Microsemi and so the vendor ID is 0x11f8. EFAR is also part of > Microchip. For PCI100x product parts we intended to use the VID 0x1055 > which is of EFAR's VID. PMC, Microsemi and EFAR product parts all are > now part of Microchip. > > The reason why EFAR VID (0x1055) is not listed in the website is > becuase the site can list only one VID per company. > > For further information, please refer to Steen's reply of > https://lore.kernel.org/linux-pci/f695618054232c5f43c2148c5e6551f3ab318792.camel@xxxxxxxxxxxxx/to > https://lore.kernel.org/linux-pci/20240621184923.GA1398370@bhelgaas/ Thanks. What an annoyance that the PCI-SIG can't expose Vendor IDs that are allocated but no longer correspond to the original vendor. I'll complain to the SIG again. Bjorn
