On Thu, Dec 14, 2017 at 05:12:59AM -0800, John Strader wrote:
> Submitting a device for quirk_broken_intx_masking, as requested by
> modinfo for vfio-pci "nointxmask" parameter.
>
> The device is a Ceton InfiniTV4 [1b7c:0004]. I can confirm that this
> device operates correctly with vfio when passed the nointxmask
> parameter. I can also confirm that applying quirk_broken_intx_masking
> via drivers/pci/quirks.c in a custom compiled kernel also works. I no
> longer compile the custom kernel and seem to have misplaced the patch,
> so I won't bother remaking a one-liner, but it was of the standard form:
>
> DECLARE_PCI_FIXUP_FINAL(0x1b7c, 0x0004,
> quirk_broken_intx_masking);
Thanks for your report! I applied the patch below to pci/virtualization
for v4.16.
commit 7f50dcad8de026d7533a0d1295a27d843ce965aa
Author: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
Date: Fri Dec 15 14:51:44 2017 -0600
PCI: Mark Ceton InfiniTV4 INTx masking as broken
PCI_COMMAND_INTX_DISABLE is writable on the Ceton InfiniTV4, indicating
that the device supports disabling the INTx# signal, but it apparently
doesn't work.
Mark the device so we know we can't use PCI_COMMAND_INTX_DISABLE to disable
its interrupts.
Link: https://lkml.kernel.org/r/92a65068-60b2-c1a8-9e17-ac41fe3c5c93@xxxxxxxxxxxxxxx
Reported-by: John Strader <strader.john@xxxxxxxxxxxxxxx>
Signed-off-by: Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
index 10684b17d0bd..bb9737ec64b5 100644
--- a/drivers/pci/quirks.c
+++ b/drivers/pci/quirks.c
@@ -3215,6 +3215,8 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_CHELSIO, 0x0030,
quirk_broken_intx_masking);
DECLARE_PCI_FIXUP_FINAL(0x1814, 0x0601, /* Ralink RT2800 802.11n PCI */
quirk_broken_intx_masking);
+DECLARE_PCI_FIXUP_FINAL(0x1b7c, 0x0004, /* Ceton InfiniTV4 */
+ quirk_broken_intx_masking);
/*
* Realtek RTL8169 PCI Gigabit Ethernet Controller (rev 10)
