Re: 5.19.9 regression: DMAR error on i7-8650U

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Mon, Sep 19, 2022 at 08:27:20AM +0300, Mika Westerberg wrote:
> Hi George,
> 
> On Mon, Sep 19, 2022 at 08:53:48AM +0800, Baolu Lu wrote:
> > On 9/18/22 8:00 AM, George Hilliard wrote:
> > > I'm experiencing a regression on 5.19.9 from 5.19.8. Not sure if you are
> > > already aware of this, apologies if it's noise.  My hunch is that this
> > > is an iommu problem since there were quite a few iommu changes in .9,
> > > but I have also Cc'd Firewire maintainers on the chance that I'm wrong.
> > > 
> > > My Apple Thunderbolt to Firewire adapter has quit working on 5.19.9 on
> > > my Thinkpad T480s (i7-8650U), with the following log message while
> > > enumerating the device (reformatted for email):
> > > 
> > >      [  318.263546] thunderbolt 0-1: new device found, vendor=0x1 device=0x8004
> > >      [  318.263559] thunderbolt 0-1: Apple, Inc. Thunderbolt to FireWire Adapter
> > >      [  318.342416] pcieport 0000:05:01.0: pciehp: Slot(1): Card present
> > >      [  318.342428] pcieport 0000:05:01.0: pciehp: Slot(1): Link Up
> > >      (snip)
> > >      [  318.503177] pcieport 0000:08:00.0: enabling device (0000 -> 0003)
> > >      [  318.504976] firewire_ohci 0000:09:00.0: enabling device (0000 -> 0002)
> > >      [  318.562000] firewire_ohci 0000:09:00.0: added OHCI v1.10 device
> > >                     as card 4, 8 IR + 8 IT contexts, quirks 0x0, physUB
> > >      [  318.562109] DMAR: DRHD: handling fault status reg 2
> > >      [  318.562114] DMAR: [INTR-REMAP] Request device [09:00.0] fault
> > >                     index 0x8080 [fault reason 0x25] Blocked a
> > >                     compatibility format interrupt request
> > > 
> > > The last two lines are the new error.
> > > 
> > > Happy to provide more details or experiment; this happens nearly every time.
> > > 
> > > George
> > > 
> > 
> > + Mika.
> 
> Since this is a regression, I wonder if you are able to bisect it to the
> commit that caused the issue? As you say there are a couple of IOMMU
> related changes between v5.19.8 and v5.19.9 so one option is to revert
> them one by one and see if the problem goes away:
> 
> f7f13b8d49bf3eff2c98230532e0671195277216 swiotlb: avoid potential left shift overflow
> 9516acba29e322202674d18f4dc383879f7813a5 iommu/vt-d: Fix possible recursive locking in intel_iommu_init()
> 420d2045715588ad44cd7d806864a223b4fc05c9 iommu/vt-d: Correctly calculate sagaw value of IOMMU
> 77a2d61334f715293c20f4fa30df035fd870fba4 iommu/virtio: Fix interaction with VFIO
> 32e4b8dbc9119c07ea1888e72c8a5681b82e180c iommu: Fix false ownership failure on AMD systems with PASID activated

Here's bugzilla report that looks to have similar symptoms:

https://bugzilla.kernel.org/show_bug.cgi?id=216497