Re: [PATCH v4 05/16] iommu: Move bus setup to IOMMU device registration

[Baolu Lu]

Hi Alex,

On 2022/10/13 0:28, Alex Williamson wrote:
> On Mon, 15 Aug 2022 17:20:06 +0100
> Robin Murphy<robin.murphy@xxxxxxx>  wrote:
>
> > Move the bus setup to iommu_device_register(). This should allow
> > bus_iommu_probe() to be correctly replayed for multiple IOMMU instances,
> > and leaves bus_set_iommu() as a glorified no-op to be cleaned up next.
> >
> > At this point we can also handle cleanup better than just rolling back
> > the most-recently-touched bus upon failure - which may release devices
> > owned by other already-registered instances, and still leave devices on
> > other buses with dangling pointers to the failed instance. Now it's easy
> > to clean up the exact footprint of a given instance, no more, no less.
> >
> > Tested-by: Marek Szyprowski<m.szyprowski@xxxxxxxxxxx>
> > Reviewed-By: Krishna Reddy<vdumpa@xxxxxxxxxx>
> > Reviewed-by: Kevin Tian<kevin.tian@xxxxxxxxx>
> > Tested-by: Matthew Rosato<mjrosato@xxxxxxxxxxxxx>  # s390
> > Tested-by: Niklas Schnelle<schnelle@xxxxxxxxxxxxx>  # s390
> > Signed-off-by: Robin Murphy<robin.murphy@xxxxxxx>
> > ---
> >
> > v4: Factor out the ops check in iommu_device_register() to keep the loop
> > even simpler, and comment the nominal change in behaviour
> >
> >   drivers/iommu/iommu.c | 55 +++++++++++++++++++++++--------------------
> >   1 file changed, 30 insertions(+), 25 deletions(-)
> This introduces the below lockdep spat regression, bisected to commit:
>
> 57365a04c921 ("iommu: Move bus setup to IOMMU device registration")
>
> This can be reproduced with simple vfio-pci device assignment to a VM
> on x86_64 with VT-d.  Thanks,

Thank you for reporting this. I have proposed below fix:

https://lore.kernel.org/all/20220927053109.4053662-1-baolu.lu@xxxxxxxxxxxxxxx/

Does it work for you?

Best regards,
baolu