On Tue, Aug 08, 2023 at 02:27:04PM -0300, Jason Gunthorpe wrote: > I missed two paths where __iommu_probe_device() can be called while > already holding the device_lock() for the device that is to be probed. > > This causes a deadlock because __iommu_probe_device() will attempt to > re-acquire the lock. > > Organize things so that these two paths can re-use the existing already > held device_lock by marking the call chains based on if the lock is held > or not. > > This is an incremental on top of Joerg's next, but it could also be handled by > respinning the last patch in that series. Please let me know. The issues this series fixes have been causing quite a bit of breakage in a range of CI systems (Arm's internal stuff, KernelCI and my personal CI). Both the KernelCI bot and my colleague Aishwarya (CCed) bisected which pointed to this series so I've tested them - I didn't cover every board but this does fix at least some boots so: Tested-by: Mark Brown <broonie@xxxxxxxxxx> It'd be great to get these fixes into -next, thanks for getting the patches out so quickly.
Attachment:
signature.asc
Description: PGP signature
