RE: [PATCH v7 1/3] iommufd: Add data structure for Intel VT-d stage-1 cache invalidation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> From: Jason Gunthorpe <jgg@xxxxxxxxxx>
> Sent: Thursday, January 4, 2024 10:37 PM
> 
> On Thu, Dec 14, 2023 at 07:26:39PM +0800, Yi Liu wrote:
> > Per the prior discussion[1], we agreed to move the error reporting into the
> > driver specific part. On Intel side, we want to report two devTLB
> > invalidation errors: ICE (invalid completion error) and ITE (invalidation
> > timeout error). Such errors have an additional SID information to tell
> > which device failed the devTLB invalidation. I've got the below structure.
> 
> IMHO all of this complexity is a consequence of the decision to hide
> the devtlb invalidation from the VM..
> 
> On the other hand I guess you want to do this because of the SIOV
> troubles where the vPCI function in the VM is entirely virtual and
> can't be trivially mapped to a real PCI function for ATC invalidation
> like ARM and AMD can do (but they also can't support SIOV because of
> this). :(
> 
> However it also makes it very confusing about how the VM would
> perceive an error - eg if it invalidates an SIOV device single PASID
> and that devtlb fails then the error should be connected back to the
> vPCI function for the SIOV's specific PASID and not back to the
> physical PCI function for the SIOV owner.
> 
> As the iommu driver itself has no idea about the vPCI functions this
> seems like it is going to get really confusing. The API I suggested in
> the other email is not entirely going to work as the vPCI function for
> SIOV cases will have to be identified by the (struct device, PASID) -
> while it would be easy enough for the iommu driver to provide the
> PASID, I'm not sure how the iommufd core will relate the PASID back to
> the iommu_device to understand SIOV without actually being aware of
> SIOV to some degree :\

we plan to add such awareness with a new binding helper:

https://lore.kernel.org/lkml/20231009085123.463179-4-yi.l.liu@xxxxxxxxx/

and with metadata to track association between PASID's and iommufd vdev.

but in reality the relation could be identified in an easy way due to a SIOV
restriction which we discussed before - shared PASID space of PF disallows
assigning sibling vdev's to a same VM (otherwise no way to identify which
sibling vdev triggering an iopf when a pasid is used on both vdev's). That
restriction implies that within an iommufd context every iommufd_device
object should contain a unique struct device pointer. So PASID can be
instead ignored in the lookup then just always do iommufd_get_dev_id()
using struct device.





[Index of Archives]     [Linux Wireless]     [Linux Kernel]     [ATH6KL]     [Linux Bluetooth]     [Linux Netdev]     [Kernel Newbies]     [Share Photos]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Samba]     [Device Mapper]

  Powered by Linux