On Wed, Oct 16, 2024 at 09:02:39AM +0000, Peng Fan wrote: > All, > > > Subject: Re: [PATCH RFC 2/2] iommu/arm-smmu-v3: Bypass SID0 for > > NXP i.MX95 > > Thanks for the discussion on this topic to show much information > that I not foresee. > > > > > On Tue, Oct 15, 2024 at 04:37:25PM +0100, Robin Murphy wrote: > > > On 2024-10-15 4:31 pm, Jason Gunthorpe wrote: > > > > On Tue, Oct 15, 2024 at 04:13:13PM +0100, Robin Murphy wrote: > > > > > On 2024-10-15 1:47 pm, Jason Gunthorpe wrote: > > > > > > On Tue, Oct 15, 2024 at 08:13:28AM +0000, Pranjal Shrivastava > > wrote: > > > > > > > > > > > > > Umm.. this was specific for rmr not a generic thing. I'd > > > > > > > suggest to avoid meddling with the STEs directly for acheiving > > > > > > > bypass. Playing with the iommu domain type could be neater. > > > > > > > Perhaps, modify the > > > > > > > ops->def_domain_type to return an appropriate domain? > > > > > > > > > > > > Yeah, that is the expected way, to force the def_domain_type to > > > > > > IDENTITY and refuse to attach a PAGING/BLOCKED domain. > > > > > > > > > > There is no domain, this is bypassing an arbitrary StreamID not > > > > > associated with any device. > > > > > > > > If the stream ID is going to flow traffic shouldn't it have a DT > > > > node for it? Something must be driving the DMA on this SID, and > > the > > > > kernel does need to know what that is in some way, even for basic > > > > security things like making sure VFIO doesn't get a hold of it :\ > > > > > > Exactly, hence this RFC is definitely not the right approach. > > > > Agreed. I assumed the bypass was needed for a registered SID. > > I just reply here, not reply each thread. Apologies, I responded to the other thread before looking at this one > > The SID is not a registered SID. > > Considering the security things, except "iommus = <&smmu 0>" > being added, is there other method for this issue? I can only think of RMRs if you can ensure/restrict eDMA3 to access a fixed region of memory. Something like a DMA zone if feasible. > > Thanks, > Peng. > > > > > > > > > Thanks, > > > Robin. > > Thanks, Pranjal
