Hi Bjorn, On Sun, Jul 05, 2026 at 05:26:22PM -0500, Bjorn Andersson wrote: > On Fri, Jul 03, 2026 at 05:31:11AM -0700, Ananthu C V wrote: > > Most Qualcomm platforms feature Gunyah hypervisor, which typically > > handles Stage 2 IOMMU configuration. > > This statement is true, but the IOMMU isn't involved in the interaction > between OS and TZ, so it doesn't add value to your problem description. > > > Additionally, SHM bridge setup is required to enable memory protection > > for both remoteproc metadata and its memory regions. > > "Additionally"? Isn't this the entire problem? > > > When the aforementioned hypervisor is absent, the operating system > > must perform these configurations instead. > > Please rephrase these three sentences to make sure we have a clear > problem description. > > Then, at this point in the text you have established the problem, so > break the paragraph with an empty line. > > > We've been relying on the iommu property being present for this, but > > for remoteprocs that are already running like SoCCP the mappings are > > already in place, and any attempt to recreate them while active would > > lead to smmu faults and a non-functional remoteproc. > > This is a significant detail. Please expand the description around > "attempt to recreate them while active" to ensure that it's clear to the > reader what the problem (what does it mean that mappings are already in > place). > > Then another </p><p> here, as we're once again in need of some breathing > and thinking-room. > > > Fix this by adding a needs_tzmem > > flag which ensures tzmem and SHM bridge setup is performed independent to > > the iommu property being present. > > "flag which ensures...", the flag doesn't ensure anything, it is > propagates as an alternative trigger to the PAS helpers to ensure that > the shmbridge is established. > > Regards, > Bjorn
ACK to all, will take in the suggestions and improve the commit message. Thanks for the review. Best, Ananthu
