On Mon, 17 Oct 2022 at 16:11, Thierry Reding <thierry.reding@xxxxxxxxx> wrote: > > On Thu, Oct 13, 2022 at 03:43:18PM +0200, Ulf Hansson wrote: > > On Thu, 13 Oct 2022 at 08:33, Prathamesh Shete <pshete@xxxxxxxxxx> wrote: > > > > > > Hi Ulf, > > > > > > >> In that case, perhaps we can add a "depends on IOMMU_API" in the Kconfig > > > >> instead? Or is the tegra driver used on platforms where IOMMU_API could be > > > >> unset? > > > Yes it can/will work with IOMMU disabled so its not recommended to add a "depends on" condition in Kconfig. > > > > Alright, in that case it looks to me that there are two other options > > to move forward. > > > > 1) Add proper definitions of the struct iommu_fwspec in > > include/linux/iommu.h even when CONFIG_IOMMU_API is unset. In a way it > > seems a bit silly to me, to have the iommu stubs around, unless those > > can be used for cases like this, right!? > > I recall that I had proposed a patch for this a long time ago: > > https://lore.kernel.org/all/20191209120005.2254786-3-thierry.reding@xxxxxxxxx/ That looks exactly what we would need! > > Given that Joerg had acked it at the time, I think the only reason why > it never ended up getting merged is because the rest of the series did > not get enough traction. I wonder if I should peel it out of the series > and propose it separately. Yes, please. > > I agree it doesn't make any sense to have the stubs to allow compilation > and then break compilation because users of the stubs will end up > wanting to dereference the structure. Thanks for sharing your thoughts around this! If you submit a new version of the old patch, I would certainly give it my blessing. Kind regards Uffe