On Wed, Jan 15, 2020 at 04:42:56PM +0000, Will Deacon wrote: > On Wed, Jan 15, 2020 at 01:52:26PM +0100, Jean-Philippe Brucker wrote: > > Since v4 [1] I addressed some of Will's comment. > > Thanks! I've tentatively managed to queue all of this apart from the last > patch, since that relies on some pasid symbols being exported from the PCI > core when building the driver as a module with PCI_PASID=y. > > Please can you look at the rest of things here?: > > https://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git/log/?h=for-joerg/arm-smmu/updates Sorry for the delay, I had to rebase and rework the SVA patchset a little in order to test. Your changes look good to me and run fine on the hisilicon KunPeng920 (although I haven't tried building as a module yet, I'll do that next). Thanks, Jean > > > Still missing and will be submitted as follow-up patches: > > * write STE.V with WRITE_ONCE() (patch 7) > > I've hacked that one up myself. > > > * batch submission of CD invalidation (patch 7) > > That can be 5.7 material. > > > * Remove WARN_ON_ONCE() in add_device() (patch 13) > > Pending Robin's input. > > I've written that one too, but we'll see what Robin says. The question now > is which commit do I tag in the branch above :) > > Will
