On 29 September 2016 at 17:38, Peter Maydell <peter.mayd...@linaro.org> wrote: > On 23 September 2016 at 00:43, Eric Auger <eric.au...@redhat.com> wrote: >> From: Pavel Fedin <p.fe...@samsung.com> >> >> The ITS control frame is in-kernel emulated while accesses to the >> GITS_TRANSLATER are mediated through the KVM_SIGNAL_MSI ioctl (MSI >> direct MSI injection advertised by the CAP_SIGNAL_MSI capability) >> >> the kvm_gsi_direct_mapping is explicitly set to false to emphasize the >> difference with GICv2M. Direct mapping cannot work with ITS since >> the content of the MSI data is not the target interrupt ID but an >> eventd id. >> >> GSI routing is advertised (kvm_gsi_routing_allowed) as well as >> msi/irqfd signaling (kvm_msi_via_irqfd_allowed). >> >> The MSI frame (GITS_TRANSLATER) absolute GPA is computed on first >> kvm_its_send_msi() call. It is then passed through KVM_SIGNAL_MSI >> ioctl. >> >> Signed-off-by: Pavel Fedin <p.fe...@samsung.com> >> Signed-off-by: Eric Auger <eric.au...@redhat.com> > > > > Applied to target-arm.next, thanks.
Oops, meant to send that as a followup to the cover letter! thanks -- PMM