Wei Liu
2011-May-26 03:06 UTC
[Xen-devel] [PATCH 0/2] x86: Enabling emulated MSI injection
This patch series enables Xen to deliver emulated MSI for upstream QEMU. To achieve this goal, a new HVM operation -- HVMOP_inj_msi -- is added. Also the VMSI handling code for x86 is refactored. Libxc is modified as well. Tested with VirtIO net and VirtIO disk. Signed-off-by: Wei Liu <liuw@liuw.name> -- Best regards Wei Liu Twitter: @iliuw Site: http://liuw.name _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2011-May-26 04:49 UTC
Re: [Xen-devel] [PATCH 0/2] x86: Enabling emulated MSI injection
On 26/05/2011 04:06, "Wei Liu" <liuw@liuw.name> wrote:> This patch series enables Xen to deliver emulated MSI for upstream QEMU. > > To achieve this goal, a new HVM operation -- HVMOP_inj_msi -- is > added. Also the VMSI handling code for x86 is refactored. Libxc is > modified as well. > > Tested with VirtIO net and VirtIO disk. > > Signed-off-by: Wei Liu <liuw@liuw.name>Don''t renumber existing HVMOPs, some of the numbers are baked into our guest ABI. Define your new HVMOP at the end, and enclose in if defined(__XEN__) || defined(__XEN_TOOLS__) -- it is only used by toolstack (qemu) and this gives us flexibility to change your new interface in future if we need to, as this tells us it is not baked into the guest ABI. Apart from that, collect an Ack from Stefano that this integrates okay with his pirq/emuirq stuff for delivering HVM interrupts on event channels. It may be that you''re going in at a slightly too raw level. Stefano wrote a bunch of the pirq stuff so he can advise on that. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Wei Liu
2011-May-26 05:05 UTC
Re: [Xen-devel] [PATCH 0/2] x86: Enabling emulated MSI injection
On Thu, May 26, 2011 at 12:49 PM, Keir Fraser <keir.xen@gmail.com> wrote:> On 26/05/2011 04:06, "Wei Liu" <liuw@liuw.name> wrote: > >> This patch series enables Xen to deliver emulated MSI for upstream QEMU. >> >> To achieve this goal, a new HVM operation -- HVMOP_inj_msi -- is >> added. Also the VMSI handling code for x86 is refactored. Libxc is >> modified as well. >> >> Tested with VirtIO net and VirtIO disk. >> >> Signed-off-by: Wei Liu <liuw@liuw.name> > > Don''t renumber existing HVMOPs, some of the numbers are baked into our guest > ABI. Define your new HVMOP at the end, and enclose in if defined(__XEN__) || > defined(__XEN_TOOLS__) -- it is only used by toolstack (qemu) and this gives > us flexibility to change your new interface in future if we need to, as this > tells us it is not baked into the guest ABI. >Oops... I mistaken that it should go into the guest ABI. Change it later.> Apart from that, collect an Ack from Stefano that this integrates okay with > his pirq/emuirq stuff for delivering HVM interrupts on event channels. It > may be that you''re going in at a slightly too raw level. Stefano wrote a > bunch of the pirq stuff so he can advise on that. > > -- Keir > > >OK. Waiting for his reply. -- Best regards Wei Liu Twitter: @iliuw Site: http://liuw.name _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel