On 09.03.2026 13:31, Julian Vetter wrote:
> In pt_irq_create_bind() and _hvm_dpci_msi_eoi() replace the bare
> MASK_EXTR(..., XEN_DOMCTL_VMSI_X86_DEST_ID_MASK) calls with
> XEN_DOMCTL_VMSI_X86_FULL_DEST() so that the high 7 destination ID bits
> stored in XEN_DOMCTL_VMSI_X86_EXT_DEST_ID_MASK are included when
> computing the target vCPU for MSI passthrough IRQs. Increase the local
> dest and delivery_mode variables to uint32_t to match.
Why exactly would the latter also need widening?
> --- a/xen/drivers/passthrough/x86/hvm.c
> +++ b/xen/drivers/passthrough/x86/hvm.c
> @@ -281,7 +281,7 @@ int pt_irq_create_bind(
> {
> case PT_IRQ_TYPE_MSI:
> {
> - uint8_t dest, delivery_mode;
> + uint32_t dest, delivery_mode;
Please prefer here (for dest) ...
> @@ -807,8 +806,7 @@ static int cf_check _hvm_dpci_msi_eoi(
> if ( (pirq_dpci->flags & HVM_IRQ_DPCI_MACH_MSI) &&
> (pirq_dpci->gmsi.gvec == vector) )
> {
> - unsigned int dest = MASK_EXTR(pirq_dpci->gmsi.gflags,
> - XEN_DOMCTL_VMSI_X86_DEST_ID_MASK);
> + unsigned int dest =
> XEN_DOMCTL_VMSI_X86_FULL_DEST(pirq_dpci->gmsi.gflags);
... the type you found already being in use here. See ./CODING_STYLE as to
the use of fixed-width types.
Jan
