Dear QEMU developers, I hope this message finds you well.
I would like to propose a feature request regarding the TCG plugin interface, aiming to enhance its usability as a golden model in CPU development environments. It would be extremely valuable for me if the TCG plugin framework could provide hooks or mechanisms to capture and propagate interruption or exception events ― including synchronous exceptions (e.g., traps, faults) and asynchronous interrupts (e.g., external IRQs, timer interrupts). This would allow plugins to gain observability into CPU control flow changes that are not directly visible through basic instruction instrumentation. At present, the QEMU plugin interface offers the following event types, as defined in `enum qemu_plugin_event`: QEMU_PLUGIN_EV_VCPU_INIT QEMU_PLUGIN_EV_VCPU_EXIT QEMU_PLUGIN_EV_VCPU_TB_TRANS QEMU_PLUGIN_EV_VCPU_IDLE QEMU_PLUGIN_EV_VCPU_RESUME QEMU_PLUGIN_EV_VCPU_SYSCALL QEMU_PLUGIN_EV_VCPU_SYSCALL_RET QEMU_PLUGIN_EV_FLUSH QEMU_PLUGIN_EV_ATEXIT QEMU_PLUGIN_EV_MAX To better support exception and interrupt observability in plugins, I propose introducing a new plugin event, for example: QEMU_PLUGIN_EV_VCPU_INT This event would be triggered when the virtual CPU receives or processes an interrupt or exception. Ideally, the associated callback should receive detailed context ― such as the exception or interrupt type, current PC, privilege level, and cause ― and should support a return value, such as bool, to indicate whether the propagation of the event should continue or be blocked to have the control of the interrupting timing. Thank you very much for your time and for the amazing work on QEMU.
