Re: [Qemu-devel] [PATCH v5 1/6] hw/arm/boot: load_dtb becomes non static arm_load_dtb
On 12/05/2014 07:16 PM, Peter Maydell wrote: > On 5 December 2014 at 16:38, Peter Maydell wrote: >> On 30 November 2014 at 18:19, Eric Auger wrote: >>> load_dtb is renamed into arm_load_dtb and becomes non static. >>> it will be used by machvirt for dynamic instantiation of >>> platform devices >> >> 'virt' shouldn't be a special case -- we should always >> handle setting up the DTB in guest memory in the same >> way, whether there happens to be a vfio platform device >> available or not. > > ...this probably means that a bunch of the work currently > done in arm_load_kernel() should be deferred to a 'machine > init complete' hook (perhaps all of it?). Hi Peter, OK I moved the arm_load_kernel code into a machine init done notify and arm_load_kernel now only registers the notifier. For machine files willing to support platform bus, the arm_load_kernel must happen before the registration of the notifier that adds platform bus nodes and after CPU init (notifiers are executed in registration reverse order). Best Regards Eric > > -- PMM >
Re: [Qemu-devel] [PATCH v5 1/6] hw/arm/boot: load_dtb becomes non static arm_load_dtb
On 5 December 2014 at 16:38, Peter Maydell wrote: > On 30 November 2014 at 18:19, Eric Auger wrote: >> load_dtb is renamed into arm_load_dtb and becomes non static. >> it will be used by machvirt for dynamic instantiation of >> platform devices > > 'virt' shouldn't be a special case -- we should always > handle setting up the DTB in guest memory in the same > way, whether there happens to be a vfio platform device > available or not. ...this probably means that a bunch of the work currently done in arm_load_kernel() should be deferred to a 'machine init complete' hook (perhaps all of it?). -- PMM
Re: [Qemu-devel] [PATCH v5 1/6] hw/arm/boot: load_dtb becomes non static arm_load_dtb
On 30 November 2014 at 18:19, Eric Auger wrote: > load_dtb is renamed into arm_load_dtb and becomes non static. > it will be used by machvirt for dynamic instantiation of > platform devices 'virt' shouldn't be a special case -- we should always handle setting up the DTB in guest memory in the same way, whether there happens to be a vfio platform device available or not. thanks -- PMM