Eduardo Habkost <ehabk...@redhat.com> writes:
> On Fri, Sep 18, 2015 at 08:42:54PM +0200, Thomas Huth wrote:
>> On 18/09/15 14:00, Markus Armbruster wrote:
>> > Several devices don't survive object_unref(object_new(T)): they crash
>> > or hang during cleanup, or they leave dangling pointers behind.
>> >
>> > This breaks at least device-list-properties, because
>> > qmp_device_list_properties() needs to create a device to find its
>> > properties. Broken in commit f4eb32b "qmp: show QOM properties in
>> > device-list-properties", v2.1. Example reproducer:
>> >
>> > $ qemu-system-aarch64 -nodefaults -display none -machine none -S -qmp
>> > stdio
>> > {"QMP": {"version": {"qemu": {"micro": 50, "minor": 4, "major": 2},
>> > "package": ""}, "capabilities": []}}
>> > { "execute": "qmp_capabilities" }
>> > {"return": {}}
>> > { "execute": "device-list-properties", "arguments": { "typename":
>> > "pxa2xx-pcmcia" } }
>> > qemu-system-aarch64: /home/armbru/work/qemu/memory.c:1307:
>> > memory_region_finalize: Assertion `((&mr->subregions)->tqh_first == ((void
>> > *)0))' failed.
>> > Aborted (core dumped)
>> > [Exit 134 (SIGABRT)]
>> >
>> > Unfortunately, I can't fix the problems in these devices right now.
>> > Instead, add DeviceClass member cannot_even_create_with_object_new_yet
>> > to mark them:
>> >
>> > * Crash or hang during cleanup (didn't debug them, so I can't say
>> > why): "pxa2xx-pcmcia", "realview_pci", "versatile_pci",
>> > "s390-sclp-event-facility", "sclp"
>> >
>> > * Dangling pointers: all CPUs, plus "allwinner-a10", "digic",
>> > "fsl,imx25", "fsl,imx31", "xlnx,zynqmp", because they create CPUs
>> >
>> > * Assert kvm_enabled(): "host-x86_64-cpu", host-i386-cpu",
>> > "host-powerpc64-cpu", "host-embedded-powerpc-cpu",
>> > "host-powerpc-cpu"
>>
>> I just had a look at the powerpc code - you're likely talking about
>> the "assert(kvm_enabled());" in the kvmppc_host_cpu_initfn() in
>> target-ppc/kvm.c ? That should be fine, I think, because
>> kvm_ppc_register_host_cpu_type() is only done on ppc when KVM has been
>> enabled.
Easy to verify on a KVM-capable PPC host: try -device C,help with KVM on
and off, where C is the appropriate host CPU.
> It's true that currently the assert() will never trigger, but we will
> have to eventually move class registration to type_init if we want to
> make a generic query-cpu-definitions implementation work properly
> without depending on global machine accel configuration.
Good point. These CPUs need a TODO marker.
> It won't hurt to set cannot_even_create_with_object_new_yet properly to
> reflect that the class isn't ready yet.
Marking cannot_even_create_with_object_new_yet() is the obvious TODO
marker. It has an unnecessary side effect: we disable introspection for
these CPUs needlessly. Tolerable?
