On 31/03/17 21:33, Max Reitz wrote:
> On 31.03.2017 06:47, Alexey Kardashevskiy wrote:
>> Normally pci_init_bus_master() would be called either via
>> bus->machine_done.notify or directly from do_pci_register_device().
>>
>> However if a device's realize() failed, pci_init_bus_master() is not
>> called, and do_pci_unregister_device() fails on
>> memory_region_del_subregion() as it was not mapped.
>>
>> This adds a check that subregion was mapped before unmapping it.
>>
>> Fixes: c53598ed18e4 ("pci: Add missing drop of bus master AS reference")
>> Signed-off-by: Alexey Kardashevskiy <a...@ozlabs.ru>
>> ---
>> This is to pass iotest 051 which does run QEMU like this:
>>
>> ppc64-softmmu/qemu-system-ppc64 -nodefaults -machine accel=qtest -nographic
>> -monitor stdio -serial none -drive if=virtio
>>
>> which normally fails with:
>>
>> qemu-system-ppc64: -drive if=virtio: Device needs media, but drive is empty
>>
>> or asserts (without this patch):
>>
>> qemu-system-ppc64: /home/aik/p/qemu-kvm/memory.c:2118:
>> memory_region_del_subregion: Assertion `subregion->container == mr' failed.
>> ---
>> hw/pci/pci.c | 6 ++++--
>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/hw/pci/pci.c b/hw/pci/pci.c
>> index bd8043c460..259483b1c0 100644
>> --- a/hw/pci/pci.c
>> +++ b/hw/pci/pci.c
>> @@ -869,8 +869,10 @@ static void do_pci_unregister_device(PCIDevice *pci_dev)
>> pci_dev->bus->devices[pci_dev->devfn] = NULL;
>> pci_config_free(pci_dev);
>>
>> - memory_region_del_subregion(&pci_dev->bus_master_container_region,
>> - &pci_dev->bus_master_enable_region);
>> + if (memory_region_is_mapped(&pci_dev->bus_master_enable_region)) {
>> + memory_region_del_subregion(&pci_dev->bus_master_container_region,
>> + &pci_dev->bus_master_enable_region);
>> + }
>> address_space_destroy(&pci_dev->bus_master_as);
>> }
>
> I'm not sure whether it's generally assumed to be safe to call
> memory_region_is_mapped() on "uninitialized" memory regions
> (bus_master_enable_region is just completely zeroed out before
> pci_init_bus_master(), as far as I can see and guess), but it certainly
> works in practice, so:
>
> Reviewed-by: Max Reitz <mre...@redhat.com>PCI device is also initialized via object_initialize() which does memset(0) so I assume it is quite safe. -- Alexey
signature.asc
Description: OpenPGP digital signature
