[libvirt] [PATCH v2 1/2] qemu: Check for existing hostdev address for cold attach device

[John Ferlan]

On 06/13/2018 04:15 AM, Ján Tomko wrote:
> On Tue, Jun 12, 2018 at 10:32:06AM -0400, John Ferlan wrote:
>> Add a check during qemuDomainAttachDeviceConfig whether the
>> new/incoming <hostdev ...> device would have an existing
>> <address> already and if so fail the attach. This can happen
>> if two hostdev's are added supplying the same address or
>> if the new hostdev address could possibly be a duplicate
>> of an existing SCSI <disk>.
>>
>> Signed-off-by: John Ferlan <jferlan at redhat.com>
>> ---
>> src/qemu/qemu_driver.c | 6 ++++++
>> 1 file changed, 6 insertions(+)
>>
>> diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
>> index f0fb806fcd..ae8e0e898a 100644
>> --- a/src/qemu/qemu_driver.c
>> +++ b/src/qemu/qemu_driver.c
>> @@ -8015,6 +8015,12 @@ qemuDomainAttachDeviceConfig(virDomainDefPtr
>> vmdef,
>>                            _("device is already in the domain
>> configuration"));
>>             return -1;
>>         }
>> +        if (dev->data.hostdev->info->type !=
>> VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE &&
>> +            virDomainDefHasDeviceAddress(vmdef,
>> dev->data.hostdev->info)) {
>> +            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
>> +                           _("a device with the same address already
>> exists "));
>> +            return -1;
>> +        }
> 
> This check feels out of place here. We should be checking for that in
> postParse. Do we have the same problem on domain startup?
> 

I've avoided new post parse checks due to the domain disappearing phenomena.

But no, I don't believe the same problem exists there. How would you
suggest reproducing that?

If one virsh edit's their domain adding two hostdev's with the same
<address>, then virDomainDefValidateInternal catches that with the call
to virDomainDefCheckDuplicateDriveAddresses.  With this patch if one
calls attach-device --config with a duplicated <address>, then this
check will catch that.

This is similar to VIR_DOMAIN_DEVICE_RNG processing in the same code.

John

> Jano
> 
>>         if (virDomainHostdevInsert(vmdef, hostdev))
>>             return -1;
>>         dev->data.hostdev = NULL;
>> -- 
>> 2.14.4
>>
>> -- 
>> libvir-list mailing list
>> libvir-list at redhat.com
>> https://www.redhat.com/mailman/listinfo/libvir-list