[Qemu-devel] Re: [RFC][PATCH v6 07/23] virtagent: base server definitions

[Michael Roth <mdroth@linux.vnet.ibm.com>]

On 01/21/2011 10:38 AM, Jes Sorensen wrote:
>> diff --git a/virtagent-server.h b/virtagent-server.h
>> new file mode 100644
>> index 0000000..9f68921
>> --- /dev/null
>> +++ b/virtagent-server.h
>> @@ -0,0 +1,34 @@
>> +/*
>> + * virt-agent - host/guest RPC daemon functions
>> + *
>> + * Copyright IBM Corp. 2010
>> + *
>> + * Authors:
>> + *  Michael Roth<mdroth@linux.vnet.ibm.com>
>> + *
>> + * This work is licensed under the terms of the GNU GPL, version 2 or later.
>> + * See the COPYING file in the top-level directory.
>> + *
>> + */
>> +
>> +#include<xmlrpc-c/base.h>
>> +#include<xmlrpc-c/server.h>
>> +
>> +#define GUEST_AGENT_SERVICE_ID "virtagent"
>> +#define GUEST_AGENT_PATH "/tmp/virtagent-guest.sock"
>> +#define HOST_AGENT_SERVICE_ID "virtagent-host"
>> +#define HOST_AGENT_PATH "/tmp/virtagent-host.sock"
>> +#define VA_GETFILE_MAX 1<<  30
>> +#define VA_FILEBUF_LEN 16384
>> +#define VA_DMESG_LEN 16384
>
> I really don't like these hard coded constants - you you have a command
> line interface allowing for the change of the sockets and file names?
> Otherwise you'll hit problems on the host side with concurrent runs of qemu.

Yup, that's one of the TODOs. In terms of configuration we can add 
parameters to the chardev to override these, but the goal here is sane 
defaults to avoid unnecessarily complicated invocations.

>
> I really would like to see the dmesg stuff removed too for now as we
> discussed earlier.

I think as a development/support tool it has a recently strong use case, 
even given it's limitations (which are not so bad....we retrieve up to a 
max of 16KB, possibly less depending on guest configuration, so it's not 
entirely predictable, but it's not dangerous. It's platform-specific, 
but that's handled by capabilities negotiation).

I just don't really see the downside to keeping it in.

>
> Cheers,
> Jes