[Qemu-devel] Re: virtio-serial semantics for binary data and guest agents

[Anthony Liguori <aliguori@linux.vnet.ibm.com>]

My basic understanding here is--there's no way for a virtio-serial 
client (in-guest) to reliably reset a session such that if the client in 
the guest crashes, the protocol needs to have some recovery mechanism.

This seems like a huge oversight in the design of virtio-serial.  Are we 
missing something here?

Regards,

Anthony Liguori

On 02/22/2011 04:40 PM, Michael Roth wrote:
> Hi everyone,
>
> As some of you are aware we've been working for a few months now 
> towards creating a qemu-specific guest agent to implement 
> bi-directional RPCs between the host and the guest to support certain 
> operations like copy/paste, guest-initiated shutdown, and basic file 
> transfer.
>
> Currently the protocol uses the xmlrpc-c library to marshall functions 
> and arguments into a data format we can transport over 
> virtio-serial/isa-serial via an HTTP-like protocol. Recently some 
> concerns have been raised over pulling an external dependency such as 
> xmlrpc-c into qemu, and we've been looking at some alternatives.
>
> Some clear, well-defined approaches such as ASN.1/BER show much 
> promise, but we currently lack a way to implement them cleanly due to 
> the following drawbacks with virtio-serial which prevent us from being 
> able to implement connection/session-oriented protocols:
>
> If something in the guest is attempting to read/write from the 
> virtio-serial device, and nothing is connected to virtio-serial's host 
> character device (say, a socket)
>
> 1. writes will block until something connect()s, at which point the 
> write will succeed
>
> 2. reads will always return 0 until something connect()s, at which 
> point the reads will block until there's data
>
> This makes it difficult (impossible?) to implement the notion of 
> connect/disconnect or open/close over virtio-serial without layering 
> another protocol on top using hackish things like length-encoded 
> payloads or sentinel values to determine the end of one 
> RPC/request/response/session and the start of the next.
>
> For instance, if the host side disconnects, then reconnects before we 
> read(), we may never get the read()=0, and our FD remains valid. 
> Whereas with a tcp/unix socket our FD is no longer valid, and the 
> read()=0 is an event we can check for at any point after the other end 
> does a close/disconnect.
>
> Or if the host side disconnects/closes before/while we write(), we 
> block. If they reconnect, our write() succeeds, and they potentially 
> end up with garbage meant for the previous process. With a tcp/unix 
> socket, the write() will return an EPIPE indicating the FD is no 
> longer valid.
>
> Since virtio-serial is meant to be a general-purpose transport for raw 
> binary data as well as a pv serial console, I wonder if a 
> virtio-serial mode with semantics closer to tcp/unix sockets is 
> necessary? Any thoughts?
>
> Thanks,
>
> Mike