On 01/07/2013 06:55 AM, Gerd Hoffmann wrote:
> Add support for file chardevs. Output file is mandatory,
> input file is optional. Both file names and file descriptor
> passing is supported.
>
> Signed-off-by: Gerd Hoffmann <kra...@redhat.com>
> ---
> qapi-schema.json | 9 +++++-
> qemu-char.c | 77
> +++++++++++++++++++++++++++++++++++++++++++++++++++++-
> 2 files changed, 84 insertions(+), 2 deletions(-)
>
> diff --git a/qapi-schema.json b/qapi-schema.json
> index e3f0d44..8904d36 100644
> --- a/qapi-schema.json
> +++ b/qapi-schema.json
> @@ -3030,9 +3030,16 @@
> #
> # Since: 1.4
> ##
> +{ 'union': 'ChardevFileSource', 'data': { 'path' : 'str',
> + 'fd' : 'str' } }
> +
> +{ 'type': 'ChardevFile', 'data': { '*in' : 'ChardevFileSource',
> + 'out' : 'ChardevFileSource' } }
> +
> { 'type': 'ChardevDummy', 'data': { } }
>
> -{ 'union': 'ChardevBackend', 'data': { 'null' : 'ChardevDummy' } }
> +{ 'union': 'ChardevBackend', 'data': { 'file' : 'ChardevFile',
> + 'null' : 'ChardevDummy' } }
>
> { 'command': 'chardev-add', 'data': {'id' : 'str',
> 'backend' : 'ChardevBackend' } }
An example in qmp-commands.hx would be helpful; am I correct that an
example would be:
-> { "execute" : "chardev-add",
"arguments" : { "id" : "foo",
"backend" : { "type" : "file", "data" : {
"in" : { "type" : "fd", "data" : "namedfd" },
"out" : { "type" : "path", "data" : "/path/to/file" } } } } }
<- { "return": {} }
where namedfd was previously given via 'getfd'?
Do we need the complexity of supporting fd passing explicitly? 'getfd'
is less than ideal compared to 'add-fd', and for 'add-fd', we would pass
via "path":"/dev/fdset/nnn". That is, why do we need to bend over
backwards to support an alternate syntax for fd passing in a new
command, when we can already use existing commands to get fd passing for
free?
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
