Re: [Qemu-devel] [PATCH v2 2/2] qapi: allow blockdev-add for NFS

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 3492 bytes --]

On 10/24/2016 02:27 PM, Ashijeet Acharya wrote:
> Introduce new object 'BlockdevOptionsNFS' in qapi/block-core.json to
> support blockdev-add for NFS network protocol driver. Also make a new
> struct NFSServer to support tcp connection.
> 
> Signed-off-by: Ashijeet Acharya <ashijeetacharya@gmail.com>
> ---
>  qapi/block-core.json | 56 ++++++++++++++++++++++++++++++++++++++++++++++++----
>  1 file changed, 52 insertions(+), 4 deletions(-)
> 
> diff --git a/qapi/block-core.json b/qapi/block-core.json
> index 9d797b8..3ab028d 100644
> --- a/qapi/block-core.json
> +++ b/qapi/block-core.json
> @@ -1714,9 +1714,9 @@
>  { 'enum': 'BlockdevDriver',
>    'data': [ 'archipelago', 'blkdebug', 'blkverify', 'bochs', 'cloop',
>              'dmg', 'file', 'ftp', 'ftps', 'gluster', 'host_cdrom',
> -            'host_device', 'http', 'https', 'luks', 'null-aio', 'null-co',
> -            'parallels', 'qcow', 'qcow2', 'qed', 'quorum', 'raw',
> -	    'replication', 'tftp', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }
> +            'host_device', 'http', 'https', 'luks', 'nfs', 'null-aio',
> +            'null-co', 'parallels', 'qcow', 'qcow2', 'qed', 'quorum', 'raw',
> +            'replication', 'tftp', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }

Missing a comment that 'nfs' is since 2.8.

>  ##
> +# @NFSServer
> +#
> +# Captures the address of the socket
> +#
> +# @type:        transport type used for NFS (only TCP supported)
> +#
> +# @host:        host part of the address
> +#
> +# Since 2.8
> +##
> +{ 'struct': 'NFSServer',
> +  'data': { 'type': 'str',

Please make this an enum, instead of an open-coded string. It's okay if
the enum only has one value 'tcp' for now; but using an enum will make
it introspectable if we later add a second transport, unlike what we get
with an open-coded string.

Must 'type' be mandatory if it must always be 'tcp'?

> +            'host': 'str' } }
> +
> +##
> +# @BlockdevOptionsNfs
> +#
> +# Driver specific block device option for NFS
> +#
> +# @server:        host address
> +#
> +# @path:          path of the image on the host
> +#
> +# @uid:           #optional UID value to use when talking to the server
> +#
> +# @gid:           #optional GID value to use when talking to the server

Do we want to allow string names in addition to numeric uid/gid values?
I'm not sure if NFS has name-based id mapping, but it's food for thought
on whether we need to use an alternate type here (alternate between
integer id and string name), or leave this as is.

> +#
> +# @tcp-syncnt:    #optional number of SYNs during the session establishment

Would tcp-syn-count be any more legible?  What is the default when omitted?

> +#
> +# @readahead:     #optional set the readahead size in bytes

What's the default when omitted?

> +#
> +# @pagecache:     #optional set the pagecache size in bytes

Default?

> +#
> +# @debug:         #optional set the NFS debug level (max 2)

Presumably default 0?

> +#
> +# Since 2.8
> +##
> +{ 'struct': 'BlockdevOptionsNfs',
> +  'data': { 'server': 'NFSServer',
> +            'path': 'str',
> +            '*uid': 'int',
> +            '*gid': 'int',
> +            '*tcp-syncnt': 'int',
> +            '*readahead': 'int',
> +            '*pagecache': 'int',
> +            '*debug': 'int' } }
> +

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]