Re: [PATCH 02/18] fuse: Allow exporting BDSs via FUSE

[Markus Armbruster]

Kevin Wolf <address@hidden> writes:

> Am 20.12.2019 um 11:48 hat Max Reitz geschrieben:
>> On 20.12.19 11:26, Kevin Wolf wrote:
>> > Am 19.12.2019 um 15:38 hat Max Reitz geschrieben:
>> >> fuse-export-add allows mounting block graph nodes via FUSE on some
>> >> existing regular file.  That file should then appears like a raw disk
>> >> image, and accesses to it result in accesses to the exported BDS.
>> >>
>> >> Right now, we only set up the mount point and tear all mount points down
>> >> in bdrv_close_all().  We do not implement any access functions, so
>> >> accessing the mount point only results in errors.  This will be
>> >> addressed by a followup patch.
>> >>
>> >> The set of exported nodes is kept in a hash table so we can later add a
>> >> fuse-export-remove that allows unmounting.
>> >>
>> >> Signed-off-by: Max Reitz <address@hidden>
>> > 
>> >> diff --git a/qapi/block.json b/qapi/block.json
>> >> index 145c268bb6..03f8d1b537 100644
>> >> --- a/qapi/block.json
>> >> +++ b/qapi/block.json
>> >> @@ -317,6 +317,29 @@
>> >>  ##
>> >>  { 'command': 'nbd-server-stop' }
>> >>  
>> >> +##
>> >> +# @fuse-export-add:
>> >> +#
>> >> +# Exports a block graph node on some (file) mountpoint as a raw image.
>> >> +#
>> >> +# @node-name: Node to be exported
>> >> +#
>> >> +# @mountpoint: Path on which to export the block device via FUSE.
>> >> +#              This must point to an existing regular file.
>> >> +#
>> >> +# @writable: Whether clients should be able to write to the block
>> >> +#            device via the FUSE export. (default: false)
>> >> +#
>> >> +# Since: 5.0
>> >> +##
>> >> +{ 'command': 'fuse-export-add',
>> >> +  'data': {
>> >> +      'node-name': 'str',
>> >> +      'mountpoint': 'str',
>> >> +      '*writable': 'bool'
>> >> +  },
>> >> +  'if': 'defined(CONFIG_FUSE)' }
>> > 
>> > Can this use a BlockExport union from the start like I'm introducing in
>> > the storage daemon series, together with a generic block-export-add?
>> 
>> Hm, you mean still adding a FuseExport structure that would be part of
>> BlockExport and then dropping fuse-export-add in favor of a
>> block-export-add that we want anyway?
>
> Yes.
>
>> > It also looks like node-name and writable should be part of the common
>> > base of BlockExport.
>> 
>> node-name definitely, I’m not so sure about writable.  Or, to be more
>> precise, I think that if we want writable to be in the base, we also
>> want growable to be there: Both are primarily options for the
>> BlockBackend that the exports use.
>> 
>> But both of course also need to be supported by the export
>> implementation.  nbd can make its BB growable all it wants, but that
>> doesn’t make it work.
>
> Right. Pragmatically, I think exports are very like to support writable,
> but probably rather unlikely to support growable. So I do think there
> would be a point for making writable part of the common base, but not
> growable.
>
>> So if we kept writable and growable in the common base, then the schema
>> would give no information about what exports actually support them.
>> 
>> On one hand, I don’t know whether it’s important to have this
>> information in a static form, or whether it’s sufficient to learn at
>> runtime.
>> 
>> On the other, I don’t know whether it’s important to have those fields
>> in the base or not.  Would it make a difference on the wire?
>
> Not for the command itself, so I think we're free to change it later. It
> might make a difference for introspection, though, not sure. Markus?

QAPI schema introspection is designed to hide the difference between
local members and base members.  You can move members to or from a base
type freely without affecting introspection.  Even if that creates or
deletes the base type.

Example: DriveBackup

    { 'struct': 'DriveBackup',
      'base': 'BackupCommon',
      'data': { 'target': 'str',
                '*format': 'str',
                '*mode': 'NewImageMode' } }

where BackupCommon is

    { 'struct': 'BackupCommon',
      'data': { '*job-id': 'str', 'device': 'str',
                'sync': 'MirrorSyncMode', '*speed': 'int',
                '*bitmap': 'str', '*bitmap-mode': 'BitmapSyncMode',
                '*compress': 'bool',
                '*on-source-error': 'BlockdevOnError',
                '*on-target-error': 'BlockdevOnError',
                '*auto-finalize': 'bool', '*auto-dismiss': 'bool',
                '*filter-node-name': 'str' } }

query-qmp-schema describes DriveBackup like this:

    {"name": "30",
     "members": [
         {"name": "job-id", "default": null, "type": "str"},
         {"name": "device", "type": "str"},
         {"name": "sync", "type": "235"},
         {"name": "speed", "default": null, "type": "int"},
         {"name": "bitmap", "default": null, "type": "str"},
         {"name": "bitmap-mode", "default": null, "type": "236"},
         {"name": "compress", "default": null, "type": "bool"},
         {"name": "on-source-error", "default": null, "type": "237"},
         {"name": "on-target-error", "default": null, "type": "237"},
         {"name": "auto-finalize", "default": null, "type": "bool"},
         {"name": "auto-dismiss", "default": null, "type": "bool"},
         {"name": "filter-node-name", "default": null, "type": "str"},
         {"name": "target", "type": "str"},
         {"name": "format", "default": null, "type": "str"},
         {"name": "mode", "default": null, "type": "234"}],
     "meta-type": "object"}

[...]