[Spice-devel] [Qemu-devel] [Xen-devel] [PATCH] libxl: Spice usbredirection support for upstream qemu

[Andreas Färber]

Am 04.07.2013 14:46, schrieb Stefano Stabellini:
> On Thu, 4 Jul 2013, Fabio Fantoni wrote:
>> Il 04/07/2013 12:32, Wei Liu ha scritto:
>>
>> On Thu, Jul 04, 2013 at 12:16:43PM +0200, Fabio Fantoni wrote:
>>
>> Il 04/07/2013 12:12, Wei Liu ha scritto:
>>
>> On Thu, Jul 04, 2013 at 12:05:59PM +0200, Fabio Fantoni wrote:
>>
>> Usage: spiceusbredirection=1|0 (default=0)
>> Enables spice usbredirection. The Spice usbredirection creates usb2
>> controller and 4 usbredirection channels for redirection of up to 4
>> usb devices from spice client to domU's qemu.
>>
>> Signed-off-by: Fabio Fantoni <fabio.fantoni at m2r.biz>
>> ---
>>  docs/man/xl.cfg.pod.5       |    8 ++++++++
>>  tools/libxl/libxl_create.c  |    1 +
>>  tools/libxl/libxl_dm.c      |   18 ++++++++++++++++++
>>  tools/libxl/libxl_types.idl |    1 +
>>  tools/libxl/xl_cmdimpl.c    |    2 ++
>>  5 files changed, 30 insertions(+)
>>
>> diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
>> index 766862d..a450800 100644
>> --- a/docs/man/xl.cfg.pod.5
>> +++ b/docs/man/xl.cfg.pod.5
>> @@ -1134,6 +1134,14 @@ requires vdagent service installed on domU o.s. to work. The default is 0.
>>  =back
>> +=item B<spiceusbredirection=BOOLEAN>
>> +
>> +Enables spice usbredirection. The Spice usbredirection creates usb2
>> +controller and 4 usbredirection channels for redirection of up to 4 usb
>> +devices from spice client to domU's qemu. The default is 0.
>> +
>> +=back
>> +
>>  =head3 Miscellaneous Emulated Hardware
>>  =over 4
>> diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
>> index 8db5460..58df106 100644
>> --- a/tools/libxl/libxl_create.c
>> +++ b/tools/libxl/libxl_create.c
>> @@ -289,6 +289,7 @@ int libxl__domain_build_info_setdefault(libxl__gc *gc,
>>                                       false);
>>              libxl_defbool_setdefault(&b_info->u.hvm.spice.agent_mouse, true);
>>              libxl_defbool_setdefault(&b_info->u.hvm.spice.vdagent, false);
>> +            libxl_defbool_setdefault(&b_info->u.hvm.spice.usbredirection, false);
>>          }
>>          libxl_defbool_setdefault(&b_info->u.hvm.nographic, false);
>> diff --git a/tools/libxl/libxl_dm.c b/tools/libxl/libxl_dm.c
>> index bc605e4..4f625e0 100644
>> --- a/tools/libxl/libxl_dm.c
>> +++ b/tools/libxl/libxl_dm.c
>> @@ -471,6 +471,24 @@ static char ** libxl__build_device_model_args_new(libxl__gc *gc,
>>                      "virtserialport,chardev=vdagent,name=com.redhat.spice.0",
>>                      NULL);
>>              }
>> +
>> +            if (libxl_defbool_val(b_info->u.hvm.spice.usbredirection)) {
>> +                flexarray_vappend(dm_args, "-device","ich9-usb-ehci1,id=usb,"
>> +                    "bus=pci.0,addr=0x1d.0x7", "-device","ich9-usb-uhci1,"
>> +                    "masterbus=usb.0,firstport=0,bus=pci.0,multifunction=on,"
>> +                    "addr=0x1d.0x0", "-device","ich9-usb-uhci2,masterbus=usb.0,"
>> +                    "firstport=2,bus=pci.0,addr=0x1d.0x1", "-device",
>> +                    "ich9-usb-uhci3,masterbus=usb.0,firstport=4,bus=pci.0,"
>> +                    "addr=0x1d.0x2", "-chardev","spicevmc,name=usbredir,"
>> +                    "id=usbrc1","-device","usb-redir,chardev=usbrc1,id=usbrc1,"
>> +                    "bus=usb.0", "-chardev","spicevmc,name=usbredir,id=usbrc2",
>> +                    "-device","usb-redir,chardev=usbrc2,id=usbrc2,bus=usb.0",
>> +                    "-chardev","spicevmc,name=usbredir,id=usbrc3","-device",
>> +                    "usb-redir,chardev=usbrc3,id=usbrc3,bus=usb.0", "-chardev",
>> +                    "spicevmc,name=usbredir,id=usbrc4","-device","usb-redir,"
>> +                    "chardev=usbrc4,id=usbrc4,bus=usb.0", NULL);
>>
>> Any reason for so many hardcoded options?
>>
>> I searched and requested for one year on spice-devel and qemu-devel
>> about alternative methods but nothing found for now.
>> Already tried usb=1 which creates usb1 controller that is not
>> working with usb redirection.
>>
>> What if QEMU upstream changes and these options don't work any more? In
>> that case this functionality is broken and users have no way to
>> workaround it. IMHO unless they are clearly documented we should not
>> consider adding in theses hardcoded options in libxl.
>>
>>
>> Added to cc spice-devel and qemu-devel for ask again if there is a better solution to do this.
>>
>> @spice-devel and qemu-devel:
>> Can someone help to improve qemu options above for enable usb redirection please?
>> Thanks for any reply.
> 
> It's not about "improving" qemu options for usb redirection (even though
> they could use a simplification), it's whether they are guaranteed to be
> stable. Are you sure that a future QEMU release is going to work with
> these options?
> For example, is libvirt using something similar to this?
> 
> Usually QEMU cmdline options are considered a stable interface, so I
> wouldn't worry too much about it.

I'm not sure if pci.0 bus name of i440fx should be considered stable?

Also if at some point you try to use, e.g., q35 rather than i440fx
machine, it might have EHCI by default and the snippet would be
instantiating a duplicate one. I.e., the three levels of dependency
could be separated more clearly: PCI, USB, redirection.

Stating the obvious, a loop would be a more readable solution for adding
4 devices only differing in device/chardev id. ;)

CC'ing Hans.

Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg