Re: [PATCH 0/4] USB hostdev: allow addressing by port

On Thu, Jun 26, 2025 at 06:01:30PM +0200, Maximilian Martin wrote: > > Am 23.06.2025 um 16:50 schrieb Daniel P. Berrangé: >> On Mon, Apr 21, 2025 at 09:38:34PM +0200, Maximilian Martin via Devel wrote: >>> This resubmission splits up the previous patch into multiple patches and >>> incorporates review comments from Michal Prívozník. >>> >>> Currently, only vendor/product and bus/device matching are supported for USB host >>> devices. Neither of these provide a stable and persistent way of assigning a guest >>> a specific host device. Vendor/product can be ambiguous. Device numbers change on >>> every enumeration. >>> >>> This patch adds a bus/port matching, which allows a specific port on the host to be >>> specified using the dotted notation found in Linux's "devpath" sysfs attribute. >> In terms of our API, we're expecting people to use the >> 'node device' APIs to identify what devices are available >> on the host and their attributes, rather than queryin >> Linux directly. This gives us a platform independant >> usage model, which also works when the mgmt app has no >> login to the virt host. >> >> >> Currently for USB devices we report bus, dev, vendor & product, >> but don't report the port: >> >> # virsh nodedev-dumpxml usb_1_10_4 >> <device> >> <name>usb_1_10_4</name> >> <path>/sys/devices/pci0000:00/0000:00:14.0/usb1/1-10/1-10.4</path> >> <devnode type='dev'>/dev/bus/usb/001/004</devnode> >> <parent>usb_1_10</parent> >> <driver> >> <name>usb</name> >> </driver> >> <capability type='usb_device'> >> <bus>1</bus> >> <device>4</device> >> <product id='0x10c0'>Dell Integrated Hub</product> >> <vendor id='0x1604'>Tascam</vendor> >> </capability> >> </device> >> >> we need to add 'port' to this xml document under the 'usb_device' >> capability. >> > Well, the port is included in the path (everything after last '-'). udev > does not report it explicitly. Do you think this is enough or should I parse > the path and add the port to capabilites? No, we would consider the path to be a platform specific opaque string, and liable to change. So it isn't something apps should be parsing to extract info from. We need the path explicitly modelled under the usb_device capability With regards, Daniel