Re: [PATH 0/4] usbip: make vhci_hcd.* objects independent of vhci_hcd.0

[Salvador FandiÃo]

On 02/21/2018 01:35 AM, Shuah Khan wrote:
> Hi Salvador,
>
> On 01/30/2018 01:36 AM, Salvador Fandino wrote:
> > Let me start by explaining the problem that have motivated me to write
> > this patches:
> >
> > I work on the QVD, a virtual desktop platform for Linux. This software
> > runs Linux desktops (i.e. XFCE, KDE) and their applications inside LXC
> > containers, and makes then available through the network to remote
> > users.
> >
> > Supporting USB devices is a common feature customers have been
> > requesting us for a long time (in order to use, for instance, remote
> > signature pads, bar-code scanners, fingerprint readers, etc.). So, we
> > have been working on that feature using the USB/IP layer on the
> > kernel.
> >
> > Connecting and disconnecting devices and transferring data works
> > seamless for the devices listed above. But we also want to make the
> > usbip operations private to the container where they are run.  For
> > instance, it is unacceptable for our product, that one user could list
> > the devices connected by other users or access them.
> >
> > We can control how can access every device using cgroups once those
> > are attached, but the usbip layer is not providing any mechanism for
> > controlling who can attach, detach or list the devices.
>
> Did you explore a solution to add a mechanism for access control to
> usbip?

Could you elaborate on that?

For "usbip", do you mean the user space tools? If that is the case, I 
don't think it would be enough. My aim is to limit vhci usage from 
containers and I have no control about what runs inside the containers. 
So, a mangled usbip tool-set could be used by a malicious user to 
circumvent any access control set there.

IMO, there is no other choice but to control access to VHCI at the 
kernel level.


> > So, we got the idea that in order to enforce that remote usbip devices
> > are only visible inside the container where they were imported, we
> > could conveniently mount-bind inside every container just one of the
> > vhci_hcd directories below /sys/devices/platform. So that it is as if
> > every container had a vhci_hcd just for itself (and also, we restrict
> > access to the matching USB ports in cgroups).
> >
> > Unfortunately, all the vhci_hcd.* devices are controlled through
> > attributes in vhci_hcd.0 making our approach fail and so... well, that
> > is what this patch series changes. It makes every vhci_hcd device
> > controllable through attributes inside its own sysfs directory.>
> > The first patch, does that in the kernel, and the second and third
> > patches change user space, adapting the libusbip and the usbip tools
> > code respectively.
> >
> > Then there is a fourth patch, that allows to create much more USB
> > hubs per machine. It was limited to 64 and we are running thousands of
> > containers (every one requiring a hub) per host.
> >
> > These changes are not completely backward compatible. In the sysfs
> > side, besides moving around the attribute files, now the port numbers
> > go from 0 to CONFIG_USBIP_VHCI_HC_PORTS * 2 - 1 and are reused for
> > every vhci_hcd device. I could have maintained the absolute numeration
> > but I think reusing the numbers is a better and simpler approach.
>
> Not being able to maintain backwards compatibility is an issue. This is
> a considerable change to the user interface.

Well, it is true that it is a considerable change to the user interface 
breaking backward compatibility, but as I had already stated, that 
interface was broken until a couple of months ago, when my coworker, 
Juan Zea, reported it and nobody had noticed it before. So, I don't 
think we are going to affect too many people.

Note also that the user interface does not change when only vhci_hcd.0 
is used.


Regards