Re: Issues with PCI-Passtrough (VT-d) in HVM with Xen 4.6

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Thu, Jun 02, 2016 at 09:59:29PM +0200, Sylwester Sosnowski wrote:
> 
> Issues with PCI-Passtrough (VT-d) in HVM with Xen 4.6
> =====================================================
> 
> Background
> ----------
> 
> I am trying to passthrough an Industrial Ethernet Interface
> (Hilscher GmbH CIFX 50E-DP(M/S)) on a HVM DomU running the
> Xen 4.6 Hypervisor. The card is being pass-trough to the HVM
> using the PCI permissive mode and VT-d is active on this platform.
> 
> The Linux driver of the card (available only with NDA) at first
> sight seems to work properly (e.g. no system stability problems,
> no call traces in dmesg).
> 
> Hilscher provides the libcifx, which is an user-level library
> for accessing the card. libcfix uses the generic UIO interface
> and some card-specific interfaces to communicate.
> 
> Issue
> -----
> 
> Whenever libcifx tries perform a reset sequence initializing
> the card peripherals, we get an empty / invalid result.
> 
> I digged deeper into this by adding some debug hooks into the
> xen-pciback kernel module.
> 
> The card performs writes to offset 0x10 and 0x30 at card
> initialization (mostly writing dwords and words).
> 
> To verify if the writes were performed successfully, I read
> back the values after writing and can see that the read data
> differs from the written one.
> 
> Temporary Fix
> -------------
> 
> After checking the source[1] of the PCI configuration space
> handling in xen-pciback, I found out that changing line 258
> to read
> 
>     if (handled && !err) {
> 
> instead of:
> 
>     if (!handled && !err) {
> 
> solves the issue and I can successfully write to the interface.
> 
> I am unsure why this works and if it's the right way to do it
> or possibly a Xen bug, so I would like to ask for feedback
> for this.

did you try the permissive module parameter?
> 
> 
> Please let me know whenever I can supply additional logfiles or
> info. I'd be happy if we can resolve whenever this is a hardware-
> specific, driver-specific or Xen issue.
> 
> Thank you,
> Sylwester
> 
> PS: I tried to CC Ryan Wilson and Konrad Rzeszutek Wilk, but the
> E-Mail seems to be no longer active. This is where the changes
> originated.
> 
> Refereces
> ---------
> 1.
> https://github.com/google/kasan/blob/master/drivers/xen/xen-pciback/conf_space.c

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel