[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH 2/3] xen/vt-d: mask interrupt message generation

On 2015/5/4 16:57, Jan Beulich wrote:
On 04.05.15 at 07:08, <tiejun.chen@xxxxxxxxx> wrote:
At first I doubted this is issued by some improper cache behaviors.
Because as you see, "root_entry[0] = 80f5001" indicates we already set
that present bit. But Caching Mode bit is zero in BDW so this means
remapping hardware doesn't own contest cache. And I also check xen
always calls clflush to write back memory on CPU side. (Even I also
tried to use wbinvd to replace clflush). And plus, you can see this
guest fault address is a weird zero, so I turned to concern if this is
just an unexpected fault which is pending because of some BIOS cleanup,
or undefined hardware behaviors. Here I mean clear_fault_bits() is
already not enough to clear this kind of pending interrupt before we
complete all VT-D initializations.

This last aspect in particular needs more explanation, but I'm
wary of the change in general (see Yang's response).


As we discussed offline, Yang though at most this is a way to workaround our problem. So I would ping BIOS team to double check if this is an issue specific to some given BIOSs or platforms since this doesn't meet the current specification


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.