Re: [Xen-devel] [RFC] x86: possible problem with guest_walk_tables()

["Jan Beulich" <JBeulich@xxxxxxxxxx>]

>>> On 21.03.11 at 14:57, Tim Deegan <Tim.Deegan@xxxxxxxxxx> wrote:
> At 13:10 +0000 on 21 Mar (1300713029), Jan Beulich wrote:
>> Yes, if this is really only GFN space, then it would "just" result in
>> bad translations getting installed, possibly conflicting with others
>> (e.g. in cache attributes).
> 
> Yep; that's no worse that anything else a guest can do. I'm pretty sure
> no guest installs unaligned PAE entries and then tries to use them. (The
> only case I _know_ of where the OS relies on h/w to enforce invalid-bit 
> detection is Xen itself).

... and luckily Xen itself no longer runs on non-PAE (so this doesn't
become an issue with nested virtualization).

>> Leaving aside the non-PAE case, does the fix presented look
>> reasonable?
>  
> Yes, it looks correct to me.  Give me a Signed-off-by and I'll apply it. 

Going with the page-fault-less case then for non-PAE? If we want
that case to produce reserved bit faults too, I admit I'd prefer to
submit a complete patch (which is half the reason why this was an
RFC, non-signed-off one)... If we want no fault there, I'd like to
comment that way in the code (so the or-ing with zero won't
prompt later readers to wonder whether this isn't a bug/oversight).

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel