Re: [PATCH RFC] xen/console: buffer and show origin of guest PV writes

[Daniel De Graaf <dgdegra@tycho.nsa.gov>]

On 08/13/2013 12:52 PM, Jan Beulich wrote:
>>>> On 13.08.13 at 18:34, Daniel De Graaf <dgdegra@tycho.nsa.gov> wrote:
>> 1. HVM guests' output is currently limited to printable characters; do
>> we want to implement the same restriction on PV guests?
>
> Would seem to make sense.

It seems only dom0 will try to use control characters under normal
conditions, so this shouldn't cause any issues (hopefully).

>> 2. The prefix on the serial console for PV output is "(XEN) d5: ", while
>> HVM output is still "(XEN) HVM5: "; should these be made consistent?
>
> Uniformly using e.g. "(Dom%d) " instead of "(XEN) " for domain output
> might be best?

I think this makes sense as a way to distinguish guest output from Xen
output regarding a guest. This will probably require introducing a
vprintk-ish function to add the guest domid argument.

The original version of this patch used "(%d) ", but I think "(d%d) " or
"(dom%d) " may end up being clearer.

>> 3. Should we change to allowing console output by default, since it is
>> now controlled by log levels?
>
> I don't think we should, not the least because guest output doesn't
> really specify a log level, and default log level messages make it
> through with default options.

The existing HVM guest output is done with XENLOG_G_DEBUG, which was
copied for the PV guest output. The default value of guest_loglvl on a
debug build includes this output, but on a non-debug build hides it.

> Will have to look at the patch itself later, but you should have Cc-ed
> Keir in any case (as he will eventually need to ack it).
>
> Jan

Ah, for some reason I didn't have him on the original list. I have added
him to this message to highlight the thread and will fix that on a resend.

-- 
Daniel De Graaf
National Security Agency