Re: [Xen-devel] What's the benefit and design purpose of Xen Console for HVM guest.

[Wei Xu <wei.xu.prc@xxxxxxxxx>]

thanks, comments inlined.

On Sat, Mar 30, 2013 at 12:30 AM, Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx> wrote:

CC'ing back xen-devel.

On Fri, 29 Mar 2013, Wei Xu wrote:

> I have test the kgdb kernel debugging via serial, it works good when i used the physical serial to connect the 2 HVM guests(host
> and target), is it possible to debugging kernel via Xen Console?

I know very little about kgdb but I know that other people on the list
use it. I guess it should be possible.

> I know xen console will related to different file in "/dev/pts" for different HVM guest, so this is impossible by now, is it
> possible to fit it?  that means xen console is a full replacement of the emulated serial.

I don't understand what you mean here. Xen PV consoles also correspond to
some /dev/pts devices in dom0.

 

yes, i know, kgdb need a pipe-like serial pair to pass data between the 2 guests, while xenconsoled tells the PV console was linked to different files in /dev/pts bundled with only one guest, so i don't know if it is possible to link/pipe them together.

> what i am wondering most is why did Xen Console come to mind in the early date, that's why did it raise? Was it going to replace
> the emulated serial, or just an additional way to access the guest with high performance or sth other.

Two main reasons come to mind: if you need additional console for
logging, PV consoles provide you with a solution.
If you need to debug an PV on HVM kernel very early on at resume time,
PV consoles can be re-initialized much earlier than emulated serials.

Why emulated serial can't be re-initialized during resume time? i'm a little confused.

beyond the 2 reasons, why performance improvement isn't the main concern? or some other reasones unmentioned here?

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