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

Re: [PATCH v5 09/21] libxl: add save/restore support for qemu-xen in stubdomain

  • To: Jason Andryuk <jandryuk@xxxxxxxxx>
  • From: Ian Jackson <ian.jackson@xxxxxxxxxx>
  • Date: Mon, 18 May 2020 15:15:17 +0100
  • Authentication-results: esa6.hc3370-68.iphmx.com; dkim=none (message not signed) header.i=none
  • Cc: Anthony Perard <anthony.perard@xxxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>, Marek Marczykowski-Górecki <marmarek@xxxxxxxxxxxxxxxxxxxxxx>, Wei Liu <wl@xxxxxxx>
  • Delivery-date: Mon, 18 May 2020 14:15:58 +0000
  • Ironport-sdr: BY2xgvnlCalyodz+iXE45gv4FW9rpW8N0PDth5UsoT4uc30prNkruBZA3qODL4gOUoSOIxLr7p fPgAAF2jB/gA82R0n06my6dq7XVTaoifauTrrHHzBmA66m/Hd5/j8xjFGCMST2kBJDMroWlzWH eQPn/nCkyOWqUGylu6XlmHI952EiWzXrcZN57wqTrewWqWdaxYtjRVnSZkX88nklouo/+brlpZ VPEBZpeKQAC6jEyKIpWdfet1wcKDFnot9Use0tq32PR+ndLY1/nfuGP6Qq9SMfs/Y7eXCwmXBG YbE=
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

Jason Andryuk writes ("Re: [PATCH v5 09/21] libxl: add save/restore support for 
qemu-xen in stubdomain"):
> On Thu, May 14, 2020 at 12:35 PM Ian Jackson <ian.jackson@xxxxxxxxxx> wrote:
> > I suggest randomly allocating one in the range [64,192>.  My random
> > number generator picked 119.  So 118 and 119 ?
> This makes sense and would be the easiest change.


> > Also, why couldn't your wrapper script add this argument ?  If you do
> > that there then there is one place that knows the fd number and a
> > slightly less tortuous linkage between libxl and the script...
> I like this idea, but there is a complication.  "-incoming" is only
> added when performing a restore, so it cannot just be blindly added to
> all qemu command lines in the stubdom.  Two options I see are to
> either communicate a restore some other way (so the stubdom scripts
> can add the appropriate option), or pass something though dm_args, but
> let the script convert it into something usable.
> There is "-incoming defer" where we can later specify
> "migrate_incoming fd:119".  Another option is to `sed
> s/defer/fd:119/`, but that is a little tricky since we need to look at
> the preceding key to know if we should sed the second.  We could pass
> only "-incoming" and require the stubdom script to modify that option.
> I haven't tested any of this.

Erk.  I see now why you did it the way you did !

> > It's not stated anywhere here that I can see but I think what is
> > happening here is that your wrapper script knows the qemu savefile
> > pathname and reads it directly.  Maybbe a comment would be
> > worthwhile ?
> The existing comment "Linux stubdomain connects specific FD to
> STUBDOM_CONSOLE_RESTORE" is trying to state that.
> STUBDOM_CONSOLE_RESTORE is defined as 2 for console 2 (/dev/hvc2), but
> it is only a libxl_internal.h define.

Err, by "the qemu savefile pathname" I meant the pathname in dom0.
I assume your wrapper script opens that and feeds it to the console.
Is that right ?




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