Re: port-xen/51080: Support pvfb on xen domu (patch)

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

The following reply was made to PR port-xen/51080; it has been noted by GNATS.

From: Manuel Bouyer <bouyer%antioche.eu.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: port-xen-maintainer%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
        netbsd-bugs%netbsd.org@localhost
Subject: Re: port-xen/51080: Support pvfb on xen domu (patch)
Date: Thu, 19 May 2016 11:17:36 +0200

 On Wed, Apr 20, 2016 at 10:20:00PM +0000, pablo%polvorin.com.ar@localhost wrote:
 > >Description:
 > The patch is available here http://blog.polvorin.com.ar/patches/pvfb.patch
 > 
 > is an attempt at supporting the virtual framebuffer on xen guest
 > ( http://wiki.netbsd.org/projects/project/xen-domu-pvfb/ )
 > 
 > Basic functionality is in place, but reached a point where guidance/feedback would be very welcome in order to push this forward,
 > this is the first piece of kernel code I attempt to write, and I'm not
 > that familiar with netbsd to start with.
 > 
 > Need to review, at least
 > 
 > * Memory management (conversion to frame number to pass references to xen host).  Currently save/restore isn't working,  but neither do it  without this patch on my install.
 > 
 > * code just act as a kind of adapter between xenbus and genfb, wskbddev and wsmousedev, letting them attach to xenbus. Is that the way to go?
 > 
 > * The way I added the drivers to be used by autoconf
 
 Hello,
 this looks good, good work !
 For the framebuffer resolution/depth, how does linux do ? Is it something it
 gets from the domU config file via the store ? Or via the command line ?
 AFAIK NetBSD's genfb doens't have a way to change the resolution from userland.
 If there's no provision to set the parameters in the xenstore you could pass
 it from the kernel command line, e.g. xenfb=1280x1024x32
 This will have to be parsed in xen_parse_cmdline(), which then can be called
 from xfb_xenbus.c.
 
 Are you sure the framebuffer has a protocol entry in the xenstore ?
 If not, you should probably remove it from xfb_xenbus_resume().
 
 For the race condition you mention in xfb_xenbus_attach() I don't understant
 how it can happens. The
 genfb framebuffer at 0xffff,  size  , depth,  stride
 line should always be there.
 Are you sure it's not some uninitialised variable ?
 
 -- 
 Manuel Bouyer <bouyer%antioche.eu.org@localhost>
      NetBSD: 26 ans d'experience feront toujours la difference
 --