# HG changeset patch
# User Paul Durrant <paul.durrant@citrix.com>
# Date 1321880835 0
# Node ID ff6518e294a7daa552e33c4d40be7597667c67db
# Parent dbdc840f8f62db58321b5009e5e0f7833066386f
Add missing modprobe to xencommons
Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
diff -r dbdc840f8f62 -r ff6518e294a7 tools/hotplug/Linux/init.d/xencommons
--- a/tools/hotplug/Linux/init.d/xencommons Wed Nov 16 18:21:14 2011 +0000
+++ b/tools/hotplug/Linux/init.d/xencommons Mon Nov 21 13:07:15 2011 +0000
@@ -52,6 +52,7 @@ do_start () {
test -z "$XENSTORED_TRACE" || XENSTORED_ARGS=" -T
/var/log/xen/xenstored-trace.log"
echo -n Starting xenstored...
+ modprobe xen-evtchn
xenstored --pid-file=/var/run/xenstored.pid $XENSTORED_ARGS
# Wait for xenstored to actually come up, timing out after 30 seconds
Paul Durrant writes ("[Xen-devel] [PATCH] Add missing modprobe to
xencommons"):> Add missing modprobe to xencommons
...> + modprobe xen-evtchn
Is this really the way we want to do things ? I guess it''s a
plausible thing to do which is harmless if the module name doesn''t
change.
But perhaps something else in the system should be organising that
evtchn gets loaded ? Also, this isn''t the only module we need.
Ian.
On Thu, 2011-11-24 at 18:31 +0000, Ian Jackson wrote:> Paul Durrant writes ("[Xen-devel] [PATCH] Add missing modprobe to xencommons"): > > Add missing modprobe to xencommons > ... > > + modprobe xen-evtchn > > Is this really the way we want to do things ? I guess it''s a > plausible thing to do which is harmless if the module name doesn''t > change.I thought we already did this sort of thing somewhere (I thought it was xencommons) but I guess not.> But perhaps something else in the system should be organising that > evtchn gets loaded ?The problem is, I think, that there is no actual device which would trigger the Linux hotplug subsystem to try and load this module. There is a similar problem with things like /dev/loop. This has been a problem since udev removed (or at least obfuscated) the old behaviour of having a /dev node with no backing driver and modprobing on open. Debian has /etc/udev/links.conf which says: # This file does not exist. Please do not ask the Debian maintainer about it. # If you need manually created devices, create them in /lib/udev/devices/ . /lib/udev/devices is empty on my system and I can''t see any documentation about what one should create there (actual dev nodes?) In any case it appears to be Debian specific. We could perhaps cause the xenbus code or some other suitably core thing to generate a hotplug event to cause these things to be autoloaded? Also this is the sort of thing which (lib)xl could and should trivially sanity check for at start of day.> Also, this isn''t the only module we need. > > Ian. > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel
On Fri, 2011-11-25 at 09:33 +0000, Ian Campbell wrote:> /lib/udev/devices is empty on my system and I can''t see any > documentation about what one should create there (actual dev nodes?) In > any case it appears to be Debian specific.Actually it turns out it is a core udevd feature and is documented in udevd(8). So we could usefully create things here in our install target, if we wanted. That''s assuming that opening /dev/xen/evtchn turns into a suitable modprobe it should do, I think. Ian.
Ian Campbell writes ("Re: [Xen-devel] [PATCH] Add missing modprobe to
xencommons"):> On Fri, 2011-11-25 at 09:33 +0000, Ian Campbell wrote:
> > /lib/udev/devices is empty on my system and I can''t see any
> > documentation about what one should create there (actual dev nodes?)
In
> > any case it appears to be Debian specific.
>
> Actually it turns out it is a core udevd feature and is documented in
> udevd(8). So we could usefully create things here in our install target,
> if we wanted.
>
> That''s assuming that opening /dev/xen/evtchn turns into a suitable
> modprobe it should do, I think.
I think I will take the patch to modprobe in the init script.
Ian.
On Fri, Nov 25, Ian Jackson wrote:> Ian Campbell writes ("Re: [Xen-devel] [PATCH] Add missing modprobe to xencommons"): > > On Fri, 2011-11-25 at 09:33 +0000, Ian Campbell wrote: > > > /lib/udev/devices is empty on my system and I can''t see any > > > documentation about what one should create there (actual dev nodes?) In > > > any case it appears to be Debian specific. > > > > Actually it turns out it is a core udevd feature and is documented in > > udevd(8). So we could usefully create things here in our install target, > > if we wanted. > > > > That''s assuming that opening /dev/xen/evtchn turns into a suitable > > modprobe it should do, I think. > > I think I will take the patch to modprobe in the init script.This is what I have in a similar patch, it loads the relevant modules for pvops and xenlinux kernel: +modprobe xen-evtchn 2>/dev/null || : +modprobe xen-gntdev 2>/dev/null || : +modprobe evtchn 2>/dev/null || : +modprobe gntdev 2>/dev/null || : Olaf
Olaf Hering writes ("Re: [Xen-devel] [PATCH] Add missing modprobe to
xencommons"):> On Fri, Nov 25, Ian Jackson wrote:
> > I think I will take the patch to modprobe in the init script.
>
> This is what I have in a similar patch, it loads the relevant modules
> for pvops and xenlinux kernel:
>
> +modprobe xen-evtchn 2>/dev/null || :
> +modprobe xen-gntdev 2>/dev/null || :
> +modprobe evtchn 2>/dev/null || :
> +modprobe gntdev 2>/dev/null || :
Would you care to post that patch with a proper signed-off-by and a
commit comment ? That seems more comprehensive than Paul''s.
Thanks,
Ian.