diff mbox

[2/2] systemd: drop dependency on proc-xen.mount in xendriverdomain.service

Message ID 20160724192713.GB22529@mail-itl (mailing list archive)
State New, archived
Headers show

Commit Message

Marek Marczykowski-Górecki July 24, 2016, 7:27 p.m. UTC
It is no longer required since xl devd use /dev/xen interface.

Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
---
 tools/hotplug/Linux/systemd/xendriverdomain.service.in | 2 --
 1 file changed, 2 deletions(-)

Comments

Wei Liu July 25, 2016, 9:27 a.m. UTC | #1
On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
> It is no longer required since xl devd use /dev/xen interface.
> 

How would this unit work when there is no /dev/xen interface?

To be precise, we prefer /dev/xen interfaces whenever possible but there
is a fallback to /proc/xen. Note that a lot of other unit files have
this dependency on proc-xen.mount.

I'm inclined to say we should keep this dependency but I'm not sure if I
missed some obvious things.

> Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> Cc: Wei Liu <wei.liu2@citrix.com>
> Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
> ---
>  tools/hotplug/Linux/systemd/xendriverdomain.service.in | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/tools/hotplug/Linux/systemd/xendriverdomain.service.in b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> index 0afb54d..a100309 100644
> --- a/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> +++ b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> @@ -1,7 +1,5 @@
>  [Unit]
>  Description=Xen driver domain device daemon
> -Requires=proc-xen.mount
> -After=proc-xen.mount
>  ConditionVirtualization=xen
>  
>  [Service]
> -- 
> 2.5.5
Marek Marczykowski-Górecki July 25, 2016, 9:30 a.m. UTC | #2
On Mon, Jul 25, 2016 at 10:27:47AM +0100, Wei Liu wrote:
> On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
> > It is no longer required since xl devd use /dev/xen interface.
> > 
> 
> How would this unit work when there is no /dev/xen interface?

Does it happen in reality? I thought /proc/xen is deprecated for a long
time...

> To be precise, we prefer /dev/xen interfaces whenever possible but there
> is a fallback to /proc/xen. Note that a lot of other unit files have
> this dependency on proc-xen.mount.
> 
> I'm inclined to say we should keep this dependency but I'm not sure if I
> missed some obvious things.
> 
> > Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> > Cc: Wei Liu <wei.liu2@citrix.com>
> > Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
> > ---
> >  tools/hotplug/Linux/systemd/xendriverdomain.service.in | 2 --
> >  1 file changed, 2 deletions(-)
> > 
> > diff --git a/tools/hotplug/Linux/systemd/xendriverdomain.service.in b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > index 0afb54d..a100309 100644
> > --- a/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > +++ b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > @@ -1,7 +1,5 @@
> >  [Unit]
> >  Description=Xen driver domain device daemon
> > -Requires=proc-xen.mount
> > -After=proc-xen.mount
> >  ConditionVirtualization=xen
> >  
> >  [Service]
> > -- 
> > 2.5.5
> 
>
Wei Liu July 25, 2016, 9:34 a.m. UTC | #3
On Mon, Jul 25, 2016 at 11:30:32AM +0200, Marek Marczykowski-Górecki wrote:
> On Mon, Jul 25, 2016 at 10:27:47AM +0100, Wei Liu wrote:
> > On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
> > > It is no longer required since xl devd use /dev/xen interface.
> > > 
> > 
> > How would this unit work when there is no /dev/xen interface?
> 
> Does it happen in reality? I thought /proc/xen is deprecated for a long
> time...
> 

I can't tell whether it happens in reality or not. I also hope to get
rid of deprecated interface but they are here to stay for as long as it
takes.

I guess the question is, do you have a compelling reason against keeping
this dependency?

Wei.

> > To be precise, we prefer /dev/xen interfaces whenever possible but there
> > is a fallback to /proc/xen. Note that a lot of other unit files have
> > this dependency on proc-xen.mount.
> > 
> > I'm inclined to say we should keep this dependency but I'm not sure if I
> > missed some obvious things.
> > 
> > > Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> > > Cc: Wei Liu <wei.liu2@citrix.com>
> > > Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
> > > ---
> > >  tools/hotplug/Linux/systemd/xendriverdomain.service.in | 2 --
> > >  1 file changed, 2 deletions(-)
> > > 
> > > diff --git a/tools/hotplug/Linux/systemd/xendriverdomain.service.in b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > > index 0afb54d..a100309 100644
> > > --- a/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > > +++ b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
> > > @@ -1,7 +1,5 @@
> > >  [Unit]
> > >  Description=Xen driver domain device daemon
> > > -Requires=proc-xen.mount
> > > -After=proc-xen.mount
> > >  ConditionVirtualization=xen
> > >  
> > >  [Service]
> > > -- 
> > > 2.5.5
> > 
> > 
> 
> -- 
> Best Regards,
> Marek Marczykowski-Górecki
> Invisible Things Lab
> A: Because it messes up the order in which people normally read text.
> Q: Why is top-posting such a bad thing?
Marek Marczykowski-Górecki July 25, 2016, 9:40 a.m. UTC | #4
On Mon, Jul 25, 2016 at 10:34:21AM +0100, Wei Liu wrote:
> On Mon, Jul 25, 2016 at 11:30:32AM +0200, Marek Marczykowski-Górecki wrote:
> > On Mon, Jul 25, 2016 at 10:27:47AM +0100, Wei Liu wrote:
> > > On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
> > > > It is no longer required since xl devd use /dev/xen interface.
> > > > 
> > > 
> > > How would this unit work when there is no /dev/xen interface?
> > 
> > Does it happen in reality? I thought /proc/xen is deprecated for a long
> > time...
> > 
> 
> I can't tell whether it happens in reality or not. I also hope to get
> rid of deprecated interface but they are here to stay for as long as it
> takes.
> 
> I guess the question is, do you have a compelling reason against keeping
> this dependency?

Just cleaning up old stuff, so - no.
Andrew Cooper July 25, 2016, 9:45 a.m. UTC | #5
On 25/07/16 10:34, Wei Liu wrote:
> On Mon, Jul 25, 2016 at 11:30:32AM +0200, Marek Marczykowski-Górecki wrote:
>> On Mon, Jul 25, 2016 at 10:27:47AM +0100, Wei Liu wrote:
>>> On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
>>>> It is no longer required since xl devd use /dev/xen interface.
>>>>
>>> How would this unit work when there is no /dev/xen interface?
>> Does it happen in reality? I thought /proc/xen is deprecated for a long
>> time...

It really should be, but isn't.

/proc/xen/capabilities and /proc/xen/allsyms (?) don't have any
equivalent in /dev/

>>
> I can't tell whether it happens in reality or not. I also hope to get
> rid of deprecated interface but they are here to stay for as long as it
> takes.
>
> I guess the question is, do you have a compelling reason against keeping
> this dependency?

We really should be making xenfs deprecated and able to be compiled out
of Linux.  It is a hangover of the early classic kernels which isn't
necessary, and causes subtle/weird failures if it isn't mounted.

~Andrew
Wei Liu July 25, 2016, 9:49 a.m. UTC | #6
On Mon, Jul 25, 2016 at 10:45:54AM +0100, Andrew Cooper wrote:
> On 25/07/16 10:34, Wei Liu wrote:
> > On Mon, Jul 25, 2016 at 11:30:32AM +0200, Marek Marczykowski-Górecki wrote:
> >> On Mon, Jul 25, 2016 at 10:27:47AM +0100, Wei Liu wrote:
> >>> On Sun, Jul 24, 2016 at 09:27:13PM +0200, Marek Marczykowski-Górecki wrote:
> >>>> It is no longer required since xl devd use /dev/xen interface.
> >>>>
> >>> How would this unit work when there is no /dev/xen interface?
> >> Does it happen in reality? I thought /proc/xen is deprecated for a long
> >> time...
> 
> It really should be, but isn't.
> 
> /proc/xen/capabilities and /proc/xen/allsyms (?) don't have any
> equivalent in /dev/
> 
> >>
> > I can't tell whether it happens in reality or not. I also hope to get
> > rid of deprecated interface but they are here to stay for as long as it
> > takes.
> >
> > I guess the question is, do you have a compelling reason against keeping
> > this dependency?
> 
> We really should be making xenfs deprecated and able to be compiled out
> of Linux.  It is a hangover of the early classic kernels which isn't
> necessary, and causes subtle/weird failures if it isn't mounted.
> 

+1 from me of course!

When the kernel is ready, I shall remove the fallback in libraries and
then we can remove all references to proc-xen.mount and even
proc-xen.mount itself.

Wei.

> ~Andrew
diff mbox

Patch

diff --git a/tools/hotplug/Linux/systemd/xendriverdomain.service.in b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
index 0afb54d..a100309 100644
--- a/tools/hotplug/Linux/systemd/xendriverdomain.service.in
+++ b/tools/hotplug/Linux/systemd/xendriverdomain.service.in
@@ -1,7 +1,5 @@ 
 [Unit]
 Description=Xen driver domain device daemon
-Requires=proc-xen.mount
-After=proc-xen.mount
 ConditionVirtualization=xen
 
 [Service]