[PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Razvan Cojocaru]

VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
not available in Xen 4.6, hence the bump.

Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
---
 xen/include/public/xen-compat.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/xen/include/public/xen-compat.h b/xen/include/public/xen-compat.h
index 1e62dc1..590de76 100644
--- a/xen/include/public/xen-compat.h
+++ b/xen/include/public/xen-compat.h
@@ -27,7 +27,7 @@
 #ifndef __XEN_PUBLIC_XEN_COMPAT_H__
 #define __XEN_PUBLIC_XEN_COMPAT_H__
 
-#define __XEN_LATEST_INTERFACE_VERSION__ 0x00040600
+#define __XEN_LATEST_INTERFACE_VERSION__ 0x00040700
 
 #if defined(__XEN__) || defined(__XEN_TOOLS__)
 /* Xen is built with matching headers and implements the latest interface. */
-- 
1.7.9.5

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Jan Beulich]

>>> On 30.09.15 at 17:16, <rcojocaru@bitdefender.com> wrote:
> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
> not available in Xen 4.6, hence the bump.

I don't follow: These are additions, not changes that require
consumers to adapt their code.

Jan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Julien Grall]

Hi,

On 30/09/15 16:16, Razvan Cojocaru wrote:
> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
> not available in Xen 4.6, hence the bump.

Shouldn't you bump XEN_DOMCTL_INTERFACE_VERSION and
VM_EVENT_INTERFACE_VERSION instead?

The former for xc_monitor_emulate_each_read as it's a DOMCTL and the
latter for VM_EVENT_FLAG_SET_REGISTERS?

Regards,

-- 
Julien Grall

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Razvan Cojocaru]

On 09/30/2015 06:23 PM, Jan Beulich wrote:
>>>> On 30.09.15 at 17:16, <rcojocaru@bitdefender.com> wrote:
>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>> not available in Xen 4.6, hence the bump.
> 
> I don't follow: These are additions, not changes that require
> consumers to adapt their code.

I have code that checks for __XEN_LATEST_INTERFACE_VERSION__ and if it
is 0x00040700 then it uses xc_monitor_emulate_each_rep(), otherwise it
knows it can't - I thought that consumers that want to make use of the
latest API should be able to tell that they're allowed to do so.

But if the Xen convention is to only bump it when the interface is no
longer backward compatible then I've misunderstood (in which case, sorry
for the noise, and also, is there another way to tell which Xen version
I'm compiling against?).


Thanks,
Razvan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Razvan Cojocaru]

On 09/30/2015 06:25 PM, Julien Grall wrote:
> Hi,
> 
> On 30/09/15 16:16, Razvan Cojocaru wrote:
>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>> not available in Xen 4.6, hence the bump.
> 
> Shouldn't you bump XEN_DOMCTL_INTERFACE_VERSION and
> VM_EVENT_INTERFACE_VERSION instead?
> 
> The former for xc_monitor_emulate_each_read as it's a DOMCTL and the
> latter for VM_EVENT_FLAG_SET_REGISTERS?

Yes, judging also by Jan's previous comment that would appear to be the
preferred way to go.


Thanks,
Razvan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Jan Beulich]

>>> On 30.09.15 at 17:28, <rcojocaru@bitdefender.com> wrote:
> On 09/30/2015 06:25 PM, Julien Grall wrote:
>> Hi,
>> 
>> On 30/09/15 16:16, Razvan Cojocaru wrote:
>>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>>> not available in Xen 4.6, hence the bump.
>> 
>> Shouldn't you bump XEN_DOMCTL_INTERFACE_VERSION and
>> VM_EVENT_INTERFACE_VERSION instead?
>> 
>> The former for xc_monitor_emulate_each_read as it's a DOMCTL and the
>> latter for VM_EVENT_FLAG_SET_REGISTERS?
> 
> Yes, judging also by Jan's previous comment that would appear to be the
> preferred way to go.

I can't talk about VM_EVENT_INTERFACE_VERSION, but I also
don't see any reason to bump XEN_DOMCTL_INTERFACE_VERSION
with the given change. Again - it's just an addition to the interface,
not an incompatible change.

Jan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Razvan Cojocaru]

On 09/30/2015 06:34 PM, Jan Beulich wrote:
>>>> On 30.09.15 at 17:28, <rcojocaru@bitdefender.com> wrote:
>> On 09/30/2015 06:25 PM, Julien Grall wrote:
>>> Hi,
>>>
>>> On 30/09/15 16:16, Razvan Cojocaru wrote:
>>>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>>>> not available in Xen 4.6, hence the bump.
>>>
>>> Shouldn't you bump XEN_DOMCTL_INTERFACE_VERSION and
>>> VM_EVENT_INTERFACE_VERSION instead?
>>>
>>> The former for xc_monitor_emulate_each_read as it's a DOMCTL and the
>>> latter for VM_EVENT_FLAG_SET_REGISTERS?
>>
>> Yes, judging also by Jan's previous comment that would appear to be the
>> preferred way to go.
> 
> I can't talk about VM_EVENT_INTERFACE_VERSION, but I also
> don't see any reason to bump XEN_DOMCTL_INTERFACE_VERSION
> with the given change. Again - it's just an addition to the interface,
> not an incompatible change.

The same is true for VM_EVENT_FLAG_SET_REGISTERS I'm afraid. Nothing at
all changes for existing vm_event users. In that case, I'll just leave
them all alone.

But the question still remains, if these markers are unreliable in my
case, could you please suggest another which I can check in my scenario?
__XEN_LATEST_INTERFACE_VERSION__ has been fine so far, but I now see
that this has only been a coincidence.


Thanks,
Razvan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Jan Beulich]

>>> On 30.09.15 at 17:27, <rcojocaru@bitdefender.com> wrote:
> On 09/30/2015 06:23 PM, Jan Beulich wrote:
>>>>> On 30.09.15 at 17:16, <rcojocaru@bitdefender.com> wrote:
>>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>>> not available in Xen 4.6, hence the bump.
>> 
>> I don't follow: These are additions, not changes that require
>> consumers to adapt their code.
> 
> I have code that checks for __XEN_LATEST_INTERFACE_VERSION__ and if it
> is 0x00040700 then it uses xc_monitor_emulate_each_rep(), otherwise it
> knows it can't - I thought that consumers that want to make use of the
> latest API should be able to tell that they're allowed to do so.
> 
> But if the Xen convention is to only bump it when the interface is no
> longer backward compatible then I've misunderstood (in which case, sorry
> for the noise, and also, is there another way to tell which Xen version
> I'm compiling against?).

Just check for the relevant #define to be there.

Jan

Re: [PATCH] xen: Bump __XEN_LATEST_INTERFACE_VERSION__ to 0x00040700

[Razvan Cojocaru]

On 09/30/2015 06:39 PM, Jan Beulich wrote:
>>>> On 30.09.15 at 17:27, <rcojocaru@bitdefender.com> wrote:
>> On 09/30/2015 06:23 PM, Jan Beulich wrote:
>>>>>> On 30.09.15 at 17:16, <rcojocaru@bitdefender.com> wrote:
>>>> VM_EVENT_FLAG_SET_REGISTERS and xc_monitor_emulate_each_rep() are
>>>> not available in Xen 4.6, hence the bump.
>>>
>>> I don't follow: These are additions, not changes that require
>>> consumers to adapt their code.
>>
>> I have code that checks for __XEN_LATEST_INTERFACE_VERSION__ and if it
>> is 0x00040700 then it uses xc_monitor_emulate_each_rep(), otherwise it
>> knows it can't - I thought that consumers that want to make use of the
>> latest API should be able to tell that they're allowed to do so.
>>
>> But if the Xen convention is to only bump it when the interface is no
>> longer backward compatible then I've misunderstood (in which case, sorry
>> for the noise, and also, is there another way to tell which Xen version
>> I'm compiling against?).
> 
> Just check for the relevant #define to be there.

That does work for VM_EVENT_FLAG_SET_REGISTERS in this case, but it
wouldn't have worked if the only addition would have been
xc_monitor_emulate_each_rep(). I like the XEN_VERSION and XEN_SUBVERSION
#defines in compile.h, but those don't make it into the public headers
(that end up in dist/install/usr/include after make dist).

But yes, in this case that #define is helpful, and then with autotools
or something similar libxc can be checked for xc_monitor_emulate_each_rep().

Again, sorry for the noise and thanks for the help.


Thanks,
Razvan