Re: [PATCH] xen: don't print error message in case of missing Xenstore entry

[Juergen Gross]

On 30/05/17 19:08, Boris Ostrovsky wrote:
> On 05/30/2017 11:03 AM, Juergen Gross wrote:
>> On 30/05/17 15:25, Boris Ostrovsky wrote:
>>> On 05/29/2017 05:13 AM, Juergen Gross wrote:
>>>> When registering for the Xenstore watch of the node control/sysrq the
>>>> handler will be called at once. Don't issue an error message if the
>>>> Xenstore node isn't there, as it will be created only when an event
>>>> is being triggered.
>>>>
>>>> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
>>>> ---
>>>>  drivers/xen/manage.c | 7 +++++--
>>>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/drivers/xen/manage.c b/drivers/xen/manage.c
>>>> index c1ec8ee80924..7ddd0803da23 100644
>>>> --- a/drivers/xen/manage.c
>>>> +++ b/drivers/xen/manage.c
>>>> @@ -277,8 +277,11 @@ static void sysrq_handler(struct xenbus_watch *watch, const char *path,
>>>>  	err = xenbus_transaction_start(&xbt);
>>>>  	if (err)
>>>>  		return;
>>>> -	if (xenbus_scanf(xbt, "control", "sysrq", "%c", &sysrq_key) < 0) {
>>>> -		pr_err("Unable to read sysrq code in control/sysrq\n");
>>>> +	err = xenbus_scanf(xbt, "control", "sysrq", "%c", &sysrq_key);
>>>> +	if (err < 0) {
>>>> +		if (err != -ENOENT)
>>> Can we distinguish initialization invocation from actual watch firing?
>>> E.g. '|| (system_state >= SYSTEM_RUNNING)'?
>> The watch will fire again after suspend/resume (e.g. live migration).
> 
> 
> That's unfortunate. (And system_state check would also not be a good
> solution btw since the watch might be processed by the watch thread
> after we enter SYSTEM_RUNNING).
> 
> Can you add a comment explaining why we are ignoring ENOENT?

Sure.


Juergen