[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [XEN PATCH] libxc/migration: Abort migration on precopy policy request



On 4/7/20 10:22 PM, Wei Liu wrote:
> On Tue, Apr 07, 2020 at 02:52:22PM +0000, Andrew Panyakin wrote:
>> libxc defines XGS_POLICY_ABORT for precopy policy to signal that migration
>> should be aborted (eg. if the estimated pause time is too huge for the
>> instance). Default simple precopy policy never returns that, but it could be
>> overriden with a custom one.
>>
> 
> Right. I think this is a real problem.
> 
>> Signed-off-by: Andrew Panyakin <apanyaki@xxxxxxxxxx>
>> ---
>>  tools/libxc/xc_sr_save.c | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/tools/libxc/xc_sr_save.c b/tools/libxc/xc_sr_save.c
>> index fa736a311f..507274ce22 100644
>> --- a/tools/libxc/xc_sr_save.c
>> +++ b/tools/libxc/xc_sr_save.c
>> @@ -560,6 +560,12 @@ static int send_memory_live(struct xc_sr_context *ctx)
>>
>>      }
>>
>> +    if ( policy_decision == XGS_POLICY_ABORT ) {
> 
> The { should be on a new line.
> 
>> +        PERROR("Abort precopy loop");
>> +        rc = -1;
>> +        goto out;
> 
> There is no need to have "goto out" here.

I was considering two more examples of "goto out" in a branch right before the 
label:
- send_domain_memory_nonlive,
- send_domain_memory_live.

Isn't it done this way to simplify the function extension: you won't need to 
add "goto out" to previous branch when adding new code?

> 
> These can be fixed easily while committing, so no need to resend yet. I
> will give other people a chance to comment.
> 
> Wei.
> 
>> +    }
>> +
>>   out:
>>      xc_set_progress_prefix(xch, NULL);
>>      free(progress_str);
>> --
>> 2.16.6
>>


-- 
Andrew



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.