[libvirt] [PATCH 1/4] parallels: Fix connecting to parallels driver

[Maxim Nestratov]

09.02.2015 16:02, Michal Privoznik пишет:
> On 30.01.2015 20:58, Maxim Nestratov wrote:
>> After 55ea7be7, failures of parallelsLoadNetworks and parallelsStorageOpen
>> started to fail establishing a new connection to parallels driver.
>> Before this patch we simply ignored such errors.
>> This patch restores the ability to connect to parallels driver
>> even if parallelsLoadNetworks fails to get information about
>> Parallels Bridged networks. This can happen when it is unconfigured
>> (which is a common case on developers hosts for instance) or configured
>> incompletely. In any case it is not a severe error and we can ignore it.
>> Failures of parallelsStorageOpen occured because we incorrectly treated
>> path to VM' configuration file as a directory. Now initialization of
>> parallels VM domains home directory is fixed.
>>
>> Signed-off-by: Maxim Nestratov <mnestratov at parallels.com>
>> ---
>>   src/parallels/parallels_network.c |    2 +-
>>   src/parallels/parallels_sdk.c     |    8 ++++++++
>>   2 files changed, 9 insertions(+), 1 deletions(-)
>>
>> diff --git a/src/parallels/parallels_network.c b/src/parallels/parallels_network.c
>> index 3e7087d..033b347 100644
>> --- a/src/parallels/parallels_network.c
>> +++ b/src/parallels/parallels_network.c
>> @@ -304,7 +304,7 @@ static int parallelsLoadNetworks(parallelsConnPtr privconn)
>>   
>>           net = parallelsLoadNetwork(privconn, jobj2);
>>           if (!net)
>> -            goto cleanup;
>> +            continue;
> I'm not sure this is correct. I mean, parallelsLoadNetwork() could have
> failed for so many reasons. And some of them are show stoppers (e.g. OOM).
Here I only wanted to say that failure of loading Bridged Virtual 
Network information is not a show stopper.
Maybe I had better sort out things inside parallelsLoadNetworks.
>>   
>>       }
>>   
>> diff --git a/src/parallels/parallels_sdk.c b/src/parallels/parallels_sdk.c
>> index d0d2ce2..b9d5ddf 100644
>> --- a/src/parallels/parallels_sdk.c
>> +++ b/src/parallels/parallels_sdk.c
>> @@ -1247,6 +1247,14 @@ prlsdkLoadDomain(parallelsConnPtr privconn,
>>       pret = PrlVmCfg_GetHomePath(sdkdom, pdom->home, &buflen);
>>       prlsdkCheckRetGoto(pret, error);
>>   
>> +    // For VMs pdom->home is actually /directory/config.pvs
>> +    if(!IS_CT(def)){
>> +        // Get rid of /config.pvs in path string
>> +        char *s = strrchr(pdom->home, '/');
>> +        if (s)
>> +            *s = '\0';
>> +    }
>> +
> We like comments in old way /* */, and we even like spaces around
> brackets and after 'if'-s.
Ok. I'll fix it.
>>       if (olddom) {
>>           /* assign new virDomainDef without any checks */
>>           /* we can't use virDomainObjAssignDef, because it checks
>>
> Michal