On Fri, Jul 27, 2018 at 02:34:41PM +0200, Michal Privoznik wrote:
On 07/27/2018 10:37 AM, Erik Skultety wrote:
> On Thu, Jul 26, 2018 at 05:36:28PM +0200, Michal Privoznik wrote:
>> This way it will be easier to use autofree.
>>
>> Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com>
>> ---
>> src/lxc/lxc_process.c | 24 +++++++++---------------
>> 1 file changed, 9 insertions(+), 15 deletions(-)
>>
>> diff --git a/src/lxc/lxc_process.c b/src/lxc/lxc_process.c
>> index d021a890f7..3ac39d598c 100644
>> --- a/src/lxc/lxc_process.c
>> +++ b/src/lxc/lxc_process.c
>> @@ -514,8 +514,7 @@ static int virLXCProcessSetupNamespaces(virConnectPtr conn,
>> * virLXCProcessSetupInterfaces:
>> * @conn: pointer to connection
>> * @def: pointer to virtual machine structure
>> - * @nveths: number of interfaces
>> - * @veths: interface names
>> + * @veths: string list of interface names
>> *
>> * Sets up the container interfaces by creating the veth device pairs and
>> * attaching the parent end to the appropriate bridge. The container end
>> @@ -525,7 +524,6 @@ static int virLXCProcessSetupNamespaces(virConnectPtr conn,
>> */
>> static int virLXCProcessSetupInterfaces(virConnectPtr conn,
>> virDomainDefPtr def,
>> - size_t *nveths,
>> char ***veths)
>> {
>> int ret = -1;
>> @@ -534,6 +532,8 @@ static int virLXCProcessSetupInterfaces(virConnectPtr conn,
>> virDomainNetDefPtr net;
>> virDomainNetType type;
>>
>> + *veths = NULL;
>> +
>> for (i = 0; i < def->nnets; i++) {
>> char *veth = NULL;
>> virNetDevBandwidthPtr actualBandwidth;
>> @@ -549,9 +549,6 @@ static int virLXCProcessSetupInterfaces(virConnectPtr conn,
>> if (virDomainNetAllocateActualDevice(def, net) < 0)
>> goto cleanup;
>>
>> - if (VIR_EXPAND_N(*veths, *nveths, 1) < 0)
>> - goto cleanup;
>
> Contrary to what I said in the previous patch, I hadn't realized we were
> expanding a list by 1 in a loop before I looked at this patch. That is very
> inefficient and we'll keep doing that. Now I'm biased towards tracking the
size
> of the list so that we can do VIR_EXPAND_N(*veths, 0, def->nnets) and then just
> add new elements, of course that would require tweaking the virStringListAdd
> more.
>
Hold on. I'm not quite sure I follow. Are you suggesting to pre-allocate
the string list and ditch virStringListAdd completely? Something like
this?
Yes, that's the idea. Originally, I didn't think dropping virStringListAdd would
be necessary, but it's clear from the snippet below that it would.
Erik
diff --git i/src/lxc/lxc_process.c w/src/lxc/lxc_process.c
index 6eef17d1ce..33c806630b 100644
--- i/src/lxc/lxc_process.c
+++ w/src/lxc/lxc_process.c
@@ -532,7 +532,8 @@ static int virLXCProcessSetupInterfaces(virConnectPtr conn,
virDomainNetDefPtr net;
virDomainNetType type;
- *veths = NULL;
+ if (VIR_ALLOC_N(*veths, def->nnets + 1) < 0)
+ return -1;
for (i = 0; i < def->nnets; i++) {
char *veth = NULL;
@@ -601,8 +602,7 @@ static int virLXCProcessSetupInterfaces(virConnectPtr conn,
}
}
- if (virStringListAdd(veths, veth) < 0)
- goto cleanup;
+ (*veths)[i] = veth;
if (VIR_STRDUP(def->nets[i]->ifname_guest_actual, veth) < 0)
goto cleanup;
Michal
--
libvir-list mailing list
libvir-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list