Re: [libvirt] [PATCH 39/41] remote: refactor the code for choosing the UNIX socket path
On Tue, 2019-07-23 at 17:03 +0100, Daniel P. Berrangé wrote:
[...]
+++ b/src/remote/remote_driver.c
@@ -739,34 +739,35 @@ remoteConnectSupportsFeatureUnlocked(virConnectPtr conn,
+static char *
+remoteGetUNIXSocket(remoteDriverTransport transport,
+ unsigned int flags)
{
char *sockname = NULL;
- char *userdir = virGetUserRuntimeDirectory();
-
- if (!userdir)
- return NULL;
+ VIR_AUTOFREE(char *userdir);
This should be
VIR_AUTOFREE(char *) userdir = NULL;
@@ -964,6 +965,17 @@ doRemoteOpen(virConnectPtr conn,
+ if ((transport == REMOTE_DRIVER_TRANSPORT_UNIX ||
+ transport == REMOTE_DRIVER_TRANSPORT_SSH ||
+ transport == REMOTE_DRIVER_TRANSPORT_LIBSSH ||
+ transport == REMOTE_DRIVER_TRANSPORT_LIBSSH2) &&
+ !sockname &&
+ !(sockname = remoteGetUNIXSocket(transport, flags)))
+ goto failed;
The check for transport type should be implemented as a switch
statement for maximum futureproofness.
With at least the "userdir" declaration fixed,
Reviewed-by: Andrea Bolognani <abologna(a)redhat.com>
--
Andrea Bolognani / Red Hat / Virtualization