[libvirt] [PATCH] conf: Remove the actual hostdev when removing a network
Commit 50348e6edfa reused the code to remove the hostdev portion of a network definition on multiple places but forgot to take into account that sometimes the "actual" network is passed and in some cases the parent of that. This patch uses the virDomainNetGetActualHostdev() helper to acquire the correct pointer all the time while removing the hostdev portion from the list. --- src/conf/domain_conf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index e412f28..cef4cf9 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -10002,14 +10002,14 @@ void virDomainNetRemoveHostdev(virDomainDefPtr def, virDomainNetDefPtr net) { - if (net->type == VIR_DOMAIN_NET_TYPE_HOSTDEV) { - /* hostdev net devices are normally also be in the hostdevs - * array, but might have already been removed by the time we - * get here. - */ - virDomainHostdevDefPtr hostdev = &net->data.hostdev.def; - size_t i; + /* hostdev net devices are normally also be in the hostdevs + * array, but might have already been removed by the time we + * get here. + */ + virDomainHostdevDefPtr hostdev = virDomainNetGetActualHostdev(net); + size_t i; + if (hostdev) { for (i = 0; i < def->nhostdevs; i++) { if (def->hostdevs[i] == hostdev) { virDomainHostdevRemove(def, i); -- 1.8.3.2
On 09/03/2013 09:29 AM, Peter Krempa wrote:
Commit 50348e6edfa reused the code to remove the hostdev portion of a network definition on multiple places but forgot to take into account that sometimes the "actual" network is passed and in some cases the parent of that.
This patch uses the virDomainNetGetActualHostdev() helper to acquire the correct pointer all the time while removing the hostdev portion from the list. --- src/conf/domain_conf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
ACK. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
On 09/03/13 19:14, Eric Blake wrote:
On 09/03/2013 09:29 AM, Peter Krempa wrote:
Commit 50348e6edfa reused the code to remove the hostdev portion of a network definition on multiple places but forgot to take into account that sometimes the "actual" network is passed and in some cases the parent of that.
This patch uses the virDomainNetGetActualHostdev() helper to acquire the correct pointer all the time while removing the hostdev portion from the list. --- src/conf/domain_conf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
ACK.
Pushed; Thanks. Peter
On 09/03/2013 11:29 AM, Peter Krempa wrote:
Commit 50348e6edfa reused the code to remove the hostdev portion of a network definition on multiple places but forgot to take into account that sometimes the "actual" network is passed and in some cases the parent of that.
This patch uses the virDomainNetGetActualHostdev() helper to acquire the correct pointer all the time while removing the hostdev portion from the list. --- src/conf/domain_conf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index e412f28..cef4cf9 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -10002,14 +10002,14 @@ void virDomainNetRemoveHostdev(virDomainDefPtr def, virDomainNetDefPtr net) { - if (net->type == VIR_DOMAIN_NET_TYPE_HOSTDEV) { - /* hostdev net devices are normally also be in the hostdevs - * array, but might have already been removed by the time we - * get here. - */ - virDomainHostdevDefPtr hostdev = &net->data.hostdev.def; - size_t i; + /* hostdev net devices are normally also be in the hostdevs + * array, but might have already been removed by the time we + * get here. + */ + virDomainHostdevDefPtr hostdev = virDomainNetGetActualHostdev(net); + size_t i;
+ if (hostdev) { for (i = 0; i < def->nhostdevs; i++) { if (def->hostdevs[i] == hostdev) { virDomainHostdevRemove(def, i);
This causes a crash for me on failed guest startup. XML attached, guest is legitimately failing because libvirt doesn't know about the CPU model it specifies. Reverting this patch fixes the crash. Snipped backtrace: #0 virDomainNetGetActualHostdev (iface=iface@entry=0x7fffd41d7930) at conf/domain_conf.c:17879 #1 0x00007ffff7562f10 in virDomainNetRemoveHostdev ( def=def@entry=0x7fffd41c7400, net=net@entry=0x7fffd41d7930) at conf/domain_conf.c:10009 #2 0x00007fffddda5d65 in qemuProcessStop (driver=driver@entry=0x7fffd409f0c0, vm=vm@entry=0x7fffd41c7ea0, reason=reason@entry=VIR_DOMAIN_SHUTOFF_FAILED, flags=flags@entry=2) at qemu/qemu_process.c:4270 #3 0x00007fffddda7e01 in qemuProcessStart (conn=conn@entry=0x7fffd0000a80, driver=driver@entry=0x7fffd409f0c0, vm=vm@entry=0x7fffd41c7ea0, migrateFrom=migrateFrom@entry=0x0, stdin_fd=stdin_fd@entry=-1, stdin_path=stdin_path@entry=0x0, snapshot=snapshot@entry=0x0, vmop=vmop@entry=VIR_NETDEV_VPORT_PROFILE_OP_CREATE, flags=<optimized out>, flags@entry=1) at qemu/qemu_process.c:4051 - Cole
On 09/04/13 17:17, Cole Robinson wrote:
On 09/03/2013 11:29 AM, Peter Krempa wrote:
Commit 50348e6edfa reused the code to remove the hostdev portion of a network definition on multiple places but forgot to take into account that sometimes the "actual" network is passed and in some cases the parent of that.
This patch uses the virDomainNetGetActualHostdev() helper to acquire the correct pointer all the time while removing the hostdev portion from the list. --- src/conf/domain_conf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index e412f28..cef4cf9 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -10002,14 +10002,14 @@ void virDomainNetRemoveHostdev(virDomainDefPtr def, virDomainNetDefPtr net) { - if (net->type == VIR_DOMAIN_NET_TYPE_HOSTDEV) { - /* hostdev net devices are normally also be in the hostdevs - * array, but might have already been removed by the time we - * get here. - */ - virDomainHostdevDefPtr hostdev = &net->data.hostdev.def; - size_t i; + /* hostdev net devices are normally also be in the hostdevs + * array, but might have already been removed by the time we + * get here. + */ + virDomainHostdevDefPtr hostdev = virDomainNetGetActualHostdev(net); + size_t i;
+ if (hostdev) { for (i = 0; i < def->nhostdevs; i++) { if (def->hostdevs[i] == hostdev) { virDomainHostdevRemove(def, i);
This causes a crash for me on failed guest startup. XML attached, guest is legitimately failing because libvirt doesn't know about the CPU model it specifies. Reverting this patch fixes the crash. Snipped backtrace:
#0 virDomainNetGetActualHostdev (iface=iface@entry=0x7fffd41d7930) at conf/domain_conf.c:17879 #1 0x00007ffff7562f10 in virDomainNetRemoveHostdev ( def=def@entry=0x7fffd41c7400, net=net@entry=0x7fffd41d7930) at conf/domain_conf.c:10009 #2 0x00007fffddda5d65 in qemuProcessStop (driver=driver@entry=0x7fffd409f0c0, vm=vm@entry=0x7fffd41c7ea0, reason=reason@entry=VIR_DOMAIN_SHUTOFF_FAILED, flags=flags@entry=2) at qemu/qemu_process.c:4270 #3 0x00007fffddda7e01 in qemuProcessStart (conn=conn@entry=0x7fffd0000a80, driver=driver@entry=0x7fffd409f0c0, vm=vm@entry=0x7fffd41c7ea0, migrateFrom=migrateFrom@entry=0x0, stdin_fd=stdin_fd@entry=-1, stdin_path=stdin_path@entry=0x0, snapshot=snapshot@entry=0x0, vmop=vmop@entry=VIR_NETDEV_VPORT_PROFILE_OP_CREATE, flags=<optimized out>, flags@entry=1) at qemu/qemu_process.c:4051
- Cole
d'oh, virDomainNetGetActualHostdev isn't bulletproof in case the actual network interface isn't allocated. I'll post a patch asap. Peter
participants (3)
-
Cole Robinson -
Eric Blake -
Peter Krempa