[libvirt PATCH 0/3] Support for restarting passt backend
If the passt process that implements the backend of a QEMU emulated network device terminates, QEMU itself is incapable of restarting a new process so that the networking can begin... working again. However, what QEMU can and does do is: 1) send the NETDEV_STREAM_DISCONNECTED event to libvirt when it sees that the socket has been closed (this event has been in QEMU for as long as support for "-netdev stream", which is used for connecting to passt) 2) as of QEMU 8.0.0, the qemu commandline for -netdev stream accepts the "reconnect" option, which tells QEMU to attempt reconnecting to the same socket it previously used, repeating the attempt every "n" seconds (the only argument to reconnect) until it is successful. If libvirt adds the reconnect option to the qemu commandline, and then responds to a NETDEV_STREAM_DISCONNECTED event by re-running the same passt command that it ran when the device was originally connected, then a guest will be able to recover in the (very unlikely, according to Stefano!) event that the original passt process unexpectedly exits, or is killed by some external entity. Patch 2/3 handles (2) above, while patch 3/3 handles (1). (patch 1/3 is a short guest appearance by pkrempa. Thanks pkrempa!). This resolves https://bugzilla.redhat.com/2172098 Along with Stefano's series fixing up selinux issues related to running the passt process, they make the passt backend very usable. Laine Stump (2): qemu: add reconnect=5 to passt qemu commandline options when available qemu: respond to NETDEV_STREAM_DISCONNECTED event Peter Krempa (1): qemu: capabilities: Introduce QEMU_CAPS_NETDEV_STREAM_RECONNECT src/qemu/qemu_capabilities.c | 4 + src/qemu/qemu_capabilities.h | 3 + src/qemu/qemu_domain.c | 1 + src/qemu/qemu_domain.h | 1 + src/qemu/qemu_driver.c | 82 +++++++++++++++++++ src/qemu/qemu_monitor.c | 11 +++ src/qemu/qemu_monitor.h | 6 ++ src/qemu/qemu_monitor_json.c | 16 ++++ src/qemu/qemu_passt.c | 11 +++ src/qemu/qemu_process.c | 18 ++++ .../caps_8.0.0.x86_64.xml | 1 + .../net-user-passt.x86_64-7.2.0.args | 37 +++++++++ .../net-user-passt.x86_64-latest.args | 2 +- tests/qemuxml2argvtest.c | 1 + 14 files changed, 193 insertions(+), 1 deletion(-) create mode 100644 tests/qemuxml2argvdata/net-user-passt.x86_64-7.2.0.args -- 2.39.2
From: Peter Krempa
QEMU's "reconnect" option of "-netdev stream" tells QEMU to
periodically (period is given in seconds as an argument to the option)
attempt to reconnect to the same passt socket to which it had
originally connected to. This is useful in cases where the passt
process terminates, and libvirtd starts a new passt process in its
place (which doesn't happen yet, but will happen automatically after
an upcoming patch in this series).
Since there is no real hueristic for determining the "best" value of
the reconnect interval, rather than clutter up config with a knob that
nobody knows how to properly twiddle, we just set the reconnect timer
to 5 seconds.
"-netdev stream" first appeared in QEMU 7.2.0, but the reconnect
option won't be available until QEMU 8.0.0, so we need to check QEMU
capabilities just in case someone is using QEMU 7.2.0 (and thus can
support passt backend, but not reconnect)
Signed-off-by: Laine Stump
When a QEMU netdev is of type "stream", if the socket it uses for
connectivity to the host network gets closed, then QEMU will send a
NETDEV_STREAM_DISCONNECTED event. We know that any stream netdev we've
created is backed by a passt process, and if the socket was closed,
that means the passt process has disappeared.
When we receive this event, we can respond by starting a new passt
process with the same options (including socket path) we originally
used. If we have previously created the stream netdev device with a
"reconnect" option, then QEMU will automatically reconnect to this new
passt process. (If we hadn't used "reconnect", then QEMU will never
try to reconnect to the new passt process, so there's no point in
starting it.)
Note that NETDEV_STREAM_DISCONNECTED is an event sent for the netdev
(ie "host side") of the network device, and so it sends the
"netdev-id" to specify which device was disconnected. But libvirt's
virDomainNetDef (the object used to keep track of network devices) is
the internal representation of both the host-side "netdev", and the
guest side device, and virDomainNetDef doesn't directly keep track of
the netdev-id, only of the device's "alias" (which is the "id"
parameter of the *guest* side of the device). Fortunately, by convention
libvirt always names the host-side of devices as "host" + alias, so in
order to search for the affected NetDef, all we need to do is trim the
1st 4 characters from the netdev-id and look for the NetDef having
that resulting trimmed string as its alias. (Contrast this to
NIC_RX_FILTER_CHANGED, which is an event received for the guest side
of the device, and so directly contains the device alias.)
Resolves: https://bugzilla.redhat.com/2172098
Signed-off-by: Laine Stump
On 2/22/23 01:35, Laine Stump wrote:
When a QEMU netdev is of type "stream", if the socket it uses for connectivity to the host network gets closed, then QEMU will send a NETDEV_STREAM_DISCONNECTED event. We know that any stream netdev we've created is backed by a passt process, and if the socket was closed, that means the passt process has disappeared.
When we receive this event, we can respond by starting a new passt process with the same options (including socket path) we originally used. If we have previously created the stream netdev device with a "reconnect" option, then QEMU will automatically reconnect to this new passt process. (If we hadn't used "reconnect", then QEMU will never try to reconnect to the new passt process, so there's no point in starting it.)
Note that NETDEV_STREAM_DISCONNECTED is an event sent for the netdev (ie "host side") of the network device, and so it sends the "netdev-id" to specify which device was disconnected. But libvirt's virDomainNetDef (the object used to keep track of network devices) is the internal representation of both the host-side "netdev", and the guest side device, and virDomainNetDef doesn't directly keep track of the netdev-id, only of the device's "alias" (which is the "id" parameter of the *guest* side of the device). Fortunately, by convention libvirt always names the host-side of devices as "host" + alias, so in order to search for the affected NetDef, all we need to do is trim the 1st 4 characters from the netdev-id and look for the NetDef having that resulting trimmed string as its alias. (Contrast this to NIC_RX_FILTER_CHANGED, which is an event received for the guest side of the device, and so directly contains the device alias.)
Resolves: https://bugzilla.redhat.com/2172098 Signed-off-by: Laine Stump
--- src/qemu/qemu_domain.c | 1 + src/qemu/qemu_domain.h | 1 + src/qemu/qemu_driver.c | 82 ++++++++++++++++++++++++++++++++++++ src/qemu/qemu_monitor.c | 11 +++++ src/qemu/qemu_monitor.h | 6 +++ src/qemu/qemu_monitor_json.c | 16 +++++++ src/qemu/qemu_process.c | 18 ++++++++ 7 files changed, 135 insertions(+) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index e9bc0f375d..4cf9a259ea 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -11238,6 +11238,7 @@ qemuProcessEventFree(struct qemuProcessEvent *event) break; case QEMU_PROCESS_EVENT_WATCHDOG: case QEMU_PROCESS_EVENT_DEVICE_DELETED: + case QEMU_PROCESS_EVENT_NETDEV_STREAM_DISCONNECTED: case QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED: case QEMU_PROCESS_EVENT_SERIAL_CHANGED: case QEMU_PROCESS_EVENT_MONITOR_EOF: diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h index 1053d1d4cb..6adc067681 100644 --- a/src/qemu/qemu_domain.h +++ b/src/qemu/qemu_domain.h @@ -447,6 +447,7 @@ typedef enum { QEMU_PROCESS_EVENT_WATCHDOG = 0, QEMU_PROCESS_EVENT_GUESTPANIC, QEMU_PROCESS_EVENT_DEVICE_DELETED, + QEMU_PROCESS_EVENT_NETDEV_STREAM_DISCONNECTED, QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED, QEMU_PROCESS_EVENT_SERIAL_CHANGED, QEMU_PROCESS_EVENT_JOB_STATUS_CHANGE, diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 6154fe9bfe..47d6a0dd95 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -40,6 +40,7 @@ #include "qemu_hostdev.h" #include "qemu_hotplug.h" #include "qemu_monitor.h" +#include "qemu_passt.h" #include "qemu_process.h" #include "qemu_migration.h" #include "qemu_migration_params.h" @@ -3622,6 +3623,84 @@ processDeviceDeletedEvent(virQEMUDriver *driver, }
+static void +processNetdevStreamDisconnectedEvent(virDomainObj *vm, + const char *netdevId) +{ + virDomainDeviceDef dev; + virDomainNetDef *def; + qemuDomainObjPrivate *priv; + virQEMUCaps *qemuCaps; + const char *devAlias = NULL; + + /* The event sends us the "netdev-id", but we don't store the + * netdev-id in the NetDef and thus can't use it to find the + * correct NetDef. We *do* keep the device alias in the NetDef. + * By definition, the netdev-id is "host" + devAlias, so we just + * need to remove "host" from the front of netdev-id to get + * something we can use to find the proper NetDef. + */ + if (STREQLEN(netdevId, "host", 4)) + devAlias = &netdevId[4];
This is open coding STRSKIP(): devAlias = STRSKIP(netdevId, "host");
+ + if (!devAlias) { + VIR_WARN("Received NETDEV_STREAM_DISCONNECTED event for unrecognized netdev %s from domain %p %s", + netdevId, vm, vm->def->name); + return; + } + + VIR_DEBUG("Received NETDEV_STREAM_DISCONNECTED event for device %s from domain %p %s", + devAlias, vm, vm->def->name); + + if (virDomainObjBeginJob(vm, VIR_JOB_QUERY) < 0) + return; + + if (!virDomainObjIsActive(vm)) { + VIR_DEBUG("Domain is not running"); + goto endjob; + } + + if (virDomainDefFindDevice(vm->def, devAlias, &dev, true) < 0) { + VIR_WARN("NETDEV_STREAM_DISCONNECTED event received for non-existent device %s in domain %s", + devAlias, vm->def->name); + goto endjob; + } + if (dev.type != VIR_DOMAIN_DEVICE_NET) { + VIR_WARN("NETDEV_STREAM_DISCONNECTED event received for non-network device %s in domain %s", + devAlias, vm->def->name); + goto endjob; + } + def = dev.data.net; + + if (def->backend.type != VIR_DOMAIN_NET_BACKEND_PASST) { + VIR_DEBUG("ignore NETDEV_STREAM_DISCONNECTED event for non-passt network device %s in domain %s", + def->info.alias, vm->def->name); + goto endjob; + } + + priv = vm->privateData; + qemuCaps = priv->qemuCaps; +
These two can be done in variable declaration. Alternatively, if you want to lose @priv variable completely you can go with: virQEMUCaps *caps = QEMU_DOMAIN_PRIVATE(vm)->qemuCaps;
+ if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_NETDEV_STREAM_RECONNECT)) { + VIR_WARN("ignore NETDEV_STREAM_DISCONNECTED event for passt network device %s in domain %s - QEMU binary does not support reconnect", + def->info.alias, vm->def->name); + goto endjob; + } + + /* handle the event - restart the passt process with its original + * parameters + */ + VIR_DEBUG("process NETDEV_STREAM_DISCONNECTED event for network device %s in domain %s", + def->info.alias, vm->def->name); + + if (qemuPasstStart(vm, def) < 0) + goto endjob; + + endjob: + virDomainObjEndJob(vm); +} +
Michal
On 2/22/23 5:21 AM, Michal Prívozník wrote:
On 2/22/23 01:35, Laine Stump wrote:
When a QEMU netdev is of type "stream", if the socket it uses for connectivity to the host network gets closed, then QEMU will send a NETDEV_STREAM_DISCONNECTED event. We know that any stream netdev we've created is backed by a passt process, and if the socket was closed, that means the passt process has disappeared.
When we receive this event, we can respond by starting a new passt process with the same options (including socket path) we originally used. If we have previously created the stream netdev device with a "reconnect" option, then QEMU will automatically reconnect to this new passt process. (If we hadn't used "reconnect", then QEMU will never try to reconnect to the new passt process, so there's no point in starting it.)
Note that NETDEV_STREAM_DISCONNECTED is an event sent for the netdev (ie "host side") of the network device, and so it sends the "netdev-id" to specify which device was disconnected. But libvirt's virDomainNetDef (the object used to keep track of network devices) is the internal representation of both the host-side "netdev", and the guest side device, and virDomainNetDef doesn't directly keep track of the netdev-id, only of the device's "alias" (which is the "id" parameter of the *guest* side of the device). Fortunately, by convention libvirt always names the host-side of devices as "host" + alias, so in order to search for the affected NetDef, all we need to do is trim the 1st 4 characters from the netdev-id and look for the NetDef having that resulting trimmed string as its alias. (Contrast this to NIC_RX_FILTER_CHANGED, which is an event received for the guest side of the device, and so directly contains the device alias.)
Resolves: https://bugzilla.redhat.com/2172098 Signed-off-by: Laine Stump
--- src/qemu/qemu_domain.c | 1 + src/qemu/qemu_domain.h | 1 + src/qemu/qemu_driver.c | 82 ++++++++++++++++++++++++++++++++++++ src/qemu/qemu_monitor.c | 11 +++++ src/qemu/qemu_monitor.h | 6 +++ src/qemu/qemu_monitor_json.c | 16 +++++++ src/qemu/qemu_process.c | 18 ++++++++ 7 files changed, 135 insertions(+) diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c index e9bc0f375d..4cf9a259ea 100644 --- a/src/qemu/qemu_domain.c +++ b/src/qemu/qemu_domain.c @@ -11238,6 +11238,7 @@ qemuProcessEventFree(struct qemuProcessEvent *event) break; case QEMU_PROCESS_EVENT_WATCHDOG: case QEMU_PROCESS_EVENT_DEVICE_DELETED: + case QEMU_PROCESS_EVENT_NETDEV_STREAM_DISCONNECTED: case QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED: case QEMU_PROCESS_EVENT_SERIAL_CHANGED: case QEMU_PROCESS_EVENT_MONITOR_EOF: diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h index 1053d1d4cb..6adc067681 100644 --- a/src/qemu/qemu_domain.h +++ b/src/qemu/qemu_domain.h @@ -447,6 +447,7 @@ typedef enum { QEMU_PROCESS_EVENT_WATCHDOG = 0, QEMU_PROCESS_EVENT_GUESTPANIC, QEMU_PROCESS_EVENT_DEVICE_DELETED, + QEMU_PROCESS_EVENT_NETDEV_STREAM_DISCONNECTED, QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED, QEMU_PROCESS_EVENT_SERIAL_CHANGED, QEMU_PROCESS_EVENT_JOB_STATUS_CHANGE, diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index 6154fe9bfe..47d6a0dd95 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -40,6 +40,7 @@ #include "qemu_hostdev.h" #include "qemu_hotplug.h" #include "qemu_monitor.h" +#include "qemu_passt.h" #include "qemu_process.h" #include "qemu_migration.h" #include "qemu_migration_params.h" @@ -3622,6 +3623,84 @@ processDeviceDeletedEvent(virQEMUDriver *driver, }
+static void +processNetdevStreamDisconnectedEvent(virDomainObj *vm, + const char *netdevId) +{ + virDomainDeviceDef dev; + virDomainNetDef *def; + qemuDomainObjPrivate *priv; + virQEMUCaps *qemuCaps; + const char *devAlias = NULL; + + /* The event sends us the "netdev-id", but we don't store the + * netdev-id in the NetDef and thus can't use it to find the + * correct NetDef. We *do* keep the device alias in the NetDef. + * By definition, the netdev-id is "host" + devAlias, so we just + * need to remove "host" from the front of netdev-id to get + * something we can use to find the proper NetDef. + */ + if (STREQLEN(netdevId, "host", 4)) + devAlias = &netdevId[4];
This is open coding STRSKIP():
devAlias = STRSKIP(netdevId, "host");
Ah yes, right you are! Changed.
+ + if (!devAlias) { + VIR_WARN("Received NETDEV_STREAM_DISCONNECTED event for unrecognized netdev %s from domain %p %s", + netdevId, vm, vm->def->name); + return; + } + + VIR_DEBUG("Received NETDEV_STREAM_DISCONNECTED event for device %s from domain %p %s", + devAlias, vm, vm->def->name); + + if (virDomainObjBeginJob(vm, VIR_JOB_QUERY) < 0) + return; + + if (!virDomainObjIsActive(vm)) { + VIR_DEBUG("Domain is not running"); + goto endjob; + } + + if (virDomainDefFindDevice(vm->def, devAlias, &dev, true) < 0) { + VIR_WARN("NETDEV_STREAM_DISCONNECTED event received for non-existent device %s in domain %s", + devAlias, vm->def->name); + goto endjob; + } + if (dev.type != VIR_DOMAIN_DEVICE_NET) { + VIR_WARN("NETDEV_STREAM_DISCONNECTED event received for non-network device %s in domain %s", + devAlias, vm->def->name); + goto endjob; + } + def = dev.data.net; + + if (def->backend.type != VIR_DOMAIN_NET_BACKEND_PASST) { + VIR_DEBUG("ignore NETDEV_STREAM_DISCONNECTED event for non-passt network device %s in domain %s", + def->info.alias, vm->def->name); + goto endjob; + } + + priv = vm->privateData; + qemuCaps = priv->qemuCaps; +
These two can be done in variable declaration. Alternatively, if you want to lose @priv variable completely you can go with:
virQEMUCaps *caps = QEMU_DOMAIN_PRIVATE(vm)->qemuCaps;
I had originally had them initialized, but moved the assignment down until it was actually needed because of some warped idea about saving cycles, or avoiding a double de-reference in case one of them might be NULL or something. But in retrospect, just delaying wouldn't have solved that problem anyway. I'm changing it to use QEMU_DOMAIN_PRIVATE before pushing. Thanks for the review!
+ if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_NETDEV_STREAM_RECONNECT)) { + VIR_WARN("ignore NETDEV_STREAM_DISCONNECTED event for passt network device %s in domain %s - QEMU binary does not support reconnect", + def->info.alias, vm->def->name); + goto endjob; + } + + /* handle the event - restart the passt process with its original + * parameters + */ + VIR_DEBUG("process NETDEV_STREAM_DISCONNECTED event for network device %s in domain %s", + def->info.alias, vm->def->name); + + if (qemuPasstStart(vm, def) < 0) + goto endjob; + + endjob: + virDomainObjEndJob(vm); +} +
Michal
On 2/22/23 01:35, Laine Stump wrote:
If the passt process that implements the backend of a QEMU emulated network device terminates, QEMU itself is incapable of restarting a new process so that the networking can begin... working again.
However, what QEMU can and does do is:
1) send the NETDEV_STREAM_DISCONNECTED event to libvirt when it sees that the socket has been closed (this event has been in QEMU for as long as support for "-netdev stream", which is used for connecting to passt)
2) as of QEMU 8.0.0, the qemu commandline for -netdev stream accepts the "reconnect" option, which tells QEMU to attempt reconnecting to the same socket it previously used, repeating the attempt every "n" seconds (the only argument to reconnect) until it is successful.
If libvirt adds the reconnect option to the qemu commandline, and then responds to a NETDEV_STREAM_DISCONNECTED event by re-running the same passt command that it ran when the device was originally connected, then a guest will be able to recover in the (very unlikely, according to Stefano!) event that the original passt process unexpectedly exits, or is killed by some external entity.
Patch 2/3 handles (2) above, while patch 3/3 handles (1). (patch 1/3 is a short guest appearance by pkrempa. Thanks pkrempa!).
This resolves https://bugzilla.redhat.com/2172098
Along with Stefano's series fixing up selinux issues related to running the passt process, they make the passt backend very usable.
Laine Stump (2): qemu: add reconnect=5 to passt qemu commandline options when available qemu: respond to NETDEV_STREAM_DISCONNECTED event
Peter Krempa (1): qemu: capabilities: Introduce QEMU_CAPS_NETDEV_STREAM_RECONNECT
src/qemu/qemu_capabilities.c | 4 + src/qemu/qemu_capabilities.h | 3 + src/qemu/qemu_domain.c | 1 + src/qemu/qemu_domain.h | 1 + src/qemu/qemu_driver.c | 82 +++++++++++++++++++ src/qemu/qemu_monitor.c | 11 +++ src/qemu/qemu_monitor.h | 6 ++ src/qemu/qemu_monitor_json.c | 16 ++++ src/qemu/qemu_passt.c | 11 +++ src/qemu/qemu_process.c | 18 ++++ .../caps_8.0.0.x86_64.xml | 1 + .../net-user-passt.x86_64-7.2.0.args | 37 +++++++++ .../net-user-passt.x86_64-latest.args | 2 +- tests/qemuxml2argvtest.c | 1 + 14 files changed, 193 insertions(+), 1 deletion(-) create mode 100644 tests/qemuxml2argvdata/net-user-passt.x86_64-7.2.0.args
Reviewed-by: Michal Privoznik
participants (2)
-
Laine Stump
-
Michal Prívozník