On 13.07.26 14:17, Daniel P. Berrangé wrote:
On Fri, Jul 10, 2026 at 10:18:01PM +0300, Vladimir Sementsov-Ogievskiy wrote:
Support transferring of TAP state (including open fd).
Add new property "local-migration-supported", which defines
whether local-migration is actually supported for this TAP device.
Starting from 11.2 QEMU Machine Types it's enabled by default.
Note, that local-migration (including migrating opened FDs
through migration channel, which must be UNIX socket), is
enabled by global "local" migration parameters. But individual
devices may have additional options to enable/disable it
per device.
The tricky thing, is that we need to know, should we do
call open/connect in TAP initialization code. That means,
that on incoming migration we want to know status of
"local" migration parameter at time of creating TAP device.
So, currently, for incoming local migration only TAP device
created through QMP is supported, as you need to set migration
parameter first (through QMP as well).
So the full picture is:
On source, to start outgoing "local" migration you need:
- migration parameter "local" set to true
- "local-migration-supported" TAP option set to true (the
default, starting from 11.2 QEMU Machine Types)
If at least one of these options is not set, TAP backend
doesn't participate in migration.
On target, things are more difficult:
Same, you need both "local" and "local-migration-supported"
be set. And same, if one of them is not set, TAP backend
is initialized as usual, and doesn't accept any incoming
state.
Additionally, if you are going to set "local", it must be
set before creating the TAP device. If TAP device created
with "local" unset, it initializes as usual. If you enable
"local" after it and start incoming migration, it will fail
in .pre_load handler of TAP backend.
Moreover, there are interface restrictions: if you create TAP
device when QEMU is in INCOMING state, and both "local"
and "local-migration-supported" set, most of TAP options are
not allowed, and script/downscript are required to be explicitly
unset (set to "" or "no").
Signed-off-by: Vladimir Sementsov-Ogievskiy <[email protected]>
---
In v17 thread we considered making new qapi parameters "unstable".
Still, it was because of doubtful "incoming-fds" parameter. In this
v18 it's completely removed, and we may live only with "local"
migration parameter + "local-migration-supported" TAP option.
This way the series doesn't conflict with suggested by Peter
[PATCH RFC 0/2] migration/vl: new -incoming config:* for early migration
parameters
So, I think, worth making one more iteration as "stable candidate".
If still some doubts, not problem for me to resend with "unstable" features.
hw/core/machine.c | 7 ++
net/tap.c | 162 ++++++++++++++++++++++++++++++++++++++++++++--
qapi/net.json | 23 ++++++-
3 files changed, 185 insertions(+), 7 deletions(-)
diff --git a/hw/core/machine.c b/hw/core/machine.c
index 15886a56b19..2a1abb29041 100644
--- a/hw/core/machine.c
+++ b/hw/core/machine.c
@@ -36,10 +36,16 @@
#include "hw/virtio/virtio-pci.h"
#include "hw/virtio/virtio-net.h"
#include "hw/virtio/virtio-iommu.h"
+#include "net/tap.h"
#include "hw/acpi/generic_event_device.h"
#include "qemu/audio.h"
#include "hw/arm/smmuv3.h"
+/*
+ * TODO: When hw_compat_11_1 appears, local-migration-supported=false
+ * for TYPE_TAP_NETDEV should be move to it from hw_compat_11_0.
+ */
+
GlobalProperty hw_compat_11_0[] = {
{ "chardev-vc", "encoding", "cp437" },
{ "tpm-crb", "cap-chunk", "off" },
@@ -49,6 +55,7 @@ GlobalProperty hw_compat_11_0[] = {
{ TYPE_ARM_SMMUV3, "ssidsize", "0" },
{ TYPE_ARM_SMMUV3, "oas", "44" },
{ "migration", "switchover-ack-legacy", "on" },
+ { TYPE_TAP_NETDEV, "local-migration-supported", "false" },
So disabled by default for historical machines type versions, but....
+static void tap_instance_init(Object *obj)
+{
+ TAPState *s = TAP_NETDEV(obj);
+ s->local_migration_supported = true;
+}
..enabled by default for all future machine type versions, but...
@@ -1005,6 +1123,23 @@ int net_init_tap(const Netdev *netdev, const char *name,
return -1;
}
+ incoming_fds = tap->local_migration_supported && migrate_local() &&
+ runstate_check(RUN_STATE_INMIGRATE);
+
+ if (incoming_fds &&
+ (tap->fd || tap->fds || tap->helper || tap->br || tap->ifname ||
+ tap->has_sndbuf || tap->has_vnet_hdr ||
+ !tap_is_explicit_no_script(tap->script) ||
+ !tap_is_explicit_no_script(tap->downscript))) {
+ error_setg(errp, "Local incoming migration of TAP device (-incoming, "
+ "migration parameter @local is set, "
+ "TAP parameter @local-migration-supported is set) "
+ "is incompatible with "
+ "fd=, fds=, helper=, br=, ifname=, sndbuf= and vnet_hdr=, "
+ "and requires explicit empty script= and downscript=");
+ return -1;
+ }
...not usable by default unless the mgmt app knows to avoid these settings,
but the user won't find out about this limit until they try to migrate and
it fails at runtime, months or years after the VM was first configured &
launched :-(
No, it fail only for RUN_STATE_INMIGRATE, and only if "local" is enabled.
If mgmt enables "local" it must know about the restrictions of using it.
AFAICT, there's also no way for mgmt apps to determine which parameters
are forbidden, because it is merely exposed in API docs text, no way to
introspect
Unfortunately yes.
IMHO the sane way to do this is to not have any "local-migration-supported"
device flag at all, and also no "local" migration parameter. The src & dst
QEMU's so badly need to be able to auto-negotiate and just do the right
thing, otherwise this local migration optimization becomes a mgmt nightmare
long term :-(
"local" is needed because we can't know, does transfer support fd-passing.
To auto-negotiate it, we probably need to pass a test FD to the target, as
simply checking "is it a UNIX socket" may not be enough.
"local-migration-supported" may be auto negotiated of course, if we have
auto negotiation.
And yet we don't have any auto-negotiate mechanism, and I don't want to
block your work :-(
Thanks) But finally, in short term, could we proceed with it as is, or not?
Would it be better, if I drop the default =true for new machine types, keeping
it false?
--
Best regards,
Vladimir
