The history of USB controller for ppc64 guest is complex and goes
back to libvirt 1.3.1 where the fun started.
Prior Libvirt 1.3.1 if no model for USB controller was specified
we've simply passed "-usb" on QEMU command line.
Since Libvirt 1.3.1 there is a patch (8156493d8db) that fixes this
issue by using "-device pci-ohci,..." but it breaks migration with
older Libvirts which was agreed that's acceptable. However this
patch didn't reflect this change in the domain XML and the model
was still missing.
Since Libvirt 2.2.0 there is a patch (f55eaccb0c5) that fixes the
issue with not setting the USB model into domain XML which we need
to know about to not break the migration and since the default
model was *pci-ohci* it was used as default in this patch as well.
This patch tries to take all the previous changes into account and
also change the default for newly defined domains that don't specify
any model for USB controller.
The VIR_DOMAIN_DEF_PARSE_ABI_UPDATE is set only if new domain is
defined or new device is added into a domain which means that in
all other cases we will use the old *pci-ohci* model instead of the
better and not broken *nec-usb-xhci* model.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1373184
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
---
src/qemu/qemu_domain.c | 16 ++++++++++++----
1 file changed, 12 insertions(+), 4 deletions(-)
diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index 10e831b779..64dbd0a491 100644
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -3012,7 +3012,8 @@ qemuDomainShmemDefPostParse(virDomainShmemDefPtr shm)
static int
qemuDomainControllerDefPostParse(virDomainControllerDefPtr cont,
const virDomainDef *def,
- virQEMUCapsPtr qemuCaps)
+ virQEMUCapsPtr qemuCaps,
+ unsigned int parseFlags)
{
switch ((virDomainControllerType)cont->type) {
case VIR_DOMAIN_CONTROLLER_TYPE_SCSI:
@@ -3040,9 +3041,16 @@ qemuDomainControllerDefPostParse(virDomainControllerDefPtr cont,
* address is found */
cont->model = VIR_DOMAIN_CONTROLLER_MODEL_USB_NONE;
} else if (ARCH_IS_PPC64(def->os.arch)) {
- /* Default USB controller for ppc64 is pci-ohci */
- if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_PCI_OHCI))
+ /* To not break migration we need to set default USB controller
+ * for ppc64 to pci-ohci if we cannot change ABI of the VM.
+ * The nec-usb-xhci controller is used as default only for
+ * newly defined domains or devices. */
+ if ((parseFlags & VIR_DOMAIN_DEF_PARSE_ABI_UPDATE) &&
+ virQEMUCapsGet(qemuCaps, QEMU_CAPS_NEC_USB_XHCI)) {
+ cont->model = VIR_DOMAIN_CONTROLLER_MODEL_USB_NEC_XHCI;
+ } else if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_PCI_OHCI)) {
cont->model = VIR_DOMAIN_CONTROLLER_MODEL_USB_PCI_OHCI;
+ }
} else {
/* Default USB controller for anything else is piix3-uhci */
if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_PIIX3_USB_UHCI))
@@ -3241,7 +3249,7 @@ qemuDomainDeviceDefPostParse(virDomainDeviceDefPtr dev,
if (dev->type == VIR_DOMAIN_DEVICE_CONTROLLER &&
qemuDomainControllerDefPostParse(dev->data.controller, def,
- qemuCaps) < 0)
+ qemuCaps, parseFlags) < 0)
goto cleanup;
if (dev->type == VIR_DOMAIN_DEVICE_SHMEM &&
--
2.12.0
--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list
On Thu, 2017-03-02 at 09:48 +0100, Pavel Hrdina wrote: > The history of USB controller for ppc64 guest is complex and goes > back to libvirt 1.3.1 where the fun started. > > Prior Libvirt 1.3.1 if no model for USB controller was specified > we've simply passed "-usb" on QEMU command line. > > Since Libvirt 1.3.1 there is a patch (8156493d8db) that fixes this > issue by using "-device pci-ohci,..." but it breaks migration with > older Libvirts which was agreed that's acceptable. However this > patch didn't reflect this change in the domain XML and the model > was still missing. Unfortunately, migrating a guest with <controller type='usb' index='0'> <address type='pci' .../> </controller> from libvirt 2.1.0 (which still allows not specifying the controller model) to a build of libvirt including this patch doesn't work as expected. On the source host, I run # virsh migrate \ mig-usb \ qemu+ssh://root@destination/system \ --live --persistent --copy-storage-all --compressed While migration is still running, I can check on the destination host and verify that: * the <controller> element has model='pci-ohci'; * the QEMU command line contains -device pci-ohci. However, after migration is complete, the <controller> element has model='nec-xhci' instead of model='pci-ohci', which means that power cycling the guest results in breaking the guest ABI. -- Andrea Bolognani / Red Hat / Virtualization -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Mon, Mar 06, 2017 at 08:11:40PM +0100, Andrea Bolognani wrote: > On Thu, 2017-03-02 at 09:48 +0100, Pavel Hrdina wrote: > > The history of USB controller for ppc64 guest is complex and goes > > back to libvirt 1.3.1 where the fun started. > > > > Prior Libvirt 1.3.1 if no model for USB controller was specified > > we've simply passed "-usb" on QEMU command line. > > > > Since Libvirt 1.3.1 there is a patch (8156493d8db) that fixes this > > issue by using "-device pci-ohci,..." but it breaks migration with > > older Libvirts which was agreed that's acceptable. However this > > patch didn't reflect this change in the domain XML and the model > > was still missing. > > Unfortunately, migrating a guest with > > <controller type='usb' index='0'> > <address type='pci' .../> > </controller> > > from libvirt 2.1.0 (which still allows not specifying the > controller model) to a build of libvirt including this patch > doesn't work as expected. > > On the source host, I run > > # virsh migrate \ > mig-usb \ > qemu+ssh://root@destination/system \ > --live --persistent --copy-storage-all --compressed > > While migration is still running, I can check on the > destination host and verify that: > > * the <controller> element has model='pci-ohci'; > * the QEMU command line contains -device pci-ohci. > > However, after migration is complete, the <controller> > element has model='nec-xhci' instead of model='pci-ohci', > which means that power cycling the guest results in > breaking the guest ABI. I'm not so sure that this is an ABI change. The guest ABI is to ensure that the same guest XML will always start the same QEMU guest. However the PERSISTENT migration can make ABI changes because it is the same as virsh dumpxml $domain > $domain.xml && copy the XML onto remote host and virsh define $domain.xml. This would also change the *model*. If this would be considered to be guest ABI stable it would mean that other changes done by using this flag would be wrong because they also modifies the persistent XML during migration. Pavel -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > However, after migration is complete, the <controller> > > element has model='nec-xhci' instead of model='pci-ohci', > > which means that power cycling the guest results in > > breaking the guest ABI. > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > that the same guest XML will always start the same QEMU guest. However > the PERSISTENT migration can make ABI changes because it is the same as > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > and virsh define $domain.xml. This would also change the *model*. > > If this would be considered to be guest ABI stable it would mean that > other changes done by using this flag would be wrong because they also > modifies the persistent XML during migration. I assume there are very good reasons for persistent migration to behave differently, but as a user I find it extremely surprising. Do you have any insight on the rationale behind allowing ABI changes when performing persistent migration? The only other example I could find of ABI update being used is in virDomainDefPostParseMemory(), and AFAIU adding or removing memory from a guest is not an ABI change. Did I miss other uses? -- Andrea Bolognani / Red Hat / Virtualization -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, Mar 07, 2017 at 02:55:43PM +0100, Andrea Bolognani wrote: > On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > > However, after migration is complete, the <controller> > > > element has model='nec-xhci' instead of model='pci-ohci', > > > which means that power cycling the guest results in > > > breaking the guest ABI. > > > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > > that the same guest XML will always start the same QEMU guest. However > > the PERSISTENT migration can make ABI changes because it is the same as > > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > > and virsh define $domain.xml. This would also change the *model*. > > > > If this would be considered to be guest ABI stable it would mean that > > other changes done by using this flag would be wrong because they also > > modifies the persistent XML during migration. > > I assume there are very good reasons for persistent > migration to behave differently, but as a user I find > it extremely surprising. Do you have any insight on the > rationale behind allowing ABI changes when performing > persistent migration? Well it's a different kind of migration which is almost similar to the virsh dumpxml --inactive && virsh define, with one more feature, migration hook that can update the XML before it's defined on the destination. I agree that it may seem that we should handle this type of migration to be guest ABI stable, on the other hand this basically defines new domain on the destination where we allow to do the ABI updates. I'm adding Dan to CC to get his opinion about the guest ABI. > The only other example I could find of ABI update being > used is in virDomainDefPostParseMemory(), and AFAIU adding > or removing memory from a guest is not an ABI change. Did > I miss other uses? There is also an update of USB addresses in qemuDomainAssignUSBAddresses. Pavel -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, Mar 07, 2017 at 03:32:37PM +0100, Pavel Hrdina wrote: > On Tue, Mar 07, 2017 at 02:55:43PM +0100, Andrea Bolognani wrote: > > On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > > > However, after migration is complete, the <controller> > > > > element has model='nec-xhci' instead of model='pci-ohci', > > > > which means that power cycling the guest results in > > > > breaking the guest ABI. > > > > > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > > > that the same guest XML will always start the same QEMU guest. However > > > the PERSISTENT migration can make ABI changes because it is the same as > > > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > > > and virsh define $domain.xml. This would also change the *model*. > > > > > > If this would be considered to be guest ABI stable it would mean that > > > other changes done by using this flag would be wrong because they also > > > modifies the persistent XML during migration. > > > > I assume there are very good reasons for persistent > > migration to behave differently, but as a user I find > > it extremely surprising. Do you have any insight on the > > rationale behind allowing ABI changes when performing > > persistent migration? > > Well it's a different kind of migration which is almost similar to the > virsh dumpxml --inactive && virsh define, with one more feature, migration > hook that can update the XML before it's defined on the destination. > > I agree that it may seem that we should handle this type of migration to > be guest ABI stable, on the other hand this basically defines new domain > on the destination where we allow to do the ABI updates. > > I'm adding Dan to CC to get his opinion about the guest ABI. So, there's no particular reason why things must be different for live vs persistent migration. The only fundamental requirement is that we must never change ABI underneath a running guest, which mandates the strong ABI guarantee for live migration & save/restore. For persistent/offline migration, there's no running guest, so from a technical POV we have more freedom in some sense. That said, a change in guest hardaware that can cause a guest to become unbootable is not a nice thing todo. For this reason, when you use 'defineXML' libvirt aims to immediately fill in (most) XML elements that would affect guest ABI. ie we assign addresses at define time, not start time, and we expand non-versioned machine types into versioned machine types, etc. So, yes, we should try to *not* change ABI across a dumpxml & define operational. When we first introduced PCI device addressing in libvirt, we had an old state where XML contained no addresses, and upon starting in new libvirt we have to assign addresses which might not be the same as those seen with previous libvirt before addressing was recorded. Once the addresses are assigned though, we must never change them. So... - if we recorded 'pci-ohci' in the XML, we must never change that to 'nec-xhci'. - if we didn't record 'pci-ochi' in the XML, we should fill in the model that is most likely to match what the previous default behaviour was, which would be 'pci-ochi' not 'nec-xhci' IIUC. Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://entangle-photo.org -o- http://search.cpan.org/~danberr/ :| -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, Mar 07, 2017 at 02:40:55PM +0000, Daniel P. Berrange wrote: > On Tue, Mar 07, 2017 at 03:32:37PM +0100, Pavel Hrdina wrote: > > On Tue, Mar 07, 2017 at 02:55:43PM +0100, Andrea Bolognani wrote: > > > On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > > > > However, after migration is complete, the <controller> > > > > > element has model='nec-xhci' instead of model='pci-ohci', > > > > > which means that power cycling the guest results in > > > > > breaking the guest ABI. > > > > > > > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > > > > that the same guest XML will always start the same QEMU guest. However > > > > the PERSISTENT migration can make ABI changes because it is the same as > > > > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > > > > and virsh define $domain.xml. This would also change the *model*. > > > > > > > > If this would be considered to be guest ABI stable it would mean that > > > > other changes done by using this flag would be wrong because they also > > > > modifies the persistent XML during migration. > > > > > > I assume there are very good reasons for persistent > > > migration to behave differently, but as a user I find > > > it extremely surprising. Do you have any insight on the > > > rationale behind allowing ABI changes when performing > > > persistent migration? > > > > Well it's a different kind of migration which is almost similar to the > > virsh dumpxml --inactive && virsh define, with one more feature, migration > > hook that can update the XML before it's defined on the destination. > > > > I agree that it may seem that we should handle this type of migration to > > be guest ABI stable, on the other hand this basically defines new domain > > on the destination where we allow to do the ABI updates. > > > > I'm adding Dan to CC to get his opinion about the guest ABI. > > So, there's no particular reason why things must be different for > live vs persistent migration. The only fundamental requirement is > that we must never change ABI underneath a running guest, which > mandates the strong ABI guarantee for live migration & save/restore. > > For persistent/offline migration, there's no running guest, so from > a technical POV we have more freedom in some sense. That said, a > change in guest hardaware that can cause a guest to become unbootable > is not a nice thing todo. > > For this reason, when you use 'defineXML' libvirt aims to immediately > fill in (most) XML elements that would affect guest ABI. ie we assign > addresses at define time, not start time, and we expand non-versioned > machine types into versioned machine types, etc. > > So, yes, we should try to *not* change ABI across a dumpxml & define > operational. > > When we first introduced PCI device addressing in libvirt, we had an > old state where XML contained no addresses, and upon starting in new > libvirt we have to assign addresses which might not be the same as > those seen with previous libvirt before addressing was recorded. Once > the addresses are assigned though, we must never change them. This is basically the same issue like with the PCI addresses. We used "-usb" instead of "-device ...", but when we switched to using the "-device ..." way we didn't record the default model in the XML. The default model is recorded in the XML since Libvirt 2.2.0. > So... > > - if we recorded 'pci-ohci' in the XML, we must never change that > to 'nec-xhci'. This will never happen. > - if we didn't record 'pci-ochi' in the XML, we should fill in the > model that is most likely to match what the previous default > behaviour was, which would be 'pci-ochi' not 'nec-xhci' IIUC. This patch tries to move from 'pci-ohci' to 'nec-xhci' as a default USB controller because the current default model is probably broken, see [1], and will never be fixed. In addition I would like to do this change in Libvirt so all other tools that uses Libvirt don't have to implement this change by themselves. In this case I think it's better to change the default from 'pci-ohci' to 'nec-xhci' if you define a new domain where the model is not specified. There is a concern whether we should allow ABI change for persistent migration which is more general issue that will affect some other changes that are already "allowed" for persistent migration. I'm starting to feel that we should remove the "VIR_DOMAIN_DEF_PARSE_ABI_UPDATE" from "virDomainDefParseNode" call in "qemuMigrationCookieXMLParse" function which currently allows to do ABI changes for persistent migration. Pavel [1] <https://bugzilla.redhat.com/show_bug.cgi?id=1212275> -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, Mar 07, 2017 at 04:15:03PM +0100, Pavel Hrdina wrote: > On Tue, Mar 07, 2017 at 02:40:55PM +0000, Daniel P. Berrange wrote: > > On Tue, Mar 07, 2017 at 03:32:37PM +0100, Pavel Hrdina wrote: > > > On Tue, Mar 07, 2017 at 02:55:43PM +0100, Andrea Bolognani wrote: > > > > On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > > > > > However, after migration is complete, the <controller> > > > > > > element has model='nec-xhci' instead of model='pci-ohci', > > > > > > which means that power cycling the guest results in > > > > > > breaking the guest ABI. > > > > > > > > > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > > > > > that the same guest XML will always start the same QEMU guest. However > > > > > the PERSISTENT migration can make ABI changes because it is the same as > > > > > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > > > > > and virsh define $domain.xml. This would also change the *model*. > > > > > > > > > > If this would be considered to be guest ABI stable it would mean that > > > > > other changes done by using this flag would be wrong because they also > > > > > modifies the persistent XML during migration. > > > > > > > > I assume there are very good reasons for persistent > > > > migration to behave differently, but as a user I find > > > > it extremely surprising. Do you have any insight on the > > > > rationale behind allowing ABI changes when performing > > > > persistent migration? > > > > > > Well it's a different kind of migration which is almost similar to the > > > virsh dumpxml --inactive && virsh define, with one more feature, migration > > > hook that can update the XML before it's defined on the destination. > > > > > > I agree that it may seem that we should handle this type of migration to > > > be guest ABI stable, on the other hand this basically defines new domain > > > on the destination where we allow to do the ABI updates. > > > > > > I'm adding Dan to CC to get his opinion about the guest ABI. > > > > So, there's no particular reason why things must be different for > > live vs persistent migration. The only fundamental requirement is > > that we must never change ABI underneath a running guest, which > > mandates the strong ABI guarantee for live migration & save/restore. > > > > For persistent/offline migration, there's no running guest, so from > > a technical POV we have more freedom in some sense. That said, a > > change in guest hardaware that can cause a guest to become unbootable > > is not a nice thing todo. > > > > For this reason, when you use 'defineXML' libvirt aims to immediately > > fill in (most) XML elements that would affect guest ABI. ie we assign > > addresses at define time, not start time, and we expand non-versioned > > machine types into versioned machine types, etc. > > > > So, yes, we should try to *not* change ABI across a dumpxml & define > > operational. > > > > When we first introduced PCI device addressing in libvirt, we had an > > old state where XML contained no addresses, and upon starting in new > > libvirt we have to assign addresses which might not be the same as > > those seen with previous libvirt before addressing was recorded. Once > > the addresses are assigned though, we must never change them. > > This is basically the same issue like with the PCI addresses. We used "-usb" > instead of "-device ...", but when we switched to using the "-device ..." > way we didn't record the default model in the XML. The default model > is recorded in the XML since Libvirt 2.2.0. Ok, that makes sense then. Until we are able to use -device we can't gaurantee stability for an arch. > > - if we recorded 'pci-ohci' in the XML, we must never change that > > to 'nec-xhci'. > > This will never happen. > > > - if we didn't record 'pci-ochi' in the XML, we should fill in the > > model that is most likely to match what the previous default > > behaviour was, which would be 'pci-ochi' not 'nec-xhci' IIUC. > > This patch tries to move from 'pci-ohci' to 'nec-xhci' as a default USB > controller because the current default model is probably broken, see [1], > and will never be fixed. In addition I would like to do this change in > Libvirt so all other tools that uses Libvirt don't have to implement > this change by themselves. > > In this case I think it's better to change the default from 'pci-ohci' to > 'nec-xhci' if you define a new domain where the model is not specified. If I look at qemu-system-ppc64 -usb, that provides an 'nec-xhci' device. I'm curious why libvirt picked pci-ohci ? Was -usb previously providing a pci-ohci device instead of nec-xhci ? I guess the thing that could justify the switch is that we don't guarantee what default devices you'll get for any given guest. It is dependent on the particular version of the machine type used. So in that sense if you have an XML doc which is not fully populated with device info, you are already liable to see changes in default devices based on which QEMU you happen to define the XML against first. > There is a concern whether we should allow ABI change for persistent > migration which is more general issue that will affect some other changes > that are already "allowed" for persistent migration. I'm starting to feel > that we should remove the "VIR_DOMAIN_DEF_PARSE_ABI_UPDATE" from > "virDomainDefParseNode" call in "qemuMigrationCookieXMLParse" function > which currently allows to do ABI changes for persistent migration. Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://entangle-photo.org -o- http://search.cpan.org/~danberr/ :| -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, Mar 07, 2017 at 03:26:55PM +0000, Daniel P. Berrange wrote: > On Tue, Mar 07, 2017 at 04:15:03PM +0100, Pavel Hrdina wrote: > > On Tue, Mar 07, 2017 at 02:40:55PM +0000, Daniel P. Berrange wrote: > > > On Tue, Mar 07, 2017 at 03:32:37PM +0100, Pavel Hrdina wrote: > > > > On Tue, Mar 07, 2017 at 02:55:43PM +0100, Andrea Bolognani wrote: > > > > > On Tue, 2017-03-07 at 09:19 +0100, Pavel Hrdina wrote: > > > > > > > However, after migration is complete, the <controller> > > > > > > > element has model='nec-xhci' instead of model='pci-ohci', > > > > > > > which means that power cycling the guest results in > > > > > > > breaking the guest ABI. > > > > > > > > > > > > I'm not so sure that this is an ABI change. The guest ABI is to ensure > > > > > > that the same guest XML will always start the same QEMU guest. However > > > > > > the PERSISTENT migration can make ABI changes because it is the same as > > > > > > virsh dumpxml $domain > $domain.xml && copy the XML onto remote host > > > > > > and virsh define $domain.xml. This would also change the *model*. > > > > > > > > > > > > If this would be considered to be guest ABI stable it would mean that > > > > > > other changes done by using this flag would be wrong because they also > > > > > > modifies the persistent XML during migration. > > > > > > > > > > I assume there are very good reasons for persistent > > > > > migration to behave differently, but as a user I find > > > > > it extremely surprising. Do you have any insight on the > > > > > rationale behind allowing ABI changes when performing > > > > > persistent migration? > > > > > > > > Well it's a different kind of migration which is almost similar to the > > > > virsh dumpxml --inactive && virsh define, with one more feature, migration > > > > hook that can update the XML before it's defined on the destination. > > > > > > > > I agree that it may seem that we should handle this type of migration to > > > > be guest ABI stable, on the other hand this basically defines new domain > > > > on the destination where we allow to do the ABI updates. > > > > > > > > I'm adding Dan to CC to get his opinion about the guest ABI. > > > > > > So, there's no particular reason why things must be different for > > > live vs persistent migration. The only fundamental requirement is > > > that we must never change ABI underneath a running guest, which > > > mandates the strong ABI guarantee for live migration & save/restore. > > > > > > For persistent/offline migration, there's no running guest, so from > > > a technical POV we have more freedom in some sense. That said, a > > > change in guest hardaware that can cause a guest to become unbootable > > > is not a nice thing todo. > > > > > > For this reason, when you use 'defineXML' libvirt aims to immediately > > > fill in (most) XML elements that would affect guest ABI. ie we assign > > > addresses at define time, not start time, and we expand non-versioned > > > machine types into versioned machine types, etc. > > > > > > So, yes, we should try to *not* change ABI across a dumpxml & define > > > operational. > > > > > > When we first introduced PCI device addressing in libvirt, we had an > > > old state where XML contained no addresses, and upon starting in new > > > libvirt we have to assign addresses which might not be the same as > > > those seen with previous libvirt before addressing was recorded. Once > > > the addresses are assigned though, we must never change them. > > > > This is basically the same issue like with the PCI addresses. We used "-usb" > > instead of "-device ...", but when we switched to using the "-device ..." > > way we didn't record the default model in the XML. The default model > > is recorded in the XML since Libvirt 2.2.0. > > Ok, that makes sense then. Until we are able to use -device we can't > gaurantee stability for an arch. > > > > > - if we recorded 'pci-ohci' in the XML, we must never change that > > > to 'nec-xhci'. > > > > This will never happen. > > > > > - if we didn't record 'pci-ochi' in the XML, we should fill in the > > > model that is most likely to match what the previous default > > > behaviour was, which would be 'pci-ochi' not 'nec-xhci' IIUC. > > > > This patch tries to move from 'pci-ohci' to 'nec-xhci' as a default USB > > controller because the current default model is probably broken, see [1], > > and will never be fixed. In addition I would like to do this change in > > Libvirt so all other tools that uses Libvirt don't have to implement > > this change by themselves. > > > > In this case I think it's better to change the default from 'pci-ohci' to > > 'nec-xhci' if you define a new domain where the model is not specified. > > If I look at qemu-system-ppc64 -usb, that provides an 'nec-xhci' device. > > I'm curious why libvirt picked pci-ohci ? Was -usb previously providing > a pci-ohci device instead of nec-xhci ? We switched from -usb to -device in Libvirt 1.3.1 but we start to record the default model in XML in Libvirt 2.2.0 (yes it should have been done in the Libvirt 1.3.1). So probably in the time of Libvirt 1.3.1 the default in Qemu was pci-ohci but they've probably switched to nex-xhci as they decided that they will not fix the issues with pci-ohci. > I guess the thing that could justify the switch is that we don't guarantee > what default devices you'll get for any given guest. It is dependent on > the particular version of the machine type used. So in that sense if you > have an XML doc which is not fully populated with device info, you are > already liable to see changes in default devices based on which QEMU > you happen to define the XML against first. I agree with that, if you let Libvirt to chose some default you probably don't care that much what the default will be. > > There is a concern whether we should allow ABI change for persistent > > migration which is more general issue that will affect some other changes > > that are already "allowed" for persistent migration. I'm starting to feel > > that we should remove the "VIR_DOMAIN_DEF_PARSE_ABI_UPDATE" from > > "virDomainDefParseNode" call in "qemuMigrationCookieXMLParse" function > > which currently allows to do ABI changes for persistent migration. Pavel -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
On Tue, 2017-03-07 at 16:35 +0100, Pavel Hrdina wrote: > > If I look at qemu-system-ppc64 -usb, that provides an 'nec-xhci' device. > > > > I'm curious why libvirt picked pci-ohci ? Was -usb previously providing > > a pci-ohci device instead of nec-xhci ? > > We switched from -usb to -device in Libvirt 1.3.1 but we start to record > the default model in XML in Libvirt 2.2.0 (yes it should have been done > in the Libvirt 1.3.1). So probably in the time of Libvirt 1.3.1 the default > in Qemu was pci-ohci but they've probably switched to nex-xhci as they > decided that they will not fix the issues with pci-ohci. Yup, the switch from pci-ohci to nec-usb-xhci as QEMU default happened about a year ago for that very reason, see https://bugzilla.redhat.com/show_bug.cgi?id=1284333 > > I guess the thing that could justify the switch is that we don't guarantee > > what default devices you'll get for any given guest. It is dependent on > > the particular version of the machine type used. So in that sense if you > > have an XML doc which is not fully populated with device info, you are > > already liable to see changes in default devices based on which QEMU > > you happen to define the XML against first. > > I agree with that, if you let Libvirt to chose some default you probably > don't care that much what the default will be. I completely agree with the argument in general. That said, libvirt should only make the choice *once* and store it in the guest XML, which is something that releases between 1.3.1 and 2.1.0 unfortunately neglected to do. If at all possible, guests created using one of the affected releases should not be changed at all, even though switching from pci-ohci to nec-xhci is something that I don't expect would actually break any guest. > > > There is a concern whether we should allow ABI change for persistent > > > migration which is more general issue that will affect some other changes > > > that are already "allowed" for persistent migration. I'm starting to feel > > > that we should remove the "VIR_DOMAIN_DEF_PARSE_ABI_UPDATE" from > > > "virDomainDefParseNode" call in "qemuMigrationCookieXMLParse" function > > > which currently allows to do ABI changes for persistent migration. I think we should really think hard about this and possibly change our current approach. While it might be true that offline migration doesn't have quite as restrictive requirements as live migration, that doesn't mean we should be updating the guest ABI willy-nilly. -- Andrea Bolognani / Red Hat / Virtualization -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
© 2016 - 2026 Red Hat, Inc.