- Devel - libvirt List Archives

[libvirt] PATCH: Support memory balloon device in QEMU (incomplete)
by Daniel P. Berrange 26 Feb '09

26 Feb '09

This is an incomplete patch starting to support the memory balloon device in QEMU guests. This requires the VirtIO Balloon driver be present in the guest. In much the same way as with Xen guests, this lets you adjust the memory allocation of the guest on the fly, ceiling limited by the initial memory allocation when booted. The missing bit is to use 'info balloon' to fetch the current allocation for the virDomainGetInfo() API call. Regards, Daniel diff -r 2662c220efc8 src/qemu_driver.c --- a/src/qemu_driver.c Thu Feb 26 15:31:11 2009 +0000 +++ b/src/qemu_driver.c Thu Feb 26 15:31:23 2009 +0000 @@ -2223,6 +2223,46 @@ cleanup: return ret; } + +static int qemudDomainSetMemoryBalloon(virConnectPtr conn, + virDomainObjPtr vm, + unsigned long newmem) { + char *cmd; + char *reply; + + /* + * 'newmem' is in KB, QEMU monitor works in MB, and we all wish + * we just worked in bytes with unsigned long long everywhere. + */ + if (virAsprintf(&cmd, "balloon %lu", (newmem / 1024)) < 0) { + virReportOOMError(conn); + return -1; + } + + if (qemudMonitorCommand(vm, cmd, &reply) < 0) { + qemudReportError(conn, dom, NULL, VIR_ERR_OPERATION_FAILED, + "%s", _("could not balloon memory allocation")); + VIR_FREE(cmd); + return -1; + } + VIR_FREE(cmd); + + /* If the command failed qemu prints: + * device not found, device is locked ... + * No message is printed on success it seems */ + DEBUG ("balloon reply: %s", reply); + if (strstr(reply, "\nunknown command:")) { + qemudReportError (conn, dom, NULL, VIR_ERR_OPERATION_FAILED, + _("memory ballooning is not supported %s"), reply); + VIR_FREE(reply); + return -1; + } + VIR_FREE(reply); + + return 0; +} + + static int qemudDomainSetMemory(virDomainPtr dom, unsigned long newmem) { struct qemud_driver *driver = dom->conn->privateData; virDomainObjPtr vm; @@ -2240,20 +2280,18 @@ static int qemudDomainSetMemory(virDomai goto cleanup; } + if (newmem > vm->def->maxmem) { + qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG, + "%s", _("cannot set memory higher than max memory")); + goto cleanup; + } + if (virDomainIsActive(vm)) { - qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT, - "%s", _("cannot set memory of an active domain")); - goto cleanup; - } - - if (newmem > vm->def->maxmem) { - qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG, - "%s", _("cannot set memory higher than max memory")); - goto cleanup; - } - - vm->def->memory = newmem; - ret = 0; + ret = qemudDomainSetMemoryBalloon(dom->conn, vm, newmem); + } else { + vm->def->memory = newmem; + ret = 0; + } cleanup: if (vm) -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

1 0

[libvirt] [PATCH]: Fix non-live migration failure case
by Chris Lalancette 26 Feb '09

26 Feb '09

All, There was a logic error in the Qemu driver when doing a non-live migrate. During a non-live migrate, on the source host during the Perform step, we pause the domain; however, if there was ever a failure, we were forgetting to unpause the domain, meaning that the domain was paused forever. Add a flag to tell us when we should unpause the domain after a failure. Signed-off-by: Chris Lalancette <clalance(a)redhat.com>

2 2

[libvirt] [PATCH 0/1] Resend of example program
by David Allan 26 Feb '09

26 Feb '09

I'm just resending this patch, as nobody objected to it after I incorporated the feedback I got, but it never got applied, either. Dave

2 2

[libvirt] PATCH: Fix autostart of session daemon
by Daniel P. Berrange 26 Feb '09

26 Feb '09

When connecting qemu:///session we can autostart the libvirtd daemon. With the node device APIs present, it can take upto a second to start up while talking to dbus. The remote driver was only waiting for a fraction of a second. This changes it to wait for upto 2 seconds, and also not to spawn multiple instances of the daemon. Daniel diff -rup libvirt-0.6.0.orig/src/remote_internal.c libvirt-0.6.0.new/src/remote_internal.c --- libvirt-0.6.0.orig/src/remote_internal.c 2009-02-18 10:56:34.000000000 +0000 +++ libvirt-0.6.0.new/src/remote_internal.c 2009-02-18 13:35:26.000000000 +0000 @@ -654,12 +654,13 @@ doRemoteOpen (virConnectPtr conn, */ if (errno == ECONNREFUSED && flags & VIR_DRV_OPEN_REMOTE_AUTOSTART && - trials < 5) { + trials < 20) { close(priv->sock); priv->sock = -1; - if (remoteForkDaemon(conn) == 0) { + if (trials > 0 || + remoteForkDaemon(conn) == 0) { trials++; - usleep(5000 * trials * trials); + usleep(1000 * 100 * trials); goto autostart_retry; } } -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

2 2

[libvirt] storage_backend_fs.c Bug
by Russell 26 Feb '09

26 Feb '09

Line 1133 of storage_backend_fs.c in 0.6.0: if (vol->target.backingStore != NULL) { virStorageReportError(conn, VIR_ERR_NO_SUPPORT, _("copy-on-write image not supported with " "qcow-create")); return -1; } virStorageVolTarget has no member called backingStore, resulting in compile errors.

5 5

[libvirt] [PATCH] Don't generate python bindings multiple times w/ parallel build
by Cole Robinson 26 Feb '09

26 Feb '09

The attached patch was posted by Michael Marineau at: https://bugzilla.redhat.com/show_bug.cgi?id=472702 It prevents generating the python bindings multiple times in a parallel build. Works fine in my testing, so I figured I'd pass it along here for better exposure, since it has been languishing in bugzilla. Thanks, Cole

3 2

[libvirt] [PATCH 0/5] Add device assignment related node device methods
by Mark McLoughlin 25 Feb '09

25 Feb '09

Hi, The following patches implement three new node device methods that are needed for KVM PCI device assignment. The dettach method unbinds the device from its device driver and binds it to the pci-stub driver, if available. This ensures the device is not being used by the host. The reattach method undos this operation. It unbinds the device from pci-stub and attempts to re-bind it to the original device driver. This allows the device to be used by the host again. The reset method is intended to be used on a dettached device before it is used by the guest or host. This operation may in future affect other devices, so it is intended that all devices to be assigned to a guest are first dettached before any device is reset. A writeup of some more background details are available here: http://marc.info/?l=kvm&m=123454366317045 Cheers, Mark.

2 13

[libvirt] [PATCH] document static host IP assignment
by Charles Duffy 25 Feb '09

25 Feb '09

Back in August 2008, DV added support for providing static IP/hostname assignments to dnsmasq via <host> elements in the network definition. Since this functionality isn't covered in the documentation, I wrote up a quick patch, attached.

3 2

[libvirt] [PATCH 0/2] Dettach and reset devices before assigning to guests
by Mark McLoughlin 25 Feb '09

25 Feb '09

Hi, The following two patches build on the previous series. The idea is simple - when starting a guest, we should automatically dettach and reset any devices assigned to it. Rather than change the semantics of the existing hostdev source, we only do this when the node device name is used as the hostdev source. Directly specifying the device address rather than its name can be seen as an option for people who know what they're doing. Cheers, Mark.

2 6

[libvirt] "type 'hvm' & arch 'i686' combination is not supported" error
by Frederik Himpe 25 Feb '09

25 Feb '09

I recently updated for libvirt 0.4.6 to libvirt 0.6.0. Now several of the virtual machines I defined, are not shown in virt-manager nor by virsh list --all, although the config files are still in /etc/libvirt/qemu. I tried to run virsh create, and this error is shown: libvir: Domain Config error : internal error os type 'hvm' & arch 'i686' combination is not supported This configuration was working perfectly with libvirt 0.4.6. I needed to change i686 to x86_64 in all config files of 32 bit systems to make them work again. Is there no way libvirt could migrate those configuration files automatically? -- Frederik Himpe

2 5