1.217. xen
1.217.1. RHBA-2009:1514: bug fix update
Note
This update has already been released (prior to the GA of this release) as errata RHBA-2009:1514
Xen is a high performance and secure open source virtualization framework. Virtualization allows users to run guest operating systems in virtual machines on top of a host operating system.
These updated xen packages fix the following bug:
* issuing an "xm info" command or any virsh commands caused xend to leak memory. As well, xend's memory consumption grew very quickly in combination with cluster suite, which uses virsh to monitor running guests. This update fixes the underlying errors in the PyList_Append and PyDict_SetItemString functions that caused the memory leaks. (BZ#528163)
All xen users are advised to upgrade to these updated packages, which resolve this issue. Note: after installation, the xend service must be restarted for this update to take effect.
1.217.2. RHBA-2010:0010: bug fix update
Note
This update has already been released (prior to the GA of this release) as errata RHBA-2010:0010
Xen is a high-performance and secure open source virtualization framework. Virtualization allows users to run guest operating systems in virtual machines on top of a host operating system.
These updated xen packages fix the following bug:
* following a guest shutdown, attempting to recreate that guest may have failed due to timing issues, which caused the xend daemon to think that the guest name was still in use, even though it was not. Also, the "xm list" command (correctly) did not show the name of the guest which had shut down. As a temporary workaround, running the "xm list" command once would rectify this situation so that the guest with the same name as the one which had been shut down could be recreated. This update resolves this issue so that a guest with the same name as a guest which has recently been shut down can be recreated immediately, without having to run an interim command. (BZ#547289)
Xen users are advised to upgrade to these updated packages, which resolve this issue. Note: after installation, the xend service must be restarted for this update to take effect.
1.217.3. RHSA-2009:1472: Moderate security and bug fix update
Important
This update has already been released (prior to the GA of this release) as the security errata RHSA-2009:1472
This update has been rated as having moderate security impact by the Red Hat Security Response Team.
Xen is an open source virtualization framework. Virtualization allows users to run guest operating systems in virtual machines on top of a host operating system.
The pyGrub boot loader did not honor the "password" option in the grub.conf file for para-virtualized guests. Users with access to a guest's console could use this flaw to bypass intended access restrictions and boot the guest with arbitrary kernel boot options, allowing them to get root privileges in the guest's operating system. With this update, pyGrub correctly honors the "password" option in grub.conf for para-virtualized guests. (CVE-2009-3525)
This update also fixes the following bugs:
* rebooting para-virtualized guests sometimes caused those guests to crash due to a race condition in the xend node control daemon. This update fixes this race condition so that rebooting guests no longer potentially causes them to crash and fail to reboot. (BZ#525141)
* due to a race condition in the xend daemon, a guest could disappear from the list of running guests following a reboot, even though the guest rebooted successfully and was running. This update fixes this race condition so that guests always reappear in the guest list following a reboot. (BZ#525143)
* attempting to use PCI pass-through to para-virtualized guests on certain kernels failed with a "Function not implemented" error message. As a result, users requiring PCI pass-through on para-virtualized guests were not able to update the xen packages without also updating the kernel and thus requiring a reboot. These updated packages enable PCI pass-through for para-virtualized guests so that users do not need to upgrade the kernel in order to take advantage of PCI pass-through functionality. (BZ#525149)
All Xen users should upgrade to these updated packages, which contain backported patches to correct these issues. After installing the updated packages, the xend service must be restarted for this update to take effect.
1.217.4. RHBA-2010:0294: bug fix and enhancement update
The
xen
packages contain tools for managing the virtual machine monitor in Red Hat Enterprise Linux Virtualization.
These updated packages fix the following bugs:
- Cause: Fully-virtualized Fedora 10 and newer guests would freeze during the boot sequence. The guest would repeatedly receive an error similar to the following error:
ata2.00: configured for MWDMA2 ata2: EH complete ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen ata2.00: cmd a0/00:00:00:00:00/00:00:00:00:00/a0 tag 0 cdb 1e 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 res 41/20:03:00:00:00/00:00:00:00:00/a0 Emask 0x3 (HSM violation) ata2.00: status: { DRDY ERR } ata2: soft resetting link
The updated packages resolve this issue by fixing the virtualized ATA driver. The updated packages cause the following issue on the guest:ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen ata2.00: BMDMA stat 0x5 ata2.00: cmd a0/01:00:00:80:00/00:00:00:00:00/a0 tag 0 dma 16512 in ata2.00: status: { DRDY DRQ } ata2: soft resetting link ata2.00: configured for MWDMA2 ata2: EH complete
After this error the guest should successfully boot. (BZ#480317) - The Xen daemon would hang if a user attempted to shut down or save a paused guest. Guests cannot be saved or shut down if the guest is in a paused state as those operations can only be conducted on running guests. The updated Xen daemon will no longer hang but the user is presented with the following warning:
Error: Can't shutdown/save the domain since the domain is paused; unpaused it first if you want to shutdown/save
To shut down or save a guest in a paused state, the guest must first be unpaused by the user. (BZ#504910) - Sufficient memory was not reserved for fully virtualized guests when creating new guests. Memory listed as available by the
xm info
command could not be fully allocated to the guest. When attempting create a fully virtualized guest with more than or equal to the maximum available memory value, the following error message would appear:Error: (1, 'Internal error', 'Could not allocate memory for HVM guest.\n (16 = Device or resource busy)')
This issue has been resolved by improving the memory balloon driver to correctly allocate memory for fully virtualized guests and fully virtualized guests using the Intel VT-d extensions for PCI passthrough. (BZ#512041) - An error message appeared when attempting to detach a PCI device from a para-virtualized guest. The error is corrected in the updated packages. PCI passthrough is only supported by Xen full virtualization and not by para-virtualization. (BZ#512307)
- Scheduling was reset if a guest was saved and restored. Scheduling data is now preserved when a guest is saved and restored. (BZ#513211)
- A race condition between the
xend
daemon and the hotplug scripts caused a recently rebooted guest to disappear from thexm list
command output. The race condition is fixed and guests will now always appear in thexm list
command output after the guest is rebooted. (BZ#513604) - The
xm pci-list
command would fail due to an unhandled error in the output parsing functions for para-virtualized guests. The output is now handled correctly by thexm pci-list
command which resolves this issue in the updated packages. (BZ#514025) - An uninformative error message appeared when the
xm pci-list-assignable-devices
command was executed. Thepciback
kernel module must be loaded before executing thexm pci-list-assignable-devices
command. The updated package provides an error that informs the user the module is not loaded. (BZ#514448) - A race condition rarely occurred when a para-virtualized guest was rebooted. When the race condition occured it caused para-virtualized guests to stop or crash after rebooting. The race condition no longer occurs in the updated packages and para-virtualized guests should always run after rebooting. (BZ#518104)
- The python environment has been changed to explicitly use the system python. This allows ysers to install other versions of python for testing. (BZ#521333)
- Fixed a regression which prevented para-virtualized guests from being started with attached PCI devices. PCI passthrough now works with para-virtualized guests. (BZ#521346)
- Using the value
rename-restart
for theon_reboot
parameter caused guests to not start when the guest was rebooted. Usingrename-restart
now causes guests to start successfully after a reboot with a different name. (BZ#521799) - Simultaneous migrations from two hosts to a third host and two hosts simultaneously swapping guests between them would fail. The updated packages fix this issue and simultaneous migrations now work between multiple hosts. (BZ#522850)
- Memory leaks in the
libxc
and thelibxenstore
python binding libraries caused increasing memory usage over time. The memory leaks in thelibxc
and thelibxenstore
libraries are fixed in the updated package and the libraries memory usage should remain steady. (BZ#524308) - A guest name conflict caused a running guest to appear shut down after restarting the guest. This issue is resolved in the updated package and running guest do not erroneously report that they are shut down. (BZ#529880)
- The
xm dump-core
command could not dump memory while a guest is running. The memory dump functionality has been improved, however, there are still issues with generating live dumps. The guest core dump image may be in an inconsistent state if the guest is under load. Users are now warned about this issue. (BZ#382591)
These updated packages add the following enhancements:
- PCI device hotplugging works in the updated packages. This feature is supported by the libvirt and virt-manager tools.(BZ#512315)
- The
xendomains
service now has less confusing and better formatted output. (BZ#251666) - Support for the
maxvcpus
parameter for Xen configuration files. Themaxvcpus
parameter sets a maximum number of VCPUs available that is separate from the initial VCPUs assigned to the guest. (BZ#453042) - The SMBIOS device, the BIOS device for virtualized guests, now presents the device manufacturer as "Red Hat" and the product family as "Red Hat Enterprise Linux". (BZ#499453)
- SCSI emulated devices were not supported in prior versions. Emulated SCSI devices are now supported. (BZ#515757)
- Support for assigning more than 12 SR-IOV Virtual Functions (VFs) to a single fully virtualized Xen guest. (BZ#511403)
- Support for BZIP2 and LZMA compressed kernels for para-virtualized guests. (BZ#517049)
- PCIe switches allow peer to peer transactions that are routed by the switch and could bypass the Intel VT-d translation hardware, potentially causing unexpected behavior in the system. Access Control Services (ACS) allows the system to force the PCIe switch to route all traffic upstream so that the VT-d hardware can validate all transactions. The updated package prevents assigning PCI devices below a non-ACS PCIe switch. (BZ#523819)
- Support for using the EXT4 file system for a boot partition for para-virtualized guests. (BZ#524611)
- SMBIOS now complies with the Microsoft Server Virtualization Validation Program (SVVP). (BZ#540161)
- Xen now automatically pins guests to a smallest possible number of NUMA nodes. This provides better performance on NUMA based systems. (BZ#543199)
Users of xen are advised to upgrade to these updated packages, which resolve these issues and add these enhancements.