Appendix B. Linux VM Release Notes

B.1. Release Notes

Most modern Linux distributions support Xen paravirtualization directly, but have different installation mechanisms and some kernel limitations.

B.1.1. Red Hat Enterprise Linux 4.5 to 4.8

The following issues have been reported to Red Hat and are already fixed in the Xen kernel (which can be installed by using the /mnt/Linux/install.sh script in the built-in xs-tools.isoCD image):

The Xen kernel in RHEL 4.8 can occasionally enter tickless mode when an RCU is pending. When this triggers, it is usually in synchronize_kernel() which means the guest essentially hangs until some external event (such as a SysRQ) releases it (Red Hat Bugzilla 427998)

Live migration can occasionally crash the kernel under low memory conditions (Red Hat Bugzilla 249867)

Guest kernel can occasionally hang due to other XenStore activity (Red Hat Bugzilla 250381)

RHEL 4.7 contains a bug which normally prevents it from booting on a host with more than 64GiB of RAM (Red Hat Bugzilla 311431). For this reason XenServer RHEL 4.7 guests are only allocated RAM addresses in the range below 64GiB by default. This may cause RHEL 4.7 guests to fail to start even if RAM appears to be available, in which case rebooting or shutting down other guests can cause suitable RAM to become available. If all else fails, temporarily shut down other guests until your RHEL 4.7 VM can boot.

Once you have succeeded in booting your RHEL 4.7 VM, install the XenServer Tools and run the command:

xe vm-param-remove uuid=<vm_uuid> param-name=other-config \ param-key=machine-address-size

to remove the memory restriction.

On some hardware (generally newer systems), the CPU will generate occasional spurious page faults which the OS should ignore. Unfortunately all versions of RHEL 4 fail to ignore the spurious fault and it causes them to crash (Red Hat Bugzilla 465914).

This has been fixed in our kernel. The RHEL 4 VM templates have been set with the suppress-spurious-page-faultsparameter. This assures that the installation will continue safely to the point that the standard kernel is replaced with the Citrix-provided kernel.

There is a performance impact with this parameter set, so, after the VM installation is complete, at the VM command prompt, run the command:

xe vm-param-remove uuid=<vm_uuid> other-config: \ param-key=suppress-spurious-page-faults

In RHEL 4.5 to 4.7, if a xenbus transaction end command fails it is possible for the suspend_mutex to remain locked preventing any further xenbus traffic. Applying the Citrix RHEL 4.8 kernel resolves this issue. [EXT-5]

In RHEL 4.5 to 4.8, use of the XFS filesystem can lead to kernel panic under exceptional circumstances. Applying the Citrix RHEL 4.8 kernel resolves this issue. [EXT-16 ]

In RHEL 4.5 to RHEL 4.8, the kernel can enter no tick idle mode with RCU pending; this leads to a guest operating system lock up. Applying the Citrix RHEL 4.8 kernel resolves this issue. [EXT-21]

In RHEL 4.7, 4.8, VMs may crash when a host has 64GiB RAM or higher configured. Applying the Citrix RHEL 4.8 kernel resolves this issue. [EXT-30]

In RHEL 4.5 to 4.8 and 5.0 to 5.3, the network driver contains an issue that can, in rare circumstances, lead to a kernel deadlock. Applying the Citrix RHEL 4.8 kernel resolves this issue. [EXT-45]

Additional Notes:

56

Page 63
Image 63
Citrix Systems 6.2.0 manual Appendix B. Linux VM Release Notes, Red Hat Enterprise Linux 4.5 to