Known Issues

Support limited to text-mode environments and server software

Oracle Linux 7.6 (aarch64) is engineered and tested for server-related usage only. Although packages for desktop and productivity features are built and provided, the testing of these packages is limited and support for graphical mode packages is not provided in this release. You can install any of these packages on your platform, but it is possible that some of these applications may not work or may have issues that are not documented here.

Oracle does not provide support for these packages, and any assistance is community-based. If you choose to run a desktop environment or any desktop applications, you should direct any questions to the Oracle Linux for Arm community forum at https://community.oracle.com/community/technology_network_community/server_%26_storage_systems/linux/oracle-linux-for-arm.

Installation Issues

The following issues might be encountered during installation.

Alternate installation options

When performing a text based installation, the installer detects the video controller and offers a graphical installation on the system video console by default. The boot menu may not offer options to perform text-based or VNC-based installation.

To perform a text-based installation, you must specify the inst.text option on the boot command line. If you intend to perform a remote graphical installation over VNC, you must use the inst.vnc option, as well as specify boot options to configure the network. Typically ip=dhcp is used for this purpose.

For more information about boot options, see Oracle Linux 7: Installation Guide. (Bug ID 27581120)

Graphical installer allows users to edit kickstart settings

When performing a graphical installation, where some installation options are already set by using a kickstart configuration file, it is still possible to modify these settings by clicking the various fields during the installation to edit the predefined content. These types of edits during the installation process requires a user to intentionally attempt to modify the setting, effectively enabling an interactive installation, where options that are set in the kickstart configuration are not secured by any policy.

Note that this type of change is not possible when performing a text installation. During a text installation, the user can only modify fields that have not already been defined in the kickstart configuration file. (Bug ID 28642357)

Kernel panic for QLogic cnic module on ThunderX2 during installation

The QLogic cnic driver module is unsupported for 64-bit Arm platforms. The Cavium ThunderX2 servers include hardware that can cause the cnic driver module to load, triggering a kernel panic.

To work around the issue, blacklist the cnic module at boot by using the module_blacklist=cnic boot option in the kernel command line for the installer. To prevent the module from installing or loading in the future, create the file /etc/modprobe.d/cnic.conf and ensure that it contains the following lines:

blacklist cnic
install cnic /bin/true

(Bug IDs 27011806, 28109733)

Uninstalling libpcap can result in the removal of a large number of `libvirt` packages

The libpcap package is updated to enable functionality for future technologies. If you install this package and then attempt to uninstall it, a large number of libvirt packages may also be uninstalled due to dependency relationships. The libvirt package has a dependency on the libvirt-daemon-driver-nwfiler package, and this package has a dependency on libpcap. Removing libpcap removes the entire libvirt family of packages. (Bug ID 28582266)

Remove fence-agents-azure-arm-4.0.11-86.el7_5.3.aarch64 package before upgrade from Oracle Linux 7.5 (aarch64)

When upgrading from Oracle Linux 7.5 (aarch64) to Oracle Linux 7.6 (aarch64), the following package conflict is encountered while installing packages from the ol7_latest channel:

Transaction check error:
file /usr/share/fence/azure_fence.py conflicts between attempted installs
of fence-agents-azure-arm-4.0.11-86.el7_5.3.aarch64 and
fence-agents-common-4.2.1-11.el7.aarch64

The workaround for this issue is to remove the fence-agents-azure-arm-4.0.11-86.el7_5.3.aarch64 package before upgrading from Oracle Linux 7.5 (aarch64).

Note:

The older package (fence-agents-azure-arm-4.0.11-86.el7_5.3.aarch64) will be removed from yum and ULN in the next update.

Bug ID (28888799)

Remove libvirt-autonuma-config-hook and other associated packages before upgrade from Oracle Linux 7.5 (aarch64)

The libvirt-autonuma-config-hook, virt-v2v and virt-p2v-maker packages were distributed with the initial release of Oracle Linux 7.5 (aarch64). These packages are not required and can cause some dependency problems if they are installed when you attempt to upgrade to Oracle Linux 7.6 (aarch64).

If any of these packages are installed on your system, remove them as follows prior to an upgrade:

# yum remove libvirt-autonuma-config-hook virt-v2v virt-p2v-maker

(Bug ID 28681615, 28807367)

Upgrade failure if open files limit is too low and rpm-plugin-systemd-inhibit is installed

Upgrade from Oracle Linux 7.5 can fail if the login session open files limit is set too low and the system that is being upgraded includes many packages from many channels or repositories. The issue can be triggered if the rpm-plugin-systemd-inhibit package is installed and the session is configured for a maximum open file limit below 4096. This issue typically results in yum failing to update and error messages similar to:

Verifying  : glib2-static-2.56.1-1.el7.i686
glib2-static-2.56.1-1.el7.i686 was supposed to be installed but is not!

To resolve this issue, set the open file limit to 4096 before running yum update:

sudo ulimit -n 4096
sudo yum update -y

(Bug ID 28720235)

File System Bugs

The following file systems related issues pertain to Oracle Linux 7.6 (aarch64).

btrfs: Convert operation results in a file system that cannot be mounted

The btrfs-convert tool can be used to convert an ext4 file system to btrfs. However, if the tool attempts to convert a file system that has not been created with a sector block size that matches the system default page size, which is set to 64 KB for aarch64, the resulting file system cannot be mounted. This issue may occur if the original ext4 file system is not created by using the -b 65536 option to specify a 64 KB block size. (Bug ID 28200561)

Automatic Bug Reporting Tool

The automated reporting daemons and features provided by the Red Hat Automatic Bug Reporting Tool (ABRT) are not supported with Oracle Linux

ABRT packages and associated files, such as libreport, are included in the distribution to satisfy package dependencies and can be used to generate local bug reports but the features to automatically upload these reports are not supported. For technical assistance, contact Oracle Support by using the My Oracle Support portal or by telephone.

Manual execution of shim first-stage boot loader may fail in the UEFI shell

Manually loading shim binaries from the UEFI shell may fail in some scenarios. The following cases are known to fail:

FS0:\EFI\redhat\shimaa64.efi
FS0:\EFI\redhat\> shimaa64.efi

FS0:\> \EFI\redhat\shimaa64.efi is known to work correctly.

A standard installation is unaffected by this bug and the problem is limited to the manual execution of shim in the UEFI shell. (Bug ID 27962691)

Kdump tools fail to create vmcore-dmesg.txt on systems with many CPUs

On systems with many CPUs, the crash dump tools that are included with Kdump, fail to create the vmcore-dmesg.txt file, which is created with the vmcore file. This may result in a segmentation fault:

...
kdump: saving to /sysroot//var/crash/127.0.0.1-2018-05-22-12:34:45/
kdump: saving vmcore-dmesg.txt
/lib/kdump-lib-initramfs.sh: line 118:   459 Segmentation fault      
$_dmesg_collector /proc/vmcore > ${_path}/vmcore-dmesg-incomplete.txt
kdump: saving vmcore-dmesg.txt failed
kdump: saving vmcore
Copying data                                      : [100.0 %] \          
eta: 0s
kdump: saving vmcore complete

This issue is the result of a log buffer that is dynamically allocated by the kernel. vmcore-dmesg does not know how to access memory allocated in this way. The issue is typically triggered on systems with 64 or more CPUs, but has also been observed on a 32-core Ampere X-Gene 3 system.

The dmesg output can be retrieved manually by running crash against the vmcore and using the dmesg command once in the crash shell. (Bug ID 28064675, 28670960)

Snapshots of KVM guests using UEFI fail and are unsupported

You cannot do snapshots of KVM guests if they use UEFI. In older versions of QEMU and libvirt, the tools might allow you to create the snapshot without an error or warning, but the snapshot could be corrupted. More recent versions of these tools prevent snapshot creation with an error similar to the following:

virsh # snapshot-create-as OL7-seboot

error: Operation not supported: internal snapshots of a VM with pflash based
firmware are not supported

(Bug ID 26826800)

Virtual machine with copied disk fails to boot

Attempting to start a virtual machine that has been created to use a copied virtual machine hard disk can fail with the messages:

Section 0 has negative size
Failed to load image: Unsupported
start_image() returned Unsupported
Error: Image at 002384AD000 start failed: Unsupported
Unloading driver at 0x002384AD000

The boot loader drops back into the UEFI Interactive Shell after it fails.

The issue is caused by a problem with the default EFI boot option that is used by the aarch64 virtual machine firmware when it attempts to boot the disk.

To work around the issue, when the virtual machine drops to the UEFI shell after it fails to boot, you can enter the following to manually load the appropriate shim binary:

Shell> sudo FS0:\EFI\BOOT\fbaa64.efi

This is a one-time requirement and future reboots of the same virtual machine do not require manual intervention. (Bug ID 27972230)

Error with PXE-based installation of a UEFI virtual machine

An error appears when performing an operating system installation on a UEFI based virtual machine using PXE boot where there is no ramfb device present. Typically, the error may appear as:

Error: Image at 0023F1EB000 start failed: Not Found

The error does not prevent installation and is reported as debugging output. The error does not appear in earlier versions of the package, where support for the QemuRamfbDxe driver was not present. The error message can be ignored. (Bug ID 28868674)

grubby fatal error during kernel upgrade when /boot is on a btrfs subvolume

If /boot is hosted on a btrfs subvolume, GRUB 2 is unable to correctly process the initramfs and vmlinuz pathnames. This problem occurs when you update or install a new kernel and grubby attempts to update the GRUB 2 configuration. In the case where you are running a fresh installation of Oracle Linux 7.6 and you upgrade the UEK kernel, the following error is displayed:

grubby fatal error: unable to find a suitable template

When the system is rebooted after the kernel update, the system boots to the old kernel.

The workaround for this problem is to use grub2-mkconfig to regenerate the /etc/grub2/grub.cfg file, or /etc/grub2-efi.cfg file on a UEFI booted system, immediately after the kernel has been installed or upgraded, for example:

sudo grub2-mkconfig -o /boot/grub2/grub.cfg

Obtain a listing of the kernel menu entries in the generated configuration as follows:

grep -P "submenu|^menuentry" /boot/grub2/grub.cfg | cut -d "'" -f2

From the listing, select the kernel entry that you want to run as the default kernel and set this entry as the default by using the following command:

sudo grub2-set-default "menu entry title"

where for menu entry title, you substitute with the title of the kernel entry that you identified in the listing.

You can use the grub2-editenv list command to check that the saved_entry has been updated with the selected kernel menu title.

Reboot and use uname -a to check that the correct kernel is now running.

(Bug ID 22750169)

Hebrew LaTeX fonts

Installing the tex-fonts-hebrew package fails unless you first install all texlive* packages. (Bug ID 19059949)

Unable to create Oracle Linux 7 LXC containers on NFS

The creation of Oracle Linux 7 containers fails when the root file system (/container) is hosted on an NFS share. This problem occurs because the iputils package in Oracle Linux 7 releases, (Updates 4 and 5) is built to use the Linux file extended attributes [xattr(7)] security capabilities(7). Because the NFS protocol does not support these file capabilities, the iputils package might not be installed into an NFS files system. For example, when attempting to create an Oracle Linux 7.4 container, the installation fails while installing the iputils package, producing the following error:

Error unpacking rpm package iputils-20121221-7.el7.x86_64
error: unpacking of archive failed on file /usr/bin/ping: cpio: cap_set_file
error: iputils-20121221-7.el7.x86_64: install failed

Similar issues are seen when attempting to install the initscripts and systemd packages while creating an Oracle Linux 7.3 container.

This issue occurs on both NFSv3 and NFSv4.

Oracle Linux 6 containers are not affected. LXC is available as a technology preview in this release and some functionality may not work correctly. (Bug ID 25024258)

Segmentation fault when running bluedevil-wizard

The bluedevil-wizard, available in the bluedevil package, is unable to connect or locate bluetooth devices and fails with a segmentation fault when it is run. (Bug ID 27101618)

dsktune command fails with an unsupported error message

The dsktune command that is included with the 389 Directory Server base package, 389-ds-base, fails with an error message notifying you that the system does not have support for cx16, a feature commonly available on x86 platforms:

ERROR: This system does not support CMPXCHG16B instruction (cpuflag cx16).
nsslapd-enable-nunc-stans must be set to "off" on this system.
In a future release of Directory Server this platform will NOT be
supported.

ERROR  : The above errors MUST be corrected before proceeding.

The dsktune command checks that a system meets requirements and can provide information that helps with the configuration, but is not required to run the 389 Directory Server. (Bug ID 26861135)

dmstats report output does not return any data

The dmstats report command does not recognize any of the device mapper targets that are present on a system and returns no output when it is run. The issue is a regression in the upstream package and can be resolved by downgrading the device-mapper packages, if required. (Bug ID 28642757)