SUSE Linux Enterprise Server 12 SP2

Release Notes

This document provides guidance and an overview to high level general features and updates for SUSE Linux Enterprise Server 12 SP2. Besides architecture or product-specific information, it also describes the capabilities and limitations of SUSE Linux Enterprise Server 12 SP2.

If you are skipping one or more service packs, check the release notes of the skipped service packs as well. Release notes usually only list changes that happened between two subsequent releases. If you are only reading the release notes of the current release, you could miss important changes.

General documentation can be found at: http://www.suse.com/documentation/sles-12/.

Publication Date: 2017-07-21, Version: 12.2.20170721
1 SUSE Linux Enterprise Server
1.1 Interoperability and Hardware Support
1.2 Support and Life Cycle
1.3 What Is New?
1.4 Documentation and Other Information
1.5 How to Obtain Source Code
1.6 Support Statement for SUSE Linux Enterprise Server
1.7 Modules, Extensions, and Related Products
1.8 Security, Standards, and Certification
2 Installation and Upgrade
2.1 Updating the Installer at the Beginning of the Installation or Upgrade
2.2 Installation
2.3 Upgrade-Related Notes
2.4 For More Information
3 Architecture Independent Information
3.1 Kernel
3.2 Kernel Modules
3.3 Security
3.4 Networking
3.5 Systems Management
3.6 Performance Related Information
3.7 Storage
3.8 Virtualization
4 AMD64/Intel 64 (x86_64) Specific Information
4.1 Support for intel_idle and Hardware P States on Intel Skylake Processors Can Lead to Decreased Performance
4.2 Kernel NOHZ_FULL Process Scheduler Mode
4.3 System and Vendor Specific Information
5 POWER (ppc64le) Specific Information
5.1 Ceph Client Support on z Systems and POWER
5.2 Cluster Support and High Availability for POWER
5.3 The libcxl Userspace Library for CAPI Has Been Added
5.4 Enhanced Support for System Call Filtering on POWER
5.5 Hardware Transactional Memory (HTM) support in glibc for POWER
5.6 Support for CXL Flash Storage Device Driver
6 IBM z Systems (s390x) Specific Information
6.1 Hardware
6.2 Virtualization
6.3 Storage
6.4 Network
6.5 Security
6.6 Reliability, Availability, Serviceability (RAS)
6.7 Performance
6.8 Miscellaneous
7 ARM 64-Bit (AArch64) Specific Information
7.1 KVM on AArch64
7.2 Toolchain Module Enabled in Default Installation
7.3 GICv2 and GICv3 Interrupt Controller Support in QEMU
7.4 Boot Requirements for AppliedMicro X-Gene 1
7.5 ARM AArch64 System-on-Chip Platform Driver Enablement
8 Driver Updates
8.1 Network Drivers
8.2 Other Drivers
9 Packages and Functionality Changes
9.1 New Packages
9.2 Updated Packages
9.3 Removed and Deprecated Functionality
9.4 Changes in Packaging and Delivery
9.5 Modules
9.6 SDK
10 Technical Information
10.1 Virtualization: Network Devices Supported
10.2 Virtualization: Devices Supported for Booting
10.3 Virtualization: Supported Disks Formats and Protocols
10.4 Kernel Limits
10.5 KVM Limits
10.6 Xen Limits
10.7 File Systems
11 Legal Notices
12 Colophon

1 SUSE Linux Enterprise Server

SUSE Linux Enterprise Server is a highly reliable, scalable, and secure server operating system, built to power mission-critical workloads in both physical and virtual environments. It is an affordable, interoperable, and manageable open source foundation. With it, enterprises can cost-effectively deliver core business services, enable secure networks, and simplify the management of their heterogeneous IT infrastructure, maximizing efficiency and value.

The only enterprise Linux recommended by Microsoft and SAP, SUSE Linux Enterprise Server is optimized to deliver high-performance mission-critical services, as well as edge of network, and web infrastructure workloads.

1.1 Interoperability and Hardware Support

Designed for interoperability, SUSE Linux Enterprise Server integrates into classical Unix as well as Windows environments, supports open standard interfaces for systems management, and has been certified for IPv6 compatibility.

This modular, general purpose operating system runs on four processor architectures and is available with optional extensions that provide advanced capabilities for tasks such as real time computing and high availability clustering.

SUSE Linux Enterprise Server is optimized to run as a high performing guest on leading hypervisors and supports an unlimited number of virtual machines per physical system with a single subscription, making it the perfect guest operating system for virtual computing.

1.2 Support and Life Cycle

SUSE Linux Enterprise Server is backed by award-winning support from SUSE, an established technology leader with a proven history of delivering enterprise-quality support services.

SUSE Linux Enterprise Server 12 has a 13-year life cycle, with 10 years of General Support and 3 years of Extended Support. The current version (SP2) will be fully maintained and supported until 6 months after the release of SUSE Linux Enterprise Server 12 SP3.

If you need additional time to design, validate and test your upgrade plans, Long Term Service Pack Support can extend the support you get an additional 12 to 36 months in twelve month increments, providing a total of 3 to 5 years of support on any given service pack.

For more information, check our Support Policy page https://www.suse.com/support/policy.html or the Long Term Service Pack Support Page https://www.suse.com/support/programs/long-term-service-pack-support.html.

1.3 What Is New?

SUSE Linux Enterprise Server 12 introduces a number of innovative changes. Here are some of the highlights:

  • Robustness on administrative errors and improved management capabilities with full system rollback based on Btrfs as the default file system for the operating system partition and the Snapper technology of SUSE.

  • An overhaul of the installer introduces a new workflow that allows you to register your system and receive all available maintenance updates as part of the installation.

  • SUSE Linux Enterprise Server Modules offer a choice of supplemental packages, ranging from tools for Web Development and Scripting, through a Cloud Management module, all the way to a sneak preview of upcoming management tooling called Advanced Systems Management. Modules are part of your SUSE Linux Enterprise Server subscription, are technically delivered as online repositories, and differ from the base of SUSE Linux Enterprise Server only by their life cycle. For more information about modules, see Section 1.7.1, “Available Modules”.

  • New core technologies like systemd (replacing the time-honored System V-based init process) and Wicked (introducing a modern, dynamic network configuration infrastructure).

  • The open-source database system MariaDB is fully supported now.

  • Support for open-vm-tools together with VMware for better integration into VMware-based hypervisor environments.

  • Linux Containers are integrated into the virtualization management infrastructure (libvirt). Docker is provided as a fully supported technology. For more details, see https://www.suse.com/promo/sle/docker/.

  • Support for the AArch64 architecture (64-bit ARMv8) and the 64-bit Little-Endian variant of the IBM POWER architecture. Additionally, we continue to support the Intel 64/AMD64 and IBM z Systems architectures.

  • GNOME 3.20 gives users a modern desktop environment with a choice of several different look and feel options, including a special SUSE Linux Enterprise Classic mode for easier migration from earlier SUSE Linux Enterprise Desktop environments.

  • For users wishing to use the full range of productivity applications of a Desktop with their SUSE Linux Enterprise Server, we are now offering SUSE Linux Enterprise Workstation Extension (requires a SUSE Linux Enterprise Desktop subscription).

  • Integration with the new SUSE Customer Center, the new central web portal from SUSE to manage Subscriptions, Entitlements, and provide access to Support.

If you are upgrading from a previous SUSE Linux Enterprise Server release, you should review at least the following sections:

1.4 Documentation and Other Information

1.4.1 Available on the Product Media

  • Read the READMEs on the media.

  • Get the detailed change log information about a particular package from the RPM (where <FILENAME>.rpm is the name of the RPM):

    rpm --changelog -qp <FILENAME>.rpm
  • Check the ChangeLog file in the top level of the media for a chronological log of all changes made to the updated packages.

  • Find more information in the docu directory of the media of SUSE Linux Enterprise Server 12 SP2. This directory includes PDF versions of the SUSE Linux Enterprise Server 12 SP2 Installation Quick Start and Deployment Guides. Documentation (if installed) is available below the /usr/share/doc/ directory of an installed system.

  • These Release Notes are identical across all architectures, and the most recent version is always available online at http://www.suse.com/releasenotes/. Some entries are listed twice, if they are important and belong to more than one section.

1.4.2 Externally Provided Documentation

1.5 How to Obtain Source Code

This SUSE product includes materials licensed to SUSE under the GNU General Public License (GPL). The GPL requires SUSE to provide the source code that corresponds to the GPL-licensed material. The source code is available for download at http://www.suse.com/download-linux/source-code.html. Also, for up to three years after distribution of the SUSE product, upon request, SUSE will mail a copy of the source code. Requests should be sent by e-mail to mailto:sle_source_request@suse.com or as otherwise instructed at http://www.suse.com/download-linux/source-code.html. SUSE may charge a reasonable fee to recover distribution costs.

1.6 Support Statement for SUSE Linux Enterprise Server

To receive support, customers need an appropriate subscription with SUSE; for more information, see http://www.suse.com/products/server/services-and-support/.

1.6.1 General Support Statement

The following definitions apply:

L1

Problem determination, which means technical support designed to provide compatibility information, usage support, ongoing maintenance, information gathering and basic troubleshooting using available documentation.

L2

Problem isolation, which means technical support designed to analyze data, duplicate customer problems, isolate problem area and provide resolution for problems not resolved by Level 1 or alternatively prepare for Level 3.

L3

Problem resolution, which means technical support designed to resolve problems by engaging engineering to resolve product defects which have been identified by Level 2 Support.

For contracted customers and partners, SUSE Linux Enterprise Server 12 SP2 and its Modules are delivered with L3 support for all packages, except the following:

  • Technology Previews

  • sound, graphics, fonts and artwork

  • packages that require an additional customer contract

  • packages provided as part of the Software Development Kit (SDK)

SUSE will only support the usage of original (e.g., unchanged or un-recompiled) packages.

1.6.1.1 Docker Orchestration Is Not Supported

Starting with Docker 1.12, orchestration (swarm) is now a part of the Docker engine, as available from the SLES Containers module. This feature is not supported.

1.6.2 Technology Previews

Technology previews are packages, stacks, or features delivered by SUSE. These features are not supported. They may be functionally incomplete, unstable or in other ways not suitable for production use. They are mainly included for customer convenience and give customers a chance to test new technologies within an enterprise environment.

Technology previews can be dropped at any time and SUSE does not commit to providing a supported version of such technologies in the future.

Give your SUSE representative feedback, including your experience and use case.

1.6.2.1 Support for Current AMD Radeon GPUs

As a technical preview, SUSE Linux Enterprise ships the graphics driver xf86-video-amdgpu for current AMD Radeon GPUs.

Since this driver is still in an experimental state, it is not installed by default. By default, it is only enabled for one GPU on which it was tested successfully.

Important: At this stage, this driver is not supported.

To be able to use the driver, first install the package xf86-video-amd. Then, enable it for your GPU by editing /etc/X11/xorg_pci_ids.

The required format is: \<VendorID\>\<DeviceID\>. It is also described in the configuration file itself.

To find vendor ID and device ID, use the command:

lspci -n | grep 0300

All supported vendor IDs/device IDs are already in the file but are commented out. For your vendor ID/device ID combination, remove the comment character # from the beginning of the line.

1.6.2.2 KVM Nested Virtualization

KVM Nested Virtualization is available in SLE 12 as a technology preview. For more information about nested virtualization, see nested-vmx.txt (https://github.com/torvalds/linux/blob/master/Documentation/virtual/kvm/nested-vmx.txt).

1.6.2.3 Converting Physical Machines to KVM Virtual Machines

libguestfs has the tool virt-v2v to convert virtual machines from Xen to KVM. However, previously, it was not possible to convert physical installations to virtual machine installations.

As a technology preview, SLES 12 SP2 now ships the tool virt-p2v in libguestfs. virt-p2v allows converting physical machines into KVM guests.

This also means that libguestfs has been updated to a more recent version, bringing new features and fixes.

1.6.2.4 Technology Previews: AArch64 (ARMv8)
1.6.2.4.1 GNOME Desktop Environment as a Technology Preview on AArch64

The GNOME desktop environment (including GNOME Shell and GDM) is now available on the AArch64 architecture as an unsupported technology preview.

The only supported graphical environment on the AArch64 architecture is IceWM with XDM as the display manager.

1.6.2.5 Technology Previews: POWER (ppc64le)
1.6.2.5.1 Device Driver ibmvnic Has Been Added

vNIC (Virtual Network Interface Controller) is a new PowerVM virtual networking technology that delivers enterprise capabilities and simplifies network management. It is a high-performance, efficient technology that when combined with SR-IOV NIC provides bandwidth control Quality of Service (QoS) capabilities at the virtual NIC level. vNIC significantly reduces virtualization overhead resulting in lower latencies and fewer server resources (CPU, memory) required for network virtualization.

This driver is a Technology Preview in SLES 12 SP2

1.6.2.6 Technology Previews: AMD64/Intel 64 64-Bit (x86_64)
1.6.2.6.1 NVDIMM Support

In SLES 12 SP2, NVDIMM support has been added as a Technology Preview. While many of its subsystems are stable, we recommend testing your specific use case and workload before using it in production environments.

NVDIMMs have two major use cases:

  • NVDIMM as a disk device for high-performance tier, metadata in memory, and caching

  • NVDIMM as system memory to process storage data and for volatile caching

Usage of NVDIMM as a disk device has been tested by SUSE on HPE Gen9 servers and there are currently no known issues. Therefore, we plan to support customers running this scenario on certified systems which includes HPE Gen9 servers.

SUSE will work together with partners to support additional use cases in the future.

1.6.2.6.2 Guest 3D Acceleration With virtio-gpu

Prior to QEMU version 2.5, virtual graphical cards had no 3D support. Therefore, QEMU guests could not use 3D acceleration.

From the perspective of the host, QEMU 2.5 and later include virtio-gpu. virtio-gpu allows rendering OpenGL commands from the guest on the GPU of the host. This results in a large improvement of the OpenGL 3D performance of the guest.

From the perspective of the guest, the Linux kernel 4.4 and higher include the virtio-gpu driver.

When attaching a virtio-gpu to a guest which has the Linux kernel 4.4 or higher and supports OpenGL 3D 3.x acceleration, the guest can use 3D acceleration and should achieve approximately 50 percent of native performance.

Unlike VGA pass-through or using an NVIDIA GRID card, virtio-gpu does not need a dedicated graphical card or special hardware. Depending on the performance of the GPU of the host, virtio-gpu can also provide OpenGL 3D acceleration for multiple guests.

1.6.3 Software Requiring Specific Contracts

The following packages require additional support contracts to be obtained by the customer in order to receive full support:

  • PostgreSQL Database

1.7 Modules, Extensions, and Related Products

This section comprises information about modules and extensions for SUSE Linux Enterprise Server 12 SP2. Modules and extensions add parts or functionality to the system.

1.7.1 Available Modules

Modules are fully supported parts of SUSE Linux Enterprise Server with a different life cycle and update timeline. They are a set of packages, have a clearly defined scope and are delivered via an online channel only. Release notes for modules are contained in this document, see Section 9.5, “Modules”.

The following modules are available for SUSE Linux Enterprise Server 12 SP2:

NameContentLife Cycle
Advanced Systems Management ModuleCFEngine, Puppet, Salt and the Machinery toolFrequent releases
Certifications Module* FIPS 140-2 certification-specific packagesCertification-dependant
Containers ModuleDocker, tools, prepackaged imagesFrequent releases
HPC ModuleTools and libraries related to High Performance Computing (HPC)Frequent releases
Legacy Module* Sendmail, old IMAP stack, old Java, …Until September 2017
Public Cloud ModulePublic cloud initialization code and toolsFrequent releases
Toolchain ModuleGNU Compiler Collection (GCC)Yearly delivery
Web and Scripting ModulePHP, Python, Ruby on Rails3 years, ~18 months overlap

* Module is not available for the AArch64 architecture.

For more information about the life cycle of packages contained in modules, see https://scc.suse.com/docs/lifecycle/sle/12/modules.

1.7.2 Available Extensions

Extensions add extra functionality to the system and require their own registration key, usually at additional cost. Extensions are delivered via an online channel or physical media. In many cases, extensions have their own release notes documents that are available from https://www.suse.com/releasenotes/.

The following extensions are available for SUSE Linux Enterprise Server 12 SP2:

Additionally, there are the following extensions which are not covered by SUSE support agreements, available at no additional cost and without an extra registration key:

1.7.3 Derived and Related Products

This sections lists derived and related products. In many cases, these products have their own release notes documents that are available from https://www.suse.com/releasenotes/.

1.8 Security, Standards, and Certification

SUSE Linux Enterprise Server 12 SP2 has been submitted to the certification bodies for:

For more information about certification, see https://www.suse.com/security/certificates.html.

2 Installation and Upgrade

SUSE Linux Enterprise Server can be deployed in several ways:

  • Physical machine

  • Virtual host

  • Virtual machine

  • System containers

  • Application containers

2.1 Updating the Installer at the Beginning of the Installation or Upgrade

Until SLES 12 SP1, the only method of updating the installer was through the use of a driver update disk. This required manual work such as downloading the driver update and explicitly pointing the installer at it.

Starting with the SLES 12 SP2 installer, at the beginning of the installation or upgrade, the installer can contact the update server to find out whether updates for the installer are available. If there are, they are automatically applied and YaST is restarted. The installer is able to download the updates from the regular update server, a local SMT server, or a custom URL.

By default, this functionality is off. Enable this feature using the boot option self_update=1.

For more information, see the documentation at https://github.com/yast/yast-installation/blob/SLE-12-SP2/doc/SELF_UPDATE.md (https://github.com/yast/yast-installation/blob/SLE-12-SP2/doc/SELF_UPDATE.md).

2.2 Installation

This section includes information related to the initial installation of SUSE Linux Enterprise Server 12 SP2. For information about installing, see Deployment Guide at https://www.suse.com/documentation/sles-12/book_sle_deployment/data/book_sle_deployment.html.

2.2.1 Installer Crashes When Set to Mount by Label by Default

When setting the default mount value to By Label during partitioning, the installer will report an error and crash.

As a workaround, use another option for installation. If needed, switch back to By Label on the running system.

2.2.2 Network Interfaces Configured via linuxrc Take Precendence

For some configurations with many network interfaces, it can take several hours until all network interfaces are initialized (see https://bugzilla.suse.com/show_bug.cgi?id=988157 (https://bugzilla.suse.com/show_bug.cgi?id=988157)). In such cases, the installation is blocked. SLE 12 SP1 and earlier did not offer a workaround for this behavior.

Starting with SLE 12 SP2, you can speed up interactive installations on systems with many network interfaces by configuring them via linuxrc. When a network interface is configured via linuxrc, YaST will not perform automatic DHCP configuration for any interface. Instead, YaST will continue to use the configuration from linuxrc.

To configure a particular interface via linuxrc, add the following to the boot command line before starting the installation:

ifcfg=eth0=dhcp

In the parameter, replace eth0 with the name of the appropriate network interface. The ifcfg option can be used multiple times.

2.2.3 Media-based Sources Are Disabled After Installation If They Are Not Needed

Previously, when installing from local media, like a CD/DVD or USB drive, these sources remained enabled after the installation.

This could cause problems during software installation, upgrade or migration because an old or obsolete installation source remained there. Additionally, if the source was physically removed (for instance, by ejecting the CD/DVD), Zypper would complain about the source not being available.

After the installation, YaST will now check every local source to determine if the product they provide is also available through a remote repository. In that case, the local source will be disabled.

2.2.4 Partitioning Proposal: "Flexible Partitioning" Feature Has Been Removed

YaST is a highly configurable installer that allows setting very different behaviors for each product using it (SUSE Enterprise Linux, openSUSE, etc.). In previous versions of YaST, it was possible to use a feature called "Flexible Partitioning". This feature has become obsolete, as the more standard proposal mechanism has been used by SLE and openSUSE in all recent releases.

The new version of YaST detects when a (modified) installer tries to use the obsolete "Flexible Partitioning" feature, alerts the user and falls back to the standard proposal mechanism automatically.

2.2.5 YaST Clears New Partitions

Previously, when YaST created a new partition, there could be signatures of previous MD RAIDs on the partition. That caused the MD RAID to be auto-assembled which made the partition busy. Thus, subsequent commands on the new partition failed.

When creating partitions with YaST, storage signatures are now deleted before auto-assembly takes place.

2.2.6 Host Name Setting During Installation

During installation, the hostname is set to install, the DHCP-provided value, if any, or the value of the boot option hostname. The host name used during installation is not propagated to /etc/hostname of the installed system except when set using the boot option hostname.

2.2.7 More Explicit and Configurable Importing of SSH Host Keys

During an installation of SUSE Linux Enterprise, existing SSH host keys from a previous installation were imported into the new system. This is convenient in some network scenarios, but as it was done without explicitly informing the user, it could lead to undesired situations.

The installer no longer silently imports SSH host keys from the most recent Linux installation on the disk. It now allows you to choose whether to import SSH host keys and from which partition they should be imported. It is now also possible to import the rest of the SSH configuration in addition to the keys.

To import previous SSH host keys and configuration during the installation, proceed until the page Installation Summary, then choose Import SSH Host Keys and Configuration.

2.2.8 Option to Create AutoYaST Profile During Installation Has Been Removed

In earlier versions of SUSE Linux Enterprise, you could clone the system configuration as an AutoYaST profile during installation. However, many services and system parameters can only be configured after the installation process has completed and the system is up and running. This can result in an AutoYaST profile missing parts of the desired configuration.

The option of creating an AutoYaST profile during installation has been removed. However, you can still create an AutoYaST profile from the running system, after you have made sure that the system configuration fits your needs.

2.2.9 Reading Registration Codes from a USB Drive

During the installation of SUSE products, it can be tedious to remember and type in registration codes.

You can now save the registration codes to a USB drive and have YaST read them automatically.

For more information, see: https://github.com/yast/yast-registration/wiki/Loading-Registration-Codes-From-an-USB-Storage-%28Flash-Drive-HDD%29.

2.3 Upgrade-Related Notes

This section includes upgrade-related information for SUSE Linux Enterprise Server 12 SP2. For information about general preparations and supported upgrade methods and paths, see the documentation at https://www.suse.com/documentation/sles-12/book_sle_deployment/data/cha_update_sle.html.

2.3.1 Online Migration with Live Patching Enabled

The SLES online migration process reports package conflicts when Live Patching is enabled and the kernel is being upgraded. This applies when crossing the boundary between two Service Packs.

To prevent the conflicts, before starting the migration, execute the following as a super user:

zypper rm $(rpm -qa kgraft-patch-*)

2.3.2 Support for PIDs cgroup Controller

The version of systemd shipped in SLES 12 SP2 uses the PIDs cgroup controller. This provides some per-service fork() bomb protection, leading to a safer system.

However, under certain circumstances you may notice regressions. The limits have already been raised above the upstream default values to avoid this but the risk remains.

If you notice regressions, you can change a number of TasksMax settings.

To control the default TasksMax= setting for services and scopes running on the system, use the system.conf setting DefaultTasksMax=. This setting defaults to 512, which means services that are not explicitly configured otherwise will only be able to create 512 processes or threads at maximum.

For thread- or process-heavy services, you may need to set a higher TasksMax value. In such cases, set TasksMax directly in the specific unit files. Either choose a numeric value or even infinity.

Similarly, you can limit the total number of processes or tasks each user can own concurrently. To do so, use the logind.conf setting UserTasksMax (the default is 12288).

nspawn containers now also have a TasksMax value set, with a default of 16384.

2.4 For More Information

For more information, see Section 3, “Architecture Independent Information” and the sections relating to your respective hardware architecture.

3 Architecture Independent Information

Information in this section pertains to all architectures supported by SUSE Linux Enterprise Server 12 SP2.

3.1 Kernel

3.1.1 MSR Cannot Be Modified When UEFI Secure Boot Is On

Write access to MSRs (model-specific registers) allows userspace applications to modify the running kernel which runs counter to the goals of UEFI Secure Boot. Therefore, MSRs cannot be written to by tools like cpupower to control processor performance when UEFI Secure Boot is enabled on a system.

This also prevents tuning tools, such as saptune and sapconf from working correctly.

The only current workaround is disabling Secure Boot.

3.1.2 ACPI Power Meter Driver Is Disabled by Default

The ACPI power meter device acpi_power_meter requires processing of AML code from the ACPI tables to update the average power measurement. This can interrupt the CPU at relatively high frequency and has a noticeable impact to latency-sensitive applications.

There are cluster monitoring applications that consume information from acpi_power_meter, so the driver is not removed. However, in SLE 12 SP2, it is blacklisted by default.

In the event that a monitoring application requires it, it can be re-enabled by removing the driver from the blacklist file /etc/modprobe.d/50-blacklist.conf.

3.1.3 Transparent Huge Page Defragmentation Disabled by Default

Transparent Huge Pages (THP) are an important alternative to hugetlbfs that boosts performance for some applications by reducing the amount of work a CPU must do when translating virtual to physical addresses. It is particularly important for virtual machine performance where there are two translation layers.

Early in the lifetime of the system, there is enough free memory that these pages can be allocated cheaply. When the system is running for long enough, memory must be reclaimed and compacted to allocate the THP. This forces applications to stall for potentially long periods of time which many applications cannot tolerate. Many tuning guides recommend disabling THP in these types of cases.

SLE 12 SP2 disables THP defragmentation by default. THPs will only be used if they are available, instead of stalling on defragmentation. Normally, the defragmentation work is deferred and THPs will be created in the future. However, if an application explicitly requests such behavior via madvise(), it will stall.

If a system has many applications that are willing to stall while allocating THP, it is possible to restore the previous behavior of SLE via sysfs:

echo always > /sys/kernel/mm/transparent_hugepage/defrag

3.1.4 Enabling Enhanced Information About Physical Memory Page Ownership and Status

Detailed information about physical memory pages can help answer questions such as:

  • Which kernel subsystem or driver has allocated which pages?

  • What page status flags are set?

This is useful for L3 support of the kernel and during development and testing of out-of-tree kernel modules, for example, to debug memory leaks. Previously, kernel interfaces could only provide a subset of the page status flags, and only provide a summary about generic memory usage categories.

The Linux kernel shipped with SLE 12 SP2 can provide more detailed information. However, tracking extra information about each page that the kernel allocates creates overhead in terms of code to be executed and memory used. Therefore, this feature is disabled by default.

This feature is shipped with all kernel versions of SLE 12 SP2 and can be enabled during boot using the kernel parameter page_owner=on.

To obtain the status of all pages, use:

cat /sys/kernel/debug/page_owner > file

The file contains the following for each physical page:

  • Allocation flags

  • Status flags

  • Page migration status

  • Backtrace leading to the allocation

Additional postprocessing of the output can be used, for example, to count the number of pages for each unique backtrace which can help discover a code path that leaks memory.

3.1.5 Subset of Scheduler Debugging Statistics Disabled by Default

The CPU scheduler maintains a number of statistics for the purposes of debugging, some tracepoints and sleep profiling. They are only useful for detailed analysis but they incur an overhead for all users. They may be disabled at kernel build time but they are enabled as debugging in the field is important and tools like latencytop depend on them.

Some expensive scheduler debugging statistics are disabled by default. Enabling sleep profiling or running latencytop will activate them automatically but activating the tracepoints will require user intervention. The affected tracepoints are sched_stat_wait, sched_stat_sleep, sched_stat_iowait, sched_stat_blocked and sched_stat_runtime.

They can be activated at runtime using:

echo 1 > /sys/kernel/debug/tracing/events/sched/enable

They can be disabled at runtime using:

echo 0 > /sys/kernel/debug/tracing/events/sched/enable

The first number of tracepoint activations may contain stale data until the necessary data is collected. If this is undesirable, it is possible to activate them at boot time via the kernel parameter schedstats=enable.

3.1.6 Incompatible Changes in the New 4.4 Kernel

The following minor changes have been identified in the 4.4 kernel:

  • Support for TCP Limited Slow Start (RFC3742) has been removed. This feature had multiple drawbacks and questionable benefit. Its implementation was inefficient and difficult to configure. The problem that Limited Slow Start was trying to solve is now better covered by the Hybrid Slow Start algorithm which is part of default congestion control algorithm, CUBIC.

  • The kernel.blk_iopoll sysctl has been removed. This setting allowed toggling some block device drivers between iopoll and non-iopoll mode. This allowed for easier debugging of these drivers during early development. Since using this toggle was dangerous and the toggle is not needed for production setups, it has been removed.

  • The cgroup.event_control file is only available in cgroups with a memcg attached to it. There was no code using this interface outside of memcg, so this change is considered harmless.

  • The vm.scan_unevictable_pages sysctl has been removed because the functionality it was backing had been removed in 2011. Any usage of the file has been reported to the kernel log with an explanation that the file has no effect. There were no reports about a use case requiring this functionality.

  • The /sys/devices/system/memory/memory%d/end_phys_index file has been removed, because the information it exposed is considered internal to the kernel and an implementation detail. This information is not required for the memory hotplug functionality.

3.1.7 Partial Memory Mirroring

Memory mirroring offers increased system reliability. However, full memory mirroring also dramatically decreases available memory size.

Partial memory mirroring addresses this issue by setting up a smaller mirrored memory range and using this range for kernel code and data structures. The remaining memory operates in regular mode which leaves more room for applications. This feature requires support in hardware and EFI firmware and is currently supported on Fujitsu PRIMEQUEST 2000 series systems and its successor models.

3.1.8 Paravirtualization Layer for Spinlocks

To overcome issues like vCPU starvation (where a busy task waits on a scheduled out lock owner), paravirtualized spinlocks allow virtual environments, such as KVM and Xen, to replace the native spinlock implementation. This hypervisor replacement is tailored to be virtualization-friendly, for example, with it, after a period of busy-waiting, tasks yield the CPU. This could be enabled using the kernel parameter CONFIG_PARAVIRT_SPINLOCK.

However, in the past, this incurred a considerable performance overhead on native systems due to the extra indirection layer. If enabled, virtual systems would perform better, but native systems would suffer. When the parameter was disabled, the opposite was true.

With new features in the SLE 12 SP2 kernel, such as queued spinlocks, the overhead of the kernel parameter CONFIG_PARAVIRT_SPINLOCK is now negligible across systems and loads. Therefore, enabling this option by default allows the virtual environments to overcome the lock holder preemption challenges without impacting the native case. This is particularly useful in CPU overcommitment configurations, which are common, for example, in cloud-based solutions.

3.1.9 Enhanced Accounting and Reporting of shmem Swap Usage

There was a request to provide information about how much of Linux-kernel shared memory (shmem) is swapped out, for processes using such memory segments. shmem mappings are either System V shared memory segments, mappings created by mmap() with MAP_ANONYMOUS / MAP_SHARED flags, and shared mmap() mappings of files residing on the tmpfs RAM disk file system. Prior to the implemented changes, in /proc/pid/smaps, swap usage for these segments would have been shown as 0.

The kernel has been modified to show swap usage of shmem segments properly in /proc/pid/smaps files. Due to shmem implementation limitations, this value will also count swapped-out pages that the process has mapped, but never touched, which differs from anonymous memory accounting. Due to the same limitations and to prevent excessive CPU overhead, the VmSwap field in /proc/pid/status is unaffected and will not account for swapped-out portions of shmem mappings. In addition, the /proc/pid/status file has been enhanced to include three new Rss* fields as a breakdown of the VmRSS field to anonymous, file and shmem mappings. Example excerpt:

VmRSS:              5108 kB
RssAnon:              92 kB
RssFile:            1324 kB
RssShmem:           3692 kB

3.2 Kernel Modules

An important requirement for every enterprise operating system is the level of support customers receive for their environment. Kernel modules are the most relevant connector between hardware (controllers) and the operating system.

For more information about the handling of kernel modules, see the SUSE Linux Enterprise Administration Guide.

3.2.1 NVDIMM Kernel Subsystem

Non-volatile DIMMs are byte-addressable memory chips that fit inside a computer's normal memory slot but are, in contrast to DRAM chips, persistent and thus can be used as an enhancement or replacement for a computer's hard disk drives. This imposes several challenges, namely:

  • Discovery of hardware

  • Mapping and addressing of this new memory type

  • Atomic semantics as with traditional storage media

  • Page frame addressing like with traditional memory

The Linux kernel shipped with SLE now includes several drivers to address these challenges:

  • Hardware discovery is initiated via the ACPI NFIT (Non-Volatile Memory Firmware Interface Table) mechanism and realized with the device driver nfit.ko.

  • Mapping and addressing of NVDIMMs is accomplished by the device driver nd_pmem.ko.

  • The driver nd_btt.ko takes care of (optional) atomic read/write semantics to the underlying hardware.

  • The pfn portion of nd_pmem.ko provides the ability to address NVDIMM memory just like any other DRAM type memory.

3.2.2 Direct Access to Files in Non-Volatile DIMMs

The page cache is usually used to buffer reads and writes to files. It is also used to provide the pages which are mapped into userspace by a call to mmap. For block devices that are memory-like, the page cache pages would be unnecessary copies of the original storage.

The Direct Access (DAX) kernel code avoids the extra copy by directly reading from and writing to the storage device. For file mappings, the storage device is mapped directly into userspace. This functionality is implemented in the XFS and Ext4 file systems.

3.2.3 ZRAM Block Device

The ZRAM module creates RAM-based block devices. Pages written to these disks are compressed and stored in memory itself. Such disks allow for very fast I/O. Additionally, compression provides memory savings.

ZRAM devices can be managed and configured with the help of the tool zramctl (see the man page of zramctl(8)). Configuration persistence is ensured by zramcfg system service.

3.2.4 Memory Compression with zswap

Usually, when a system's physical memory is exceeded, the system moves some memory onto reserved space on a hard drive, called "swap" space. This frees physical memory space for additional use. However, this process of "swapping" memory onto (and off a hard drive is much slower than direct memory access, so it can slow down the entire system.

The zswap driver inserts itself between the system and the swap hard drive, and instead of writing memory to a hard drive, it compresses memory. This speeds up both writing to swap and reading from swap, which results in better overall system performance while using swap.

To enable the zswap driver, write 1 or Y to the file /sys/module/zswap/parameters/enabled.

Storage Back-ends

There are two back-ends available for storing compressed pages, zbud (the default), and zsmalloc. The two back-ends each have their own advantages and disadvantages:

  • The effective compression ratio of zbud cannot exceed 50 percent. That is, it can at most store two uncompressed pages in one compressed page. If the workload's compression ratio exceeds 50% for all pages, zbud will not be able to save any memory.

  • zsmalloc can achieve better compression ratios. However, it is more complex and its performance is less predictable.

  • zsmalloc does not free pages when the limit set in /sys/module/zswap/parameters/max_pool_percent is reached. This is reflected by the counter /sys/kernel/debug/zswap/reject_reclaim_fail.

It is not possible to give a general recommendation on which storage back-end should be used, as the decision is highly dependent on workload. To change the storage back-end, write either zbud or zsmalloc to the file /sys/module/zswap/parameters/zpool. Pick the back-end before enabling zswap. Changing it later is unsupported.

Setting zswap Memory

Compressed memory still uses a certain amount of memory, so zswap has a limit to the amount of memory which will be stored compressed, which is controllable through the file /sys/module/zswap/parameters/max_pool_percent. By default, this is set to 20, which indicates zswap will use 20 percent of the total system physical memory to store compressed memory.

The zswap memory limit has to be carefully configured. Setting the limit too high can lead to premature out-of-memory situations that would not exist without zswap, if the memory is filled by non-swappable non-reclaimable pages. This includes mlocked memory and pages locked by drivers and other kernel users.

For the same reason, performance can also be hurt by compression/decompression if the current workload's workset would, for example, fit into 90 percent of the available RAM, but 20 percent of RAM is already occupied by zswap. This means that the missing 10 percent of uncompressed RAM would constantly be swapped out of/in to the memory area compressed by zswap, while the rest of the memory compressed by zswap would hold pages that were swapped out earlier which are currently unused. There is no mechanism that would result in gradual writeback of those unused pages to let the uncompressed memory grow.

Freeing zswap Memory

zswap will only free its pages in certain situations:

  • The processes using the pages free the pages or exit

  • When the storage back-end zbud is in use, zswap will also free memory when its configured memory limit is exceeded. In this case, the oldest zswap pages are written back to disk-based swap.

Memory Allocation Issues

In theory, it can happen that zswap is not yet exceeding its memory limit, but already fails to allocate memory to store compressed pages. In that case, it will refuse to compress any new pages and they will be swapped to disk immediately. For confirmation whether this issue is occurring, check the value of /sys/kernel/debug/zswap/reject_alloc_fail.

3.3 Security

3.3.1 iSCSI with CHAP Is Not Supported in FIPS Mode

iSCSI's use of the Challenge-Handshake Authentication Protocol (CHAP) is not supported in FIPS mode. The protocol uses a digest algorithm (MD5) that is not FIPS-compliant. If FIPS mode is enabled, iSCSI will not be able to use CHAP.

If operation in FIPS mode is required, discontinue use of CHAP for iSCSI and secure the network by other means such as IPSec.

3.3.2 SELinux Enablement

SELinux capabilities have been added to SUSE Linux Enterprise Server (in addition to other frameworks, such as AppArmor). While SELinux is not enabled by default, customers can run SELinux with SUSE Linux Enterprise Server if they choose to.

SELinux Enablement includes the following:

  • The kernel ships with SELinux support.

  • We will apply SELinux patches to all “common” userland packages.

  • The libraries required for SELinux (libselinux, libsepol, libsemanage, etc.) have been added SUSE Linux Enterprise.

  • Quality Assurance is performed with SELinux disabled—to make sure that SELinux patches do not break the default delivery and the majority of packages.

  • The SELinux-specific tools are shipped as part of the default distribution delivery.

  • SELinux policies are not provided by SUSE. Supported policies may be available from the repositories in the future.

  • Customers and Partners who have an interest in using SELinux in their solutions are encouraged to contact SUSE to evaluate their necessary level of support and how support and services for their specific SELinux policies will be granted.

By enabling SELinux in our code base, we add community code to offer customers the option to use SELinux without replacing significant parts of the distribution.

3.4 Networking

3.4.1 Improved Bridge Handling in YaST

The configuration UI for bridges in YaST did not always show all information and did not always convert parameters properly when editing older configurations.

In SLE 12 SP2, this behavior has been improved upon:

  • The information about bridge ports and bridges is shown for each interface.

  • In the case of old configuration, upon reading the configuration, the bootproto static will be converted to none and the parameter zero IPADDR will be removed.

Additionally, to improve the user experience, the management of bridges and bonding has been unified and the interface is now updated after any change.

3.4.2 No Support for Samba as Active Directory-Style Domain Controller

The version of Samba shipped with SLE 12 GA and newer does not include support to operate as an Active Directory-style domain controller. This functionality is currently disabled, as it lacks integration with system-wide MIT Kerberos.

3.4.3 Better Information About Physical Port IDs Used by Network Interfaces with NPAR/SR-IOV Capabilities

Previously, YaST offered no way to know whether two interfaces with NPAR/SR-IOV capabilities were sharing the same physical port. As a result, users could bond them without realizing that they were not getting the desired effect in terms of redundancy.

Information about the physical port ID has been added to Interface Overview and also for each entry of the Bond Slaves table, so you can now inspect the physical port ID when selecting an interface.

Additionally, you will be alerted when trying to bond devices sharing the same physical port.

3.5 Systems Management

3.5.1 The YaST Module for SSH Server Configuration Has Been Removed

The YaST module for configuring an SSH server which was present in SLE 11, is not a part of SLE 12. It does not have any direct successor.

The module SSH Server only supported configuring a small subset of all SSH server capabilities. Therefore, the functionality of the module can be replaced by using a combination of 2 YaST modules: The /etc/sysconfig Editor and the Services Manager. This also applies to system configuration via AutoYaST.

3.5.2 SASL Integration in sudo

When SUSE Linux Enterprise 12 was first released, the sudo binary did not correctly support SASL authentication for LDAP because the package was built without a build dependency on the package cyrus-sasl-devel.

To be able to use sudo with SASL, update to the latest version of the package sudo. For information about enabling SASL authentication for sudo, see man 5 sudoers.ldap.

3.5.3 systemd: Support for System V and LSB Init Scripts Has Been Moved Out of Core Daemon

To ease future maintenance, in SLE 12 SP2, systemd was updated to version 228. This version does not support using System V and LSB init scripts from the systemd daemon itself any more.

This functionality is now implemented as a generator that creates systemd unit files from System V/LSB init scripts. These unit files are generated at boot or when systemd is reloaded. Therefore, to have changed System V init scripts recognized by systemd, run systemctl daemon-reload or reboot the machine.

For more information, see the man page of systemd-sysv-generator (man systemd-sysv-generator).

If you are packaging software that ships System V init scripts, use the RPM macros documented at https://en.opensuse.org/openSUSE:Systemd_packaging_guidelines (https://en.opensuse.org/openSUSE:Systemd_packaging_guidelines#Register_services_in_install_scripts) (Section "Register Services in Install Scripts").

3.5.4 AutoYaST: Applying the First-Stage Network Configuration to the Installed System

Due to a problem in the AutoYaST version shipped with SLE 12 SP1, the network configuration used during the first stage was always copied to the installed system. This happened regardless of the value of keep_install_network in the AutoYaST profile.

SLE 12 SP2 behaves as expected and keep_install_network will be set to true by default.

3.5.5 New YaST VPN module

The new YaST VPN module provides an intuitive and easy to use interface for setting up VPN gateways and clients. It simplifies the setup of typical IPSec VPN gateways and clients.

IPSec is an open and standardized VPN protocol, natively supported by most operating systems and devices, including Linux, Unix, Windows, Android, Blackberry, Apple iOS and MacOS, without the need for third-party software solution.

Using the YaST VPN module, you can create VPN gateways for the following scenarios:

  • Provide network access to Linux clients authenticated via a pre-shared key or certificate.

  • Provide network access to Windows 7, 8, 10, and Blackberry clients authenticated via a combination of certificate and username/password.

  • Provide network access to Android, iOS, and MacOS clients authenticated via a combination of a pre-shared key and username/password.

Additionally, you can set up connections to remote VPN gateways, for the following scenarios:

  • Prove client identity with a pre-shared key.

3.5.6 Enrolling in a Microsoft Active Directory Domain via YaST

You can configure a SLES computer to become a member in Microsoft Active Directory to leverage its user account and group management. In previous versions of SLES, enrolling a computer in a Microsoft Active Directory was a lengthy and error-prone procedure.

In SLES 12 SP2, YaST ships with the new configuration tool User Logon Management (previously Authentication Client) which offers a powerful yet simple user interface for joining an Active Directory domain and allows authenticating users using those domain accounts. In addition to Active Directory, the editor can also set up authentication against a generic Kerberos or LDAP service.

3.5.7 ntp 4.2.8

ntp was updated to version 4.2.8.

  • The ntp server ntpd does not synchronize with its peers anymore and the peers are specified by their host name in /etc/ntp.conf.

  • The output of ntpq --peers lists IP numbers of the remote servers instead of their host names.

Name resolution for the affected hosts works otherwise.

Parameter changes

The meaning of some parameters for the sntp command-line tool have changed or have been dropped, for example sntp -s is now sntp -S. Please review any sntp usage in your own scripts for required changes.

After having been deprecated for several years, ntpdc is now disabled by default for security reasons. It can be re-enabled by adding the line enable mode7 to /etc/ntp.conf, but preferably ntpq should be used instead.

3.5.8 Installing kGraft Patches with Weak Package Dependency Resolution Disabled

In environments with a clearly defined list of packages to be installed on the system and weak package dependency resolution disabled via solver.onlyRequires=true in /etc/zypp/zypp.conf, automatic installation of the initial kGraft patch is broken.

As an aid in this situation, the package kernel-$FLAVOR-kgraft is provided. Installing this package pulls the associated kGraft patch into the system.

3.5.9 Sudo Now Respects Groups Added by the pam_group Module

Sudo now respects groups added by the pam_group module and adds these groups to the target user.

If there is a user tux, you can now use the following to add it to the group games:

  1. Open /etc/security/group.conf and add: sudo;*;tux;Al0000-2400;games

  2. Open /etc/pam.d/sudo and add the following line at the beginning of the file: auth required pam_group.so

  3. Then run: sudo -iu tux id

In SLE 12 SP1 and before, the user tux would not have been added to the group games:

uid=1002(tux) gid=100(users) groups=100(users)

In SLE 12 SP2, the user tux is added to the group games:

uid=1002(tux) gid=100(users) groups=100(users),40(games)

3.6 Performance Related Information

3.6.1 perf Provides Guest Exit Statistics

This feature enables perf to collect guest exit statistics based on the kvm_exits made by the threads of a guest-to-host context. The statistics report is grouped by exit reason. This can used as an indicator of the performance of a VM under a certain workload.

Besides kvm_exits, hypervisor calls are also reported and grouped by hcall reason. The statistics can be shown for an individual guest or all guests running on a system.

3.6.2 Deferred and Parallelized Initialization of Page Structures in Memory Management

Page initialization takes a very long time on large-memory systems. This is one of the reasons why large machines take a long time to boot.

The kernel now provides deferred initialization of page structures on the x86_64 architecture. Only approximately 2 GB per memory node are initialized during boot, the rest is initialized in parallel with the boot process by kernel threads named pgdatinitX, where X indicates the node ID.

3.7 Storage

3.7.1 Unloading device_handler Modules Not Possible Anymore

With SLES 12 SP2, device_handler modules cannot be unloaded anymore. This functionality has been removed upstream because of the dangers associated with it.

If the device_handler module is loaded, it is not possible to switch it to another. This was possible in earlier versions of SLES 12.

There is no workaround that allows unloading device_handler modules. However, the SLES 12 SP2 kernel has much improved algorithms for checking which device handler needs to be loaded for a given device. This will accurately reflect the capabilities of the device.

3.7.2 Root File System Conversion to Btrfs Not Supported

If it is not the root file system and if the file system has at least 20 % free space available, in-place conversion of an existing Ext2/Ext3/Ext4 or ReiserFS file system is supported for data mount points.

SUSE does not recommend or support in-place conversion of OS root file systems. In-place conversion to Btrfs of root file systems requires manual subvolume configuration and additional configuration changes that are not automatically applied for all use cases.

To ensure data integrity and the highest level of customer satisfaction, when upgrading, maintain existing root file systems. Alternatively, reinstall the entire operating system.

3.7.3 /var/cache on an Own Subvolume for Snapshots and Rollback

/var/cache contains very volatile data, like the Zypper cache with RPM packages in different versions for each update. As a result of storing data that is mostly redundant but highly volatile, the amount of disk space a snapshot occupies can increase very fast.

To solve this, move /var/cache to a separate subvolume. On fresh installations of SLE 12 SP2 or newer, this is done automatically. To convert an existing root file system, perform the following steps:

  1. Find out the device name (/dev/sda2, /dev/sda3 etc.) of the root file system: df /

  2. Identify the parent subvolume of all the other subvolumes. For SLE 12 installations, this is a subvolume named @. To check if you have a @ subvolume, use: btrfs subvolume list / | grep '@'. If the output of this command is empty, you do not have a subvolume named @. In that case, you may be able to proceed with subvolume ID 5 which was used in older versions of SLE.

  3. Now mount the requisite subvolume.

    • If you have a @ subvolume, mount that subvolume to a temporary mount point: mount <root_device> -o subvol=@ /mnt

    • If you don't have a @ subvolume, mount subvolume ID 5 instead: mount <root_device> -o subvolid=5 /mnt

  4. /mnt/var/cache can already exist and could be the same directory as /var/cache. To avoid data loss, move it: mv /mnt/var/cache /mnt/var/cache.old

  5. In either case, create a new subvolume: btrfs subvol create /mnt/var/cache

  6. If there is now a directory /var/cache.old, move it to the new location: mv /var/cache.old/* /mnt/var/cache. If that is not the case, instead do: mv /var/cache/* /mnt/var/cache/

  7. Optionally, remove /mnt/var/cache.old: rm -rf /mnt/var/cache.old

  8. Unmount the subvolume from the temporary mount point: umount /mnt

  9. Add an entry to /etc/fstab for the new /var/cache subvolume. Use an existing subvolume as a template to copy from. Make sure to leave the UUID untouched (this is the root file system's UUID) and change the subvolume name and its mount point consistently to /var/cache.

  10. Mount the new subvolume as specified in /etc/fstab: mount /var/cache

3.7.4 nvme-cli: A User-Space Tool to Manage NVMe Devices on Linux

The tool nvme-cli provides management features to NVMe devices, such as adapter information retrieval, namespace creation/formatting and adapter firmware update.

3.7.5 systemd: The NFS Mount Option bg Is Deprecated

The upstream developers of systemd do not support the NFS mount option bg any more. While this mount option is still supported in SLE 12 SP2, it will be removed in the next version of SLE.

It will be replaced by the systemd mount option nofail.

3.7.6 Snapper: Cleanup Rules Based on Fill Level

Some programs do not respect the special disk space characteristics of a Btrfs file system containing snapshots. This can result in unexpected situations where no free space is left on a Btrfs filesystem.

Snapper can watch the disk space of snapshots that have automatic cleanup enabled and can try to keep the amount of disk space used below a threshold.

If snapshots are enabled, the feature is enabled for the root file system by default on new installations.

For existing installations, the system administrator must enable quota and set limits for the cleanup algorithm to use this new feature. This can be done using the following commands:

  1. snapper setup-quota

  2. snapper set-config NUMBER_LIMIT=2-10 NUMBER_LIMIT_IMPORTANT=4-10

For more information, see the man pages of snapper and snapper-configs.

3.8 Virtualization

3.8.1 Virtual Machine Driver Pack 2.4 (VMDP 2.4)

SUSE Linux Enterprise Virtual Machine Driver Pack is a set of paravirtualized device drivers for Microsoft Windows operating systems. These drivers improve the performance of unmodified Windows guest operating systems that are run in virtual environments created using Xen or KVM hypervisors with SUSE Linux Enterprise Server 11 SP4 and SUSE Linux Enterprise Server 12 SP2. Paravirtualized device drivers are installed in virtual machine instances of operating systems and represent hardware and functionality similar to the underlying physical hardware used by the system virtualization software layer.

The new features of SUSE Linux Enterprise Virtual Machine Driver Pack 2.4 include:

  • Support for SUSE Linux Enterprise Server 12 SP2

  • Drivers for Windows Server 2016

  • Drivers are no longer dependent on pvvxbn.sys being loaded

  • Support Windows Multipoint Server

New driver and utility features:

  • pvvxbn.sys: Issues a Xen shutdown/reboot at the end of the power down sequence unless the PV control flag dfs ("disable forced shutdown") is enabled.

  • pvvxblk.sys: VirtIO: MSI vectors can now be used. Xen: support for indirect descriptors. Queuing, queue depth, and max_segs are tunable.

  • pvvxscsi.sys: VirtIO: MSI vectors can now be used.

  • setup.exe: Has enhanced support for virt-v2v.

  • pvctrl.exe: Can now modify NIC parameters. Enable/disable Xen pvvxblk queuing/queue depth (qdepth). Set Xen pvvxblk maximum number of segments (max_segs). Set debug print mask (dpm). Enable/disable Xen force shutdown after power-down sequence (dfs). Enable/disable virtio_serial MSI usage (vserial_msi).

3.8.2 KVM

3.8.2.1 KVM Legacy Device Assignment Was Disabled

The legacy device assignment feature of KVM was disabled.

As a replacement, use VFIO. VFIO provides the same functionality and has the following advantages:

  • It is actively maintained upstream while the legacy code is not.

  • It is more secure.

  • It supports new hardware features such as interrupt virtualization.

3.8.2.2 Support for UEFI in QEMU Virtual Machines

libvirt and KVM/QEMU now support UEFI for virtual machines. UEFI firmware is provided through the qemu-ovmf-x86_64 package.

3.8.2.3 Obtaining Addresses with libvirt-nss

With libvirt-nss, you can obtain addresses of dnsmasq -backed KVM guests. For more information, see the Virtualization Guide, Chapter "Obtaining IP Addresses with nsswitch for NAT Networks".

3.8.2.4 Post-Copy Live Migration Support in libvirt and QEMU/KVM

Pre-copy live migration can take a lot of time depending on the workload and page dirtying rate of the virtual machine.

libvirt and QEMU/KVM now support post-copy live migration. This means that the virtual machine starts running on the destination host as soon as possible and the RAM from the source host is pagefaulted into the destination over time. This ensures minimal downtime for the virtual machine.

The guest will run on target host immediately, only CPU state and device state are transferred to target host. If the network is down before all missing memory pages are copied from the original guest, the new guest will crash.

3.8.3 Xen

3.8.3.1 qemu-xen Has Been Dropped From the Xen Package

QEMU is a large software project that sees many bug and security fixes. Providing several different qemu binaries is challenging for maintenance, requiring bug and security fixes to be backported to all the different qemu sources.

The Xen package now uses qemu-system-x86_64 from the qemu package instead of providing its own qemu binary.

3.8.3.2 Support UEFI in Xen HVM Virtual Machines

libvirt and Xen now support UEFI for virtual machines. UEFI firmware is provided through the qemu-ovmf-x86_64 package.

3.8.3.3 GRUB Does Not Support vfb/vkbd Any More

The version of GRUB shipped with SLES 12 SP1 and SP2 does not support vfb/vkbd any more. This means that in Xen paravirtualized machines, there is no graphical display available while GRUB is active.

To be able to see and interact with GRUB, switch to the text-based xencon protocol: Modify the kernel parameter of the PV guest, add console=hvc0 xencons=tty, and connect with the command console DOMAINNAME of the libvirt toolstack.

3.8.3.4 libvirt XML Now Supports the External Block Scripts of Xen

The external block scripts of Xen, such as block-drbd, block-dmmd could formerly only be used with xl / libxl using the disk configuration syntax script=. libvirt did not support such external scripts and thus could not be used with disks configured with the block scripts.

External block scripts of Xen can now be used with libvirt by specifying base name of the block script in the <source> element of the disk. For example:

<source dev='dmmd:md;/dev/md0;lvm;/dev/vgxen/lv-vm01'/>
3.8.3.5 Support for the PVUSB Driver in Xen and the libvirt Xen Driver

libxl now has a PVUSB API which supports passing a USB device from the host to the guest domain via PVUSB. This functionality is also supported by the command line tool xl.

PVUSB support was also added to the libvirt libxl driver to use PVUSB functionality from the libvirt toolstack.

3.8.3.6 Xen: PV-OPS Kernel Supersedes kernel-xen

The Xen hypervisor functions have been ported over to the standard PV-OPS mechanism and are now included in the default kernel. As everything necessary is now provided by the default kernel, the kernel-xen package was removed.

3.8.4 Others

3.8.4.1 virt-convert: Support for Compressed Files in Within an OVA

According to the OVF 1.1.0 specification, OVA files can contain files compressed using gzip, for example, vmdk files. This case was previously not handled correctly.

In SLE 12 SP2, virt-convert will now correctly decompress gz files first and then convert them using qemu-img.

3.8.4.2 libiscsi Integration with QEMU

QEMU now integrates with libiscsi. This allows QEMU to access iSCSI resources directly and use them as virtual machine block devices. iSCSI-based disk devices can also be specified in the libvirt XML configuration. This feature is only available using the RAW image format, as the iSCSI protocol has some technical limitations.

3.8.4.3 DPDK Support for vhost-user Live Migration

Currently, the common back-end implementation to vhost-user is dpdk. To support vhost-user live migration, a feature bit called VHOST_USER_PROTOCOL_F_LOG_SHMFD is required on both the QEMU side and the vhost-user back-end side.

On the QEMU side, upstream version 2.6 already provides the required functionality. But on the DPDK side, the upstream release of DPDK 2.2.0 does not provide it.

The version of DPDK 2.2.0 shipped with SLE 12 SP2 is patched to provide the ability of vhost-user live migration.

3.8.4.4 wbemcli Now Allows Configuring the SSL/TLS version

Previously, it could be impossible to monitor certain servers that used very specific versions of the SSL/TLS protocols using wbemcli.

wbemcli can now be configured to use a specific SSL/TLS protocol version. To do so, use the environment variable WBEMCLI_CURL_SSLVERSION. Possible values are: SSLv2, SSLv3, TLSv1, TLSv1_0 (TLSv1.0), TLSv1_1 (TLSv1.1), TLSv1_2 (TLSv1.2).

3.8.4.5 Support for 3D Graphics in VMware Guest

The vmwgfx driver supports 3D with VMware hardware version 11.

4 AMD64/Intel 64 (x86_64) Specific Information

Information in this section pertains to the version of SUSE Linux Enterprise Server 12 SP2 for the AMD64/Intel 64 architectures.

4.1 Support for intel_idle and Hardware P States on Intel Skylake Processors Can Lead to Decreased Performance

On Intel processors from the generation code-named Skylake, some workloads can run slower on SLE 12 SP2 than they run on SLE 12 SP1.

SLE 12 SP1 and before, when running on Intel processors from the generation code-named Skylake, did not leverage hardware P states (HWP) or the intel_idle driver to save power. Instead, these processors ran at the maximum CPU frequency even when idle.

With SLE 12 SP2, hardware P states and the intel_idle driver are now supported on Skylake processors. This means that because the processor will not run at full speed at all times, some workloads can perform worse on SLE 12 SP2 than they do on SLE 12 SP1.

4.2 Kernel NOHZ_FULL Process Scheduler Mode

Under normal operation, the kernel interrupts process execution several hundred times per second for statistics collection and kernel internal maintenance tasks. Despite the interruptions being brief, they add up. This adds an unpredictable amount of time to process run time. Highly timing sensitive applications may be disturbed by this activity.

The SLE kernel now ships with adaptive tick mode (NOHZ_FULL) enabled by default to reduce the number of kernel interrupts. With this option enabled and conditions for adaptive tick mode fulfilled, the number of interrupts goes down to ones per second.

4.3 System and Vendor Specific Information

4.3.1 Support for Run-Time Allocation of Huge Pages With 1 GB Size

In previous versions of SLE, huge pages with a size of 1 GB could only be allocated via a kernel parameter at boot. This has the following drawbacks:

  • You cannot specify the NUMA node for allocation.

  • You cannot free these pages later without a reboot.

On the x86-64 architecture, SLE can now allocate and free 1 GB huge pages at system run time, using the same methods that are also used for regular huge pages.

However, you should still allocate 1 GB huge pages as early as possible during the run time. Otherwise, physical memory can become fragmented by other uses and the risk of allocation failure grows.

5 POWER (ppc64le) Specific Information

Information in this section pertains to the version of SUSE Linux Enterprise Server 12 SP2 for the POWER architecture.

5.1 Ceph Client Support on z Systems and POWER

On SLES 12 SP2 and SLES 12 SP3, z Systems and POWER machines can now function as SUSE Enterprise Storage (Ceph) clients.

This support is possible because the kernels for z Systems and POWER now have the relevant modules for CephFS and RBD enabled. The Ceph client RPMs for z Systems and POWER are included in SLE 12 SP3. Additionally, the QEMU packages for z Systems and POWER are now built against librbd.

5.2 Cluster Support and High Availability for POWER

Packages to facilitate cluster setup and to enable HA have been added to SUSE Linux High Availability Extension for POWER (LE).

5.3 The libcxl Userspace Library for CAPI Has Been Added

SLES now ships with the package libcxl. It provides the library of the same name that can be used for userspace CAPI.

The SLE SDK contains the corresponding development package, libcxl-devel.

5.4 Enhanced Support for System Call Filtering on POWER

Mode 2 of seccomp is now supported on POWER, allowing for fine-grained filtering of system calls. Support is available in both the kernel and in libseccomp.

5.5 Hardware Transactional Memory (HTM) support in glibc for POWER

Lock elision in the GNU C Library is available, but disabled by default. To enable it, set the environment variable GLIBC_ELISION_ENABLE to the value "yes".

5.6 Support for CXL Flash Storage Device Driver

The CXL flash storage device provides persistent, flash-based storage using CAPI technology.

6 IBM z Systems (s390x) Specific Information

Information in this section pertains to the version of SUSE Linux Enterprise Server 12 SP2 for the IBM z Systems architecture. For more information, see http://www.ibm.com/developerworks/linux/linux390/documentation_novell_suse.html

IBM zEnterprise 196 (z196) and IBM zEnterprise 114 (z114) further on referred to as z196 and z114.

6.1 Hardware

6.1.1 Improved Auto LUN Scan

Optimized tools and configuration enable improved WWPN and FCP LUN scanning. Manual intervention during installation or setup has been eliminated wherever possible.

6.1.2 Support for IPL Device in Any Subchannel Set

IPL devices are no longer restricted to subchannel set 0. The limitation is removed as of IBM zEnterprise 196 GA2.

6.1.3 Bus Awareness for z Systems in systemd

systemd now provides full and correct support for driver model buses specific to Linux on z Systems, such as ccw, ccwgroup, and zfcp.

6.2 Virtualization

6.2.1 Executing Hypervisor-Specific Actions During Boot

Depending on the hypervisor that a system runs on (such as z/VM, zKVM, or LPAR), during boot, differerent actions can be needed.

The service virtsetup is preconfigured to do that. To activate it, execute the following command:

systemctl enable virtsetup.service

To configure this service in more detail, see the file /etc/sysconfig/virtsetup. You can also edit the file through YaST:

yast2 sysconfig

6.2.2 VMUR Print Spool Options for Linux

Linux guests are now better integrated into the z/VM print solution. It is now possible to specify the spool options CLASS and FORM together with the print command of the VMUR tool.

6.2.3 zKVM: SIE Capability Exposed to User Space

Userspace applications can now query whether the Linux instance can act as a hypervisor by checking for the SIE (Start Interpretive Execution) capability. This is useful, for example, in continuous integration (CI) environments.

6.3 Storage

6.3.1 iSCSI Devices Not Enabled After Installation

When installing SLES 12 SP2, iSCSI devices may not be enabled after installation.

When configuring iSCSI volumes, make sure to set start mode to automatic. onboot is only valid for iSCSI devices which are supposed to be activated from the initrd, that is, when the system is booted from iSCSI. However, that is currently not supported on z Systems.

6.3.2 Ceph Client Support on z Systems and POWER

On SLES 12 SP2 and SLES 12 SP3, z Systems and POWER machines can now function as SUSE Enterprise Storage (Ceph) clients.

This support is possible because the kernels for z Systems and POWER now have the relevant modules for CephFS and RBD enabled. The Ceph client RPMs for z Systems and POWER are included in SLE 12 SP3. Additionally, the QEMU packages for z Systems and POWER are now built against librbd.

6.3.3 Query Host Access to Volume Support

You can now concurrently access DASD volumes from different operating system instances. Applications can now query whether a DASD volume is online within another operating system instance by querying the storage server for the online status of all attached hosts. The command lsdasd can display this information, and the commands zdsfs, fdasd, and dasdfmt can evaluate it.

6.3.4 Disk Mirroring with Real-Time Enhancement for z Systems

This functionality is included in SLES 12 SP2 as a technology preview.

6.4 Network

6.4.1 Port Name for Open Systems Adapter (OSA) Is No Longer Needed

For some time, systems sharing an OSA with the z/OS operating system needed to specify a port name that matched what was used on z/OS. Not providing a port name, or providing a non-matching port name would result in the network interface not activating.

IBM has modified the microcode on all the OSAs currently available so that a port name is not needed. As a result, SUSE has removed the prompt for it from the installer. If a port name is provided via the installation parameter file, an informational message is displayed:

**
** The Portname parameter is no longer needed. Please do not specify it. **
**

6.4.2 10GbE RoCE Express Feature for RDMA

SLES 12 SP2 supports the 10GbE RoCE Express feature on zEC12, zBC12 and IBM z13 via the Ethernet device using TCP/IP traffic without restrictions. Before using this feature on an IBM z13, make sure that the minimum required service is applied: z/VM APAR UM34525 and HW ycode N98778.057 (bundle 14). Use the default MTU size (1500).

SLES 12 SP2 now includes support for RDMA enablement and DAPL/OFED for z Systems. With the Mellanox virtualization support (SR-IOV) the limitation for LPAR use only on an IBM zEC12 or zBC12 is removed and RDMA can be used on an IBM z13.

6.4.3 Bridging HiperSockets to Ethernet

A HiperSocket port can now be configured to accept Ethernet frames to unknown MAC addresses. This enables it to be used as a member of a software bridge. Control and report of the bridge port status of the HiperSocket port and the udev events are performed via new sysfs attributes.

6.4.4 IPv6 Priority Queuing Added to qeth Device Driver

Priority queuing is now supported for IPv6, similarly to IPv4. This especially improves Linux Live Guest Migration by using IPv6 to minimize impact on workload traffic and enables priority queuing for all applications that use IPv6 QoS traffic operations.

6.4.5 Layer 2 Offloads Enabled

Classic OSA operation in layer 3 mode provides numerous offload operations, exchanging larger amounts of data between the operating system and the OSA adapter. The qeth device driver now also provides large send/receive and checksum offload operations for layer 2 mode.

6.4.6 IPv6 Support in snIPL

The tool for remote systems management for Linux, snIPL, now includes IPv6 support. This broadens the set of environments that snIPL supports and simplifies moving from IPv4 to IPv6.

6.4.7 Enhanced OSA Network to Receive All Frames Through a Network Interface

Enhancements in the OSA device driver enable setting network interfaces into promiscuous mode. The mode can provide outside connectivity for virtual servers by receiving all frames through a network interface.

In OpenStack environments, Open vSwitch is one of the connectivity options that use this feature.

6.5 Security

6.5.1 Support for DBRG in libica

The libica support for the generation of pseudo-random numbers for the "Deterministic Random Bit Generator" (DRBG) was enhanced to comply with updated security specifications (NIST SP 800-90A).

6.5.2 Monitoring CPACF Crypto Activity

This feature enables the monitoring of CPACF crypto activity in the Linux image, in the kernel, and in userspace. A configurable crypto-activity counter allows switching monitoring of CPACF crypto activity on or off for selected areas to verify and monitor specific needs in the crypto stack.

6.5.3 Support for Dynamic Traces in openCryptoki

Dynamic tracing in openCryptoki now allows starting and stopping tracing of all openCryptoki API calls and the related tokens while the application is running. This also allows using cryptography in the Java Security Architecture (JCA/JCE) which transparently falls back to software cryptography. Enhanced tracing can now identify whether cryptographic hardware is actually used.

6.5.4 CPACF MSA 4: Support for the GCM mechanism in openCryptoki

The openCryptoki ICA includes support for a new mechanism supported by CPACF MSA 4. GCM is a highly recommended mechanism for use with TLS 1.2.

6.5.5 Support for CCA Master Key Change for openCryptoki CCA Token

We now provide a tool to change master keys on the CCA co-processor without losing the encrypted data. This helps to stay compliant with enhanced industry regulations and company policies.

6.6 Reliability, Availability, Serviceability (RAS)

6.6.1 CUIR: Enhanced Scope Detection

The Linux support for CUIR (Control Unit Initiated Reconfiguration), which enables concurrent storage service with no or minimized down time, has been extended to include Linux running as a z/VM guest.

6.7 Performance

6.7.1 Extended CPU Performance Metrics in HYPFS for Linux z/VM guests

The HYPFS has been extended to provide the "diag 0C data" also for Linux z/VM guests that distinguish "management time" spent as part of CPU load.

6.7.2 GCC SIMD Performance Tuning

Enhanced instruction support in GCC improves application performance. Optimized applications can now also use SIMD instructions.

6.7.3 IBM z13 Hardware Instructions in glibc

Support of the IBM z13 hardware instructions in glibc provides improved application performance.

6.7.4 Fake NUMA Support

Splitting the system memory into multiple NUMA nodes and distributing memory without using real topology information about the physical memory can improve performance. This is especially true for large systems. This feature is turned off by default but can be enabled for a system from the command line.

6.8 Miscellaneous

6.8.1 Enable Boot Parameter quiet for Better Visibility of Password Prompts

In the default configuration of SLES 12 SP2 for z Systems, the boot parameter quiet is disabled, so the system console shows more useful log messages. This has the drawback that the increased amount of log messages can hide a password prompt, such as the prompt for decrypting devices at boot.

To make the password prompt more visible among the system messages, add the boot parameter quiet when there are encrypted devices that need to be activated at system boot.

6.8.2 Installing From DVD/USB Drive of the HMC

You can now install from media in the DVD/USB drive of the Hardware Management Console (HMC).

To do so:

  • Add install=hmc:/ to the parm file or kernel options.

  • Alternatively, in manual mode, in linuxrc, choose Start Installation > Installation > Hardware Management Console. The installation medium must be inserted in the HMC.

There are two .ins files available which you can install with:

  • suse.ins to install with network access. When using this option, do not forget to configure the network in linuxrc before starting the installation. There is no way to pass boot parameters later and it is very likely that you will need network access. In linuxrc, go to Start Installation > Network Setup.

  • susehmc.ins which allows installing without network access.

Important: Wait until the Linux system is booting before granting access to the DVD in the HMC. IPLing seems to disrupt the connection between the HMC and the LPAR in some way. If the first attempt to use it fails, you can grant the access and retry the option HMC.

Note: Because of the transitory nature of the assignment, the DVD that was used during installation will not be kept as a repository. If you need an installation repository there, register and use the online repository.

7 ARM 64-Bit (AArch64) Specific Information

Information in this section pertains to the version of SUSE Linux Enterprise Server 12 SP2 for the AArch64 architecture.

7.1 KVM on AArch64

KVM virtualization has been enabled and is supported on some system-on-chip platforms for mutually agreed-upon partner-specific use cases. It is only supported on partner certified hardware and firmware. Not all QEMU options and backends are available on AArch64. The same statement is applicable for other virtualization tools shipped on AArch64.

7.2 Toolchain Module Enabled in Default Installation

The system compiler (gcc4.8) is not supported on the AArch64 architecture. To work around this issue, you previously had to enable the Toolchain module manually and use the GCC version from that module.

On AArch64, the Toolchain Module is now automatically pre-selected after registering SLES during installation. This makes the latest SLE compilers available on all installations. You now only need to make sure to also use that compiler.

Important
Important: When Using AutoYaST, Make Sure to Enable Toolchain Module

Be aware that when using AutoYaST to install, you have to explicitly add the Toolchain module into the XML installation profile.

7.3 GICv2 and GICv3 Interrupt Controller Support in QEMU

KVM/QEMU now works with GICv2 and GICv3 interrupt controllers that implement virtualization capabilities.

7.4 Boot Requirements for AppliedMicro X-Gene 1

The AppliedMicro X-C1 Server Development Platform (Mustang) ships with U-Boot based firmware. To install SUSE Linux Enterprise Server 12 SP2, the firmware needs to be updated to the UEFI based firmware version 3.06.15 or newer.

Other server systems, such as Gigabyte MP30, may also require a firmware update for an optimal experience. For details, contact your vendor.

7.5 ARM AArch64 System-on-Chip Platform Driver Enablement

For ARM based systems to boot SUSE Linux Enterprise Server, some chipset-specific drivers are needed.

The following System-on-Chip (SoC) platforms have been enabled for SP2:

  • AMD Opteron A1100

  • AppliedMicro X-Gene 1

  • AppliedMicro X-Gene 2

  • Cavium ThunderX

  • NXP QorIQ LS2085A / LS2045A, LS2080A / LS2040A

  • Xilinx UltraScale+ MPSoC

8 Driver Updates

8.1 Network Drivers

8.1.1 Support Status of Ethernet Drivers

Ethernet drivers have been added between kernel versions 3.12 (SLES 12 GA) and 4.4 (SLES 12 SP2).

The support status of Ethernet drivers has been updated for SLE 12 SP2 and below is the list of newly supported drivers.

  • Agere Systems ET1310 (et131x)

  • Qualcomm Atheros AR816x/AR817x PCI-E (alx)

  • Broadcom BCM573xx (bnxt_en)

  • JMicron JMC2x0 PCI-E (jme)

  • QLogic FastLinQ 4xxxx (qede)

  • SMC 83c170 EPIC series (epic100)

  • SMSC LAN911x/LAN921x (smsc911x)

  • SMSC LAN9420 PCI (smsc9420)

  • STMMAC 10/100/1000 PCI (stmmac-pci)

  • WIZnet W5100 (w5100)

  • WIZnet W5300 (w5300)

  • FUJITSU Extended Socket Network (fjes)

  • SMSC95XX USB (smsc95xx)

  • Xilinx LL TEMAC (ll_temac)

  • APM X-Gene (xgene-enet)

  • Cavium Thunder (nicpf, nicvf, thunder_bgx)

8.2 Other Drivers

8.2.1 Support for New Intel Processors

This Service Pack adds support for the following Intel processors:

  • Intel® Xeon® Processor E3-1200/1500 v5 Product Family

  • Intel® Xeon Phi™ Product Family x200

9 Packages and Functionality Changes

This section comprises changes to packages, such as additions, updates, removals and changes to the package layout of software. It also contains information about modules available for SUSE Linux Enterprise Server. For information about changes to package management tools, such as Zypper or RPM, see Section 3.5, “Systems Management”.

9.1 New Packages

9.1.1 Mutt Has Been Updated to 1.6.0

Mutt has been updated to version 1.6.0. This version has the following new features:

  • Better internationalization support: UTF-8 mailbox support for IMAP and improved support for internationalized email and SMTPUTF8

  • $use_idn has been renamed to $idn_decode

  • Expandos for comma-separated lists of To (%r) and CC recipients (%R).

  • Improved handling of drafts: -E command-line argument for edit draft or include files, $resume_draft_files and $resume_edited_draft_files to control processing of draft files, and support for multipart draft files

  • $reflow_space_quotes allows format=flowed email quotes to be displayed with spacing between them.

  • The S/MIME message digest algorithm is now specified using the option $smime_sign_digest_alg. $smime_sign_command should be modified to include -md %d.

  • For classic GPG mode, set $pgp_decryption_okay to verify that multipart/encrypted mails are actually encrypted.

  • By default, mailto URL header parameters are restricted to body and subject. To add or remove allowed mailto URL header parameters, use mailto_allow and unmailto_allow.

  • $hostname is set differently: the domain will now be determined using DNS calls

9.1.2 targetcli-fb Has Been Added

In addition to the established tool targetcli, there is now also its enhanced version targetcli-fb available. New users are encouraged to deploy targetcli-fb.

9.1.3 Devilspie 2 Has Been Added

Desktop users often want the size and position of windows to remain the same, even across application restarts. Such functionality usually has to be implemented at the application level but not all applications do so.

In SUSE Linux Enterprise 12 SP2, Devilspie 2 (package devilspie2) has been added. Desvilspie 2 is a window matching utility that allow you to script actions on windows as they are created, such as maximizing windows or setting their size and position.

9.1.4 openldap2-ppolicy-check-password Has Been Added: OpenLDAP Password Strength Policy Enforcer

To allow evaluating and enforcing password strength in an OpenLDAP deployment, the package openldap2-ppolicy-check-password has been added. It is an OpenLDAP password policy plugin which evaluates and enforces strength in new user passwords, and denies weak passwords in password change operations. Configuration options of the plugin allow system administrators to adjust password strength requirements.

9.2 Updated Packages

9.2.1 Ceph Client Enablement Has Been Upgraded to Ceph Jewel

SUSE Enterprise Storage 3 and later versions expose additional functionality and performance to upgraded clients, such as the use of advanced RBD features and improved CephFS integration. While SUSE Enterprise Storage 3 is backwards-compatible with older clients, the full benefits are only available to newer clients.

As part of SUSE Linux Enterprise Server 12 Service Pack 2, the Ceph client code, as provided by ceph-common and the related library packages, has been upgraded to match the latest SUSE Enterprise Storage release.

This update also includes rebuilt versions of the KVM integration to take advantage of these.

9.2.2 Upgrade of libStorageMgmt to Version 1.3.2

libStorageMgmt allows programmatically managing storage hardware in a vendor-neutral way.

In SLES 12 SP2, libStorageMgmt was upgraded to version 1.3.2. This version fixes several bugs and adds the ability to more retrieve disk information, such as information on batteries and the list of local disks.

9.2.3 Glibc Has Been Upgraded to Version 2.22

glibc has been upgraded to meet demands in transactional memory handling and memory protection and to gain performance optimizations for modern platforms.

9.2.4 lsof Has Been Updated to Version 4.89

lsof has been updated from version 4.84 to 4.89. The changelog can be found in the file /usr/share/doc/packages/lsof/DIST.

9.2.5 Qt 5 Has Been Updated to 5.6.1

The Qt 5 libraries were updated to 5.6.1, a Qt 5.6 LTS based release. Qt 5.6.1 includes new features and security fixes for known vulnerabilities over Qt 5.5.1 (the version shipped in an upgrade to SP1).

This release includes many bug fixes and changes that improve performance and reduce memory consumption.

For security reasons, the MNG and JPEG2000 image format plugins are not shipped anymore, because the underlying MNG and JPEG2000 libraries have known security issues.

New features include:

  • Better support for high-DPI screens

  • Update of QtWebEngine which updates the included Chromium snapshot to version 45 and now uses many of the system libraries instead of bundled ones

  • New Qt WebEngineCore module for new low-level APIs

  • The Qt Location module is not fully supported.

  • Improved compatibility with C++11 and the STL

  • New QVersionNumber class

  • Added support for HTTP redirection in QNetworkAccessManager

  • Improved support for OpenGL ES 3

  • Qt Multimedia got a new PlayList QML type and an audio role API for the media player

  • Qt Canvas 3D now supports Qt Quick Items as textures and can directly render to the QML scenes foreground or background

  • Qt 3D has received many improvements and new functionality

  • Many other features and bugfixes

As part of this update, Qt Creator has been updated to 4.0.1 (from Qt Creator 3.5.1 shipped as an update to SP1).

New features of Qt Creator include:

  • Clang static analyzer integration, extended QML profiler features, path editor of Qt Quick Designer and auto test integration (experimental) are now available

  • The Clang code model is now automatically used if the (experimental) plugin is turned on

  • Improved workflow for CMake-based projects

  • The Analyze mode was merged with Debug mode, so that the new unified Debug mode includes the Debugger, Clang Static Analyzer, Memcheck, Callgrind and QML Profiler tools

  • Many other features and bugfixes

9.2.6 RPM Ignores the BuildRoot Directive in Spec Files

In versions of RPM greater than 4.6.0, the behavior of the BuildRoot directive was changed compared to prior versions. RPM now enforces using a build root for all packages and ignores the BuildRoot directive in spec files. By default, rpmbuild places the build root inside %{_topdir}. However, this can be changed through macro configuration.

In the version of RPM shipped with SUSE Linux Enterprise 12 (and later), the BuildRoot directive of spec files is silently ignored. However, it is recommended to keep the BuildRoot directive in spec files for backward compatibility with earlier versions of SUSE Linux Enterprise (and RPM).

For more information, see the RPM 4.6.0 release notes at http://rpm.org/wiki/Releases/4.6.0 (http://rpm.org/wiki/Releases/4.6.0).

9.2.7 OpenSSH Has Been Updated to Version 7.2

OpenSSH received numerous changes and improvements in the last years. To ease further maintenance, OpenSSH was upgraded to a more current release.

Note that the SSHv1 protocol is no longer supported.

9.2.8 Puppet Has Been Updated from 3.6.2 to 3.8.5

Puppet has been updated from 3.6.2 to 3.8.5. All releases between these two versions should only bring Puppet 3 backward-compatible features and bug and security fixes.

For more information, read the following release notes:

In particular, you should pay attention to the following upgrade notes and warnings:

9.2.9 Changes in Behavior Between coreutils 8.22 and 8.25

SLE 12 SP1 shipped with coreutils 8.22. SLE 12 SP2 ships with coreutils 8.25. This new release brings a number of changes in behavior:

  • base64: base64 no longer supports --wrap parameters in hexal or octal format. This improves support for decimals with leading zeros.

  • chroot: Using / as the argument no longer implicitly changes the current directory to /. This allows changing user credentials for a single command only.

  • chroot: --userspec will now unset supplemental groups associated with root and instead use the supplemental groups of the specified user.

  • cut: Using -d$'\n' will again output lines identified in the --fields list (this behavior had been changed in version 8.21 and 8.22). Note that this functionality is non-portable and will result in the delayed output of lines.

  • date: The option --iso-8601 now uses the timezone format +00:00 rather than +0000. This "extended" format is preferred by the ISO 8601 standard.

  • df: df now prefers sources towards the root of a device when eliding duplicate bind -mounted entries.

  • df: df no longer suppresses separate exports of the same remote device, as these are generally explicitly mounted. The --total option does still suppress duplicate remote file systems.

  • join, sort, uniq: When called with --zero-terminated, these commands now treat \n as a field delimiter.

  • ls: If neither of the environment variables LS_COLORS and COLORTERM is set and the environment variable TERM is empty or unknown, ls will now not output colors even with --colors=always.

  • ls: When outputting to a terminal, ls now quotes file names unambiguously and appropriate for use in a shell.

  • mv: mv no longer supports moving a file to a hard link. If you try, it issues an error. The prior implementation was susceptible to races in the presence of multiple mv instances which could result in both hard links being deleted. Also, on case-insensitive file systems like HFS, mv would remove a hardlinked file if called like mv file File.

  • numfmt: The options --from-unit and --to-unit now interpret suffixes as SI units, and IEC (power of 2) units are now specified by appending i.

  • tee: If there are no more writable outputs, tee will exit early.

  • tee: tee does not treat the file operand - as meaning standard output any longer. This allows for better POSIX conformance.

  • timeout: The option --foreground no longer sends SIGCONT to the monitored process, as this was seen to cause intermittent issues with GDB for example.

9.2.10 openSSL Has Been Updated to Version 1.0.2

openSSL has been updated from version 1.0.1 to 1.0.2 which is a compatible minor version update. This will help future maintenance, and also brings many bug fixes.

The update to openSSL 1.0.2 should be transparent to existing programs.

However, there were some functional changes were done: SSL 2 support is now fully disabled and certain weak ciphers are no longer built in.

9.3 Removed and Deprecated Functionality

9.3.1 Perl Bindings for Cyrus Have Been Removed

With SLE 12 SP2, the packages perl-Cyrus-IMAP and perl-Cyrus-SIEVE-managesieve have been removed from the media.

9.3.2 librpcsecgss3 Has Been Removed

librpcsecgss (packages: librpcsecgss3, librpcsecgss-devel) has been removed. With the release of libtirpc, the development of libsecgss stopped and it fell out of use. We recommend using libtirpc instead.

9.3.3 Docker Compose Has Been Removed from the Containers Module

Docker Compose is not supported as a part of SUSE Linux Enterprise Server 12. While it was temporarily included as a Technology Preview, testing showed that the technology was not ready for enterprise use.

SUSE's focus is on Kubernetes which provides better value in terms of features, extensibility, stability and performance.

9.3.4 libusnic_verbs-rdmav2 and libusnic_verbs-rdmav2-pingpong Are Now Obsolete

Functionality previously shipped in the packages libusnic_verbs-rdmav2 and libusnic_verbs-rdmav2-pingpong has been integrated into libibverbs.

9.3.5 Packages Removed with SUSE Linux Enterprise Server 12

The packages listed below were removed with the major release of SUSE Linux Enterprise Server 12.

9.3.5.1 Nagios Server Now Part of a SUSE Manager Subscription

Support for Icinga (a successor of Nagios) will not be part of the SUSE Linux Enterprise Server 12 subscription.

Fully supported Icinga packages for SUSE Linux Enterprise Server 12 will be available as part of a SUSE Manager subscription. In the SUSE Manager context we will be able to deliver better integration into the monitoring frameworks.

More frequent updates on the monitoring server parts than in the past are planned.

9.3.6 Packages Removed with SUSE Linux Enterprise Server 12 SP1

The packages listed below were removed with the release of SUSE Linux Enterprise Server 12 SP1.

9.3.6.1 wpa_supplicant Replaces xsupplicant

In SUSE Linux Enterprise 12 SP1 and 12 SP2, xsupplicant was removed entirely.

For pre-authentication of systems via network (including RADIUS) and specifically wireless connections, install the wpa_supplicant package. wpa_supplicant now replaces xsupplicant. wpa_supplicant provides better stability, security and a broader range of authentication options.

9.3.7 Packages and Features to Be Removed in the Future

9.3.7.1 Server Component of Puppet Is Deprecated

Puppet is shipped as part of the Advanced Systems Management module for SLES. Currently, this module contains both the Puppet client and the Puppet server.

Starting with the packages for Puppet 4 which will be released in early 2017, SUSE will only ship the Puppet client (currently packaged as puppet, in the future packaged as rubygem-puppet) but not the Puppet server (package puppet-server).

The Puppet Server package for SLES is provided by upstream now and it can be downloaded at https://yum.puppetlabs.com/sles/12/PC1/x86_64/ (https://yum.puppetlabs.com/sles/12/PC1/x86_64/) (puppetserver-2.6.0-1.sles12.noarch.rpm).

For a period of 6 months after the release of the Puppet 4 package, SUSE will continue to provide and support packages for Puppet 3. SUSE will not support the migration of the server package.

Before updating the client packages to version 4, see the prerequisites listed at https://docs.puppet.com/puppet/4.8/ (https://docs.puppet.com/puppet/4.8/) (Section "Installing and upgrading", Subsection "Upgrade: From Puppet 3.x").

9.4 Changes in Packaging and Delivery

9.4.1 GNOME Desktop: Clicking "Open in Terminal" on the Desktop Now Opens the Home Directory

When right-clicking the GNOME desktop and selecting Open in Terminal, GNOME Terminal will now open with the working directory set to the home directory (~) and not set to the Desktop directory (~/Desktop). This happens because the package nautilus-extension-terminal is now installed by default.

To switch to the former behavior, first uninstall nautilus-extension-terminal and then install nautilus-open-terminal. However, note that the package nautilus-open-terminal may not be provided in future service packs.

9.4.2 Change of OpenMPI Behavior for Plugin Developers

To be compliant with the upstream version of OpenMPI, the source configuration option --with-devel-header has been removed. This only affects developers of OpenMPI plugins outside of the source tree.

Developers of plugins outside of the source tree need to recompile the source with the option --with-devel-header added. All other users are not affected.

9.4.3 Support for Intel OPA Fabrics Moved to mvapich2-psm2 Package

The version of the package mvapich2-psm originally shipped with SLES 12 SP2 and SLES 12 SP3 exclusively supported Intel Omni-Path Architecture (OPA) fabrics. In SLES 12 SP1 and earlier, this package supported the use of Intel True Scale fabrics instead.

This issue is fixed by a maintenance update providing an additional package named mvapich2-psm2 which only supports Intel OPA, whereas the original package mvapich2-psm only supports Intel True Scale fabrics again.

If you are currently using mvapich2-psm together with Intel OPA fabrics, make sure to switch to the new package mvapich2-psm2 after this maintenance update.

9.5 Modules

This section contains information about important changes to modules. For more information about available modules, see Section 1.7.1, “Available Modules”.

9.5.1 libgcrypt11 Available from the Legacy Module

The Legacy module now provides a package for libgcrypt11. This enables running applications built on SLES 11 against libgcrypt11 on SLES 12.

9.5.2 PHP 7 Packages Have Been Added to the Web and Scripting Module

So far, the Web and Scripting module for SLES contained packages for PHP 5 only.

The Web and Scripting module for SLES now additionally contains packages for PHP 7. For a detailed overview of changes over PHP 5, see https://secure.php.net/releases/7_0_0.php (https://secure.php.net/releases/7_0_0.php).

9.6 SDK

9.6.1 Byebug Has Been Added

Byebug is a simple-to-use, feature-rich Ruby 2 debugger that is also used to debug YaST. It uses the TracePoint API and the Debug Inspector API. For speed, it is implemented as a C extension.

It allows you to see what is going on inside a Ruby program while it executes and offers traditional debugging features such as stepping, breaking, evaluating, and tracking.

10 Technical Information

This section contains information about system limits, a number of technical changes and enhancements for the experienced user.

When talking about CPUs, we use the following terminology:

CPU Socket

The visible physical entity, as it is typically mounted to a motherboard or an equivalent.

CPU Core

The (usually not visible) physical entity as reported by the CPU vendor.

On IBM z Systems, this is equivalent to an IFL.

Logical CPU

This is what the Linux Kernel recognizes as a "CPU".

We avoid the word "thread" (which is sometimes used), as the word "thread" would also become ambiguous subsequently.

Virtual CPU

A logical CPU as seen from within a Virtual Machine.

10.1 Virtualization: Network Devices Supported

SLES 12 supports the following virtualized network drivers:

  • Full virtualization: Intel e1000

  • Full virtualization: Realtek 8139

  • Paravirtualized: QEMU Virtualized NIC Card (virtio, KVM only)

10.2 Virtualization: Devices Supported for Booting

SLE12 support VM guest to boot from:

  • Parallel ATA (PATA/IDE)

  • Advanced Host Controller Interface (AHCI)

  • Floppy Disk Drive (FDD)

  • virtio-blk

  • virtio-scsi

  • Preboot eXecution Environment (PXE) ROMs (for supported Network Interface Cards)

Boot from USB and PCI pass-through devices are not supported.

10.3 Virtualization: Supported Disks Formats and Protocols

The following disk formats support read-write access (RW):

  • raw

  • qed (KVM only)

  • qcow2

The following disk formats support read-only access (RO):

  • vmdk

  • vpc

  • vhd / vhdx

The following protocols can be used for read-only access (RO) to images:

  • http, https

  • ftp, ftps, tftp

When using Xen, the qed format will not be displayed as a selectable storage in virt-manager.

Note
Note: Parameter Unprivileged SG_IO (unpriv_sgio) Is Not Supported

The parameter for unprivileged SG_IO (unpriv_sgio) depends on non-standard kernel patches that are not included in the SLES 12 kernel. Trying to attach a disk using this parameter will result in an error.

10.4 Kernel Limits

http://www.suse.com/products/server/technical-information/#Kernel

This table summarizes the various limits which exist in our recent kernels and utilities (if related) for SUSE Linux Enterprise Server 12 SP2.

SLES 12 SP2 (Linux 4.4) AMD64/Intel 64 (x86_64)IBM z Systems (s390x)POWER (ppc64le)AArch64 (ARMv8)

CPU bits

64

64

64

64

Maximum number of logical CPUs

8192

256

2048

128

Maximum amount of RAM (theoretical/certified)

> 1 PiB/64 TiB

10 TiB/256 GiB

1 PiB/64 TiB

256 TiB/n.a.

Maximum amount of user space/kernel space

128 TiB/128 TiB

n.a.

2 TiB/2 EiB

256 TiB/128 TiB

Maximum amount of swap space

Up to 29 * 64 GB (x86_64) or 30 * 64 GB (other architectures)

Maximum number of processes

1048576

Maximum number of threads per process

Upper limit depends on memory and other parameters (tested with more than 120,000).

Maximum size per block device

Up to 8 EiB on all 64-bit architectures

FD_SETSIZE

1024

10.5 KVM Limits

SLES 12 SP2 Virtual Machine (VM) Limits

Maximum VMs per host

Unlimited (total number of virtual CPUs in all guests being no greater than 8 times the number of CPU cores in the host)

Maximum Virtual CPUs per VM

240

Maximum Memory per VM

4 TiB

Virtual Host Server (VHS) limits are identical to those of SUSE Linux Enterprise Server.

10.6 Xen Limits

Since SUSE Linux Enterprise Server 11 SP2, we removed the 32-bit hypervisor as a virtualization host. 32-bit virtual guests are not affected and are fully supported with the provided 64-bit hypervisor.

SLES 12 SP2 Virtual Machine (VM) Limits

Maximum number of virtual CPUs per VM

64

Maximum amount of memory per VM

16 GiB x86_32, 511 GiB x86_64

SLES 12 SP2 Virtual Host Server (VHS) Limits

Maximum number of physical CPUs

256

Maximum number of virtual CPUs

256

Maximum amount of physical memory

5 TiB

Maximum amount of Dom0 physical memory

500 GiB

Maximum number of block devices

12,000 SCSI logical units

  • PV:  Paravirtualization

  • FV:  Full virtualization

For more information about acronyms, see the virtualization documentation provided at https://www.suse.com/documentation/sles-12/.

10.7 File Systems

https://www.suse.com/products/server/technical-information/#FileSystem

10.7.1 Btrfs File System Going Read-only When Executing Balance Operation

When executing a balance operation on a Btrfs file system on which there is almost no free space available, the file system may go into a forced read-only mode.

Balancing a Btrfs file system involves relocating extents. When there is not enough free space available, this fails. The error is unconditionally overwritten and success is returned, but the extent has not actually been relocated. There is no data loss at this point, but balancing will fail on every invocation.

This has been fixed via a maintenance update. If you are seeing this issue, make sure your system is up-to-date.

For more information, see https://www.suse.com/support/kb/doc?id=7018233 (https://www.suse.com/support/kb/doc?id=7018233).

10.7.2 Comparison of Supported File Systems

SUSE Linux Enterprise was the first enterprise Linux distribution to support journaling file systems and logical volume managers back in 2000. Later, we introduced XFS to Linux, which today is seen as the primary work horse for large-scale file systems, systems with heavy load and multiple parallel reading and writing operations. With SUSE Linux Enterprise 12, we went the next step of innovation and started using the copy-on-write file system Btrfs as the default for the operating system, to support system snapshots and rollback.

+ supported
unsupported

FeatureBtrfsXFSExt4ReiserFS **OCFS 2 ***

Data/metadata journaling

N/A *

– / +

 

– / +

– / +

Journal internal/external

N/A *

+ / +

+ / –

  

Offline extend/shrink

+ / +

– / –

+ / +

+ / –

 

Online extend/shrink

+ / +

+ / –

+ / –

+ / –

+ / –

Inode allocation map

B-tree

B+-tree

table

u. B*-tree

table

Sparse files

+

    

Tail packing

+

+

 

Defrag

+

   

ExtAttr/ACLs

+ / +

    

Quotas

+

    

Dump/restore

+

  

Block size default

4 KiB

Maximum file system size

16 EiB

8 EiB

1 EiB

16 TiB

4 PiB

Maximum file size

16 EiB

8 EiB

1 EiB

1 EiB

4 PiB

Support in products

SLE

SLE

SLE

SLE

SLE HA

  • * Btrfs is a copy-on-write file system. Rather than journaling changes before writing them in-place, it writes them to a new location and then links the new location in. Until the last write, the new changes are not committed. Due to the nature of the file system, quotas are implemented based on subvolumes (qgroups).

    The block size default varies with different host architectures. 64 KiB is used on POWER, 4 KiB on most other systems. The actual size used can be checked with the command getconf PAGE_SIZE.

  • ** ReiserFS is supported for existing file systems. The creation of new ReiserFS file systems is discouraged.

  • *** OCFS2 is fully supported as part of the SUSE Linux Enterprise High Availability Extension.

The maximum file size above can be larger than the file system's actual size due to usage of sparse blocks. Note that unless a file system comes with large file support (LFS), the maximum file size on a 32-bit system is 2 GB (2^31 bytes). Currently all of our standard file systems (including Ext3 and ReiserFS) have LFS, which gives a maximum file size of 2^63 bytes in theory. The numbers in the above tables assume that the file systems are using 4 KiB block size. When using different block sizes, the results are different, but 4 KiB reflects the most common standard.

In this document: 1024 Bytes = 1 KiB; 1024 KiB = 1 MiB; 1024 MiB = 1 GiB; 1024 GiB = 1 TiB; 1024 TiB = 1 PiB; 1024 PiB = 1 EiB. See also http://physics.nist.gov/cuu/Units/binary.html.

NFSv4 with IPv6 is only supported for the client side. An NFSv4 server with IPv6 is not supported.

The version of Samba shipped with SUSE Linux Enterprise Server 12 SP2 delivers integration with Windows 7 Active Directory domains. In addition, we provide the clustered version of Samba as part of SUSE Linux Enterprise High Availability Extension 12 SP2.

10.7.3 Supported Btrfs Features

The following table lists supported and unsupported Btrfs features across multiple SLES versions.

+ supported
unsupported

FeatureSLES 11 SP4SLES 12 GASLES 12 SP1SLES 12 SP2
Copy on Write++++
Snapshots/Subvolumes++++
Metadata Integrity++++
Data Integrity++++
Online Metadata Scrubbing++++
Automatic Defragmentation
Manual Defragmentation++++
In-band Deduplication
Out-of-band Deduplication++++
Quota Groups++++
Metadata Duplication++++
Multiple Devices+++
RAID 0+++
RAID 1+++
RAID 10+++
RAID 5
RAID 6
Hot Add/Remove+++
Device Replace
Seeding Devices
Compression++
Big Metadata Blocks+++
Skinny Metadata+++
Send Without File Data+++
Send/Receive+
Inode Cache
Fallocate with Hole Punch+

12 Colophon

Thanks for using SUSE Linux Enterprise Server in your business.

The SUSE Linux Enterprise Server Team.

Print this page