ASMLIB provides a range of features that are dependent on facilities in the system kernel and on the version of the ASMLIB software installed. This table provides a reference that you can use to identify the appropriate minimum software version and platform and kernel combination that you need to take advantage of the different features in ASMLIB.

^Footnote 1 I/O Filter is available for x86_64 only.

^Footnote 2 kernel-uek-5.4.17-2136.342.1 or later is required to support Automatic Block Size Selection on UEKR6.

^Footnote 3 io_uring interface marked Tech Preview.

^Footnote 4 io_uring interface marked Tech Preview.

^Footnote 5 I/O Filter is available for x86_64 only.

^Footnote 6 oracleasm-kmp-default-2.0.8-150600.16.13.2 or later is required to support Automatic Block Size Selection.

Oracle ASMLIB is updated to version 3.1.1. The following notable changes are included:

• ASMLIB can now discover ASM disk labels on unpartitioned block devices.
• ASMLIB attempts additional retries when the kernel reports transient I/O errors.
• ASMLIB no longer lowers maxio for devices which report an invalid optimal I/O size.

ASMLIB is updated to version 3.1. The following notable changes are included:

• End-to-End Data Integrity: ASMLIB can use end-to-end data integrity protection when running on UEK R8 or later. This feature adds a layer of protection against data corruption by attaching integrity metadata to each I/O. The attached integrity metadata can be validated by I/O controller and storage device on write operations, and by ASM on read operations.

  End-to-end data integrity protection requires ASM disks to be placed on storage devices formatted with the T10 Protection Information, Type 1. Associated SCSI storage controllers must support Data Integrity Extensions (DIX). NVMe storage controllers must support the separate metadata buffer feature.

• Automatic Block Size Selection: ASM can now select whether to use the logical or the physical block size on a per-I/O basis. This feature makes it possible for ASM to migrate or access disks that were originally configured using different block sizes. In earlier releases, the block size selection was global and applied to all disks on a system.

  To use per-I/O block size selection, the ORACLEASM_USE_LOGICAL_BLOCK_SIZE configuration option must be set to false and the Oracle Database must be patched to implement the per-I/O block size capability. See Known Issues.

• Thin Provisioning: Unmap support is included, making it possible to reclaim unused disk space on thin-provisioned storage devices. This capability requires KABI_V3 (UEK R7 or later). Use oracleasm querydisk -i or oracleasm discover -l to query whether an ASM disk supports thin provisioning.

• When using KABI_V3 (UEK R7 or later) a universally unique identifier (UUID) is reported for those SCSI and NVMe storage devices that support it. Use oracleasm querydisk -i or oracleasm discover -l to get the UUID that uniquely identifies a storage device to ASM across all nodes in a cluster.

The following notable changes are included in oracleasm-support version 3.1:

• Disk labeling operations now inhibit writing ASM disks that are marked as being members of an ASM disk group. This is done to prevent inadvertently deleting or renaming ASM disks that are actively in use by other nodes in a cluster.

• Checks were added to the oracleasm status command to validate that the oracleasm service is fully operational. The checks include verifying that ASMLIB is installed and that the io_uring interface is accessible to the configured ORACLEASM_UID/ORACLEASM_GID.

• The -i option is added to the oracleasm querydisk command to print detailed information about the specified ASM disk.

  The -l option is added to the oracleasm discover command to print detailed information about all ASM disks attached to the system.

  The detailed information printed in each case includes device node, disk UUID, ASM disk group name, disk size, logical and physical block sizes, maximum I/O size, and whether the disk supports end-to-end data integrity and thin provisioning. Note that some information is only available when using KABI_V3 interface (UEK R7 or later).

Previous releases of Oracle ASMLIB supported several versions of interfaces implemented by the oracleasm driver over the years. The library picked the correct interface based on the oracleasm version reported by the kernel at runtime. Therefore, a single library binary could be used with various kernel releases.

Starting with ASMLIB version 3.0, another I/O submission interface is added to the library. The new I/O submission interface takes advantage of the high performance io_uring interface available in modern Linux kernels. If ASMLIB version 3.0 is loaded on a system that doesn't have an oracleasm driver loaded, is running a kernel that has io_uring enabled, and that supports a recent enough version of io_uring, the io_uring interface is used to submit I/O to the kernel instead of oracleasm.

The single ASMLIB version 3.0 binary handles all previous I/O submission interface versions in addition to io_uring and therefore no configuration changes are required when switching between kernels that don't have io_uring enabled and those that do.

The library automatically uses the appropriate interface for the running kernel.

The oracleasm-support package has been enhanced to work with version 3.0 of ASMLIB, which uses the generic io_uring interface to manage ASM disks on kernels that include this functionality. The updated oracleasm-support package continues to work with older kernels with backward compatibility. The command syntax remains the same irrespective of the kernel that's running.

I/O Filter: On systems running UEK R7 or Oracle Linux 9 with RHCK, oracleasm-support automatically adds an I/O filter to protect ASM disks against accidental overwrites. The filter rejects any write operations that aren't started by ASM and prevents writes to ASM disks by admin commands such as dd after disks have been instantiated. No new user level commands are required to manage the I/O filter map. If a disk device is found to have a valid ASM disk label, a filter map entry is automatically added.

The oracleasm configuration has a new parameter ORACLEASM_CONFIG_MAX_DISKS, which specifies the maximum number of ASM disks that can be used in the system. This parameter is used to calculate the size of I/O filter map.

I/O filtering depends on BPF (Berkeley Packet Filter) functionality within the kernel.