PCI hot plug management

Edit online

You can insert a new PCI hot plug adapter into an available PCI slot while the operating system is running.

This can be another adapter of the same type that is currently installed or of a different type of PCI adapter. New resources are made available to the operating system and applications without having to restart the operating system. Some reasons for adding a hot plug adapter might include:
  • Adding additional function or capacity to your existing hardware and firmware.
  • Migrating PCI adapters from a system that no longer requires the function provided by those adapters.
  • Installing a new system for which adapter cards become available after the initial configuration of optional hardware subsystems, including PCI adapters, and the installation and start of the operating system.
Note: If you add an adapter using a PCI hot plug replace or add operation, or using Dynamic Logical Partitioning, it and its child devices may not be available for specification as a boot device using the bootlist command. You may have to restart the machine to make all potential boot devices known to the operating system. An adapter already listed in the boot list, which is then replaced by the exact type of adapter, is still a valid boot device.
You can also remove a defective or failing PCI hot plug adapter or exchange it with another of the same type without shutting down or powering off the system. When you exchange the adapter, the existing device driver supports the adapter because it is of the same type. Device configuration and configuration information about devices below the adapter are retained for the replacement device. Some reasons for replacing an adapter might include:
  • Temporarily replacing the card to aid in determining a problem or to isolate a failing FRU.
  • Replacing a flawed, failing, or intermittently failing adapter with a functional card.
  • Replacing a failing redundant adapter in an HACMP or multipath I/O configuration.
When you remove a hot plug adapter, the resources provided by the adapter become unavailable to the operating system and applications. Some reasons for removing an adapter might include:
  • Removing existing I/O subsystems.
  • Removing an adapter that is no longer required or is failing and a replacement card is not available.
  • Migrating an adapter to another system when the function is no longer required on the system from which it is being removed.

Before you can remove or replace a hot plug device, it must be unconfigured. The associated device driver must free any system resources that it has allocated for the device. This includes unpinning and freeing memory, undefining interrupt and EPOW handlers, releasing DMA and timer resources, and any other required steps. The driver must also ensure that interrupts, bus memory, and bus I/O are disabled on the device.

The system administrator must perform the following tasks before and after removing an adapter:
  • Terminate and restore applications, daemons, or processes using the device.
  • Unmount and remount file systems.
  • Remove and recreate device definitions and perform other operations necessary to free up a device in use.
  • Put the system into a safe state to be serviced.
  • Obtain and install any required device drivers.

The remove and replace operations fail unless the device connected to the identified slot has been unconfigured and is in the defined state. You can do this with the rmdev command. Before placing the adapter in the defined state, close all applications that are using the adapter, otherwise, the command will be unsuccessful. For more information about the rmdev command, see rmdev.

In some cases, the you can also perform the following tasks:
  • Prepare a PCI hot plug adapter to be inserted, removed, or replaced.
  • Identify slots or PCI adapters that are involved in the hot plug operation.
  • Remove or insert PCI hot plug adapters.
Note: During the hot plug operations, the Object Data Manager (ODM) remains locked. Hence, the other tasks that require the ODM might hang or fail. The cluster-wide configuration changes that are initiated by other nodes might also hang or fail in a cluster. Therefore, ensure that such tasks are not performed until the hot plug operation is complete.
Attention: Although PCI hot plug management provides the capability of adding, removing, and replacing PCI adapters without powering off the system or restarting the operating system, not all devices in hot plug slots can be managed in this fashion. For example, the hard disk that makes up the rootvg volume group or the I/O controller to which it is attached cannot be removed or replaced without powering off the system because it is necessary for running the operating system. If the rootvg volume group is mirrored, you can use the chpv command to take the disks offline. If the rootvg volume group resides on one or more disks that are Multi-Path I/O (MPIO) enabled and connected to multiple I/O controllers, one of these I/O controllers can be removed (or replaced) without rebooting the system. In this situation, all paths from the I/O controller to be removed (or replaced) should be unconfigured using the rmdev -R command on the adapter. This will unconfigure the paths and the adapter. You can then proceed with Hot Plug Management. Before you attempt to remove or insert PCI hot plug adapters, refer to the PCI Adapter Placement Reference, (shipped with system units that support hot plug), to determine whether your adapter can be hot-swapped. Refer to your system unit documentation for instructions for installing or removing adapters.