• Red Hat Enterprise Linux 5, 6
• CentOS 5, 6
• SUSE Linux Enterprise Server 11, 11sp1, 11sp2
• Fedora 17, 18
• openSUSE Linux 11, 12

• i386 (with or without "CONFIG_REGPARM")
• x86_64 (AMD64, Intel EM64T)

NOTE: The kernel header files, make, and gcc must be installed. For details on how to do this, click here.

1. Unzip and untar the driver files with tar xfz lnx_drv_esas2hba_XXX.tgz
2. Enter the lnx_drv_esas2hba_XXX directory and run the install script install.sh.
3. After the driver is installed, it can be loaded automatically if desired.

NOTE: In Linux kernel 2.6.33 and higher, there is built-in support for the PMC 8001 chip in a driver named pm8001.ko. If present, this driver will load automatically for the ExpressSAS 6Gb HBA and prevent the esas2hba driver from loading. The driver installation script will detect the presence of the pm8001 and provide the option to unload that driver and rename it to prevent this conflict.

Alternatively, the pm8001 driver can be unloaded using rmmod pm8001. To permanently remove the pm8001 driver, you can delete the pm8001 kernel module. To find the built in driver, use the command modinfo pm8001 and look for the "filename" field, then remove that file from the filesystem.

Also, see Recommendations for Fedora to reconfigure the initrd image.

Type the following to manually load the module:

$ modprobe esas2hba

You may receive a warning that the module will taint the kernel.

On SUSE 11, you may also receive an error message about "unsupported" drivers:

$ modprobe esas2hba
FATAL: module '/lib/modules/2.6.27.19-5-default/kernel/drivers/scsi/esas2hba.ko' is unsupported
Use --allow-unsupported or set allow_unsupported_modules to 1 in
/etc/modprobe.d/unsupported-modules


You can fix this problem by loading the driver with modprobe --allow-unsupported esas2hba or editing the file /etc/modprobe.d/unsupported-modules as described.

Unzip and untar the driver files:

$ tar xfz lnx_drv_esas2hba_XXX.tgz

NOTE: Make sure there are no spaces in the path in which you extract the archive. The Linux kernel Makefile may fail if the path name contains a space character.

Make and install the modules (must be done as root):

$ cd lnx_drv_esas2hba_XXX.tgz/src
$ make install


NOTE: You need the kernel header files installed to build this driver. If the header files are in a non-standard location, you may need to modify the KDIR variable on the make command line, eg. make install KDIR=/path/to/kernel

The modules will now be installed and ready to use.

On some Linux distributions, the driver may not load automatically when the system is booted. To enable this behavior, try the following suggestions:

Recommended for Red Hat 4 & 5:

Add the following line to /etc/modprobe.conf after installing the driver:

alias scsi_hostadapterX esas2hba

Where X is the next available number.

Recommended for SUSE 10:

Add the following line to /etc/init.d/boot.local:

modprobe esas2hba

Recommended for Fedora:

Fedora distributions may include the built-in pm8001 driver as part of the default initrd image. After running the driver install script, or manually removing the pm8001 driver, you may also need to update the initrd image using the following command. Consult the Fedora documentation for complete details of this command.

mkinitrd -f <initrd-image> <kernel-version>

(ex. mkinitrd -f /boot/initramfs-3.6.10-4.fc18.i686.PAE.img 3.6.10-4.fc18.i686.PAE)

The initrd-image is typically found in the /boot directory and is named initramfs-<kernel-version>.img. The kernel-version can be found by running uname -r.

WARNING: An incorrect or invalid initrd image can render a system unbootable. Care should be taken when modifying this file.

The following module parameter is supported by this driver:

• event_log_mask (default 0) Logs error and informative messages to the kernel ring buffer. This value is a bit-mask of message catergories; set to 0xFFFFFFFF to log all possible messages. This feature is mostly for debugging purposes and is not recommended for normal use.

There are also several other parameters available for tweaking. For documentation on these settings, read the file oswrap.c (look for "Module parameter definitions") or the output of the command modinfo esas2hba.ko.

On certain 64-bit platforms, the driver Makefile may be unable to detect the correct CPU architecture. In this instance, you will see an error similar to the following when attempting to compile the driver:

cc1 : error : CPU you selected does not support x86_64 instruction set

This can be resolved by specifying the correct architecture when running the make command, such as:

$ make install ARCH=x86_64

This driver requires that the kernel header files, make, and gcc be installed on the system.

For SUSE, use the YAST utility's "Software Management" module to install the "kernel-source", "gcc", and "make" packages.

For Red Hat, use the "Add/Remove Applications" utility to install the "Development Tools" packages. For details on installing the kernel source package:

• For Red Hat 4, click here
• For Red Hat 5, click here

Refer to your system documentation for further details.

Event Type	Value (hexadecimal)	Description
Fatal	0x00000001	Fatal errors, always reported
SCSI	0x00000004	SCSI command errors
Protocol	0x00000008	SAS protocol errors
Login	0x00000010	Device login/logout events
Resource	0x00000040	Resource usage failures
Info	0x00000080	Informational messages

 

In general, only errors that are output for event type 'Fatal' are significant; all others are mostly for informational or diagnostic purposes. When turning on event logging, be aware that certain events reported as errors are expected. For example, a SCSI Check Condition error will be reported for the first command sent to a device after power on or a reset condition. Also, certain data underrun errors are expected by the software and are normal.

All events report the adapter number on which the event occurred. The adapter number is simply the order in which Linux initialized the adapter channels and has no relevance to what slot the adapter is in or which channel of the adapter is being reported.

The following events are classified as 'Fatal' and are always reported.

Event Message	Description
More than 3 attempts to recover from PCI errors in the last minute. Channel is now inoperative.	After 3 attempts to recover from PCI errors within a one- minute period, the driver gives up and renders the adapter channel inoperative to avoid affecting the operation of the rest of the system. This error typically indicates that either the adapter or the slot into which it is plugged has a serious problem.
Heartbeat failed.	The adapter failed to respond with a heartbeat within the expected time. The driver will attempt to reset the adapter. This could indicate a hardware failure.
Configuration parameters not found or corrupt. Please reflash card.	This event indicates that the driver could not read the configuration parameters for the indicated adapter. Default configuration values will be used. The adapter needs to have its flash memory rewritten.
Firmware completed a reset sequence.	The driver recovered the adapter from a hardware or firmware fault.
Unknown Event Code. Code is {Code}. Least Significant Word is {Info}.	This event indicates that an unknown event code was generated by the driver code that the decoder does not know about. This event should never occur in a released driver.
Event Buffer Overflow. Number of Events Lost is {Count}.	This event indicates that events occurred faster than the OS could get them dumped to the event log on disk. This doesn't happen very often.

The following events are classified as 'SCSI'.

Event Message	Description
SCSI Error. Error Code {Code}, Sense Key {Key}, Add'l Sense Code {ASC}, Command Code {Cmd}, Path ID {Path}, Target ID {TargetID}, LUN {Lun}.	An error has been reported by the target device. The Error Code is usually 0x02 (Check Condition), in which case the Sense Key and Additional Sense Code displayed are valid. It is normal for the first command sent to a SCSI device after power on or reset to report a SCSI error, in which case the Sense Key would be 0x06 (Unit Attention) and the Additional Sense Code would be 0x29 (Power-On/Reset).
SCSI Bus Reset Requested.	The OS is requesting that a bus reset be performed. This may happen in cases where a request times out and the Abort that the driver attempts is not successful. Any requests outstanding at the time will be completed with a status of Aborted due to Bus Reset.

The following events are classified as 'Protocol'.

Event Message	Description
IOC Error (Initiator). IOC Status is {n} {Desc}. Command Code {Cmd}, Path ID {Path}, Target ID {Targ}, LUN {Lun} LBA {Logical Block Address}.	A transport or other error has been detected on an Initiator mode request. Note that the Logical Block Address is only valid if the command is a read (0x28) or write (0x2A).

The following events are classified as 'Login'.

Event Message	Description
Device {SAS Addr} attached to port {#} at {Speed}.	The indicated SAS device has been attached.
Device {SAS Addr} removed from port {#}.	The indicated SAS device has been removed.

The following events are classified as 'Resource'.

Event Message

Description

Resource Failure. {ResourceType}. Unavailable Count is {count}.

The driver encountered a situation where not enough of the indicated resource is available. The ResourceType will be one of the following. (The module parameter controlling the number of each resource is indicated in parentheses.) I/O Requests (num_ioreq) S/G Lists (num_sg_lists, sgl_page_size) When running out of S/G Lists, nothing bad will happen; the driver will retry the operation when the resources become available.

The following events are classified as 'Info'.

Event Message	Description
Bus Reset Request Completed. Status: {n} ({Desc}).	This event indicates that a requested Bus Reset has completed. If there was no error during the reset procedure, this event will only be reported if the EVENT_TYPE_INFO bit is set in EventLogMask, otherwise it will be reported regardless of the EventLogMask.

ATTO Technology, Inc.
155 CrossPoint Parkway
Amherst, New York 14068
Phone: (716) 691-1999
Fax: (716) 691-9353
www.attotech.com

ATTO Technology can also be reached via e-mail at the following addresses:

Sales Support: sls@attotech.com
Technical Support: techsupp@attotech.com