------------------------------------------------------------------ RAIDCore BC4000 Series LINUX README Copyright (c) 2006. Broadcom Corporation. All rights reserved. ------------------------------------------------------------------ Welcome to the RAIDCore BC4000 Series Controller. This document provides important release information that should be read before using the RAIDCore product. CONTENTS -------- * ABOUT THIS RELEASE * SUPPORTED OPERATING SYSTEMS * DRIVERS * DOCUMENTATION * INSTALLATION INFORMATION * KNOWN ISSUES * CONTACT INFORMATION ABOUT THIS RELEASE ------------------ 1. Several drivers have been released to support various errata kernels. See SUPPORTED OPERATING SYSTEMS section and visit our website. 2. An unsupported Software Developers Kit (SDK) is being released with this version of the software. 3. On 2.6.x Linux Kernels greater than 2 terabyte array sizes are supported. 4. Compatibility issues with Mozilla 1.7 have been addressed with this release. 5. A fatal system error has been fixed. This happened when trying to transform without enough disk space. 6. A new controller option has been added to the BIOS to allow the user to modify the PCI-ID of the controller. This option was added for newer versions of Linux that recognize the RAID controller as a SATA controller. Of the currently released drivers this only affects Fedora Core 2 errata kernels 2.6.9-1.3_FC2 and later. However, since we are also releasing an SDK with this version please be aware that you may experience this issue with other Linux kernels. To change the PCI-ID, enter the BIOS. From the Main Menu select Controller Options. Toggle Legacy Board ID by default is set to ON. This is the old PCI-ID. To change to the new ID change the value to OFF. After toggling the ID the system will automatically reboot. NOTE: If you are dual booted with any version of Windows and the Windows boot partition is on the RAIDCore Controller, the Windows version will not recognize the new PCI-ID. If this is the case, you can set the PCI-ID back to its default value and you will be able to boot your Windows OS. You should contact support to get a workaround. 7. A change was made to the installation between v1.3 and v1.4. This does not affect the user when upgrading from v1.3 to v1.4, but when going from v1.4 to v1.3 the user will need to execute the following commands: cd /lib/modules find -name bcraid.o | xargs rm -f 8. Performance Enhancements Performance on machines with more than 1 gigabyte of memory is substantially better than previous releases. 9. A change has been made to the firmware (BIOS) regarding Acoustic Management and Power Management. It is believed that an earlier change may have been too aggressive in setting these feature levels. SUPPORTED OPERATING SYSTEMS --------------------------- Opteron, Athlon, and i686 architectures are supported in this release on stock as well as many errata kernels for: Red Hat 9, 32-bit Red Hat Enterprise Linux 3, 32-bit Red Hat Enterprise Linux 4, 32-bit and 64-bit (x86-64) Fedora Core 1, 32-bit Fedora Core 2, 32-bit Fedora Core 3, 32-bit and 64-bit (x86-64) Fedora Core 4, 32-bit and 64-bit (x86-64) SuSE Linux Enterprise Server 9, 32-bit and 64-bit (x86-64) SuSE Linux Enterprise Server 9, 64-bit (x86_64) SuSE Linux Enterprise Server 10, 64-bit (x86_64) SuSE 9.1 Professional, 32-bit SuSE 9.2 Professional, 32-bit SuSE 9.3 Professional, 32-bit and 64-bit (x86-64) SuSE 10.0, 32-bit and 64-bit (x86-64) Please check our web site periodically for other versions of Linux and errata kernel drivers. DRIVERS ------- CAUTION! You must install the same level of driver as the XelCore™ application. This means that version 1.x drivers will not work with version 2.x of the software, and vice-versa. DOCUMENTATION ------------- For information on how to install and use this product, launch the web-based help system on the CD from /[mnt point]/docs/onlinehelp.htm. You can also view or print the PDF file of the same content from docs/User_Manual.pdf. INSTALLATION INFORMATION ------------------------ Installation instructions for the application: NOTE: You need the g++ compiler and the pthread library on your system to install the bcadm RPM. From the directory where the kits were downloaded, execute the following commands to install the RPMs: rpm -ivh bcadm-{version}.i386.rpm rpm -ivh bcraid-onlinehelp-{version}.i386.rpm The docs install to: /usr/local/share/bcraid-onlinehelp The application (bcadm) installs to: /bin The man page installs to: /usr/share/man/man8 Installation instructions for the driver: The driver kit is delivered as a compressed floppy diskette image. Uncompress the image and either copy to a floppy using dd or mount the image using the loop-back driver. More specific instructions for extracting and mounting the floppy image can be found in the online help, available either on the website or from the online help RPM. Specific instructions for installing the driver can be found in the README on the driver diskette. SuSE LINUX USERS ONLY: For all SuSE Linux versions, Broadcom is now providing ISO images with all the kernel variants on them (install, uni, smp). Due to space considerations, these must be loaded from a CD-ROM. Installation instructions for updating to an errata kernel: When updating the linux kernel version it is necessary to copy the bcraid drivers into the /lib/modules tree before upgrading the kernel RPM. An upgrade script, available on the distribution disk, takes a single optional argument to specify the type of kernel. Kernel types are "smp", "bigmem", and "hugemem", if the type is not given the uniprocessor version of the driver will be copied. See section "Installation instructions for the driver" above for instructions on creating the driver floppy. Insert the driver disk, log in to the system as root and run the following commands: mount /dev/fd0 /mnt/floppy cd /mnt/floppy ./upgrade [smp | bigmem | hugemem] Install the kernel update RPM package. Special instruction for kernel upgrades on SUSE 9.1: For SUSE 9.1 kernel version 2.6.4-52 upgrades will fail. This is because the rpm post-install scriplet does not run depmod prior to making the new initrd. It is therefore necessary to manually perform these steps after upgrading the kernel. After mounting the correct bcraid driver disk for the kernel do the following: ./upgrade rpm -ivh kernel-[default, smp, bigsmp]-2.6.4-52.i586.rpm /sbin/depmod 2.6.4-52-[default, smp, bigsmp] -a /sbin/mkinitrd -k vmlinuz-2.6.4-52-[default, smp, bigsmp] -i /boot/initrd-2.6.4-52-[default, smp, bigsmp] You can now reboot your system to the new kernel KNOWN ISSUES ------------ 1. Online help has display issues with Konqueror. When viewing the online help with Knoqueror, there are several display issues that can be resolved by either using a different browser (Firefox for example), or by changing the default browser Identification. For example, Tools->Change Browser Identification->other-> and select "FireFox 1.0". 2. Some files not updated properly when using rpm -U with bcraid-*.rpm. To upgrade bcraid-*.rpm please use the following proceedure: rpm --erase bcraid rpm -i bcraid*.rpm If you upgraded using rpm -U, the GUI may stop working at some time in the future depending on the setting of prelink. In this case erase the rpm and reinstall. 3. Multiple Bootable Devices May Be Incompatible If there is more than one BIOS in a system, you may not be able to boot. On motherboards with embedded controllers, or when using the BC4000 Series RAID Controller with other I/O controllers, Broadcom requires disabling INT13 support on all controllers except the boot controller. 4. Intel Hance Rapids Chipset PCI-X slots not supported The supported versions of Linux do not support the Intel Hance Rapids PCI Bridge Chipset PCI-X slots. Putting any PCI-X card in these slots will not work with Linux. It will, however, work in the 32-bit PCI slots. The symptom is that interrupts are not delivered to Linux for these PCI-X slots and the controller will appear to be inactive. The 'lspci -l' command displays the Hance Rapids Chipset as an unknown pci bridge device. 5. Disk drives previously used with 3Ware controller may not work If you have disk drives that were previously connected to a 3Ware controller with arrays, they may be left password protected against reads and writes by non-3Ware controllers. If this occurs, plug the disk drives back into the 3Ware controller and delete all the arrays on the drives (which will also delete all the data), then set all of the drives to be accessible as JBODs. This will clear the password protection so that the drive can be used with any controller. 6. Inconsistent Array and Disk Numbers Array and disk numbers in the BIOS may not match those shown in applications running in the operating system. 7. Anaconda Installer "check for bad blocks" option not supported There have been problems reported using the Anaconda installer with the "Check for bad blocks" option set during installation. The option tries to read beyond the end of the disk and fails. We do not support using this option. More details can be found in the Red Hat Bugzilla database. 8. Installation Consideration For performance reasons, it is not recommended that the RAID controller be installed in slots which share interrupts with high interrupt rate devices. Multiple RAIDCore controllers may share interrupts without a performance impact. 9. RAIDCore arrays are presented as SCSI devices. RAIDCore arrays are presented to the OS as SCSI devices. Therefore, when other SCSI devices are added to or removed from the system, such as a USB floppy drive, other removable device, or other SCSI device, the ordering and operating system name of the devices may change; for instance, sdb may become sda. 10. Known issues for arrays greater than 2 terabytes -- Linux limitations. While this is not a RAIDCore issue, we are compelled to document this issue. Logical Block Addresses: The largest logical block address (LBA) which can be represented by a 32-bit unsigned integer is 2^32 - 1. Various utilities, file systems, and partition table formats use 32-bit LBAs. Even though the Linux 2.6 kernel addressed and resolved the 32-bit LBA limit in the block and SCSI IO subsystems, there other limits which may restrict the size of the largest usable array for your system. The following table illustrates the largest file system size given a 32-bit LBA and various file system block sizes. Note that other factors for a given file system type may reduce the maximum allowable file system size. Please consult your operating system documentation for specific file system limits. lba block size size limit bits size bits size bits size 32 4G 9 512 41 2TB 32 4G 10 1024 42 4TB 32 4G 11 2048 43 8TB 32 4G 12 4096 44 16TB Fdisk, sfdisk, and partition tables: The DOS partition table uses a 32-bit unsigned integer to represent the starting offset and the size of each partition. As a result, the DOS partition table can not represent a disk greater than 2 TB nor a partition with a starting offset or size greater than 2 TB. Some versions of fdisk will produce a partition table for devices greater than 2 TB, but any LBA which should be greater than 2 TB will be silently truncated to 32 bits and an erroneous partition will be written to the disk. Sfdisk appears to create a zero length, zero offset, partition table entry, or may create no entry at all. Filesystems: Mke2fs may create a file system smaller than the array size if the -b option for block size is not used and the array is greater than 2TB. Using 'mke2fs -b 4096 /dev/xxx' works correctly. Again, consult your operating system documentation and check the size of the file system after creating it. Alternatives to using partition tables. If you want or need to use arrays greater than 2 TB, there are at least two alternatives. The first alternative is to simply forgo using a partition table and use the base device entry for the array, i.e., instead of creating a partition table and using /dev/sdc1, simply use /dev/sdc. NOTE: you can not do this for your boot partition, as the PC BIOS needs the partition table and the Master Boot Record to boot. Another, more versatile, alternative is to use the Logical Volume Manager. It handles arrays greater than 2 TB. Use the LVM to create (pvcreate) Physical Volumes using each array. LVM can aggregate (vgcreate) these Volumes into Volume Groups. LVM can then create (lvcreate) Logical Volumes, carving out space from the Volume Group in a manner analogous to a partition. Logical Volumes can be expanded, allowing for dynamic management of storage space. Consult your operating system software for more information. CONTACT INFORMATION ------------------- sales@raidcore.com support@raidcore.com www.broadcom.com/storage Product Version: 2.1-0 Product Build: RC-210-2006044.3