Emulex Driver for Linux Release Notes
Contents
1. brctl setageing lt bridge gt 0 This causes the bridge to behave like a hub and flood the packet to all the ports except the one on which the packet arrived every time This may impact performance If you have only two interfaces on this bridge one NIC interface and virbr0 nic there is no performance impact b Another option may be to use the MacVTap interface to the guest instead of the bridge interface 2 In certain SKEWs of controllers when SR IOV is enabled NIC priority group PG and priority flow control PFC are not supported Workaround None 3 In certain SKEWs of controllers when the driver is loaded with num_vfs 32 the initialization of two of the VFs fails and only 30 VF interfaces are created Workaround None 4 Incertain configurations timeout errors may occur during maximum performance MPx socket testing Workaround For Linux Xen e Only configure interfaces created with the netfront source model and ignore the second set of interfaces created with 8139cp For Linux KVM e Use the virtio device model instead of Hypervisor Default 5 When using bonding on top of VLANs on a RHEL based system networking appears to hang during system boot or when restarting the network On a RHEL based system having a bond on top of a VLAN while using ONBOOT yes can cause the system to hang during boot or when restarting the network The system will appear to hang while trying to start one of th2. When the RoCE maximum transmission unit MTU on the OCe14102 adapter is set to 4096 the MTU on the switch port in use must be set to at least 5000 otherwise cyclic redundancy check CRC errors will be reported When PFC settings are enabled on Emulex OCe14000 series adapters the 8021q ko kernel module must be loaded if not already loaded by default This module must be loaded before the Emulex drivers are loaded using the following command modprobe 8021q Emulex Driver for Linux Release Notes 16 P011386 01A Rev A gt G EMULEX An Avago Technologies Company 12 For better performance of NFS over RDMA enable multiple queue pair feature on the RoCE device using the following command echo 1 gt sys class infiniband ocrdmax qp expansion 13 When unloading the ocrdma driver an error message module ocrdma is in use may appear Linux does not allow unloading of modules when they are in use In such cases it is recommended that you terminate the application s that may be using ocrdma and try again 14 It is recommended that you use the RoCE 2 NIC profile with NFS RDMA For information on setting the profile refer to Section Basic Configuration in the most recent Emulex Drivers for Linux User Manual available on the Emulex website When you successfully update the profile the Active Profile ID must display 0x15 15 The recommended RoCE profile for Linux iSER NFS or VM Migration is now RoCE 2 16 Ref
3. from Oracle s public yum at http public yum oracle com repo OracleLinux OL5 latest Emulex Driver for Linux Release Notes 13 P011386 01A Rev A gt G EMULEX An Avago Technologies Company Note Package updates are not required in Oracle Linux 6 Unbreakable Enterprise Kernel systems You can perform this update when the UCNA is online and passing network or storage traffic To update the Ethernet firmware image a Copy the latest firmware image to the lib firmware directory using the following command cp be3flash ufi lib firmware b Start the update process using the following command ethtool f eth lt xX gt be3flash ufi 0 c Reboot the system to enable the new firmware image 4 Upgrading the be2iscsi driver from proprietary to Open iSCSI in SLES 11 SP3 a Ensure that the Open iSCSI package is installed If it is not mount the installation CD and install it linux oe3a media SLES 11 SP3 DVD x86 6407031 SuSE x86 64 rpm ihv open iscsi 2 0 873 0 21 1 x86 64 rpm b Edit etc iSCSI initiatorname iscsi to add the initiator iqn name This name can be the same name as in iSCSIselect or a new unique one Note The name must conform to the standard iqn and for iSCSI boot it has to be the same name used in iSCSIselect One method to obtain the current adapter iqn when running the proprietary be2iscsi is cat sys class scsi_ host host4 adapter igqn etc iscsi initiator name iscsi Example Initiato
4. gt G EMULEX An Avago Technologies Company Emulex Driver for Linux Release Notes Versions Driver for Linux FC FCoE Version 10 6 144 21 RHEL 6 RHEL 7 SLES OL 6 OL 7 Driver for Linux FC FCoE Version 10 4 255 16 RHEL 5 OL 5 Driver for Linux NIC Version 10 6 144 21 RHEL 6 RHEL 7 SLES OL 6 OL 7 Debian Ubuntu Driver for Linux NIC Version 10 4 255 20 RHEL 5 OL 5 Driver for Linux Open iSCSI Version 10 6 144 21 RHEL 7 RHEL 6 SLES 12 SLES 11 OL Driver for Linux Proprietary iSCSI Version 10 0 719 1009 RHEL 5 OL 5 Driver for Linux RoCE Version 10 6 144 21 See Table 1 on page 2 Date July 2015 Purpose and Contact Information These release notes describe the new features resolved known issues current known issues and technical tips associated with this Emulex drivers for Linux release For the latest product documentation go to www Emulex com If you have questions or require additional information contact an authorized Emulex technical support representative at tech support elx avagotech com 800 854 7112 US Canada toll free 1 714 885 3402 US International or 44 1189 772929 Europe Middle East and Africa New Features e Supports OCel14000B series adapters e Supports the Red Hat Enterprise Linux 7 1 RHEL 7 1 operating system e Enhanced performance of FCoE NIC iSCSI drivers on OCe14000 series adapters e Discontinued support for LPe11000 series and OCe10100 series adapters e Dis
5. SCSI be2iscsi driver installed but not used as bootable any iSCSI LUN configured in iSCSISelect will not get detected automatically When the open iSCSI be2iscsi driver is installed as bootable any non bootable iSCSI LUNs configured in iSCSISelect will not get detected automatically Configure the isCSI LUNs using iscsiadm or the OneCommand Manager application 4 When Oracle UEK 6 5 x64 and RHEL 6 x and 5 x systems are used with the OCe14102 UM adapter with iSCSI NIC enabled iSCSI BIOS can fail to boot from the iSCSI LUN and kernel panic can occur Workaround a Boot the system with the rdshell option in the kernel parameter When the system cannot find or mount the boot partition it will exit to the basic command shell b Manually run iscsistart b This will add the boot target to the system c Type exit and the system will continue to boot normally 5 The system may crash when logging into a target after injecting a PDU error Emulex Driver for Linux Release Notes 11 P011386 01A Rev A gt G EMULEX An Avago Technologies Company Workaround None 6 The storage path may go down while running converged I O on bond interface and multipath devices Workaround None RoCE 1 SLES 11 platforms may prevent loading of unsupported modules by default In such cases the installer may fail to load the modules even though the RPMs are installed Workaround Try manually loading the ocrdma module after doing o
6. al physical NIC to the XenServer http support citrix com article CTX121615 The Open iSCSI driver persistent targets are maintained in the Host Nodes database The driver recognizes only the boot target that is persistent and is not aware of any non boot persistent targets on the adapter Recommendation Do not use iSCSISelect to configure persistent non boot targets for all variants of RHEL 6 RHEL 7 and SLES 11 releases When the SLES 11u1x32 or SLES 11u2x32 operating systems are used with the OCe11100 series adapter with the open iSCSI driver the OEMSEAaddSendTargetPortal command does not work Workaround When installing a new SLES 11u1x32 or SLES 11u2x32 operating system ensure that the open iSCSI rpm package that came on the installation CD is used during the installation Other rpm packages may not work Setting PFC as the default mode is recommended when possible while using RoCE If CEE or DCB mode is enabled on the switch enable PFC for priority 5 configure ROCE traffic on priority 5 In addition VLAN interfaces must be configured and used for RoCE traffic in order for PFC to work correctly VLANs greater than 1 must be used for best interoperability If CEE or DCB is not enabled generic pause must be enabled on the switch and the NIC port to prevent loss of packets On NFS servers where fsid needs to be specified while exporting file systems ensure that the fsid is unique for each exported file system
7. cal preview No No Download 10 6 NIC and RoCE driver SLES 12 3 12 1 Technical preview No No Download 10 6 NIC and Note Currently RoCE is not supported on SLES 11 SP4 Resolved Issues FC FCoE There are no resolved issues NIC RoCE driver 1 The VF interface when in VLAN promiscuous mode no longer sees multicast and broadcast traffic intended for other VMs on other VLANs 2 The status of the Virtual Ethernet Bridge VEB is exposed correctly to the operating system 3 Fixed external loop back test ethtool t lt eth gt external_lb on OCe14000 series adapters 4 The VF interface displays link status correctly when the state is changed from disable to auto iSCSI There are no resolved issues Emulex Driver for Linux Release Notes P011386 01A Rev A gt G EMULEX An Avago Technologies Company RoCE There are no resolved issues Known Issues 1 On OCe14000 series adapters firmware versions 10 6 xxx xx and later include new features that require new flash regions to support them Firmware versions earlier than 10 0 803 37 did not have the ability to configure the flash regions to support these new features If you are updating from a firmware version earlier than 10 0 803 37 use one of these methods to update the firmware to 10 6 xxx xx e Use the ISO flash tool e Use the released 10 6 version of the ethtool utility You must perform the firmware update proced
8. cause applications such as the OneCommand Manager application or third party applications that use the Emulex libraries for example HBAAPI to misbehave or malfunction Workaround a Stop all applications that are accessing LPFC s HBAAPI interface OneCommand Manager application or third party applications before performing PCI Hot Plug of an LPFC adapter b Use the following command to stop the OneCommand Manager application usr sbin ocmanager stop ocmanager c After performing PCI Hot Plug of the adapter you can restart the applications 3 Deletion of Vports PCI Hot Unplug On occasion the kernel might report SCSI errors when deleting Vports via the sysfs interface or performing a PCI Hot Unplug of an Emulex adapter kernel Synchronizing SCSI cache for disk kernel FAILED Or SCSI error return code 0x00010000 Workaround None These messages do not indicate a functional failure and can be ignored 4 Deleting Vports while devices are in use Emulex provides management utilities that allow you to delete Vports However there is no mechanism for the LPFC driver to detect whether devices accessed through that Vport are in use This means you can delete a Vport when devices accessible through the Vport are mounted or when I O is outstanding to the device When file systems are mounted on Vports and Vports are deleted the file systems still appear to be mounted however they will be inaccessible Workaround Befo
9. continued support for RHEL 6 4 SLES 11 SP1 SLES 11 SP1e and SLES 11 SP2 e Support for receive side scaling RSS is enabled for virtual functions VFs in OCe14000 series adapters when SR IOV is enabled and requested number of VFs are less than maximum supported e Supports enabling or disabling number of VFs using sysfs e Fully supports RDMA over Converged Ethernet RoCE on SUSE Linux Enterprise Server 12 SLES 12 operating system with OFED 3 12 1 e Enhanced NFS RDMA performance with multiple queue pairs e Refer to Table 1 on page 2 for details on operating system support for RoCE Emulex Driver for Linux Release Notes 1 P011386 01A Rev A gt G EMULEX An Avago Technologies Company Note All listed new features are not supported on RHEL 5 x operating systems Table 1 Operating System Support for RoCE DFE Or NFS over RDMA SER BER Required Emulex Os Native Initiator Target Software Stack RHEL 6 5 3 12 1 Technical preview No No Download 10 6 NIC and RoCE driver RHEL 6 6 Native Technical preview No No Download 10 6 NIC stack driver Use inbox RoCE driver RHEL 6 7 Native Technical preview No No Download 10 6 NIC stack driver Use inbox RoCE driver RHEL 7 0 Native Technical preview No No Download 10 6 NIC and stack RoCE driver RHEL 7 1 Native Client Production Technical Technical Download 10 6 NIC and stack Server Technical preview preview RoCE driver preview SLES 11 SP3 3 12 1 Techni
10. d with Cisco Nexus 5000 series FCoE capable Ethernet switch firmware 4 1 3 N2 1 or earlier in NPV mode that may cause the loss or interruption of SCSI connections when used with the Emulex OneConnect UCNAs The switch incorrectly sends out a Discovery Advertisement to AIl ENode MACs from the FCF Emulex Driver for Linux Release Notes 6 P011386 01A Rev A gt G EMULEX An Avago Technologies Company MAC with which the FC uplink was down The end result is that sometimes the UCNA hangs on to an offline FCF or experiences back to back FCF failover and it may potentially lead to a Linux SCSI mid layer devloss timeout Workaround It has been verified that Cisco s IOS 4 2 1 N1 1 release has corrected the issue Emulex highly recommends that you upgrade your Cisco Nexus 5000 series FCoE switch firmware to 4 2 1 N1 1 or later to avoid this issue If you decide to use 4 1 3 N2 1 or earlier firmware with your Cisco Nexus 5000 series FCoE switch and this issue is encountered increase the FC transport dev_loss_tmo parameter to 60 seconds This can be accomplished in one of two ways e Update the FC transport dev_loss_tmo parameter For example echo 60 gt sys class fc_remote ports rport 3 0 1 dev_loss_tmo or e Update the LPFC driver s lpfc_nodev_tmo parameter For example echo 60 gt sys class scsi_host host3 lpfc_ nodev_tmo 12 Potential multipath path loss on the SLES 11 SP3 operating system During the controller failback test mult
11. e slave interfaces Emulex Driver for Linux Release Notes 8 P011386 01A Rev A gt G EMULEX An Avago Technologies Company For example if you have a configuration similar to the following DEVICE bond0 ROADCAST 10 255 255 255 ATEWAY 10 0 0 254 ETMASK 255 0 0 0 B G IPADDR 10 0 0 14 N ONBOOT yes BONDING OPTS mode 0 miimon 100 DEVICE eth2 2 BOOTPROTO none ONBOOT yes MASTER bond0 SLAVE yes VLAN yes DEVICE eth3 2 BOOT PROTO none ONBOOT yes MASTER bond0 SLAVE yes VLAN yes This system hangs because the RHEL networking scripts try to bring up the eth2 2 and eth3 2 interfaces twice It will bring them up once because the bonding interface specifies them as slave interfaces and then it tries to bring them up a second time because it sees the ONBOOT yes parameter on a VLAN device The issue occurs because after the devices are first brought up the MAC addresses of the interfaces changes Usually the first interface s MAC address is set on the rest of the interfaces When the device is brought up a second time the MAC address of the interface no longer matches the HWADDR field The network scripts will wait for an interface with a matching MAC address to appear which does not occur Workaround Set the ONBOOT parameter to no on the slave interfaces This prevents the system network scripts from trying to bring up the interface a second
12. ectory to boot copy the default initrd to a new name and create a boot entry in boot grub menu Ist for this new initrd name using the same kernel This step ensures that the system can still boot if something goes wrong Install the Emulex Open iSCSI driver rpm file If this is not an iSCSI boot system reboot and follow the Open iSCSI instructions to log into targets you do not need to continue following these steps If this is an iSCSI boot system that previously used the proprietary iSCSI driver do not reboot Instead continue with the next step Change directory to boot rebuild the initrd to include the iscsi feature The output of this command lists the capabilities included in the initrd Make sure iscsi is listed This option includes capabilities such as iscsiadm in the initrd For example linux fiai boot mkinitrd i initrd 3 0 76 0 11 default k vmlinuz 3 0 76 0 11 default f iscsi The previous step did not include the iSCSI script to log into the boot iSCSI target To include it automatically boot the system using the Emulex Open iSCSI driver Reboot the system and wait until it fails to mount root and exit into a command shell Run the following command from the command shell to log into the boot target iscsiadm m fw l Exit the shell using exit for the system to continue to boot with the iSCSI target Regenerate the boot initrd to include iscsi sh iscsi sh is needed for the initialization
13. er to ofed 3 12 1 release notes for ofed issues Copyright 2012 2015 Emulex All rights reserved worldwide This document refers to various companies and products by their trade names In most if not all cases their respective companies claim these designations as trademarks or registered trademarks This information is provided for reference only Although this information is believed to be accurate and reliable at the time of publication Emulex assumes no responsibility for errors or omissions Emulex reserves the right to make changes or corrections without notice This report is the property of Emulex and may not be duplicated without permission from the Company Emulex Driver for Linux Release Notes 17 P011386 01A Rev A
14. ipath cannot recover all paths For SLES 11 SP3 there is a new uid_attribute method to obtain the LUN serial number which is defined in the multipath conf file This method will call scsi_id to send the sg_io If any IOs are returned with a busy status multipath will not reinstate that path Workaround Edit the multipath conf file and replace uid attribute ID SERIAL with getuid callout lib dev scsi_id whitelisted device dev n 13 On SLES 12 systems a cable pull before tmo of the guest IO causes IO errors Workaround None 14 LP16000 series adapters connected to an I T target with LUNs via a fabric may not see the expected number of targets Workaround None 15 After unloading the driver an rport dump shows it is deleted but the counter is not freed Workaround None Emulex Driver for Linux Release Notes 7 P011386 01A Rev A gt G EMULEX An Avago Technologies Company NIC 1 In certain SKEWSs of controllers PING is not working when attempting to bridge the 1G or 10G ports to the virtual machines when SR IOV is enabled for 10G ports in the BIOS This issue occurs due to limitations of the virtual Ethernet bridge All transmitted broadcast packets are looped back by the controller This affects the functionality of the Linux bridge as it appears as if the same ARP broadcast packets are received on two different interfaces Workaround a Set the aging of the bridge to 0 using the following command
15. ne of the following e set allow_unsupported_modules to 1 in etc modprobe d unsupported modules e specify allow unsupported on the command line 2 RHEL 6 6 RHEL 6 7 and RHEL 7 x do not have ibdev2netdev included in the ibutils package Workaround As a root user run the following commands on the shell git clone git flatbed openfabrics org compat rdma compat rdma git tmp compat rdma cp tmp compat rdma ofed_scripts ibdev2netdev usr local bin rm rf tmpo compat rdma 3 Current implementations of NFS RDMA and KVM VM migration over RDMA do not support IPv6 Workaround None Technical Tips 1 On systems running RHEL 6 5 or later when upgrading or downgrading from the proprietary be2iscsi driver to the open be2iscsi driver you must install the Dracut network RPM package and pass the netroot command parameter into Dracut to build the initramfs image To update the proprietary iSCSI driver to the Open iSCSI driver for SLES 11 SP3 and later you must do the following a Make sure the Open iSCSI package is installed If it is not mount the installation CD and install it b Edit etc iscsi initiatorname iscsi to add the initiator IQN name This name can be the same name as in iSCS Select or a new unique one Emulex Driver for Linux Release Notes 12 P011386 01A Rev A k gt G EMULEX An Avago Technologies Company Note The name must conform to the standard IQN Optional step Change dir
16. oes not check for the existence of the VLAN field to exclude the field as a real interface Therefore you may see unpredictable behavior when including the HWADDR field in the ifcfg ethX Y file for a VLAN The setting may or may not work When this setting does not work it is because the scripts have created or renamed the main interface as ethX Y instead of just ethX Later when adding the VLAN interface it fails because ethX does not exist Workaround Remove the HWADDR line in ifcfg ethX Y files that refer to VLAN configurations You must also clean up the etc udev rules d 70 persistent net rules file to remove all the incorrect references to ethX Y For more information see https bugzilla redhat com show_bug cgi id 723936 7 Devloss timeout after swapping ports The driver may not finish discovery when two initiator ports are swapped This causes all devices accessible through one or both of these initiator ports to time out and all I O to fail Workaround Do one of the following e When swapping cables replace each cable one at a time and allow discovery to finish before replacing the next cable To determine if discovery is finished read the state sysfs parameter e When swapping cables allow devloss timeout to fire before replacing the cables this fails all outstanding I O 8 LILO Boot Loader is not supported on i386 and x86_64 architectures The LILO Boot Loader on i386 and x86_64 architectures is not su
17. ot and exits into a command shell h Run the following commands from the command shell to login to the boot target iscsid n iscsiadm m fw l i Type exit to exit the shell and the system will continue to boot with the iSCSI target j After the system has booted up with the open be2iscsi driver using the workaround the boot initrd must be regenerated to include iscsi sh Note iscsi sh is needed for the init script to login to the iSCSI boot target This workaround is required because mkinitrd will only include iscsi sh in the initrd if there is an entry such as sys firmware iscsi_boot which can only be created if Open iSCSI boot is enabled In addition f multipath must be added to the mkinitrd command if multipath is being used linux oe3a boot mkinitrd i initrd 3 0 76 0 11 default k vmlinuz 3 0 76 0 11 default f multipath Features e acpi e dm e multipathd e kms e block e usb e network e iSCSI e multipath e kpartx e resume userspace Emulex Driver for Linux Release Notes 15 P011386 01A Rev A 10 11 gt G EMULEX An Avago Technologies Company e resume kernel k Reboot the system again it should boot normally 1 If the system has other non boot iSCSI targets follow the Open iSCSI instructions to log in to those targets Additional physical NICs added to the XenServer do not show up in the XenCenter Workaround See the following link for instructions on adding an addition
18. pported for this driver If the LILO boot loader is used after the LPFC driver package is installed and upon reboot an incorrect initial ramdisk is used and the system might not boot correctly Workaround The boot loader supported with this driver is GRUB which is the default boot loader for most of the Linux distributions LILO is an older boot loader used on i386 and x86_64 architectures only GRUB works correctly with the driver package s installation script Emulex Driver for Linux Release Notes 5 P011386 01A Rev A gt G EMULEX An Avago Technologies Company 9 Suspend to disk command results in a kernel Oops If you attempt to suspend to disk using the command echo disk gt sys power state the LPFC driver encounters a kernel Oops The sysfs parameter sys power state is used to suspend and resume the system The LPFC driver does not support the suspend to disk and resume command Do not attempt to use this sysfs parameter when the LPFC driver is loaded Workaround None 10 Potential error messages during the driver kit removal process As part of the driver kit removal process initiated via the Ipfc install u command the previous in box LPFC driver version which was saved as part of the current driver kit install process is restored and becomes the active driver However the driver kit un install process does not remove any entries in the Linux distribution configuration file modprobe conf As such pa
19. rName iqn 1996 04 de suse 01 619c53c296b2 c Optional step Change the directory to boot copy the default initrd to a new name and create a boot entry in boot grub menu I st for this new initrd name using the same kernel This step ensures that the system can still be rebooted if something unexpected occurs d Install the be2iscsi Open iSCSI driver rpm linux oe3a tmp rpm Uhv be2iscsi kmp default 10 0 713 0 3 0 76 0 11 1 slesllsp3 x86 64 rpm e If this is anon iSCSI boot system reboot and follow the Open iSCSI instructions to log in to targets If this is an iSCSI boot system that previously used the proprietary be2iscsi do not reboot continue to the next step f Change directory to boot and rebuild the initrd to include the iSCSI feature The output of this command will list the features included in the initrd Note Make sure iSCSI is listed This feature will include iscsiadm in the initrd linux oe3a boot mkinitrd i initrd 3 0 76 0 11 default k vmlinuz 3 0 76 0 11 default f iscsi Emulex Driver for Linux Release Notes 14 P011386 01A Rev A gt G EMULEX An Avago Technologies Company Features e acpi e dm e multipathd e kms e block e usb e network e multipath e kpartx e resume userspace e resume kernel e kdump g To automatically include the iSCSI script to login to the boot iSCSI target boot the system using open be2iscsi Reboot the system and wait until it fails to mount ro
20. rameters that would have been valid for the just removed driver versions and entered in the modprobe conf file are also used to load the just restored in box driver version This can potentially create problems when e The just removed driver version might include module parameters that did not exist in the older just restored driver version e One or more of these module parameters are included in the configuration file modprobe conf If the above criteria are met an error message may be displayed during the uninstallation process of the driver kit such as Loading LPFC Driver FATAL Error inserting lpfc lib modules lt kernel_revision gt kernel drivers scsi lpfc lpfc ko Unknown symbol in module or unknown parameter s dmesg For example this issue can be observed when uninstalling an 8 2 0 x driver kit which had DH CHAP functionality enabled on a Linux distribution with an older 8 1 10 x in box driver version Workaround If such an error is seen during the kit removal process edit the Linux configuration file modprobe conf and remove all entries that list LPFC driver module parameters that is entries that start with options lpfc Then attempt to uninstall the driver kit again Note To find the module parameters supported by an LPFC driver module type modinfo lt driver dir gt lpfc ko 11 Potential connection loss due to an FCF Failover issue with a Cisco FCoE switch An issue was discovere
21. re deleting Vports you must prepare the system affected by the Vport deletion accordingly by unmounting all the devices accessible through the Vports and ensuring there is no outstanding I O 5 Support of 4Gbps adapters in Direct I O virtualized environments Default driver configuration fails to initialize 4Gbps adapters in virtualized environments that use Direct I O or SFPT This may result in a system hang or an uninitialized LPFC adapter in Intel VT d and AMD V IOMMU systems Workaround In these virtualized environments that use Direct I O or SFPT and 4Gbps adapters you must load the LPFC driver with the following driver parameter e lpfc_sli_mode 2 For example modprobe lpfc lpfc sli _ mode 2 Emulex Driver for Linux Release Notes 4 P011386 01A Rev A gt G EMULEX An Avago Technologies Company Note A consequence of this workaround is that virtual ports are no longer supported by the LPFC driver when the lpfc_sli_mode parameter is set to 2 6 The physical interface can improperly name eth0 123 when the etc sysconfig network scripts ifcfg eth0 123 file contains a HWADDR declaration When using VLAN on RHEL 6 x the main interface is not created and the VLAN interface does not actually work correctly as a VLAN This is an issue with the udev script in RHEL 6 x In RHEL 6 x the udev scripts interpret the HWADDR field in an ifcfg ethX Y configuration file to mean that the configuration belongs to a real interface It d
22. roughput is observed with LPe16000 series adapters when only one Tx queue is created on the interface Workaround None 11 In UMC configuration with 4 port OCe14000 series adapters minimum bandwidth is not working as expected with default number of 7 queues Workaround Reduce the number of queues to 4 via ethtool L eth lt X gt combined 4 command for minimum bandwidth configured to work as expected 12 When VxLAN offload is enabled in a OCe14000 series adapter TCP connections through other tunnel interfaces like GRE or IPIP in the adapter may not work This is because checksum offload is not supported for tunnels other than VxLAN When VxLAN offload is enabled the driver message Enabled VxLAN offloads for UDP port lt port num gt is displayed as the output of dmesg command Workaround Disable TX checksum offload in the adapter port on which other tunnel interfaces like GRE IPIP are configured ethtool K eth lt X gt tx off 13 In some of the supported operating systems when SR IOV is enabled in the adapter the value of Number of VFs in the PCI config space is same as the Total VFs although VFs have not been enabled in the driver This issue is fixed in RHEL 7 0 SLES 12 and later kernels Workaround None Emulex Driver for Linux Release Notes 10 P011386 01A Rev A gt G EMULEX An Avago Technologies Company 14 In OCe14000 series adapters throughput does not adhere to the VF Transmit Tx rate set in the h
23. script to log into the iSCSI boot target This is necessary because mkinitrd will include iscsi sh in the initrd only if there is an entry such as sys firmware iscsi_boot which can only be created under Open iSCSI boot Reboot the system again It should boot normally If the system has other non boot iSCSI targets follow the Open iSCSI instruction to log into those targets 2 Red Hat does not officially support saving kdump files to iSCSI targets In RHEL 6 5 and later to save kdump files to iSCSI targets a b c The system must reserve at least 256 MB for the crash dump Issue the kernel command with the option crashkernel 256M Reboot the system for this change to take effect Copy the initramfs image to the initrd kdump image or build a new initramfs image using Dracut and rename it to initrd kdump Restart the kdump service 3 Updating OneConnect UCNA firmware in Oracle Linux 5 Unbreakable Enterprise Kernel systems The Emulex Ethernet driver for OneConnect adapters supports updating the firmware image in the UCNA flash through the request_firmware interface in Linux For Oracle Linux 5 Unbreakable Enterprise Kernel systems an updated version of the ethtool package ethtool 6 4 0 1 el5 lt arch gt rpm is required If this package is not already installed on your system it is available from the Oracle Linux 5 LATEST channel on the Unbreakable Linux Network Advisory ELBA 2013 2544 released Aug 29 2013 and
24. time This will not negatively affect the operation of the bonded interface because the ifcfg bond0 file still has ONBOOT set to yes The networking scripts will attempt to bring up the bond0 interface on boot which will cause the slave interfaces to be brought up and configured correctly If you are only using bonding without VLAN the network scripts do not bring up slave interfaces so the issue is avoided The issue occurs only when bonding is on top of VLAN interfaces Emulex Driver for Linux Release Notes 9 P011386 01A Rev A gt G EMULEX An Avago Technologies Company 6 Firmware dump using ethtool W is not supported for OCe14000 series adapters Workaround None 7 When a CNA is configured for a NIC iSCSI profile the NIC and iSCSI traffic can be configured to share the total bandwidth In such a scenario the NIC traffic will be assured a minimum bandwidth In such a configuration a VF interface will inherit the minimum bandwidth of the PF while the OneCommand Manager and BIOS components will display the minimum and maximum bandwidths configured and the PF will display the total bandwidth available 8 Low throughput is observed for the VLAN bonded interface on SLES 12 Workaround Obtain the latest kernel patch from SUSE 9 Acall trace in dmesg is observed when running Tx traffic on LPe16000 series adapters with bonded VLAN interface with SLES 12 Workaround Obtain the latest kernel patch from SUSE 10 Low th
25. ure twice to ensure that the flash regions are properly configured and you must reboot the system after each firmware update Note Do not use Linux inbox drivers when performing the firmware update Note After you have updated the firmware you must not downgrade the firmware to a version earlier than 10 0 803 37 If you are updating from firmware version 10 0 803 37 or later use one of these methods to update the firmware to 10 6 xxx xx e Use the ISO flash tool e Use the released 10 6 version of the ethtool utility and reboot the system after the firmware update It is not necessary to perform the firmware update process twice Note Do not use Linux inbox drivers when performing the firmware update Note After you have updated the firmware you must not downgrade the firmware to a version earlier than 10 0 803 37 2 The host may fail to boot on certain vendor s BIOS versions with a single port adapter This is most commonly observed when a 1 port adapter is using the RoCE protocol but it can also occur when a 1 port adapter uses the NIC protocol with SR IOV disabled Affected vendor systems include e SuperMicro X9DR3 F i e Cisco USC 240 m3 Workaround None FC FCoE 1 Link Aggregation Control Protocol LACP cannot be used on the same port as FCoE or iSCSI Workaround None Emulex Driver for Linux Release Notes 3 P011386 01A Rev A gt G EMULEX An Avago Technologies Company 2 PCI Hot Plug may
26. ypervisor when multiple Tx queues are created on the VF Workaround Decrease the TX queue count on the VF interface to 1 using the command ethtool L eth lt X gt combined 1 where eth lt X gt is the VF interface iSCSI 1 The Open iSCSI administration utility binds the MAC address of each iSCSI port as an identity to create a configuration database If a MAC address for an iSCSI port changes the configuration data will be invalid MAC addresses can be changed via firmware updates or by changing the profile or personality of the adapter Non boot persistent sessions will no longer work Therefore all previously mounted partitions will not be found When iscsid attempts to open sessions through the iSCSI port with a changed MAC address expect to see the following message beiscsi_ep connect shost is NULL Workaround Clean up the saved configuration and then recreate a new interface discovery and login 2 There is a known Dracut issue that occurs on some systems with specific network settings where iSCSI BIOS can fail to boot from the iSCSI LUN and kernel panic can occur Workaround a Boot the system with the rdshell option in the kernel parameter When the system cannot find or mount the boot partition it will exit to the basic command shell b Manually run iscsistart b This will add the boot target to the system c Type exit and the system will continue to boot normally 3 In systems with an open i
Download Pdf Manuals
Related Search