Chapter 3. Kernel
Prior to Red Hat Enterprise Linux 6.2, the qla4xxx
adapter firmware managed discovery and login to iSCSI targets. A new feature in Red Hat Enterprise Linux 6.2 allows you to use open-iscsi to manage the qla4xxx
discovery and login process. This can result in a more uniform management process.
qla4xxx
iSCSI firmware settings are accessible via:
~]# iscsiadm -m fw
ql4xdisablesysfsboot=1
parameter as follows:
- Set the parameter in the
/etc/modprobe.d
file:~]#
echo "options qla4xxx ql4xdisablesysfsboot=1" >> /etc/modprobe.d/qla4xxx.conf
- Reload the
qla4xxx
module either by executing the following set of commands:~]#
rmmod qla4xxx
~]#modprobe qla4xxx
or, if you are booted off theqla4xxx
device, by rebooting your system.
qla4xxx
device, upgrading from Red Hat Enterprise Linux 6.1 to Red Hat Enterprise Linux 6.2 will cause the system to fail to boot up with the new kernel. For more information on this known issue, refer to the Technical Notes.
Kdump (a kexec-based crash dumping mechanism) now supports dumping of the core on the following file systems on Red Hat Enterprise Linux 6:
- Btrfs (Note that this file system is a Technology Preview)
- ext4
- XFS (Note that XFS is a layer product and must be installed to enable this feature)
The pkgtemp
module has been merged with the coretemp
module. The pkgtemp
module is now deprecated. The coretemp
module now supports all the features it previously did plus the features that were supported by the pkgtemp
module.
coretemp
previously only provided per core temperatures, while the pkgtemp
module provided the temperatures of the CPU package. In Red Hat Enterprise Linux 6.2, the coretemp
module allows you to read the temperatures of the cores, the uncore, and the package.
In Red Hat Enterprise Linux 6.2, the SCSI midlayer supports optional lockless dispatching of SCSI driver queuecommand
functions.
scsi_host_template
structure is used by SCSI drivers to indicate to the SCSI midlayer that driver queuecommand
will be dispatched without the SCSI host bus lock held.
Scsi_Host
lock will be held during a driver queuecommand
dispatch. Setting the scsi_host_template
lockless bit prior to scsi_host_alloc
will cause the driver queuecommand
function to be dispatched without the Scsi_Host
lock being held. In such a case, the responsibility for any lock protection required is pushed down into the driver queuecommand
code path.
queuecommand
in Red Hat Enterprise Linux 6.2 are listed below:
- iscsi_iser
- be2iscsi
- bnx2fc
- bnx2i
- cxgb3i
- cxgb4i
- fcoe (software fcoe)
- qla2xxx
- qla4xxx
Red Hat Enterprise Linux 6.2 includes support for Fiber Channel over Ethernet (FCoE) target mode, as a Technology Preview. This kernel feature is configurable via targetadmin, supplied by the fcoe-target-utils package. FCoE is designed to be used on a network supporting Data Center Bridging (DCB). Further details are available in the dcbtool(8)
and targetadmin(8)
man pages.
Important
In Red Hat Enterprise Linux 6.1, with BZ#605786, the crashkernel=auto
boot parameter was deprecated. However, in Red Hat Enterprise Linux 6.2, support for crashkernel=auto
is continued on all Red Hat Enterprise Linux 6 systems.
The mdadm
and mdmon
utilities have been updated to support Array Auto-Rebuild, RAID Level Migrations, RAID 5 support limitation, and SAS-SATA drive roaming.
Red Hat Enterprise Linux 6.2 supports merging of flush requests to assist devices which are slow to perform a flush.
Red Hat Enterprise Linux 6.2 adds UV2 Hub support. UV2 is the UVhub chip that is the successor to the current UV1 hub chip. UV2 uses the HARP hub chip that is currently in development. UV2 provides support for new Intel sockets. It provides new features to improve performance. UV2 is being designed to support 64 TB of memory in a Single System Image (SSI). Additionally, the node controller MMRs have been updated for UV systems.
Red Hat Enterprise Linux 6.2 introduces the acpi_rsdp
boot parameter for kdump to pass an ACPI RSDP address, so that the kdump kernel can boot without Extensible Firmware Interface (EFI).
The following enhancements have been added to the QETH network device driver:
- Support for
af_iucv
HiperSockets transport - Support for forced signal adapter indications
- Support for asynchronous delivery of storage blocks
- New Ethernet Protocol ID added to the
if_ether
module
Support for the new CPACF (CP Assist for Cryptographic Function) algorithms, supported by IBM zEnterprise 196, has been added. The new hardware accelerated algorithms are:
- CTR mode for AES
- CTR mode for DES and 3DES
- XTS mode for AES with key lengths of 128 and 256 bits
- GHASH message digest for GCM mode
Red Hat Enterprise Linux 6.2 supports conditional resource-reallocation through the pci=realloc
kernel parameter. This feature provides an interim solution for adding a dynamically reallocatable PCI resource without causing any regressions. It disables dynamic reallocation by default, but adds the ability to enable it through the pci=realloc
kernel command line parameter.
Dynamic reallocation is disabled by default. It can be enabled with the pci=realloc
kernel command line parameter. In addition, bridge resources have been updated to provide larger ranges in the PCI assign unassigned
call.
Red Hat Enterprise Linux 6.2 enables SMEP (Supervision Mode Execution Protection) in the kernel. SMEP provides an enforcement mechanism, allowing the system to set a requirement that is not intended to be executed from user pages while in the supervisor mode. This requirement is then enforced by the CPU. This feature is able to prevent all attacks, irrespective of the vulnerability in the system code, that are executed from user mode pages while the CPU is in the supervisor mode.
Support for enhanced fast string REP
MOVSB
/STORESB
instructions for the latest Intel platform has been added.
The USB 3.0 xHCI host side driver has been updated to add split-hub support, allowing the xHCI host controller to act as an external USB 3.0 hub by registering a USB 3.0 roothub and a USB 2.0 roothub.
The ACPI, APEI, and EINJ parameter support is now disabled by default.
Red Hat Enterprise Linux 6.2 adds support for pstore—a file system interface for platform dependent persistent storage.
Support for printk based APEI (ACPI Platform Error Interface) hardware error reporting has been added, providing a way to unify errors from various sources and send them to the system console.
The ioatdma
driver (dma
engine driver) has been updated to support Intel processors with a dma
engine.
Support for the Digi/IBM PCIe 2-port Async EIA-232 Adapter has been added to the 8250 PCI serial driver. Additionally, EEH (Enhanced Error Handling) support for the Digi/IBM PCIe 2-port Async EIA-232 Adapter has been added to the 8250 PCI serial driver.
ARI (Alternative Routing- ID Interpretation) support, a PCIe v2 feature, has been to Red Hat Enterprise Linux 6.2.
PCIe OBFF (Optimized Buffer Flush/Fill) enable/disable support has been added for Intel's latest platform. OBFF provides devices with information on interrupts and memory activity and their potentially reduced power impact, ultimately improving energy efficiency.
In Red Hat Enterprise Linux 6.2, the kernel is enabled to capture kernel oops/panic reports from the dmesg
buffer into NVRAM on PowerPC architectures.
The MXM driver, responsible for handling graphics switching on NVIDIA platforms, has been backported to Red Hat Enterprise Linux 6.2.
Red Hat Enterprise Linux 6.2 introduces page coalescing, a feature on IBM Power servers which allows for coalescing identical pages between logical partitions.
Support for L3 Cache Partitioning has been added to the latest AMD family CPUs.
The thinkpad_acpi
module has been updated to add support for new ThinkPad models.
Latest Intel processor C-State support has been added to intel_idle.
Red Hat Enterprise Linux 6.2 now displays warnings for IOMMU (Input/Output Memory Management Unit) on AMD systems.
Logging of board, system, and BIOS information to dmesg
during boot has been added.
cputable
entries have been added to the kernel, providing support for the latest IBM PowerPC processor family.
The VPHN (Virtual Processor Home Node) feature has been disabled on IBM System p.
The following drivers now support the latest Intel chipset:
i2c-i801
SMBus driverahci
AHCI-mode SATAata_piix
IDE-mode SATA driver- TCO Watchdog driver
- LPC Controller driver
On IBM PowerPC systems, the exec-shield
value in sysctl or in the /proc/sys/kernel/exec-shield
parameter is no longer enforced.
Additional checks and fixes have been added to support kdump
on 64-bit PowerPC and 64-bit IBM POWER Series systems.
The UV MMTIMER module (uv_mmtimer
) has been enabled on SGI platforms. The uv_mmtimer
module allows direct userland access to the UV system's real time clock which is synchronized across all hubs.
Support for the IB700
module has been added in Red Hat Enterprise Linux 6.2
The aer_mask_override
module parameter has been added, providing a way to override the corrected or uncorrected masks for a PCI device. The mask will have the bit corresponding to the status passed into the aer_inject()
function.
USB 3.0 host controller support has been added to 64-bit PowerPC and 64-bit IBM POWER Series systems.
An improved upstream Out-of-Memory (OOM) killer implementation has been backported to Red Hat Enterprise Linux 6.2. The improvements include:
- Processes which are about to exit are preferred by the OOM killer.
- The OOM kill process also kills the children of the selected processes.
- A heuristic has been added to kill the
forkbomb
processes.
oom_score_adj
/proc
tunable parameter adds the value stored in each process's oom_score_adj
variable, which can be adjusted via /proc
. This allows for an adjustment of each process's attractiveness to the OOM killer in user space; setting it to -1000
will disable OOM kills entirely, while setting it to +1000
marks this process as OOM's primary kill target.
Red Hat Enterprise Linux 6.2 provides an updated zram
driver (creates generic RAM based compressed block devices).
In Red Hat Enterprise Linux 6.2, the taskstat utility in the kernel, which prints the status of ASET tasks, has been enhanced by providing microsecond CPU time resolution for the top utility to use.
Red Hat Enterprise Linux 6.2 updates the perf utility to upstream version 3.1 along with the kernel upgrade to v 3.1. Refer to BZ#725524 for newly supported kernel features provided by the perf utility. The updated version of the perf utility includes:
- Added cgroup support
- Added handling of
/proc/sys/kernel/kptr_restrict
- Added more cache-miss percentage printouts
- Added the
-d -d
and-d -d -d
options to show more CPU events - Added the
--sync/-S
option - Added support for the
PERF_TYPE_RAW
parameter - Added more documentation about the
-f/--fields
option - The python-perf package has been added for python binding support.
Red Hat Enterprise Linux 6.2 adds OProfile support for the latest Intel processors.
The number of interrupt requests (IRQ) is now counted in a sum of all irq counter, reducing the cost of the look-up in the /proc/stat
file.
Red Hat Enterprise Linux 6.2 introduces a scheduling improvement where a hint is provided to the scheduler on the next buddy hint on sleep and preempt path. This hint enhancement helps the workload of multiple tasks in multiple task groups.
In Red Hat Enterprise Linux 6.2, Transparent Huge Pages are now supported in several places by the kernel:
- The system calls of mremap, mincore, and mprotect
/proc
tunable parameters:/proc/<pid>/smaps
and/proc/vmstat
Red Hat Enterprise Linux 6.2 adds XTS (XEX-based Tweaked CodeBook) AES256 self-tests to meet the FIPS-140 requirements.
Previously, the SELinux netfilter hooks returned NF_DROP
if they dropped a packet. In Red Hat Enterprise Linux 6.2, a drop in the netfilter hooks is signaled as a permanent fatal error and is not transient. By doing this, the error is passed back up the stack, and in some situations applications will get a faster indication that something went wrong.
In Red Hat Enterprise Linux 6.2, the remount mount options (mount -o remount
) are passed to a new LSM hook.
Red Hat Enterprise Linux 6.0 and 6.1 defaulted to running UEFI systems in a physical addressing mode. Red Hat Enterprise Linux 6.2 defaults to running UEFI systems in a virtual addressing mode. The previous behavior may be obtained by passing the physefi
kernel parameter.
In Red Hat Enterprise Linux 6, the default core_collector
method for kdumping the core over SSH has been changed from scp
to makedumpfile
, which helps shrink the size of the core file when copying over the network link, resulting in faster copying.
/etc/kdump.conf
file:
core_collector /usr/bin/scp