The kernel packages contain the Linux kernel, the core of any Linux operating system.
These updated packages fix the following security issues:
a possible kernel memory leak was found in the Linux kernel Simple Internet Transition (SIT) INET6 implementation. This could allow a local unprivileged user to cause a denial of service. (CVE-2008-2136, Important)
a flaw was found in the Linux kernel setrlimit system call, when setting RLIMIT_CPU to a certain value. This could allow a local unprivileged user to bypass the CPU time limit. (CVE-2008-1294, Moderate)
multiple NULL pointer dereferences were found in various Linux kernel network drivers. These drivers were missing checks for terminal validity, which could allow privilege escalation. (CVE-2008-2812, Moderate)
These updated packages fix the following bugs:
The Linux kernel has been updated to implement random UDP source ports where none are specified by an application. This allows applications, such as those using the GNU libc stub resolver, to use random UDP source ports, helping to make DNS spoofing attacks harder.
when using certain hardware, a bug in UART_BUG_TXEN may have caused incorrect hardware detection, causing data flow to "/dev/ttyS1" to hang.
a 50-75% drop in NFS server rewrite performance, compared to Red Hat Enterprise Linux 4.6, has been resolved.
due a bug in the fast userspace mutex code, while one thread fetched a pointer, another thread may have removed it, causing the first thread to fetch the wrong pointer, possibly causing a system crash.
on certain Hitachi hardware, removing the "uhci_hcd" module caused a kernel oops, and the following error:
BUG: warning at arch/ia64/kernel/iosapic.c:1001/iosapic_unregister_intr()
Even after the "uhci_hcd" module was reloaded, there was no access to USB devices. As well, on systems that have legacy interrupts, "acpi_unregister_gsi" incorrectly called "iosapci_unregister_intr()", causing warning messages to be logged.
when a page was mapped with mmap(), and "PROT_WRITE" was the only "prot" argument, the first read of that page caused a segmentation fault. If the page was read after it was written to, no fault occurred. This was incompatible with the Red Hat Enterprise Linux 4 behavior.
due to a NULL pointer dereference in powernowk8_init(), a panic may have occurred.
certain error conditions handled by the bonding sysfs interface could have left rtnl_lock() unbalanced, either by locking and returning without unlocking, or by unlocking when it did not lock, possibly causing a "kernel: RTNL: assertion failed at net/core/fib_rules.c" error.
the kernel currently expects a maximum of six Machine Check Exception (MCE) banks to be exposed by a CPU. Certain CPUs have 7 or more, which may have caused the MCE to be incorrectly reported.
a race condition in UNIX domain sockets may have caused recv() to return zero. For clusters, this may have caused unexpected failovers.
msgrcv() frequently returned an incorrect "ERESTARTNOHAND (514)" error number.
on certain Intel Itanium-based systems, when kdump was configured to halt the system after a dump operation, after the "System halted." output, the kernel continued to output endless "soft lockup" messages.
Red Hat Enterprise Linux 5 users are advised to upgrade to these updated packages, which contain backported patches to resolve these issues.