Rapid7 Vulnerability & Exploit Database

SUSE-SA:2005:005: kernel bugfixes and SP1 merge

Free InsightVM Trial No credit card necessary
Watch Demo See how it all works
Back to Search

SUSE-SA:2005:005: kernel bugfixes and SP1 merge



The linux kernel is the core of the SUSE Linux based products.

Two weeks ago we released the Service Pack 1 for our SUSE Linux Enterprise Server 9 product. Due to the strict code freeze we were not able to merge all the security fixes from the last kernel update on Jan23rd (SUSE-SA:2005:003) into this kernel.

This update merges those missed security fixes and also included critical bug fixes for the SP1 kernel.

Other SUSE Linux versions are not included in this update.

For our SUSE Linux 9.1 Box customers this update includes an upgrade of the kernel to the kernel level we use with SUSE Linux Enterprise Server 9 + Service Pack 1.

Changes for SUSE Linux Enterprise Server 9 customers:

The following security bugs were fixed for the Service Pack 1 kernel:

  • A NFS Direct I/O local denial of service could allow a local attacker to crash the machine.
  • A previous smbfs security fix was faulty, writes did no longer work on smbfs shares.
  • Unsigned vs signed problems in the generic SCSI ioctl handler were reported by grsecurity. They have no impact due to the compiler using unsigned arithmetic, but are fixed nevertheless.
  • ppos /proc file race conditions in the mapped_base and oom_adjust proc files were fixed.

Also following critical bugs were fixed:

  • A bug in the pciconfig sysfs interface could cause incorrect values to be read from and written to PCI config space.
  • A locking problem in CKRM could lead to a crash.
  • In low memory situations, large writes would not be serviced in a timely fashion.
  • The initialization of the bio->bi_bdev field was incorrect.
  • VFS callouts for flock were added.
  • MD on top of Device Mapper based devices was not working and lead to kernel crashes.
  • Non-fatal errors encountered during read ahead operations caused the device mapper multipath to fail the hardware path.
  • A race condition in the kernel timer code could lead to kernel crashes under high load.
  • An inetaddr notification problem with SCTP could lead to machine crashes.

For SUSE Linux 9.1 customers additionally the Service Pack 1 kernel changes apply (long list):

Platform / Hardware / Driver support

  • Introduce concept of "cloned" drivers to minimize impact on HW certifications In cases where updating driver would have impacted too many hardware certifications we introduced a cloned driver which supports only the new PCI IDs by default.
  • Support many new hardware components via driver and PCI ID updates:

    + cloned tg3-new with version 3.10 to support Broadcom 5721 and 5751 + cloned bcm-new with version 7.3.5 to support Broadcom 5721 + updated e1000-new with version 5.3.19 to support one new PCI ID + updated ixgb to version 1.0.82 to support 10 GB Ethernet + updated ipr to version to support new RAID types + updated IBM ServeRAID driver ips to 7.10.18 to fix bugs + updated s2io driver to version + updated qlogic to version 8.00.00 to use official release which is better and binary compatible to version 8.00.00b14 which was used in SLES 9 GA. + updated megaide to version 5.07r to support LSI controllers + updated megaraid_mbox to version for new hardware support + updated cciss driver to version 2.6.4 to support SAS + updated gdth to version 3.04 for new hardware support + updated Emulex lpfc driver to version 2.10g for bug fixes + added driver jsm version 1.1 to support Digi Neo PCI serial cards + updated avm_fcdsl driver to support Fritz!Card DSL USB analog driver and Fritz!Card DSL USB 2.0 driver + updated avmfritzcapi to support new Eumex devices + included support for Intel i915 chipset (hwinfo, sax2, xf86) + enhanced driver update dialog to also support USB disks/sticks + fixed 4-port SATA support in the ICH6 driver + updated MPT fusion driver to version + fixed aic7?xx driver probe info + added Altix system controller communication driver + cloned aic79xx-new with version 2.0.12 to support AIC7901 and 39320 + added Qlogic iSCSI support (qla4xxx) + added patches to Infiniband Gen1 code + back ported dpt_i2o from 2.6.8 + updated aacraid driver to version 1.1.2-lk2 from 2.6.9 + added TIO support for SGI Altix

  • allow modules to use virtual IRQs
  • e1000 EEH error must not remove device
  • fix bad cciss unknown ioctl return
  • update e1000 drivers according to new information from Intel
  • fix veth dying on ppc64
  • fix oops in e1000 driver on x86_64
  • fix oops with Nvidia Nforce4
  • enable APIC on ES7000 architecture and increase MAX_MP_BUSSES
  • fix oops with gdth controller on x86_64
  • enable new DASD CCW IDs (S/390)
  • fix some hipersockets bugs and a dasd dbf oops on s390

Filesystems and I/O subsystem

  • Fixed files > 2 GB in isofs
  • Merged new Lustre hooks
  • Updated XFS filesystem and tools to latest CVS snapshot
  • Updated CIFS to 1.22
  • Improved iSCSI and SAN/NAS support with patches from EMC, NetAPP and others
  • Back ported several NFS bug fixes from upstream/mainline kernel
  • Integrated patches to allow enabling ext3 reservation code
  • fix file locking for 32 bit apps running on 64 bit systems
  • improve NFS performance by avoiding unnecessary silly renames
  • allow swapfiles > 2G on x86
  • fix reiserfs oops on small file systems (< 128 MB)
  • fix panic and deadlock in XFS direct IO
  • allow reading from zeropage with O_DIRECT/rawio
  • add reiserfs performance improvements
  • fix ACL umask handling over nfs
  • add several autofs4 fixes
  • infiniband can also be used on ppc64
  • fix oops in aio_free_ring
  • fix spinlock problem in infiniband drivers
  • fix ext2/ext3 memory leak
  • Fix ext3 directIO when extending the journal
  • kernel statd should accept NOTIFY calls from high ports


  • Added multipath fixes for barrier handling
  • With SP1 we now disabled by default the multipathing fail over support in the QLogic driver as it caused many problems. We print a warning that it is depreciated and how one can still turn it on if needed using ql2xfailover=1
  • Provide code to enable recovery from PCI EEH errors
  • Added CPU hotplug support for S/390

Power management

  • Added powernow K8 cpufreq support for CG stepping K8
  • Added cpufreq support for SMP systems
  • fix centrino speedstep on x86-64


  • Integrated bugfix to SHPC PCI hotplug driver
  • Updated CKRM to E16 and added CPU controller
  • Updated kdb to version 4.4
  • Updated Linux kernel crash dump (lkcd) and lkcdutils
  • Added SGI Altix hardware performance monitoring API
  • Exported some symbols needed by ES7000 Service Processor
  • Added tg3 ethtool stats
  • Added modular kdb support for x86_64
  • Added PAGG support on IPF
  • fix problem with monitored processes going to sleep on ia64
  • correctly display per process CPU utilization
  • fix double echo on x86-64 KDB
  • check PROM version on ia64 Altix machines and print early warning
  • update CKRM to newer revision
  • fix breakpoints on x86-64 KDB
  • allow for producing reliable backtraces with lkcd

Scalability / Performance

  • Improved RCU scalability
  • Fixed scalability problem in dnotify_parent
  • Assorted scalability improvement for large machines
  • Support SGI Altix and 512 CPUs with Linux kernel crash dump (LKCD)
  • Added CPUSET support for IPF
  • Added scalability enhancements for big IPF machines
  • Added support for systems with many IRQ resources
  • Added clustered APIC support for x86_64
  • Default readahead to 512KB (instead of 128KB)
  • fix TPC-C performance problems on x86_64 (caused by vsyscall gettimeofday)
  • don't waste memory for hashes on huge machines
  • fix possible cpuset race
  • avoid memory allocation problem on machines where still enough memory is available
  • allow memory holes on S/390
  • enlarge max number of CPUs on x86-64 and number of IO-APIC
  • Disable clustered APIC mode on AMD systems
  • fix perfmon assertion failure in pfm_load_regs


  • Back ported epoll fixes from 2.6.9
  • Integrated numerous other bugfixes from upstream/mainline kernel
  • Several backports from upstream/mainline kernel:

    + unmap_mapping_range() from 2.6.6 + generic_file_direct_write() and generic_file_buffered_write() from 2.6.9-rc4 + backport mapping_mapped() + export sync_page_range

  • Support official variable name INSTALL_MOD_DIR in addition to our MOD_DIR
  • Fixed hooks to enable CA
  • Added kernel support for POSIX message queue
  • fix sys_stime() in 31-bit compatibility mode on S/390
  • fix CD/DVD writing for non-root users
  • fix device special files and permission cache
  • fix problem with pclose() sometimes hanging
  • crbce should provide time stamp in msec rather than in jiffies
  • allow IRQ0 to be used as a legal PCI device IRQ
  • fix possible EEH or memory corruption when DMA crosses a 64k boundary
  • fix iSeries Linux on legacy systems not reporting PURR
  • do not create unnamed directory under /rcfs/taskclass
  • fix hang caused by loopback TX
  • limit max number of concurrent khelper processes
  • allow ACPI PCI hotplug callbacks to set/get_attention_status()
  • fix IMM highmem oops
  • avoid deadlocks with non-ram under mlockall
  • always add credentials to inodes for NFS
  • fix race condition in unix_dgram_recvmsg()
  • fix vfree() with interrupts disabled in sg driver
  • do not fill up process table when many events occur
  • fix MCA during cross-partition MPI (ia64)
  • fix random kernel memory corruption if openfirmware stdin device is an usb controller
  • Fix CPU time reporting for single processes
  • fix kernel hang in __getblk_slow()
  • sunrpc - don't crash on unknown program numbers
  • fix memory leak in pageattr code (x86 and x86-64 only)
  • fix RAID1 device failure resulting in kernel crash
  • don't lose edge triggered IRQ when delivered while IRQ disabled
  • fix USB HID driver parsing usage IDs
  • allow keyboard to survive if any key (e.g. F2) is pressed early


With Rapid7 live dashboards, I have a clear view of all the assets on my network, which ones can be exploited, and what I need to do in order to reduce the risk in my environment in real-time. No other tool gives us that kind of value and insight.

– Scott Cheney, Manager of Information Security, Sierra View Medical Center