Scientific Linux Security Update : kernel on SL4.x i386/x86_64


This update addresses the following security issues : - a buffer overflow was found in the Linux kernel Partial Reliable Stream Control Transmission Protocol (PR-SCTP) implementation. This could, potentially, lead to a denial of service if a Forward-TSN chunk is received with a large stream ID. (CVE-2009-0065, Important) - a memory leak was found in keyctl handling. A local, unprivileged user could use this flaw to deplete kernel memory, eventually leading to a denial of service. (CVE-2009-0031, Important) - a deficiency was found in the Remote BIOS Update (RBU) driver for Dell systems. This could allow a local, unprivileged user to cause a denial of service by reading zero bytes from the image_type or packet_size file in '/sys/devices/platform/dell_rbu/'. (CVE-2009-0322, Important) - a deficiency was found in the libATA implementation. This could, potentially, lead to a denial of service. Note: by default, '/dev/sg*' devices are accessible only to the root user. (CVE-2008-5700, Low) This update also fixes the following bugs : - when the hypervisor changed a page table entry (pte) mapping from read-only to writable via a make_writable hypercall, accessing the changed page immediately following the change caused a spurious page fault. When trying to install a para-virtualized Scientific Linux 4 guest on a Scientific Linux 5.3 dom0 host, this fault crashed the installer with a kernel backtrace. With this update, the 'spurious' page fault is handled properly. (BZ#483748) - net_rx_action could detect its cpu poll_list as non-empty, but have that same list reduced to empty by the poll_napi path. This resulted in garbage data being returned when net_rx_action calls list_entry, which subsequently resulted in several possible crash conditions. The race condition in the network code which caused this has been fixed. (BZ#475970, BZ#479681 & BZ#480741) - a misplaced memory barrier at unlock_buffer() could lead to a concurrent h_refcounter update which produced a reference counter leak and, later, a double free in ext3_xattr_release_block(). Consequent to the double free, ext3 reported an error ext3_free_blocks_sb: bit already cleared for block [block number] and mounted itself as read-only. With this update, the memory barrier is now placed before the buffer head lock bit, forcing the write order and preventing the double free. (BZ#476533) - when the iptables module was unloaded, it was assumed the correct entry for removal had been found if 'wrapper->ops->pf' matched the value passed in by 'reg->pf'. If several ops ranges were registered against the same protocol family, however, (which was likely if you had both ip_conntrack and ip_contrack_* loaded) this assumption could lead to NULL list pointers and cause a kernel panic. With this update, 'wrapper->ops' is matched to pointer values 'reg', which ensures the correct entry is removed and results in no NULL list pointers. (BZ#477147) - when the pidmap page (used for tracking process ids, pids) incremented to an even page (ie the second, fourth, sixth, etc. pidmap page), the alloc_pidmap() routine skipped the page. This resulted in 'holes' in the allocated pids. For example, after pid 32767, you would expect 32768 to be allocated. If the page skipping behavior presented, however, the pid allocated after 32767 was 65536. With this update, alloc_pidmap() no longer skips alternate pidmap pages and allocated pid holes no longer occur. This fix also corrects an error which allowed pid_max to be set higher than the pid_max limit has been corrected. (BZ#479182) The system must be rebooted for this update to take effect.