This is the OpenBSD 3.8 release errata & patch list:

The patches below are available in CVS via the OPENBSD_3_8 patch branch.

For more detailed information on how to install patches to OpenBSD, please consult the OpenBSD FAQ.

• 020: SECURITY FIX: October 12, 2006   All architectures
  Fix 2 security bugs found in OpenSSH. A pre-authentication denial of service (found by Tavis Ormandy) that would cause sshd(8) to spin until the login grace time expired. An unsafe signal handler (found by Mark Dowd) that is vulnerable to a race condition that could be exploited to perform a pre-authentication denial of service. CVE-2006-4924, CVE-2006-5051
  A source code patch exists which remedies this problem.

• 019: SECURITY FIX: October 7, 2006   All architectures
  Fix for an integer overflow in systrace(4)'s STRIOCREPLACE support, found by Chris Evans. This could be exploited for DoS, limited kmem reads or local privilege escalation.
  A source code patch exists which remedies this problem.

• 018: SECURITY FIX: October 7, 2006   All architectures
  Several problems have been found in OpenSSL. While parsing certain invalid ASN.1 structures an error condition is mishandled, possibly resulting in an infinite loop. A buffer overflow exists in the SSL_get_shared_ciphers function. A NULL pointer may be dereferenced in the SSL version 2 client code. In addition, many applications using OpenSSL do not perform any validation of the lengths of public keys being used. CVE-2006-2937, CVE-2006-3738, CVE-2006-4343, CVE-2006-2940
  A source code patch exists which remedies this problem.

• 017: SECURITY FIX: October 7, 2006   All architectures
  httpd(8) does not sanitize the Expect header from an HTTP request when it is reflected back in an error message, which might allow cross-site scripting (XSS) style attacks. CVE-2006-3918
  A source code patch exists which remedies this problem.

• 016: SECURITY FIX: September 8, 2006   All architectures
  Due to incorrect PKCS#1 v1.5 padding validation in OpenSSL, it is possible for an attacker to construct an invalid signature which OpenSSL would accept as a valid PKCS#1 v1.5 signature. CVE-2006-4339
  A source code patch exists which remedies this problem.

• 015: SECURITY FIX: September 8, 2006   All architectures
  Two Denial of Service issues have been found with BIND. An attacker who can perform recursive lookups on a DNS server and is able to send a sufficiently large number of recursive queries, or is able to get the DNS server to return more than one SIG(covered) RRsets can stop the functionality of the DNS service. An attacker querying an authoritative DNS server serving a RFC 2535 DNSSEC zone may be able to crash the DNS server. CVE-2006-4095 CVE-2006-4096
  A source code patch exists which remedies this problem.

• 014: SECURITY FIX: September 2, 2006   All architectures
  Due to the failure to correctly validate LCP configuration option lengths, it is possible for an attacker to send LCP packets via an sppp(4) connection causing the kernel to panic. CVE-2006-4304
  A source code patch exists which remedies this problem.

• 013: SECURITY FIX: August 25, 2006   All architectures
  A problem in isakmpd(8) caused IPsec to run partly without replay protection. If isakmpd(8) was acting as responder during SA negotiation, SA's with a replay window of size 0 were created. An attacker could reinject sniffed IPsec packets, which will be accepted without checking the replay counter.
  A source code patch exists which remedies this problem.

• 012: SECURITY FIX: August 25, 2006   All architectures
  It is possible to cause the kernel to panic when more than the default number of sempahores have been allocated.
  A source code patch exists which remedies this problem.

• 011: SECURITY FIX: August 25, 2006   All architectures
  Due to an off-by-one error in dhcpd(8), it is possible to cause dhcpd(8) to exit by sending a DHCPDISCOVER packet with a 32-byte client identifier option. CVE-2006-3122
  A source code patch exists which remedies this problem.

• 010: SECURITY FIX: August 25, 2006   All architectures
  A potential denial of service problem has been found in sendmail. A message with really long header lines could trigger a use-after-free bug causing sendmail to crash.
  A source code patch exists which remedies this problem.

• 009: SECURITY FIX: July 30, 2006   All architectures
  httpd(8)'s mod_rewrite has a potentially exploitable off-by-one buffer overflow. The buffer overflow may result in a vulnerability which, in combination with certain types of Rewrite rules in the web server configuration files, could be triggered remotely. The default install is not affected by the buffer overflow. CVE-2006-3747
  A source code patch exists which remedies this problem.

• 008: SECURITY FIX: June 15, 2006   All architectures
  A potential denial of service problem has been found in sendmail. A malformed MIME message could trigger excessive recursion which will lead to stack exhaustion. This denial of service attack only affects delivery of mail from the queue and delivery of a malformed message. Other incoming mail is still accepted and delivered. However, mail messages in the queue may not be reattempted if a malformed MIME message exists.
  A source code patch exists which remedies this problem.

• 007: SECURITY FIX: May 2, 2006   All architectures
  A security vulnerability has been found in the X.Org server -- CVE-2006-1526. Clients authorized to connect to the X server are able to crash it and to execute malicious code within the X server.
  A source code patch exists which remedies this problem.

• 006: SECURITY FIX: March 25, 2006   All architectures
  A race condition has been reported to exist in the handling by sendmail of asynchronous signals. A remote attacker may be able to execute arbitrary code with the privileges of the user running sendmail, typically root.
  A source code patch exists which remedies this problem.

• 005: SECURITY FIX: February 12, 2006   All architectures
  Josh Bressers has reported a weakness in OpenSSH caused due to the insecure use of the system(3) function in scp(1) when performing copy operations using filenames that are supplied by the user from the command line. This can be exploited to execute shell commands with privileges of the user running scp(1).
  A source code patch exists which remedies this problem.

• 004: RELIABILITY FIX: January 13, 2006   i386 architecture
  Constrain i386_set_ioperm(2) so even root is blocked from accessing the ioports unless the machine is running at lower securelevels or with an open X11 aperture.
  A source code patch exists which remedies this problem.

• 003: RELIABILITY FIX: January 13, 2006   i386 architecture
  Change the implementation of i386 W^X so that the "execute line" can move around. Before it was limited to being either at 512MB (below which all code normally lands) or at the top of the stack. Now the line can float as mprotect(2) and mmap(2) requests need it to. This is now implemented using only GDT selectors instead of the LDT so that it is more robust as well.
  A source code patch exists which remedies this problem.

• 002: SECURITY FIX: January 5, 2006   All architectures
  Do not allow users to trick suid programs into re-opening files via /dev/fd.
  A source code patch exists which remedies this problem.

• 001: SECURITY FIX: January 5, 2006   All architectures
  A buffer overflow has been found in the Perl interpreter with the sprintf function which may be exploitable under certain conditions.
  A source code patch exists which remedies this problem.