relnotes-sparc64.txt

标准模板库自修教程与参考手册 STL进行C++编程.我的代码发完了

   result, it is now possible to obtain crash dumps from panics during the
   late stages of kernel initialization (before the system enters into
   single-user mode). [MERGED]

   The kernel memory allocator is now a slab memory allocator, similar to
   that used in Solaris. This is a SMP-safe memory allocator that has
   near-linear performance as the number of CPUs increases. It also allows
   for reduced memory fragmentation.

     ----------------------------------------------------------------------

  2.1.1 Processor/Motherboard Support

   SMP support has been largely reworked, incorporating code from BSD/OS 5.0.
   One of the main features of SMPng (``SMP Next Generation'') is to allow
   more processes to run in kernel, without the need for spin locks that can
   dramatically reduce the efficiency of multiple processors. Interrupt
   handlers now have contexts associated with them that allow them to be
   blocked, which reduces the need to lock out interrupts.

   The UltraSPARC platform is now supported by FreeBSD. The following
   machines are supported to at least some degree: Ultra 1/2/5/10/30/60,
   Enterprise 220R/420R, Netra T1 AC200/DC200, Netra T 105, and Blade 100.
   SMP is supported, and has been tested on the Ultra 2, Ultra 60, Enterprise
   220R, and Enterprise 420R.

     ----------------------------------------------------------------------

  2.1.2 Bootloader Changes

   The kernel and modules have been moved to the directory /boot/kernel, so
   they can be easily manipulated together. The boot loader has been updated
   to make this change as seamless as possible.

     ----------------------------------------------------------------------

  2.1.3 Network Interface Support

   The an(4) driver for Cisco Aironet cards now supports Wired Equivalent
   Privacy (WEP) encryption, settable via ancontrol(8). [MERGED]

   The an(4) driver now supports the Cisco Aironet 350 series of adaptors.
   [MERGED]

   The an(4) driver now supports ``monitor'' mode, settable via the -M option
   to ancontrol(8). [MERGED]

   The an(4) driver now supports Cisco LEAP, as well as the ``Home'' WEP key.
   The Linux Aironet utilities are now supported under emulation. [MERGED]

   The dc(4) driver now supports NICs based on the Xircom 3201 and Conexant
   LANfinity RS7112 chips.

   The dc(4) driver now has support for VLANs. [MERGED]

   The de(4) driver now performs round-robin arbitration between the transmit
   and receive units of the 21143, instead of giving priority to the receive
   unit. This gives a 10-15% performance improvement in the forwarding rate
   under heavy load. [MERGED]

   A bug in the ed(4) driver that could cause panics with very short packets
   and BPF or bridging active has been fixed. [MERGED]

   The ed(4) driver now has support for D-Link DL10022 chips, necessary for
   the NetGear FA-410TX and other cards. As a result, device miibus is
   required in kernel configurations using the ed(4) driver. [MERGED]

   The faith(4) device is now loadable, unloadable, and clonable. [MERGED]

   The fxp(4) driver now requires a device miibus entry in the kernel
   configuration file. [MERGED]

   The fxp(4) driver now contains a workaround for PCI protocol violations
   caused by defects in some systems based on the Intel ICH2/ICH2-M chip. The
   workaround is to rewrite the EEPROM on the interface to disable Dynamic
   Standby Mode; once the EEPROM is rewritten, the system needs to be
   rebooted for the new settings to take effect. [MERGED]

   The fxp(4) driver now supports Intel's loadable microcode to implement
   receive-side interrupt coalescing and packet bundling, on NICs that
   support these features. This support can be activated by the use of the
   link0 option to ifconfig(8). [MERGED]

   The gem driver has been added to support the Sun GEM Gigabit Ethernet and
   ERI Fast Ethernet adapters.

   The gx(4) driver has been added to support NICs based on the Intel 82542
   and 82543 Gigabit Ethernet controller chips. Both fiber and copper
   variants of the cards are supported. Both boards support VLAN
   tagging/insertion, and the 82543 additionally supports TCP/IP checksum
   offload. [MERGED]

   The hme driver has been added to support the Sun HME Fast Ethernet
   adapter, onboard on many Sun Ultra series machines.

   The lge(4) driver has been added to support the Level 1 LXT1001
   NetCellerator Gigabit Ethernet controller chip. This device is used on
   some fiber optic GigE cards from SMC, D-Link and Addtron. Jumbograms and
   TCP/IP checksum offload on receive are supported, although hardware VLAN
   filtering is not. [MERGED]

   The my driver, which supports the Myson Fast Ethernet and Gigabit Ethernet
   adapters, has been added. [MERGED]

   Added the nge(4) driver, which supports PCI Gigabit Ethernet adapters
   based on the National Semiconductor DP83820 and DP83821 Gigabit Ethernet
   controller chips, including the D-Link DGE-500T, SMC EZ Card 1000
   (SMC9462TX), Asante FriendlyNet GigaNIC 1000TA and 1000TPC and Addtron
   AEG320T. This driver supports transmit and receive checksum offloading.
   [MERGED]

   The pcn(4) driver, which supports the AMD PCnet/FAST, PCnet/FAST+,
   PCnet/FAST III, PCnet/PRO, PCnet/Home, and HomePNA adapters, has been
   added. Although these cards are already supported by the lnc(4) driver,
   the pcn(4) driver runs these chips in 32-bit mode and uses the RX
   alignment feature to achieve zero-copy receive. This driver is also
   machine-independent, so it will work on both the i386 and Alpha platforms.
   The lnc(4) driver is still needed to support non-PCI cards. [MERGED]

   The ray(4) driver, which supports the Webgear Aviator wireless network
   cards, has been committed. The operation of ray(4) interfaces can be
   modified by raycontrol(8). [MERGED]

   Added support for PCI Ethernet adapters based on the SiS 900 and SiS 7016
   Fast Ethernet controller chips (for example, as seen on the SiS 635 and
   735 motherboard chipsets), as well as the National Semiconductor DP83815
   chipset (including the NetGear FA311-TX and FA312-TX) in the form of the
   sis(4) driver. This device has support for VLANs. [MERGED]

   The stf(4) device is now clonable.

   The tap(4) driver, a virtual Ethernet device driver for bridged
   configurations, has been added. This device is clonable. [MERGED]

   The ti(4) driver now supports the Alteon AceNIC 1000baseT Gigabit Ethernet
   and Netgear GA620T 1000baseT Gigabit cards. [MERGED]

   The ti(4) driver correctly masks VLAN tags. [MERGED]

   The txp(4) driver has been added to support NICs based on the 3Com 3XP
   Typhoon/Sidewinder (3CR990) chipset. [MERGED]

   vlan(4) devices are now loadable, unloadable, and clonable. [MERGED]

   The wi(4) driver now has support for Prism II and Prism 2.5-based NICs.
   104/128-bit WEP now works on Prism cards. [MERGED]

   The xl(4) driver now supports the 3Com 3C556 and 3C556B MiniPCI adapters
   used on some laptops. [MERGED]

   The xl(4) driver now supports reception of VLAN tagged frames (on the
   ``Cyclone'' or newer chipsets). [MERGED]

   The xl(4) driver now supports send- and receive-side TCP/IP checksum
   offloading for NICs implementing this feature, such as the 3C905B, 3C905C,
   and 3C980C. [MERGED]

   A bug in the xl(4) driver, related to statistics overflow interrupt
   handling, was causing slowdowns at medium to high packet rates; this has
   been fixed. [MERGED]

   The per-interface ifnet structure now has the ability to indicate a set of
   capabilities supported by a network interface, and which ones are enabled.
   ifconfig(8) has support for querying these capabilities. [MERGED]

   Performance with hosts having a large number of IP aliases has been
   improved, by replacing the per-interface if_inaddr linear list with a hash
   table. [MERGED]

   Network devices now automatically appear as special files in /dev/net.
   Interface hardware ioctls (not protocol or routing) can be performed on
   these devices. The SIOCGIFCONF ioctl may be performed on the special
   /dev/network node.

   Selected network drivers now implement a semi-polling mode, which makes
   systems much more resilient to attacks and overloads. To enable polling,
   the following options are required in a kernel configuration file:

     options DEVICE_POLLING
     options HZ=1000 # not compulsory but strongly recommended

   The kern.polling.enable sysctl variable will then activate polling mode;
   with the kern.polling.user_frac sysctl indicating the percentage of CPU
   time to be reserved for userland. The devices initially supporting polling
   are dc(4), fxp(4), rl(4), and sis(4). More details can be found in the
   polling(4) manual page. [MERGED]

     ----------------------------------------------------------------------

  2.1.4 Network Protocols

   accept_filter(9), a kernel feature to reduce overheads when accepting and
   reading new connections on listening sockets, has been added. [MERGED]

   The proxy modifier to arp(8)'s -d option has been renamed to pub, for
   consistency with the -s option. The only keyword has been added to the -s
   and -S flags, to be used in creating ``proxy-only'' published entries.
   [MERGED]

   The read timeout feature of bpf(4) now works more correctly with
   select(2)/poll(2), and therefore with pthreads. [MERGED]

   bridge(4) and dummynet(4) have received some enhancements and bug fixes,
   and are now loadable modules. [MERGED]

   bridge(4) now has better support for multiple, fully-independent bridging
   clusters, and is much more stable in the presence of dynamic attachments
   and detatchments. Full support for VLANs is also supported. [MERGED]

   ICMP ECHO and TSTAMP replies are now rate limited. TCP RSTs generated due
   to packets sent to open and unopen ports are now limited by separate
   counters. Each rate limiting queue now has its own description.

   ICMP UNREACH_FILTER_PROHIB messages can now RST TCP connections in the
   SYN_SENT state if the correct sequence numbers are sent back, as
   controlled by the net.inet.tcp.icmp_may_rst sysctl. [MERGED]

   IP multicast now works on VLAN devices. Several other bugs in the VLAN
   code have also been fixed.

   A bug in the IPSec processing for IPv4, which caused the inbound SPD
   checks to be ignored, has been fixed. [MERGED]

   ipfw(4) now filters correctly in the presence of ECN bits in TCP segments.
   [MERGED]

   A new ng_etf(4) netgraph node allows Ethernet type packets to be filtered
   to different hooks depending on ethertype. [MERGED]

   The ng_gif(4) and ng_gif_demux(4) netgraph nodes, for operating on gif(4)
   devices, have been added.

   The ng_ip_input(4) netgraph node, for queueing IP packets into the main IP
   input processing code, has been added.

   The ng_mppc(4) and ng_bridge(4) node types have been added to the
   netgraph(4) subsystem. The ng_ether(4) node is now dynamically loadable.
   Miscellaneous bug fixes and enhancements have also been made. [MERGED]

   A new netgraph node type ng_one2many(4) for multiplexing and
   demultiplexing packets over multiple links has been added. [MERGED]

   A new sysctl net.inet.ip.check_interface, which is on by default, causes
   IP to verify that an incoming packet arrives on an interface that has an
   address matching the packet's destination address. [MERGED]

   A new sysctl net.link.ether.inet.log_arp_wrong_iface has been added to
   control the suppression of logging when ARP replies arrive on the wrong
   interface. [MERGED]

   A new options RANDOM_IP_ID kernel option causes the ID field of IP packets
   to be randomized. This closes a minor information leak which allows a
   remote observer to determine the rate at which the machine is generating
   packets, since the default behavior is to increment a counter for each
   packet sent. [MERGED]

   TCP has received some bug fixes for its delayed ACK behavior. [MERGED]

   TCP now supports the NewReno modification to the TCP Fast Recovery
   algorithm. This behavior can be controlled via the net.inet.tcp.newreno
   sysctl variable. [MERGED]

   TCP now uses a more aggressive timeout for initial SYN segments; this
   allows initial connection attempts to be dropped much faster. [MERGED]

   The TCP_COMPAT_42 kernel option has been removed. [MERGED]

   The TCP_RESTRICT_RST kernel option has been removed. Similar functionality
   can be achieved with the net.inet.tcp.blackhole sysctl variable. [MERGED]

   TCP now has RFC 1323 extensions enabled by default in rc.conf(5). [MERGED]

   RFC 1323 and RFC 1644 TCP extensions are now disabled for a connection in
   progress if no response has been received by the third SYN segment sent.
   This behavior tries to work around (very old) terminal servers with buggy
   VJ header compression implementations. [MERGED]

   The TCP implementation no longer requires the allocation of a TCP template
   structure for each connection; this should reduce the buffer usage on
   large systems handling many connections. [MERGED]

   TCP's default buffer sizes, controlled by the net.inet.tcp.sendspace and
   net.inet.tcp.recvspace sysctl variables, have been increased to 32K and
   64K respectively. Previously, the default for both buffer sizes was 16K.