FreeBSD 8.1-RELEASE Release Notes

2.2.1 Boot Loader Changes

[pc98] The boot2 bootcode has been reimplemented based on the i386 counterpart. It now supports ELF binary, UFS2 file system, and larger number of slices.

[ia64] The EFI loader program now supports a command-line option -dev currdev to specify the default value of currdev. This option can be set by the EFI boot manager.

[powerpc] The loader(8) program now supports U-Boot storage.

[i386] The algorithm the loader(8) uses has been improved to choose a memory range for its heap when using a range above 1MB. This fixes a symptom that the loader fails to load a kernel.

A kernel environment variable vfs.root.mountfrom now supports multiple elements for root file system in a space-separated list. Each list element will be tried in order and the first available one will be mounted.

The zfsloader has been added. This is a separate zfs(8) enabled loader. Note that a ZFS bootcode (zfsboot or gptzfsboot) need to be installed to use this new loader.

The zfsboot and gptzfsboot bootcode now fully support 64-bit LBAs for disk addresses. This allows booting from large volumes.

2.2.2 Hardware Support

[powerpc] The adb driver now supports for interpreting taps on ADB touchpads as a button click.

The amdsbwd(4) driver for AMD SB600/SB7xx watchdog timer has been added.

[powerpc] The apt driver for the Apple Touchpad present on MacBook has been added to GENERIC kernel.

[sparc64] The epic(4) driver for the front panel LEDs in Sun Fire V215/V245 has been added.

A bug in the ipmi(4) driver that caused incorrect watchdog timer setting has been fixed.

[sparc64] The pci(4) driver now supports a JBus to PCIe bridge (called as “Fire”) found in the Sun Fire V215/V245 and Sun Ultra 25/45 machines.

[powerpc] The smu(4) driver now provides thermal management and monitoring features. This allows fan control and thermal monitoring on SMU-based Apple G5 machines, as well as an led(4) interface to control the sleep LED.

The tnt4882(4) driver for IEEE-488 (GPIB) bus now supports National Instruments TNT5004 chip.

The uart(4) driver now supports NetMos NM9865 family of Serial/Parallel ports.

The uep(4) driver for USB onscreen touch panel from eGalax has been added. This driver is supported by x11-drivers/xf86-input-egalax.

A bug in the uftdi(4) driver that can allow to send a zero length packet has been fixed.

The usb(4) subsystem now reports devd(8) notify events with the device properties instead of attach events. The following is an example entry of devd.conf(5) to match a umass(4) device with a SCSI subclass and BBB protocol:

notify 100 {
    match "system"      "USB";
    match "subsystem"   "INTERFACE";
    match "type"        "ATTACH";
    match "intclass"    "0x08";
    match "intsubclass" "0x06";
    match "intprotocol" "0x50";
    action "/path/to/command -flag";
};

2.2.3 Network Protocols

FreeBSD flowtable now supports IPv6. This is for per-CPU caching flows as a means of accelerating L3 and L2 lookups as well as providing stateful load balancing when ECMP (Equal-Cost Multi-Path routing) is enabled by option RADIX_MPATH.

A new capability flag LINKSTATE has been added to struct ifnet.if_capabilities. This indicates if the interface can check the link state or not. The ifconfig(8) utility now shows this flag if supported.

A new event handler iflladdr_event has been added. This signals that the L2 address on an interface has changed, and lets stacked interfaces such as vlan(4) detect that their lower interface has changed and adjust things in order to keep working. This fixes an issue of lagg(4) and vlan(4) configuration.

IPcomp (IP Payload Compression Protocol defined in RFC 2393) protocol is now enabled by default. Note that this requires option IPSEC in the kernel configuration file and GENERIC kernel does not include it. This functionality can be disabled by using a sysctl variable net.inet.ipcomp.ipcomp_enable.

The ipfw(4) subsystem including dummynet(4) has been updated to “ipfw3” and various bugs have been fixed:

• The major enhancement is a completely restructured version of dummynet(4), with support for different packet scheduling algorithms (loadable at runtime), faster queue/pipe lookup, and a much cleaner internal architecture and kernel/userland ABI which simplifies future extensions.

• All of O(N) sequences in the firewall rule evaluation removed from the kernel critical sections. The worst case is now O(log N).

• It now supports ipfw0 pseudo interface for logging similar to pflog(4). A sysctl net.inet.ip.fw.verbose=0 enables logging to ipfw0, and net.inet.ip.fw.verbose=1 sends logging to syslog(3) as before.

• The me keyword in the ipfw(4) rule now matches any IPv6 addresses configured on an interface as well as IPv4 ones.

• A bug that keep-alive rule did not work for IPv6 packets has been fixed.

• The lookup match option has been added.

  lookup {dst-ip|src-ip|dst-port|src-port|uid|jail} N
  

  This searches the specified field in table N and sets tablearg accordingly. With dst-ip or src-ip the option replicates two existing options. When used with other arguments, the option can be useful to quickly dispatch traffic based on other fields.

• A bug in the sysctl(8) variable ip.fw.one_pass handling has been fixed. A packet which comes from a pipe without being delayed incorrectly ignored this variable.

A memory alignment issue in the ng_ksocket(4) and ng_ppp(4), Netgraph node drivers have been fixed. This fixes kernel panics due to the misalignment.

The ng_bridge(4) and ng_hub(4) Netgraph node drivers now supports a flag persistent. It disables automatic node shutdown when the last hook gets disconnected. The new control messages NGM_BRIDGE_SET_PERSISTENT and NGM_HUB_SET_PERSISTENT have been added for the flag.

The pf(4) subsystem now supports sloppy keyword to enable a TCP state machine for tracking TCP connections with no sequence number check. This feature is in the latest version of pf.

The pfil(9) framework for packet filtering in FreeBSD kernel now supports separate packet filtering instances like ipfw(4) for each VIMAGE jail.

A bug that proxy ARP entries cannot be added over point-to-point link types has been fixed.

The tap(4) pseudo interface now reports the link state properly by updating if_link_state variable in the kernel.

The vlan(4) pseudo interface has been added to GENERIC kernel.

The vlan(4) pseudo interface now supports TSO (TCP Segmentation Offloading). The capability flag is named as IFCAP_VLAN_HWTSO and it is separated from IFCAP_VLAN_HWTAGGING. The age(4), alc(4), ale(4), bce(4), bge(4), cxgb(4), jme(4), re(4), and mxge(4) driver support this feature.

The vlan(4) pseudo interface for IEEE 802.1Q VLAN now ignore renaming of the parent's interface name. The configured VLAN interfaces continue to work with the new name while previously the configurations were removed as the renaming happens.

2.2.4 Disks and Storage

The ada(4) driver now supports BIO_DELETE. For SSDs this uses TRIM feature of DATA SET MANAGEMENT command, as defined by ACS-2 specification working draft. For Compact Flash use CFA ERASE command, same as ad(4) does. This change realizes restoring write speed of SSDs which supports TRIM command by doing newfs -E /dev/ada1, for example.

The ahci(4) driver now supports SATA part of Marvell 88SE912x controllers.

The ahci(4) driver now supports FIS-based (Frame Information Structure) switching of port multiplier on supported controllers.

The ahd(4) driver now supports three separated error counters for correctable, uncorrectable, and fatal, in sysctl(8) MIB.

A new kernel option option ATA_CAM has been added. This turns ata(4) controller drivers into cam(4) interface modules. When enabled, this option deprecates all ata(4) peripheral drivers and interfaces such as ad and acd, and allows cam(4) drivers ada, and cd and interfaces to be natively used instead. Note that this is not enabled by default in the GENERIC kernel.

A bug in the ata(4) driver which can lead to interrupt storms and command timeouts has been fixed.

USB mass storage device support in the ata(4) driver has been removed. Note that this was not used in GENERIC kernel and the umass(4) driver supports such devices for a long time.

FreeBSD cam(3) SCSI framework has been improved:

• SATA and PATA support has been improved and it now recognizes more detail device capabilities. For example, the ahci(4) and siis(4) driver now reports maximum tag number to the framework to optimize the NCQ handling.

• A loader tunable kern.cam.boot_delay has been added. This controls the delay time before cam(3) probes the attached devices.

• SCSI error recovery for devices on buses without automatic sense reporting has been improved. Typical devices are on ATAPI and USB. For example, this allows cam(3) to wait, while CD drive loads disk, instead of immediately return error status.

• The cam(4) ATA transport layer now supports Power-Up In Stand-by (PUIS). The PUIS is a configuration of SATA or PATA drives to prevent them from automatic spin-up when power is applied. A typical application is staggered spin-up.

• The cam(4) ATA transport layer now supports negotiating and enabling additional SATA features such as device initiated power management, Automatic Partial to Slumber mode transition, and DMA auto-activation.

A livelock issue of the ciss(4) driver under a high load has been fixed.

A bug in the fdc(4) driver which prevents the kernel module from unloading has been fixed.

The glabel(8) now supports the following sysctl variables for each label type to enable the labeling itself:

kern.geom.label.ext2fs.enable
kern.geom.label.iso9660.enable
kern.geom.label.msdosfs.enable
kern.geom.label.ntfs.enable
kern.geom.label.reiserfs.enable
kern.geom.label.ufs.enable
kern.geom.label.ufsid.enable
kern.geom.label.gptid.enable
kern.geom.label.gpt.enable

Note that all of them are also loader tunables. They are enabled (set as 1) by default.

geom(8) providers including complex ones such as gconcat(8), gmirror(8), graid3(8), gstripe(8), and some hardware RAID device drivers like twa(4) now inform its optimal access block size to the upper layer.

The gmirror(8) utility now supports configure -p priority command to change the providers priority.

The balancing mode algorithm load used in the gmirror(8) utility has been changed and it is now the default one instead of split:

• Instead of measuring last request execution time for each drive and choosing one with smallest time, use averaged number of requests, running on each drive. This information is more accurate and timely. It allows to distribute load between drives in more even and predictable way.

• For each drive track offset of the last submitted request. If new request offset matches previous one or close for some drive, prefer that drive. It allows to significantly speedup simultaneous sequential reads.

The gmultipath(8) utility now supports destroy, rotate, getactive commands.

A bug in the graid3(8) which causes a panic when a large request arrives has been fixed. This happens when MAXPHYS is set as larger than 128k.

The default block size of gstripe(8) has been increased from 4k to 64k.

The GEOM_SCHED module has been added. This supports scheduling disk I/O requests in a device independent manner. A supported algorithm is an anticipatory scheduler gsched_rr which gives very nice performance improvements in presence of competing random access patterns. See also gsched(8) manual page for more details.

The HAST (Highly Available STorage) framework has been added:

• This is a framework to allow transparently storing data on two physically separated machines connected over the TCP/IP network. HAST works in Primary-Secondary (Master-Backup, Master-Slave) configuration, which means that only one of the cluster nodes can be active at any given time. Only Primary node is able to handle I/O requests to HAST-managed devices. Currently HAST is limited to two cluster nodes in total.

• This operates on block level; it provides disk-like devices in /dev/hast/ directory for use by file systems and/or applications. Working on block level makes it transparent for file systems and applications. There in no difference between using HAST-provided device and raw disk, partition, etc. All of them are just regular geom(8) providers in FreeBSD.

• The userland part consists of hastd(8), hastctl(8), and hast.conf(5). More details can be found at http://wiki.FreeBSD.org/HAST.

The isp(4) driver has been improved in stability.

The mvs(4) CAM ATA driver for Marvell 88SX50XX/88SX60XX/88SX70XX/SoC SATA controllers has been added. This driver supports same hardware as the ata(4) driver does, but provides many additional features, such as NCQ and PMP.

The siis(4) driver now enables MSI by default on SiI3124-based devices. This can be disabled by using a hint.siis.0.msi loader tunable.

The Max Read Request Size in the siis(4) driver for PCIe chips has been increased from 512 to 1024 bytes for better performance.

The twa(4) driver has been updated to the latest version from LSI.

2.2.5 File Systems

The msdosfs(5) subsystem is now MP-safe and a race condition when a force unmount happens has been fixed.

FreeBSD NFS subsystem now supports a timeout for the negative name cache entries in the client. This avoids a bogus negative name cache entry from persisting forever when another client creates an entry with the same name within the same NFS server time of day clock tick. The mount option negnametimeo can be used to override the default timeout interval (60 seconds) on a per-mount-point basis. a Setting negnametimeo to 0 disables negative name caching for the mount point.

A race condition in FreeBSD NFS subsystem that occurs when nfsiod(8) threads are being created has been fixed. This also fixes an interoperability issue found in combination of a FreeBSD NFS client and a Linux NFS server.

The inode number handling in ffs(7) file system is now unsigned. Previously some large inode numbers can be treated as negative, and this issue shows up at file systems with the size of more than 16Tb in 16k block case. The newfs(8) utility never create a file system with more than 2^32 inodes by cutting back on the number of inodes per cylinder group if necessary to stay under the limit.

The UFS file system ( ffs(7)) now supports NFSv4 ACL.

FreeBSD VFS(9) subsystem now supports a new sysctl variable vfs.vlru_allow_cache_src. This allow vnlru kernel thread to reclaim of the directory vnodes that are source of the namecache records. This is not enabled by default because for typical workload it would make namecache unusable, but large nested directory tree easily puts any process that accesses file system into one second wait for vnlru kernel thread.

The ZFS file system has been improved:

• It now supports NFSv4 ACL.

• The L2ARC code has been improved in stability and performance.

• The zpool version has been updated to version 14. It is now possible to use zpools created on OpenSolaris 2009.06.

• A sysctl variable vfs.zfs.txg.write_limit_override has been added. This can be used for tuning of ZFS write throttling.

• ZFS prefetch statistics has been added as a sysctl variable kstat.zfs.misc.zfetchstats.

• The zfs(8) zpool export command now supports -F flag. When exporting with this flag, zpool.cache remains untouched.

• A data corruption issue of zfs send/receive between two different platforms has been fixed. Symbolic links could be broken in the previous releases.

• A possible deadlock of zfs receive has been fixed.

• Possible panics of zfs destroy and zfs rollback have been fixed.

• A occasional failure of zfs rename due to a busy state has been fixed.

• Bugs that zfs snapshot -r fails when the file system is busy, and zfs receive can fail with an E2BIG error, have been fixed.

2.3.1 /etc/rc.d Scripts

The service(8) command has been added. This provides an easy command-line interface to the rc.d system.

The rc.d/ipfw script and /etc/rc.firewall now supports IPv6 and rc.d/ip6fw script and /etc/rc.firewall6 are obsolete. Note that ipv6_firewall_* variables in rc.conf(5) are replaced with firewall_client_net_ipv6, firewall_simple_iif_ipv6, firewall_simple_inet_ipv6, firewall_simple_oif_ipv6, firewall_simple_onet_ipv6.

A new rc.d script rc.d/rtsold has been added. This handles rtsold(8) daemon.

A new rc.d script rc.d/static_arp has been added. This allows the administrator to statically define mappings of MAC address to IPv4 at boot time. See also the rc.conf(5) manual page for more details.

The rc.d/tmp script now uses a unique directory name prefixed with /tmp/.diskless instead of /tmp/.diskless itself. This fixes an issue when /tmp/.diskless exists before the script runs.

A new rc.d script rc.d/ubthidhci has been added. This small script calls usbconfig(8) to change a USB Bluetooth controller from HID mode to HCI mode.

The rc.conf(5) now supports a firewall_coscripts variable. This should contain a list of commands which should be executed after firewall starts or stops.

The rc.conf(5) now supports configuring vlan(4) interfaces as child devices similar to wlan(4) interfaces. vlan(4) interfaces are listed via a new vlans_IF variable. If a VLAN interface is a number, then that number is treated as the VLAN tag for the interface and the interface will be named IF.tag. Otherwise, the VLAN tag must be provided via a VLAN parameter in a create_args_IF variable.