1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 Items affecting the ports and packages system can be found in
10 /usr/ports/UPDATING. Please read that file before running portupgrade.
12 NOTE TO PEOPLE WHO THINK THAT FreeBSD 10.x IS SLOW:
13 FreeBSD 10.x has many debugging features turned on, in both the kernel
14 and userland. These features attempt to detect incorrect use of
15 system primitives, and encourage loud failure through extra sanity
16 checking and fail stop semantics. They also substantially impact
17 system performance. If you want to do performance measurement,
18 benchmarking, and optimization, you'll want to turn them off. This
19 includes various WITNESS- related kernel options, INVARIANTS, malloc
20 debugging flags in userland, and various verbose features in the
21 kernel. Many developers choose to disable these features on build
22 machines to maximize performance. (To completely disable malloc
23 debugging, define MALLOC_PRODUCTION in /etc/make.conf, or to merely
24 disable the most expensive debugging functionality run
25 "ln -s 'abort:false,junk:false' /etc/malloc.conf".)
28 The pf(4) packet filter ABI has been changed. pfctl(8) and
29 snmp_pf module need to be recompiled to work with new kernel.
32 A new ZFS feature flag "com.delphix:empty_bpobj" has been merged
33 to -HEAD. Pools that have empty_bpobj in active state can not be
34 imported read-write with ZFS implementations that do not support
35 this feature. For more information read the zpool-features(5)
39 The sparc64 ZFS loader has been changed to no longer try to auto-
40 detect ZFS providers based on diskN aliases but now requires these
41 to be explicitly listed in the OFW boot-device environment variable.
44 The OpenSSL has been upgraded to 1.0.1c. Any binaries requiring
45 libcrypto.so.6 or libssl.so.6 must be recompiled. Also, there are
46 configuration changes. Make sure to merge /etc/ssl/openssl.cnf.
49 The following sysctls and tunables have been renamed for consistency
51 kern.cam.da.da_send_ordered -> kern.cam.da.send_ordered
52 kern.cam.ada.ada_send_ordered -> kern.cam.ada.send_ordered
55 The sort utility has been replaced with BSD sort. For now, GNU sort
56 is also available as "gnusort" or the default can be set back to
57 GNU sort by setting WITH_GNU_SORT. In this case, BSD sort will be
58 installed as "bsdsort".
61 A new version of ZFS (pool version 5000) has been merged to -HEAD.
62 Starting with this version the old system of ZFS pool versioning
63 is superseded by "feature flags". This concept enables forward
64 compatibility against certain future changes in functionality of ZFS
65 pools. The first read-only compatible "feature flag" for ZFS pools
66 is named "com.delphix:async_destroy". For more information
67 read the new zpool-features(5) manual page.
68 Please refer to the "ZFS notes" section of this file for information
69 on upgrading boot ZFS pools.
72 The malloc(3) implementation embedded in libc now uses sources imported
73 as contrib/jemalloc. The most disruptive API change is to
74 /etc/malloc.conf. If your system has an old-style /etc/malloc.conf,
75 delete it prior to installworld, and optionally re-create it using the
76 new format after rebooting. See malloc.conf(5) for details
77 (specifically the TUNING section and the "opt.*" entries in the MALLCTL
81 Big-endian MIPS TARGET_ARCH values no longer end in "eb". mips64eb
82 is now spelled mips64. mipsn32eb is now spelled mipsn32. mipseb is
83 now spelled mips. This is to aid compatibility with third-party
84 software that expects this naming scheme in uname(3). Little-endian
85 settings are unchanged.
88 Disable by default the option VFS_ALLOW_NONMPSAFE for all supported
92 Now unix domain sockets behave "as expected" on nullfs(5). Previously
93 nullfs(5) did not pass through all behaviours to the underlying layer,
94 as a result if we bound to a socket on the lower layer we could connect
95 only to the lower path; if we bound to the upper layer we could connect
96 only to the upper path. The new behavior is one can connect to both the
97 lower and the upper paths regardless what layer path one binds to.
100 The getifaddrs upgrade path broken with 20111215 has been restored.
101 If you have upgraded in between 20111215 and 20120209 you need to
102 recompile libc again with your kernel. You still need to recompile
103 world to be able to configure CARP but this restriction already
107 The set_rcvar() function has been removed from /etc/rc.subr. All
108 base and ports rc.d scripts have been updated, so if you have a
109 port installed with a script in /usr/local/etc/rc.d you can either
110 hand-edit the rcvar= line, or reinstall the port.
112 An easy way to handle the mass-update of /etc/rc.d:
113 rm /etc/rc.d/* && mergemaster -i
116 panic(9) now stops other CPUs in the SMP systems, disables interrupts
117 on the current CPU and prevents other threads from running.
118 This behavior can be reverted using the kern.stop_scheduler_on_panic
120 The new behavior can be incompatible with kern.sync_on_panic.
123 The carp(4) facility has been changed significantly. Configuration
124 of the CARP protocol via ifconfig(8) has changed, as well as format
125 of CARP events submitted to devd(8) has changed. See manual pages
126 for more information. The arpbalance feature of carp(4) is currently
127 not supported anymore.
129 Size of struct in_aliasreq, struct in6_aliasreq has changed. User
130 utilities using SIOCAIFADDR, SIOCAIFADDR_IN6, e.g. ifconfig(8),
131 need to be recompiled.
134 The acpi_wmi(4) status device /dev/wmistat has been renamed to
138 The option VFS_ALLOW_NONMPSAFE option has been added in order to
139 explicitely support non-MPSAFE filesystems.
140 It is on by default for all supported platform at this present
144 The broken amd(4) driver has been replaced with esp(4) in the amd64,
145 i386 and pc98 GENERIC kernel configuration files.
148 sysinstall has been removed
151 The stable/9 branch created in subversion. This corresponds to the
152 RELENG_9 branch in CVS.
155 This commit modifies vfs_register() so that it uses a hash
156 calculation to set vfc_typenum, which is enabled by default.
157 The first time a system is booted after this change, the
158 vfc_typenum values will change for all file systems. The
159 main effect of this is a change to the NFS server file handles
160 for file systems that use vfc_typenum in their fsid, such as ZFS.
161 It will, however, prevent vfc_typenum from changing when file
162 systems are loaded in a different order for subsequent reboots.
163 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
164 until you are ready to remount all NFS clients after a reboot.
167 Bump the shared library version numbers for libraries that
168 do not use symbol versioning, have changed the ABI compared
169 to stable/8 and which shared library version was not bumped.
170 Done as part of 9.0-RELEASE cycle.
173 During the merge of Capsicum features, the fget(9) KPI was modified.
174 This may require the rebuilding of out-of-tree device drivers --
175 issues have been reported specifically with the nVidia device driver.
176 __FreeBSD_version is bumped to 900041.
178 Also, there is a period between 20110811 and 20110814 where the
179 special devices /dev/{stdin,stdout,stderr} did not work correctly.
180 Building world from a kernel during that window may not work.
183 The packet filter (pf) code has been updated to OpenBSD 4.5.
184 You need to update userland tools to be in sync with kernel.
185 This update breaks backward compatibility with earlier pfsync(4)
186 versions. Care must be taken when updating redundant firewall setups.
189 The following sysctls and tunables are retired on x86 platforms:
191 machdep.hlt_logical_cpus
192 The following sysctl is retired:
193 machdep.hyperthreading_allowed
194 The sysctls were supposed to provide a way to dynamically offline and
195 online selected CPUs on x86 platforms, but the implementation has not
196 been reliable especially with SCHED_ULE scheduler.
197 machdep.hyperthreading_allowed tunable is still available to ignore
198 hyperthreading CPUs at OS level.
199 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
200 where X is an APIC ID of a CPU. Be advised, though, that disabling
201 CPUs in non-uniform fashion will result in non-uniform topology and
202 may lead to sub-optimal system performance with SCHED_ULE, which is
206 cpumask_t type is retired and cpuset_t is used in order to describe
210 Changes to ifconfig(8) for dynamic address family detection mandate
211 that you are running a kernel of 20110525 or later. Make sure to
212 follow the update procedure to boot a new kernel before installing
216 Support for sun4v architecture is officially dropped
219 Several KPI breaking changes have been committed to the mii(4) layer,
220 the PHY drivers and consequently some Ethernet drivers using mii(4).
221 This means that miibus.ko and the modules of the affected Ethernet
222 drivers need to be recompiled.
224 Note to kernel developers: Given that the OUI bit reversion problem
225 was fixed as part of these changes all mii(4) commits related to OUIs,
226 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
227 handling, no longer can be merged verbatim to stable/8 and previous
231 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
232 into their kernel configurations along with 'device pci'.
235 The default NFS client is now the new NFS client, so fstype "newnfs"
236 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
237 Although mounts via fstype "nfs" will usually work without userland
238 changes, it is recommended that the mount(8) and mount_nfs(8)
239 commands be rebuilt from sources and that a link to mount_nfs called
240 mount_oldnfs be created. The new client is compiled into the
241 kernel with "options NFSCL" and this is needed for diskless root
242 file systems. The GENERIC kernel configs have been changed to use
243 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
244 To use the regular/old client, you can "mount -t oldnfs ...". For
245 a diskless root file system, you must also include a line like:
247 vfs.root.mountfrom="oldnfs:"
249 in the boot/loader.conf on the root fs on the NFS server to make
250 a diskless root fs use the old client.
253 The GENERIC kernels for all architectures now default to the new
254 CAM-based ATA stack. It means that all legacy ATA drivers were
255 removed and replaced by respective CAM drivers. If you are using
256 ATA device names in /etc/fstab or other places, make sure to update
257 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
258 where 'Y's are the sequential numbers starting from zero for each type
259 in order of detection, unless configured otherwise with tunables,
260 see cam(4)). There will be symbolic links created in /dev/ to map
261 old adX devices to the respective adaY. They should provide basic
262 compatibility for file systems mounting in most cases, but they do
263 not support old user-level APIs and do not have respective providers
264 in GEOM. Consider using updated management tools with new device names.
266 It is possible to load devices ahci, ata, siis and mvs as modules,
267 but option ATA_CAM should remain in kernel configuration to make ata
268 module work as CAM driver supporting legacy ATA controllers. Device ata
269 still can be used in modular fashion (atacore + ...). Modules atadisk
270 and atapi* are not used and won't affect operation in ATA_CAM mode.
271 Note that to use CAM-based ATA kernel should include CAM devices
272 scbus, pass, da (or explicitly ada), cd and optionally others. All of
273 them are parts of the cam module.
275 ataraid(4) functionality is now supported by the RAID GEOM class.
276 To use it you can load geom_raid kernel module and use graid(8) tool
277 for management. Instead of /dev/arX device names, use /dev/raid/rX.
279 No kernel config options or code have been removed, so if a problem
280 arises, please report it and optionally revert to the old ATA stack.
281 In order to do it you can remove from the kernel config:
286 , and instead add back:
287 device atadisk # ATA disk drives
288 device ataraid # ATA RAID drives
289 device atapicd # ATAPI CDROM drives
290 device atapifd # ATAPI floppy drives
291 device atapist # ATAPI tape drives
294 The default NFS server has been changed to the new server, which
295 was referred to as the experimental server. If you need to switch
296 back to the old NFS server, you must now put the "-o" option on
297 both the mountd and nfsd commands. This can be done using the
298 mountd_flags and nfs_server_flags rc.conf variables until an
299 update to the rc scripts is committed, which is coming soon.
302 The GNU Objective-C runtime library (libobjc), and other Objective-C
303 related components have been removed from the base system. If you
304 require an Objective-C library, please use one of the available ports.
307 ath(4) has been split into bus- and device- modules. if_ath contains
308 the HAL, the TX rate control and the network device code. if_ath_pci
309 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
310 contains the AHB glue. Users need to load both if_ath_pci and if_ath
311 in order to use ath on everything else.
313 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
314 need to load if_ath and if_ath_pci for ath(4) operation.
317 As part of the replacement of sysinstall, the process of building
318 release media has changed significantly. For details, please re-read
319 release(7), which has been updated to reflect the new build process.
322 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
323 is the last available version under GPLv2. It brings a number of new
324 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
325 4.1 and SSE 4.2), better support for powerpc64, a number of new
326 directives, and lots of other small improvements. See the ChangeLog
327 file in contrib/binutils for the full details.
330 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
331 compliant, and will now use half of hash for authentication.
332 This will break interoperability with all stacks (including all
333 actual FreeBSD versions) who implement
334 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
336 The only workaround with such peers is to use another HMAC
337 algorithm for IPsec ("phase 2") authentication.
340 Remove the uio_yield prototype and symbol. This function has
341 been misnamed since it was introduced and should not be
342 globally exposed with this name. The equivalent functionality
343 is now available using kern_yield(curthread->td_user_pri).
344 The function remains undocumented.
347 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
348 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
349 for scalar sysctls is defined but disabled. Code that needs
350 UQUAD to pass the type checking that must compile on older
351 systems where the define is not present can check against
352 __FreeBSD_version >= 900030.
354 The system dialog(1) has been replaced with a new version previously
355 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
356 with the previous version, but the libdialog associated with it has
357 a largely incompatible API. As such, the original version of libdialog
358 will be kept temporarily as libodialog, until its base system consumers
359 are replaced or updated. Bump __FreeBSD_version to 900030.
362 If you are trying to run make universe on a -stable system, and you get
363 the following warning:
364 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
365 unknown. config(8) likely too old."
366 or something similar to it, then you must upgrade your -stable system
367 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
368 stable/8) or build the config from the latest stable/8 branch and
369 install it on your system.
371 Prior to this date, building a current universe on 8-stable system from
372 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
373 error in the first kernel build phase. A new config on those old
374 systems will fix that problem for older versions of -current.
377 The TCP stack has been modified to allow Khelp modules to interact with
378 it via helper hook points and store per-connection data in the TCP
379 control block. Bump __FreeBSD_version to 900029. User space tools that
380 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
384 Generic IEEE 802.3 annex 31B full duplex flow control support has been
385 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
386 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
387 to take advantage of it instead of using custom implementations. This
388 means that these drivers now no longer unconditionally advertise
389 support for flow control but only do so if flow control is a selected
390 media option. This was implemented in the generic support that way in
391 order to allow flow control to be switched on and off via ifconfig(8)
392 with the PHY specific default to typically off in order to protect
393 from unwanted effects. Consequently, if you used flow control with
394 one of the above mentioned drivers you now need to explicitly enable
396 ifconfig bge0 media auto mediaopt flowcontrol
398 Along with the above mentioned changes generic support for setting
399 1000baseT master mode also has been added and brgphy(4), ciphy(4),
400 e1000phy(4) as well as ip1000phy(4) have been converted to take
401 advantage of it. This means that these drivers now no longer take the
402 link0 parameter for selecting master mode but the master media option
403 has to be used instead, for example like in the following:
404 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
406 Selection of master mode now is also available with all other PHY
407 drivers supporting 1000baseT.
410 The TCP stack has received a significant update to add support for
411 modularised congestion control and generally improve the clarity of
412 congestion control decisions. Bump __FreeBSD_version to 900025. User
413 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
414 sockstat) need to be recompiled.
417 The man(1) utility has been replaced by a new version that no longer
418 uses /etc/manpath.config. Please consult man.conf(5) for how to
419 migrate local entries to the new format.
422 The copyright strings printed by login(1) and sshd(8) at the time of a
423 new connection have been removed to follow other operating systems and
427 A workaround for a fixed ld bug has been removed in kernel code,
428 so make sure that your system ld is built from sources after
429 revision 210245 from 2010-07-19 (r211583 if building head kernel
430 on stable/8, r211584 for stable/7; both from 2010-08-21).
431 A symptom of incorrect ld version is different addresses for
432 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
435 The $ipv6_prefer variable in rc.conf(5) has been split into
436 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
438 The $ip6addrctl_policy is a variable to choose a pre-defined
439 address selection policy set by ip6addrctl(8). A value
440 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
443 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
444 flag (see an entry of 20090926) is set on an interface with no
445 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
446 security reason. If you want IPv6 link-local address on all
447 interfaces by default, set this to "YES".
449 The old ipv6_prefer="YES" is equivalent to
450 ipv6_activate_all_interfaces="YES" and
451 ip6addrctl_policy="ipv6_prefer".
454 DTrace has grown support for userland tracing. Due to this, DTrace is
455 now i386 and amd64 only.
456 dtruss(1) is now installed by default on those systems and a new
457 kernel module is needed for userland tracing: fasttrap.
458 No changes to your kernel config file are necessary to enable
459 userland tracing, but you might consider adding 'STRIP=' and
460 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
461 to have informative userland stack traces in DTrace (ustack).
464 The acpi_aiboost(4) driver has been removed in favor of the new
465 aibs(4) driver. You should update your kernel configuration file.
468 BSD grep has been imported to the base system and it is built by
469 default. It is completely BSD licensed, highly GNU-compatible, uses
470 less memory than its GNU counterpart and has a small codebase.
471 However, it is slower than its GNU counterpart, which is mostly
472 noticeable for larger searches, for smaller ones it is measurable
473 but not significant. The reason is complex, the most important factor
474 is that we lack a modern and efficient regex library and GNU
475 overcomes this by optimizing the searches internally. Future work
476 on improving the regex performance is planned, for the meantime,
477 users that need better performance, can build GNU grep instead by
478 setting the WITH_GNU_GREP knob.
481 Due to the import of powerpc64 support, all existing powerpc kernel
482 configuration files must be updated with a machine directive like this:
483 machine powerpc powerpc
485 In addition, an updated config(8) is required to build powerpc kernels
489 A new version of ZFS (version 15) has been merged to -HEAD.
490 This version uses a python library for the following subcommands:
491 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
492 For full functionality of these commands the following port must
493 be installed: sysutils/py-zfs
496 'vm_page's are now hashed by physical address to an array of mutexes.
497 Currently this is only used to serialize access to hold_count. Over
498 time the page queue mutex will be peeled away. This changes the size
499 of pmap on every architecture. And requires all callers of vm_page_hold
500 and vm_page_unhold to be updated.
503 WITH_CTF can now be specified in src.conf (not recommended, there
504 are some problems with static executables), make.conf (would also
505 affect ports which do not use GNU make and do not override the
506 compile targets) or in the kernel config (via "makeoptions
508 When WITH_CTF was specified there before this was silently ignored,
509 so make sure that WITH_CTF is not used in places which could lead
510 to unwanted behavior.
513 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
514 to allow 32-bit compatibility on non-x86 platforms. All kernel
515 configurations on amd64 and ia64 platforms using these options must
516 be modified accordingly.
519 The utmp user accounting database has been replaced with utmpx,
520 the user accounting interface standardized by POSIX.
521 Unfortunately the semantics of utmp and utmpx don't match,
522 making it practically impossible to support both interfaces.
523 The user accounting database is used by tools like finger(1),
524 last(1), talk(1), w(1) and ac(8).
526 All applications in the base system use utmpx. This means only
527 local binaries (e.g. from the ports tree) may still use these
528 utmp database files. These applications must be rebuilt to make
531 After the system has been upgraded, it is safe to remove the old
532 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
533 assuming their contents is of no importance anymore. Old wtmp
534 databases can only be used by last(1) and ac(8) after they have
535 been converted to the new format using wtmpcvt(1).
538 Introduce the kernel thread "deadlock resolver" (which can be enabled
539 via the DEADLKRES option, see NOTES for more details) and the
540 sleepq_type() function for sleepqueues.
543 The rc.firewall and rc.firewall6 were unified, and
544 rc.firewall6 and rc.d/ip6fw were removed.
545 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
546 variables are obsoleted. Instead, the following new rc
547 variables are added to rc.d/ipfw:
549 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
550 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
551 firewall_simple_onet_ipv6, firewall_trusted_ipv6
553 The meanings correspond to the relevant IPv4 variables.
559 The default terminal emulation for syscons(4) has been changed
560 from cons25 to xterm on all platforms except pc98. This means
561 that the /etc/ttys file needs to be updated to ensure correct
562 operation of applications on the console.
564 The terminal emulation style can be toggled per window by using
565 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
566 options can be used to change the compile-time default back to
569 To prevent graphical artifacts, make sure the TERM environment
570 variable is set to match the terminal emulation that is being
571 performed by syscons(4).
574 The layout of the structure ieee80211req_scan_result has changed.
575 Applications that require wireless scan results (e.g. ifconfig(8))
576 from net80211 need to be recompiled.
578 Applications such as wpa_supplicant(8) may require a full world
579 build without using NO_CLEAN in order to get synchronized with the
583 The iwn(4) driver has been updated to support the 5000 and 5150 series.
584 There's one kernel module for each firmware. Adding "device iwnfw"
585 to the kernel configuration file means including all three firmware
586 images inside the kernel. If you want to include just the one for
587 your wireless card, use the devices iwn4965fw, iwn5000fw or
591 The rc.d/network_ipv6, IPv6 configuration script has been integrated
592 into rc.d/netif. The changes are the following:
594 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
595 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
596 Note that both variables need the "inet6" keyword at the head.
598 Do not set $ipv6_network_interfaces manually if you do not
599 understand what you are doing. It is not needed in most cases.
601 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
604 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
605 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
607 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
608 all configured interfaces have "inet6 accept_rtadv" in the
609 $ifconfig_IF_ipv6. These are for backward compatibility.
611 3. A new variable $ipv6_prefer has been added. If NO, IPv6
612 functionality of interfaces with no corresponding
613 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
614 and the default address selection policy of ip6addrctl(8)
615 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
616 Note that if you want to configure IPv6 functionality on the
617 disabled interfaces after boot, first you need to clear the flag by
618 using ifconfig(8) like:
620 ifconfig em0 inet6 -ifdisabled
622 If YES, the default address selection policy is set as
625 The default value of $ipv6_prefer is NO.
627 4. If your system need to receive Router Advertisement messages,
628 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
629 scripts automatically invoke rtsol(8) when the interface becomes
630 UP. The Router Advertisement messages are used for SLAAC
631 (State-Less Address AutoConfiguration).
634 802.11s D3.03 support was committed. This is incompatible with the
635 previous code, which was based on D3.0.
638 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
639 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
640 control whether accepting Router Advertisement messages or not.
641 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
642 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
643 The ifconfig(8) utility now supports these flags.
646 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
647 mount(8) and -a option for df(1) to see them.
650 The old tunable hw.bus.devctl_disable has been superseded by
651 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
652 replaced by hw.bus.devctl_queue=0. The default for this new tunable
656 Remove the option STOP_NMI. The default action is now to use NMI only
657 for KDB via the newly introduced function stop_cpus_hard() and
658 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
661 The stable/8 branch created in subversion. This corresponds to the
662 RELENG_8 branch in CVS.
665 Bump the shared library version numbers for all libraries that do not
666 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
667 __FreeBSD_version to 800105.
670 Due to changes in the implementation of virtual network stack support,
671 all network-related kernel modules must be recompiled. As this change
672 breaks the ABI, bump __FreeBSD_version to 800104.
675 The TOE interface to the TCP syncache has been modified to remove
676 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
677 The cxgb driver is the only TOE consumer affected by this change, and
678 needs to be recompiled along with the kernel. As this change breaks
679 the ABI, bump __FreeBSD_version to 800103.
682 Padding has been added to struct tcpcb, sackhint and tcpstat in
683 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
684 maintaining the ABI. However, this change breaks the ABI, so bump
685 __FreeBSD_version to 800102. User space tools that rely on the size of
686 any of these structs (e.g. sockstat) need to be recompiled.
689 The NFS_LEGACYRPC option has been removed along with the old kernel
690 RPC implementation that this option selected. Kernel configurations
691 may need to be adjusted.
694 The network interface device nodes at /dev/net/<interface> have been
695 removed. All ioctl operations can be performed the normal way using
696 routing sockets. The kqueue functionality can generally be replaced
697 with routing sockets.
700 The documentation from the FreeBSD Documentation Project (Handbook,
701 FAQ, etc.) is now installed via packages by sysinstall(8) and under
702 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
705 The ABI of various structures related to the SYSV IPC API have been
706 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
707 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
711 Layout of struct vnet has changed as routing related variables were
712 moved to their own Vimage module. Modules need to be recompiled. Bump
713 __FreeBSD_version to 800099.
716 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
717 respectively. As long as no more than 16 groups per process are used,
718 no changes should be visible. When more than 16 groups are used, old
719 binaries may fail if they call getgroups() or getgrouplist() with
720 statically sized storage. Recompiling will work around this, but
721 applications should be modified to use dynamically allocated storage
722 for group arrays as POSIX.1-2008 does not cap an implementation's
723 number of supported groups at NGROUPS_MAX+1 as previous versions did.
725 NFS and portalfs mounts may also be affected as the list of groups is
726 truncated to 16. Users of NFS who use more than 16 groups, should
727 take care that negative group permissions are not used on the exported
728 file systems as they will not be reliable unless a GSSAPI based
729 authentication method is used.
732 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
733 option compiles in the support for adaptive spinning for lockmgrs
734 which want to enable it. The lockinit() function now accepts the flag
735 LK_ADAPTIVE in order to make the lock object subject to adaptive
736 spinning when both held in write and read mode.
739 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
740 changed. User applications that use this ioctl need to be rebuilt.
743 The layout of struct thread has changed. Kernel and modules need to
747 The layout of structs ifnet, domain, protosw and vnet_net has changed.
748 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
751 window(1) has been removed from the base system. It can now be
752 installed from ports. The port is called misc/window.
755 The way we are storing and accessing `routing table' entries has
756 changed. Programs reading the FIB, like netstat, need to be
760 A new netisr implementation has been added for FreeBSD 8. Network
761 file system modules, such as igmp, ipdivert, and others, should be
763 Bump __FreeBSD_version to 800096.
766 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
770 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
771 Bump __FreeBSD_version to 800094.
774 Add mnt_xflag field to 'struct mount'. File system modules need to be
776 Bump __FreeBSD_version to 800093.
779 The compiling option ADAPTIVE_SX has been retired while it has been
780 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
781 The KPI for sx_init_flags() changes as accepting flags:
782 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
783 been introduced in order to handle the reversed logic.
784 Bump __FreeBSD_version to 800092.
787 Add support for hierarchical jails. Remove global securelevel.
788 Bump __FreeBSD_version to 800091.
791 The layout of struct vnet_net has changed, therefore modules
793 Bump __FreeBSD_version to 800090.
796 The newly imported zic(8) produces a new format in the output. Please
797 run tzsetup(8) to install the newly created data to /etc/localtime.
800 The sysctl tree for the usb stack has renamed from hw.usb2.* to
801 hw.usb.* and is now consistent again with previous releases.
804 802.11 monitor mode support was revised and driver api's were changed.
805 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
806 of DLT_IEEE802_11. No user-visible data structures were changed but
807 applications that use DLT_IEEE802_11 may require changes.
808 Bump __FreeBSD_version to 800088.
811 The layout of the following structs has changed: sysctl_oid,
812 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
813 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
814 panics may be experienced. World rebuild is required for
815 correctly checking networking state from userland.
816 Bump __FreeBSD_version to 800085.
819 MLDv2 and Source-Specific Multicast (SSM) have been merged
820 to the IPv6 stack. VIMAGE hooks are in but not yet used.
821 The implementation of SSM within FreeBSD's IPv6 stack closely
822 follows the IPv4 implementation.
824 For kernel developers:
826 * The most important changes are that the ip6_output() and
827 ip6_input() paths no longer take the IN6_MULTI_LOCK,
828 and this lock has been downgraded to a non-recursive mutex.
830 * As with the changes to the IPv4 stack to support SSM, filtering
831 of inbound multicast traffic must now be performed by transport
832 protocols within the IPv6 stack. This does not apply to TCP and
833 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
835 * The KPIs used by IPv6 multicast are similar to those used by
836 the IPv4 stack, with the following differences:
837 * im6o_mc_filter() is analogous to imo_multicast_filter().
838 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
839 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
840 * IN6_LOOKUP_MULTI() has been deprecated and removed.
841 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
842 for MLDv1 have an additional 'timer' argument which is used to
843 jitter the initial membership report for the solicited-node
844 multicast membership on-link.
845 * This is not strictly needed for MLDv2, which already jitters
846 its report transmissions. However, the 'timer' argument is
847 preserved in case MLDv1 is active on the interface.
849 * The KAME linked-list based IPv6 membership implementation has
850 been refactored to use a vector similar to that used by the IPv4
852 Code which maintains a list of its own multicast memberships
853 internally, e.g. carp, has been updated to reflect the new
856 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
857 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
858 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
859 implementation constraint which needs to be addressed in HEAD.
861 For application developers:
863 * The changes are broadly similar to those made for the IPv4
866 * The use of IPv4 and IPv6 multicast socket options on the same
867 socket, using mapped addresses, HAS NOT been tested or supported.
869 * There are a number of issues with the implementation of various
870 IPv6 multicast APIs which need to be resolved in the API surface
871 before the implementation is fully compatible with KAME userland
872 use, and these are mostly to do with interface index treatment.
874 * The literature available discusses the use of either the delta / ASM
875 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
876 using setsourcefilter(3)/getsourcefilter(3). For more information
877 please refer to RFC 3768, 'Socket Interface Extensions for
878 Multicast Source Filters'.
880 * Applications which use the published RFC 3678 APIs should be fine.
882 For systems administrators:
884 * The mtest(8) utility has been refactored to support IPv6, in
885 addition to IPv4. Interface addresses are no longer accepted
886 as arguments, their names must be used instead. The utility
887 will map the interface name to its first IPv4 address as
888 returned by getifaddrs(3).
890 * The ifmcstat(8) utility has also been updated to print the MLDv2
891 endpoint state and source filter lists via sysctl(3).
893 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
894 loopback of IPv6 multicast datagrams by default; it defaults to 1
895 to preserve the existing behaviour. Disabling multicast loopback is
896 recommended for optimal system performance.
898 * The IPv6 MROUTING code has been changed to examine this sysctl
899 instead of attempting to perform a group lookup before looping
900 back forwarded datagrams.
902 Bump __FreeBSD_version to 800084.
905 Implement low-level Bluetooth HCI API.
906 Bump __FreeBSD_version to 800083.
909 The layout of struct malloc_type, used by modules to register new
910 memory allocation types, has changed. Most modules will need to
911 be rebuilt or panics may be experienced.
912 Bump __FreeBSD_version to 800081.
915 Anticipate overflowing inp_flags - add inp_flags2.
916 This changes most offsets in inpcb, so checking v4 connection
917 state will require a world rebuild.
918 Bump __FreeBSD_version to 800080.
921 Add an llentry to struct route and struct route_in6. Modules
922 embedding a struct route will need to be recompiled.
923 Bump __FreeBSD_version to 800079.
926 The size of rt_metrics_lite and by extension rtentry has changed.
927 Networking administration apps will need to be recompiled.
928 The route command now supports show as an alias for get, weighting
929 of routes, sticky and nostick flags to alter the behavior of stateful
931 Bump __FreeBSD_version to 800078.
934 Do not use Giant for kbdmux(4) locking. This is wrong and
935 apparently causing more problems than it solves. This will
936 re-open the issue where interrupt handlers may race with
937 kbdmux(4) in polling mode. Typical symptoms include (but
938 not limited to) duplicated and/or missing characters when
939 low level console functions (such as gets) are used while
940 interrupts are enabled (for example geli password prompt,
941 mountroot prompt etc.). Disabling kbdmux(4) may help.
944 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
945 kernel modules referencing any of the above need to be recompiled.
946 Bump __FreeBSD_version to 800075.
949 GEOM_PART has become the default partition slicer for storage devices,
950 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
951 introduces some changes:
953 MSDOS/EBR: the devices created from MSDOS extended partition entries
954 (EBR) can be named differently than with GEOM_MBR and are now symlinks
955 to devices with offset-based names. fstabs may need to be modified.
957 BSD: the "geometry does not match label" warning is harmless in most
958 cases but it points to problems in file system misalignment with
959 disk geometry. The "c" partition is now implicit, covers the whole
960 top-level drive and cannot be (mis)used by users.
962 General: Kernel dumps are now not allowed to be written to devices
963 whose partition types indicate they are meant to be used for file
964 systems (or, in case of MSDOS partitions, as something else than
967 Most of these changes date approximately from 200812.
970 The uscanner(4) driver has been removed from the kernel. This follows
971 Linux removing theirs in 2.6 and making libusb the default interface
975 The multicast forwarding code has been cleaned up. netstat(1)
976 only relies on KVM now for printing bandwidth upcall meters.
977 The IPv4 and IPv6 modules are split into ip_mroute_mod and
978 ip6_mroute_mod respectively. The config(5) options for statically
979 compiling this code remain the same, i.e. 'options MROUTING'.
982 Support for the IFF_NEEDSGIANT network interface flag has been
983 removed, which means that non-MPSAFE network device drivers are no
984 longer supported. In particular, if_ar, if_sr, and network device
985 drivers from the old (legacy) USB stack can no longer be built or
989 POSIX.1 Native Language Support (NLS) has been enabled in libc and
990 a bunch of new language catalog files have also been added.
991 This means that some common libc messages are now localized and
992 they depend on the LC_MESSAGES environmental variable.
995 The k8temp(4) driver has been renamed to amdtemp(4) since
996 support for Family 10 and Family 11 CPU families was added.
999 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1000 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1002 For kernel developers, the most important changes are that the
1003 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1004 and this lock has been downgraded to a non-recursive mutex.
1006 Transport protocols (UDP, Raw IP) are now responsible for filtering
1007 inbound multicast traffic according to group membership and source
1008 filters. The imo_multicast_filter() KPI exists for this purpose.
1009 Transports which do not use multicast (SCTP, TCP) already reject
1010 multicast by default. Forwarding and receive performance may improve
1011 as a mutex acquisition is no longer needed in the ip_input()
1012 low-level input path. in_addmulti() and in_delmulti() are shimmed
1013 to new KPIs which exist to support SSM in-kernel.
1015 For application developers, it is recommended that loopback of
1016 multicast datagrams be disabled for best performance, as this
1017 will still cause the lock to be taken for each looped-back
1018 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1019 be tuned to 0 to disable loopback by default; it defaults to 1
1020 to preserve the existing behaviour.
1022 For systems administrators, to obtain best performance with
1023 multicast reception and multiple groups, it is always recommended
1024 that a card with a suitably precise hash filter is used. Hash
1025 collisions will still result in the lock being taken within the
1026 transport protocol input path to check group membership.
1028 If deploying FreeBSD in an environment with IGMP snooping switches,
1029 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1030 enabled; this forces 224.0.0.0/24 group membership to be announced
1033 The size of 'struct igmpstat' has changed; netstat needs to be
1034 recompiled to reflect this.
1035 Bump __FreeBSD_version to 800070.
1038 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1039 updated to use it. This requires a buildworld/installworld in order to
1040 update the library and dependencies (usbconfig, etc). Its advisable to
1041 rebuild all ports which uses libusb. More specific directions are given
1042 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1043 libusb are no longer required and can be removed.
1046 A workaround is committed to allow the creation of System V shared
1047 memory segment of size > 2 GB on the 64-bit architectures.
1048 Due to a limitation of the existing ABI, the shm_segsz member
1049 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1050 wrong for large segments. Note that limits must be explicitly
1051 raised to allow such segments to be created.
1054 The layout of struct ifnet has changed, requiring a rebuild of all
1055 network device driver modules.
1058 The /dev handling for the new USB stack has changed, a
1059 buildworld/installworld is required for libusb20.
1062 The new USB2 stack has now been permanently moved in and all kernel and
1063 module names reverted to their previous values (eg, usb, ehci, ohci,
1064 ums, ...). The old usb stack can be compiled in by prefixing the name
1065 with the letter 'o', the old usb modules have been removed.
1066 Updating entry 20090216 for xorg and 20090215 for libmap may still
1070 The rc.conf(5) option if_up_delay has been renamed to
1071 defaultroute_delay to better reflect its purpose. If you have
1072 customized this setting in /etc/rc.conf you need to update it to
1076 xorg 7.4 wants to configure its input devices via hald which does not
1077 yet work with USB2. If the keyboard/mouse does not work in xorg then
1079 Option "AllowEmptyInput" "off"
1080 to your ServerLayout section. This will cause X to use the configured
1081 kbd and mouse sections from your xorg.conf.
1084 The GENERIC kernels for all architectures now default to the new USB2
1085 stack. No kernel config options or code have been removed so if a
1086 problem arises please report it and optionally revert to the old USB
1087 stack. If you are loading USB kernel modules or have a custom kernel
1088 that includes GENERIC then ensure that usb names are also changed over,
1089 eg uftdi -> usb2_serial_ftdi.
1091 Older programs linked against the ports libusb 0.1 need to be
1092 redirected to the new stack's libusb20. /etc/libmap.conf can
1094 # Map old usb library to new one for usb2 stack
1095 libusb-0.1.so.8 libusb20.so.1
1098 All USB ethernet devices now attach as interfaces under the name ueN
1099 (eg. ue0). This is to provide a predictable name as vendors often
1100 change usb chipsets in a product without notice.
1103 The ichsmb(4) driver has been changed to require SMBus slave
1104 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1105 All of the other SMBus controller drivers require left-justified
1106 slave addresses, so this change makes all the drivers provide the
1110 INET6 statistics (struct ip6stat) was updated.
1111 netstat(1) needs to be recompiled.
1114 NTFS has been removed from GENERIC kernel on amd64 to match
1115 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1116 will load ntfs.ko module automatically when NTFS support is
1117 actually needed, unless ntfs.ko is not installed or security
1118 level prohibits loading kernel modules. If either is the case,
1119 "options NTFS" has to be added into kernel config.
1122 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1123 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1124 800061. User space tools that rely on the size of struct tcpcb in
1125 tcp_var.h (e.g. sockstat) need to be recompiled.
1128 ng_tty(4) module updated to match the new TTY subsystem.
1129 Due to API change, user-level applications must be updated.
1130 New API support added to mpd5 CVS and expected to be present
1131 in next mpd5.3 release.
1134 With __FreeBSD_version 800060 the makefs tool is part of
1135 the base system (it was a port).
1138 The afdata and ifnet locks have been changed from mutexes to
1139 rwlocks, network modules will need to be re-compiled.
1142 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1143 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1144 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1145 architecture and 40 bytes on 64-bit architecture. The userland
1146 applications "arp" and "ndp" have been updated accordingly.
1147 The output from "netstat -r" shows only routing entries and
1148 none of the L2 information.
1151 __FreeBSD_version 800057 marks the switchover from the
1152 binary ath hal to source code. Users must add the line:
1154 options AH_SUPPORT_AR5416
1156 to their kernel config files when specifying:
1160 The ath_hal module no longer exists; the code is now compiled
1161 together with the driver in the ath module. It is now
1162 possible to tailor chip support (i.e. reduce the set of chips
1163 and thereby the code size); consult ath_hal(4) for details.
1166 __FreeBSD_version 800054 adds memory barriers to
1167 <machine/atomic.h>, new interfaces to ifnet to facilitate
1168 multiple hardware transmit queues for cards that support
1169 them, and a lock-less ring-buffer implementation to
1170 enable drivers to more efficiently manage queueing of
1174 A new version of ZFS (version 13) has been merged to -HEAD.
1175 This version has zpool attribute "listsnapshots" off by
1176 default, which means "zfs list" does not show snapshots,
1177 and is the same as Solaris behavior.
1180 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1183 The uhci, ohci, ehci and slhci USB Host controller drivers have
1184 been put into separate modules. If you load the usb module
1185 separately through loader.conf you will need to load the
1186 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1187 controller add the following to loader.conf:
1193 The ABI used by the PMC toolset has changed. Please keep
1194 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1198 atapci kernel module now includes only generic PCI ATA
1199 driver. AHCI driver moved to ataahci kernel module.
1200 All vendor-specific code moved into separate kernel modules:
1201 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1202 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1203 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1204 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1207 The TTY subsystem of the kernel has been replaced by a new
1208 implementation, which provides better scalability and an
1209 improved driver model. Most common drivers have been migrated to
1210 the new TTY subsystem, while others have not. The following
1211 drivers have not yet been ported to the new TTY layer:
1214 cy, digi, rc, rp, sio
1220 ng_h4, ng_tty, ppp, sl, snp
1222 Adding these drivers to your kernel configuration file shall
1223 cause compilation to fail.
1226 ntpd has been upgraded to 4.2.4p5.
1229 OpenSSH has been upgraded to 5.1p1.
1231 For many years, FreeBSD's version of OpenSSH preferred DSA
1232 over RSA for host and user authentication keys. With this
1233 upgrade, we've switched to the vendor's default of RSA over
1234 DSA. This may cause upgraded clients to warn about unknown
1235 host keys even for previously known hosts. Users should
1236 follow the usual procedure for verifying host keys before
1237 accepting the RSA key.
1239 This can be circumvented by setting the "HostKeyAlgorithms"
1240 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1243 Please note that the sequence of keys offered for
1244 authentication has been changed as well. You may want to
1245 specify IdentityFile in a different order to revert this
1249 The sio(4) driver has been removed from the i386 and amd64
1250 kernel configuration files. This means uart(4) is now the
1251 default serial port driver on those platforms as well.
1253 To prevent collisions with the sio(4) driver, the uart(4) driver
1254 uses different names for its device nodes. This means the
1255 onboard serial port will now most likely be called "ttyu0"
1256 instead of "ttyd0". You may need to reconfigure applications to
1257 use the new device names.
1259 When using the serial port as a boot console, be sure to update
1260 /boot/device.hints and /etc/ttys before booting the new kernel.
1261 If you forget to do so, you can still manually specify the hints
1262 at the loader prompt:
1264 set hint.uart.0.at="isa"
1265 set hint.uart.0.port="0x3F8"
1266 set hint.uart.0.flags="0x10"
1267 set hint.uart.0.irq="4"
1271 The gpt(8) utility has been removed. Use gpart(8) to partition
1275 The version that Linuxulator emulates was changed from 2.4.2
1276 to 2.6.16. If you experience any problems with Linux binaries
1277 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1278 if it fixes the problem contact emulation mailing list.
1281 ISDN4BSD (I4B) was removed from the src tree. You may need to
1282 update a your kernel configuration and remove relevant entries.
1285 I have checked in code to support multiple routing tables.
1286 See the man pages setfib(1) and setfib(2).
1287 This is a hopefully backwards compatible version,
1288 but to make use of it you need to compile your kernel
1289 with options ROUTETABLES=2 (or more up to 16).
1292 The 802.11 wireless support was redone to enable multi-bss
1293 operation on devices that are capable. The underlying device
1294 is no longer used directly but instead wlanX devices are
1295 cloned with ifconfig. This requires changes to rc.conf files.
1296 For example, change:
1297 ifconfig_ath0="WPA DHCP"
1300 ifconfig_wlan0="WPA DHCP"
1301 see rc.conf(5) for more details. In addition, mergemaster of
1302 /etc/rc.d is highly recommended. Simultaneous update of userland
1303 and kernel wouldn't hurt either.
1305 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1306 modules were merged into the base wlan module. All references
1307 to these modules (e.g. in kernel config files) must be removed.
1310 psm(4) has gained write(2) support in native operation level.
1311 Arbitrary commands can be written to /dev/psm%d and status can
1312 be read back from it. Therefore, an application is responsible
1313 for status validation and error recovery. It is a no-op in
1314 other operation levels.
1317 Support for KSE threading has been removed from the kernel. To
1318 run legacy applications linked against KSE libmap.conf may
1319 be used. The following libmap.conf may be used to ensure
1320 compatibility with any prior release:
1322 libpthread.so.1 libthr.so.1
1323 libpthread.so.2 libthr.so.2
1324 libkse.so.3 libthr.so.3
1327 The layout of struct vmspace has changed. This affects libkvm
1328 and any executables that link against libkvm and use the
1329 kvm_getprocs() function. In particular, but not exclusively,
1330 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1331 The effects are minimal, but it's advisable to upgrade world
1335 The latest em driver no longer has support in it for the
1336 82575 adapter, this is now moved to the igb driver. The
1337 split was done to make new features that are incompatible
1338 with older hardware easier to do.
1341 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1342 likewise the kernel option is now GEOM_LINUX_LVM.
1345 The default NFS mount mode has changed from UDP to TCP for
1346 increased reliability. If you rely on (insecurely) NFS
1347 mounting across a firewall you may need to update your
1351 Belatedly note the addition of m_collapse for compacting
1355 The fts(3) structures have been changed to use adequate
1356 integer types for their members and so to be able to cope
1357 with huge file trees. The old fts(3) ABI is preserved
1358 through symbol versioning in libc, so third-party binaries
1359 using fts(3) should still work, although they will not take
1360 advantage of the extended types. At the same time, some
1361 third-party software might fail to build after this change
1362 due to unportable assumptions made in its source code about
1363 fts(3) structure members. Such software should be fixed
1364 by its vendor or, in the worst case, in the ports tree.
1365 FreeBSD_version 800015 marks this change for the unlikely
1366 case that a portable fix is impossible.
1369 To upgrade to -current after this date, you must be running
1370 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1371 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1374 The ADAPTIVE_GIANT kernel option has been retired because its
1375 functionality is the default now.
1378 The AT keyboard emulation of sunkbd(4) has been turned on
1379 by default. In order to make the special symbols of the Sun
1380 keyboards driven by sunkbd(4) work under X these now have
1381 to be configured the same way as Sun USB keyboards driven
1382 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1384 Option "XkbLayout" "us"
1385 Option "XkbRules" "xorg"
1386 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1389 It has been decided that it is desirable to provide ABI
1390 backwards compatibility to the FreeBSD 4/5/6 versions of the
1391 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1392 broken with the introduction of PCI domain support (see the
1393 20070930 entry). Unfortunately, this required the ABI of
1394 PCIOCGETCONF to be broken again in order to be able to
1395 provide backwards compatibility to the old version of that
1396 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1397 again. As for prominent ports this affects neither pciutils
1398 nor xorg-server this time, the hal port needs to be rebuilt
1402 The misnamed kthread_create() and friends have been renamed
1403 to kproc_create() etc. Many of the callers already
1404 used kproc_start()..
1405 I will return kthread_create() and friends in a while
1406 with implementations that actually create threads, not procs.
1407 Renaming corresponds with version 800002.
1416 Avoid using make -j when upgrading. While generally safe, there are
1417 sometimes problems using -j to upgrade. If your upgrade fails with
1418 -j, please try again without -j. From time to time in the past there
1419 have been problems using -j with buildworld and/or installworld. This
1420 is especially true when upgrading between "distant" versions (eg one
1421 that cross a major release boundary or several minor releases, or when
1422 several months have passed on the -current branch).
1424 Sometimes, obscure build problems are the result of environment
1425 poisoning. This can happen because the make utility reads its
1426 environment when searching for values for global variables. To run
1427 your build attempts in an "environmental clean room", prefix all make
1428 commands with 'env -i '. See the env(1) manual page for more details.
1430 When upgrading from one major version to another it is generally best
1431 to upgrade to the latest code in the currently installed branch first,
1432 then do an upgrade to the new branch. This is the best-tested upgrade
1433 path, and has the highest probability of being successful. Please try
1434 this approach before reporting problems with a major version upgrade.
1438 When upgrading the boot ZFS pool to a new version, always follow
1441 1.) recompile and reinstall the ZFS boot loader and boot block
1442 (this is part of "make buildworld" and "make installworld")
1444 2.) update the ZFS boot block on your boot drive
1446 The following example updates the ZFS boot block on the first
1447 partition (freebsd-boot) of a GPT partitioned drive ad0:
1448 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1450 Non-boot pools do not need these updates.
1454 If you are updating from a prior version of FreeBSD (even one just
1455 a few days old), you should follow this procedure. It is the most
1456 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1458 make kernel-toolchain
1459 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1460 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1462 To test a kernel once
1463 ---------------------
1464 If you just want to boot a kernel once (because you are not sure
1465 if it works, or if you want to boot a known bad kernel to provide
1466 debugging information) run
1467 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1468 nextboot -k testkernel
1470 To just build a kernel when you know that it won't mess you up
1471 --------------------------------------------------------------
1472 This assumes you are already running a CURRENT system. Replace
1473 ${arch} with the architecture of your machine (e.g. "i386",
1474 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1476 cd src/sys/${arch}/conf
1477 config KERNEL_NAME_HERE
1478 cd ../compile/KERNEL_NAME_HERE
1483 If this fails, go to the "To build a kernel" section.
1485 To rebuild everything and install it on the current system.
1486 -----------------------------------------------------------
1487 # Note: sometimes if you are running current you gotta do more than
1488 # is listed here if you are upgrading from a really old current.
1490 <make sure you have good level 0 dumps>
1492 make kernel KERNCONF=YOUR_KERNEL_HERE
1494 <reboot in single user> [3]
1502 To cross-install current onto a separate partition
1503 --------------------------------------------------
1504 # In this approach we use a separate partition to hold
1505 # current's root, 'usr', and 'var' directories. A partition
1506 # holding "/", "/usr" and "/var" should be about 2GB in
1509 <make sure you have good level 0 dumps>
1512 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1513 <maybe newfs current's root partition>
1514 <mount current's root partition on directory ${CURRENT_ROOT}>
1515 make installworld DESTDIR=${CURRENT_ROOT}
1516 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1517 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1518 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1519 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1520 <reboot into current>
1521 <do a "native" rebuild/install as described in the previous section>
1522 <maybe install compatibility libraries from ports/misc/compat*>
1526 To upgrade in-place from 8.x-stable to current
1527 ----------------------------------------------
1528 <make sure you have good level 0 dumps>
1530 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1532 <reboot in single user> [3]
1539 Make sure that you've read the UPDATING file to understand the
1540 tweaks to various things you need. At this point in the life
1541 cycle of current, things change often and you are on your own
1542 to cope. The defaults can also change, so please read ALL of
1543 the UPDATING entries.
1545 Also, if you are tracking -current, you must be subscribed to
1546 freebsd-current@freebsd.org. Make sure that before you update
1547 your sources that you have read and understood all the recent
1548 messages there. If in doubt, please track -stable which has
1549 much fewer pitfalls.
1551 [1] If you have third party modules, such as vmware, you
1552 should disable them at this point so they don't crash your
1555 [3] From the bootblocks, boot -s, and then do
1560 adjkerntz -i # if CMOS is wall time
1561 Also, when doing a major release upgrade, it is required that
1562 you boot into single user mode to do the installworld.
1564 [4] Note: This step is non-optional. Failure to do this step
1565 can result in a significant reduction in the functionality of the
1566 system. Attempting to do it by hand is not recommended and those
1567 that pursue this avenue should read this file carefully, as well
1568 as the archives of freebsd-current and freebsd-hackers mailing lists
1569 for potential gotchas. The -U option is also useful to consider.
1570 See mergemaster(8) for more information.
1572 [5] Usually this step is a noop. However, from time to time
1573 you may need to do this if you get unknown user in the following
1574 step. It never hurts to do it all the time. You may need to
1575 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1576 install) after the buildworld before this step if you last updated
1577 from current before 20020224 or from -stable before 20020408.
1579 [6] This only deletes old files and directories. Old libraries
1580 can be deleted by "make delete-old-libs", but you have to make
1581 sure that no program is using those libraries anymore.
1583 [8] In order to have a kernel that can run the 4.x binaries needed to
1584 do an installworld, you must include the COMPAT_FREEBSD4 option in
1585 your kernel. Failure to do so may leave you with a system that is
1586 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1587 required to run the 5.x binaries on more recent kernels. And so on
1588 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1590 Make sure that you merge any new devices from GENERIC since the
1591 last time you updated your kernel config file.
1593 [9] When checking out sources, you must include the -P flag to have
1594 cvs prune empty directories.
1596 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1597 "?=" instead of the "=" assignment operator, so that buildworld can
1598 override the CPUTYPE if it needs to.
1600 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1601 not on the command line, or in /etc/make.conf. buildworld will
1602 warn if it is improperly defined.
1605 This file contains a list, in reverse chronological order, of major
1606 breakages in tracking -current. It is not guaranteed to be a complete
1607 list of such breakages, and only contains entries since October 10, 2007.
1608 If you need to see UPDATING entries from before that date, you will need
1609 to fetch an UPDATING file from an older FreeBSD release.
1611 Copyright information:
1613 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1615 Redistribution, publication, translation and use, with or without
1616 modification, in full or in part, in any form or format of this
1617 document are permitted without further permission from the author.
1619 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1620 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1621 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1622 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1623 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1624 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1625 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1626 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1627 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1628 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1629 POSSIBILITY OF SUCH DAMAGE.
1631 Contact Warner Losh if you have any questions about your use of