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 http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/makeworld.html
11 Items affecting the ports and packages system can be found in
12 /usr/ports/UPDATING. Please read that file before running portupgrade.
15 The nve(4) driver for NVIDIA nForce MCP Ethernet adapters has
16 been deprecated and will not be part of FreeBSD 11.0 and later
17 releases. If you use this driver, please consider switching to
18 the nfe(4) driver instead.
21 The behavior of gss_pseudo_random() for the krb5 mechanism
22 has changed, for applications requesting a longer random string
23 than produced by the underlying enctype's pseudo-random() function.
24 In particular, the random string produced from a session key of
25 enctype aes256-cts-hmac-sha1-96 or aes256-cts-hmac-sha1-96 will
26 be different at the 17th octet and later, after this change.
27 The counter used in the PRF+ construction is now encoded as a
28 big-endian integer in accordance with RFC 4402.
29 __FreeBSD_version is bumped to 902505.
35 Behavior of devfs rules path matching has been changed.
36 Pattern is now always matched against fully qualified devfs
37 path and slash characters must be explicitly matched by
38 slashes in pattern (FNM_PATHNAME). Rulesets involving devfs
39 subdirectories must be reviewed.
42 hastctl(8)'s `status' command output changed to terse one-liner format.
43 Scripts using this should switch to `list' command or be rewritten.
46 Fix a bug that allowed a tracing process (e.g. gdb) to write
47 to a memory-mapped file in the traced process's address space
48 even if neither the traced process nor the tracing process had
49 write access to that file.
52 Added ZFS TRIM support which is enabled by default. To disable
53 ZFS TRIM support set vfs.zfs.trim.enabled=0 in loader.conf.
55 Creating new ZFS pools and adding new devices to existing pools
56 first performs a full device level TRIM which can take a significant
57 amount of time. The sysctl vfs.zfs.vdev.trim_on_init can be set to 0
58 to disable this behaviour.
60 ZFS TRIM requires the underlying device support BIO_DELETE which
61 is currently provided by methods such as ATA TRIM and SCSI UNMAP
62 via CAM, which are typically supported by SSD's.
64 Stats for ZFS TRIM can be monitored by looking at the sysctl's
65 under kstat.zfs.misc.zio_trim.
68 `list' command has been added to hastctl(8). For now, it is full
69 equivalent of `status' command.
70 WARNING: in the near future the output of hastctl's status command
71 will change to more terse format. If you use `hastctl status'
72 for parsing in your scripts, switch to `hastctl list'.
75 The mergemaster command now uses the default MAKEOBJDIRPREFIX
76 rather than creating it's own in the temporary directory in
77 order allow access to bootstrapped versions of tools such as
78 install and mtree. When upgrading from version of FreeBSD where
79 the install command does not support -l, you will need to
80 install a new mergemaster command if mergemaster -p is required.
81 This can be accomplished with the command (cd src/usr.sbin/mergemaster
84 Due to the use of the new -l option to install(1) during build
85 and install, you must take care not to directly set the INSTALL
86 make variable in your /etc/make.conf, /etc/src.conf, or on the
87 command line. If you wish to use the -C flag for all installs
88 you may be able to add INSTALL+=-C to /etc/make.conf or
92 Fix a bug that allows NFS clients to issue READDIR on files.
95 The install(1) option -M has changed meaning and now takes an
96 argument that is a file or path to append logs to. In the
97 unlikely event that -M was the last option on the command line
98 and the command line contained at least two files and a target
99 directory the first file will have logs appended to it. The -M
100 option served little practical purpose in the last decade so it's
101 used expected to be extremely rare.
104 A new compression method (lz4) has been merged to. Please refer to
105 zpool-features(7) for more information.
107 Please refer to the "ZFS notes" section of this file for information
108 on upgrading boot ZFS pools.
111 The VFS KBI was changed with the merge of several nullfs
112 optimizations and fixes. All filesystem modules must be
116 With the addition of auditdistd(8), a new auditdistd user is now
117 depended on during installworld. "mergemaster -p" can be used to add
118 the user prior to installworld, as documented in the handbook.
124 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
125 Starting with this version the old system of ZFS pool versioning
126 is superseded by "feature flags". This concept enables forward
127 compatibility against certain future changes in functionality of ZFS
128 pools. The first two read-only compatible "feature flags" for ZFS
129 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
130 For more information read the new zpool-features(7) manual page.
131 Please refer to the "ZFS notes" section of this file for information
132 on upgrading boot ZFS pools.
135 The commit introducing bsd.compiler.mk breaks the traditional
136 building of kernels before this point. Add -m ${SRC}/share/mk
137 (for the right value of SRC) to your command lines to work
138 around; update your useland to a point after this; or use the
139 buildkernel/installkernel top-level targets. See also 20120829.
142 The IPFIREWALL_FORWARD kernel option has been removed. Its
143 functionality now turned on by default.
146 The random(4) support for the VIA hardware random number
147 generator (`PADLOCK') is no longer enabled unconditionally.
148 Add the PADLOCK_RNG option in the custom kernel config if
149 needed. The GENERIC kernels on i386 and amd64 do include the
150 option, so the change only affects the custom kernel
154 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
155 the traditional method, you must update your system with an installworld
156 before the kernel will build. The documented make buildkernel/installkernel
157 interfaces (coupled with fresh make kernel-toolchain) continue to work.
160 The sparc64 ZFS loader has been changed to no longer try to auto-
161 detect ZFS providers based on diskN aliases but now requires these
162 to be explicitly listed in the OFW boot-device environment variable.
165 Now unix domain sockets behave "as expected" on nullfs(5). Previously
166 nullfs(5) did not pass through all behaviours to the underlying layer,
167 as a result if we bound to a socket on the lower layer we could connect
168 only to the lower path; if we bound to the upper layer we could connect
169 only to the upper path. The new behavior is one can connect to both the
170 lower and the upper paths regardless what layer path one binds to.
173 The acpi_wmi(4) status device /dev/wmistat has been renamed to
177 A new VOP_ADVISE() was added to support posix_fadvise(2). All
178 filesystem modules must be recompiled.
181 The interface of the VOP_VPTOCNP(9) changed, now the returned
182 vnode shall be referenced, previously it was required to be
183 only held. All in-tree filesystems are converted.
189 The broken amd(4) driver has been replaced with esp(4) in the amd64,
190 i386 and pc98 GENERIC kernel configuration files.
193 This commit modifies vfs_register() so that it uses a hash
194 calculation to set vfc_typenum, which is enabled by default.
195 The first time a system is booted after this change, the
196 vfc_typenum values will change for all file systems. The
197 main effect of this is a change to the NFS server file handles
198 for file systems that use vfc_typenum in their fsid, such as ZFS.
199 It will, however, prevent vfc_typenum from changing when file
200 systems are loaded in a different order for subsequent reboots.
201 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
202 until you are ready to remount all NFS clients after a reboot.
205 Bump the shared library version numbers for libraries that
206 do not use symbol versioning, have changed the ABI compared
207 to stable/8 and which shared library version was not bumped.
208 Done as part of 9.0-RELEASE cycle.
211 During the merge of Capsicum features, the fget(9) KPI was modified.
212 This may require the rebuilding of out-of-tree device drivers --
213 issues have been reported specifically with the nVidia device driver.
214 __FreeBSD_version is bumped to 900041.
216 Also, there is a period between 20110811 and 20110814 where the
217 special devices /dev/{stdin,stdout,stderr} did not work correctly.
218 Building world from a kernel during that window may not work.
221 The packet filter (pf) code has been updated to OpenBSD 4.5.
222 You need to update userland tools to be in sync with kernel.
223 This update breaks backward compatibility with earlier pfsync(4)
224 versions. Care must be taken when updating redundant firewall setups.
227 The following sysctls and tunables are retired on x86 platforms:
229 machdep.hlt_logical_cpus
230 The following sysctl is retired:
231 machdep.hyperthreading_allowed
232 The sysctls were supposed to provide a way to dynamically offline and
233 online selected CPUs on x86 platforms, but the implementation has not
234 been reliable especially with SCHED_ULE scheduler.
235 machdep.hyperthreading_allowed tunable is still available to ignore
236 hyperthreading CPUs at OS level.
237 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
238 where X is an APIC ID of a CPU. Be advised, though, that disabling
239 CPUs in non-uniform fashion will result in non-uniform topology and
240 may lead to sub-optimal system performance with SCHED_ULE, which is
244 cpumask_t type is retired and cpuset_t is used in order to describe
248 Changes to ifconfig(8) for dynamic address family detection mandate
249 that you are running a kernel of 20110525 or later. Make sure to
250 follow the update procedure to boot a new kernel before installing
254 Support for sun4v architecture is officially dropped
257 Several KPI breaking changes have been committed to the mii(4) layer,
258 the PHY drivers and consequently some Ethernet drivers using mii(4).
259 This means that miibus.ko and the modules of the affected Ethernet
260 drivers need to be recompiled.
262 Note to kernel developers: Given that the OUI bit reversion problem
263 was fixed as part of these changes all mii(4) commits related to OUIs,
264 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
265 handling, no longer can be merged verbatim to stable/8 and previous
269 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
270 into their kernel configurations along with 'device pci'.
273 The default NFS client is now the new NFS client, so fstype "newnfs"
274 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
275 Although mounts via fstype "nfs" will usually work without userland
276 changes, it is recommended that the mount(8) and mount_nfs(8)
277 commands be rebuilt from sources and that a link to mount_nfs called
278 mount_oldnfs be created. The new client is compiled into the
279 kernel with "options NFSCL" and this is needed for diskless root
280 file systems. The GENERIC kernel configs have been changed to use
281 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
282 To use the regular/old client, you can "mount -t oldnfs ...". For
283 a diskless root file system, you must also include a line like:
285 vfs.root.mountfrom="oldnfs:"
287 in the boot/loader.conf on the root fs on the NFS server to make
288 a diskless root fs use the old client.
291 The GENERIC kernels for all architectures now default to the new
292 CAM-based ATA stack. It means that all legacy ATA drivers were
293 removed and replaced by respective CAM drivers. If you are using
294 ATA device names in /etc/fstab or other places, make sure to update
295 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
296 where 'Y's are the sequential numbers starting from zero for each type
297 in order of detection, unless configured otherwise with tunables,
298 see cam(4)). There will be symbolic links created in /dev/ to map
299 old adX devices to the respective adaY. They should provide basic
300 compatibility for file systems mounting in most cases, but they do
301 not support old user-level APIs and do not have respective providers
302 in GEOM. Consider using updated management tools with new device names.
304 It is possible to load devices ahci, ata, siis and mvs as modules,
305 but option ATA_CAM should remain in kernel configuration to make ata
306 module work as CAM driver supporting legacy ATA controllers. Device ata
307 still can be used in modular fashion (atacore + ...). Modules atadisk
308 and atapi* are not used and won't affect operation in ATA_CAM mode.
309 Note that to use CAM-based ATA kernel should include CAM devices
310 scbus, pass, da (or explicitly ada), cd and optionally others. All of
311 them are parts of the cam module.
313 ataraid(4) functionality is now supported by the RAID GEOM class.
314 To use it you can load geom_raid kernel module and use graid(8) tool
315 for management. Instead of /dev/arX device names, use /dev/raid/rX.
317 No kernel config options or code have been removed, so if a problem
318 arises, please report it and optionally revert to the old ATA stack.
319 In order to do it you can remove from the kernel config:
324 , and instead add back:
325 device atadisk # ATA disk drives
326 device ataraid # ATA RAID drives
327 device atapicd # ATAPI CDROM drives
328 device atapifd # ATAPI floppy drives
329 device atapist # ATAPI tape drives
332 The default NFS server has been changed to the new server, which
333 was referred to as the experimental server. If you need to switch
334 back to the old NFS server, you must now put the "-o" option on
335 both the mountd and nfsd commands. This can be done using the
336 mountd_flags and nfs_server_flags rc.conf variables until an
337 update to the rc scripts is committed, which is coming soon.
340 The GNU Objective-C runtime library (libobjc), and other Objective-C
341 related components have been removed from the base system. If you
342 require an Objective-C library, please use one of the available ports.
345 ath(4) has been split into bus- and device- modules. if_ath contains
346 the HAL, the TX rate control and the network device code. if_ath_pci
347 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
348 contains the AHB glue. Users need to load both if_ath_pci and if_ath
349 in order to use ath on everything else.
351 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
352 need to load if_ath and if_ath_pci for ath(4) operation.
355 As part of the replacement of sysinstall, the process of building
356 release media has changed significantly. For details, please re-read
357 release(7), which has been updated to reflect the new build process.
360 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
361 is the last available version under GPLv2. It brings a number of new
362 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
363 4.1 and SSE 4.2), better support for powerpc64, a number of new
364 directives, and lots of other small improvements. See the ChangeLog
365 file in contrib/binutils for the full details.
368 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
369 compliant, and will now use half of hash for authentication.
370 This will break interoperability with all stacks (including all
371 actual FreeBSD versions) who implement
372 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
374 The only workaround with such peers is to use another HMAC
375 algorithm for IPsec ("phase 2") authentication.
378 Remove the uio_yield prototype and symbol. This function has
379 been misnamed since it was introduced and should not be
380 globally exposed with this name. The equivalent functionality
381 is now available using kern_yield(curthread->td_user_pri).
382 The function remains undocumented.
385 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
386 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
387 for scalar sysctls is defined but disabled. Code that needs
388 UQUAD to pass the type checking that must compile on older
389 systems where the define is not present can check against
390 __FreeBSD_version >= 900030.
392 The system dialog(1) has been replaced with a new version previously
393 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
394 with the previous version, but the libdialog associated with it has
395 a largely incompatible API. As such, the original version of libdialog
396 will be kept temporarily as libodialog, until its base system consumers
397 are replaced or updated. Bump __FreeBSD_version to 900030.
400 If you are trying to run make universe on a -stable system, and you get
401 the following warning:
402 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
403 unknown. config(8) likely too old."
404 or something similar to it, then you must upgrade your -stable system
405 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
406 stable/8) or build the config from the latest stable/8 branch and
407 install it on your system.
409 Prior to this date, building a current universe on 8-stable system from
410 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
411 error in the first kernel build phase. A new config on those old
412 systems will fix that problem for older versions of -current.
415 The TCP stack has been modified to allow Khelp modules to interact with
416 it via helper hook points and store per-connection data in the TCP
417 control block. Bump __FreeBSD_version to 900029. User space tools that
418 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
422 Generic IEEE 802.3 annex 31B full duplex flow control support has been
423 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
424 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
425 to take advantage of it instead of using custom implementations. This
426 means that these drivers now no longer unconditionally advertise
427 support for flow control but only do so if flow control is a selected
428 media option. This was implemented in the generic support that way in
429 order to allow flow control to be switched on and off via ifconfig(8)
430 with the PHY specific default to typically off in order to protect
431 from unwanted effects. Consequently, if you used flow control with
432 one of the above mentioned drivers you now need to explicitly enable
434 ifconfig bge0 media auto mediaopt flowcontrol
436 Along with the above mentioned changes generic support for setting
437 1000baseT master mode also has been added and brgphy(4), ciphy(4),
438 e1000phy(4) as well as ip1000phy(4) have been converted to take
439 advantage of it. This means that these drivers now no longer take the
440 link0 parameter for selecting master mode but the master media option
441 has to be used instead, for example like in the following:
442 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
444 Selection of master mode now is also available with all other PHY
445 drivers supporting 1000baseT.
448 The TCP stack has received a significant update to add support for
449 modularised congestion control and generally improve the clarity of
450 congestion control decisions. Bump __FreeBSD_version to 900025. User
451 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
452 sockstat) need to be recompiled.
455 The man(1) utility has been replaced by a new version that no longer
456 uses /etc/manpath.config. Please consult man.conf(5) for how to
457 migrate local entries to the new format.
460 The copyright strings printed by login(1) and sshd(8) at the time of a
461 new connection have been removed to follow other operating systems and
465 A workaround for a fixed ld bug has been removed in kernel code,
466 so make sure that your system ld is built from sources after
467 revision 210245 from 2010-07-19 (r211583 if building head kernel
468 on stable/8, r211584 for stable/7; both from 2010-08-21).
469 A symptom of incorrect ld version is different addresses for
470 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
473 The $ipv6_prefer variable in rc.conf(5) has been split into
474 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
476 The $ip6addrctl_policy is a variable to choose a pre-defined
477 address selection policy set by ip6addrctl(8). A value
478 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
481 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
482 flag (see an entry of 20090926) is set on an interface with no
483 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
484 security reason. If you want IPv6 link-local address on all
485 interfaces by default, set this to "YES".
487 The old ipv6_prefer="YES" is equivalent to
488 ipv6_activate_all_interfaces="YES" and
489 ip6addrctl_policy="ipv6_prefer".
492 DTrace has grown support for userland tracing. Due to this, DTrace is
493 now i386 and amd64 only.
494 dtruss(1) is now installed by default on those systems and a new
495 kernel module is needed for userland tracing: fasttrap.
496 No changes to your kernel config file are necessary to enable
497 userland tracing, but you might consider adding 'STRIP=' and
498 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
499 to have informative userland stack traces in DTrace (ustack).
502 The acpi_aiboost(4) driver has been removed in favor of the new
503 aibs(4) driver. You should update your kernel configuration file.
506 BSD grep has been imported to the base system and it is built by
507 default. It is completely BSD licensed, highly GNU-compatible, uses
508 less memory than its GNU counterpart and has a small codebase.
509 However, it is slower than its GNU counterpart, which is mostly
510 noticeable for larger searches, for smaller ones it is measurable
511 but not significant. The reason is complex, the most important factor
512 is that we lack a modern and efficient regex library and GNU
513 overcomes this by optimizing the searches internally. Future work
514 on improving the regex performance is planned, for the meantime,
515 users that need better performance, can build GNU grep instead by
516 setting the WITH_GNU_GREP knob.
519 Due to the import of powerpc64 support, all existing powerpc kernel
520 configuration files must be updated with a machine directive like this:
521 machine powerpc powerpc
523 In addition, an updated config(8) is required to build powerpc kernels
527 A new version of ZFS (version 15) has been merged to -HEAD.
528 This version uses a python library for the following subcommands:
529 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
530 For full functionality of these commands the following port must
531 be installed: sysutils/py-zfs
534 'vm_page's are now hashed by physical address to an array of mutexes.
535 Currently this is only used to serialize access to hold_count. Over
536 time the page queue mutex will be peeled away. This changes the size
537 of pmap on every architecture. And requires all callers of vm_page_hold
538 and vm_page_unhold to be updated.
541 WITH_CTF can now be specified in src.conf (not recommended, there
542 are some problems with static executables), make.conf (would also
543 affect ports which do not use GNU make and do not override the
544 compile targets) or in the kernel config (via "makeoptions
546 When WITH_CTF was specified there before this was silently ignored,
547 so make sure that WITH_CTF is not used in places which could lead
548 to unwanted behavior.
551 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
552 to allow 32-bit compatibility on non-x86 platforms. All kernel
553 configurations on amd64 and ia64 platforms using these options must
554 be modified accordingly.
557 The utmp user accounting database has been replaced with utmpx,
558 the user accounting interface standardized by POSIX.
559 Unfortunately the semantics of utmp and utmpx don't match,
560 making it practically impossible to support both interfaces.
561 The user accounting database is used by tools like finger(1),
562 last(1), talk(1), w(1) and ac(8).
564 All applications in the base system use utmpx. This means only
565 local binaries (e.g. from the ports tree) may still use these
566 utmp database files. These applications must be rebuilt to make
569 After the system has been upgraded, it is safe to remove the old
570 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
571 assuming their contents is of no importance anymore. Old wtmp
572 databases can only be used by last(1) and ac(8) after they have
573 been converted to the new format using wtmpcvt(1).
576 Introduce the kernel thread "deadlock resolver" (which can be enabled
577 via the DEADLKRES option, see NOTES for more details) and the
578 sleepq_type() function for sleepqueues.
581 The rc.firewall and rc.firewall6 were unified, and
582 rc.firewall6 and rc.d/ip6fw were removed.
583 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
584 variables are obsoleted. Instead, the following new rc
585 variables are added to rc.d/ipfw:
587 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
588 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
589 firewall_simple_onet_ipv6, firewall_trusted_ipv6
591 The meanings correspond to the relevant IPv4 variables.
597 The default terminal emulation for syscons(4) has been changed
598 from cons25 to xterm on all platforms except pc98. This means
599 that the /etc/ttys file needs to be updated to ensure correct
600 operation of applications on the console.
602 The terminal emulation style can be toggled per window by using
603 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
604 options can be used to change the compile-time default back to
607 To prevent graphical artifacts, make sure the TERM environment
608 variable is set to match the terminal emulation that is being
609 performed by syscons(4).
612 The layout of the structure ieee80211req_scan_result has changed.
613 Applications that require wireless scan results (e.g. ifconfig(8))
614 from net80211 need to be recompiled.
616 Applications such as wpa_supplicant(8) may require a full world
617 build without using NO_CLEAN in order to get synchronized with the
621 The iwn(4) driver has been updated to support the 5000 and 5150 series.
622 There's one kernel module for each firmware. Adding "device iwnfw"
623 to the kernel configuration file means including all three firmware
624 images inside the kernel. If you want to include just the one for
625 your wireless card, use the devices iwn4965fw, iwn5000fw or
629 The rc.d/network_ipv6, IPv6 configuration script has been integrated
630 into rc.d/netif. The changes are the following:
632 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
633 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
634 Note that both variables need the "inet6" keyword at the head.
636 Do not set $ipv6_network_interfaces manually if you do not
637 understand what you are doing. It is not needed in most cases.
639 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
642 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
643 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
645 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
646 all configured interfaces have "inet6 accept_rtadv" in the
647 $ifconfig_IF_ipv6. These are for backward compatibility.
649 3. A new variable $ipv6_prefer has been added. If NO, IPv6
650 functionality of interfaces with no corresponding
651 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
652 and the default address selection policy of ip6addrctl(8)
653 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
654 Note that if you want to configure IPv6 functionality on the
655 disabled interfaces after boot, first you need to clear the flag by
656 using ifconfig(8) like:
658 ifconfig em0 inet6 -ifdisabled
660 If YES, the default address selection policy is set as
663 The default value of $ipv6_prefer is NO.
665 4. If your system need to receive Router Advertisement messages,
666 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
667 scripts automatically invoke rtsol(8) when the interface becomes
668 UP. The Router Advertisement messages are used for SLAAC
669 (State-Less Address AutoConfiguration).
672 802.11s D3.03 support was committed. This is incompatible with the
673 previous code, which was based on D3.0.
676 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
677 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
678 control whether accepting Router Advertisement messages or not.
679 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
680 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
681 The ifconfig(8) utility now supports these flags.
684 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
685 mount(8) and -a option for df(1) to see them.
688 The old tunable hw.bus.devctl_disable has been superseded by
689 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
690 replaced by hw.bus.devctl_queue=0. The default for this new tunable
694 Remove the option STOP_NMI. The default action is now to use NMI only
695 for KDB via the newly introduced function stop_cpus_hard() and
696 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
699 The stable/8 branch created in subversion. This corresponds to the
700 RELENG_8 branch in CVS.
703 Bump the shared library version numbers for all libraries that do not
704 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
705 __FreeBSD_version to 800105.
708 Due to changes in the implementation of virtual network stack support,
709 all network-related kernel modules must be recompiled. As this change
710 breaks the ABI, bump __FreeBSD_version to 800104.
713 The TOE interface to the TCP syncache has been modified to remove
714 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
715 The cxgb driver is the only TOE consumer affected by this change, and
716 needs to be recompiled along with the kernel. As this change breaks
717 the ABI, bump __FreeBSD_version to 800103.
720 Padding has been added to struct tcpcb, sackhint and tcpstat in
721 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
722 maintaining the ABI. However, this change breaks the ABI, so bump
723 __FreeBSD_version to 800102. User space tools that rely on the size of
724 any of these structs (e.g. sockstat) need to be recompiled.
727 The NFS_LEGACYRPC option has been removed along with the old kernel
728 RPC implementation that this option selected. Kernel configurations
729 may need to be adjusted.
732 The network interface device nodes at /dev/net/<interface> have been
733 removed. All ioctl operations can be performed the normal way using
734 routing sockets. The kqueue functionality can generally be replaced
735 with routing sockets.
738 The documentation from the FreeBSD Documentation Project (Handbook,
739 FAQ, etc.) is now installed via packages by sysinstall(8) and under
740 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
743 The ABI of various structures related to the SYSV IPC API have been
744 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
745 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
749 Layout of struct vnet has changed as routing related variables were
750 moved to their own Vimage module. Modules need to be recompiled. Bump
751 __FreeBSD_version to 800099.
754 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
755 respectively. As long as no more than 16 groups per process are used,
756 no changes should be visible. When more than 16 groups are used, old
757 binaries may fail if they call getgroups() or getgrouplist() with
758 statically sized storage. Recompiling will work around this, but
759 applications should be modified to use dynamically allocated storage
760 for group arrays as POSIX.1-2008 does not cap an implementation's
761 number of supported groups at NGROUPS_MAX+1 as previous versions did.
763 NFS and portalfs mounts may also be affected as the list of groups is
764 truncated to 16. Users of NFS who use more than 16 groups, should
765 take care that negative group permissions are not used on the exported
766 file systems as they will not be reliable unless a GSSAPI based
767 authentication method is used.
770 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
771 option compiles in the support for adaptive spinning for lockmgrs
772 which want to enable it. The lockinit() function now accepts the flag
773 LK_ADAPTIVE in order to make the lock object subject to adaptive
774 spinning when both held in write and read mode.
777 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
778 changed. User applications that use this ioctl need to be rebuilt.
781 The layout of struct thread has changed. Kernel and modules need to
785 The layout of structs ifnet, domain, protosw and vnet_net has changed.
786 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
789 window(1) has been removed from the base system. It can now be
790 installed from ports. The port is called misc/window.
793 The way we are storing and accessing `routing table' entries has
794 changed. Programs reading the FIB, like netstat, need to be
798 A new netisr implementation has been added for FreeBSD 8. Network
799 file system modules, such as igmp, ipdivert, and others, should be
801 Bump __FreeBSD_version to 800096.
804 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
808 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
809 Bump __FreeBSD_version to 800094.
812 Add mnt_xflag field to 'struct mount'. File system modules need to be
814 Bump __FreeBSD_version to 800093.
817 The compiling option ADAPTIVE_SX has been retired while it has been
818 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
819 The KPI for sx_init_flags() changes as accepting flags:
820 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
821 been introduced in order to handle the reversed logic.
822 Bump __FreeBSD_version to 800092.
825 Add support for hierarchical jails. Remove global securelevel.
826 Bump __FreeBSD_version to 800091.
829 The layout of struct vnet_net has changed, therefore modules
831 Bump __FreeBSD_version to 800090.
834 The newly imported zic(8) produces a new format in the output. Please
835 run tzsetup(8) to install the newly created data to /etc/localtime.
838 The sysctl tree for the usb stack has renamed from hw.usb2.* to
839 hw.usb.* and is now consistent again with previous releases.
842 802.11 monitor mode support was revised and driver api's were changed.
843 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
844 of DLT_IEEE802_11. No user-visible data structures were changed but
845 applications that use DLT_IEEE802_11 may require changes.
846 Bump __FreeBSD_version to 800088.
849 The layout of the following structs has changed: sysctl_oid,
850 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
851 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
852 panics may be experienced. World rebuild is required for
853 correctly checking networking state from userland.
854 Bump __FreeBSD_version to 800085.
857 MLDv2 and Source-Specific Multicast (SSM) have been merged
858 to the IPv6 stack. VIMAGE hooks are in but not yet used.
859 The implementation of SSM within FreeBSD's IPv6 stack closely
860 follows the IPv4 implementation.
862 For kernel developers:
864 * The most important changes are that the ip6_output() and
865 ip6_input() paths no longer take the IN6_MULTI_LOCK,
866 and this lock has been downgraded to a non-recursive mutex.
868 * As with the changes to the IPv4 stack to support SSM, filtering
869 of inbound multicast traffic must now be performed by transport
870 protocols within the IPv6 stack. This does not apply to TCP and
871 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
873 * The KPIs used by IPv6 multicast are similar to those used by
874 the IPv4 stack, with the following differences:
875 * im6o_mc_filter() is analogous to imo_multicast_filter().
876 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
877 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
878 * IN6_LOOKUP_MULTI() has been deprecated and removed.
879 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
880 for MLDv1 have an additional 'timer' argument which is used to
881 jitter the initial membership report for the solicited-node
882 multicast membership on-link.
883 * This is not strictly needed for MLDv2, which already jitters
884 its report transmissions. However, the 'timer' argument is
885 preserved in case MLDv1 is active on the interface.
887 * The KAME linked-list based IPv6 membership implementation has
888 been refactored to use a vector similar to that used by the IPv4
890 Code which maintains a list of its own multicast memberships
891 internally, e.g. carp, has been updated to reflect the new
894 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
895 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
896 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
897 implementation constraint which needs to be addressed in HEAD.
899 For application developers:
901 * The changes are broadly similar to those made for the IPv4
904 * The use of IPv4 and IPv6 multicast socket options on the same
905 socket, using mapped addresses, HAS NOT been tested or supported.
907 * There are a number of issues with the implementation of various
908 IPv6 multicast APIs which need to be resolved in the API surface
909 before the implementation is fully compatible with KAME userland
910 use, and these are mostly to do with interface index treatment.
912 * The literature available discusses the use of either the delta / ASM
913 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
914 using setsourcefilter(3)/getsourcefilter(3). For more information
915 please refer to RFC 3768, 'Socket Interface Extensions for
916 Multicast Source Filters'.
918 * Applications which use the published RFC 3678 APIs should be fine.
920 For systems administrators:
922 * The mtest(8) utility has been refactored to support IPv6, in
923 addition to IPv4. Interface addresses are no longer accepted
924 as arguments, their names must be used instead. The utility
925 will map the interface name to its first IPv4 address as
926 returned by getifaddrs(3).
928 * The ifmcstat(8) utility has also been updated to print the MLDv2
929 endpoint state and source filter lists via sysctl(3).
931 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
932 loopback of IPv6 multicast datagrams by default; it defaults to 1
933 to preserve the existing behaviour. Disabling multicast loopback is
934 recommended for optimal system performance.
936 * The IPv6 MROUTING code has been changed to examine this sysctl
937 instead of attempting to perform a group lookup before looping
938 back forwarded datagrams.
940 Bump __FreeBSD_version to 800084.
943 Implement low-level Bluetooth HCI API.
944 Bump __FreeBSD_version to 800083.
947 The layout of struct malloc_type, used by modules to register new
948 memory allocation types, has changed. Most modules will need to
949 be rebuilt or panics may be experienced.
950 Bump __FreeBSD_version to 800081.
953 Anticipate overflowing inp_flags - add inp_flags2.
954 This changes most offsets in inpcb, so checking v4 connection
955 state will require a world rebuild.
956 Bump __FreeBSD_version to 800080.
959 Add an llentry to struct route and struct route_in6. Modules
960 embedding a struct route will need to be recompiled.
961 Bump __FreeBSD_version to 800079.
964 The size of rt_metrics_lite and by extension rtentry has changed.
965 Networking administration apps will need to be recompiled.
966 The route command now supports show as an alias for get, weighting
967 of routes, sticky and nostick flags to alter the behavior of stateful
969 Bump __FreeBSD_version to 800078.
972 Do not use Giant for kbdmux(4) locking. This is wrong and
973 apparently causing more problems than it solves. This will
974 re-open the issue where interrupt handlers may race with
975 kbdmux(4) in polling mode. Typical symptoms include (but
976 not limited to) duplicated and/or missing characters when
977 low level console functions (such as gets) are used while
978 interrupts are enabled (for example geli password prompt,
979 mountroot prompt etc.). Disabling kbdmux(4) may help.
982 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
983 kernel modules referencing any of the above need to be recompiled.
984 Bump __FreeBSD_version to 800075.
987 GEOM_PART has become the default partition slicer for storage devices,
988 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
989 introduces some changes:
991 MSDOS/EBR: the devices created from MSDOS extended partition entries
992 (EBR) can be named differently than with GEOM_MBR and are now symlinks
993 to devices with offset-based names. fstabs may need to be modified.
995 BSD: the "geometry does not match label" warning is harmless in most
996 cases but it points to problems in file system misalignment with
997 disk geometry. The "c" partition is now implicit, covers the whole
998 top-level drive and cannot be (mis)used by users.
1000 General: Kernel dumps are now not allowed to be written to devices
1001 whose partition types indicate they are meant to be used for file
1002 systems (or, in case of MSDOS partitions, as something else than
1005 Most of these changes date approximately from 200812.
1008 The uscanner(4) driver has been removed from the kernel. This follows
1009 Linux removing theirs in 2.6 and making libusb the default interface
1010 (supported by sane).
1013 The multicast forwarding code has been cleaned up. netstat(1)
1014 only relies on KVM now for printing bandwidth upcall meters.
1015 The IPv4 and IPv6 modules are split into ip_mroute_mod and
1016 ip6_mroute_mod respectively. The config(5) options for statically
1017 compiling this code remain the same, i.e. 'options MROUTING'.
1020 Support for the IFF_NEEDSGIANT network interface flag has been
1021 removed, which means that non-MPSAFE network device drivers are no
1022 longer supported. In particular, if_ar, if_sr, and network device
1023 drivers from the old (legacy) USB stack can no longer be built or
1027 POSIX.1 Native Language Support (NLS) has been enabled in libc and
1028 a bunch of new language catalog files have also been added.
1029 This means that some common libc messages are now localized and
1030 they depend on the LC_MESSAGES environmental variable.
1033 The k8temp(4) driver has been renamed to amdtemp(4) since
1034 support for Family 10 and Family 11 CPU families was added.
1037 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1038 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1040 For kernel developers, the most important changes are that the
1041 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1042 and this lock has been downgraded to a non-recursive mutex.
1044 Transport protocols (UDP, Raw IP) are now responsible for filtering
1045 inbound multicast traffic according to group membership and source
1046 filters. The imo_multicast_filter() KPI exists for this purpose.
1047 Transports which do not use multicast (SCTP, TCP) already reject
1048 multicast by default. Forwarding and receive performance may improve
1049 as a mutex acquisition is no longer needed in the ip_input()
1050 low-level input path. in_addmulti() and in_delmulti() are shimmed
1051 to new KPIs which exist to support SSM in-kernel.
1053 For application developers, it is recommended that loopback of
1054 multicast datagrams be disabled for best performance, as this
1055 will still cause the lock to be taken for each looped-back
1056 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1057 be tuned to 0 to disable loopback by default; it defaults to 1
1058 to preserve the existing behaviour.
1060 For systems administrators, to obtain best performance with
1061 multicast reception and multiple groups, it is always recommended
1062 that a card with a suitably precise hash filter is used. Hash
1063 collisions will still result in the lock being taken within the
1064 transport protocol input path to check group membership.
1066 If deploying FreeBSD in an environment with IGMP snooping switches,
1067 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1068 enabled; this forces 224.0.0.0/24 group membership to be announced
1071 The size of 'struct igmpstat' has changed; netstat needs to be
1072 recompiled to reflect this.
1073 Bump __FreeBSD_version to 800070.
1076 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1077 updated to use it. This requires a buildworld/installworld in order to
1078 update the library and dependencies (usbconfig, etc). Its advisable to
1079 rebuild all ports which uses libusb. More specific directions are given
1080 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1081 libusb are no longer required and can be removed.
1084 A workaround is committed to allow the creation of System V shared
1085 memory segment of size > 2 GB on the 64-bit architectures.
1086 Due to a limitation of the existing ABI, the shm_segsz member
1087 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1088 wrong for large segments. Note that limits must be explicitly
1089 raised to allow such segments to be created.
1092 The layout of struct ifnet has changed, requiring a rebuild of all
1093 network device driver modules.
1096 The /dev handling for the new USB stack has changed, a
1097 buildworld/installworld is required for libusb20.
1100 The new USB2 stack has now been permanently moved in and all kernel and
1101 module names reverted to their previous values (eg, usb, ehci, ohci,
1102 ums, ...). The old usb stack can be compiled in by prefixing the name
1103 with the letter 'o', the old usb modules have been removed.
1104 Updating entry 20090216 for xorg and 20090215 for libmap may still
1108 The rc.conf(5) option if_up_delay has been renamed to
1109 defaultroute_delay to better reflect its purpose. If you have
1110 customized this setting in /etc/rc.conf you need to update it to
1114 xorg 7.4 wants to configure its input devices via hald which does not
1115 yet work with USB2. If the keyboard/mouse does not work in xorg then
1117 Option "AllowEmptyInput" "off"
1118 to your ServerLayout section. This will cause X to use the configured
1119 kbd and mouse sections from your xorg.conf.
1122 The GENERIC kernels for all architectures now default to the new USB2
1123 stack. No kernel config options or code have been removed so if a
1124 problem arises please report it and optionally revert to the old USB
1125 stack. If you are loading USB kernel modules or have a custom kernel
1126 that includes GENERIC then ensure that usb names are also changed over,
1127 eg uftdi -> usb2_serial_ftdi.
1129 Older programs linked against the ports libusb 0.1 need to be
1130 redirected to the new stack's libusb20. /etc/libmap.conf can
1132 # Map old usb library to new one for usb2 stack
1133 libusb-0.1.so.8 libusb20.so.1
1136 All USB ethernet devices now attach as interfaces under the name ueN
1137 (eg. ue0). This is to provide a predictable name as vendors often
1138 change usb chipsets in a product without notice.
1141 The ichsmb(4) driver has been changed to require SMBus slave
1142 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1143 All of the other SMBus controller drivers require left-justified
1144 slave addresses, so this change makes all the drivers provide the
1148 INET6 statistics (struct ip6stat) was updated.
1149 netstat(1) needs to be recompiled.
1152 NTFS has been removed from GENERIC kernel on amd64 to match
1153 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1154 will load ntfs.ko module automatically when NTFS support is
1155 actually needed, unless ntfs.ko is not installed or security
1156 level prohibits loading kernel modules. If either is the case,
1157 "options NTFS" has to be added into kernel config.
1160 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1161 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1162 800061. User space tools that rely on the size of struct tcpcb in
1163 tcp_var.h (e.g. sockstat) need to be recompiled.
1166 ng_tty(4) module updated to match the new TTY subsystem.
1167 Due to API change, user-level applications must be updated.
1168 New API support added to mpd5 CVS and expected to be present
1169 in next mpd5.3 release.
1172 With __FreeBSD_version 800060 the makefs tool is part of
1173 the base system (it was a port).
1176 The afdata and ifnet locks have been changed from mutexes to
1177 rwlocks, network modules will need to be re-compiled.
1180 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1181 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1182 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1183 architecture and 40 bytes on 64-bit architecture. The userland
1184 applications "arp" and "ndp" have been updated accordingly.
1185 The output from "netstat -r" shows only routing entries and
1186 none of the L2 information.
1189 __FreeBSD_version 800057 marks the switchover from the
1190 binary ath hal to source code. Users must add the line:
1192 options AH_SUPPORT_AR5416
1194 to their kernel config files when specifying:
1198 The ath_hal module no longer exists; the code is now compiled
1199 together with the driver in the ath module. It is now
1200 possible to tailor chip support (i.e. reduce the set of chips
1201 and thereby the code size); consult ath_hal(4) for details.
1204 __FreeBSD_version 800054 adds memory barriers to
1205 <machine/atomic.h>, new interfaces to ifnet to facilitate
1206 multiple hardware transmit queues for cards that support
1207 them, and a lock-less ring-buffer implementation to
1208 enable drivers to more efficiently manage queueing of
1212 A new version of ZFS (version 13) has been merged to -HEAD.
1213 This version has zpool attribute "listsnapshots" off by
1214 default, which means "zfs list" does not show snapshots,
1215 and is the same as Solaris behavior.
1218 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1221 The uhci, ohci, ehci and slhci USB Host controller drivers have
1222 been put into separate modules. If you load the usb module
1223 separately through loader.conf you will need to load the
1224 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1225 controller add the following to loader.conf:
1231 The ABI used by the PMC toolset has changed. Please keep
1232 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1236 atapci kernel module now includes only generic PCI ATA
1237 driver. AHCI driver moved to ataahci kernel module.
1238 All vendor-specific code moved into separate kernel modules:
1239 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1240 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1241 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1242 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1245 The TTY subsystem of the kernel has been replaced by a new
1246 implementation, which provides better scalability and an
1247 improved driver model. Most common drivers have been migrated to
1248 the new TTY subsystem, while others have not. The following
1249 drivers have not yet been ported to the new TTY layer:
1252 cy, digi, rc, rp, sio
1258 ng_h4, ng_tty, ppp, sl, snp
1260 Adding these drivers to your kernel configuration file shall
1261 cause compilation to fail.
1264 ntpd has been upgraded to 4.2.4p5.
1267 OpenSSH has been upgraded to 5.1p1.
1269 For many years, FreeBSD's version of OpenSSH preferred DSA
1270 over RSA for host and user authentication keys. With this
1271 upgrade, we've switched to the vendor's default of RSA over
1272 DSA. This may cause upgraded clients to warn about unknown
1273 host keys even for previously known hosts. Users should
1274 follow the usual procedure for verifying host keys before
1275 accepting the RSA key.
1277 This can be circumvented by setting the "HostKeyAlgorithms"
1278 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1281 Please note that the sequence of keys offered for
1282 authentication has been changed as well. You may want to
1283 specify IdentityFile in a different order to revert this
1287 The sio(4) driver has been removed from the i386 and amd64
1288 kernel configuration files. This means uart(4) is now the
1289 default serial port driver on those platforms as well.
1291 To prevent collisions with the sio(4) driver, the uart(4) driver
1292 uses different names for its device nodes. This means the
1293 onboard serial port will now most likely be called "ttyu0"
1294 instead of "ttyd0". You may need to reconfigure applications to
1295 use the new device names.
1297 When using the serial port as a boot console, be sure to update
1298 /boot/device.hints and /etc/ttys before booting the new kernel.
1299 If you forget to do so, you can still manually specify the hints
1300 at the loader prompt:
1302 set hint.uart.0.at="isa"
1303 set hint.uart.0.port="0x3F8"
1304 set hint.uart.0.flags="0x10"
1305 set hint.uart.0.irq="4"
1309 The gpt(8) utility has been removed. Use gpart(8) to partition
1313 The version that Linuxulator emulates was changed from 2.4.2
1314 to 2.6.16. If you experience any problems with Linux binaries
1315 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1316 if it fixes the problem contact emulation mailing list.
1319 ISDN4BSD (I4B) was removed from the src tree. You may need to
1320 update a your kernel configuration and remove relevant entries.
1323 I have checked in code to support multiple routing tables.
1324 See the man pages setfib(1) and setfib(2).
1325 This is a hopefully backwards compatible version,
1326 but to make use of it you need to compile your kernel
1327 with options ROUTETABLES=2 (or more up to 16).
1330 The 802.11 wireless support was redone to enable multi-bss
1331 operation on devices that are capable. The underlying device
1332 is no longer used directly but instead wlanX devices are
1333 cloned with ifconfig. This requires changes to rc.conf files.
1334 For example, change:
1335 ifconfig_ath0="WPA DHCP"
1338 ifconfig_wlan0="WPA DHCP"
1339 see rc.conf(5) for more details. In addition, mergemaster of
1340 /etc/rc.d is highly recommended. Simultaneous update of userland
1341 and kernel wouldn't hurt either.
1343 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1344 modules were merged into the base wlan module. All references
1345 to these modules (e.g. in kernel config files) must be removed.
1348 psm(4) has gained write(2) support in native operation level.
1349 Arbitrary commands can be written to /dev/psm%d and status can
1350 be read back from it. Therefore, an application is responsible
1351 for status validation and error recovery. It is a no-op in
1352 other operation levels.
1355 Support for KSE threading has been removed from the kernel. To
1356 run legacy applications linked against KSE libmap.conf may
1357 be used. The following libmap.conf may be used to ensure
1358 compatibility with any prior release:
1360 libpthread.so.1 libthr.so.1
1361 libpthread.so.2 libthr.so.2
1362 libkse.so.3 libthr.so.3
1365 The layout of struct vmspace has changed. This affects libkvm
1366 and any executables that link against libkvm and use the
1367 kvm_getprocs() function. In particular, but not exclusively,
1368 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1369 The effects are minimal, but it's advisable to upgrade world
1373 The latest em driver no longer has support in it for the
1374 82575 adapter, this is now moved to the igb driver. The
1375 split was done to make new features that are incompatible
1376 with older hardware easier to do.
1379 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1380 likewise the kernel option is now GEOM_LINUX_LVM.
1383 The default NFS mount mode has changed from UDP to TCP for
1384 increased reliability. If you rely on (insecurely) NFS
1385 mounting across a firewall you may need to update your
1389 Belatedly note the addition of m_collapse for compacting
1393 The fts(3) structures have been changed to use adequate
1394 integer types for their members and so to be able to cope
1395 with huge file trees. The old fts(3) ABI is preserved
1396 through symbol versioning in libc, so third-party binaries
1397 using fts(3) should still work, although they will not take
1398 advantage of the extended types. At the same time, some
1399 third-party software might fail to build after this change
1400 due to unportable assumptions made in its source code about
1401 fts(3) structure members. Such software should be fixed
1402 by its vendor or, in the worst case, in the ports tree.
1403 FreeBSD_version 800015 marks this change for the unlikely
1404 case that a portable fix is impossible.
1407 To upgrade to -current after this date, you must be running
1408 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1409 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1412 The ADAPTIVE_GIANT kernel option has been retired because its
1413 functionality is the default now.
1416 The AT keyboard emulation of sunkbd(4) has been turned on
1417 by default. In order to make the special symbols of the Sun
1418 keyboards driven by sunkbd(4) work under X these now have
1419 to be configured the same way as Sun USB keyboards driven
1420 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1422 Option "XkbLayout" "us"
1423 Option "XkbRules" "xorg"
1424 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1427 It has been decided that it is desirable to provide ABI
1428 backwards compatibility to the FreeBSD 4/5/6 versions of the
1429 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1430 broken with the introduction of PCI domain support (see the
1431 20070930 entry). Unfortunately, this required the ABI of
1432 PCIOCGETCONF to be broken again in order to be able to
1433 provide backwards compatibility to the old version of that
1434 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1435 again. As for prominent ports this affects neither pciutils
1436 nor xorg-server this time, the hal port needs to be rebuilt
1440 The misnamed kthread_create() and friends have been renamed
1441 to kproc_create() etc. Many of the callers already
1442 used kproc_start()..
1443 I will return kthread_create() and friends in a while
1444 with implementations that actually create threads, not procs.
1445 Renaming corresponds with version 800002.
1454 Avoid using make -j when upgrading. While generally safe, there are
1455 sometimes problems using -j to upgrade. If your upgrade fails with
1456 -j, please try again without -j. From time to time in the past there
1457 have been problems using -j with buildworld and/or installworld. This
1458 is especially true when upgrading between "distant" versions (eg one
1459 that cross a major release boundary or several minor releases, or when
1460 several months have passed on the -current branch).
1462 Sometimes, obscure build problems are the result of environment
1463 poisoning. This can happen because the make utility reads its
1464 environment when searching for values for global variables. To run
1465 your build attempts in an "environmental clean room", prefix all make
1466 commands with 'env -i '. See the env(1) manual page for more details.
1468 When upgrading from one major version to another it is generally best
1469 to upgrade to the latest code in the currently installed branch first,
1470 then do an upgrade to the new branch. This is the best-tested upgrade
1471 path, and has the highest probability of being successful. Please try
1472 this approach before reporting problems with a major version upgrade.
1476 When upgrading the boot ZFS pool to a new version, always follow
1479 1.) recompile and reinstall the ZFS boot loader and boot block
1480 (this is part of "make buildworld" and "make installworld")
1482 2.) update the ZFS boot block on your boot drive
1484 The following example updates the ZFS boot block on the first
1485 partition (freebsd-boot) of a GPT partitioned drive ad0:
1486 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1488 Non-boot pools do not need these updates.
1492 If you are updating from a prior version of FreeBSD (even one just
1493 a few days old), you should follow this procedure. It is the most
1494 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1496 make kernel-toolchain
1497 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1498 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1500 To test a kernel once
1501 ---------------------
1502 If you just want to boot a kernel once (because you are not sure
1503 if it works, or if you want to boot a known bad kernel to provide
1504 debugging information) run
1505 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1506 nextboot -k testkernel
1508 To just build a kernel when you know that it won't mess you up
1509 --------------------------------------------------------------
1510 This assumes you are already running a CURRENT system. Replace
1511 ${arch} with the architecture of your machine (e.g. "i386",
1512 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1514 cd src/sys/${arch}/conf
1515 config KERNEL_NAME_HERE
1516 cd ../compile/KERNEL_NAME_HERE
1521 If this fails, go to the "To build a kernel" section.
1523 To rebuild everything and install it on the current system.
1524 -----------------------------------------------------------
1525 # Note: sometimes if you are running current you gotta do more than
1526 # is listed here if you are upgrading from a really old current.
1528 <make sure you have good level 0 dumps>
1530 make kernel KERNCONF=YOUR_KERNEL_HERE
1532 <reboot in single user> [3]
1540 To cross-install current onto a separate partition
1541 --------------------------------------------------
1542 # In this approach we use a separate partition to hold
1543 # current's root, 'usr', and 'var' directories. A partition
1544 # holding "/", "/usr" and "/var" should be about 2GB in
1547 <make sure you have good level 0 dumps>
1550 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1551 <maybe newfs current's root partition>
1552 <mount current's root partition on directory ${CURRENT_ROOT}>
1553 make installworld DESTDIR=${CURRENT_ROOT}
1554 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1555 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1556 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1557 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1558 <reboot into current>
1559 <do a "native" rebuild/install as described in the previous section>
1560 <maybe install compatibility libraries from ports/misc/compat*>
1564 To upgrade in-place from stable to current
1565 ----------------------------------------------
1566 <make sure you have good level 0 dumps>
1568 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1570 <reboot in single user> [3]
1577 Make sure that you've read the UPDATING file to understand the
1578 tweaks to various things you need. At this point in the life
1579 cycle of current, things change often and you are on your own
1580 to cope. The defaults can also change, so please read ALL of
1581 the UPDATING entries.
1583 Also, if you are tracking -current, you must be subscribed to
1584 freebsd-current@freebsd.org. Make sure that before you update
1585 your sources that you have read and understood all the recent
1586 messages there. If in doubt, please track -stable which has
1587 much fewer pitfalls.
1589 [1] If you have third party modules, such as vmware, you
1590 should disable them at this point so they don't crash your
1593 [3] From the bootblocks, boot -s, and then do
1598 adjkerntz -i # if CMOS is wall time
1599 Also, when doing a major release upgrade, it is required that
1600 you boot into single user mode to do the installworld.
1602 [4] Note: This step is non-optional. Failure to do this step
1603 can result in a significant reduction in the functionality of the
1604 system. Attempting to do it by hand is not recommended and those
1605 that pursue this avenue should read this file carefully, as well
1606 as the archives of freebsd-current and freebsd-hackers mailing lists
1607 for potential gotchas. The -U option is also useful to consider.
1608 See mergemaster(8) for more information.
1610 [5] Usually this step is a noop. However, from time to time
1611 you may need to do this if you get unknown user in the following
1612 step. It never hurts to do it all the time. You may need to
1613 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1614 install) after the buildworld before this step if you last updated
1615 from [78]-stable or 9-stable before 20130430.
1617 [6] This only deletes old files and directories. Old libraries
1618 can be deleted by "make delete-old-libs", but you have to make
1619 sure that no program is using those libraries anymore.
1621 [8] In order to have a kernel that can run the 4.x binaries needed to
1622 do an installworld, you must include the COMPAT_FREEBSD4 option in
1623 your kernel. Failure to do so may leave you with a system that is
1624 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1625 required to run the 5.x binaries on more recent kernels. And so on
1626 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1628 Make sure that you merge any new devices from GENERIC since the
1629 last time you updated your kernel config file.
1631 [9] When checking out sources, you must include the -P flag to have
1632 cvs prune empty directories.
1634 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1635 "?=" instead of the "=" assignment operator, so that buildworld can
1636 override the CPUTYPE if it needs to.
1638 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1639 not on the command line, or in /etc/make.conf. buildworld will
1640 warn if it is improperly defined.
1643 This file contains a list, in reverse chronological order, of major
1644 breakages in tracking -current. It is not guaranteed to be a complete
1645 list of such breakages, and only contains entries since October 10, 2007.
1646 If you need to see UPDATING entries from before that date, you will need
1647 to fetch an UPDATING file from an older FreeBSD release.
1649 Copyright information:
1651 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1653 Redistribution, publication, translation and use, with or without
1654 modification, in full or in part, in any form or format of this
1655 document are permitted without further permission from the author.
1657 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1658 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1659 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1660 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1661 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1662 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1663 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1664 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1665 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1666 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1667 POSSIBILITY OF SUCH DAMAGE.
1669 Contact Warner Losh if you have any questions about your use of