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 On i386 and amd64 systems, the onifconsole flag is now set by default
16 in /etc/ttys for ttyu0. This causes ttyu0 to be automatically enabled
17 as a login TTY if it is set in the bootloader as an active kernel
18 console. No changes in behavior should result otherwise. To revert to
19 the previous behavior, set ttyu0 to "off" in /etc/ttys.
22 Clang and llvm have been upgraded to 3.4.1 release.
25 Clang and llvm have been upgraded to 3.4 release.
28 The nve(4) driver for NVIDIA nForce MCP Ethernet adapters has
29 been deprecated and will not be part of FreeBSD 11.0 and later
30 releases. If you use this driver, please consider switching to
31 the nfe(4) driver instead.
34 The behavior of gss_pseudo_random() for the krb5 mechanism
35 has changed, for applications requesting a longer random string
36 than produced by the underlying enctype's pseudo-random() function.
37 In particular, the random string produced from a session key of
38 enctype aes256-cts-hmac-sha1-96 or aes256-cts-hmac-sha1-96 will
39 be different at the 17th octet and later, after this change.
40 The counter used in the PRF+ construction is now encoded as a
41 big-endian integer in accordance with RFC 4402.
42 __FreeBSD_version is bumped to 902505.
48 Behavior of devfs rules path matching has been changed.
49 Pattern is now always matched against fully qualified devfs
50 path and slash characters must be explicitly matched by
51 slashes in pattern (FNM_PATHNAME). Rulesets involving devfs
52 subdirectories must be reviewed.
55 hastctl(8)'s `status' command output changed to terse one-liner format.
56 Scripts using this should switch to `list' command or be rewritten.
59 Fix a bug that allowed a tracing process (e.g. gdb) to write
60 to a memory-mapped file in the traced process's address space
61 even if neither the traced process nor the tracing process had
62 write access to that file.
65 Added ZFS TRIM support which is enabled by default. To disable
66 ZFS TRIM support set vfs.zfs.trim.enabled=0 in loader.conf.
68 Creating new ZFS pools and adding new devices to existing pools
69 first performs a full device level TRIM which can take a significant
70 amount of time. The sysctl vfs.zfs.vdev.trim_on_init can be set to 0
71 to disable this behaviour.
73 ZFS TRIM requires the underlying device support BIO_DELETE which
74 is currently provided by methods such as ATA TRIM and SCSI UNMAP
75 via CAM, which are typically supported by SSD's.
77 Stats for ZFS TRIM can be monitored by looking at the sysctl's
78 under kstat.zfs.misc.zio_trim.
81 `list' command has been added to hastctl(8). For now, it is full
82 equivalent of `status' command.
83 WARNING: in the near future the output of hastctl's status command
84 will change to more terse format. If you use `hastctl status'
85 for parsing in your scripts, switch to `hastctl list'.
88 The mergemaster command now uses the default MAKEOBJDIRPREFIX
89 rather than creating it's own in the temporary directory in
90 order allow access to bootstrapped versions of tools such as
91 install and mtree. When upgrading from version of FreeBSD where
92 the install command does not support -l, you will need to
93 install a new mergemaster command if mergemaster -p is required.
94 This can be accomplished with the command (cd src/usr.sbin/mergemaster
97 Due to the use of the new -l option to install(1) during build
98 and install, you must take care not to directly set the INSTALL
99 make variable in your /etc/make.conf, /etc/src.conf, or on the
100 command line. If you wish to use the -C flag for all installs
101 you may be able to add INSTALL+=-C to /etc/make.conf or
105 Fix a bug that allows NFS clients to issue READDIR on files.
108 The install(1) option -M has changed meaning and now takes an
109 argument that is a file or path to append logs to. In the
110 unlikely event that -M was the last option on the command line
111 and the command line contained at least two files and a target
112 directory the first file will have logs appended to it. The -M
113 option served little practical purpose in the last decade so it's
114 used expected to be extremely rare.
117 A new compression method (lz4) has been merged to. Please refer to
118 zpool-features(7) for more information.
120 Please refer to the "ZFS notes" section of this file for information
121 on upgrading boot ZFS pools.
124 The VFS KBI was changed with the merge of several nullfs
125 optimizations and fixes. All filesystem modules must be
129 With the addition of auditdistd(8), a new auditdistd user is now
130 depended on during installworld. "mergemaster -p" can be used to add
131 the user prior to installworld, as documented in the handbook.
137 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
138 Starting with this version the old system of ZFS pool versioning
139 is superseded by "feature flags". This concept enables forward
140 compatibility against certain future changes in functionality of ZFS
141 pools. The first two read-only compatible "feature flags" for ZFS
142 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
143 For more information read the new zpool-features(7) manual page.
144 Please refer to the "ZFS notes" section of this file for information
145 on upgrading boot ZFS pools.
148 The commit introducing bsd.compiler.mk breaks the traditional
149 building of kernels before this point. Add -m ${SRC}/share/mk
150 (for the right value of SRC) to your command lines to work
151 around; update your useland to a point after this; or use the
152 buildkernel/installkernel top-level targets. See also 20120829.
155 The IPFIREWALL_FORWARD kernel option has been removed. Its
156 functionality now turned on by default.
159 The random(4) support for the VIA hardware random number
160 generator (`PADLOCK') is no longer enabled unconditionally.
161 Add the PADLOCK_RNG option in the custom kernel config if
162 needed. The GENERIC kernels on i386 and amd64 do include the
163 option, so the change only affects the custom kernel
167 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
168 the traditional method, you must update your system with an installworld
169 before the kernel will build. The documented make buildkernel/installkernel
170 interfaces (coupled with fresh make kernel-toolchain) continue to work.
173 The sparc64 ZFS loader has been changed to no longer try to auto-
174 detect ZFS providers based on diskN aliases but now requires these
175 to be explicitly listed in the OFW boot-device environment variable.
178 Now unix domain sockets behave "as expected" on nullfs(5). Previously
179 nullfs(5) did not pass through all behaviours to the underlying layer,
180 as a result if we bound to a socket on the lower layer we could connect
181 only to the lower path; if we bound to the upper layer we could connect
182 only to the upper path. The new behavior is one can connect to both the
183 lower and the upper paths regardless what layer path one binds to.
186 The acpi_wmi(4) status device /dev/wmistat has been renamed to
190 A new VOP_ADVISE() was added to support posix_fadvise(2). All
191 filesystem modules must be recompiled.
194 The interface of the VOP_VPTOCNP(9) changed, now the returned
195 vnode shall be referenced, previously it was required to be
196 only held. All in-tree filesystems are converted.
202 The broken amd(4) driver has been replaced with esp(4) in the amd64,
203 i386 and pc98 GENERIC kernel configuration files.
206 This commit modifies vfs_register() so that it uses a hash
207 calculation to set vfc_typenum, which is enabled by default.
208 The first time a system is booted after this change, the
209 vfc_typenum values will change for all file systems. The
210 main effect of this is a change to the NFS server file handles
211 for file systems that use vfc_typenum in their fsid, such as ZFS.
212 It will, however, prevent vfc_typenum from changing when file
213 systems are loaded in a different order for subsequent reboots.
214 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
215 until you are ready to remount all NFS clients after a reboot.
218 Bump the shared library version numbers for libraries that
219 do not use symbol versioning, have changed the ABI compared
220 to stable/8 and which shared library version was not bumped.
221 Done as part of 9.0-RELEASE cycle.
224 During the merge of Capsicum features, the fget(9) KPI was modified.
225 This may require the rebuilding of out-of-tree device drivers --
226 issues have been reported specifically with the nVidia device driver.
227 __FreeBSD_version is bumped to 900041.
229 Also, there is a period between 20110811 and 20110814 where the
230 special devices /dev/{stdin,stdout,stderr} did not work correctly.
231 Building world from a kernel during that window may not work.
234 The packet filter (pf) code has been updated to OpenBSD 4.5.
235 You need to update userland tools to be in sync with kernel.
236 This update breaks backward compatibility with earlier pfsync(4)
237 versions. Care must be taken when updating redundant firewall setups.
240 The following sysctls and tunables are retired on x86 platforms:
242 machdep.hlt_logical_cpus
243 The following sysctl is retired:
244 machdep.hyperthreading_allowed
245 The sysctls were supposed to provide a way to dynamically offline and
246 online selected CPUs on x86 platforms, but the implementation has not
247 been reliable especially with SCHED_ULE scheduler.
248 machdep.hyperthreading_allowed tunable is still available to ignore
249 hyperthreading CPUs at OS level.
250 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
251 where X is an APIC ID of a CPU. Be advised, though, that disabling
252 CPUs in non-uniform fashion will result in non-uniform topology and
253 may lead to sub-optimal system performance with SCHED_ULE, which is
257 cpumask_t type is retired and cpuset_t is used in order to describe
261 Changes to ifconfig(8) for dynamic address family detection mandate
262 that you are running a kernel of 20110525 or later. Make sure to
263 follow the update procedure to boot a new kernel before installing
267 Support for sun4v architecture is officially dropped
270 Several KPI breaking changes have been committed to the mii(4) layer,
271 the PHY drivers and consequently some Ethernet drivers using mii(4).
272 This means that miibus.ko and the modules of the affected Ethernet
273 drivers need to be recompiled.
275 Note to kernel developers: Given that the OUI bit reversion problem
276 was fixed as part of these changes all mii(4) commits related to OUIs,
277 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
278 handling, no longer can be merged verbatim to stable/8 and previous
282 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
283 into their kernel configurations along with 'device pci'.
286 The default NFS client is now the new NFS client, so fstype "newnfs"
287 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
288 Although mounts via fstype "nfs" will usually work without userland
289 changes, it is recommended that the mount(8) and mount_nfs(8)
290 commands be rebuilt from sources and that a link to mount_nfs called
291 mount_oldnfs be created. The new client is compiled into the
292 kernel with "options NFSCL" and this is needed for diskless root
293 file systems. The GENERIC kernel configs have been changed to use
294 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
295 To use the regular/old client, you can "mount -t oldnfs ...". For
296 a diskless root file system, you must also include a line like:
298 vfs.root.mountfrom="oldnfs:"
300 in the boot/loader.conf on the root fs on the NFS server to make
301 a diskless root fs use the old client.
304 The GENERIC kernels for all architectures now default to the new
305 CAM-based ATA stack. It means that all legacy ATA drivers were
306 removed and replaced by respective CAM drivers. If you are using
307 ATA device names in /etc/fstab or other places, make sure to update
308 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
309 where 'Y's are the sequential numbers starting from zero for each type
310 in order of detection, unless configured otherwise with tunables,
311 see cam(4)). There will be symbolic links created in /dev/ to map
312 old adX devices to the respective adaY. They should provide basic
313 compatibility for file systems mounting in most cases, but they do
314 not support old user-level APIs and do not have respective providers
315 in GEOM. Consider using updated management tools with new device names.
317 It is possible to load devices ahci, ata, siis and mvs as modules,
318 but option ATA_CAM should remain in kernel configuration to make ata
319 module work as CAM driver supporting legacy ATA controllers. Device ata
320 still can be used in modular fashion (atacore + ...). Modules atadisk
321 and atapi* are not used and won't affect operation in ATA_CAM mode.
322 Note that to use CAM-based ATA kernel should include CAM devices
323 scbus, pass, da (or explicitly ada), cd and optionally others. All of
324 them are parts of the cam module.
326 ataraid(4) functionality is now supported by the RAID GEOM class.
327 To use it you can load geom_raid kernel module and use graid(8) tool
328 for management. Instead of /dev/arX device names, use /dev/raid/rX.
330 No kernel config options or code have been removed, so if a problem
331 arises, please report it and optionally revert to the old ATA stack.
332 In order to do it you can remove from the kernel config:
337 , and instead add back:
338 device atadisk # ATA disk drives
339 device ataraid # ATA RAID drives
340 device atapicd # ATAPI CDROM drives
341 device atapifd # ATAPI floppy drives
342 device atapist # ATAPI tape drives
345 The default NFS server has been changed to the new server, which
346 was referred to as the experimental server. If you need to switch
347 back to the old NFS server, you must now put the "-o" option on
348 both the mountd and nfsd commands. This can be done using the
349 mountd_flags and nfs_server_flags rc.conf variables until an
350 update to the rc scripts is committed, which is coming soon.
353 The GNU Objective-C runtime library (libobjc), and other Objective-C
354 related components have been removed from the base system. If you
355 require an Objective-C library, please use one of the available ports.
358 ath(4) has been split into bus- and device- modules. if_ath contains
359 the HAL, the TX rate control and the network device code. if_ath_pci
360 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
361 contains the AHB glue. Users need to load both if_ath_pci and if_ath
362 in order to use ath on everything else.
364 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
365 need to load if_ath and if_ath_pci for ath(4) operation.
368 As part of the replacement of sysinstall, the process of building
369 release media has changed significantly. For details, please re-read
370 release(7), which has been updated to reflect the new build process.
373 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
374 is the last available version under GPLv2. It brings a number of new
375 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
376 4.1 and SSE 4.2), better support for powerpc64, a number of new
377 directives, and lots of other small improvements. See the ChangeLog
378 file in contrib/binutils for the full details.
381 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
382 compliant, and will now use half of hash for authentication.
383 This will break interoperability with all stacks (including all
384 actual FreeBSD versions) who implement
385 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
387 The only workaround with such peers is to use another HMAC
388 algorithm for IPsec ("phase 2") authentication.
391 Remove the uio_yield prototype and symbol. This function has
392 been misnamed since it was introduced and should not be
393 globally exposed with this name. The equivalent functionality
394 is now available using kern_yield(curthread->td_user_pri).
395 The function remains undocumented.
398 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
399 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
400 for scalar sysctls is defined but disabled. Code that needs
401 UQUAD to pass the type checking that must compile on older
402 systems where the define is not present can check against
403 __FreeBSD_version >= 900030.
405 The system dialog(1) has been replaced with a new version previously
406 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
407 with the previous version, but the libdialog associated with it has
408 a largely incompatible API. As such, the original version of libdialog
409 will be kept temporarily as libodialog, until its base system consumers
410 are replaced or updated. Bump __FreeBSD_version to 900030.
413 If you are trying to run make universe on a -stable system, and you get
414 the following warning:
415 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
416 unknown. config(8) likely too old."
417 or something similar to it, then you must upgrade your -stable system
418 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
419 stable/8) or build the config from the latest stable/8 branch and
420 install it on your system.
422 Prior to this date, building a current universe on 8-stable system from
423 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
424 error in the first kernel build phase. A new config on those old
425 systems will fix that problem for older versions of -current.
428 The TCP stack has been modified to allow Khelp modules to interact with
429 it via helper hook points and store per-connection data in the TCP
430 control block. Bump __FreeBSD_version to 900029. User space tools that
431 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
435 Generic IEEE 802.3 annex 31B full duplex flow control support has been
436 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
437 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
438 to take advantage of it instead of using custom implementations. This
439 means that these drivers now no longer unconditionally advertise
440 support for flow control but only do so if flow control is a selected
441 media option. This was implemented in the generic support that way in
442 order to allow flow control to be switched on and off via ifconfig(8)
443 with the PHY specific default to typically off in order to protect
444 from unwanted effects. Consequently, if you used flow control with
445 one of the above mentioned drivers you now need to explicitly enable
447 ifconfig bge0 media auto mediaopt flowcontrol
449 Along with the above mentioned changes generic support for setting
450 1000baseT master mode also has been added and brgphy(4), ciphy(4),
451 e1000phy(4) as well as ip1000phy(4) have been converted to take
452 advantage of it. This means that these drivers now no longer take the
453 link0 parameter for selecting master mode but the master media option
454 has to be used instead, for example like in the following:
455 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
457 Selection of master mode now is also available with all other PHY
458 drivers supporting 1000baseT.
461 The TCP stack has received a significant update to add support for
462 modularised congestion control and generally improve the clarity of
463 congestion control decisions. Bump __FreeBSD_version to 900025. User
464 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
465 sockstat) need to be recompiled.
468 The man(1) utility has been replaced by a new version that no longer
469 uses /etc/manpath.config. Please consult man.conf(5) for how to
470 migrate local entries to the new format.
473 The copyright strings printed by login(1) and sshd(8) at the time of a
474 new connection have been removed to follow other operating systems and
478 A workaround for a fixed ld bug has been removed in kernel code,
479 so make sure that your system ld is built from sources after
480 revision 210245 from 2010-07-19 (r211583 if building head kernel
481 on stable/8, r211584 for stable/7; both from 2010-08-21).
482 A symptom of incorrect ld version is different addresses for
483 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
486 The $ipv6_prefer variable in rc.conf(5) has been split into
487 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
489 The $ip6addrctl_policy is a variable to choose a pre-defined
490 address selection policy set by ip6addrctl(8). A value
491 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
494 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
495 flag (see an entry of 20090926) is set on an interface with no
496 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
497 security reason. If you want IPv6 link-local address on all
498 interfaces by default, set this to "YES".
500 The old ipv6_prefer="YES" is equivalent to
501 ipv6_activate_all_interfaces="YES" and
502 ip6addrctl_policy="ipv6_prefer".
505 DTrace has grown support for userland tracing. Due to this, DTrace is
506 now i386 and amd64 only.
507 dtruss(1) is now installed by default on those systems and a new
508 kernel module is needed for userland tracing: fasttrap.
509 No changes to your kernel config file are necessary to enable
510 userland tracing, but you might consider adding 'STRIP=' and
511 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
512 to have informative userland stack traces in DTrace (ustack).
515 The acpi_aiboost(4) driver has been removed in favor of the new
516 aibs(4) driver. You should update your kernel configuration file.
519 BSD grep has been imported to the base system and it is built by
520 default. It is completely BSD licensed, highly GNU-compatible, uses
521 less memory than its GNU counterpart and has a small codebase.
522 However, it is slower than its GNU counterpart, which is mostly
523 noticeable for larger searches, for smaller ones it is measurable
524 but not significant. The reason is complex, the most important factor
525 is that we lack a modern and efficient regex library and GNU
526 overcomes this by optimizing the searches internally. Future work
527 on improving the regex performance is planned, for the meantime,
528 users that need better performance, can build GNU grep instead by
529 setting the WITH_GNU_GREP knob.
532 Due to the import of powerpc64 support, all existing powerpc kernel
533 configuration files must be updated with a machine directive like this:
534 machine powerpc powerpc
536 In addition, an updated config(8) is required to build powerpc kernels
540 A new version of ZFS (version 15) has been merged to -HEAD.
541 This version uses a python library for the following subcommands:
542 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
543 For full functionality of these commands the following port must
544 be installed: sysutils/py-zfs
547 'vm_page's are now hashed by physical address to an array of mutexes.
548 Currently this is only used to serialize access to hold_count. Over
549 time the page queue mutex will be peeled away. This changes the size
550 of pmap on every architecture. And requires all callers of vm_page_hold
551 and vm_page_unhold to be updated.
554 WITH_CTF can now be specified in src.conf (not recommended, there
555 are some problems with static executables), make.conf (would also
556 affect ports which do not use GNU make and do not override the
557 compile targets) or in the kernel config (via "makeoptions
559 When WITH_CTF was specified there before this was silently ignored,
560 so make sure that WITH_CTF is not used in places which could lead
561 to unwanted behavior.
564 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
565 to allow 32-bit compatibility on non-x86 platforms. All kernel
566 configurations on amd64 and ia64 platforms using these options must
567 be modified accordingly.
570 The utmp user accounting database has been replaced with utmpx,
571 the user accounting interface standardized by POSIX.
572 Unfortunately the semantics of utmp and utmpx don't match,
573 making it practically impossible to support both interfaces.
574 The user accounting database is used by tools like finger(1),
575 last(1), talk(1), w(1) and ac(8).
577 All applications in the base system use utmpx. This means only
578 local binaries (e.g. from the ports tree) may still use these
579 utmp database files. These applications must be rebuilt to make
582 After the system has been upgraded, it is safe to remove the old
583 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
584 assuming their contents is of no importance anymore. Old wtmp
585 databases can only be used by last(1) and ac(8) after they have
586 been converted to the new format using wtmpcvt(1).
589 Introduce the kernel thread "deadlock resolver" (which can be enabled
590 via the DEADLKRES option, see NOTES for more details) and the
591 sleepq_type() function for sleepqueues.
594 The rc.firewall and rc.firewall6 were unified, and
595 rc.firewall6 and rc.d/ip6fw were removed.
596 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
597 variables are obsoleted. Instead, the following new rc
598 variables are added to rc.d/ipfw:
600 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
601 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
602 firewall_simple_onet_ipv6, firewall_trusted_ipv6
604 The meanings correspond to the relevant IPv4 variables.
610 The default terminal emulation for syscons(4) has been changed
611 from cons25 to xterm on all platforms except pc98. This means
612 that the /etc/ttys file needs to be updated to ensure correct
613 operation of applications on the console.
615 The terminal emulation style can be toggled per window by using
616 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
617 options can be used to change the compile-time default back to
620 To prevent graphical artifacts, make sure the TERM environment
621 variable is set to match the terminal emulation that is being
622 performed by syscons(4).
625 The layout of the structure ieee80211req_scan_result has changed.
626 Applications that require wireless scan results (e.g. ifconfig(8))
627 from net80211 need to be recompiled.
629 Applications such as wpa_supplicant(8) may require a full world
630 build without using NO_CLEAN in order to get synchronized with the
634 The iwn(4) driver has been updated to support the 5000 and 5150 series.
635 There's one kernel module for each firmware. Adding "device iwnfw"
636 to the kernel configuration file means including all three firmware
637 images inside the kernel. If you want to include just the one for
638 your wireless card, use the devices iwn4965fw, iwn5000fw or
642 The rc.d/network_ipv6, IPv6 configuration script has been integrated
643 into rc.d/netif. The changes are the following:
645 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
646 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
647 Note that both variables need the "inet6" keyword at the head.
649 Do not set $ipv6_network_interfaces manually if you do not
650 understand what you are doing. It is not needed in most cases.
652 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
655 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
656 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
658 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
659 all configured interfaces have "inet6 accept_rtadv" in the
660 $ifconfig_IF_ipv6. These are for backward compatibility.
662 3. A new variable $ipv6_prefer has been added. If NO, IPv6
663 functionality of interfaces with no corresponding
664 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
665 and the default address selection policy of ip6addrctl(8)
666 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
667 Note that if you want to configure IPv6 functionality on the
668 disabled interfaces after boot, first you need to clear the flag by
669 using ifconfig(8) like:
671 ifconfig em0 inet6 -ifdisabled
673 If YES, the default address selection policy is set as
676 The default value of $ipv6_prefer is NO.
678 4. If your system need to receive Router Advertisement messages,
679 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
680 scripts automatically invoke rtsol(8) when the interface becomes
681 UP. The Router Advertisement messages are used for SLAAC
682 (State-Less Address AutoConfiguration).
685 802.11s D3.03 support was committed. This is incompatible with the
686 previous code, which was based on D3.0.
689 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
690 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
691 control whether accepting Router Advertisement messages or not.
692 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
693 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
694 The ifconfig(8) utility now supports these flags.
697 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
698 mount(8) and -a option for df(1) to see them.
701 The old tunable hw.bus.devctl_disable has been superseded by
702 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
703 replaced by hw.bus.devctl_queue=0. The default for this new tunable
707 Remove the option STOP_NMI. The default action is now to use NMI only
708 for KDB via the newly introduced function stop_cpus_hard() and
709 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
712 The stable/8 branch created in subversion. This corresponds to the
713 RELENG_8 branch in CVS.
716 Bump the shared library version numbers for all libraries that do not
717 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
718 __FreeBSD_version to 800105.
721 Due to changes in the implementation of virtual network stack support,
722 all network-related kernel modules must be recompiled. As this change
723 breaks the ABI, bump __FreeBSD_version to 800104.
726 The TOE interface to the TCP syncache has been modified to remove
727 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
728 The cxgb driver is the only TOE consumer affected by this change, and
729 needs to be recompiled along with the kernel. As this change breaks
730 the ABI, bump __FreeBSD_version to 800103.
733 Padding has been added to struct tcpcb, sackhint and tcpstat in
734 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
735 maintaining the ABI. However, this change breaks the ABI, so bump
736 __FreeBSD_version to 800102. User space tools that rely on the size of
737 any of these structs (e.g. sockstat) need to be recompiled.
740 The NFS_LEGACYRPC option has been removed along with the old kernel
741 RPC implementation that this option selected. Kernel configurations
742 may need to be adjusted.
745 The network interface device nodes at /dev/net/<interface> have been
746 removed. All ioctl operations can be performed the normal way using
747 routing sockets. The kqueue functionality can generally be replaced
748 with routing sockets.
751 The documentation from the FreeBSD Documentation Project (Handbook,
752 FAQ, etc.) is now installed via packages by sysinstall(8) and under
753 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
756 The ABI of various structures related to the SYSV IPC API have been
757 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
758 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
762 Layout of struct vnet has changed as routing related variables were
763 moved to their own Vimage module. Modules need to be recompiled. Bump
764 __FreeBSD_version to 800099.
767 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
768 respectively. As long as no more than 16 groups per process are used,
769 no changes should be visible. When more than 16 groups are used, old
770 binaries may fail if they call getgroups() or getgrouplist() with
771 statically sized storage. Recompiling will work around this, but
772 applications should be modified to use dynamically allocated storage
773 for group arrays as POSIX.1-2008 does not cap an implementation's
774 number of supported groups at NGROUPS_MAX+1 as previous versions did.
776 NFS and portalfs mounts may also be affected as the list of groups is
777 truncated to 16. Users of NFS who use more than 16 groups, should
778 take care that negative group permissions are not used on the exported
779 file systems as they will not be reliable unless a GSSAPI based
780 authentication method is used.
783 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
784 option compiles in the support for adaptive spinning for lockmgrs
785 which want to enable it. The lockinit() function now accepts the flag
786 LK_ADAPTIVE in order to make the lock object subject to adaptive
787 spinning when both held in write and read mode.
790 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
791 changed. User applications that use this ioctl need to be rebuilt.
794 The layout of struct thread has changed. Kernel and modules need to
798 The layout of structs ifnet, domain, protosw and vnet_net has changed.
799 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
802 window(1) has been removed from the base system. It can now be
803 installed from ports. The port is called misc/window.
806 The way we are storing and accessing `routing table' entries has
807 changed. Programs reading the FIB, like netstat, need to be
811 A new netisr implementation has been added for FreeBSD 8. Network
812 file system modules, such as igmp, ipdivert, and others, should be
814 Bump __FreeBSD_version to 800096.
817 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
821 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
822 Bump __FreeBSD_version to 800094.
825 Add mnt_xflag field to 'struct mount'. File system modules need to be
827 Bump __FreeBSD_version to 800093.
830 The compiling option ADAPTIVE_SX has been retired while it has been
831 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
832 The KPI for sx_init_flags() changes as accepting flags:
833 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
834 been introduced in order to handle the reversed logic.
835 Bump __FreeBSD_version to 800092.
838 Add support for hierarchical jails. Remove global securelevel.
839 Bump __FreeBSD_version to 800091.
842 The layout of struct vnet_net has changed, therefore modules
844 Bump __FreeBSD_version to 800090.
847 The newly imported zic(8) produces a new format in the output. Please
848 run tzsetup(8) to install the newly created data to /etc/localtime.
851 The sysctl tree for the usb stack has renamed from hw.usb2.* to
852 hw.usb.* and is now consistent again with previous releases.
855 802.11 monitor mode support was revised and driver api's were changed.
856 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
857 of DLT_IEEE802_11. No user-visible data structures were changed but
858 applications that use DLT_IEEE802_11 may require changes.
859 Bump __FreeBSD_version to 800088.
862 The layout of the following structs has changed: sysctl_oid,
863 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
864 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
865 panics may be experienced. World rebuild is required for
866 correctly checking networking state from userland.
867 Bump __FreeBSD_version to 800085.
870 MLDv2 and Source-Specific Multicast (SSM) have been merged
871 to the IPv6 stack. VIMAGE hooks are in but not yet used.
872 The implementation of SSM within FreeBSD's IPv6 stack closely
873 follows the IPv4 implementation.
875 For kernel developers:
877 * The most important changes are that the ip6_output() and
878 ip6_input() paths no longer take the IN6_MULTI_LOCK,
879 and this lock has been downgraded to a non-recursive mutex.
881 * As with the changes to the IPv4 stack to support SSM, filtering
882 of inbound multicast traffic must now be performed by transport
883 protocols within the IPv6 stack. This does not apply to TCP and
884 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
886 * The KPIs used by IPv6 multicast are similar to those used by
887 the IPv4 stack, with the following differences:
888 * im6o_mc_filter() is analogous to imo_multicast_filter().
889 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
890 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
891 * IN6_LOOKUP_MULTI() has been deprecated and removed.
892 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
893 for MLDv1 have an additional 'timer' argument which is used to
894 jitter the initial membership report for the solicited-node
895 multicast membership on-link.
896 * This is not strictly needed for MLDv2, which already jitters
897 its report transmissions. However, the 'timer' argument is
898 preserved in case MLDv1 is active on the interface.
900 * The KAME linked-list based IPv6 membership implementation has
901 been refactored to use a vector similar to that used by the IPv4
903 Code which maintains a list of its own multicast memberships
904 internally, e.g. carp, has been updated to reflect the new
907 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
908 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
909 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
910 implementation constraint which needs to be addressed in HEAD.
912 For application developers:
914 * The changes are broadly similar to those made for the IPv4
917 * The use of IPv4 and IPv6 multicast socket options on the same
918 socket, using mapped addresses, HAS NOT been tested or supported.
920 * There are a number of issues with the implementation of various
921 IPv6 multicast APIs which need to be resolved in the API surface
922 before the implementation is fully compatible with KAME userland
923 use, and these are mostly to do with interface index treatment.
925 * The literature available discusses the use of either the delta / ASM
926 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
927 using setsourcefilter(3)/getsourcefilter(3). For more information
928 please refer to RFC 3768, 'Socket Interface Extensions for
929 Multicast Source Filters'.
931 * Applications which use the published RFC 3678 APIs should be fine.
933 For systems administrators:
935 * The mtest(8) utility has been refactored to support IPv6, in
936 addition to IPv4. Interface addresses are no longer accepted
937 as arguments, their names must be used instead. The utility
938 will map the interface name to its first IPv4 address as
939 returned by getifaddrs(3).
941 * The ifmcstat(8) utility has also been updated to print the MLDv2
942 endpoint state and source filter lists via sysctl(3).
944 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
945 loopback of IPv6 multicast datagrams by default; it defaults to 1
946 to preserve the existing behaviour. Disabling multicast loopback is
947 recommended for optimal system performance.
949 * The IPv6 MROUTING code has been changed to examine this sysctl
950 instead of attempting to perform a group lookup before looping
951 back forwarded datagrams.
953 Bump __FreeBSD_version to 800084.
956 Implement low-level Bluetooth HCI API.
957 Bump __FreeBSD_version to 800083.
960 The layout of struct malloc_type, used by modules to register new
961 memory allocation types, has changed. Most modules will need to
962 be rebuilt or panics may be experienced.
963 Bump __FreeBSD_version to 800081.
966 Anticipate overflowing inp_flags - add inp_flags2.
967 This changes most offsets in inpcb, so checking v4 connection
968 state will require a world rebuild.
969 Bump __FreeBSD_version to 800080.
972 Add an llentry to struct route and struct route_in6. Modules
973 embedding a struct route will need to be recompiled.
974 Bump __FreeBSD_version to 800079.
977 The size of rt_metrics_lite and by extension rtentry has changed.
978 Networking administration apps will need to be recompiled.
979 The route command now supports show as an alias for get, weighting
980 of routes, sticky and nostick flags to alter the behavior of stateful
982 Bump __FreeBSD_version to 800078.
985 Do not use Giant for kbdmux(4) locking. This is wrong and
986 apparently causing more problems than it solves. This will
987 re-open the issue where interrupt handlers may race with
988 kbdmux(4) in polling mode. Typical symptoms include (but
989 not limited to) duplicated and/or missing characters when
990 low level console functions (such as gets) are used while
991 interrupts are enabled (for example geli password prompt,
992 mountroot prompt etc.). Disabling kbdmux(4) may help.
995 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
996 kernel modules referencing any of the above need to be recompiled.
997 Bump __FreeBSD_version to 800075.
1000 GEOM_PART has become the default partition slicer for storage devices,
1001 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
1002 introduces some changes:
1004 MSDOS/EBR: the devices created from MSDOS extended partition entries
1005 (EBR) can be named differently than with GEOM_MBR and are now symlinks
1006 to devices with offset-based names. fstabs may need to be modified.
1008 BSD: the "geometry does not match label" warning is harmless in most
1009 cases but it points to problems in file system misalignment with
1010 disk geometry. The "c" partition is now implicit, covers the whole
1011 top-level drive and cannot be (mis)used by users.
1013 General: Kernel dumps are now not allowed to be written to devices
1014 whose partition types indicate they are meant to be used for file
1015 systems (or, in case of MSDOS partitions, as something else than
1018 Most of these changes date approximately from 200812.
1021 The uscanner(4) driver has been removed from the kernel. This follows
1022 Linux removing theirs in 2.6 and making libusb the default interface
1023 (supported by sane).
1026 The multicast forwarding code has been cleaned up. netstat(1)
1027 only relies on KVM now for printing bandwidth upcall meters.
1028 The IPv4 and IPv6 modules are split into ip_mroute_mod and
1029 ip6_mroute_mod respectively. The config(5) options for statically
1030 compiling this code remain the same, i.e. 'options MROUTING'.
1033 Support for the IFF_NEEDSGIANT network interface flag has been
1034 removed, which means that non-MPSAFE network device drivers are no
1035 longer supported. In particular, if_ar, if_sr, and network device
1036 drivers from the old (legacy) USB stack can no longer be built or
1040 POSIX.1 Native Language Support (NLS) has been enabled in libc and
1041 a bunch of new language catalog files have also been added.
1042 This means that some common libc messages are now localized and
1043 they depend on the LC_MESSAGES environmental variable.
1046 The k8temp(4) driver has been renamed to amdtemp(4) since
1047 support for Family 10 and Family 11 CPU families was added.
1050 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1051 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1053 For kernel developers, the most important changes are that the
1054 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1055 and this lock has been downgraded to a non-recursive mutex.
1057 Transport protocols (UDP, Raw IP) are now responsible for filtering
1058 inbound multicast traffic according to group membership and source
1059 filters. The imo_multicast_filter() KPI exists for this purpose.
1060 Transports which do not use multicast (SCTP, TCP) already reject
1061 multicast by default. Forwarding and receive performance may improve
1062 as a mutex acquisition is no longer needed in the ip_input()
1063 low-level input path. in_addmulti() and in_delmulti() are shimmed
1064 to new KPIs which exist to support SSM in-kernel.
1066 For application developers, it is recommended that loopback of
1067 multicast datagrams be disabled for best performance, as this
1068 will still cause the lock to be taken for each looped-back
1069 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1070 be tuned to 0 to disable loopback by default; it defaults to 1
1071 to preserve the existing behaviour.
1073 For systems administrators, to obtain best performance with
1074 multicast reception and multiple groups, it is always recommended
1075 that a card with a suitably precise hash filter is used. Hash
1076 collisions will still result in the lock being taken within the
1077 transport protocol input path to check group membership.
1079 If deploying FreeBSD in an environment with IGMP snooping switches,
1080 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1081 enabled; this forces 224.0.0.0/24 group membership to be announced
1084 The size of 'struct igmpstat' has changed; netstat needs to be
1085 recompiled to reflect this.
1086 Bump __FreeBSD_version to 800070.
1089 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1090 updated to use it. This requires a buildworld/installworld in order to
1091 update the library and dependencies (usbconfig, etc). Its advisable to
1092 rebuild all ports which uses libusb. More specific directions are given
1093 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1094 libusb are no longer required and can be removed.
1097 A workaround is committed to allow the creation of System V shared
1098 memory segment of size > 2 GB on the 64-bit architectures.
1099 Due to a limitation of the existing ABI, the shm_segsz member
1100 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1101 wrong for large segments. Note that limits must be explicitly
1102 raised to allow such segments to be created.
1105 The layout of struct ifnet has changed, requiring a rebuild of all
1106 network device driver modules.
1109 The /dev handling for the new USB stack has changed, a
1110 buildworld/installworld is required for libusb20.
1113 The new USB2 stack has now been permanently moved in and all kernel and
1114 module names reverted to their previous values (eg, usb, ehci, ohci,
1115 ums, ...). The old usb stack can be compiled in by prefixing the name
1116 with the letter 'o', the old usb modules have been removed.
1117 Updating entry 20090216 for xorg and 20090215 for libmap may still
1121 The rc.conf(5) option if_up_delay has been renamed to
1122 defaultroute_delay to better reflect its purpose. If you have
1123 customized this setting in /etc/rc.conf you need to update it to
1127 xorg 7.4 wants to configure its input devices via hald which does not
1128 yet work with USB2. If the keyboard/mouse does not work in xorg then
1130 Option "AllowEmptyInput" "off"
1131 to your ServerLayout section. This will cause X to use the configured
1132 kbd and mouse sections from your xorg.conf.
1135 The GENERIC kernels for all architectures now default to the new USB2
1136 stack. No kernel config options or code have been removed so if a
1137 problem arises please report it and optionally revert to the old USB
1138 stack. If you are loading USB kernel modules or have a custom kernel
1139 that includes GENERIC then ensure that usb names are also changed over,
1140 eg uftdi -> usb2_serial_ftdi.
1142 Older programs linked against the ports libusb 0.1 need to be
1143 redirected to the new stack's libusb20. /etc/libmap.conf can
1145 # Map old usb library to new one for usb2 stack
1146 libusb-0.1.so.8 libusb20.so.1
1149 All USB ethernet devices now attach as interfaces under the name ueN
1150 (eg. ue0). This is to provide a predictable name as vendors often
1151 change usb chipsets in a product without notice.
1154 The ichsmb(4) driver has been changed to require SMBus slave
1155 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1156 All of the other SMBus controller drivers require left-justified
1157 slave addresses, so this change makes all the drivers provide the
1161 INET6 statistics (struct ip6stat) was updated.
1162 netstat(1) needs to be recompiled.
1165 NTFS has been removed from GENERIC kernel on amd64 to match
1166 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1167 will load ntfs.ko module automatically when NTFS support is
1168 actually needed, unless ntfs.ko is not installed or security
1169 level prohibits loading kernel modules. If either is the case,
1170 "options NTFS" has to be added into kernel config.
1173 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1174 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1175 800061. User space tools that rely on the size of struct tcpcb in
1176 tcp_var.h (e.g. sockstat) need to be recompiled.
1179 ng_tty(4) module updated to match the new TTY subsystem.
1180 Due to API change, user-level applications must be updated.
1181 New API support added to mpd5 CVS and expected to be present
1182 in next mpd5.3 release.
1185 With __FreeBSD_version 800060 the makefs tool is part of
1186 the base system (it was a port).
1189 The afdata and ifnet locks have been changed from mutexes to
1190 rwlocks, network modules will need to be re-compiled.
1193 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1194 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1195 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1196 architecture and 40 bytes on 64-bit architecture. The userland
1197 applications "arp" and "ndp" have been updated accordingly.
1198 The output from "netstat -r" shows only routing entries and
1199 none of the L2 information.
1202 __FreeBSD_version 800057 marks the switchover from the
1203 binary ath hal to source code. Users must add the line:
1205 options AH_SUPPORT_AR5416
1207 to their kernel config files when specifying:
1211 The ath_hal module no longer exists; the code is now compiled
1212 together with the driver in the ath module. It is now
1213 possible to tailor chip support (i.e. reduce the set of chips
1214 and thereby the code size); consult ath_hal(4) for details.
1217 __FreeBSD_version 800054 adds memory barriers to
1218 <machine/atomic.h>, new interfaces to ifnet to facilitate
1219 multiple hardware transmit queues for cards that support
1220 them, and a lock-less ring-buffer implementation to
1221 enable drivers to more efficiently manage queueing of
1225 A new version of ZFS (version 13) has been merged to -HEAD.
1226 This version has zpool attribute "listsnapshots" off by
1227 default, which means "zfs list" does not show snapshots,
1228 and is the same as Solaris behavior.
1231 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1234 The uhci, ohci, ehci and slhci USB Host controller drivers have
1235 been put into separate modules. If you load the usb module
1236 separately through loader.conf you will need to load the
1237 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1238 controller add the following to loader.conf:
1244 The ABI used by the PMC toolset has changed. Please keep
1245 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1249 atapci kernel module now includes only generic PCI ATA
1250 driver. AHCI driver moved to ataahci kernel module.
1251 All vendor-specific code moved into separate kernel modules:
1252 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1253 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1254 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1255 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1258 The TTY subsystem of the kernel has been replaced by a new
1259 implementation, which provides better scalability and an
1260 improved driver model. Most common drivers have been migrated to
1261 the new TTY subsystem, while others have not. The following
1262 drivers have not yet been ported to the new TTY layer:
1265 cy, digi, rc, rp, sio
1271 ng_h4, ng_tty, ppp, sl, snp
1273 Adding these drivers to your kernel configuration file shall
1274 cause compilation to fail.
1277 ntpd has been upgraded to 4.2.4p5.
1280 OpenSSH has been upgraded to 5.1p1.
1282 For many years, FreeBSD's version of OpenSSH preferred DSA
1283 over RSA for host and user authentication keys. With this
1284 upgrade, we've switched to the vendor's default of RSA over
1285 DSA. This may cause upgraded clients to warn about unknown
1286 host keys even for previously known hosts. Users should
1287 follow the usual procedure for verifying host keys before
1288 accepting the RSA key.
1290 This can be circumvented by setting the "HostKeyAlgorithms"
1291 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1294 Please note that the sequence of keys offered for
1295 authentication has been changed as well. You may want to
1296 specify IdentityFile in a different order to revert this
1300 The sio(4) driver has been removed from the i386 and amd64
1301 kernel configuration files. This means uart(4) is now the
1302 default serial port driver on those platforms as well.
1304 To prevent collisions with the sio(4) driver, the uart(4) driver
1305 uses different names for its device nodes. This means the
1306 onboard serial port will now most likely be called "ttyu0"
1307 instead of "ttyd0". You may need to reconfigure applications to
1308 use the new device names.
1310 When using the serial port as a boot console, be sure to update
1311 /boot/device.hints and /etc/ttys before booting the new kernel.
1312 If you forget to do so, you can still manually specify the hints
1313 at the loader prompt:
1315 set hint.uart.0.at="isa"
1316 set hint.uart.0.port="0x3F8"
1317 set hint.uart.0.flags="0x10"
1318 set hint.uart.0.irq="4"
1322 The gpt(8) utility has been removed. Use gpart(8) to partition
1326 The version that Linuxulator emulates was changed from 2.4.2
1327 to 2.6.16. If you experience any problems with Linux binaries
1328 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1329 if it fixes the problem contact emulation mailing list.
1332 ISDN4BSD (I4B) was removed from the src tree. You may need to
1333 update a your kernel configuration and remove relevant entries.
1336 I have checked in code to support multiple routing tables.
1337 See the man pages setfib(1) and setfib(2).
1338 This is a hopefully backwards compatible version,
1339 but to make use of it you need to compile your kernel
1340 with options ROUTETABLES=2 (or more up to 16).
1343 The 802.11 wireless support was redone to enable multi-bss
1344 operation on devices that are capable. The underlying device
1345 is no longer used directly but instead wlanX devices are
1346 cloned with ifconfig. This requires changes to rc.conf files.
1347 For example, change:
1348 ifconfig_ath0="WPA DHCP"
1351 ifconfig_wlan0="WPA DHCP"
1352 see rc.conf(5) for more details. In addition, mergemaster of
1353 /etc/rc.d is highly recommended. Simultaneous update of userland
1354 and kernel wouldn't hurt either.
1356 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1357 modules were merged into the base wlan module. All references
1358 to these modules (e.g. in kernel config files) must be removed.
1361 psm(4) has gained write(2) support in native operation level.
1362 Arbitrary commands can be written to /dev/psm%d and status can
1363 be read back from it. Therefore, an application is responsible
1364 for status validation and error recovery. It is a no-op in
1365 other operation levels.
1368 Support for KSE threading has been removed from the kernel. To
1369 run legacy applications linked against KSE libmap.conf may
1370 be used. The following libmap.conf may be used to ensure
1371 compatibility with any prior release:
1373 libpthread.so.1 libthr.so.1
1374 libpthread.so.2 libthr.so.2
1375 libkse.so.3 libthr.so.3
1378 The layout of struct vmspace has changed. This affects libkvm
1379 and any executables that link against libkvm and use the
1380 kvm_getprocs() function. In particular, but not exclusively,
1381 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1382 The effects are minimal, but it's advisable to upgrade world
1386 The latest em driver no longer has support in it for the
1387 82575 adapter, this is now moved to the igb driver. The
1388 split was done to make new features that are incompatible
1389 with older hardware easier to do.
1392 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1393 likewise the kernel option is now GEOM_LINUX_LVM.
1396 The default NFS mount mode has changed from UDP to TCP for
1397 increased reliability. If you rely on (insecurely) NFS
1398 mounting across a firewall you may need to update your
1402 Belatedly note the addition of m_collapse for compacting
1406 The fts(3) structures have been changed to use adequate
1407 integer types for their members and so to be able to cope
1408 with huge file trees. The old fts(3) ABI is preserved
1409 through symbol versioning in libc, so third-party binaries
1410 using fts(3) should still work, although they will not take
1411 advantage of the extended types. At the same time, some
1412 third-party software might fail to build after this change
1413 due to unportable assumptions made in its source code about
1414 fts(3) structure members. Such software should be fixed
1415 by its vendor or, in the worst case, in the ports tree.
1416 FreeBSD_version 800015 marks this change for the unlikely
1417 case that a portable fix is impossible.
1420 To upgrade to -current after this date, you must be running
1421 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1422 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1425 The ADAPTIVE_GIANT kernel option has been retired because its
1426 functionality is the default now.
1429 The AT keyboard emulation of sunkbd(4) has been turned on
1430 by default. In order to make the special symbols of the Sun
1431 keyboards driven by sunkbd(4) work under X these now have
1432 to be configured the same way as Sun USB keyboards driven
1433 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1435 Option "XkbLayout" "us"
1436 Option "XkbRules" "xorg"
1437 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1440 It has been decided that it is desirable to provide ABI
1441 backwards compatibility to the FreeBSD 4/5/6 versions of the
1442 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1443 broken with the introduction of PCI domain support (see the
1444 20070930 entry). Unfortunately, this required the ABI of
1445 PCIOCGETCONF to be broken again in order to be able to
1446 provide backwards compatibility to the old version of that
1447 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1448 again. As for prominent ports this affects neither pciutils
1449 nor xorg-server this time, the hal port needs to be rebuilt
1453 The misnamed kthread_create() and friends have been renamed
1454 to kproc_create() etc. Many of the callers already
1455 used kproc_start()..
1456 I will return kthread_create() and friends in a while
1457 with implementations that actually create threads, not procs.
1458 Renaming corresponds with version 800002.
1467 Avoid using make -j when upgrading. While generally safe, there are
1468 sometimes problems using -j to upgrade. If your upgrade fails with
1469 -j, please try again without -j. From time to time in the past there
1470 have been problems using -j with buildworld and/or installworld. This
1471 is especially true when upgrading between "distant" versions (eg one
1472 that cross a major release boundary or several minor releases, or when
1473 several months have passed on the -current branch).
1475 Sometimes, obscure build problems are the result of environment
1476 poisoning. This can happen because the make utility reads its
1477 environment when searching for values for global variables. To run
1478 your build attempts in an "environmental clean room", prefix all make
1479 commands with 'env -i '. See the env(1) manual page for more details.
1481 When upgrading from one major version to another it is generally best
1482 to upgrade to the latest code in the currently installed branch first,
1483 then do an upgrade to the new branch. This is the best-tested upgrade
1484 path, and has the highest probability of being successful. Please try
1485 this approach before reporting problems with a major version upgrade.
1489 When upgrading the boot ZFS pool to a new version, always follow
1492 1.) recompile and reinstall the ZFS boot loader and boot block
1493 (this is part of "make buildworld" and "make installworld")
1495 2.) update the ZFS boot block on your boot drive
1497 The following example updates the ZFS boot block on the first
1498 partition (freebsd-boot) of a GPT partitioned drive ad0:
1499 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1501 Non-boot pools do not need these updates.
1505 If you are updating from a prior version of FreeBSD (even one just
1506 a few days old), you should follow this procedure. It is the most
1507 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1509 make kernel-toolchain
1510 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1511 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1513 To test a kernel once
1514 ---------------------
1515 If you just want to boot a kernel once (because you are not sure
1516 if it works, or if you want to boot a known bad kernel to provide
1517 debugging information) run
1518 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1519 nextboot -k testkernel
1521 To just build a kernel when you know that it won't mess you up
1522 --------------------------------------------------------------
1523 This assumes you are already running a CURRENT system. Replace
1524 ${arch} with the architecture of your machine (e.g. "i386",
1525 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1527 cd src/sys/${arch}/conf
1528 config KERNEL_NAME_HERE
1529 cd ../compile/KERNEL_NAME_HERE
1534 If this fails, go to the "To build a kernel" section.
1536 To rebuild everything and install it on the current system.
1537 -----------------------------------------------------------
1538 # Note: sometimes if you are running current you gotta do more than
1539 # is listed here if you are upgrading from a really old current.
1541 <make sure you have good level 0 dumps>
1543 make kernel KERNCONF=YOUR_KERNEL_HERE
1545 <reboot in single user> [3]
1553 To cross-install current onto a separate partition
1554 --------------------------------------------------
1555 # In this approach we use a separate partition to hold
1556 # current's root, 'usr', and 'var' directories. A partition
1557 # holding "/", "/usr" and "/var" should be about 2GB in
1560 <make sure you have good level 0 dumps>
1563 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1564 <maybe newfs current's root partition>
1565 <mount current's root partition on directory ${CURRENT_ROOT}>
1566 make installworld DESTDIR=${CURRENT_ROOT}
1567 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1568 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1569 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1570 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1571 <reboot into current>
1572 <do a "native" rebuild/install as described in the previous section>
1573 <maybe install compatibility libraries from ports/misc/compat*>
1577 To upgrade in-place from stable to current
1578 ----------------------------------------------
1579 <make sure you have good level 0 dumps>
1581 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1583 <reboot in single user> [3]
1590 Make sure that you've read the UPDATING file to understand the
1591 tweaks to various things you need. At this point in the life
1592 cycle of current, things change often and you are on your own
1593 to cope. The defaults can also change, so please read ALL of
1594 the UPDATING entries.
1596 Also, if you are tracking -current, you must be subscribed to
1597 freebsd-current@freebsd.org. Make sure that before you update
1598 your sources that you have read and understood all the recent
1599 messages there. If in doubt, please track -stable which has
1600 much fewer pitfalls.
1602 [1] If you have third party modules, such as vmware, you
1603 should disable them at this point so they don't crash your
1606 [3] From the bootblocks, boot -s, and then do
1611 adjkerntz -i # if CMOS is wall time
1612 Also, when doing a major release upgrade, it is required that
1613 you boot into single user mode to do the installworld.
1615 [4] Note: This step is non-optional. Failure to do this step
1616 can result in a significant reduction in the functionality of the
1617 system. Attempting to do it by hand is not recommended and those
1618 that pursue this avenue should read this file carefully, as well
1619 as the archives of freebsd-current and freebsd-hackers mailing lists
1620 for potential gotchas. The -U option is also useful to consider.
1621 See mergemaster(8) for more information.
1623 [5] Usually this step is a noop. However, from time to time
1624 you may need to do this if you get unknown user in the following
1625 step. It never hurts to do it all the time. You may need to
1626 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1627 install) after the buildworld before this step if you last updated
1628 from [78]-stable or 9-stable before 20130430.
1630 [6] This only deletes old files and directories. Old libraries
1631 can be deleted by "make delete-old-libs", but you have to make
1632 sure that no program is using those libraries anymore.
1634 [8] In order to have a kernel that can run the 4.x binaries needed to
1635 do an installworld, you must include the COMPAT_FREEBSD4 option in
1636 your kernel. Failure to do so may leave you with a system that is
1637 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1638 required to run the 5.x binaries on more recent kernels. And so on
1639 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1641 Make sure that you merge any new devices from GENERIC since the
1642 last time you updated your kernel config file.
1644 [9] When checking out sources, you must include the -P flag to have
1645 cvs prune empty directories.
1647 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1648 "?=" instead of the "=" assignment operator, so that buildworld can
1649 override the CPUTYPE if it needs to.
1651 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1652 not on the command line, or in /etc/make.conf. buildworld will
1653 warn if it is improperly defined.
1656 This file contains a list, in reverse chronological order, of major
1657 breakages in tracking -current. It is not guaranteed to be a complete
1658 list of such breakages, and only contains entries since October 10, 2007.
1659 If you need to see UPDATING entries from before that date, you will need
1660 to fetch an UPDATING file from an older FreeBSD release.
1662 Copyright information:
1664 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1666 Redistribution, publication, translation and use, with or without
1667 modification, in full or in part, in any form or format of this
1668 document are permitted without further permission from the author.
1670 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1671 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1672 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1673 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1674 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1675 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1676 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1677 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1678 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1679 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1680 POSSIBILITY OF SUCH DAMAGE.
1682 Contact Warner Losh if you have any questions about your use of