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 behavior of gss_pseudo_random() for the krb5 mechanism
16 has changed, for applications requesting a longer random string
17 than produced by the underlying enctype's pseudo-random() function.
18 In particular, the random string produced from a session key of
19 enctype aes256-cts-hmac-sha1-96 or aes256-cts-hmac-sha1-96 will
20 be different at the 17th octet and later, after this change.
21 The counter used in the PRF+ construction is now encoded as a
22 big-endian integer in accordance with RFC 4402.
23 __FreeBSD_version is bumped to 902505.
29 Behavior of devfs rules path matching has been changed.
30 Pattern is now always matched against fully qualified devfs
31 path and slash characters must be explicitly matched by
32 slashes in pattern (FNM_PATHNAME). Rulesets involving devfs
33 subdirectories must be reviewed.
36 hastctl(8)'s `status' command output changed to terse one-liner format.
37 Scripts using this should switch to `list' command or be rewritten.
40 Fix a bug that allowed a tracing process (e.g. gdb) to write
41 to a memory-mapped file in the traced process's address space
42 even if neither the traced process nor the tracing process had
43 write access to that file.
46 Added ZFS TRIM support which is enabled by default. To disable
47 ZFS TRIM support set vfs.zfs.trim.enabled=0 in loader.conf.
49 Creating new ZFS pools and adding new devices to existing pools
50 first performs a full device level TRIM which can take a significant
51 amount of time. The sysctl vfs.zfs.vdev.trim_on_init can be set to 0
52 to disable this behaviour.
54 ZFS TRIM requires the underlying device support BIO_DELETE which
55 is currently provided by methods such as ATA TRIM and SCSI UNMAP
56 via CAM, which are typically supported by SSD's.
58 Stats for ZFS TRIM can be monitored by looking at the sysctl's
59 under kstat.zfs.misc.zio_trim.
62 `list' command has been added to hastctl(8). For now, it is full
63 equivalent of `status' command.
64 WARNING: in the near future the output of hastctl's status command
65 will change to more terse format. If you use `hastctl status'
66 for parsing in your scripts, switch to `hastctl list'.
69 The mergemaster command now uses the default MAKEOBJDIRPREFIX
70 rather than creating it's own in the temporary directory in
71 order allow access to bootstrapped versions of tools such as
72 install and mtree. When upgrading from version of FreeBSD where
73 the install command does not support -l, you will need to
74 install a new mergemaster command if mergemaster -p is required.
75 This can be accomplished with the command (cd src/usr.sbin/mergemaster
78 Due to the use of the new -l option to install(1) during build
79 and install, you must take care not to directly set the INSTALL
80 make variable in your /etc/make.conf, /etc/src.conf, or on the
81 command line. If you wish to use the -C flag for all installs
82 you may be able to add INSTALL+=-C to /etc/make.conf or
86 Fix a bug that allows NFS clients to issue READDIR on files.
89 The install(1) option -M has changed meaning and now takes an
90 argument that is a file or path to append logs to. In the
91 unlikely event that -M was the last option on the command line
92 and the command line contained at least two files and a target
93 directory the first file will have logs appended to it. The -M
94 option served little practical purpose in the last decade so it's
95 used expected to be extremely rare.
98 A new compression method (lz4) has been merged to. Please refer to
99 zpool-features(7) for more information.
101 Please refer to the "ZFS notes" section of this file for information
102 on upgrading boot ZFS pools.
105 The VFS KBI was changed with the merge of several nullfs
106 optimizations and fixes. All filesystem modules must be
110 With the addition of auditdistd(8), a new auditdistd user is now
111 depended on during installworld. "mergemaster -p" can be used to add
112 the user prior to installworld, as documented in the handbook.
118 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
119 Starting with this version the old system of ZFS pool versioning
120 is superseded by "feature flags". This concept enables forward
121 compatibility against certain future changes in functionality of ZFS
122 pools. The first two read-only compatible "feature flags" for ZFS
123 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
124 For more information read the new zpool-features(7) manual page.
125 Please refer to the "ZFS notes" section of this file for information
126 on upgrading boot ZFS pools.
129 The commit introducing bsd.compiler.mk breaks the traditional
130 building of kernels before this point. Add -m ${SRC}/share/mk
131 (for the right value of SRC) to your command lines to work
132 around; update your useland to a point after this; or use the
133 buildkernel/installkernel top-level targets. See also 20120829.
136 The IPFIREWALL_FORWARD kernel option has been removed. Its
137 functionality now turned on by default.
140 The random(4) support for the VIA hardware random number
141 generator (`PADLOCK') is no longer enabled unconditionally.
142 Add the PADLOCK_RNG option in the custom kernel config if
143 needed. The GENERIC kernels on i386 and amd64 do include the
144 option, so the change only affects the custom kernel
148 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
149 the traditional method, you must update your system with an installworld
150 before the kernel will build. The documented make buildkernel/installkernel
151 interfaces (coupled with fresh make kernel-toolchain) continue to work.
154 The sparc64 ZFS loader has been changed to no longer try to auto-
155 detect ZFS providers based on diskN aliases but now requires these
156 to be explicitly listed in the OFW boot-device environment variable.
159 Now unix domain sockets behave "as expected" on nullfs(5). Previously
160 nullfs(5) did not pass through all behaviours to the underlying layer,
161 as a result if we bound to a socket on the lower layer we could connect
162 only to the lower path; if we bound to the upper layer we could connect
163 only to the upper path. The new behavior is one can connect to both the
164 lower and the upper paths regardless what layer path one binds to.
167 The acpi_wmi(4) status device /dev/wmistat has been renamed to
171 A new VOP_ADVISE() was added to support posix_fadvise(2). All
172 filesystem modules must be recompiled.
175 The interface of the VOP_VPTOCNP(9) changed, now the returned
176 vnode shall be referenced, previously it was required to be
177 only held. All in-tree filesystems are converted.
183 The broken amd(4) driver has been replaced with esp(4) in the amd64,
184 i386 and pc98 GENERIC kernel configuration files.
187 This commit modifies vfs_register() so that it uses a hash
188 calculation to set vfc_typenum, which is enabled by default.
189 The first time a system is booted after this change, the
190 vfc_typenum values will change for all file systems. The
191 main effect of this is a change to the NFS server file handles
192 for file systems that use vfc_typenum in their fsid, such as ZFS.
193 It will, however, prevent vfc_typenum from changing when file
194 systems are loaded in a different order for subsequent reboots.
195 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
196 until you are ready to remount all NFS clients after a reboot.
199 Bump the shared library version numbers for libraries that
200 do not use symbol versioning, have changed the ABI compared
201 to stable/8 and which shared library version was not bumped.
202 Done as part of 9.0-RELEASE cycle.
205 During the merge of Capsicum features, the fget(9) KPI was modified.
206 This may require the rebuilding of out-of-tree device drivers --
207 issues have been reported specifically with the nVidia device driver.
208 __FreeBSD_version is bumped to 900041.
210 Also, there is a period between 20110811 and 20110814 where the
211 special devices /dev/{stdin,stdout,stderr} did not work correctly.
212 Building world from a kernel during that window may not work.
215 The packet filter (pf) code has been updated to OpenBSD 4.5.
216 You need to update userland tools to be in sync with kernel.
217 This update breaks backward compatibility with earlier pfsync(4)
218 versions. Care must be taken when updating redundant firewall setups.
221 The following sysctls and tunables are retired on x86 platforms:
223 machdep.hlt_logical_cpus
224 The following sysctl is retired:
225 machdep.hyperthreading_allowed
226 The sysctls were supposed to provide a way to dynamically offline and
227 online selected CPUs on x86 platforms, but the implementation has not
228 been reliable especially with SCHED_ULE scheduler.
229 machdep.hyperthreading_allowed tunable is still available to ignore
230 hyperthreading CPUs at OS level.
231 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
232 where X is an APIC ID of a CPU. Be advised, though, that disabling
233 CPUs in non-uniform fashion will result in non-uniform topology and
234 may lead to sub-optimal system performance with SCHED_ULE, which is
238 cpumask_t type is retired and cpuset_t is used in order to describe
242 Changes to ifconfig(8) for dynamic address family detection mandate
243 that you are running a kernel of 20110525 or later. Make sure to
244 follow the update procedure to boot a new kernel before installing
248 Support for sun4v architecture is officially dropped
251 Several KPI breaking changes have been committed to the mii(4) layer,
252 the PHY drivers and consequently some Ethernet drivers using mii(4).
253 This means that miibus.ko and the modules of the affected Ethernet
254 drivers need to be recompiled.
256 Note to kernel developers: Given that the OUI bit reversion problem
257 was fixed as part of these changes all mii(4) commits related to OUIs,
258 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
259 handling, no longer can be merged verbatim to stable/8 and previous
263 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
264 into their kernel configurations along with 'device pci'.
267 The default NFS client is now the new NFS client, so fstype "newnfs"
268 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
269 Although mounts via fstype "nfs" will usually work without userland
270 changes, it is recommended that the mount(8) and mount_nfs(8)
271 commands be rebuilt from sources and that a link to mount_nfs called
272 mount_oldnfs be created. The new client is compiled into the
273 kernel with "options NFSCL" and this is needed for diskless root
274 file systems. The GENERIC kernel configs have been changed to use
275 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
276 To use the regular/old client, you can "mount -t oldnfs ...". For
277 a diskless root file system, you must also include a line like:
279 vfs.root.mountfrom="oldnfs:"
281 in the boot/loader.conf on the root fs on the NFS server to make
282 a diskless root fs use the old client.
285 The GENERIC kernels for all architectures now default to the new
286 CAM-based ATA stack. It means that all legacy ATA drivers were
287 removed and replaced by respective CAM drivers. If you are using
288 ATA device names in /etc/fstab or other places, make sure to update
289 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
290 where 'Y's are the sequential numbers starting from zero for each type
291 in order of detection, unless configured otherwise with tunables,
292 see cam(4)). There will be symbolic links created in /dev/ to map
293 old adX devices to the respective adaY. They should provide basic
294 compatibility for file systems mounting in most cases, but they do
295 not support old user-level APIs and do not have respective providers
296 in GEOM. Consider using updated management tools with new device names.
298 It is possible to load devices ahci, ata, siis and mvs as modules,
299 but option ATA_CAM should remain in kernel configuration to make ata
300 module work as CAM driver supporting legacy ATA controllers. Device ata
301 still can be used in modular fashion (atacore + ...). Modules atadisk
302 and atapi* are not used and won't affect operation in ATA_CAM mode.
303 Note that to use CAM-based ATA kernel should include CAM devices
304 scbus, pass, da (or explicitly ada), cd and optionally others. All of
305 them are parts of the cam module.
307 ataraid(4) functionality is now supported by the RAID GEOM class.
308 To use it you can load geom_raid kernel module and use graid(8) tool
309 for management. Instead of /dev/arX device names, use /dev/raid/rX.
311 No kernel config options or code have been removed, so if a problem
312 arises, please report it and optionally revert to the old ATA stack.
313 In order to do it you can remove from the kernel config:
318 , and instead add back:
319 device atadisk # ATA disk drives
320 device ataraid # ATA RAID drives
321 device atapicd # ATAPI CDROM drives
322 device atapifd # ATAPI floppy drives
323 device atapist # ATAPI tape drives
326 The default NFS server has been changed to the new server, which
327 was referred to as the experimental server. If you need to switch
328 back to the old NFS server, you must now put the "-o" option on
329 both the mountd and nfsd commands. This can be done using the
330 mountd_flags and nfs_server_flags rc.conf variables until an
331 update to the rc scripts is committed, which is coming soon.
334 The GNU Objective-C runtime library (libobjc), and other Objective-C
335 related components have been removed from the base system. If you
336 require an Objective-C library, please use one of the available ports.
339 ath(4) has been split into bus- and device- modules. if_ath contains
340 the HAL, the TX rate control and the network device code. if_ath_pci
341 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
342 contains the AHB glue. Users need to load both if_ath_pci and if_ath
343 in order to use ath on everything else.
345 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
346 need to load if_ath and if_ath_pci for ath(4) operation.
349 As part of the replacement of sysinstall, the process of building
350 release media has changed significantly. For details, please re-read
351 release(7), which has been updated to reflect the new build process.
354 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
355 is the last available version under GPLv2. It brings a number of new
356 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
357 4.1 and SSE 4.2), better support for powerpc64, a number of new
358 directives, and lots of other small improvements. See the ChangeLog
359 file in contrib/binutils for the full details.
362 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
363 compliant, and will now use half of hash for authentication.
364 This will break interoperability with all stacks (including all
365 actual FreeBSD versions) who implement
366 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
368 The only workaround with such peers is to use another HMAC
369 algorithm for IPsec ("phase 2") authentication.
372 Remove the uio_yield prototype and symbol. This function has
373 been misnamed since it was introduced and should not be
374 globally exposed with this name. The equivalent functionality
375 is now available using kern_yield(curthread->td_user_pri).
376 The function remains undocumented.
379 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
380 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
381 for scalar sysctls is defined but disabled. Code that needs
382 UQUAD to pass the type checking that must compile on older
383 systems where the define is not present can check against
384 __FreeBSD_version >= 900030.
386 The system dialog(1) has been replaced with a new version previously
387 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
388 with the previous version, but the libdialog associated with it has
389 a largely incompatible API. As such, the original version of libdialog
390 will be kept temporarily as libodialog, until its base system consumers
391 are replaced or updated. Bump __FreeBSD_version to 900030.
394 If you are trying to run make universe on a -stable system, and you get
395 the following warning:
396 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
397 unknown. config(8) likely too old."
398 or something similar to it, then you must upgrade your -stable system
399 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
400 stable/8) or build the config from the latest stable/8 branch and
401 install it on your system.
403 Prior to this date, building a current universe on 8-stable system from
404 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
405 error in the first kernel build phase. A new config on those old
406 systems will fix that problem for older versions of -current.
409 The TCP stack has been modified to allow Khelp modules to interact with
410 it via helper hook points and store per-connection data in the TCP
411 control block. Bump __FreeBSD_version to 900029. User space tools that
412 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
416 Generic IEEE 802.3 annex 31B full duplex flow control support has been
417 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
418 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
419 to take advantage of it instead of using custom implementations. This
420 means that these drivers now no longer unconditionally advertise
421 support for flow control but only do so if flow control is a selected
422 media option. This was implemented in the generic support that way in
423 order to allow flow control to be switched on and off via ifconfig(8)
424 with the PHY specific default to typically off in order to protect
425 from unwanted effects. Consequently, if you used flow control with
426 one of the above mentioned drivers you now need to explicitly enable
428 ifconfig bge0 media auto mediaopt flowcontrol
430 Along with the above mentioned changes generic support for setting
431 1000baseT master mode also has been added and brgphy(4), ciphy(4),
432 e1000phy(4) as well as ip1000phy(4) have been converted to take
433 advantage of it. This means that these drivers now no longer take the
434 link0 parameter for selecting master mode but the master media option
435 has to be used instead, for example like in the following:
436 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
438 Selection of master mode now is also available with all other PHY
439 drivers supporting 1000baseT.
442 The TCP stack has received a significant update to add support for
443 modularised congestion control and generally improve the clarity of
444 congestion control decisions. Bump __FreeBSD_version to 900025. User
445 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
446 sockstat) need to be recompiled.
449 The man(1) utility has been replaced by a new version that no longer
450 uses /etc/manpath.config. Please consult man.conf(5) for how to
451 migrate local entries to the new format.
454 The copyright strings printed by login(1) and sshd(8) at the time of a
455 new connection have been removed to follow other operating systems and
459 A workaround for a fixed ld bug has been removed in kernel code,
460 so make sure that your system ld is built from sources after
461 revision 210245 from 2010-07-19 (r211583 if building head kernel
462 on stable/8, r211584 for stable/7; both from 2010-08-21).
463 A symptom of incorrect ld version is different addresses for
464 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
467 The $ipv6_prefer variable in rc.conf(5) has been split into
468 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
470 The $ip6addrctl_policy is a variable to choose a pre-defined
471 address selection policy set by ip6addrctl(8). A value
472 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
475 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
476 flag (see an entry of 20090926) is set on an interface with no
477 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
478 security reason. If you want IPv6 link-local address on all
479 interfaces by default, set this to "YES".
481 The old ipv6_prefer="YES" is equivalent to
482 ipv6_activate_all_interfaces="YES" and
483 ip6addrctl_policy="ipv6_prefer".
486 DTrace has grown support for userland tracing. Due to this, DTrace is
487 now i386 and amd64 only.
488 dtruss(1) is now installed by default on those systems and a new
489 kernel module is needed for userland tracing: fasttrap.
490 No changes to your kernel config file are necessary to enable
491 userland tracing, but you might consider adding 'STRIP=' and
492 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
493 to have informative userland stack traces in DTrace (ustack).
496 The acpi_aiboost(4) driver has been removed in favor of the new
497 aibs(4) driver. You should update your kernel configuration file.
500 BSD grep has been imported to the base system and it is built by
501 default. It is completely BSD licensed, highly GNU-compatible, uses
502 less memory than its GNU counterpart and has a small codebase.
503 However, it is slower than its GNU counterpart, which is mostly
504 noticeable for larger searches, for smaller ones it is measurable
505 but not significant. The reason is complex, the most important factor
506 is that we lack a modern and efficient regex library and GNU
507 overcomes this by optimizing the searches internally. Future work
508 on improving the regex performance is planned, for the meantime,
509 users that need better performance, can build GNU grep instead by
510 setting the WITH_GNU_GREP knob.
513 Due to the import of powerpc64 support, all existing powerpc kernel
514 configuration files must be updated with a machine directive like this:
515 machine powerpc powerpc
517 In addition, an updated config(8) is required to build powerpc kernels
521 A new version of ZFS (version 15) has been merged to -HEAD.
522 This version uses a python library for the following subcommands:
523 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
524 For full functionality of these commands the following port must
525 be installed: sysutils/py-zfs
528 'vm_page's are now hashed by physical address to an array of mutexes.
529 Currently this is only used to serialize access to hold_count. Over
530 time the page queue mutex will be peeled away. This changes the size
531 of pmap on every architecture. And requires all callers of vm_page_hold
532 and vm_page_unhold to be updated.
535 WITH_CTF can now be specified in src.conf (not recommended, there
536 are some problems with static executables), make.conf (would also
537 affect ports which do not use GNU make and do not override the
538 compile targets) or in the kernel config (via "makeoptions
540 When WITH_CTF was specified there before this was silently ignored,
541 so make sure that WITH_CTF is not used in places which could lead
542 to unwanted behavior.
545 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
546 to allow 32-bit compatibility on non-x86 platforms. All kernel
547 configurations on amd64 and ia64 platforms using these options must
548 be modified accordingly.
551 The utmp user accounting database has been replaced with utmpx,
552 the user accounting interface standardized by POSIX.
553 Unfortunately the semantics of utmp and utmpx don't match,
554 making it practically impossible to support both interfaces.
555 The user accounting database is used by tools like finger(1),
556 last(1), talk(1), w(1) and ac(8).
558 All applications in the base system use utmpx. This means only
559 local binaries (e.g. from the ports tree) may still use these
560 utmp database files. These applications must be rebuilt to make
563 After the system has been upgraded, it is safe to remove the old
564 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
565 assuming their contents is of no importance anymore. Old wtmp
566 databases can only be used by last(1) and ac(8) after they have
567 been converted to the new format using wtmpcvt(1).
570 Introduce the kernel thread "deadlock resolver" (which can be enabled
571 via the DEADLKRES option, see NOTES for more details) and the
572 sleepq_type() function for sleepqueues.
575 The rc.firewall and rc.firewall6 were unified, and
576 rc.firewall6 and rc.d/ip6fw were removed.
577 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
578 variables are obsoleted. Instead, the following new rc
579 variables are added to rc.d/ipfw:
581 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
582 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
583 firewall_simple_onet_ipv6, firewall_trusted_ipv6
585 The meanings correspond to the relevant IPv4 variables.
591 The default terminal emulation for syscons(4) has been changed
592 from cons25 to xterm on all platforms except pc98. This means
593 that the /etc/ttys file needs to be updated to ensure correct
594 operation of applications on the console.
596 The terminal emulation style can be toggled per window by using
597 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
598 options can be used to change the compile-time default back to
601 To prevent graphical artifacts, make sure the TERM environment
602 variable is set to match the terminal emulation that is being
603 performed by syscons(4).
606 The layout of the structure ieee80211req_scan_result has changed.
607 Applications that require wireless scan results (e.g. ifconfig(8))
608 from net80211 need to be recompiled.
610 Applications such as wpa_supplicant(8) may require a full world
611 build without using NO_CLEAN in order to get synchronized with the
615 The iwn(4) driver has been updated to support the 5000 and 5150 series.
616 There's one kernel module for each firmware. Adding "device iwnfw"
617 to the kernel configuration file means including all three firmware
618 images inside the kernel. If you want to include just the one for
619 your wireless card, use the devices iwn4965fw, iwn5000fw or
623 The rc.d/network_ipv6, IPv6 configuration script has been integrated
624 into rc.d/netif. The changes are the following:
626 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
627 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
628 Note that both variables need the "inet6" keyword at the head.
630 Do not set $ipv6_network_interfaces manually if you do not
631 understand what you are doing. It is not needed in most cases.
633 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
636 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
637 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
639 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
640 all configured interfaces have "inet6 accept_rtadv" in the
641 $ifconfig_IF_ipv6. These are for backward compatibility.
643 3. A new variable $ipv6_prefer has been added. If NO, IPv6
644 functionality of interfaces with no corresponding
645 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
646 and the default address selection policy of ip6addrctl(8)
647 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
648 Note that if you want to configure IPv6 functionality on the
649 disabled interfaces after boot, first you need to clear the flag by
650 using ifconfig(8) like:
652 ifconfig em0 inet6 -ifdisabled
654 If YES, the default address selection policy is set as
657 The default value of $ipv6_prefer is NO.
659 4. If your system need to receive Router Advertisement messages,
660 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
661 scripts automatically invoke rtsol(8) when the interface becomes
662 UP. The Router Advertisement messages are used for SLAAC
663 (State-Less Address AutoConfiguration).
666 802.11s D3.03 support was committed. This is incompatible with the
667 previous code, which was based on D3.0.
670 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
671 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
672 control whether accepting Router Advertisement messages or not.
673 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
674 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
675 The ifconfig(8) utility now supports these flags.
678 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
679 mount(8) and -a option for df(1) to see them.
682 The old tunable hw.bus.devctl_disable has been superseded by
683 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
684 replaced by hw.bus.devctl_queue=0. The default for this new tunable
688 Remove the option STOP_NMI. The default action is now to use NMI only
689 for KDB via the newly introduced function stop_cpus_hard() and
690 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
693 The stable/8 branch created in subversion. This corresponds to the
694 RELENG_8 branch in CVS.
697 Bump the shared library version numbers for all libraries that do not
698 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
699 __FreeBSD_version to 800105.
702 Due to changes in the implementation of virtual network stack support,
703 all network-related kernel modules must be recompiled. As this change
704 breaks the ABI, bump __FreeBSD_version to 800104.
707 The TOE interface to the TCP syncache has been modified to remove
708 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
709 The cxgb driver is the only TOE consumer affected by this change, and
710 needs to be recompiled along with the kernel. As this change breaks
711 the ABI, bump __FreeBSD_version to 800103.
714 Padding has been added to struct tcpcb, sackhint and tcpstat in
715 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
716 maintaining the ABI. However, this change breaks the ABI, so bump
717 __FreeBSD_version to 800102. User space tools that rely on the size of
718 any of these structs (e.g. sockstat) need to be recompiled.
721 The NFS_LEGACYRPC option has been removed along with the old kernel
722 RPC implementation that this option selected. Kernel configurations
723 may need to be adjusted.
726 The network interface device nodes at /dev/net/<interface> have been
727 removed. All ioctl operations can be performed the normal way using
728 routing sockets. The kqueue functionality can generally be replaced
729 with routing sockets.
732 The documentation from the FreeBSD Documentation Project (Handbook,
733 FAQ, etc.) is now installed via packages by sysinstall(8) and under
734 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
737 The ABI of various structures related to the SYSV IPC API have been
738 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
739 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
743 Layout of struct vnet has changed as routing related variables were
744 moved to their own Vimage module. Modules need to be recompiled. Bump
745 __FreeBSD_version to 800099.
748 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
749 respectively. As long as no more than 16 groups per process are used,
750 no changes should be visible. When more than 16 groups are used, old
751 binaries may fail if they call getgroups() or getgrouplist() with
752 statically sized storage. Recompiling will work around this, but
753 applications should be modified to use dynamically allocated storage
754 for group arrays as POSIX.1-2008 does not cap an implementation's
755 number of supported groups at NGROUPS_MAX+1 as previous versions did.
757 NFS and portalfs mounts may also be affected as the list of groups is
758 truncated to 16. Users of NFS who use more than 16 groups, should
759 take care that negative group permissions are not used on the exported
760 file systems as they will not be reliable unless a GSSAPI based
761 authentication method is used.
764 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
765 option compiles in the support for adaptive spinning for lockmgrs
766 which want to enable it. The lockinit() function now accepts the flag
767 LK_ADAPTIVE in order to make the lock object subject to adaptive
768 spinning when both held in write and read mode.
771 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
772 changed. User applications that use this ioctl need to be rebuilt.
775 The layout of struct thread has changed. Kernel and modules need to
779 The layout of structs ifnet, domain, protosw and vnet_net has changed.
780 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
783 window(1) has been removed from the base system. It can now be
784 installed from ports. The port is called misc/window.
787 The way we are storing and accessing `routing table' entries has
788 changed. Programs reading the FIB, like netstat, need to be
792 A new netisr implementation has been added for FreeBSD 8. Network
793 file system modules, such as igmp, ipdivert, and others, should be
795 Bump __FreeBSD_version to 800096.
798 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
802 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
803 Bump __FreeBSD_version to 800094.
806 Add mnt_xflag field to 'struct mount'. File system modules need to be
808 Bump __FreeBSD_version to 800093.
811 The compiling option ADAPTIVE_SX has been retired while it has been
812 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
813 The KPI for sx_init_flags() changes as accepting flags:
814 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
815 been introduced in order to handle the reversed logic.
816 Bump __FreeBSD_version to 800092.
819 Add support for hierarchical jails. Remove global securelevel.
820 Bump __FreeBSD_version to 800091.
823 The layout of struct vnet_net has changed, therefore modules
825 Bump __FreeBSD_version to 800090.
828 The newly imported zic(8) produces a new format in the output. Please
829 run tzsetup(8) to install the newly created data to /etc/localtime.
832 The sysctl tree for the usb stack has renamed from hw.usb2.* to
833 hw.usb.* and is now consistent again with previous releases.
836 802.11 monitor mode support was revised and driver api's were changed.
837 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
838 of DLT_IEEE802_11. No user-visible data structures were changed but
839 applications that use DLT_IEEE802_11 may require changes.
840 Bump __FreeBSD_version to 800088.
843 The layout of the following structs has changed: sysctl_oid,
844 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
845 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
846 panics may be experienced. World rebuild is required for
847 correctly checking networking state from userland.
848 Bump __FreeBSD_version to 800085.
851 MLDv2 and Source-Specific Multicast (SSM) have been merged
852 to the IPv6 stack. VIMAGE hooks are in but not yet used.
853 The implementation of SSM within FreeBSD's IPv6 stack closely
854 follows the IPv4 implementation.
856 For kernel developers:
858 * The most important changes are that the ip6_output() and
859 ip6_input() paths no longer take the IN6_MULTI_LOCK,
860 and this lock has been downgraded to a non-recursive mutex.
862 * As with the changes to the IPv4 stack to support SSM, filtering
863 of inbound multicast traffic must now be performed by transport
864 protocols within the IPv6 stack. This does not apply to TCP and
865 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
867 * The KPIs used by IPv6 multicast are similar to those used by
868 the IPv4 stack, with the following differences:
869 * im6o_mc_filter() is analogous to imo_multicast_filter().
870 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
871 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
872 * IN6_LOOKUP_MULTI() has been deprecated and removed.
873 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
874 for MLDv1 have an additional 'timer' argument which is used to
875 jitter the initial membership report for the solicited-node
876 multicast membership on-link.
877 * This is not strictly needed for MLDv2, which already jitters
878 its report transmissions. However, the 'timer' argument is
879 preserved in case MLDv1 is active on the interface.
881 * The KAME linked-list based IPv6 membership implementation has
882 been refactored to use a vector similar to that used by the IPv4
884 Code which maintains a list of its own multicast memberships
885 internally, e.g. carp, has been updated to reflect the new
888 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
889 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
890 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
891 implementation constraint which needs to be addressed in HEAD.
893 For application developers:
895 * The changes are broadly similar to those made for the IPv4
898 * The use of IPv4 and IPv6 multicast socket options on the same
899 socket, using mapped addresses, HAS NOT been tested or supported.
901 * There are a number of issues with the implementation of various
902 IPv6 multicast APIs which need to be resolved in the API surface
903 before the implementation is fully compatible with KAME userland
904 use, and these are mostly to do with interface index treatment.
906 * The literature available discusses the use of either the delta / ASM
907 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
908 using setsourcefilter(3)/getsourcefilter(3). For more information
909 please refer to RFC 3768, 'Socket Interface Extensions for
910 Multicast Source Filters'.
912 * Applications which use the published RFC 3678 APIs should be fine.
914 For systems administrators:
916 * The mtest(8) utility has been refactored to support IPv6, in
917 addition to IPv4. Interface addresses are no longer accepted
918 as arguments, their names must be used instead. The utility
919 will map the interface name to its first IPv4 address as
920 returned by getifaddrs(3).
922 * The ifmcstat(8) utility has also been updated to print the MLDv2
923 endpoint state and source filter lists via sysctl(3).
925 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
926 loopback of IPv6 multicast datagrams by default; it defaults to 1
927 to preserve the existing behaviour. Disabling multicast loopback is
928 recommended for optimal system performance.
930 * The IPv6 MROUTING code has been changed to examine this sysctl
931 instead of attempting to perform a group lookup before looping
932 back forwarded datagrams.
934 Bump __FreeBSD_version to 800084.
937 Implement low-level Bluetooth HCI API.
938 Bump __FreeBSD_version to 800083.
941 The layout of struct malloc_type, used by modules to register new
942 memory allocation types, has changed. Most modules will need to
943 be rebuilt or panics may be experienced.
944 Bump __FreeBSD_version to 800081.
947 Anticipate overflowing inp_flags - add inp_flags2.
948 This changes most offsets in inpcb, so checking v4 connection
949 state will require a world rebuild.
950 Bump __FreeBSD_version to 800080.
953 Add an llentry to struct route and struct route_in6. Modules
954 embedding a struct route will need to be recompiled.
955 Bump __FreeBSD_version to 800079.
958 The size of rt_metrics_lite and by extension rtentry has changed.
959 Networking administration apps will need to be recompiled.
960 The route command now supports show as an alias for get, weighting
961 of routes, sticky and nostick flags to alter the behavior of stateful
963 Bump __FreeBSD_version to 800078.
966 Do not use Giant for kbdmux(4) locking. This is wrong and
967 apparently causing more problems than it solves. This will
968 re-open the issue where interrupt handlers may race with
969 kbdmux(4) in polling mode. Typical symptoms include (but
970 not limited to) duplicated and/or missing characters when
971 low level console functions (such as gets) are used while
972 interrupts are enabled (for example geli password prompt,
973 mountroot prompt etc.). Disabling kbdmux(4) may help.
976 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
977 kernel modules referencing any of the above need to be recompiled.
978 Bump __FreeBSD_version to 800075.
981 GEOM_PART has become the default partition slicer for storage devices,
982 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
983 introduces some changes:
985 MSDOS/EBR: the devices created from MSDOS extended partition entries
986 (EBR) can be named differently than with GEOM_MBR and are now symlinks
987 to devices with offset-based names. fstabs may need to be modified.
989 BSD: the "geometry does not match label" warning is harmless in most
990 cases but it points to problems in file system misalignment with
991 disk geometry. The "c" partition is now implicit, covers the whole
992 top-level drive and cannot be (mis)used by users.
994 General: Kernel dumps are now not allowed to be written to devices
995 whose partition types indicate they are meant to be used for file
996 systems (or, in case of MSDOS partitions, as something else than
999 Most of these changes date approximately from 200812.
1002 The uscanner(4) driver has been removed from the kernel. This follows
1003 Linux removing theirs in 2.6 and making libusb the default interface
1004 (supported by sane).
1007 The multicast forwarding code has been cleaned up. netstat(1)
1008 only relies on KVM now for printing bandwidth upcall meters.
1009 The IPv4 and IPv6 modules are split into ip_mroute_mod and
1010 ip6_mroute_mod respectively. The config(5) options for statically
1011 compiling this code remain the same, i.e. 'options MROUTING'.
1014 Support for the IFF_NEEDSGIANT network interface flag has been
1015 removed, which means that non-MPSAFE network device drivers are no
1016 longer supported. In particular, if_ar, if_sr, and network device
1017 drivers from the old (legacy) USB stack can no longer be built or
1021 POSIX.1 Native Language Support (NLS) has been enabled in libc and
1022 a bunch of new language catalog files have also been added.
1023 This means that some common libc messages are now localized and
1024 they depend on the LC_MESSAGES environmental variable.
1027 The k8temp(4) driver has been renamed to amdtemp(4) since
1028 support for Family 10 and Family 11 CPU families was added.
1031 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1032 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1034 For kernel developers, the most important changes are that the
1035 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1036 and this lock has been downgraded to a non-recursive mutex.
1038 Transport protocols (UDP, Raw IP) are now responsible for filtering
1039 inbound multicast traffic according to group membership and source
1040 filters. The imo_multicast_filter() KPI exists for this purpose.
1041 Transports which do not use multicast (SCTP, TCP) already reject
1042 multicast by default. Forwarding and receive performance may improve
1043 as a mutex acquisition is no longer needed in the ip_input()
1044 low-level input path. in_addmulti() and in_delmulti() are shimmed
1045 to new KPIs which exist to support SSM in-kernel.
1047 For application developers, it is recommended that loopback of
1048 multicast datagrams be disabled for best performance, as this
1049 will still cause the lock to be taken for each looped-back
1050 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1051 be tuned to 0 to disable loopback by default; it defaults to 1
1052 to preserve the existing behaviour.
1054 For systems administrators, to obtain best performance with
1055 multicast reception and multiple groups, it is always recommended
1056 that a card with a suitably precise hash filter is used. Hash
1057 collisions will still result in the lock being taken within the
1058 transport protocol input path to check group membership.
1060 If deploying FreeBSD in an environment with IGMP snooping switches,
1061 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1062 enabled; this forces 224.0.0.0/24 group membership to be announced
1065 The size of 'struct igmpstat' has changed; netstat needs to be
1066 recompiled to reflect this.
1067 Bump __FreeBSD_version to 800070.
1070 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1071 updated to use it. This requires a buildworld/installworld in order to
1072 update the library and dependencies (usbconfig, etc). Its advisable to
1073 rebuild all ports which uses libusb. More specific directions are given
1074 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1075 libusb are no longer required and can be removed.
1078 A workaround is committed to allow the creation of System V shared
1079 memory segment of size > 2 GB on the 64-bit architectures.
1080 Due to a limitation of the existing ABI, the shm_segsz member
1081 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1082 wrong for large segments. Note that limits must be explicitly
1083 raised to allow such segments to be created.
1086 The layout of struct ifnet has changed, requiring a rebuild of all
1087 network device driver modules.
1090 The /dev handling for the new USB stack has changed, a
1091 buildworld/installworld is required for libusb20.
1094 The new USB2 stack has now been permanently moved in and all kernel and
1095 module names reverted to their previous values (eg, usb, ehci, ohci,
1096 ums, ...). The old usb stack can be compiled in by prefixing the name
1097 with the letter 'o', the old usb modules have been removed.
1098 Updating entry 20090216 for xorg and 20090215 for libmap may still
1102 The rc.conf(5) option if_up_delay has been renamed to
1103 defaultroute_delay to better reflect its purpose. If you have
1104 customized this setting in /etc/rc.conf you need to update it to
1108 xorg 7.4 wants to configure its input devices via hald which does not
1109 yet work with USB2. If the keyboard/mouse does not work in xorg then
1111 Option "AllowEmptyInput" "off"
1112 to your ServerLayout section. This will cause X to use the configured
1113 kbd and mouse sections from your xorg.conf.
1116 The GENERIC kernels for all architectures now default to the new USB2
1117 stack. No kernel config options or code have been removed so if a
1118 problem arises please report it and optionally revert to the old USB
1119 stack. If you are loading USB kernel modules or have a custom kernel
1120 that includes GENERIC then ensure that usb names are also changed over,
1121 eg uftdi -> usb2_serial_ftdi.
1123 Older programs linked against the ports libusb 0.1 need to be
1124 redirected to the new stack's libusb20. /etc/libmap.conf can
1126 # Map old usb library to new one for usb2 stack
1127 libusb-0.1.so.8 libusb20.so.1
1130 All USB ethernet devices now attach as interfaces under the name ueN
1131 (eg. ue0). This is to provide a predictable name as vendors often
1132 change usb chipsets in a product without notice.
1135 The ichsmb(4) driver has been changed to require SMBus slave
1136 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1137 All of the other SMBus controller drivers require left-justified
1138 slave addresses, so this change makes all the drivers provide the
1142 INET6 statistics (struct ip6stat) was updated.
1143 netstat(1) needs to be recompiled.
1146 NTFS has been removed from GENERIC kernel on amd64 to match
1147 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1148 will load ntfs.ko module automatically when NTFS support is
1149 actually needed, unless ntfs.ko is not installed or security
1150 level prohibits loading kernel modules. If either is the case,
1151 "options NTFS" has to be added into kernel config.
1154 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1155 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1156 800061. User space tools that rely on the size of struct tcpcb in
1157 tcp_var.h (e.g. sockstat) need to be recompiled.
1160 ng_tty(4) module updated to match the new TTY subsystem.
1161 Due to API change, user-level applications must be updated.
1162 New API support added to mpd5 CVS and expected to be present
1163 in next mpd5.3 release.
1166 With __FreeBSD_version 800060 the makefs tool is part of
1167 the base system (it was a port).
1170 The afdata and ifnet locks have been changed from mutexes to
1171 rwlocks, network modules will need to be re-compiled.
1174 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1175 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1176 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1177 architecture and 40 bytes on 64-bit architecture. The userland
1178 applications "arp" and "ndp" have been updated accordingly.
1179 The output from "netstat -r" shows only routing entries and
1180 none of the L2 information.
1183 __FreeBSD_version 800057 marks the switchover from the
1184 binary ath hal to source code. Users must add the line:
1186 options AH_SUPPORT_AR5416
1188 to their kernel config files when specifying:
1192 The ath_hal module no longer exists; the code is now compiled
1193 together with the driver in the ath module. It is now
1194 possible to tailor chip support (i.e. reduce the set of chips
1195 and thereby the code size); consult ath_hal(4) for details.
1198 __FreeBSD_version 800054 adds memory barriers to
1199 <machine/atomic.h>, new interfaces to ifnet to facilitate
1200 multiple hardware transmit queues for cards that support
1201 them, and a lock-less ring-buffer implementation to
1202 enable drivers to more efficiently manage queueing of
1206 A new version of ZFS (version 13) has been merged to -HEAD.
1207 This version has zpool attribute "listsnapshots" off by
1208 default, which means "zfs list" does not show snapshots,
1209 and is the same as Solaris behavior.
1212 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1215 The uhci, ohci, ehci and slhci USB Host controller drivers have
1216 been put into separate modules. If you load the usb module
1217 separately through loader.conf you will need to load the
1218 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1219 controller add the following to loader.conf:
1225 The ABI used by the PMC toolset has changed. Please keep
1226 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1230 atapci kernel module now includes only generic PCI ATA
1231 driver. AHCI driver moved to ataahci kernel module.
1232 All vendor-specific code moved into separate kernel modules:
1233 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1234 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1235 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1236 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1239 The TTY subsystem of the kernel has been replaced by a new
1240 implementation, which provides better scalability and an
1241 improved driver model. Most common drivers have been migrated to
1242 the new TTY subsystem, while others have not. The following
1243 drivers have not yet been ported to the new TTY layer:
1246 cy, digi, rc, rp, sio
1252 ng_h4, ng_tty, ppp, sl, snp
1254 Adding these drivers to your kernel configuration file shall
1255 cause compilation to fail.
1258 ntpd has been upgraded to 4.2.4p5.
1261 OpenSSH has been upgraded to 5.1p1.
1263 For many years, FreeBSD's version of OpenSSH preferred DSA
1264 over RSA for host and user authentication keys. With this
1265 upgrade, we've switched to the vendor's default of RSA over
1266 DSA. This may cause upgraded clients to warn about unknown
1267 host keys even for previously known hosts. Users should
1268 follow the usual procedure for verifying host keys before
1269 accepting the RSA key.
1271 This can be circumvented by setting the "HostKeyAlgorithms"
1272 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1275 Please note that the sequence of keys offered for
1276 authentication has been changed as well. You may want to
1277 specify IdentityFile in a different order to revert this
1281 The sio(4) driver has been removed from the i386 and amd64
1282 kernel configuration files. This means uart(4) is now the
1283 default serial port driver on those platforms as well.
1285 To prevent collisions with the sio(4) driver, the uart(4) driver
1286 uses different names for its device nodes. This means the
1287 onboard serial port will now most likely be called "ttyu0"
1288 instead of "ttyd0". You may need to reconfigure applications to
1289 use the new device names.
1291 When using the serial port as a boot console, be sure to update
1292 /boot/device.hints and /etc/ttys before booting the new kernel.
1293 If you forget to do so, you can still manually specify the hints
1294 at the loader prompt:
1296 set hint.uart.0.at="isa"
1297 set hint.uart.0.port="0x3F8"
1298 set hint.uart.0.flags="0x10"
1299 set hint.uart.0.irq="4"
1303 The gpt(8) utility has been removed. Use gpart(8) to partition
1307 The version that Linuxulator emulates was changed from 2.4.2
1308 to 2.6.16. If you experience any problems with Linux binaries
1309 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1310 if it fixes the problem contact emulation mailing list.
1313 ISDN4BSD (I4B) was removed from the src tree. You may need to
1314 update a your kernel configuration and remove relevant entries.
1317 I have checked in code to support multiple routing tables.
1318 See the man pages setfib(1) and setfib(2).
1319 This is a hopefully backwards compatible version,
1320 but to make use of it you need to compile your kernel
1321 with options ROUTETABLES=2 (or more up to 16).
1324 The 802.11 wireless support was redone to enable multi-bss
1325 operation on devices that are capable. The underlying device
1326 is no longer used directly but instead wlanX devices are
1327 cloned with ifconfig. This requires changes to rc.conf files.
1328 For example, change:
1329 ifconfig_ath0="WPA DHCP"
1332 ifconfig_wlan0="WPA DHCP"
1333 see rc.conf(5) for more details. In addition, mergemaster of
1334 /etc/rc.d is highly recommended. Simultaneous update of userland
1335 and kernel wouldn't hurt either.
1337 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1338 modules were merged into the base wlan module. All references
1339 to these modules (e.g. in kernel config files) must be removed.
1342 psm(4) has gained write(2) support in native operation level.
1343 Arbitrary commands can be written to /dev/psm%d and status can
1344 be read back from it. Therefore, an application is responsible
1345 for status validation and error recovery. It is a no-op in
1346 other operation levels.
1349 Support for KSE threading has been removed from the kernel. To
1350 run legacy applications linked against KSE libmap.conf may
1351 be used. The following libmap.conf may be used to ensure
1352 compatibility with any prior release:
1354 libpthread.so.1 libthr.so.1
1355 libpthread.so.2 libthr.so.2
1356 libkse.so.3 libthr.so.3
1359 The layout of struct vmspace has changed. This affects libkvm
1360 and any executables that link against libkvm and use the
1361 kvm_getprocs() function. In particular, but not exclusively,
1362 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1363 The effects are minimal, but it's advisable to upgrade world
1367 The latest em driver no longer has support in it for the
1368 82575 adapter, this is now moved to the igb driver. The
1369 split was done to make new features that are incompatible
1370 with older hardware easier to do.
1373 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1374 likewise the kernel option is now GEOM_LINUX_LVM.
1377 The default NFS mount mode has changed from UDP to TCP for
1378 increased reliability. If you rely on (insecurely) NFS
1379 mounting across a firewall you may need to update your
1383 Belatedly note the addition of m_collapse for compacting
1387 The fts(3) structures have been changed to use adequate
1388 integer types for their members and so to be able to cope
1389 with huge file trees. The old fts(3) ABI is preserved
1390 through symbol versioning in libc, so third-party binaries
1391 using fts(3) should still work, although they will not take
1392 advantage of the extended types. At the same time, some
1393 third-party software might fail to build after this change
1394 due to unportable assumptions made in its source code about
1395 fts(3) structure members. Such software should be fixed
1396 by its vendor or, in the worst case, in the ports tree.
1397 FreeBSD_version 800015 marks this change for the unlikely
1398 case that a portable fix is impossible.
1401 To upgrade to -current after this date, you must be running
1402 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1403 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1406 The ADAPTIVE_GIANT kernel option has been retired because its
1407 functionality is the default now.
1410 The AT keyboard emulation of sunkbd(4) has been turned on
1411 by default. In order to make the special symbols of the Sun
1412 keyboards driven by sunkbd(4) work under X these now have
1413 to be configured the same way as Sun USB keyboards driven
1414 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1416 Option "XkbLayout" "us"
1417 Option "XkbRules" "xorg"
1418 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1421 It has been decided that it is desirable to provide ABI
1422 backwards compatibility to the FreeBSD 4/5/6 versions of the
1423 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1424 broken with the introduction of PCI domain support (see the
1425 20070930 entry). Unfortunately, this required the ABI of
1426 PCIOCGETCONF to be broken again in order to be able to
1427 provide backwards compatibility to the old version of that
1428 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1429 again. As for prominent ports this affects neither pciutils
1430 nor xorg-server this time, the hal port needs to be rebuilt
1434 The misnamed kthread_create() and friends have been renamed
1435 to kproc_create() etc. Many of the callers already
1436 used kproc_start()..
1437 I will return kthread_create() and friends in a while
1438 with implementations that actually create threads, not procs.
1439 Renaming corresponds with version 800002.
1448 Avoid using make -j when upgrading. While generally safe, there are
1449 sometimes problems using -j to upgrade. If your upgrade fails with
1450 -j, please try again without -j. From time to time in the past there
1451 have been problems using -j with buildworld and/or installworld. This
1452 is especially true when upgrading between "distant" versions (eg one
1453 that cross a major release boundary or several minor releases, or when
1454 several months have passed on the -current branch).
1456 Sometimes, obscure build problems are the result of environment
1457 poisoning. This can happen because the make utility reads its
1458 environment when searching for values for global variables. To run
1459 your build attempts in an "environmental clean room", prefix all make
1460 commands with 'env -i '. See the env(1) manual page for more details.
1462 When upgrading from one major version to another it is generally best
1463 to upgrade to the latest code in the currently installed branch first,
1464 then do an upgrade to the new branch. This is the best-tested upgrade
1465 path, and has the highest probability of being successful. Please try
1466 this approach before reporting problems with a major version upgrade.
1470 When upgrading the boot ZFS pool to a new version, always follow
1473 1.) recompile and reinstall the ZFS boot loader and boot block
1474 (this is part of "make buildworld" and "make installworld")
1476 2.) update the ZFS boot block on your boot drive
1478 The following example updates the ZFS boot block on the first
1479 partition (freebsd-boot) of a GPT partitioned drive ad0:
1480 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1482 Non-boot pools do not need these updates.
1486 If you are updating from a prior version of FreeBSD (even one just
1487 a few days old), you should follow this procedure. It is the most
1488 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1490 make kernel-toolchain
1491 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1492 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1494 To test a kernel once
1495 ---------------------
1496 If you just want to boot a kernel once (because you are not sure
1497 if it works, or if you want to boot a known bad kernel to provide
1498 debugging information) run
1499 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1500 nextboot -k testkernel
1502 To just build a kernel when you know that it won't mess you up
1503 --------------------------------------------------------------
1504 This assumes you are already running a CURRENT system. Replace
1505 ${arch} with the architecture of your machine (e.g. "i386",
1506 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1508 cd src/sys/${arch}/conf
1509 config KERNEL_NAME_HERE
1510 cd ../compile/KERNEL_NAME_HERE
1515 If this fails, go to the "To build a kernel" section.
1517 To rebuild everything and install it on the current system.
1518 -----------------------------------------------------------
1519 # Note: sometimes if you are running current you gotta do more than
1520 # is listed here if you are upgrading from a really old current.
1522 <make sure you have good level 0 dumps>
1524 make kernel KERNCONF=YOUR_KERNEL_HERE
1526 <reboot in single user> [3]
1534 To cross-install current onto a separate partition
1535 --------------------------------------------------
1536 # In this approach we use a separate partition to hold
1537 # current's root, 'usr', and 'var' directories. A partition
1538 # holding "/", "/usr" and "/var" should be about 2GB in
1541 <make sure you have good level 0 dumps>
1544 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1545 <maybe newfs current's root partition>
1546 <mount current's root partition on directory ${CURRENT_ROOT}>
1547 make installworld DESTDIR=${CURRENT_ROOT}
1548 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1549 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1550 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1551 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1552 <reboot into current>
1553 <do a "native" rebuild/install as described in the previous section>
1554 <maybe install compatibility libraries from ports/misc/compat*>
1558 To upgrade in-place from stable to current
1559 ----------------------------------------------
1560 <make sure you have good level 0 dumps>
1562 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1564 <reboot in single user> [3]
1571 Make sure that you've read the UPDATING file to understand the
1572 tweaks to various things you need. At this point in the life
1573 cycle of current, things change often and you are on your own
1574 to cope. The defaults can also change, so please read ALL of
1575 the UPDATING entries.
1577 Also, if you are tracking -current, you must be subscribed to
1578 freebsd-current@freebsd.org. Make sure that before you update
1579 your sources that you have read and understood all the recent
1580 messages there. If in doubt, please track -stable which has
1581 much fewer pitfalls.
1583 [1] If you have third party modules, such as vmware, you
1584 should disable them at this point so they don't crash your
1587 [3] From the bootblocks, boot -s, and then do
1592 adjkerntz -i # if CMOS is wall time
1593 Also, when doing a major release upgrade, it is required that
1594 you boot into single user mode to do the installworld.
1596 [4] Note: This step is non-optional. Failure to do this step
1597 can result in a significant reduction in the functionality of the
1598 system. Attempting to do it by hand is not recommended and those
1599 that pursue this avenue should read this file carefully, as well
1600 as the archives of freebsd-current and freebsd-hackers mailing lists
1601 for potential gotchas. The -U option is also useful to consider.
1602 See mergemaster(8) for more information.
1604 [5] Usually this step is a noop. However, from time to time
1605 you may need to do this if you get unknown user in the following
1606 step. It never hurts to do it all the time. You may need to
1607 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1608 install) after the buildworld before this step if you last updated
1609 from [78]-stable or 9-stable before 20130430.
1611 [6] This only deletes old files and directories. Old libraries
1612 can be deleted by "make delete-old-libs", but you have to make
1613 sure that no program is using those libraries anymore.
1615 [8] In order to have a kernel that can run the 4.x binaries needed to
1616 do an installworld, you must include the COMPAT_FREEBSD4 option in
1617 your kernel. Failure to do so may leave you with a system that is
1618 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1619 required to run the 5.x binaries on more recent kernels. And so on
1620 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1622 Make sure that you merge any new devices from GENERIC since the
1623 last time you updated your kernel config file.
1625 [9] When checking out sources, you must include the -P flag to have
1626 cvs prune empty directories.
1628 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1629 "?=" instead of the "=" assignment operator, so that buildworld can
1630 override the CPUTYPE if it needs to.
1632 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1633 not on the command line, or in /etc/make.conf. buildworld will
1634 warn if it is improperly defined.
1637 This file contains a list, in reverse chronological order, of major
1638 breakages in tracking -current. It is not guaranteed to be a complete
1639 list of such breakages, and only contains entries since October 10, 2007.
1640 If you need to see UPDATING entries from before that date, you will need
1641 to fetch an UPDATING file from an older FreeBSD release.
1643 Copyright information:
1645 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1647 Redistribution, publication, translation and use, with or without
1648 modification, in full or in part, in any form or format of this
1649 document are permitted without further permission from the author.
1651 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1652 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1653 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1654 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1655 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1656 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1657 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1658 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1659 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1660 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1661 POSSIBILITY OF SUCH DAMAGE.
1663 Contact Warner Losh if you have any questions about your use of