1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 Items affecting the ports and packages system can be found in
10 /usr/ports/UPDATING. Please read that file before running portupgrade.
12 NOTE TO PEOPLE WHO THINK THAT FreeBSD 10.x IS SLOW:
13 FreeBSD 10.x has many debugging features turned on, in both the kernel
14 and userland. These features attempt to detect incorrect use of
15 system primitives, and encourage loud failure through extra sanity
16 checking and fail stop semantics. They also substantially impact
17 system performance. If you want to do performance measurement,
18 benchmarking, and optimization, you'll want to turn them off. This
19 includes various WITNESS- related kernel options, INVARIANTS, malloc
20 debugging flags in userland, and various verbose features in the
21 kernel. Many developers choose to disable these features on build
22 machines to maximize performance. (To completely disable malloc
23 debugging, define MALLOC_PRODUCTION in /etc/make.conf, or to merely
24 disable the most expensive debugging functionality run
25 "ln -s 'abort:false,junk:false' /etc/malloc.conf".)
28 The IPv4 network stack has been converted to network byte
29 order. The following modules need to be recompiled together
30 with kernel: carp(4), divert(4), gif(4), siftr(4), gre(4),
31 pf(4), ipfw(4), ng_ipfw(4), stf(4).
34 Support for non-MPSAFE filesystems was removed from VFS. The
35 VFS_VERSION was bumped, all filesystem modules shall be
39 All the non-MPSAFE filesystems have been disconnected from
40 the build. The full list includes: codafs, hpfs, ntfs, nwfs,
44 The interface cloning API and ABI has changed. The following
45 modules need to be recompiled together with kernel:
46 ipfw(4), pfsync(4), pflog(4), usb(4), wlan(4), stf(4),
47 vlan(4), disc(4), edsc(4), if_bridge(4), gif(4), tap(4),
48 faith(4), epair(4), enc(4), tun(4), if_lagg(4), gre(4).
51 The sdhci driver was split in two parts: sdhci (generic SD Host
52 Controller logic) and sdhci_pci (actual hardware driver).
53 No kernel config modifications are required, but if you
54 load sdhc as a module you must switch to sdhci_pci instead.
57 Import the FUSE kernel and userland support into base system.
60 The GNU sort(1) program has been removed since the BSD-licensed
61 sort(1) has been the default for quite some time and no serious
62 problems have been reported. The corresponding WITH_GNU_SORT
66 The pfil(9) API/ABI for AF_INET family has been changed. Packet
67 filtering modules: pf(4), ipfw(4), ipfilter(4) need to be recompiled
71 The net80211(4) ABI has been changed to allow for improved driver
72 PS-POLL and power-save support. All wireless drivers need to be
73 recompiled to work with the new kernel.
76 The random(4) support for the VIA hardware random number
77 generator (`PADLOCK') is no longer enabled unconditionally.
78 Add the PADLOCK_RNG option in the custom kernel config if
79 needed. The GENERIC kernels on i386 and amd64 do include the
80 option, so the change only affects the custom kernel
84 The pf(4) packet filter ABI has been changed. pfctl(8) and
85 snmp_pf module need to be recompiled to work with new kernel.
88 A new ZFS feature flag "com.delphix:empty_bpobj" has been merged
89 to -HEAD. Pools that have empty_bpobj in active state can not be
90 imported read-write with ZFS implementations that do not support
91 this feature. For more information read the zpool-features(5)
95 The sparc64 ZFS loader has been changed to no longer try to auto-
96 detect ZFS providers based on diskN aliases but now requires these
97 to be explicitly listed in the OFW boot-device environment variable.
100 The OpenSSL has been upgraded to 1.0.1c. Any binaries requiring
101 libcrypto.so.6 or libssl.so.6 must be recompiled. Also, there are
102 configuration changes. Make sure to merge /etc/ssl/openssl.cnf.
105 The following sysctls and tunables have been renamed for consistency
106 with other variables:
107 kern.cam.da.da_send_ordered -> kern.cam.da.send_ordered
108 kern.cam.ada.ada_send_ordered -> kern.cam.ada.send_ordered
111 The sort utility has been replaced with BSD sort. For now, GNU sort
112 is also available as "gnusort" or the default can be set back to
113 GNU sort by setting WITH_GNU_SORT. In this case, BSD sort will be
114 installed as "bsdsort".
117 A new version of ZFS (pool version 5000) has been merged to -HEAD.
118 Starting with this version the old system of ZFS pool versioning
119 is superseded by "feature flags". This concept enables forward
120 compatibility against certain future changes in functionality of ZFS
121 pools. The first read-only compatible "feature flag" for ZFS pools
122 is named "com.delphix:async_destroy". For more information
123 read the new zpool-features(5) manual page.
124 Please refer to the "ZFS notes" section of this file for information
125 on upgrading boot ZFS pools.
128 The malloc(3) implementation embedded in libc now uses sources imported
129 as contrib/jemalloc. The most disruptive API change is to
130 /etc/malloc.conf. If your system has an old-style /etc/malloc.conf,
131 delete it prior to installworld, and optionally re-create it using the
132 new format after rebooting. See malloc.conf(5) for details
133 (specifically the TUNING section and the "opt.*" entries in the MALLCTL
137 Big-endian MIPS TARGET_ARCH values no longer end in "eb". mips64eb
138 is now spelled mips64. mipsn32eb is now spelled mipsn32. mipseb is
139 now spelled mips. This is to aid compatibility with third-party
140 software that expects this naming scheme in uname(3). Little-endian
141 settings are unchanged.
144 Disable by default the option VFS_ALLOW_NONMPSAFE for all supported
148 Now unix domain sockets behave "as expected" on nullfs(5). Previously
149 nullfs(5) did not pass through all behaviours to the underlying layer,
150 as a result if we bound to a socket on the lower layer we could connect
151 only to the lower path; if we bound to the upper layer we could connect
152 only to the upper path. The new behavior is one can connect to both the
153 lower and the upper paths regardless what layer path one binds to.
156 The getifaddrs upgrade path broken with 20111215 has been restored.
157 If you have upgraded in between 20111215 and 20120209 you need to
158 recompile libc again with your kernel. You still need to recompile
159 world to be able to configure CARP but this restriction already
163 The set_rcvar() function has been removed from /etc/rc.subr. All
164 base and ports rc.d scripts have been updated, so if you have a
165 port installed with a script in /usr/local/etc/rc.d you can either
166 hand-edit the rcvar= line, or reinstall the port.
168 An easy way to handle the mass-update of /etc/rc.d:
169 rm /etc/rc.d/* && mergemaster -i
172 panic(9) now stops other CPUs in the SMP systems, disables interrupts
173 on the current CPU and prevents other threads from running.
174 This behavior can be reverted using the kern.stop_scheduler_on_panic
176 The new behavior can be incompatible with kern.sync_on_panic.
179 The carp(4) facility has been changed significantly. Configuration
180 of the CARP protocol via ifconfig(8) has changed, as well as format
181 of CARP events submitted to devd(8) has changed. See manual pages
182 for more information. The arpbalance feature of carp(4) is currently
183 not supported anymore.
185 Size of struct in_aliasreq, struct in6_aliasreq has changed. User
186 utilities using SIOCAIFADDR, SIOCAIFADDR_IN6, e.g. ifconfig(8),
187 need to be recompiled.
190 The acpi_wmi(4) status device /dev/wmistat has been renamed to
194 The option VFS_ALLOW_NONMPSAFE option has been added in order to
195 explicitely support non-MPSAFE filesystems.
196 It is on by default for all supported platform at this present
200 The broken amd(4) driver has been replaced with esp(4) in the amd64,
201 i386 and pc98 GENERIC kernel configuration files.
204 sysinstall has been removed
207 The stable/9 branch created in subversion. This corresponds to the
208 RELENG_9 branch in CVS.
211 This commit modifies vfs_register() so that it uses a hash
212 calculation to set vfc_typenum, which is enabled by default.
213 The first time a system is booted after this change, the
214 vfc_typenum values will change for all file systems. The
215 main effect of this is a change to the NFS server file handles
216 for file systems that use vfc_typenum in their fsid, such as ZFS.
217 It will, however, prevent vfc_typenum from changing when file
218 systems are loaded in a different order for subsequent reboots.
219 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
220 until you are ready to remount all NFS clients after a reboot.
223 Bump the shared library version numbers for libraries that
224 do not use symbol versioning, have changed the ABI compared
225 to stable/8 and which shared library version was not bumped.
226 Done as part of 9.0-RELEASE cycle.
229 During the merge of Capsicum features, the fget(9) KPI was modified.
230 This may require the rebuilding of out-of-tree device drivers --
231 issues have been reported specifically with the nVidia device driver.
232 __FreeBSD_version is bumped to 900041.
234 Also, there is a period between 20110811 and 20110814 where the
235 special devices /dev/{stdin,stdout,stderr} did not work correctly.
236 Building world from a kernel during that window may not work.
239 The packet filter (pf) code has been updated to OpenBSD 4.5.
240 You need to update userland tools to be in sync with kernel.
241 This update breaks backward compatibility with earlier pfsync(4)
242 versions. Care must be taken when updating redundant firewall setups.
245 The following sysctls and tunables are retired on x86 platforms:
247 machdep.hlt_logical_cpus
248 The following sysctl is retired:
249 machdep.hyperthreading_allowed
250 The sysctls were supposed to provide a way to dynamically offline and
251 online selected CPUs on x86 platforms, but the implementation has not
252 been reliable especially with SCHED_ULE scheduler.
253 machdep.hyperthreading_allowed tunable is still available to ignore
254 hyperthreading CPUs at OS level.
255 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
256 where X is an APIC ID of a CPU. Be advised, though, that disabling
257 CPUs in non-uniform fashion will result in non-uniform topology and
258 may lead to sub-optimal system performance with SCHED_ULE, which is
262 cpumask_t type is retired and cpuset_t is used in order to describe
266 Changes to ifconfig(8) for dynamic address family detection mandate
267 that you are running a kernel of 20110525 or later. Make sure to
268 follow the update procedure to boot a new kernel before installing
272 Support for sun4v architecture is officially dropped
275 Several KPI breaking changes have been committed to the mii(4) layer,
276 the PHY drivers and consequently some Ethernet drivers using mii(4).
277 This means that miibus.ko and the modules of the affected Ethernet
278 drivers need to be recompiled.
280 Note to kernel developers: Given that the OUI bit reversion problem
281 was fixed as part of these changes all mii(4) commits related to OUIs,
282 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
283 handling, no longer can be merged verbatim to stable/8 and previous
287 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
288 into their kernel configurations along with 'device pci'.
291 The default NFS client is now the new NFS client, so fstype "newnfs"
292 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
293 Although mounts via fstype "nfs" will usually work without userland
294 changes, it is recommended that the mount(8) and mount_nfs(8)
295 commands be rebuilt from sources and that a link to mount_nfs called
296 mount_oldnfs be created. The new client is compiled into the
297 kernel with "options NFSCL" and this is needed for diskless root
298 file systems. The GENERIC kernel configs have been changed to use
299 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
300 To use the regular/old client, you can "mount -t oldnfs ...". For
301 a diskless root file system, you must also include a line like:
303 vfs.root.mountfrom="oldnfs:"
305 in the boot/loader.conf on the root fs on the NFS server to make
306 a diskless root fs use the old client.
309 The GENERIC kernels for all architectures now default to the new
310 CAM-based ATA stack. It means that all legacy ATA drivers were
311 removed and replaced by respective CAM drivers. If you are using
312 ATA device names in /etc/fstab or other places, make sure to update
313 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
314 where 'Y's are the sequential numbers starting from zero for each type
315 in order of detection, unless configured otherwise with tunables,
316 see cam(4)). There will be symbolic links created in /dev/ to map
317 old adX devices to the respective adaY. They should provide basic
318 compatibility for file systems mounting in most cases, but they do
319 not support old user-level APIs and do not have respective providers
320 in GEOM. Consider using updated management tools with new device names.
322 It is possible to load devices ahci, ata, siis and mvs as modules,
323 but option ATA_CAM should remain in kernel configuration to make ata
324 module work as CAM driver supporting legacy ATA controllers. Device ata
325 still can be used in modular fashion (atacore + ...). Modules atadisk
326 and atapi* are not used and won't affect operation in ATA_CAM mode.
327 Note that to use CAM-based ATA kernel should include CAM devices
328 scbus, pass, da (or explicitly ada), cd and optionally others. All of
329 them are parts of the cam module.
331 ataraid(4) functionality is now supported by the RAID GEOM class.
332 To use it you can load geom_raid kernel module and use graid(8) tool
333 for management. Instead of /dev/arX device names, use /dev/raid/rX.
335 No kernel config options or code have been removed, so if a problem
336 arises, please report it and optionally revert to the old ATA stack.
337 In order to do it you can remove from the kernel config:
342 , and instead add back:
343 device atadisk # ATA disk drives
344 device ataraid # ATA RAID drives
345 device atapicd # ATAPI CDROM drives
346 device atapifd # ATAPI floppy drives
347 device atapist # ATAPI tape drives
350 The default NFS server has been changed to the new server, which
351 was referred to as the experimental server. If you need to switch
352 back to the old NFS server, you must now put the "-o" option on
353 both the mountd and nfsd commands. This can be done using the
354 mountd_flags and nfs_server_flags rc.conf variables until an
355 update to the rc scripts is committed, which is coming soon.
358 The GNU Objective-C runtime library (libobjc), and other Objective-C
359 related components have been removed from the base system. If you
360 require an Objective-C library, please use one of the available ports.
363 ath(4) has been split into bus- and device- modules. if_ath contains
364 the HAL, the TX rate control and the network device code. if_ath_pci
365 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
366 contains the AHB glue. Users need to load both if_ath_pci and if_ath
367 in order to use ath on everything else.
369 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
370 need to load if_ath and if_ath_pci for ath(4) operation.
373 As part of the replacement of sysinstall, the process of building
374 release media has changed significantly. For details, please re-read
375 release(7), which has been updated to reflect the new build process.
378 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
379 is the last available version under GPLv2. It brings a number of new
380 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
381 4.1 and SSE 4.2), better support for powerpc64, a number of new
382 directives, and lots of other small improvements. See the ChangeLog
383 file in contrib/binutils for the full details.
386 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
387 compliant, and will now use half of hash for authentication.
388 This will break interoperability with all stacks (including all
389 actual FreeBSD versions) who implement
390 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
392 The only workaround with such peers is to use another HMAC
393 algorithm for IPsec ("phase 2") authentication.
396 Remove the uio_yield prototype and symbol. This function has
397 been misnamed since it was introduced and should not be
398 globally exposed with this name. The equivalent functionality
399 is now available using kern_yield(curthread->td_user_pri).
400 The function remains undocumented.
403 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
404 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
405 for scalar sysctls is defined but disabled. Code that needs
406 UQUAD to pass the type checking that must compile on older
407 systems where the define is not present can check against
408 __FreeBSD_version >= 900030.
410 The system dialog(1) has been replaced with a new version previously
411 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
412 with the previous version, but the libdialog associated with it has
413 a largely incompatible API. As such, the original version of libdialog
414 will be kept temporarily as libodialog, until its base system consumers
415 are replaced or updated. Bump __FreeBSD_version to 900030.
418 If you are trying to run make universe on a -stable system, and you get
419 the following warning:
420 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
421 unknown. config(8) likely too old."
422 or something similar to it, then you must upgrade your -stable system
423 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
424 stable/8) or build the config from the latest stable/8 branch and
425 install it on your system.
427 Prior to this date, building a current universe on 8-stable system from
428 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
429 error in the first kernel build phase. A new config on those old
430 systems will fix that problem for older versions of -current.
433 The TCP stack has been modified to allow Khelp modules to interact with
434 it via helper hook points and store per-connection data in the TCP
435 control block. Bump __FreeBSD_version to 900029. User space tools that
436 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
440 Generic IEEE 802.3 annex 31B full duplex flow control support has been
441 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
442 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
443 to take advantage of it instead of using custom implementations. This
444 means that these drivers now no longer unconditionally advertise
445 support for flow control but only do so if flow control is a selected
446 media option. This was implemented in the generic support that way in
447 order to allow flow control to be switched on and off via ifconfig(8)
448 with the PHY specific default to typically off in order to protect
449 from unwanted effects. Consequently, if you used flow control with
450 one of the above mentioned drivers you now need to explicitly enable
452 ifconfig bge0 media auto mediaopt flowcontrol
454 Along with the above mentioned changes generic support for setting
455 1000baseT master mode also has been added and brgphy(4), ciphy(4),
456 e1000phy(4) as well as ip1000phy(4) have been converted to take
457 advantage of it. This means that these drivers now no longer take the
458 link0 parameter for selecting master mode but the master media option
459 has to be used instead, for example like in the following:
460 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
462 Selection of master mode now is also available with all other PHY
463 drivers supporting 1000baseT.
466 The TCP stack has received a significant update to add support for
467 modularised congestion control and generally improve the clarity of
468 congestion control decisions. Bump __FreeBSD_version to 900025. User
469 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
470 sockstat) need to be recompiled.
473 The man(1) utility has been replaced by a new version that no longer
474 uses /etc/manpath.config. Please consult man.conf(5) for how to
475 migrate local entries to the new format.
478 The copyright strings printed by login(1) and sshd(8) at the time of a
479 new connection have been removed to follow other operating systems and
483 A workaround for a fixed ld bug has been removed in kernel code,
484 so make sure that your system ld is built from sources after
485 revision 210245 from 2010-07-19 (r211583 if building head kernel
486 on stable/8, r211584 for stable/7; both from 2010-08-21).
487 A symptom of incorrect ld version is different addresses for
488 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
491 The $ipv6_prefer variable in rc.conf(5) has been split into
492 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
494 The $ip6addrctl_policy is a variable to choose a pre-defined
495 address selection policy set by ip6addrctl(8). A value
496 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
499 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
500 flag (see an entry of 20090926) is set on an interface with no
501 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
502 security reason. If you want IPv6 link-local address on all
503 interfaces by default, set this to "YES".
505 The old ipv6_prefer="YES" is equivalent to
506 ipv6_activate_all_interfaces="YES" and
507 ip6addrctl_policy="ipv6_prefer".
510 DTrace has grown support for userland tracing. Due to this, DTrace is
511 now i386 and amd64 only.
512 dtruss(1) is now installed by default on those systems and a new
513 kernel module is needed for userland tracing: fasttrap.
514 No changes to your kernel config file are necessary to enable
515 userland tracing, but you might consider adding 'STRIP=' and
516 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
517 to have informative userland stack traces in DTrace (ustack).
520 The acpi_aiboost(4) driver has been removed in favor of the new
521 aibs(4) driver. You should update your kernel configuration file.
524 BSD grep has been imported to the base system and it is built by
525 default. It is completely BSD licensed, highly GNU-compatible, uses
526 less memory than its GNU counterpart and has a small codebase.
527 However, it is slower than its GNU counterpart, which is mostly
528 noticeable for larger searches, for smaller ones it is measurable
529 but not significant. The reason is complex, the most important factor
530 is that we lack a modern and efficient regex library and GNU
531 overcomes this by optimizing the searches internally. Future work
532 on improving the regex performance is planned, for the meantime,
533 users that need better performance, can build GNU grep instead by
534 setting the WITH_GNU_GREP knob.
537 Due to the import of powerpc64 support, all existing powerpc kernel
538 configuration files must be updated with a machine directive like this:
539 machine powerpc powerpc
541 In addition, an updated config(8) is required to build powerpc kernels
545 A new version of ZFS (version 15) has been merged to -HEAD.
546 This version uses a python library for the following subcommands:
547 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
548 For full functionality of these commands the following port must
549 be installed: sysutils/py-zfs
552 'vm_page's are now hashed by physical address to an array of mutexes.
553 Currently this is only used to serialize access to hold_count. Over
554 time the page queue mutex will be peeled away. This changes the size
555 of pmap on every architecture. And requires all callers of vm_page_hold
556 and vm_page_unhold to be updated.
559 WITH_CTF can now be specified in src.conf (not recommended, there
560 are some problems with static executables), make.conf (would also
561 affect ports which do not use GNU make and do not override the
562 compile targets) or in the kernel config (via "makeoptions
564 When WITH_CTF was specified there before this was silently ignored,
565 so make sure that WITH_CTF is not used in places which could lead
566 to unwanted behavior.
569 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
570 to allow 32-bit compatibility on non-x86 platforms. All kernel
571 configurations on amd64 and ia64 platforms using these options must
572 be modified accordingly.
575 The utmp user accounting database has been replaced with utmpx,
576 the user accounting interface standardized by POSIX.
577 Unfortunately the semantics of utmp and utmpx don't match,
578 making it practically impossible to support both interfaces.
579 The user accounting database is used by tools like finger(1),
580 last(1), talk(1), w(1) and ac(8).
582 All applications in the base system use utmpx. This means only
583 local binaries (e.g. from the ports tree) may still use these
584 utmp database files. These applications must be rebuilt to make
587 After the system has been upgraded, it is safe to remove the old
588 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
589 assuming their contents is of no importance anymore. Old wtmp
590 databases can only be used by last(1) and ac(8) after they have
591 been converted to the new format using wtmpcvt(1).
594 Introduce the kernel thread "deadlock resolver" (which can be enabled
595 via the DEADLKRES option, see NOTES for more details) and the
596 sleepq_type() function for sleepqueues.
599 The rc.firewall and rc.firewall6 were unified, and
600 rc.firewall6 and rc.d/ip6fw were removed.
601 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
602 variables are obsoleted. Instead, the following new rc
603 variables are added to rc.d/ipfw:
605 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
606 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
607 firewall_simple_onet_ipv6, firewall_trusted_ipv6
609 The meanings correspond to the relevant IPv4 variables.
615 The default terminal emulation for syscons(4) has been changed
616 from cons25 to xterm on all platforms except pc98. This means
617 that the /etc/ttys file needs to be updated to ensure correct
618 operation of applications on the console.
620 The terminal emulation style can be toggled per window by using
621 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
622 options can be used to change the compile-time default back to
625 To prevent graphical artifacts, make sure the TERM environment
626 variable is set to match the terminal emulation that is being
627 performed by syscons(4).
630 The layout of the structure ieee80211req_scan_result has changed.
631 Applications that require wireless scan results (e.g. ifconfig(8))
632 from net80211 need to be recompiled.
634 Applications such as wpa_supplicant(8) may require a full world
635 build without using NO_CLEAN in order to get synchronized with the
639 The iwn(4) driver has been updated to support the 5000 and 5150 series.
640 There's one kernel module for each firmware. Adding "device iwnfw"
641 to the kernel configuration file means including all three firmware
642 images inside the kernel. If you want to include just the one for
643 your wireless card, use the devices iwn4965fw, iwn5000fw or
647 The rc.d/network_ipv6, IPv6 configuration script has been integrated
648 into rc.d/netif. The changes are the following:
650 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
651 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
652 Note that both variables need the "inet6" keyword at the head.
654 Do not set $ipv6_network_interfaces manually if you do not
655 understand what you are doing. It is not needed in most cases.
657 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
660 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
661 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
663 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
664 all configured interfaces have "inet6 accept_rtadv" in the
665 $ifconfig_IF_ipv6. These are for backward compatibility.
667 3. A new variable $ipv6_prefer has been added. If NO, IPv6
668 functionality of interfaces with no corresponding
669 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
670 and the default address selection policy of ip6addrctl(8)
671 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
672 Note that if you want to configure IPv6 functionality on the
673 disabled interfaces after boot, first you need to clear the flag by
674 using ifconfig(8) like:
676 ifconfig em0 inet6 -ifdisabled
678 If YES, the default address selection policy is set as
681 The default value of $ipv6_prefer is NO.
683 4. If your system need to receive Router Advertisement messages,
684 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
685 scripts automatically invoke rtsol(8) when the interface becomes
686 UP. The Router Advertisement messages are used for SLAAC
687 (State-Less Address AutoConfiguration).
690 802.11s D3.03 support was committed. This is incompatible with the
691 previous code, which was based on D3.0.
694 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
695 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
696 control whether accepting Router Advertisement messages or not.
697 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
698 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
699 The ifconfig(8) utility now supports these flags.
702 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
703 mount(8) and -a option for df(1) to see them.
706 The old tunable hw.bus.devctl_disable has been superseded by
707 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
708 replaced by hw.bus.devctl_queue=0. The default for this new tunable
712 Remove the option STOP_NMI. The default action is now to use NMI only
713 for KDB via the newly introduced function stop_cpus_hard() and
714 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
717 The stable/8 branch created in subversion. This corresponds to the
718 RELENG_8 branch in CVS.
721 Bump the shared library version numbers for all libraries that do not
722 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
723 __FreeBSD_version to 800105.
726 Due to changes in the implementation of virtual network stack support,
727 all network-related kernel modules must be recompiled. As this change
728 breaks the ABI, bump __FreeBSD_version to 800104.
731 The TOE interface to the TCP syncache has been modified to remove
732 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
733 The cxgb driver is the only TOE consumer affected by this change, and
734 needs to be recompiled along with the kernel. As this change breaks
735 the ABI, bump __FreeBSD_version to 800103.
738 Padding has been added to struct tcpcb, sackhint and tcpstat in
739 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
740 maintaining the ABI. However, this change breaks the ABI, so bump
741 __FreeBSD_version to 800102. User space tools that rely on the size of
742 any of these structs (e.g. sockstat) need to be recompiled.
745 The NFS_LEGACYRPC option has been removed along with the old kernel
746 RPC implementation that this option selected. Kernel configurations
747 may need to be adjusted.
750 The network interface device nodes at /dev/net/<interface> have been
751 removed. All ioctl operations can be performed the normal way using
752 routing sockets. The kqueue functionality can generally be replaced
753 with routing sockets.
756 The documentation from the FreeBSD Documentation Project (Handbook,
757 FAQ, etc.) is now installed via packages by sysinstall(8) and under
758 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
761 The ABI of various structures related to the SYSV IPC API have been
762 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
763 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
767 Layout of struct vnet has changed as routing related variables were
768 moved to their own Vimage module. Modules need to be recompiled. Bump
769 __FreeBSD_version to 800099.
772 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
773 respectively. As long as no more than 16 groups per process are used,
774 no changes should be visible. When more than 16 groups are used, old
775 binaries may fail if they call getgroups() or getgrouplist() with
776 statically sized storage. Recompiling will work around this, but
777 applications should be modified to use dynamically allocated storage
778 for group arrays as POSIX.1-2008 does not cap an implementation's
779 number of supported groups at NGROUPS_MAX+1 as previous versions did.
781 NFS and portalfs mounts may also be affected as the list of groups is
782 truncated to 16. Users of NFS who use more than 16 groups, should
783 take care that negative group permissions are not used on the exported
784 file systems as they will not be reliable unless a GSSAPI based
785 authentication method is used.
788 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
789 option compiles in the support for adaptive spinning for lockmgrs
790 which want to enable it. The lockinit() function now accepts the flag
791 LK_ADAPTIVE in order to make the lock object subject to adaptive
792 spinning when both held in write and read mode.
795 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
796 changed. User applications that use this ioctl need to be rebuilt.
799 The layout of struct thread has changed. Kernel and modules need to
803 The layout of structs ifnet, domain, protosw and vnet_net has changed.
804 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
807 window(1) has been removed from the base system. It can now be
808 installed from ports. The port is called misc/window.
811 The way we are storing and accessing `routing table' entries has
812 changed. Programs reading the FIB, like netstat, need to be
816 A new netisr implementation has been added for FreeBSD 8. Network
817 file system modules, such as igmp, ipdivert, and others, should be
819 Bump __FreeBSD_version to 800096.
822 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
826 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
827 Bump __FreeBSD_version to 800094.
830 Add mnt_xflag field to 'struct mount'. File system modules need to be
832 Bump __FreeBSD_version to 800093.
835 The compiling option ADAPTIVE_SX has been retired while it has been
836 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
837 The KPI for sx_init_flags() changes as accepting flags:
838 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
839 been introduced in order to handle the reversed logic.
840 Bump __FreeBSD_version to 800092.
843 Add support for hierarchical jails. Remove global securelevel.
844 Bump __FreeBSD_version to 800091.
847 The layout of struct vnet_net has changed, therefore modules
849 Bump __FreeBSD_version to 800090.
852 The newly imported zic(8) produces a new format in the output. Please
853 run tzsetup(8) to install the newly created data to /etc/localtime.
856 The sysctl tree for the usb stack has renamed from hw.usb2.* to
857 hw.usb.* and is now consistent again with previous releases.
860 802.11 monitor mode support was revised and driver api's were changed.
861 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
862 of DLT_IEEE802_11. No user-visible data structures were changed but
863 applications that use DLT_IEEE802_11 may require changes.
864 Bump __FreeBSD_version to 800088.
867 The layout of the following structs has changed: sysctl_oid,
868 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
869 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
870 panics may be experienced. World rebuild is required for
871 correctly checking networking state from userland.
872 Bump __FreeBSD_version to 800085.
875 MLDv2 and Source-Specific Multicast (SSM) have been merged
876 to the IPv6 stack. VIMAGE hooks are in but not yet used.
877 The implementation of SSM within FreeBSD's IPv6 stack closely
878 follows the IPv4 implementation.
880 For kernel developers:
882 * The most important changes are that the ip6_output() and
883 ip6_input() paths no longer take the IN6_MULTI_LOCK,
884 and this lock has been downgraded to a non-recursive mutex.
886 * As with the changes to the IPv4 stack to support SSM, filtering
887 of inbound multicast traffic must now be performed by transport
888 protocols within the IPv6 stack. This does not apply to TCP and
889 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
891 * The KPIs used by IPv6 multicast are similar to those used by
892 the IPv4 stack, with the following differences:
893 * im6o_mc_filter() is analogous to imo_multicast_filter().
894 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
895 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
896 * IN6_LOOKUP_MULTI() has been deprecated and removed.
897 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
898 for MLDv1 have an additional 'timer' argument which is used to
899 jitter the initial membership report for the solicited-node
900 multicast membership on-link.
901 * This is not strictly needed for MLDv2, which already jitters
902 its report transmissions. However, the 'timer' argument is
903 preserved in case MLDv1 is active on the interface.
905 * The KAME linked-list based IPv6 membership implementation has
906 been refactored to use a vector similar to that used by the IPv4
908 Code which maintains a list of its own multicast memberships
909 internally, e.g. carp, has been updated to reflect the new
912 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
913 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
914 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
915 implementation constraint which needs to be addressed in HEAD.
917 For application developers:
919 * The changes are broadly similar to those made for the IPv4
922 * The use of IPv4 and IPv6 multicast socket options on the same
923 socket, using mapped addresses, HAS NOT been tested or supported.
925 * There are a number of issues with the implementation of various
926 IPv6 multicast APIs which need to be resolved in the API surface
927 before the implementation is fully compatible with KAME userland
928 use, and these are mostly to do with interface index treatment.
930 * The literature available discusses the use of either the delta / ASM
931 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
932 using setsourcefilter(3)/getsourcefilter(3). For more information
933 please refer to RFC 3768, 'Socket Interface Extensions for
934 Multicast Source Filters'.
936 * Applications which use the published RFC 3678 APIs should be fine.
938 For systems administrators:
940 * The mtest(8) utility has been refactored to support IPv6, in
941 addition to IPv4. Interface addresses are no longer accepted
942 as arguments, their names must be used instead. The utility
943 will map the interface name to its first IPv4 address as
944 returned by getifaddrs(3).
946 * The ifmcstat(8) utility has also been updated to print the MLDv2
947 endpoint state and source filter lists via sysctl(3).
949 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
950 loopback of IPv6 multicast datagrams by default; it defaults to 1
951 to preserve the existing behaviour. Disabling multicast loopback is
952 recommended for optimal system performance.
954 * The IPv6 MROUTING code has been changed to examine this sysctl
955 instead of attempting to perform a group lookup before looping
956 back forwarded datagrams.
958 Bump __FreeBSD_version to 800084.
961 Implement low-level Bluetooth HCI API.
962 Bump __FreeBSD_version to 800083.
965 The layout of struct malloc_type, used by modules to register new
966 memory allocation types, has changed. Most modules will need to
967 be rebuilt or panics may be experienced.
968 Bump __FreeBSD_version to 800081.
971 Anticipate overflowing inp_flags - add inp_flags2.
972 This changes most offsets in inpcb, so checking v4 connection
973 state will require a world rebuild.
974 Bump __FreeBSD_version to 800080.
977 Add an llentry to struct route and struct route_in6. Modules
978 embedding a struct route will need to be recompiled.
979 Bump __FreeBSD_version to 800079.
982 The size of rt_metrics_lite and by extension rtentry has changed.
983 Networking administration apps will need to be recompiled.
984 The route command now supports show as an alias for get, weighting
985 of routes, sticky and nostick flags to alter the behavior of stateful
987 Bump __FreeBSD_version to 800078.
990 Do not use Giant for kbdmux(4) locking. This is wrong and
991 apparently causing more problems than it solves. This will
992 re-open the issue where interrupt handlers may race with
993 kbdmux(4) in polling mode. Typical symptoms include (but
994 not limited to) duplicated and/or missing characters when
995 low level console functions (such as gets) are used while
996 interrupts are enabled (for example geli password prompt,
997 mountroot prompt etc.). Disabling kbdmux(4) may help.
1000 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
1001 kernel modules referencing any of the above need to be recompiled.
1002 Bump __FreeBSD_version to 800075.
1005 GEOM_PART has become the default partition slicer for storage devices,
1006 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
1007 introduces some changes:
1009 MSDOS/EBR: the devices created from MSDOS extended partition entries
1010 (EBR) can be named differently than with GEOM_MBR and are now symlinks
1011 to devices with offset-based names. fstabs may need to be modified.
1013 BSD: the "geometry does not match label" warning is harmless in most
1014 cases but it points to problems in file system misalignment with
1015 disk geometry. The "c" partition is now implicit, covers the whole
1016 top-level drive and cannot be (mis)used by users.
1018 General: Kernel dumps are now not allowed to be written to devices
1019 whose partition types indicate they are meant to be used for file
1020 systems (or, in case of MSDOS partitions, as something else than
1023 Most of these changes date approximately from 200812.
1026 The uscanner(4) driver has been removed from the kernel. This follows
1027 Linux removing theirs in 2.6 and making libusb the default interface
1028 (supported by sane).
1031 The multicast forwarding code has been cleaned up. netstat(1)
1032 only relies on KVM now for printing bandwidth upcall meters.
1033 The IPv4 and IPv6 modules are split into ip_mroute_mod and
1034 ip6_mroute_mod respectively. The config(5) options for statically
1035 compiling this code remain the same, i.e. 'options MROUTING'.
1038 Support for the IFF_NEEDSGIANT network interface flag has been
1039 removed, which means that non-MPSAFE network device drivers are no
1040 longer supported. In particular, if_ar, if_sr, and network device
1041 drivers from the old (legacy) USB stack can no longer be built or
1045 POSIX.1 Native Language Support (NLS) has been enabled in libc and
1046 a bunch of new language catalog files have also been added.
1047 This means that some common libc messages are now localized and
1048 they depend on the LC_MESSAGES environmental variable.
1051 The k8temp(4) driver has been renamed to amdtemp(4) since
1052 support for Family 10 and Family 11 CPU families was added.
1055 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1056 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1058 For kernel developers, the most important changes are that the
1059 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1060 and this lock has been downgraded to a non-recursive mutex.
1062 Transport protocols (UDP, Raw IP) are now responsible for filtering
1063 inbound multicast traffic according to group membership and source
1064 filters. The imo_multicast_filter() KPI exists for this purpose.
1065 Transports which do not use multicast (SCTP, TCP) already reject
1066 multicast by default. Forwarding and receive performance may improve
1067 as a mutex acquisition is no longer needed in the ip_input()
1068 low-level input path. in_addmulti() and in_delmulti() are shimmed
1069 to new KPIs which exist to support SSM in-kernel.
1071 For application developers, it is recommended that loopback of
1072 multicast datagrams be disabled for best performance, as this
1073 will still cause the lock to be taken for each looped-back
1074 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1075 be tuned to 0 to disable loopback by default; it defaults to 1
1076 to preserve the existing behaviour.
1078 For systems administrators, to obtain best performance with
1079 multicast reception and multiple groups, it is always recommended
1080 that a card with a suitably precise hash filter is used. Hash
1081 collisions will still result in the lock being taken within the
1082 transport protocol input path to check group membership.
1084 If deploying FreeBSD in an environment with IGMP snooping switches,
1085 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1086 enabled; this forces 224.0.0.0/24 group membership to be announced
1089 The size of 'struct igmpstat' has changed; netstat needs to be
1090 recompiled to reflect this.
1091 Bump __FreeBSD_version to 800070.
1094 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1095 updated to use it. This requires a buildworld/installworld in order to
1096 update the library and dependencies (usbconfig, etc). Its advisable to
1097 rebuild all ports which uses libusb. More specific directions are given
1098 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1099 libusb are no longer required and can be removed.
1102 A workaround is committed to allow the creation of System V shared
1103 memory segment of size > 2 GB on the 64-bit architectures.
1104 Due to a limitation of the existing ABI, the shm_segsz member
1105 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1106 wrong for large segments. Note that limits must be explicitly
1107 raised to allow such segments to be created.
1110 The layout of struct ifnet has changed, requiring a rebuild of all
1111 network device driver modules.
1114 The /dev handling for the new USB stack has changed, a
1115 buildworld/installworld is required for libusb20.
1118 The new USB2 stack has now been permanently moved in and all kernel and
1119 module names reverted to their previous values (eg, usb, ehci, ohci,
1120 ums, ...). The old usb stack can be compiled in by prefixing the name
1121 with the letter 'o', the old usb modules have been removed.
1122 Updating entry 20090216 for xorg and 20090215 for libmap may still
1126 The rc.conf(5) option if_up_delay has been renamed to
1127 defaultroute_delay to better reflect its purpose. If you have
1128 customized this setting in /etc/rc.conf you need to update it to
1132 xorg 7.4 wants to configure its input devices via hald which does not
1133 yet work with USB2. If the keyboard/mouse does not work in xorg then
1135 Option "AllowEmptyInput" "off"
1136 to your ServerLayout section. This will cause X to use the configured
1137 kbd and mouse sections from your xorg.conf.
1140 The GENERIC kernels for all architectures now default to the new USB2
1141 stack. No kernel config options or code have been removed so if a
1142 problem arises please report it and optionally revert to the old USB
1143 stack. If you are loading USB kernel modules or have a custom kernel
1144 that includes GENERIC then ensure that usb names are also changed over,
1145 eg uftdi -> usb2_serial_ftdi.
1147 Older programs linked against the ports libusb 0.1 need to be
1148 redirected to the new stack's libusb20. /etc/libmap.conf can
1150 # Map old usb library to new one for usb2 stack
1151 libusb-0.1.so.8 libusb20.so.1
1154 All USB ethernet devices now attach as interfaces under the name ueN
1155 (eg. ue0). This is to provide a predictable name as vendors often
1156 change usb chipsets in a product without notice.
1159 The ichsmb(4) driver has been changed to require SMBus slave
1160 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1161 All of the other SMBus controller drivers require left-justified
1162 slave addresses, so this change makes all the drivers provide the
1166 INET6 statistics (struct ip6stat) was updated.
1167 netstat(1) needs to be recompiled.
1170 NTFS has been removed from GENERIC kernel on amd64 to match
1171 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1172 will load ntfs.ko module automatically when NTFS support is
1173 actually needed, unless ntfs.ko is not installed or security
1174 level prohibits loading kernel modules. If either is the case,
1175 "options NTFS" has to be added into kernel config.
1178 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1179 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1180 800061. User space tools that rely on the size of struct tcpcb in
1181 tcp_var.h (e.g. sockstat) need to be recompiled.
1184 ng_tty(4) module updated to match the new TTY subsystem.
1185 Due to API change, user-level applications must be updated.
1186 New API support added to mpd5 CVS and expected to be present
1187 in next mpd5.3 release.
1190 With __FreeBSD_version 800060 the makefs tool is part of
1191 the base system (it was a port).
1194 The afdata and ifnet locks have been changed from mutexes to
1195 rwlocks, network modules will need to be re-compiled.
1198 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1199 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1200 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1201 architecture and 40 bytes on 64-bit architecture. The userland
1202 applications "arp" and "ndp" have been updated accordingly.
1203 The output from "netstat -r" shows only routing entries and
1204 none of the L2 information.
1207 __FreeBSD_version 800057 marks the switchover from the
1208 binary ath hal to source code. Users must add the line:
1210 options AH_SUPPORT_AR5416
1212 to their kernel config files when specifying:
1216 The ath_hal module no longer exists; the code is now compiled
1217 together with the driver in the ath module. It is now
1218 possible to tailor chip support (i.e. reduce the set of chips
1219 and thereby the code size); consult ath_hal(4) for details.
1222 __FreeBSD_version 800054 adds memory barriers to
1223 <machine/atomic.h>, new interfaces to ifnet to facilitate
1224 multiple hardware transmit queues for cards that support
1225 them, and a lock-less ring-buffer implementation to
1226 enable drivers to more efficiently manage queueing of
1230 A new version of ZFS (version 13) has been merged to -HEAD.
1231 This version has zpool attribute "listsnapshots" off by
1232 default, which means "zfs list" does not show snapshots,
1233 and is the same as Solaris behavior.
1236 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1239 The uhci, ohci, ehci and slhci USB Host controller drivers have
1240 been put into separate modules. If you load the usb module
1241 separately through loader.conf you will need to load the
1242 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1243 controller add the following to loader.conf:
1249 The ABI used by the PMC toolset has changed. Please keep
1250 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1254 atapci kernel module now includes only generic PCI ATA
1255 driver. AHCI driver moved to ataahci kernel module.
1256 All vendor-specific code moved into separate kernel modules:
1257 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1258 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1259 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1260 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1263 The TTY subsystem of the kernel has been replaced by a new
1264 implementation, which provides better scalability and an
1265 improved driver model. Most common drivers have been migrated to
1266 the new TTY subsystem, while others have not. The following
1267 drivers have not yet been ported to the new TTY layer:
1270 cy, digi, rc, rp, sio
1276 ng_h4, ng_tty, ppp, sl, snp
1278 Adding these drivers to your kernel configuration file shall
1279 cause compilation to fail.
1282 ntpd has been upgraded to 4.2.4p5.
1285 OpenSSH has been upgraded to 5.1p1.
1287 For many years, FreeBSD's version of OpenSSH preferred DSA
1288 over RSA for host and user authentication keys. With this
1289 upgrade, we've switched to the vendor's default of RSA over
1290 DSA. This may cause upgraded clients to warn about unknown
1291 host keys even for previously known hosts. Users should
1292 follow the usual procedure for verifying host keys before
1293 accepting the RSA key.
1295 This can be circumvented by setting the "HostKeyAlgorithms"
1296 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1299 Please note that the sequence of keys offered for
1300 authentication has been changed as well. You may want to
1301 specify IdentityFile in a different order to revert this
1305 The sio(4) driver has been removed from the i386 and amd64
1306 kernel configuration files. This means uart(4) is now the
1307 default serial port driver on those platforms as well.
1309 To prevent collisions with the sio(4) driver, the uart(4) driver
1310 uses different names for its device nodes. This means the
1311 onboard serial port will now most likely be called "ttyu0"
1312 instead of "ttyd0". You may need to reconfigure applications to
1313 use the new device names.
1315 When using the serial port as a boot console, be sure to update
1316 /boot/device.hints and /etc/ttys before booting the new kernel.
1317 If you forget to do so, you can still manually specify the hints
1318 at the loader prompt:
1320 set hint.uart.0.at="isa"
1321 set hint.uart.0.port="0x3F8"
1322 set hint.uart.0.flags="0x10"
1323 set hint.uart.0.irq="4"
1327 The gpt(8) utility has been removed. Use gpart(8) to partition
1331 The version that Linuxulator emulates was changed from 2.4.2
1332 to 2.6.16. If you experience any problems with Linux binaries
1333 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1334 if it fixes the problem contact emulation mailing list.
1337 ISDN4BSD (I4B) was removed from the src tree. You may need to
1338 update a your kernel configuration and remove relevant entries.
1341 I have checked in code to support multiple routing tables.
1342 See the man pages setfib(1) and setfib(2).
1343 This is a hopefully backwards compatible version,
1344 but to make use of it you need to compile your kernel
1345 with options ROUTETABLES=2 (or more up to 16).
1348 The 802.11 wireless support was redone to enable multi-bss
1349 operation on devices that are capable. The underlying device
1350 is no longer used directly but instead wlanX devices are
1351 cloned with ifconfig. This requires changes to rc.conf files.
1352 For example, change:
1353 ifconfig_ath0="WPA DHCP"
1356 ifconfig_wlan0="WPA DHCP"
1357 see rc.conf(5) for more details. In addition, mergemaster of
1358 /etc/rc.d is highly recommended. Simultaneous update of userland
1359 and kernel wouldn't hurt either.
1361 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1362 modules were merged into the base wlan module. All references
1363 to these modules (e.g. in kernel config files) must be removed.
1366 psm(4) has gained write(2) support in native operation level.
1367 Arbitrary commands can be written to /dev/psm%d and status can
1368 be read back from it. Therefore, an application is responsible
1369 for status validation and error recovery. It is a no-op in
1370 other operation levels.
1373 Support for KSE threading has been removed from the kernel. To
1374 run legacy applications linked against KSE libmap.conf may
1375 be used. The following libmap.conf may be used to ensure
1376 compatibility with any prior release:
1378 libpthread.so.1 libthr.so.1
1379 libpthread.so.2 libthr.so.2
1380 libkse.so.3 libthr.so.3
1383 The layout of struct vmspace has changed. This affects libkvm
1384 and any executables that link against libkvm and use the
1385 kvm_getprocs() function. In particular, but not exclusively,
1386 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1387 The effects are minimal, but it's advisable to upgrade world
1391 The latest em driver no longer has support in it for the
1392 82575 adapter, this is now moved to the igb driver. The
1393 split was done to make new features that are incompatible
1394 with older hardware easier to do.
1397 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1398 likewise the kernel option is now GEOM_LINUX_LVM.
1401 The default NFS mount mode has changed from UDP to TCP for
1402 increased reliability. If you rely on (insecurely) NFS
1403 mounting across a firewall you may need to update your
1407 Belatedly note the addition of m_collapse for compacting
1411 The fts(3) structures have been changed to use adequate
1412 integer types for their members and so to be able to cope
1413 with huge file trees. The old fts(3) ABI is preserved
1414 through symbol versioning in libc, so third-party binaries
1415 using fts(3) should still work, although they will not take
1416 advantage of the extended types. At the same time, some
1417 third-party software might fail to build after this change
1418 due to unportable assumptions made in its source code about
1419 fts(3) structure members. Such software should be fixed
1420 by its vendor or, in the worst case, in the ports tree.
1421 FreeBSD_version 800015 marks this change for the unlikely
1422 case that a portable fix is impossible.
1425 To upgrade to -current after this date, you must be running
1426 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1427 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1430 The ADAPTIVE_GIANT kernel option has been retired because its
1431 functionality is the default now.
1434 The AT keyboard emulation of sunkbd(4) has been turned on
1435 by default. In order to make the special symbols of the Sun
1436 keyboards driven by sunkbd(4) work under X these now have
1437 to be configured the same way as Sun USB keyboards driven
1438 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1440 Option "XkbLayout" "us"
1441 Option "XkbRules" "xorg"
1442 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1445 It has been decided that it is desirable to provide ABI
1446 backwards compatibility to the FreeBSD 4/5/6 versions of the
1447 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1448 broken with the introduction of PCI domain support (see the
1449 20070930 entry). Unfortunately, this required the ABI of
1450 PCIOCGETCONF to be broken again in order to be able to
1451 provide backwards compatibility to the old version of that
1452 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1453 again. As for prominent ports this affects neither pciutils
1454 nor xorg-server this time, the hal port needs to be rebuilt
1458 The misnamed kthread_create() and friends have been renamed
1459 to kproc_create() etc. Many of the callers already
1460 used kproc_start()..
1461 I will return kthread_create() and friends in a while
1462 with implementations that actually create threads, not procs.
1463 Renaming corresponds with version 800002.
1472 Avoid using make -j when upgrading. While generally safe, there are
1473 sometimes problems using -j to upgrade. If your upgrade fails with
1474 -j, please try again without -j. From time to time in the past there
1475 have been problems using -j with buildworld and/or installworld. This
1476 is especially true when upgrading between "distant" versions (eg one
1477 that cross a major release boundary or several minor releases, or when
1478 several months have passed on the -current branch).
1480 Sometimes, obscure build problems are the result of environment
1481 poisoning. This can happen because the make utility reads its
1482 environment when searching for values for global variables. To run
1483 your build attempts in an "environmental clean room", prefix all make
1484 commands with 'env -i '. See the env(1) manual page for more details.
1486 When upgrading from one major version to another it is generally best
1487 to upgrade to the latest code in the currently installed branch first,
1488 then do an upgrade to the new branch. This is the best-tested upgrade
1489 path, and has the highest probability of being successful. Please try
1490 this approach before reporting problems with a major version upgrade.
1494 When upgrading the boot ZFS pool to a new version, always follow
1497 1.) recompile and reinstall the ZFS boot loader and boot block
1498 (this is part of "make buildworld" and "make installworld")
1500 2.) update the ZFS boot block on your boot drive
1502 The following example updates the ZFS boot block on the first
1503 partition (freebsd-boot) of a GPT partitioned drive ad0:
1504 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1506 Non-boot pools do not need these updates.
1510 If you are updating from a prior version of FreeBSD (even one just
1511 a few days old), you should follow this procedure. It is the most
1512 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1514 make kernel-toolchain
1515 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1516 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1518 To test a kernel once
1519 ---------------------
1520 If you just want to boot a kernel once (because you are not sure
1521 if it works, or if you want to boot a known bad kernel to provide
1522 debugging information) run
1523 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1524 nextboot -k testkernel
1526 To just build a kernel when you know that it won't mess you up
1527 --------------------------------------------------------------
1528 This assumes you are already running a CURRENT system. Replace
1529 ${arch} with the architecture of your machine (e.g. "i386",
1530 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1532 cd src/sys/${arch}/conf
1533 config KERNEL_NAME_HERE
1534 cd ../compile/KERNEL_NAME_HERE
1539 If this fails, go to the "To build a kernel" section.
1541 To rebuild everything and install it on the current system.
1542 -----------------------------------------------------------
1543 # Note: sometimes if you are running current you gotta do more than
1544 # is listed here if you are upgrading from a really old current.
1546 <make sure you have good level 0 dumps>
1548 make kernel KERNCONF=YOUR_KERNEL_HERE
1550 <reboot in single user> [3]
1558 To cross-install current onto a separate partition
1559 --------------------------------------------------
1560 # In this approach we use a separate partition to hold
1561 # current's root, 'usr', and 'var' directories. A partition
1562 # holding "/", "/usr" and "/var" should be about 2GB in
1565 <make sure you have good level 0 dumps>
1568 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1569 <maybe newfs current's root partition>
1570 <mount current's root partition on directory ${CURRENT_ROOT}>
1571 make installworld DESTDIR=${CURRENT_ROOT}
1572 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1573 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1574 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1575 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1576 <reboot into current>
1577 <do a "native" rebuild/install as described in the previous section>
1578 <maybe install compatibility libraries from ports/misc/compat*>
1582 To upgrade in-place from 8.x-stable to current
1583 ----------------------------------------------
1584 <make sure you have good level 0 dumps>
1586 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1588 <reboot in single user> [3]
1595 Make sure that you've read the UPDATING file to understand the
1596 tweaks to various things you need. At this point in the life
1597 cycle of current, things change often and you are on your own
1598 to cope. The defaults can also change, so please read ALL of
1599 the UPDATING entries.
1601 Also, if you are tracking -current, you must be subscribed to
1602 freebsd-current@freebsd.org. Make sure that before you update
1603 your sources that you have read and understood all the recent
1604 messages there. If in doubt, please track -stable which has
1605 much fewer pitfalls.
1607 [1] If you have third party modules, such as vmware, you
1608 should disable them at this point so they don't crash your
1611 [3] From the bootblocks, boot -s, and then do
1616 adjkerntz -i # if CMOS is wall time
1617 Also, when doing a major release upgrade, it is required that
1618 you boot into single user mode to do the installworld.
1620 [4] Note: This step is non-optional. Failure to do this step
1621 can result in a significant reduction in the functionality of the
1622 system. Attempting to do it by hand is not recommended and those
1623 that pursue this avenue should read this file carefully, as well
1624 as the archives of freebsd-current and freebsd-hackers mailing lists
1625 for potential gotchas. The -U option is also useful to consider.
1626 See mergemaster(8) for more information.
1628 [5] Usually this step is a noop. However, from time to time
1629 you may need to do this if you get unknown user in the following
1630 step. It never hurts to do it all the time. You may need to
1631 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1632 install) after the buildworld before this step if you last updated
1633 from current before 20020224 or from -stable before 20020408.
1635 [6] This only deletes old files and directories. Old libraries
1636 can be deleted by "make delete-old-libs", but you have to make
1637 sure that no program is using those libraries anymore.
1639 [8] In order to have a kernel that can run the 4.x binaries needed to
1640 do an installworld, you must include the COMPAT_FREEBSD4 option in
1641 your kernel. Failure to do so may leave you with a system that is
1642 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1643 required to run the 5.x binaries on more recent kernels. And so on
1644 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1646 Make sure that you merge any new devices from GENERIC since the
1647 last time you updated your kernel config file.
1649 [9] When checking out sources, you must include the -P flag to have
1650 cvs prune empty directories.
1652 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1653 "?=" instead of the "=" assignment operator, so that buildworld can
1654 override the CPUTYPE if it needs to.
1656 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1657 not on the command line, or in /etc/make.conf. buildworld will
1658 warn if it is improperly defined.
1661 This file contains a list, in reverse chronological order, of major
1662 breakages in tracking -current. It is not guaranteed to be a complete
1663 list of such breakages, and only contains entries since October 10, 2007.
1664 If you need to see UPDATING entries from before that date, you will need
1665 to fetch an UPDATING file from an older FreeBSD release.
1667 Copyright information:
1669 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1671 Redistribution, publication, translation and use, with or without
1672 modification, in full or in part, in any form or format of this
1673 document are permitted without further permission from the author.
1675 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1676 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1677 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1678 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1679 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1680 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1681 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1682 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1683 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1684 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1685 POSSIBILITY OF SUCH DAMAGE.
1687 Contact Warner Losh if you have any questions about your use of