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 A new version of ZFS (pool version 5000) has been merged to -HEAD.
29 Starting with this version the old system of ZFS pool versioning
30 is superseded by "feature flags". This concept enables forward
31 compatibility against certain future changes in functionality of ZFS
32 pools. The first read-only compatible "feature flag" for ZFS pools
33 is named "com.delphix:async_destroy". For more information
34 read the new zpool-features(5) manual page.
35 Please refer to the "ZFS notes" section of this file for information
36 on upgrading boot ZFS pools.
39 The malloc(3) implementation embedded in libc now uses sources imported
40 as contrib/jemalloc. The most disruptive API change is to
41 /etc/malloc.conf. If your system has an old-style /etc/malloc.conf,
42 delete it prior to installworld, and optionally re-create it using the
43 new format after rebooting. See malloc.conf(5) for details
44 (specifically the TUNING section and the "opt.*" entries in the MALLCTL
48 Big-endian MIPS TARGET_ARCH values no longer end in "eb". mips64eb
49 is now spelled mips64. mipsn32eb is now spelled mipsn32. mipseb is
50 now spelled mips. This is to aid compatibility with third-party
51 software that expects this naming scheme in uname(3). Little-endian
52 settings are unchanged.
55 Disable by default the option VFS_ALLOW_NONMPSAFE for all supported
59 Now unix domain sockets behave "as expected" on nullfs(5). Previously
60 nullfs(5) did not pass through all behaviours to the underlying layer,
61 as a result if we bound to a socket on the lower layer we could connect
62 only to the lower path; if we bound to the upper layer we could connect
63 only to the upper path. The new behavior is one can connect to both the
64 lower and the upper paths regardless what layer path one binds to.
67 The getifaddrs upgrade path broken with 20111215 has been restored.
68 If you have upgraded in between 20111215 and 20120209 you need to
69 recompile libc again with your kernel. You still need to recompile
70 world to be able to configure CARP but this restriction already
74 The set_rcvar() function has been removed from /etc/rc.subr. All
75 base and ports rc.d scripts have been updated, so if you have a
76 port installed with a script in /usr/local/etc/rc.d you can either
77 hand-edit the rcvar= line, or reinstall the port.
79 An easy way to handle the mass-update of /etc/rc.d:
80 rm /etc/rc.d/* && mergemaster -i
83 panic(9) now stops other CPUs in the SMP systems, disables interrupts
84 on the current CPU and prevents other threads from running.
85 This behavior can be reverted using the kern.stop_scheduler_on_panic
87 The new behavior can be incompatible with kern.sync_on_panic.
90 The carp(4) facility has been changed significantly. Configuration
91 of the CARP protocol via ifconfig(8) has changed, as well as format
92 of CARP events submitted to devd(8) has changed. See manual pages
93 for more information. The arpbalance feature of carp(4) is currently
94 not supported anymore.
96 Size of struct in_aliasreq, struct in6_aliasreq has changed. User
97 utilities using SIOCAIFADDR, SIOCAIFADDR_IN6, e.g. ifconfig(8),
98 need to be recompiled.
101 The acpi_wmi(4) status device /dev/wmistat has been renamed to
105 The option VFS_ALLOW_NONMPSAFE option has been added in order to
106 explicitely support non-MPSAFE filesystems.
107 It is on by default for all supported platform at this present
111 The broken amd(4) driver has been replaced with esp(4) in the amd64,
112 i386 and pc98 GENERIC kernel configuration files.
115 sysinstall has been removed
118 The stable/9 branch created in subversion. This corresponds to the
119 RELENG_9 branch in CVS.
122 This commit modifies vfs_register() so that it uses a hash
123 calculation to set vfc_typenum, which is enabled by default.
124 The first time a system is booted after this change, the
125 vfc_typenum values will change for all file systems. The
126 main effect of this is a change to the NFS server file handles
127 for file systems that use vfc_typenum in their fsid, such as ZFS.
128 It will, however, prevent vfc_typenum from changing when file
129 systems are loaded in a different order for subsequent reboots.
130 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
131 until you are ready to remount all NFS clients after a reboot.
134 Bump the shared library version numbers for libraries that
135 do not use symbol versioning, have changed the ABI compared
136 to stable/8 and which shared library version was not bumped.
137 Done as part of 9.0-RELEASE cycle.
140 During the merge of Capsicum features, the fget(9) KPI was modified.
141 This may require the rebuilding of out-of-tree device drivers --
142 issues have been reported specifically with the nVidia device driver.
143 __FreeBSD_version is bumped to 900041.
145 Also, there is a period between 20110811 and 20110814 where the
146 special devices /dev/{stdin,stdout,stderr} did not work correctly.
147 Building world from a kernel during that window may not work.
150 The packet filter (pf) code has been updated to OpenBSD 4.5.
151 You need to update userland tools to be in sync with kernel.
152 This update breaks backward compatibility with earlier pfsync(4)
153 versions. Care must be taken when updating redundant firewall setups.
156 The following sysctls and tunables are retired on x86 platforms:
158 machdep.hlt_logical_cpus
159 The following sysctl is retired:
160 machdep.hyperthreading_allowed
161 The sysctls were supposed to provide a way to dynamically offline and
162 online selected CPUs on x86 platforms, but the implementation has not
163 been reliable especially with SCHED_ULE scheduler.
164 machdep.hyperthreading_allowed tunable is still available to ignore
165 hyperthreading CPUs at OS level.
166 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
167 where X is an APIC ID of a CPU. Be advised, though, that disabling
168 CPUs in non-uniform fashion will result in non-uniform topology and
169 may lead to sub-optimal system performance with SCHED_ULE, which is
173 cpumask_t type is retired and cpuset_t is used in order to describe
177 Changes to ifconfig(8) for dynamic address family detection mandate
178 that you are running a kernel of 20110525 or later. Make sure to
179 follow the update procedure to boot a new kernel before installing
183 Support for sun4v architecture is officially dropped
186 Several KPI breaking changes have been committed to the mii(4) layer,
187 the PHY drivers and consequently some Ethernet drivers using mii(4).
188 This means that miibus.ko and the modules of the affected Ethernet
189 drivers need to be recompiled.
191 Note to kernel developers: Given that the OUI bit reversion problem
192 was fixed as part of these changes all mii(4) commits related to OUIs,
193 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
194 handling, no longer can be merged verbatim to stable/8 and previous
198 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
199 into their kernel configurations along with 'device pci'.
202 The default NFS client is now the new NFS client, so fstype "newnfs"
203 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
204 Although mounts via fstype "nfs" will usually work without userland
205 changes, it is recommended that the mount(8) and mount_nfs(8)
206 commands be rebuilt from sources and that a link to mount_nfs called
207 mount_oldnfs be created. The new client is compiled into the
208 kernel with "options NFSCL" and this is needed for diskless root
209 file systems. The GENERIC kernel configs have been changed to use
210 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
211 To use the regular/old client, you can "mount -t oldnfs ...". For
212 a diskless root file system, you must also include a line like:
214 vfs.root.mountfrom="oldnfs:"
216 in the boot/loader.conf on the root fs on the NFS server to make
217 a diskless root fs use the old client.
220 The GENERIC kernels for all architectures now default to the new
221 CAM-based ATA stack. It means that all legacy ATA drivers were
222 removed and replaced by respective CAM drivers. If you are using
223 ATA device names in /etc/fstab or other places, make sure to update
224 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
225 where 'Y's are the sequential numbers starting from zero for each type
226 in order of detection, unless configured otherwise with tunables,
227 see cam(4)). There will be symbolic links created in /dev/ to map
228 old adX devices to the respective adaY. They should provide basic
229 compatibility for file systems mounting in most cases, but they do
230 not support old user-level APIs and do not have respective providers
231 in GEOM. Consider using updated management tools with new device names.
233 It is possible to load devices ahci, ata, siis and mvs as modules,
234 but option ATA_CAM should remain in kernel configuration to make ata
235 module work as CAM driver supporting legacy ATA controllers. Device ata
236 still can be used in modular fashion (atacore + ...). Modules atadisk
237 and atapi* are not used and won't affect operation in ATA_CAM mode.
238 Note that to use CAM-based ATA kernel should include CAM devices
239 scbus, pass, da (or explicitly ada), cd and optionally others. All of
240 them are parts of the cam module.
242 ataraid(4) functionality is now supported by the RAID GEOM class.
243 To use it you can load geom_raid kernel module and use graid(8) tool
244 for management. Instead of /dev/arX device names, use /dev/raid/rX.
246 No kernel config options or code have been removed, so if a problem
247 arises, please report it and optionally revert to the old ATA stack.
248 In order to do it you can remove from the kernel config:
253 , and instead add back:
254 device atadisk # ATA disk drives
255 device ataraid # ATA RAID drives
256 device atapicd # ATAPI CDROM drives
257 device atapifd # ATAPI floppy drives
258 device atapist # ATAPI tape drives
261 The default NFS server has been changed to the new server, which
262 was referred to as the experimental server. If you need to switch
263 back to the old NFS server, you must now put the "-o" option on
264 both the mountd and nfsd commands. This can be done using the
265 mountd_flags and nfs_server_flags rc.conf variables until an
266 update to the rc scripts is committed, which is coming soon.
269 The GNU Objective-C runtime library (libobjc), and other Objective-C
270 related components have been removed from the base system. If you
271 require an Objective-C library, please use one of the available ports.
274 ath(4) has been split into bus- and device- modules. if_ath contains
275 the HAL, the TX rate control and the network device code. if_ath_pci
276 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
277 contains the AHB glue. Users need to load both if_ath_pci and if_ath
278 in order to use ath on everything else.
280 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
281 need to load if_ath and if_ath_pci for ath(4) operation.
284 As part of the replacement of sysinstall, the process of building
285 release media has changed significantly. For details, please re-read
286 release(7), which has been updated to reflect the new build process.
289 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
290 is the last available version under GPLv2. It brings a number of new
291 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
292 4.1 and SSE 4.2), better support for powerpc64, a number of new
293 directives, and lots of other small improvements. See the ChangeLog
294 file in contrib/binutils for the full details.
297 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
298 compliant, and will now use half of hash for authentication.
299 This will break interoperability with all stacks (including all
300 actual FreeBSD versions) who implement
301 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
303 The only workaround with such peers is to use another HMAC
304 algorithm for IPsec ("phase 2") authentication.
307 Remove the uio_yield prototype and symbol. This function has
308 been misnamed since it was introduced and should not be
309 globally exposed with this name. The equivalent functionality
310 is now available using kern_yield(curthread->td_user_pri).
311 The function remains undocumented.
314 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
315 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
316 for scalar sysctls is defined but disabled. Code that needs
317 UQUAD to pass the type checking that must compile on older
318 systems where the define is not present can check against
319 __FreeBSD_version >= 900030.
321 The system dialog(1) has been replaced with a new version previously
322 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
323 with the previous version, but the libdialog associated with it has
324 a largely incompatible API. As such, the original version of libdialog
325 will be kept temporarily as libodialog, until its base system consumers
326 are replaced or updated. Bump __FreeBSD_version to 900030.
329 If you are trying to run make universe on a -stable system, and you get
330 the following warning:
331 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
332 unknown. config(8) likely too old."
333 or something similar to it, then you must upgrade your -stable system
334 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
335 stable/8) or build the config from the latest stable/8 branch and
336 install it on your system.
338 Prior to this date, building a current universe on 8-stable system from
339 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
340 error in the first kernel build phase. A new config on those old
341 systems will fix that problem for older versions of -current.
344 The TCP stack has been modified to allow Khelp modules to interact with
345 it via helper hook points and store per-connection data in the TCP
346 control block. Bump __FreeBSD_version to 900029. User space tools that
347 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
351 Generic IEEE 802.3 annex 31B full duplex flow control support has been
352 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
353 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
354 to take advantage of it instead of using custom implementations. This
355 means that these drivers now no longer unconditionally advertise
356 support for flow control but only do so if flow control is a selected
357 media option. This was implemented in the generic support that way in
358 order to allow flow control to be switched on and off via ifconfig(8)
359 with the PHY specific default to typically off in order to protect
360 from unwanted effects. Consequently, if you used flow control with
361 one of the above mentioned drivers you now need to explicitly enable
363 ifconfig bge0 media auto mediaopt flowcontrol
365 Along with the above mentioned changes generic support for setting
366 1000baseT master mode also has been added and brgphy(4), ciphy(4),
367 e1000phy(4) as well as ip1000phy(4) have been converted to take
368 advantage of it. This means that these drivers now no longer take the
369 link0 parameter for selecting master mode but the master media option
370 has to be used instead, for example like in the following:
371 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
373 Selection of master mode now is also available with all other PHY
374 drivers supporting 1000baseT.
377 The TCP stack has received a significant update to add support for
378 modularised congestion control and generally improve the clarity of
379 congestion control decisions. Bump __FreeBSD_version to 900025. User
380 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
381 sockstat) need to be recompiled.
384 The man(1) utility has been replaced by a new version that no longer
385 uses /etc/manpath.config. Please consult man.conf(5) for how to
386 migrate local entries to the new format.
389 The copyright strings printed by login(1) and sshd(8) at the time of a
390 new connection have been removed to follow other operating systems and
394 A workaround for a fixed ld bug has been removed in kernel code,
395 so make sure that your system ld is built from sources after
396 revision 210245 from 2010-07-19 (r211583 if building head kernel
397 on stable/8, r211584 for stable/7; both from 2010-08-21).
398 A symptom of incorrect ld version is different addresses for
399 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
402 The $ipv6_prefer variable in rc.conf(5) has been split into
403 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
405 The $ip6addrctl_policy is a variable to choose a pre-defined
406 address selection policy set by ip6addrctl(8). A value
407 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
410 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
411 flag (see an entry of 20090926) is set on an interface with no
412 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
413 security reason. If you want IPv6 link-local address on all
414 interfaces by default, set this to "YES".
416 The old ipv6_prefer="YES" is equivalent to
417 ipv6_activate_all_interfaces="YES" and
418 ip6addrctl_policy="ipv6_prefer".
421 DTrace has grown support for userland tracing. Due to this, DTrace is
422 now i386 and amd64 only.
423 dtruss(1) is now installed by default on those systems and a new
424 kernel module is needed for userland tracing: fasttrap.
425 No changes to your kernel config file are necessary to enable
426 userland tracing, but you might consider adding 'STRIP=' and
427 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
428 to have informative userland stack traces in DTrace (ustack).
431 The acpi_aiboost(4) driver has been removed in favor of the new
432 aibs(4) driver. You should update your kernel configuration file.
435 BSD grep has been imported to the base system and it is built by
436 default. It is completely BSD licensed, highly GNU-compatible, uses
437 less memory than its GNU counterpart and has a small codebase.
438 However, it is slower than its GNU counterpart, which is mostly
439 noticeable for larger searches, for smaller ones it is measurable
440 but not significant. The reason is complex, the most important factor
441 is that we lack a modern and efficient regex library and GNU
442 overcomes this by optimizing the searches internally. Future work
443 on improving the regex performance is planned, for the meantime,
444 users that need better performance, can build GNU grep instead by
445 setting the WITH_GNU_GREP knob.
448 Due to the import of powerpc64 support, all existing powerpc kernel
449 configuration files must be updated with a machine directive like this:
450 machine powerpc powerpc
452 In addition, an updated config(8) is required to build powerpc kernels
456 A new version of ZFS (version 15) has been merged to -HEAD.
457 This version uses a python library for the following subcommands:
458 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
459 For full functionality of these commands the following port must
460 be installed: sysutils/py-zfs
463 'vm_page's are now hashed by physical address to an array of mutexes.
464 Currently this is only used to serialize access to hold_count. Over
465 time the page queue mutex will be peeled away. This changes the size
466 of pmap on every architecture. And requires all callers of vm_page_hold
467 and vm_page_unhold to be updated.
470 WITH_CTF can now be specified in src.conf (not recommended, there
471 are some problems with static executables), make.conf (would also
472 affect ports which do not use GNU make and do not override the
473 compile targets) or in the kernel config (via "makeoptions
475 When WITH_CTF was specified there before this was silently ignored,
476 so make sure that WITH_CTF is not used in places which could lead
477 to unwanted behavior.
480 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
481 to allow 32-bit compatibility on non-x86 platforms. All kernel
482 configurations on amd64 and ia64 platforms using these options must
483 be modified accordingly.
486 The utmp user accounting database has been replaced with utmpx,
487 the user accounting interface standardized by POSIX.
488 Unfortunately the semantics of utmp and utmpx don't match,
489 making it practically impossible to support both interfaces.
490 The user accounting database is used by tools like finger(1),
491 last(1), talk(1), w(1) and ac(8).
493 All applications in the base system use utmpx. This means only
494 local binaries (e.g. from the ports tree) may still use these
495 utmp database files. These applications must be rebuilt to make
498 After the system has been upgraded, it is safe to remove the old
499 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
500 assuming their contents is of no importance anymore. Old wtmp
501 databases can only be used by last(1) and ac(8) after they have
502 been converted to the new format using wtmpcvt(1).
505 Introduce the kernel thread "deadlock resolver" (which can be enabled
506 via the DEADLKRES option, see NOTES for more details) and the
507 sleepq_type() function for sleepqueues.
510 The rc.firewall and rc.firewall6 were unified, and
511 rc.firewall6 and rc.d/ip6fw were removed.
512 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
513 variables are obsoleted. Instead, the following new rc
514 variables are added to rc.d/ipfw:
516 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
517 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
518 firewall_simple_onet_ipv6, firewall_trusted_ipv6
520 The meanings correspond to the relevant IPv4 variables.
526 The default terminal emulation for syscons(4) has been changed
527 from cons25 to xterm on all platforms except pc98. This means
528 that the /etc/ttys file needs to be updated to ensure correct
529 operation of applications on the console.
531 The terminal emulation style can be toggled per window by using
532 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
533 options can be used to change the compile-time default back to
536 To prevent graphical artifacts, make sure the TERM environment
537 variable is set to match the terminal emulation that is being
538 performed by syscons(4).
541 The layout of the structure ieee80211req_scan_result has changed.
542 Applications that require wireless scan results (e.g. ifconfig(8))
543 from net80211 need to be recompiled.
545 Applications such as wpa_supplicant(8) may require a full world
546 build without using NO_CLEAN in order to get synchronized with the
550 The iwn(4) driver has been updated to support the 5000 and 5150 series.
551 There's one kernel module for each firmware. Adding "device iwnfw"
552 to the kernel configuration file means including all three firmware
553 images inside the kernel. If you want to include just the one for
554 your wireless card, use the devices iwn4965fw, iwn5000fw or
558 The rc.d/network_ipv6, IPv6 configuration script has been integrated
559 into rc.d/netif. The changes are the following:
561 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
562 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
563 Note that both variables need the "inet6" keyword at the head.
565 Do not set $ipv6_network_interfaces manually if you do not
566 understand what you are doing. It is not needed in most cases.
568 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
571 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
572 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
574 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
575 all configured interfaces have "inet6 accept_rtadv" in the
576 $ifconfig_IF_ipv6. These are for backward compatibility.
578 3. A new variable $ipv6_prefer has been added. If NO, IPv6
579 functionality of interfaces with no corresponding
580 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
581 and the default address selection policy of ip6addrctl(8)
582 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
583 Note that if you want to configure IPv6 functionality on the
584 disabled interfaces after boot, first you need to clear the flag by
585 using ifconfig(8) like:
587 ifconfig em0 inet6 -ifdisabled
589 If YES, the default address selection policy is set as
592 The default value of $ipv6_prefer is NO.
594 4. If your system need to receive Router Advertisement messages,
595 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
596 scripts automatically invoke rtsol(8) when the interface becomes
597 UP. The Router Advertisement messages are used for SLAAC
598 (State-Less Address AutoConfiguration).
601 802.11s D3.03 support was committed. This is incompatible with the
602 previous code, which was based on D3.0.
605 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
606 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
607 control whether accepting Router Advertisement messages or not.
608 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
609 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
610 The ifconfig(8) utility now supports these flags.
613 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
614 mount(8) and -a option for df(1) to see them.
617 The old tunable hw.bus.devctl_disable has been superseded by
618 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
619 replaced by hw.bus.devctl_queue=0. The default for this new tunable
623 Remove the option STOP_NMI. The default action is now to use NMI only
624 for KDB via the newly introduced function stop_cpus_hard() and
625 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
628 The stable/8 branch created in subversion. This corresponds to the
629 RELENG_8 branch in CVS.
632 Bump the shared library version numbers for all libraries that do not
633 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
634 __FreeBSD_version to 800105.
637 Due to changes in the implementation of virtual network stack support,
638 all network-related kernel modules must be recompiled. As this change
639 breaks the ABI, bump __FreeBSD_version to 800104.
642 The TOE interface to the TCP syncache has been modified to remove
643 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
644 The cxgb driver is the only TOE consumer affected by this change, and
645 needs to be recompiled along with the kernel. As this change breaks
646 the ABI, bump __FreeBSD_version to 800103.
649 Padding has been added to struct tcpcb, sackhint and tcpstat in
650 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
651 maintaining the ABI. However, this change breaks the ABI, so bump
652 __FreeBSD_version to 800102. User space tools that rely on the size of
653 any of these structs (e.g. sockstat) need to be recompiled.
656 The NFS_LEGACYRPC option has been removed along with the old kernel
657 RPC implementation that this option selected. Kernel configurations
658 may need to be adjusted.
661 The network interface device nodes at /dev/net/<interface> have been
662 removed. All ioctl operations can be performed the normal way using
663 routing sockets. The kqueue functionality can generally be replaced
664 with routing sockets.
667 The documentation from the FreeBSD Documentation Project (Handbook,
668 FAQ, etc.) is now installed via packages by sysinstall(8) and under
669 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
672 The ABI of various structures related to the SYSV IPC API have been
673 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
674 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
678 Layout of struct vnet has changed as routing related variables were
679 moved to their own Vimage module. Modules need to be recompiled. Bump
680 __FreeBSD_version to 800099.
683 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
684 respectively. As long as no more than 16 groups per process are used,
685 no changes should be visible. When more than 16 groups are used, old
686 binaries may fail if they call getgroups() or getgrouplist() with
687 statically sized storage. Recompiling will work around this, but
688 applications should be modified to use dynamically allocated storage
689 for group arrays as POSIX.1-2008 does not cap an implementation's
690 number of supported groups at NGROUPS_MAX+1 as previous versions did.
692 NFS and portalfs mounts may also be affected as the list of groups is
693 truncated to 16. Users of NFS who use more than 16 groups, should
694 take care that negative group permissions are not used on the exported
695 file systems as they will not be reliable unless a GSSAPI based
696 authentication method is used.
699 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
700 option compiles in the support for adaptive spinning for lockmgrs
701 which want to enable it. The lockinit() function now accepts the flag
702 LK_ADAPTIVE in order to make the lock object subject to adaptive
703 spinning when both held in write and read mode.
706 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
707 changed. User applications that use this ioctl need to be rebuilt.
710 The layout of struct thread has changed. Kernel and modules need to
714 The layout of structs ifnet, domain, protosw and vnet_net has changed.
715 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
718 window(1) has been removed from the base system. It can now be
719 installed from ports. The port is called misc/window.
722 The way we are storing and accessing `routing table' entries has
723 changed. Programs reading the FIB, like netstat, need to be
727 A new netisr implementation has been added for FreeBSD 8. Network
728 file system modules, such as igmp, ipdivert, and others, should be
730 Bump __FreeBSD_version to 800096.
733 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
737 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
738 Bump __FreeBSD_version to 800094.
741 Add mnt_xflag field to 'struct mount'. File system modules need to be
743 Bump __FreeBSD_version to 800093.
746 The compiling option ADAPTIVE_SX has been retired while it has been
747 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
748 The KPI for sx_init_flags() changes as accepting flags:
749 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
750 been introduced in order to handle the reversed logic.
751 Bump __FreeBSD_version to 800092.
754 Add support for hierarchical jails. Remove global securelevel.
755 Bump __FreeBSD_version to 800091.
758 The layout of struct vnet_net has changed, therefore modules
760 Bump __FreeBSD_version to 800090.
763 The newly imported zic(8) produces a new format in the output. Please
764 run tzsetup(8) to install the newly created data to /etc/localtime.
767 The sysctl tree for the usb stack has renamed from hw.usb2.* to
768 hw.usb.* and is now consistent again with previous releases.
771 802.11 monitor mode support was revised and driver api's were changed.
772 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
773 of DLT_IEEE802_11. No user-visible data structures were changed but
774 applications that use DLT_IEEE802_11 may require changes.
775 Bump __FreeBSD_version to 800088.
778 The layout of the following structs has changed: sysctl_oid,
779 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
780 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
781 panics may be experienced. World rebuild is required for
782 correctly checking networking state from userland.
783 Bump __FreeBSD_version to 800085.
786 MLDv2 and Source-Specific Multicast (SSM) have been merged
787 to the IPv6 stack. VIMAGE hooks are in but not yet used.
788 The implementation of SSM within FreeBSD's IPv6 stack closely
789 follows the IPv4 implementation.
791 For kernel developers:
793 * The most important changes are that the ip6_output() and
794 ip6_input() paths no longer take the IN6_MULTI_LOCK,
795 and this lock has been downgraded to a non-recursive mutex.
797 * As with the changes to the IPv4 stack to support SSM, filtering
798 of inbound multicast traffic must now be performed by transport
799 protocols within the IPv6 stack. This does not apply to TCP and
800 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
802 * The KPIs used by IPv6 multicast are similar to those used by
803 the IPv4 stack, with the following differences:
804 * im6o_mc_filter() is analogous to imo_multicast_filter().
805 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
806 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
807 * IN6_LOOKUP_MULTI() has been deprecated and removed.
808 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
809 for MLDv1 have an additional 'timer' argument which is used to
810 jitter the initial membership report for the solicited-node
811 multicast membership on-link.
812 * This is not strictly needed for MLDv2, which already jitters
813 its report transmissions. However, the 'timer' argument is
814 preserved in case MLDv1 is active on the interface.
816 * The KAME linked-list based IPv6 membership implementation has
817 been refactored to use a vector similar to that used by the IPv4
819 Code which maintains a list of its own multicast memberships
820 internally, e.g. carp, has been updated to reflect the new
823 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
824 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
825 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
826 implementation constraint which needs to be addressed in HEAD.
828 For application developers:
830 * The changes are broadly similar to those made for the IPv4
833 * The use of IPv4 and IPv6 multicast socket options on the same
834 socket, using mapped addresses, HAS NOT been tested or supported.
836 * There are a number of issues with the implementation of various
837 IPv6 multicast APIs which need to be resolved in the API surface
838 before the implementation is fully compatible with KAME userland
839 use, and these are mostly to do with interface index treatment.
841 * The literature available discusses the use of either the delta / ASM
842 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
843 using setsourcefilter(3)/getsourcefilter(3). For more information
844 please refer to RFC 3768, 'Socket Interface Extensions for
845 Multicast Source Filters'.
847 * Applications which use the published RFC 3678 APIs should be fine.
849 For systems administrators:
851 * The mtest(8) utility has been refactored to support IPv6, in
852 addition to IPv4. Interface addresses are no longer accepted
853 as arguments, their names must be used instead. The utility
854 will map the interface name to its first IPv4 address as
855 returned by getifaddrs(3).
857 * The ifmcstat(8) utility has also been updated to print the MLDv2
858 endpoint state and source filter lists via sysctl(3).
860 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
861 loopback of IPv6 multicast datagrams by default; it defaults to 1
862 to preserve the existing behaviour. Disabling multicast loopback is
863 recommended for optimal system performance.
865 * The IPv6 MROUTING code has been changed to examine this sysctl
866 instead of attempting to perform a group lookup before looping
867 back forwarded datagrams.
869 Bump __FreeBSD_version to 800084.
872 Implement low-level Bluetooth HCI API.
873 Bump __FreeBSD_version to 800083.
876 The layout of struct malloc_type, used by modules to register new
877 memory allocation types, has changed. Most modules will need to
878 be rebuilt or panics may be experienced.
879 Bump __FreeBSD_version to 800081.
882 Anticipate overflowing inp_flags - add inp_flags2.
883 This changes most offsets in inpcb, so checking v4 connection
884 state will require a world rebuild.
885 Bump __FreeBSD_version to 800080.
888 Add an llentry to struct route and struct route_in6. Modules
889 embedding a struct route will need to be recompiled.
890 Bump __FreeBSD_version to 800079.
893 The size of rt_metrics_lite and by extension rtentry has changed.
894 Networking administration apps will need to be recompiled.
895 The route command now supports show as an alias for get, weighting
896 of routes, sticky and nostick flags to alter the behavior of stateful
898 Bump __FreeBSD_version to 800078.
901 Do not use Giant for kbdmux(4) locking. This is wrong and
902 apparently causing more problems than it solves. This will
903 re-open the issue where interrupt handlers may race with
904 kbdmux(4) in polling mode. Typical symptoms include (but
905 not limited to) duplicated and/or missing characters when
906 low level console functions (such as gets) are used while
907 interrupts are enabled (for example geli password prompt,
908 mountroot prompt etc.). Disabling kbdmux(4) may help.
911 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
912 kernel modules referencing any of the above need to be recompiled.
913 Bump __FreeBSD_version to 800075.
916 GEOM_PART has become the default partition slicer for storage devices,
917 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
918 introduces some changes:
920 MSDOS/EBR: the devices created from MSDOS extended partition entries
921 (EBR) can be named differently than with GEOM_MBR and are now symlinks
922 to devices with offset-based names. fstabs may need to be modified.
924 BSD: the "geometry does not match label" warning is harmless in most
925 cases but it points to problems in file system misalignment with
926 disk geometry. The "c" partition is now implicit, covers the whole
927 top-level drive and cannot be (mis)used by users.
929 General: Kernel dumps are now not allowed to be written to devices
930 whose partition types indicate they are meant to be used for file
931 systems (or, in case of MSDOS partitions, as something else than
934 Most of these changes date approximately from 200812.
937 The uscanner(4) driver has been removed from the kernel. This follows
938 Linux removing theirs in 2.6 and making libusb the default interface
942 The multicast forwarding code has been cleaned up. netstat(1)
943 only relies on KVM now for printing bandwidth upcall meters.
944 The IPv4 and IPv6 modules are split into ip_mroute_mod and
945 ip6_mroute_mod respectively. The config(5) options for statically
946 compiling this code remain the same, i.e. 'options MROUTING'.
949 Support for the IFF_NEEDSGIANT network interface flag has been
950 removed, which means that non-MPSAFE network device drivers are no
951 longer supported. In particular, if_ar, if_sr, and network device
952 drivers from the old (legacy) USB stack can no longer be built or
956 POSIX.1 Native Language Support (NLS) has been enabled in libc and
957 a bunch of new language catalog files have also been added.
958 This means that some common libc messages are now localized and
959 they depend on the LC_MESSAGES environmental variable.
962 The k8temp(4) driver has been renamed to amdtemp(4) since
963 support for Family 10 and Family 11 CPU families was added.
966 IGMPv3 and Source-Specific Multicast (SSM) have been merged
967 to the IPv4 stack. VIMAGE hooks are in but not yet used.
969 For kernel developers, the most important changes are that the
970 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
971 and this lock has been downgraded to a non-recursive mutex.
973 Transport protocols (UDP, Raw IP) are now responsible for filtering
974 inbound multicast traffic according to group membership and source
975 filters. The imo_multicast_filter() KPI exists for this purpose.
976 Transports which do not use multicast (SCTP, TCP) already reject
977 multicast by default. Forwarding and receive performance may improve
978 as a mutex acquisition is no longer needed in the ip_input()
979 low-level input path. in_addmulti() and in_delmulti() are shimmed
980 to new KPIs which exist to support SSM in-kernel.
982 For application developers, it is recommended that loopback of
983 multicast datagrams be disabled for best performance, as this
984 will still cause the lock to be taken for each looped-back
985 datagram transmission. The net.inet.ip.mcast.loop sysctl may
986 be tuned to 0 to disable loopback by default; it defaults to 1
987 to preserve the existing behaviour.
989 For systems administrators, to obtain best performance with
990 multicast reception and multiple groups, it is always recommended
991 that a card with a suitably precise hash filter is used. Hash
992 collisions will still result in the lock being taken within the
993 transport protocol input path to check group membership.
995 If deploying FreeBSD in an environment with IGMP snooping switches,
996 it is recommended that the net.inet.igmp.sendlocal sysctl remain
997 enabled; this forces 224.0.0.0/24 group membership to be announced
1000 The size of 'struct igmpstat' has changed; netstat needs to be
1001 recompiled to reflect this.
1002 Bump __FreeBSD_version to 800070.
1005 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1006 updated to use it. This requires a buildworld/installworld in order to
1007 update the library and dependencies (usbconfig, etc). Its advisable to
1008 rebuild all ports which uses libusb. More specific directions are given
1009 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1010 libusb are no longer required and can be removed.
1013 A workaround is committed to allow the creation of System V shared
1014 memory segment of size > 2 GB on the 64-bit architectures.
1015 Due to a limitation of the existing ABI, the shm_segsz member
1016 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1017 wrong for large segments. Note that limits must be explicitly
1018 raised to allow such segments to be created.
1021 The layout of struct ifnet has changed, requiring a rebuild of all
1022 network device driver modules.
1025 The /dev handling for the new USB stack has changed, a
1026 buildworld/installworld is required for libusb20.
1029 The new USB2 stack has now been permanently moved in and all kernel and
1030 module names reverted to their previous values (eg, usb, ehci, ohci,
1031 ums, ...). The old usb stack can be compiled in by prefixing the name
1032 with the letter 'o', the old usb modules have been removed.
1033 Updating entry 20090216 for xorg and 20090215 for libmap may still
1037 The rc.conf(5) option if_up_delay has been renamed to
1038 defaultroute_delay to better reflect its purpose. If you have
1039 customized this setting in /etc/rc.conf you need to update it to
1043 xorg 7.4 wants to configure its input devices via hald which does not
1044 yet work with USB2. If the keyboard/mouse does not work in xorg then
1046 Option "AllowEmptyInput" "off"
1047 to your ServerLayout section. This will cause X to use the configured
1048 kbd and mouse sections from your xorg.conf.
1051 The GENERIC kernels for all architectures now default to the new USB2
1052 stack. No kernel config options or code have been removed so if a
1053 problem arises please report it and optionally revert to the old USB
1054 stack. If you are loading USB kernel modules or have a custom kernel
1055 that includes GENERIC then ensure that usb names are also changed over,
1056 eg uftdi -> usb2_serial_ftdi.
1058 Older programs linked against the ports libusb 0.1 need to be
1059 redirected to the new stack's libusb20. /etc/libmap.conf can
1061 # Map old usb library to new one for usb2 stack
1062 libusb-0.1.so.8 libusb20.so.1
1065 All USB ethernet devices now attach as interfaces under the name ueN
1066 (eg. ue0). This is to provide a predictable name as vendors often
1067 change usb chipsets in a product without notice.
1070 The ichsmb(4) driver has been changed to require SMBus slave
1071 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1072 All of the other SMBus controller drivers require left-justified
1073 slave addresses, so this change makes all the drivers provide the
1077 INET6 statistics (struct ip6stat) was updated.
1078 netstat(1) needs to be recompiled.
1081 NTFS has been removed from GENERIC kernel on amd64 to match
1082 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1083 will load ntfs.ko module automatically when NTFS support is
1084 actually needed, unless ntfs.ko is not installed or security
1085 level prohibits loading kernel modules. If either is the case,
1086 "options NTFS" has to be added into kernel config.
1089 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1090 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1091 800061. User space tools that rely on the size of struct tcpcb in
1092 tcp_var.h (e.g. sockstat) need to be recompiled.
1095 ng_tty(4) module updated to match the new TTY subsystem.
1096 Due to API change, user-level applications must be updated.
1097 New API support added to mpd5 CVS and expected to be present
1098 in next mpd5.3 release.
1101 With __FreeBSD_version 800060 the makefs tool is part of
1102 the base system (it was a port).
1105 The afdata and ifnet locks have been changed from mutexes to
1106 rwlocks, network modules will need to be re-compiled.
1109 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1110 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1111 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1112 architecture and 40 bytes on 64-bit architecture. The userland
1113 applications "arp" and "ndp" have been updated accordingly.
1114 The output from "netstat -r" shows only routing entries and
1115 none of the L2 information.
1118 __FreeBSD_version 800057 marks the switchover from the
1119 binary ath hal to source code. Users must add the line:
1121 options AH_SUPPORT_AR5416
1123 to their kernel config files when specifying:
1127 The ath_hal module no longer exists; the code is now compiled
1128 together with the driver in the ath module. It is now
1129 possible to tailor chip support (i.e. reduce the set of chips
1130 and thereby the code size); consult ath_hal(4) for details.
1133 __FreeBSD_version 800054 adds memory barriers to
1134 <machine/atomic.h>, new interfaces to ifnet to facilitate
1135 multiple hardware transmit queues for cards that support
1136 them, and a lock-less ring-buffer implementation to
1137 enable drivers to more efficiently manage queueing of
1141 A new version of ZFS (version 13) has been merged to -HEAD.
1142 This version has zpool attribute "listsnapshots" off by
1143 default, which means "zfs list" does not show snapshots,
1144 and is the same as Solaris behavior.
1147 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1150 The uhci, ohci, ehci and slhci USB Host controller drivers have
1151 been put into separate modules. If you load the usb module
1152 separately through loader.conf you will need to load the
1153 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1154 controller add the following to loader.conf:
1160 The ABI used by the PMC toolset has changed. Please keep
1161 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1165 atapci kernel module now includes only generic PCI ATA
1166 driver. AHCI driver moved to ataahci kernel module.
1167 All vendor-specific code moved into separate kernel modules:
1168 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1169 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1170 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1171 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1174 The TTY subsystem of the kernel has been replaced by a new
1175 implementation, which provides better scalability and an
1176 improved driver model. Most common drivers have been migrated to
1177 the new TTY subsystem, while others have not. The following
1178 drivers have not yet been ported to the new TTY layer:
1181 cy, digi, rc, rp, sio
1187 ng_h4, ng_tty, ppp, sl, snp
1189 Adding these drivers to your kernel configuration file shall
1190 cause compilation to fail.
1193 ntpd has been upgraded to 4.2.4p5.
1196 OpenSSH has been upgraded to 5.1p1.
1198 For many years, FreeBSD's version of OpenSSH preferred DSA
1199 over RSA for host and user authentication keys. With this
1200 upgrade, we've switched to the vendor's default of RSA over
1201 DSA. This may cause upgraded clients to warn about unknown
1202 host keys even for previously known hosts. Users should
1203 follow the usual procedure for verifying host keys before
1204 accepting the RSA key.
1206 This can be circumvented by setting the "HostKeyAlgorithms"
1207 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1210 Please note that the sequence of keys offered for
1211 authentication has been changed as well. You may want to
1212 specify IdentityFile in a different order to revert this
1216 The sio(4) driver has been removed from the i386 and amd64
1217 kernel configuration files. This means uart(4) is now the
1218 default serial port driver on those platforms as well.
1220 To prevent collisions with the sio(4) driver, the uart(4) driver
1221 uses different names for its device nodes. This means the
1222 onboard serial port will now most likely be called "ttyu0"
1223 instead of "ttyd0". You may need to reconfigure applications to
1224 use the new device names.
1226 When using the serial port as a boot console, be sure to update
1227 /boot/device.hints and /etc/ttys before booting the new kernel.
1228 If you forget to do so, you can still manually specify the hints
1229 at the loader prompt:
1231 set hint.uart.0.at="isa"
1232 set hint.uart.0.port="0x3F8"
1233 set hint.uart.0.flags="0x10"
1234 set hint.uart.0.irq="4"
1238 The gpt(8) utility has been removed. Use gpart(8) to partition
1242 The version that Linuxulator emulates was changed from 2.4.2
1243 to 2.6.16. If you experience any problems with Linux binaries
1244 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1245 if it fixes the problem contact emulation mailing list.
1248 ISDN4BSD (I4B) was removed from the src tree. You may need to
1249 update a your kernel configuration and remove relevant entries.
1252 I have checked in code to support multiple routing tables.
1253 See the man pages setfib(1) and setfib(2).
1254 This is a hopefully backwards compatible version,
1255 but to make use of it you need to compile your kernel
1256 with options ROUTETABLES=2 (or more up to 16).
1259 The 802.11 wireless support was redone to enable multi-bss
1260 operation on devices that are capable. The underlying device
1261 is no longer used directly but instead wlanX devices are
1262 cloned with ifconfig. This requires changes to rc.conf files.
1263 For example, change:
1264 ifconfig_ath0="WPA DHCP"
1267 ifconfig_wlan0="WPA DHCP"
1268 see rc.conf(5) for more details. In addition, mergemaster of
1269 /etc/rc.d is highly recommended. Simultaneous update of userland
1270 and kernel wouldn't hurt either.
1272 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1273 modules were merged into the base wlan module. All references
1274 to these modules (e.g. in kernel config files) must be removed.
1277 psm(4) has gained write(2) support in native operation level.
1278 Arbitrary commands can be written to /dev/psm%d and status can
1279 be read back from it. Therefore, an application is responsible
1280 for status validation and error recovery. It is a no-op in
1281 other operation levels.
1284 Support for KSE threading has been removed from the kernel. To
1285 run legacy applications linked against KSE libmap.conf may
1286 be used. The following libmap.conf may be used to ensure
1287 compatibility with any prior release:
1289 libpthread.so.1 libthr.so.1
1290 libpthread.so.2 libthr.so.2
1291 libkse.so.3 libthr.so.3
1294 The layout of struct vmspace has changed. This affects libkvm
1295 and any executables that link against libkvm and use the
1296 kvm_getprocs() function. In particular, but not exclusively,
1297 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1298 The effects are minimal, but it's advisable to upgrade world
1302 The latest em driver no longer has support in it for the
1303 82575 adapter, this is now moved to the igb driver. The
1304 split was done to make new features that are incompatible
1305 with older hardware easier to do.
1308 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1309 likewise the kernel option is now GEOM_LINUX_LVM.
1312 The default NFS mount mode has changed from UDP to TCP for
1313 increased reliability. If you rely on (insecurely) NFS
1314 mounting across a firewall you may need to update your
1318 Belatedly note the addition of m_collapse for compacting
1322 The fts(3) structures have been changed to use adequate
1323 integer types for their members and so to be able to cope
1324 with huge file trees. The old fts(3) ABI is preserved
1325 through symbol versioning in libc, so third-party binaries
1326 using fts(3) should still work, although they will not take
1327 advantage of the extended types. At the same time, some
1328 third-party software might fail to build after this change
1329 due to unportable assumptions made in its source code about
1330 fts(3) structure members. Such software should be fixed
1331 by its vendor or, in the worst case, in the ports tree.
1332 FreeBSD_version 800015 marks this change for the unlikely
1333 case that a portable fix is impossible.
1336 To upgrade to -current after this date, you must be running
1337 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1338 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1341 The ADAPTIVE_GIANT kernel option has been retired because its
1342 functionality is the default now.
1345 The AT keyboard emulation of sunkbd(4) has been turned on
1346 by default. In order to make the special symbols of the Sun
1347 keyboards driven by sunkbd(4) work under X these now have
1348 to be configured the same way as Sun USB keyboards driven
1349 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1351 Option "XkbLayout" "us"
1352 Option "XkbRules" "xorg"
1353 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1356 It has been decided that it is desirable to provide ABI
1357 backwards compatibility to the FreeBSD 4/5/6 versions of the
1358 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1359 broken with the introduction of PCI domain support (see the
1360 20070930 entry). Unfortunately, this required the ABI of
1361 PCIOCGETCONF to be broken again in order to be able to
1362 provide backwards compatibility to the old version of that
1363 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1364 again. As for prominent ports this affects neither pciutils
1365 nor xorg-server this time, the hal port needs to be rebuilt
1369 The misnamed kthread_create() and friends have been renamed
1370 to kproc_create() etc. Many of the callers already
1371 used kproc_start()..
1372 I will return kthread_create() and friends in a while
1373 with implementations that actually create threads, not procs.
1374 Renaming corresponds with version 800002.
1383 Avoid using make -j when upgrading. While generally safe, there are
1384 sometimes problems using -j to upgrade. If your upgrade fails with
1385 -j, please try again without -j. From time to time in the past there
1386 have been problems using -j with buildworld and/or installworld. This
1387 is especially true when upgrading between "distant" versions (eg one
1388 that cross a major release boundary or several minor releases, or when
1389 several months have passed on the -current branch).
1391 Sometimes, obscure build problems are the result of environment
1392 poisoning. This can happen because the make utility reads its
1393 environment when searching for values for global variables. To run
1394 your build attempts in an "environmental clean room", prefix all make
1395 commands with 'env -i '. See the env(1) manual page for more details.
1397 When upgrading from one major version to another it is generally best
1398 to upgrade to the latest code in the currently installed branch first,
1399 then do an upgrade to the new branch. This is the best-tested upgrade
1400 path, and has the highest probability of being successful. Please try
1401 this approach before reporting problems with a major version upgrade.
1405 When upgrading the boot ZFS pool to a new version, always follow
1408 1.) recompile and reinstall the ZFS boot loader and boot block
1409 (this is part of "make buildworld" and "make installworld")
1411 2.) update the ZFS boot block on your boot drive
1413 The following example updates the ZFS boot block on the first
1414 partition (freebsd-boot) of a GPT partitioned drive ad0:
1415 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1417 Non-boot pools do not need these updates.
1421 If you are updating from a prior version of FreeBSD (even one just
1422 a few days old), you should follow this procedure. It is the most
1423 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1425 make kernel-toolchain
1426 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1427 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1429 To test a kernel once
1430 ---------------------
1431 If you just want to boot a kernel once (because you are not sure
1432 if it works, or if you want to boot a known bad kernel to provide
1433 debugging information) run
1434 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1435 nextboot -k testkernel
1437 To just build a kernel when you know that it won't mess you up
1438 --------------------------------------------------------------
1439 This assumes you are already running a CURRENT system. Replace
1440 ${arch} with the architecture of your machine (e.g. "i386",
1441 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1443 cd src/sys/${arch}/conf
1444 config KERNEL_NAME_HERE
1445 cd ../compile/KERNEL_NAME_HERE
1450 If this fails, go to the "To build a kernel" section.
1452 To rebuild everything and install it on the current system.
1453 -----------------------------------------------------------
1454 # Note: sometimes if you are running current you gotta do more than
1455 # is listed here if you are upgrading from a really old current.
1457 <make sure you have good level 0 dumps>
1459 make kernel KERNCONF=YOUR_KERNEL_HERE
1461 <reboot in single user> [3]
1469 To cross-install current onto a separate partition
1470 --------------------------------------------------
1471 # In this approach we use a separate partition to hold
1472 # current's root, 'usr', and 'var' directories. A partition
1473 # holding "/", "/usr" and "/var" should be about 2GB in
1476 <make sure you have good level 0 dumps>
1479 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1480 <maybe newfs current's root partition>
1481 <mount current's root partition on directory ${CURRENT_ROOT}>
1482 make installworld DESTDIR=${CURRENT_ROOT}
1483 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1484 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1485 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1486 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1487 <reboot into current>
1488 <do a "native" rebuild/install as described in the previous section>
1489 <maybe install compatibility libraries from ports/misc/compat*>
1493 To upgrade in-place from 8.x-stable to current
1494 ----------------------------------------------
1495 <make sure you have good level 0 dumps>
1497 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1499 <reboot in single user> [3]
1506 Make sure that you've read the UPDATING file to understand the
1507 tweaks to various things you need. At this point in the life
1508 cycle of current, things change often and you are on your own
1509 to cope. The defaults can also change, so please read ALL of
1510 the UPDATING entries.
1512 Also, if you are tracking -current, you must be subscribed to
1513 freebsd-current@freebsd.org. Make sure that before you update
1514 your sources that you have read and understood all the recent
1515 messages there. If in doubt, please track -stable which has
1516 much fewer pitfalls.
1518 [1] If you have third party modules, such as vmware, you
1519 should disable them at this point so they don't crash your
1522 [3] From the bootblocks, boot -s, and then do
1527 adjkerntz -i # if CMOS is wall time
1528 Also, when doing a major release upgrade, it is required that
1529 you boot into single user mode to do the installworld.
1531 [4] Note: This step is non-optional. Failure to do this step
1532 can result in a significant reduction in the functionality of the
1533 system. Attempting to do it by hand is not recommended and those
1534 that pursue this avenue should read this file carefully, as well
1535 as the archives of freebsd-current and freebsd-hackers mailing lists
1536 for potential gotchas. The -U option is also useful to consider.
1537 See mergemaster(8) for more information.
1539 [5] Usually this step is a noop. However, from time to time
1540 you may need to do this if you get unknown user in the following
1541 step. It never hurts to do it all the time. You may need to
1542 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1543 install) after the buildworld before this step if you last updated
1544 from current before 20020224 or from -stable before 20020408.
1546 [6] This only deletes old files and directories. Old libraries
1547 can be deleted by "make delete-old-libs", but you have to make
1548 sure that no program is using those libraries anymore.
1550 [8] In order to have a kernel that can run the 4.x binaries needed to
1551 do an installworld, you must include the COMPAT_FREEBSD4 option in
1552 your kernel. Failure to do so may leave you with a system that is
1553 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1554 required to run the 5.x binaries on more recent kernels. And so on
1555 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1557 Make sure that you merge any new devices from GENERIC since the
1558 last time you updated your kernel config file.
1560 [9] When checking out sources, you must include the -P flag to have
1561 cvs prune empty directories.
1563 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1564 "?=" instead of the "=" assignment operator, so that buildworld can
1565 override the CPUTYPE if it needs to.
1567 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1568 not on the command line, or in /etc/make.conf. buildworld will
1569 warn if it is improperly defined.
1572 This file contains a list, in reverse chronological order, of major
1573 breakages in tracking -current. It is not guaranteed to be a complete
1574 list of such breakages, and only contains entries since October 10, 2007.
1575 If you need to see UPDATING entries from before that date, you will need
1576 to fetch an UPDATING file from an older FreeBSD release.
1578 Copyright information:
1580 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1582 Redistribution, publication, translation and use, with or without
1583 modification, in full or in part, in any form or format of this
1584 document are permitted without further permission from the author.
1586 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1587 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1588 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1589 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1590 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1591 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1592 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1593 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1594 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1595 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1596 POSSIBILITY OF SUCH DAMAGE.
1598 Contact Warner Losh if you have any questions about your use of