1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/makeworld.html
11 Items affecting the ports and packages system can be found in
12 /usr/ports/UPDATING. Please read that file before running portupgrade.
15 The mergemaster command now uses the default MAKEOBJDIRPREFIX
16 rather than creating it's own in the temporary directory in
17 order allow access to bootstrapped versions of tools such as
18 install and mtree. When upgrading from version of FreeBSD where
19 the install command does not support -l, you will need to
20 install a new mergemaster command if mergemaster -p is required.
21 This can be accomplished with the command (cd src/usr.sbin/mergemaster
25 Fix a bug that allows NFS clients to issue READDIR on files.
28 The install(1) option -M has changed meaning and now takes an
29 argument that is a file or path to append logs to. In the
30 unlikely event that -M was the last option on the command line
31 and the command line contained at least two files and a target
32 directory the first file will have logs appended to it. The -M
33 option served little practical purpose in the last decade so it's
34 used expected to be extremely rare.
37 A new compression method (lz4) has been merged to. Please refer to
38 zpool-features(7) for more information.
40 Please refer to the "ZFS notes" section of this file for information
41 on upgrading boot ZFS pools.
44 The VFS KBI was changed with the merge of several nullfs
45 optimizations and fixes. All filesystem modules must be
49 With the addition of auditdistd(8), a new auditdistd user is now
50 depended on during installworld. "mergemaster -p" can be used to add
51 the user prior to installworld, as documented in the handbook.
57 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
58 Starting with this version the old system of ZFS pool versioning
59 is superseded by "feature flags". This concept enables forward
60 compatibility against certain future changes in functionality of ZFS
61 pools. The first two read-only compatible "feature flags" for ZFS
62 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
63 For more information read the new zpool-features(7) manual page.
64 Please refer to the "ZFS notes" section of this file for information
65 on upgrading boot ZFS pools.
68 The commit introducing bsd.compiler.mk breaks the traditional
69 building of kernels before this point. Add -m ${SRC}/share/mk
70 (for the right value of SRC) to your command lines to work
71 around; update your useland to a point after this; or use the
72 buildkernel/installkernel top-level targets. See also 20120829.
75 The IPFIREWALL_FORWARD kernel option has been removed. Its
76 functionality now turned on by default.
79 The random(4) support for the VIA hardware random number
80 generator (`PADLOCK') is no longer enabled unconditionally.
81 Add the PADLOCK_RNG option in the custom kernel config if
82 needed. The GENERIC kernels on i386 and amd64 do include the
83 option, so the change only affects the custom kernel
87 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
88 the traditional method, you must update your system with an installworld
89 before the kernel will build. The documented make buildkernel/installkernel
90 interfaces (coupled with fresh make kernel-toolchain) continue to work.
93 The sparc64 ZFS loader has been changed to no longer try to auto-
94 detect ZFS providers based on diskN aliases but now requires these
95 to be explicitly listed in the OFW boot-device environment variable.
98 Now unix domain sockets behave "as expected" on nullfs(5). Previously
99 nullfs(5) did not pass through all behaviours to the underlying layer,
100 as a result if we bound to a socket on the lower layer we could connect
101 only to the lower path; if we bound to the upper layer we could connect
102 only to the upper path. The new behavior is one can connect to both the
103 lower and the upper paths regardless what layer path one binds to.
106 The acpi_wmi(4) status device /dev/wmistat has been renamed to
110 A new VOP_ADVISE() was added to support posix_fadvise(2). All
111 filesystem modules must be recompiled.
114 The interface of the VOP_VPTOCNP(9) changed, now the returned
115 vnode shall be referenced, previously it was required to be
116 only held. All in-tree filesystems are converted.
122 The broken amd(4) driver has been replaced with esp(4) in the amd64,
123 i386 and pc98 GENERIC kernel configuration files.
126 This commit modifies vfs_register() so that it uses a hash
127 calculation to set vfc_typenum, which is enabled by default.
128 The first time a system is booted after this change, the
129 vfc_typenum values will change for all file systems. The
130 main effect of this is a change to the NFS server file handles
131 for file systems that use vfc_typenum in their fsid, such as ZFS.
132 It will, however, prevent vfc_typenum from changing when file
133 systems are loaded in a different order for subsequent reboots.
134 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
135 until you are ready to remount all NFS clients after a reboot.
138 Bump the shared library version numbers for libraries that
139 do not use symbol versioning, have changed the ABI compared
140 to stable/8 and which shared library version was not bumped.
141 Done as part of 9.0-RELEASE cycle.
144 During the merge of Capsicum features, the fget(9) KPI was modified.
145 This may require the rebuilding of out-of-tree device drivers --
146 issues have been reported specifically with the nVidia device driver.
147 __FreeBSD_version is bumped to 900041.
149 Also, there is a period between 20110811 and 20110814 where the
150 special devices /dev/{stdin,stdout,stderr} did not work correctly.
151 Building world from a kernel during that window may not work.
154 The packet filter (pf) code has been updated to OpenBSD 4.5.
155 You need to update userland tools to be in sync with kernel.
156 This update breaks backward compatibility with earlier pfsync(4)
157 versions. Care must be taken when updating redundant firewall setups.
160 The following sysctls and tunables are retired on x86 platforms:
162 machdep.hlt_logical_cpus
163 The following sysctl is retired:
164 machdep.hyperthreading_allowed
165 The sysctls were supposed to provide a way to dynamically offline and
166 online selected CPUs on x86 platforms, but the implementation has not
167 been reliable especially with SCHED_ULE scheduler.
168 machdep.hyperthreading_allowed tunable is still available to ignore
169 hyperthreading CPUs at OS level.
170 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
171 where X is an APIC ID of a CPU. Be advised, though, that disabling
172 CPUs in non-uniform fashion will result in non-uniform topology and
173 may lead to sub-optimal system performance with SCHED_ULE, which is
177 cpumask_t type is retired and cpuset_t is used in order to describe
181 Changes to ifconfig(8) for dynamic address family detection mandate
182 that you are running a kernel of 20110525 or later. Make sure to
183 follow the update procedure to boot a new kernel before installing
187 Support for sun4v architecture is officially dropped
190 Several KPI breaking changes have been committed to the mii(4) layer,
191 the PHY drivers and consequently some Ethernet drivers using mii(4).
192 This means that miibus.ko and the modules of the affected Ethernet
193 drivers need to be recompiled.
195 Note to kernel developers: Given that the OUI bit reversion problem
196 was fixed as part of these changes all mii(4) commits related to OUIs,
197 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
198 handling, no longer can be merged verbatim to stable/8 and previous
202 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
203 into their kernel configurations along with 'device pci'.
206 The default NFS client is now the new NFS client, so fstype "newnfs"
207 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
208 Although mounts via fstype "nfs" will usually work without userland
209 changes, it is recommended that the mount(8) and mount_nfs(8)
210 commands be rebuilt from sources and that a link to mount_nfs called
211 mount_oldnfs be created. The new client is compiled into the
212 kernel with "options NFSCL" and this is needed for diskless root
213 file systems. The GENERIC kernel configs have been changed to use
214 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
215 To use the regular/old client, you can "mount -t oldnfs ...". For
216 a diskless root file system, you must also include a line like:
218 vfs.root.mountfrom="oldnfs:"
220 in the boot/loader.conf on the root fs on the NFS server to make
221 a diskless root fs use the old client.
224 The GENERIC kernels for all architectures now default to the new
225 CAM-based ATA stack. It means that all legacy ATA drivers were
226 removed and replaced by respective CAM drivers. If you are using
227 ATA device names in /etc/fstab or other places, make sure to update
228 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
229 where 'Y's are the sequential numbers starting from zero for each type
230 in order of detection, unless configured otherwise with tunables,
231 see cam(4)). There will be symbolic links created in /dev/ to map
232 old adX devices to the respective adaY. They should provide basic
233 compatibility for file systems mounting in most cases, but they do
234 not support old user-level APIs and do not have respective providers
235 in GEOM. Consider using updated management tools with new device names.
237 It is possible to load devices ahci, ata, siis and mvs as modules,
238 but option ATA_CAM should remain in kernel configuration to make ata
239 module work as CAM driver supporting legacy ATA controllers. Device ata
240 still can be used in modular fashion (atacore + ...). Modules atadisk
241 and atapi* are not used and won't affect operation in ATA_CAM mode.
242 Note that to use CAM-based ATA kernel should include CAM devices
243 scbus, pass, da (or explicitly ada), cd and optionally others. All of
244 them are parts of the cam module.
246 ataraid(4) functionality is now supported by the RAID GEOM class.
247 To use it you can load geom_raid kernel module and use graid(8) tool
248 for management. Instead of /dev/arX device names, use /dev/raid/rX.
250 No kernel config options or code have been removed, so if a problem
251 arises, please report it and optionally revert to the old ATA stack.
252 In order to do it you can remove from the kernel config:
257 , and instead add back:
258 device atadisk # ATA disk drives
259 device ataraid # ATA RAID drives
260 device atapicd # ATAPI CDROM drives
261 device atapifd # ATAPI floppy drives
262 device atapist # ATAPI tape drives
265 The default NFS server has been changed to the new server, which
266 was referred to as the experimental server. If you need to switch
267 back to the old NFS server, you must now put the "-o" option on
268 both the mountd and nfsd commands. This can be done using the
269 mountd_flags and nfs_server_flags rc.conf variables until an
270 update to the rc scripts is committed, which is coming soon.
273 The GNU Objective-C runtime library (libobjc), and other Objective-C
274 related components have been removed from the base system. If you
275 require an Objective-C library, please use one of the available ports.
278 ath(4) has been split into bus- and device- modules. if_ath contains
279 the HAL, the TX rate control and the network device code. if_ath_pci
280 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
281 contains the AHB glue. Users need to load both if_ath_pci and if_ath
282 in order to use ath on everything else.
284 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
285 need to load if_ath and if_ath_pci for ath(4) operation.
288 As part of the replacement of sysinstall, the process of building
289 release media has changed significantly. For details, please re-read
290 release(7), which has been updated to reflect the new build process.
293 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
294 is the last available version under GPLv2. It brings a number of new
295 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
296 4.1 and SSE 4.2), better support for powerpc64, a number of new
297 directives, and lots of other small improvements. See the ChangeLog
298 file in contrib/binutils for the full details.
301 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
302 compliant, and will now use half of hash for authentication.
303 This will break interoperability with all stacks (including all
304 actual FreeBSD versions) who implement
305 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
307 The only workaround with such peers is to use another HMAC
308 algorithm for IPsec ("phase 2") authentication.
311 Remove the uio_yield prototype and symbol. This function has
312 been misnamed since it was introduced and should not be
313 globally exposed with this name. The equivalent functionality
314 is now available using kern_yield(curthread->td_user_pri).
315 The function remains undocumented.
318 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
319 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
320 for scalar sysctls is defined but disabled. Code that needs
321 UQUAD to pass the type checking that must compile on older
322 systems where the define is not present can check against
323 __FreeBSD_version >= 900030.
325 The system dialog(1) has been replaced with a new version previously
326 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
327 with the previous version, but the libdialog associated with it has
328 a largely incompatible API. As such, the original version of libdialog
329 will be kept temporarily as libodialog, until its base system consumers
330 are replaced or updated. Bump __FreeBSD_version to 900030.
333 If you are trying to run make universe on a -stable system, and you get
334 the following warning:
335 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
336 unknown. config(8) likely too old."
337 or something similar to it, then you must upgrade your -stable system
338 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
339 stable/8) or build the config from the latest stable/8 branch and
340 install it on your system.
342 Prior to this date, building a current universe on 8-stable system from
343 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
344 error in the first kernel build phase. A new config on those old
345 systems will fix that problem for older versions of -current.
348 The TCP stack has been modified to allow Khelp modules to interact with
349 it via helper hook points and store per-connection data in the TCP
350 control block. Bump __FreeBSD_version to 900029. User space tools that
351 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
355 Generic IEEE 802.3 annex 31B full duplex flow control support has been
356 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
357 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
358 to take advantage of it instead of using custom implementations. This
359 means that these drivers now no longer unconditionally advertise
360 support for flow control but only do so if flow control is a selected
361 media option. This was implemented in the generic support that way in
362 order to allow flow control to be switched on and off via ifconfig(8)
363 with the PHY specific default to typically off in order to protect
364 from unwanted effects. Consequently, if you used flow control with
365 one of the above mentioned drivers you now need to explicitly enable
367 ifconfig bge0 media auto mediaopt flowcontrol
369 Along with the above mentioned changes generic support for setting
370 1000baseT master mode also has been added and brgphy(4), ciphy(4),
371 e1000phy(4) as well as ip1000phy(4) have been converted to take
372 advantage of it. This means that these drivers now no longer take the
373 link0 parameter for selecting master mode but the master media option
374 has to be used instead, for example like in the following:
375 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
377 Selection of master mode now is also available with all other PHY
378 drivers supporting 1000baseT.
381 The TCP stack has received a significant update to add support for
382 modularised congestion control and generally improve the clarity of
383 congestion control decisions. Bump __FreeBSD_version to 900025. User
384 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
385 sockstat) need to be recompiled.
388 The man(1) utility has been replaced by a new version that no longer
389 uses /etc/manpath.config. Please consult man.conf(5) for how to
390 migrate local entries to the new format.
393 The copyright strings printed by login(1) and sshd(8) at the time of a
394 new connection have been removed to follow other operating systems and
398 A workaround for a fixed ld bug has been removed in kernel code,
399 so make sure that your system ld is built from sources after
400 revision 210245 from 2010-07-19 (r211583 if building head kernel
401 on stable/8, r211584 for stable/7; both from 2010-08-21).
402 A symptom of incorrect ld version is different addresses for
403 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
406 The $ipv6_prefer variable in rc.conf(5) has been split into
407 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
409 The $ip6addrctl_policy is a variable to choose a pre-defined
410 address selection policy set by ip6addrctl(8). A value
411 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
414 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
415 flag (see an entry of 20090926) is set on an interface with no
416 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
417 security reason. If you want IPv6 link-local address on all
418 interfaces by default, set this to "YES".
420 The old ipv6_prefer="YES" is equivalent to
421 ipv6_activate_all_interfaces="YES" and
422 ip6addrctl_policy="ipv6_prefer".
425 DTrace has grown support for userland tracing. Due to this, DTrace is
426 now i386 and amd64 only.
427 dtruss(1) is now installed by default on those systems and a new
428 kernel module is needed for userland tracing: fasttrap.
429 No changes to your kernel config file are necessary to enable
430 userland tracing, but you might consider adding 'STRIP=' and
431 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
432 to have informative userland stack traces in DTrace (ustack).
435 The acpi_aiboost(4) driver has been removed in favor of the new
436 aibs(4) driver. You should update your kernel configuration file.
439 BSD grep has been imported to the base system and it is built by
440 default. It is completely BSD licensed, highly GNU-compatible, uses
441 less memory than its GNU counterpart and has a small codebase.
442 However, it is slower than its GNU counterpart, which is mostly
443 noticeable for larger searches, for smaller ones it is measurable
444 but not significant. The reason is complex, the most important factor
445 is that we lack a modern and efficient regex library and GNU
446 overcomes this by optimizing the searches internally. Future work
447 on improving the regex performance is planned, for the meantime,
448 users that need better performance, can build GNU grep instead by
449 setting the WITH_GNU_GREP knob.
452 Due to the import of powerpc64 support, all existing powerpc kernel
453 configuration files must be updated with a machine directive like this:
454 machine powerpc powerpc
456 In addition, an updated config(8) is required to build powerpc kernels
460 A new version of ZFS (version 15) has been merged to -HEAD.
461 This version uses a python library for the following subcommands:
462 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
463 For full functionality of these commands the following port must
464 be installed: sysutils/py-zfs
467 'vm_page's are now hashed by physical address to an array of mutexes.
468 Currently this is only used to serialize access to hold_count. Over
469 time the page queue mutex will be peeled away. This changes the size
470 of pmap on every architecture. And requires all callers of vm_page_hold
471 and vm_page_unhold to be updated.
474 WITH_CTF can now be specified in src.conf (not recommended, there
475 are some problems with static executables), make.conf (would also
476 affect ports which do not use GNU make and do not override the
477 compile targets) or in the kernel config (via "makeoptions
479 When WITH_CTF was specified there before this was silently ignored,
480 so make sure that WITH_CTF is not used in places which could lead
481 to unwanted behavior.
484 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
485 to allow 32-bit compatibility on non-x86 platforms. All kernel
486 configurations on amd64 and ia64 platforms using these options must
487 be modified accordingly.
490 The utmp user accounting database has been replaced with utmpx,
491 the user accounting interface standardized by POSIX.
492 Unfortunately the semantics of utmp and utmpx don't match,
493 making it practically impossible to support both interfaces.
494 The user accounting database is used by tools like finger(1),
495 last(1), talk(1), w(1) and ac(8).
497 All applications in the base system use utmpx. This means only
498 local binaries (e.g. from the ports tree) may still use these
499 utmp database files. These applications must be rebuilt to make
502 After the system has been upgraded, it is safe to remove the old
503 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
504 assuming their contents is of no importance anymore. Old wtmp
505 databases can only be used by last(1) and ac(8) after they have
506 been converted to the new format using wtmpcvt(1).
509 Introduce the kernel thread "deadlock resolver" (which can be enabled
510 via the DEADLKRES option, see NOTES for more details) and the
511 sleepq_type() function for sleepqueues.
514 The rc.firewall and rc.firewall6 were unified, and
515 rc.firewall6 and rc.d/ip6fw were removed.
516 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
517 variables are obsoleted. Instead, the following new rc
518 variables are added to rc.d/ipfw:
520 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
521 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
522 firewall_simple_onet_ipv6, firewall_trusted_ipv6
524 The meanings correspond to the relevant IPv4 variables.
530 The default terminal emulation for syscons(4) has been changed
531 from cons25 to xterm on all platforms except pc98. This means
532 that the /etc/ttys file needs to be updated to ensure correct
533 operation of applications on the console.
535 The terminal emulation style can be toggled per window by using
536 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
537 options can be used to change the compile-time default back to
540 To prevent graphical artifacts, make sure the TERM environment
541 variable is set to match the terminal emulation that is being
542 performed by syscons(4).
545 The layout of the structure ieee80211req_scan_result has changed.
546 Applications that require wireless scan results (e.g. ifconfig(8))
547 from net80211 need to be recompiled.
549 Applications such as wpa_supplicant(8) may require a full world
550 build without using NO_CLEAN in order to get synchronized with the
554 The iwn(4) driver has been updated to support the 5000 and 5150 series.
555 There's one kernel module for each firmware. Adding "device iwnfw"
556 to the kernel configuration file means including all three firmware
557 images inside the kernel. If you want to include just the one for
558 your wireless card, use the devices iwn4965fw, iwn5000fw or
562 The rc.d/network_ipv6, IPv6 configuration script has been integrated
563 into rc.d/netif. The changes are the following:
565 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
566 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
567 Note that both variables need the "inet6" keyword at the head.
569 Do not set $ipv6_network_interfaces manually if you do not
570 understand what you are doing. It is not needed in most cases.
572 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
575 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
576 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
578 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
579 all configured interfaces have "inet6 accept_rtadv" in the
580 $ifconfig_IF_ipv6. These are for backward compatibility.
582 3. A new variable $ipv6_prefer has been added. If NO, IPv6
583 functionality of interfaces with no corresponding
584 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
585 and the default address selection policy of ip6addrctl(8)
586 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
587 Note that if you want to configure IPv6 functionality on the
588 disabled interfaces after boot, first you need to clear the flag by
589 using ifconfig(8) like:
591 ifconfig em0 inet6 -ifdisabled
593 If YES, the default address selection policy is set as
596 The default value of $ipv6_prefer is NO.
598 4. If your system need to receive Router Advertisement messages,
599 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
600 scripts automatically invoke rtsol(8) when the interface becomes
601 UP. The Router Advertisement messages are used for SLAAC
602 (State-Less Address AutoConfiguration).
605 802.11s D3.03 support was committed. This is incompatible with the
606 previous code, which was based on D3.0.
609 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
610 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
611 control whether accepting Router Advertisement messages or not.
612 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
613 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
614 The ifconfig(8) utility now supports these flags.
617 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
618 mount(8) and -a option for df(1) to see them.
621 The old tunable hw.bus.devctl_disable has been superseded by
622 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
623 replaced by hw.bus.devctl_queue=0. The default for this new tunable
627 Remove the option STOP_NMI. The default action is now to use NMI only
628 for KDB via the newly introduced function stop_cpus_hard() and
629 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
632 The stable/8 branch created in subversion. This corresponds to the
633 RELENG_8 branch in CVS.
636 Bump the shared library version numbers for all libraries that do not
637 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
638 __FreeBSD_version to 800105.
641 Due to changes in the implementation of virtual network stack support,
642 all network-related kernel modules must be recompiled. As this change
643 breaks the ABI, bump __FreeBSD_version to 800104.
646 The TOE interface to the TCP syncache has been modified to remove
647 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
648 The cxgb driver is the only TOE consumer affected by this change, and
649 needs to be recompiled along with the kernel. As this change breaks
650 the ABI, bump __FreeBSD_version to 800103.
653 Padding has been added to struct tcpcb, sackhint and tcpstat in
654 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
655 maintaining the ABI. However, this change breaks the ABI, so bump
656 __FreeBSD_version to 800102. User space tools that rely on the size of
657 any of these structs (e.g. sockstat) need to be recompiled.
660 The NFS_LEGACYRPC option has been removed along with the old kernel
661 RPC implementation that this option selected. Kernel configurations
662 may need to be adjusted.
665 The network interface device nodes at /dev/net/<interface> have been
666 removed. All ioctl operations can be performed the normal way using
667 routing sockets. The kqueue functionality can generally be replaced
668 with routing sockets.
671 The documentation from the FreeBSD Documentation Project (Handbook,
672 FAQ, etc.) is now installed via packages by sysinstall(8) and under
673 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
676 The ABI of various structures related to the SYSV IPC API have been
677 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
678 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
682 Layout of struct vnet has changed as routing related variables were
683 moved to their own Vimage module. Modules need to be recompiled. Bump
684 __FreeBSD_version to 800099.
687 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
688 respectively. As long as no more than 16 groups per process are used,
689 no changes should be visible. When more than 16 groups are used, old
690 binaries may fail if they call getgroups() or getgrouplist() with
691 statically sized storage. Recompiling will work around this, but
692 applications should be modified to use dynamically allocated storage
693 for group arrays as POSIX.1-2008 does not cap an implementation's
694 number of supported groups at NGROUPS_MAX+1 as previous versions did.
696 NFS and portalfs mounts may also be affected as the list of groups is
697 truncated to 16. Users of NFS who use more than 16 groups, should
698 take care that negative group permissions are not used on the exported
699 file systems as they will not be reliable unless a GSSAPI based
700 authentication method is used.
703 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
704 option compiles in the support for adaptive spinning for lockmgrs
705 which want to enable it. The lockinit() function now accepts the flag
706 LK_ADAPTIVE in order to make the lock object subject to adaptive
707 spinning when both held in write and read mode.
710 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
711 changed. User applications that use this ioctl need to be rebuilt.
714 The layout of struct thread has changed. Kernel and modules need to
718 The layout of structs ifnet, domain, protosw and vnet_net has changed.
719 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
722 window(1) has been removed from the base system. It can now be
723 installed from ports. The port is called misc/window.
726 The way we are storing and accessing `routing table' entries has
727 changed. Programs reading the FIB, like netstat, need to be
731 A new netisr implementation has been added for FreeBSD 8. Network
732 file system modules, such as igmp, ipdivert, and others, should be
734 Bump __FreeBSD_version to 800096.
737 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
741 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
742 Bump __FreeBSD_version to 800094.
745 Add mnt_xflag field to 'struct mount'. File system modules need to be
747 Bump __FreeBSD_version to 800093.
750 The compiling option ADAPTIVE_SX has been retired while it has been
751 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
752 The KPI for sx_init_flags() changes as accepting flags:
753 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
754 been introduced in order to handle the reversed logic.
755 Bump __FreeBSD_version to 800092.
758 Add support for hierarchical jails. Remove global securelevel.
759 Bump __FreeBSD_version to 800091.
762 The layout of struct vnet_net has changed, therefore modules
764 Bump __FreeBSD_version to 800090.
767 The newly imported zic(8) produces a new format in the output. Please
768 run tzsetup(8) to install the newly created data to /etc/localtime.
771 The sysctl tree for the usb stack has renamed from hw.usb2.* to
772 hw.usb.* and is now consistent again with previous releases.
775 802.11 monitor mode support was revised and driver api's were changed.
776 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
777 of DLT_IEEE802_11. No user-visible data structures were changed but
778 applications that use DLT_IEEE802_11 may require changes.
779 Bump __FreeBSD_version to 800088.
782 The layout of the following structs has changed: sysctl_oid,
783 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
784 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
785 panics may be experienced. World rebuild is required for
786 correctly checking networking state from userland.
787 Bump __FreeBSD_version to 800085.
790 MLDv2 and Source-Specific Multicast (SSM) have been merged
791 to the IPv6 stack. VIMAGE hooks are in but not yet used.
792 The implementation of SSM within FreeBSD's IPv6 stack closely
793 follows the IPv4 implementation.
795 For kernel developers:
797 * The most important changes are that the ip6_output() and
798 ip6_input() paths no longer take the IN6_MULTI_LOCK,
799 and this lock has been downgraded to a non-recursive mutex.
801 * As with the changes to the IPv4 stack to support SSM, filtering
802 of inbound multicast traffic must now be performed by transport
803 protocols within the IPv6 stack. This does not apply to TCP and
804 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
806 * The KPIs used by IPv6 multicast are similar to those used by
807 the IPv4 stack, with the following differences:
808 * im6o_mc_filter() is analogous to imo_multicast_filter().
809 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
810 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
811 * IN6_LOOKUP_MULTI() has been deprecated and removed.
812 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
813 for MLDv1 have an additional 'timer' argument which is used to
814 jitter the initial membership report for the solicited-node
815 multicast membership on-link.
816 * This is not strictly needed for MLDv2, which already jitters
817 its report transmissions. However, the 'timer' argument is
818 preserved in case MLDv1 is active on the interface.
820 * The KAME linked-list based IPv6 membership implementation has
821 been refactored to use a vector similar to that used by the IPv4
823 Code which maintains a list of its own multicast memberships
824 internally, e.g. carp, has been updated to reflect the new
827 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
828 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
829 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
830 implementation constraint which needs to be addressed in HEAD.
832 For application developers:
834 * The changes are broadly similar to those made for the IPv4
837 * The use of IPv4 and IPv6 multicast socket options on the same
838 socket, using mapped addresses, HAS NOT been tested or supported.
840 * There are a number of issues with the implementation of various
841 IPv6 multicast APIs which need to be resolved in the API surface
842 before the implementation is fully compatible with KAME userland
843 use, and these are mostly to do with interface index treatment.
845 * The literature available discusses the use of either the delta / ASM
846 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
847 using setsourcefilter(3)/getsourcefilter(3). For more information
848 please refer to RFC 3768, 'Socket Interface Extensions for
849 Multicast Source Filters'.
851 * Applications which use the published RFC 3678 APIs should be fine.
853 For systems administrators:
855 * The mtest(8) utility has been refactored to support IPv6, in
856 addition to IPv4. Interface addresses are no longer accepted
857 as arguments, their names must be used instead. The utility
858 will map the interface name to its first IPv4 address as
859 returned by getifaddrs(3).
861 * The ifmcstat(8) utility has also been updated to print the MLDv2
862 endpoint state and source filter lists via sysctl(3).
864 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
865 loopback of IPv6 multicast datagrams by default; it defaults to 1
866 to preserve the existing behaviour. Disabling multicast loopback is
867 recommended for optimal system performance.
869 * The IPv6 MROUTING code has been changed to examine this sysctl
870 instead of attempting to perform a group lookup before looping
871 back forwarded datagrams.
873 Bump __FreeBSD_version to 800084.
876 Implement low-level Bluetooth HCI API.
877 Bump __FreeBSD_version to 800083.
880 The layout of struct malloc_type, used by modules to register new
881 memory allocation types, has changed. Most modules will need to
882 be rebuilt or panics may be experienced.
883 Bump __FreeBSD_version to 800081.
886 Anticipate overflowing inp_flags - add inp_flags2.
887 This changes most offsets in inpcb, so checking v4 connection
888 state will require a world rebuild.
889 Bump __FreeBSD_version to 800080.
892 Add an llentry to struct route and struct route_in6. Modules
893 embedding a struct route will need to be recompiled.
894 Bump __FreeBSD_version to 800079.
897 The size of rt_metrics_lite and by extension rtentry has changed.
898 Networking administration apps will need to be recompiled.
899 The route command now supports show as an alias for get, weighting
900 of routes, sticky and nostick flags to alter the behavior of stateful
902 Bump __FreeBSD_version to 800078.
905 Do not use Giant for kbdmux(4) locking. This is wrong and
906 apparently causing more problems than it solves. This will
907 re-open the issue where interrupt handlers may race with
908 kbdmux(4) in polling mode. Typical symptoms include (but
909 not limited to) duplicated and/or missing characters when
910 low level console functions (such as gets) are used while
911 interrupts are enabled (for example geli password prompt,
912 mountroot prompt etc.). Disabling kbdmux(4) may help.
915 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
916 kernel modules referencing any of the above need to be recompiled.
917 Bump __FreeBSD_version to 800075.
920 GEOM_PART has become the default partition slicer for storage devices,
921 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
922 introduces some changes:
924 MSDOS/EBR: the devices created from MSDOS extended partition entries
925 (EBR) can be named differently than with GEOM_MBR and are now symlinks
926 to devices with offset-based names. fstabs may need to be modified.
928 BSD: the "geometry does not match label" warning is harmless in most
929 cases but it points to problems in file system misalignment with
930 disk geometry. The "c" partition is now implicit, covers the whole
931 top-level drive and cannot be (mis)used by users.
933 General: Kernel dumps are now not allowed to be written to devices
934 whose partition types indicate they are meant to be used for file
935 systems (or, in case of MSDOS partitions, as something else than
938 Most of these changes date approximately from 200812.
941 The uscanner(4) driver has been removed from the kernel. This follows
942 Linux removing theirs in 2.6 and making libusb the default interface
946 The multicast forwarding code has been cleaned up. netstat(1)
947 only relies on KVM now for printing bandwidth upcall meters.
948 The IPv4 and IPv6 modules are split into ip_mroute_mod and
949 ip6_mroute_mod respectively. The config(5) options for statically
950 compiling this code remain the same, i.e. 'options MROUTING'.
953 Support for the IFF_NEEDSGIANT network interface flag has been
954 removed, which means that non-MPSAFE network device drivers are no
955 longer supported. In particular, if_ar, if_sr, and network device
956 drivers from the old (legacy) USB stack can no longer be built or
960 POSIX.1 Native Language Support (NLS) has been enabled in libc and
961 a bunch of new language catalog files have also been added.
962 This means that some common libc messages are now localized and
963 they depend on the LC_MESSAGES environmental variable.
966 The k8temp(4) driver has been renamed to amdtemp(4) since
967 support for Family 10 and Family 11 CPU families was added.
970 IGMPv3 and Source-Specific Multicast (SSM) have been merged
971 to the IPv4 stack. VIMAGE hooks are in but not yet used.
973 For kernel developers, the most important changes are that the
974 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
975 and this lock has been downgraded to a non-recursive mutex.
977 Transport protocols (UDP, Raw IP) are now responsible for filtering
978 inbound multicast traffic according to group membership and source
979 filters. The imo_multicast_filter() KPI exists for this purpose.
980 Transports which do not use multicast (SCTP, TCP) already reject
981 multicast by default. Forwarding and receive performance may improve
982 as a mutex acquisition is no longer needed in the ip_input()
983 low-level input path. in_addmulti() and in_delmulti() are shimmed
984 to new KPIs which exist to support SSM in-kernel.
986 For application developers, it is recommended that loopback of
987 multicast datagrams be disabled for best performance, as this
988 will still cause the lock to be taken for each looped-back
989 datagram transmission. The net.inet.ip.mcast.loop sysctl may
990 be tuned to 0 to disable loopback by default; it defaults to 1
991 to preserve the existing behaviour.
993 For systems administrators, to obtain best performance with
994 multicast reception and multiple groups, it is always recommended
995 that a card with a suitably precise hash filter is used. Hash
996 collisions will still result in the lock being taken within the
997 transport protocol input path to check group membership.
999 If deploying FreeBSD in an environment with IGMP snooping switches,
1000 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1001 enabled; this forces 224.0.0.0/24 group membership to be announced
1004 The size of 'struct igmpstat' has changed; netstat needs to be
1005 recompiled to reflect this.
1006 Bump __FreeBSD_version to 800070.
1009 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1010 updated to use it. This requires a buildworld/installworld in order to
1011 update the library and dependencies (usbconfig, etc). Its advisable to
1012 rebuild all ports which uses libusb. More specific directions are given
1013 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1014 libusb are no longer required and can be removed.
1017 A workaround is committed to allow the creation of System V shared
1018 memory segment of size > 2 GB on the 64-bit architectures.
1019 Due to a limitation of the existing ABI, the shm_segsz member
1020 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1021 wrong for large segments. Note that limits must be explicitly
1022 raised to allow such segments to be created.
1025 The layout of struct ifnet has changed, requiring a rebuild of all
1026 network device driver modules.
1029 The /dev handling for the new USB stack has changed, a
1030 buildworld/installworld is required for libusb20.
1033 The new USB2 stack has now been permanently moved in and all kernel and
1034 module names reverted to their previous values (eg, usb, ehci, ohci,
1035 ums, ...). The old usb stack can be compiled in by prefixing the name
1036 with the letter 'o', the old usb modules have been removed.
1037 Updating entry 20090216 for xorg and 20090215 for libmap may still
1041 The rc.conf(5) option if_up_delay has been renamed to
1042 defaultroute_delay to better reflect its purpose. If you have
1043 customized this setting in /etc/rc.conf you need to update it to
1047 xorg 7.4 wants to configure its input devices via hald which does not
1048 yet work with USB2. If the keyboard/mouse does not work in xorg then
1050 Option "AllowEmptyInput" "off"
1051 to your ServerLayout section. This will cause X to use the configured
1052 kbd and mouse sections from your xorg.conf.
1055 The GENERIC kernels for all architectures now default to the new USB2
1056 stack. No kernel config options or code have been removed so if a
1057 problem arises please report it and optionally revert to the old USB
1058 stack. If you are loading USB kernel modules or have a custom kernel
1059 that includes GENERIC then ensure that usb names are also changed over,
1060 eg uftdi -> usb2_serial_ftdi.
1062 Older programs linked against the ports libusb 0.1 need to be
1063 redirected to the new stack's libusb20. /etc/libmap.conf can
1065 # Map old usb library to new one for usb2 stack
1066 libusb-0.1.so.8 libusb20.so.1
1069 All USB ethernet devices now attach as interfaces under the name ueN
1070 (eg. ue0). This is to provide a predictable name as vendors often
1071 change usb chipsets in a product without notice.
1074 The ichsmb(4) driver has been changed to require SMBus slave
1075 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1076 All of the other SMBus controller drivers require left-justified
1077 slave addresses, so this change makes all the drivers provide the
1081 INET6 statistics (struct ip6stat) was updated.
1082 netstat(1) needs to be recompiled.
1085 NTFS has been removed from GENERIC kernel on amd64 to match
1086 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1087 will load ntfs.ko module automatically when NTFS support is
1088 actually needed, unless ntfs.ko is not installed or security
1089 level prohibits loading kernel modules. If either is the case,
1090 "options NTFS" has to be added into kernel config.
1093 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1094 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1095 800061. User space tools that rely on the size of struct tcpcb in
1096 tcp_var.h (e.g. sockstat) need to be recompiled.
1099 ng_tty(4) module updated to match the new TTY subsystem.
1100 Due to API change, user-level applications must be updated.
1101 New API support added to mpd5 CVS and expected to be present
1102 in next mpd5.3 release.
1105 With __FreeBSD_version 800060 the makefs tool is part of
1106 the base system (it was a port).
1109 The afdata and ifnet locks have been changed from mutexes to
1110 rwlocks, network modules will need to be re-compiled.
1113 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1114 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1115 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1116 architecture and 40 bytes on 64-bit architecture. The userland
1117 applications "arp" and "ndp" have been updated accordingly.
1118 The output from "netstat -r" shows only routing entries and
1119 none of the L2 information.
1122 __FreeBSD_version 800057 marks the switchover from the
1123 binary ath hal to source code. Users must add the line:
1125 options AH_SUPPORT_AR5416
1127 to their kernel config files when specifying:
1131 The ath_hal module no longer exists; the code is now compiled
1132 together with the driver in the ath module. It is now
1133 possible to tailor chip support (i.e. reduce the set of chips
1134 and thereby the code size); consult ath_hal(4) for details.
1137 __FreeBSD_version 800054 adds memory barriers to
1138 <machine/atomic.h>, new interfaces to ifnet to facilitate
1139 multiple hardware transmit queues for cards that support
1140 them, and a lock-less ring-buffer implementation to
1141 enable drivers to more efficiently manage queueing of
1145 A new version of ZFS (version 13) has been merged to -HEAD.
1146 This version has zpool attribute "listsnapshots" off by
1147 default, which means "zfs list" does not show snapshots,
1148 and is the same as Solaris behavior.
1151 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1154 The uhci, ohci, ehci and slhci USB Host controller drivers have
1155 been put into separate modules. If you load the usb module
1156 separately through loader.conf you will need to load the
1157 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1158 controller add the following to loader.conf:
1164 The ABI used by the PMC toolset has changed. Please keep
1165 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1169 atapci kernel module now includes only generic PCI ATA
1170 driver. AHCI driver moved to ataahci kernel module.
1171 All vendor-specific code moved into separate kernel modules:
1172 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1173 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1174 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1175 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1178 The TTY subsystem of the kernel has been replaced by a new
1179 implementation, which provides better scalability and an
1180 improved driver model. Most common drivers have been migrated to
1181 the new TTY subsystem, while others have not. The following
1182 drivers have not yet been ported to the new TTY layer:
1185 cy, digi, rc, rp, sio
1191 ng_h4, ng_tty, ppp, sl, snp
1193 Adding these drivers to your kernel configuration file shall
1194 cause compilation to fail.
1197 ntpd has been upgraded to 4.2.4p5.
1200 OpenSSH has been upgraded to 5.1p1.
1202 For many years, FreeBSD's version of OpenSSH preferred DSA
1203 over RSA for host and user authentication keys. With this
1204 upgrade, we've switched to the vendor's default of RSA over
1205 DSA. This may cause upgraded clients to warn about unknown
1206 host keys even for previously known hosts. Users should
1207 follow the usual procedure for verifying host keys before
1208 accepting the RSA key.
1210 This can be circumvented by setting the "HostKeyAlgorithms"
1211 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1214 Please note that the sequence of keys offered for
1215 authentication has been changed as well. You may want to
1216 specify IdentityFile in a different order to revert this
1220 The sio(4) driver has been removed from the i386 and amd64
1221 kernel configuration files. This means uart(4) is now the
1222 default serial port driver on those platforms as well.
1224 To prevent collisions with the sio(4) driver, the uart(4) driver
1225 uses different names for its device nodes. This means the
1226 onboard serial port will now most likely be called "ttyu0"
1227 instead of "ttyd0". You may need to reconfigure applications to
1228 use the new device names.
1230 When using the serial port as a boot console, be sure to update
1231 /boot/device.hints and /etc/ttys before booting the new kernel.
1232 If you forget to do so, you can still manually specify the hints
1233 at the loader prompt:
1235 set hint.uart.0.at="isa"
1236 set hint.uart.0.port="0x3F8"
1237 set hint.uart.0.flags="0x10"
1238 set hint.uart.0.irq="4"
1242 The gpt(8) utility has been removed. Use gpart(8) to partition
1246 The version that Linuxulator emulates was changed from 2.4.2
1247 to 2.6.16. If you experience any problems with Linux binaries
1248 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1249 if it fixes the problem contact emulation mailing list.
1252 ISDN4BSD (I4B) was removed from the src tree. You may need to
1253 update a your kernel configuration and remove relevant entries.
1256 I have checked in code to support multiple routing tables.
1257 See the man pages setfib(1) and setfib(2).
1258 This is a hopefully backwards compatible version,
1259 but to make use of it you need to compile your kernel
1260 with options ROUTETABLES=2 (or more up to 16).
1263 The 802.11 wireless support was redone to enable multi-bss
1264 operation on devices that are capable. The underlying device
1265 is no longer used directly but instead wlanX devices are
1266 cloned with ifconfig. This requires changes to rc.conf files.
1267 For example, change:
1268 ifconfig_ath0="WPA DHCP"
1271 ifconfig_wlan0="WPA DHCP"
1272 see rc.conf(5) for more details. In addition, mergemaster of
1273 /etc/rc.d is highly recommended. Simultaneous update of userland
1274 and kernel wouldn't hurt either.
1276 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1277 modules were merged into the base wlan module. All references
1278 to these modules (e.g. in kernel config files) must be removed.
1281 psm(4) has gained write(2) support in native operation level.
1282 Arbitrary commands can be written to /dev/psm%d and status can
1283 be read back from it. Therefore, an application is responsible
1284 for status validation and error recovery. It is a no-op in
1285 other operation levels.
1288 Support for KSE threading has been removed from the kernel. To
1289 run legacy applications linked against KSE libmap.conf may
1290 be used. The following libmap.conf may be used to ensure
1291 compatibility with any prior release:
1293 libpthread.so.1 libthr.so.1
1294 libpthread.so.2 libthr.so.2
1295 libkse.so.3 libthr.so.3
1298 The layout of struct vmspace has changed. This affects libkvm
1299 and any executables that link against libkvm and use the
1300 kvm_getprocs() function. In particular, but not exclusively,
1301 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1302 The effects are minimal, but it's advisable to upgrade world
1306 The latest em driver no longer has support in it for the
1307 82575 adapter, this is now moved to the igb driver. The
1308 split was done to make new features that are incompatible
1309 with older hardware easier to do.
1312 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1313 likewise the kernel option is now GEOM_LINUX_LVM.
1316 The default NFS mount mode has changed from UDP to TCP for
1317 increased reliability. If you rely on (insecurely) NFS
1318 mounting across a firewall you may need to update your
1322 Belatedly note the addition of m_collapse for compacting
1326 The fts(3) structures have been changed to use adequate
1327 integer types for their members and so to be able to cope
1328 with huge file trees. The old fts(3) ABI is preserved
1329 through symbol versioning in libc, so third-party binaries
1330 using fts(3) should still work, although they will not take
1331 advantage of the extended types. At the same time, some
1332 third-party software might fail to build after this change
1333 due to unportable assumptions made in its source code about
1334 fts(3) structure members. Such software should be fixed
1335 by its vendor or, in the worst case, in the ports tree.
1336 FreeBSD_version 800015 marks this change for the unlikely
1337 case that a portable fix is impossible.
1340 To upgrade to -current after this date, you must be running
1341 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1342 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1345 The ADAPTIVE_GIANT kernel option has been retired because its
1346 functionality is the default now.
1349 The AT keyboard emulation of sunkbd(4) has been turned on
1350 by default. In order to make the special symbols of the Sun
1351 keyboards driven by sunkbd(4) work under X these now have
1352 to be configured the same way as Sun USB keyboards driven
1353 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1355 Option "XkbLayout" "us"
1356 Option "XkbRules" "xorg"
1357 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1360 It has been decided that it is desirable to provide ABI
1361 backwards compatibility to the FreeBSD 4/5/6 versions of the
1362 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1363 broken with the introduction of PCI domain support (see the
1364 20070930 entry). Unfortunately, this required the ABI of
1365 PCIOCGETCONF to be broken again in order to be able to
1366 provide backwards compatibility to the old version of that
1367 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1368 again. As for prominent ports this affects neither pciutils
1369 nor xorg-server this time, the hal port needs to be rebuilt
1373 The misnamed kthread_create() and friends have been renamed
1374 to kproc_create() etc. Many of the callers already
1375 used kproc_start()..
1376 I will return kthread_create() and friends in a while
1377 with implementations that actually create threads, not procs.
1378 Renaming corresponds with version 800002.
1387 Avoid using make -j when upgrading. While generally safe, there are
1388 sometimes problems using -j to upgrade. If your upgrade fails with
1389 -j, please try again without -j. From time to time in the past there
1390 have been problems using -j with buildworld and/or installworld. This
1391 is especially true when upgrading between "distant" versions (eg one
1392 that cross a major release boundary or several minor releases, or when
1393 several months have passed on the -current branch).
1395 Sometimes, obscure build problems are the result of environment
1396 poisoning. This can happen because the make utility reads its
1397 environment when searching for values for global variables. To run
1398 your build attempts in an "environmental clean room", prefix all make
1399 commands with 'env -i '. See the env(1) manual page for more details.
1401 When upgrading from one major version to another it is generally best
1402 to upgrade to the latest code in the currently installed branch first,
1403 then do an upgrade to the new branch. This is the best-tested upgrade
1404 path, and has the highest probability of being successful. Please try
1405 this approach before reporting problems with a major version upgrade.
1409 When upgrading the boot ZFS pool to a new version, always follow
1412 1.) recompile and reinstall the ZFS boot loader and boot block
1413 (this is part of "make buildworld" and "make installworld")
1415 2.) update the ZFS boot block on your boot drive
1417 The following example updates the ZFS boot block on the first
1418 partition (freebsd-boot) of a GPT partitioned drive ad0:
1419 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1421 Non-boot pools do not need these updates.
1425 If you are updating from a prior version of FreeBSD (even one just
1426 a few days old), you should follow this procedure. It is the most
1427 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1429 make kernel-toolchain
1430 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1431 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1433 To test a kernel once
1434 ---------------------
1435 If you just want to boot a kernel once (because you are not sure
1436 if it works, or if you want to boot a known bad kernel to provide
1437 debugging information) run
1438 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1439 nextboot -k testkernel
1441 To just build a kernel when you know that it won't mess you up
1442 --------------------------------------------------------------
1443 This assumes you are already running a CURRENT system. Replace
1444 ${arch} with the architecture of your machine (e.g. "i386",
1445 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1447 cd src/sys/${arch}/conf
1448 config KERNEL_NAME_HERE
1449 cd ../compile/KERNEL_NAME_HERE
1454 If this fails, go to the "To build a kernel" section.
1456 To rebuild everything and install it on the current system.
1457 -----------------------------------------------------------
1458 # Note: sometimes if you are running current you gotta do more than
1459 # is listed here if you are upgrading from a really old current.
1461 <make sure you have good level 0 dumps>
1463 make kernel KERNCONF=YOUR_KERNEL_HERE
1465 <reboot in single user> [3]
1473 To cross-install current onto a separate partition
1474 --------------------------------------------------
1475 # In this approach we use a separate partition to hold
1476 # current's root, 'usr', and 'var' directories. A partition
1477 # holding "/", "/usr" and "/var" should be about 2GB in
1480 <make sure you have good level 0 dumps>
1483 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1484 <maybe newfs current's root partition>
1485 <mount current's root partition on directory ${CURRENT_ROOT}>
1486 make installworld DESTDIR=${CURRENT_ROOT}
1487 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1488 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1489 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1490 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1491 <reboot into current>
1492 <do a "native" rebuild/install as described in the previous section>
1493 <maybe install compatibility libraries from ports/misc/compat*>
1497 To upgrade in-place from stable to current
1498 ----------------------------------------------
1499 <make sure you have good level 0 dumps>
1501 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1503 <reboot in single user> [3]
1510 Make sure that you've read the UPDATING file to understand the
1511 tweaks to various things you need. At this point in the life
1512 cycle of current, things change often and you are on your own
1513 to cope. The defaults can also change, so please read ALL of
1514 the UPDATING entries.
1516 Also, if you are tracking -current, you must be subscribed to
1517 freebsd-current@freebsd.org. Make sure that before you update
1518 your sources that you have read and understood all the recent
1519 messages there. If in doubt, please track -stable which has
1520 much fewer pitfalls.
1522 [1] If you have third party modules, such as vmware, you
1523 should disable them at this point so they don't crash your
1526 [3] From the bootblocks, boot -s, and then do
1531 adjkerntz -i # if CMOS is wall time
1532 Also, when doing a major release upgrade, it is required that
1533 you boot into single user mode to do the installworld.
1535 [4] Note: This step is non-optional. Failure to do this step
1536 can result in a significant reduction in the functionality of the
1537 system. Attempting to do it by hand is not recommended and those
1538 that pursue this avenue should read this file carefully, as well
1539 as the archives of freebsd-current and freebsd-hackers mailing lists
1540 for potential gotchas. The -U option is also useful to consider.
1541 See mergemaster(8) for more information.
1543 [5] Usually this step is a noop. However, from time to time
1544 you may need to do this if you get unknown user in the following
1545 step. It never hurts to do it all the time. You may need to
1546 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1547 install) after the buildworld before this step if you last updated
1548 from [78]-stable or 9-stable before 20130430.
1550 [6] This only deletes old files and directories. Old libraries
1551 can be deleted by "make delete-old-libs", but you have to make
1552 sure that no program is using those libraries anymore.
1554 [8] In order to have a kernel that can run the 4.x binaries needed to
1555 do an installworld, you must include the COMPAT_FREEBSD4 option in
1556 your kernel. Failure to do so may leave you with a system that is
1557 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1558 required to run the 5.x binaries on more recent kernels. And so on
1559 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1561 Make sure that you merge any new devices from GENERIC since the
1562 last time you updated your kernel config file.
1564 [9] When checking out sources, you must include the -P flag to have
1565 cvs prune empty directories.
1567 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1568 "?=" instead of the "=" assignment operator, so that buildworld can
1569 override the CPUTYPE if it needs to.
1571 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1572 not on the command line, or in /etc/make.conf. buildworld will
1573 warn if it is improperly defined.
1576 This file contains a list, in reverse chronological order, of major
1577 breakages in tracking -current. It is not guaranteed to be a complete
1578 list of such breakages, and only contains entries since October 10, 2007.
1579 If you need to see UPDATING entries from before that date, you will need
1580 to fetch an UPDATING file from an older FreeBSD release.
1582 Copyright information:
1584 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1586 Redistribution, publication, translation and use, with or without
1587 modification, in full or in part, in any form or format of this
1588 document are permitted without further permission from the author.
1590 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1591 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1592 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1593 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1594 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1595 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1596 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1597 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1598 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1599 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1600 POSSIBILITY OF SUCH DAMAGE.
1602 Contact Warner Losh if you have any questions about your use of