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 9.x IS SLOW:
13 FreeBSD 9.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 disable malloc debugging, run
23 ln -s aj /etc/malloc.conf.)
26 cpumask_t type is retired and cpuset_t is used in order to describe
30 Changes to ifconfig(8) for dynamic address family detection mandate
31 that you are running a kernel of 20110525 or later. Make sure to
32 follow the update procedure to boot a new kernel before installing
36 Support for sun4v architecture is officially dropped
39 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
40 into their kernel configurations along with 'device pci'.
43 The default NFS client is now the new NFS client, so fstype "newnfs"
44 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
45 Although mounts via fstype "nfs" will usually work without userland
46 changes, it is recommended that the mount(8) and mount_nfs(8)
47 commands be rebuilt from sources and that a link to mount_nfs called
48 mount_oldnfs be created. The new client is compiled into the
49 kernel with "options NFSCL" and this is needed for diskless root
50 file systems. The GENERIC kernel configs have been changed to use
51 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
52 To use the regular/old client, you can "mount -t oldnfs ...". For
53 a diskless root file system, you must also include a line like:
55 vfs.root.mountfrom="oldnfs:"
57 in the boot/loader.conf on the root fs on the NFS server to make
58 a diskless root fs use the old client.
61 The GENERIC kernels for all architectures now default to the new
62 CAM-based ATA stack. It means that all legacy ATA drivers were
63 removed and replaced by respective CAM drivers. If you are using
64 ATA device names in /etc/fstab or other places, make sure to update
65 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
66 where 'Y's are the sequential numbers starting from zero for each type
67 in order of detection, unless configured otherwise with tunables,
68 see cam(4)). There will be symbolic links created in /dev/ to map
69 old adX devices to the respective adaY. They should provide basic
70 compatibility for file systems mounting in most cases, but they do
71 not support old user-level APIs and do not have respective providers
72 in GEOM. Consider using updated management tools with new device names.
74 It is possible to load devices ahci, ata, siis and mvs as modules,
75 but option ATA_CAM should remain in kernel configuration to make ata
76 module work as CAM driver supporting legacy ATA controllers. Device ata
77 still can be used in modular fashion (atacore + ...). Modules atadisk
78 and atapi* are not used and won't affect operation in ATA_CAM mode.
79 Note that to use CAM-based ATA kernel should include CAM devices
80 scbus, pass, da (or explicitly ada), cd and optionally others. All of
81 them are parts of the cam module.
83 ataraid(4) functionality is now supported by the RAID GEOM class.
84 To use it you can load geom_raid kernel module and use graid(8) tool
85 for management. Instead of /dev/arX device names, use /dev/raid/rX.
87 No kernel config options or code have been removed, so if a problem
88 arises, please report it and optionally revert to the old ATA stack.
89 In order to do it you can remove from the kernel config:
94 , and instead add back:
95 device atadisk # ATA disk drives
96 device ataraid # ATA RAID drives
97 device atapicd # ATAPI CDROM drives
98 device atapifd # ATAPI floppy drives
99 device atapist # ATAPI tape drives
102 The default NFS server has been changed to the new server, which
103 was referred to as the experimental server. If you need to switch
104 back to the old NFS server, you must now put the "-o" option on
105 both the mountd and nfsd commands. This can be done using the
106 mountd_flags and nfs_server_flags rc.conf variables until an
107 update to the rc scripts is committed, which is coming soon.
110 The GNU Objective-C runtime library (libobjc), and other Objective-C
111 related components have been removed from the base system. If you
112 require an Objective-C library, please use one of the available ports.
115 ath(4) has been split into bus- and device- modules. if_ath contains
116 the HAL, the TX rate control and the network device code. if_ath_pci
117 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
118 contains the AHB glue. Users need to load both if_ath_pci and if_ath
119 in order to use ath on everything else.
121 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
122 need to load if_ath and if_ath_pci for ath(4) operation.
125 As part of the replacement of sysinstall, the process of building
126 release media has changed significantly. For details, please re-read
127 release(7), which has been updated to reflect the new build process.
130 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
131 is the last available version under GPLv2. It brings a number of new
132 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
133 4.1 and SSE 4.2), better support for powerpc64, a number of new
134 directives, and lots of other small improvements. See the ChangeLog
135 file in contrib/binutils for the full details.
138 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
139 compliant, and will now use half of hash for authentication.
140 This will break interoperability with all stacks (including all
141 actual FreeBSD versions) who implement
142 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
144 The only workaround with such peers is to use another HMAC
145 algorithm for IPsec ("phase 2") authentication.
148 Remove the uio_yield prototype and symbol. This function has
149 been misnamed since it was introduced and should not be
150 globally exposed with this name. The equivalent functionality
151 is now available using kern_yield(curthread->td_user_pri).
152 The function remains undocumented.
155 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
156 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
157 for scalar sysctls is defined but disabled. Code that needs
158 UQUAD to pass the type checking that must compile on older
159 systems where the define is not present can check against
160 __FreeBSD_version >= 900030.
162 The system dialog(1) has been replaced with a new version previously
163 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
164 with the previous version, but the libdialog associated with it has
165 a largely incompatible API. As such, the original version of libdialog
166 will be kept temporarily as libodialog, until its base system consumers
167 are replaced or updated. Bump __FreeBSD_version to 900030.
170 If you are trying to run make universe on a -stable system, and you get
171 the following warning:
172 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
173 unknown. config(8) likely too old."
174 or something similar to it, then you must upgrade your -stable system
175 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
176 stable/8) or build the config from the latest stable/8 branch and
177 install it on your system.
179 Prior to this date, building a current universe on 8-stable system from
180 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
181 error in the first kernel build phase. A new config on those old
182 systems will fix that problem for older versions of -current.
185 The TCP stack has been modified to allow Khelp modules to interact with
186 it via helper hook points and store per-connection data in the TCP
187 control block. Bump __FreeBSD_version to 900029. User space tools that
188 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
192 Generic IEEE 802.3 annex 31B full duplex flow control support has been
193 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
194 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
195 to take advantage of it instead of using custom implementations. This
196 means that these drivers now no longer unconditionally advertise
197 support for flow control but only do so if flow control is a selected
198 media option. This was implemented in the generic support that way in
199 order to allow flow control to be switched on and off via ifconfig(8)
200 with the PHY specific default to typically off in order to protect
201 from unwanted effects. Consequently, if you used flow control with
202 one of the above mentioned drivers you now need to explicitly enable
204 ifconfig bge0 media auto mediaopt flowcontrol
206 Along with the above mentioned changes generic support for setting
207 1000baseT master mode also has been added and brgphy(4), ciphy(4),
208 e1000phy(4) as well as ip1000phy(4) have been converted to take
209 advantage of it. This means that these drivers now no longer take the
210 link0 parameter for selecting master mode but the master media option
211 has to be used instead, for example like in the following:
212 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
214 Selection of master mode now is also available with all other PHY
215 drivers supporting 1000baseT.
218 The TCP stack has received a significant update to add support for
219 modularised congestion control and generally improve the clarity of
220 congestion control decisions. Bump __FreeBSD_version to 900025. User
221 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
222 sockstat) need to be recompiled.
225 The man(1) utility has been replaced by a new version that no longer
226 uses /etc/manpath.config. Please consult man.conf(5) for how to
227 migrate local entries to the new format.
230 The copyright strings printed by login(1) and sshd(8) at the time of a
231 new connection have been removed to follow other operating systems and
235 A workaround for a fixed ld bug has been removed in kernel code,
236 so make sure that your system ld is built from sources after
237 revision 210245 from 2010-07-19 (r211583 if building head kernel
238 on stable/8, r211584 for stable/7; both from 2010-08-21).
239 A symptom of incorrect ld version is different addresses for
240 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
243 The $ipv6_prefer variable in rc.conf(5) has been split into
244 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
246 The $ip6addrctl_policy is a variable to choose a pre-defined
247 address selection policy set by ip6addrctl(8). A value
248 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
251 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
252 flag (see an entry of 20090926) is set on an interface with no
253 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
254 security reason. If you want IPv6 link-local address on all
255 interfaces by default, set this to "YES".
257 The old ipv6_prefer="YES" is equivalent to
258 ipv6_activate_all_interfaces="YES" and
259 ip6addrctl_policy="ipv6_prefer".
262 DTrace has grown support for userland tracing. Due to this, DTrace is
263 now i386 and amd64 only.
264 dtruss(1) is now installed by default on those systems and a new
265 kernel module is needed for userland tracing: fasttrap.
266 No changes to your kernel config file are necessary to enable
267 userland tracing, but you might consider adding 'STRIP=' and
268 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
269 to have informative userland stack traces in DTrace (ustack).
272 The acpi_aiboost(4) driver has been removed in favor of the new
273 aibs(4) driver. You should update your kernel configuration file.
276 BSD grep has been imported to the base system and it is built by
277 default. It is completely BSD licensed, highly GNU-compatible, uses
278 less memory than its GNU counterpart and has a small codebase.
279 However, it is slower than its GNU counterpart, which is mostly
280 noticeable for larger searches, for smaller ones it is measurable
281 but not significant. The reason is complex, the most important factor
282 is that we lack a modern and efficient regex library and GNU
283 overcomes this by optimizing the searches internally. Future work
284 on improving the regex performance is planned, for the meantime,
285 users that need better performance, can build GNU grep instead by
286 setting the WITH_GNU_GREP knob.
289 Due to the import of powerpc64 support, all existing powerpc kernel
290 configuration files must be updated with a machine directive like this:
291 machine powerpc powerpc
293 In addition, an updated config(8) is required to build powerpc kernels
297 A new version of ZFS (version 15) has been merged to -HEAD.
298 This version uses a python library for the following subcommands:
299 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
300 For full functionality of these commands the following port must
301 be installed: sysutils/py-zfs
304 'vm_page's are now hashed by physical address to an array of mutexes.
305 Currently this is only used to serialize access to hold_count. Over
306 time the page queue mutex will be peeled away. This changes the size
307 of pmap on every architecture. And requires all callers of vm_page_hold
308 and vm_page_unhold to be updated.
311 WITH_CTF can now be specified in src.conf (not recommended, there
312 are some problems with static executables), make.conf (would also
313 affect ports which do not use GNU make and do not override the
314 compile targets) or in the kernel config (via "makeoptions
316 When WITH_CTF was specified there before this was silently ignored,
317 so make sure that WITH_CTF is not used in places which could lead
318 to unwanted behavior.
321 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
322 to allow 32-bit compatibility on non-x86 platforms. All kernel
323 configurations on amd64 and ia64 platforms using these options must
324 be modified accordingly.
327 The utmp user accounting database has been replaced with utmpx,
328 the user accounting interface standardized by POSIX.
329 Unfortunately the semantics of utmp and utmpx don't match,
330 making it practically impossible to support both interfaces.
331 The user accounting database is used by tools like finger(1),
332 last(1), talk(1), w(1) and ac(8).
334 All applications in the base system use utmpx. This means only
335 local binaries (e.g. from the ports tree) may still use these
336 utmp database files. These applications must be rebuilt to make
339 After the system has been upgraded, it is safe to remove the old
340 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
341 assuming their contents is of no importance anymore. Old wtmp
342 databases can only be used by last(1) and ac(8) after they have
343 been converted to the new format using wtmpcvt(1).
346 Introduce the kernel thread "deadlock resolver" (which can be enabled
347 via the DEADLKRES option, see NOTES for more details) and the
348 sleepq_type() function for sleepqueues.
351 The rc.firewall and rc.firewall6 were unified, and
352 rc.firewall6 and rc.d/ip6fw were removed.
353 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
354 variables are obsoleted. Instead, the following new rc
355 variables are added to rc.d/ipfw:
357 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
358 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
359 firewall_simple_onet_ipv6, firewall_trusted_ipv6
361 The meanings correspond to the relevant IPv4 variables.
367 The default terminal emulation for syscons(4) has been changed
368 from cons25 to xterm on all platforms except pc98. This means
369 that the /etc/ttys file needs to be updated to ensure correct
370 operation of applications on the console.
372 The terminal emulation style can be toggled per window by using
373 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
374 options can be used to change the compile-time default back to
377 To prevent graphical artifacts, make sure the TERM environment
378 variable is set to match the terminal emulation that is being
379 performed by syscons(4).
382 The layout of the structure ieee80211req_scan_result has changed.
383 Applications that require wireless scan results (e.g. ifconfig(8))
384 from net80211 need to be recompiled.
386 Applications such as wpa_supplicant(8) may require a full world
387 build without using NO_CLEAN in order to get synchronized with the
391 The iwn(4) driver has been updated to support the 5000 and 5150 series.
392 There's one kernel module for each firmware. Adding "device iwnfw"
393 to the kernel configuration file means including all three firmware
394 images inside the kernel. If you want to include just the one for
395 your wireless card, use the the devices iwn4965fw, iwn5000fw or
399 The rc.d/network_ipv6, IPv6 configuration script has been integrated
400 into rc.d/netif. The changes are the following:
402 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
403 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
404 Note that both variables need the "inet6" keyword at the head.
406 Do not set $ipv6_network_interfaces manually if you do not
407 understand what you are doing. It is not needed in most cases.
409 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
412 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
413 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
415 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
416 all configured interfaces have "inet6 accept_rtadv" in the
417 $ifconfig_IF_ipv6. These are for backward compatibility.
419 3. A new variable $ipv6_prefer has been added. If NO, IPv6
420 functionality of interfaces with no corresponding
421 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
422 and the default address selection policy of ip6addrctl(8)
423 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
424 Note that if you want to configure IPv6 functionality on the
425 disabled interfaces after boot, first you need to clear the flag by
426 using ifconfig(8) like:
428 ifconfig em0 inet6 -ifdisabled
430 If YES, the default address selection policy is set as
433 The default value of $ipv6_prefer is NO.
435 4. If your system need to receive Router Advertisement messages,
436 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
437 scripts automatically invoke rtsol(8) when the interface becomes
438 UP. The Router Advertisement messages are used for SLAAC
439 (State-Less Address AutoConfiguration).
442 802.11s D3.03 support was committed. This is incompatible with the
443 previous code, which was based on D3.0.
446 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
447 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
448 control whether accepting Router Advertisement messages or not.
449 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
450 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
451 The ifconfig(8) utility now supports these flags.
454 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
455 mount(8) and -a option for df(1) to see them.
458 The old tunable hw.bus.devctl_disable has been superseded by
459 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
460 replaced by hw.bus.devctl_queue=0. The default for this new tunable
464 Remove the option STOP_NMI. The default action is now to use NMI only
465 for KDB via the newly introduced function stop_cpus_hard() and
466 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
469 The stable/8 branch created in subversion. This corresponds to the
470 RELENG_8 branch in CVS.
473 Bump the shared library version numbers for all libraries that do not
474 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
475 __FreeBSD_version to 800105.
478 Due to changes in the implementation of virtual network stack support,
479 all network-related kernel modules must be recompiled. As this change
480 breaks the ABI, bump __FreeBSD_version to 800104.
483 The TOE interface to the TCP syncache has been modified to remove
484 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
485 The cxgb driver is the only TOE consumer affected by this change, and
486 needs to be recompiled along with the kernel. As this change breaks
487 the ABI, bump __FreeBSD_version to 800103.
490 Padding has been added to struct tcpcb, sackhint and tcpstat in
491 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
492 maintaining the ABI. However, this change breaks the ABI, so bump
493 __FreeBSD_version to 800102. User space tools that rely on the size of
494 any of these structs (e.g. sockstat) need to be recompiled.
497 The NFS_LEGACYRPC option has been removed along with the old kernel
498 RPC implementation that this option selected. Kernel configurations
499 may need to be adjusted.
502 The network interface device nodes at /dev/net/<interface> have been
503 removed. All ioctl operations can be performed the normal way using
504 routing sockets. The kqueue functionality can generally be replaced
505 with routing sockets.
508 The documentation from the FreeBSD Documentation Project (Handbook,
509 FAQ, etc.) is now installed via packages by sysinstall(8) and under
510 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
513 The ABI of various structures related to the SYSV IPC API have been
514 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
515 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
519 Layout of struct vnet has changed as routing related variables were
520 moved to their own Vimage module. Modules need to be recompiled. Bump
521 __FreeBSD_version to 800099.
524 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
525 respectively. As long as no more than 16 groups per process are used,
526 no changes should be visible. When more than 16 groups are used, old
527 binaries may fail if they call getgroups() or getgrouplist() with
528 statically sized storage. Recompiling will work around this, but
529 applications should be modified to use dynamically allocated storage
530 for group arrays as POSIX.1-2008 does not cap an implementation's
531 number of supported groups at NGROUPS_MAX+1 as previous versions did.
533 NFS and portalfs mounts may also be affected as the list of groups is
534 truncated to 16. Users of NFS who use more than 16 groups, should
535 take care that negative group permissions are not used on the exported
536 file systems as they will not be reliable unless a GSSAPI based
537 authentication method is used.
540 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
541 option compiles in the support for adaptive spinning for lockmgrs
542 which want to enable it. The lockinit() function now accepts the flag
543 LK_ADAPTIVE in order to make the lock object subject to adaptive
544 spinning when both held in write and read mode.
547 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
548 changed. User applications that use this ioctl need to be rebuilt.
551 The layout of struct thread has changed. Kernel and modules need to
555 The layout of structs ifnet, domain, protosw and vnet_net has changed.
556 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
559 window(1) has been removed from the base system. It can now be
560 installed from ports. The port is called misc/window.
563 The way we are storing and accessing `routing table' entries has
564 changed. Programs reading the FIB, like netstat, need to be
568 A new netisr implementation has been added for FreeBSD 8. Network
569 file system modules, such as igmp, ipdivert, and others, should be
571 Bump __FreeBSD_version to 800096.
574 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
578 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
579 Bump __FreeBSD_version to 800094.
582 Add mnt_xflag field to 'struct mount'. File system modules need to be
584 Bump __FreeBSD_version to 800093.
587 The compiling option ADAPTIVE_SX has been retired while it has been
588 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
589 The KPI for sx_init_flags() changes as accepting flags:
590 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
591 been introduced in order to handle the reversed logic.
592 Bump __FreeBSD_version to 800092.
595 Add support for hierarchical jails. Remove global securelevel.
596 Bump __FreeBSD_version to 800091.
599 The layout of struct vnet_net has changed, therefore modules
601 Bump __FreeBSD_version to 800090.
604 The newly imported zic(8) produces a new format in the output. Please
605 run tzsetup(8) to install the newly created data to /etc/localtime.
608 The sysctl tree for the usb stack has renamed from hw.usb2.* to
609 hw.usb.* and is now consistent again with previous releases.
612 802.11 monitor mode support was revised and driver api's were changed.
613 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
614 of DLT_IEEE802_11. No user-visible data structures were changed but
615 applications that use DLT_IEEE802_11 may require changes.
616 Bump __FreeBSD_version to 800088.
619 The layout of the following structs has changed: sysctl_oid,
620 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
621 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
622 panics may be experienced. World rebuild is required for
623 correctly checking networking state from userland.
624 Bump __FreeBSD_version to 800085.
627 MLDv2 and Source-Specific Multicast (SSM) have been merged
628 to the IPv6 stack. VIMAGE hooks are in but not yet used.
629 The implementation of SSM within FreeBSD's IPv6 stack closely
630 follows the IPv4 implementation.
632 For kernel developers:
634 * The most important changes are that the ip6_output() and
635 ip6_input() paths no longer take the IN6_MULTI_LOCK,
636 and this lock has been downgraded to a non-recursive mutex.
638 * As with the changes to the IPv4 stack to support SSM, filtering
639 of inbound multicast traffic must now be performed by transport
640 protocols within the IPv6 stack. This does not apply to TCP and
641 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
643 * The KPIs used by IPv6 multicast are similar to those used by
644 the IPv4 stack, with the following differences:
645 * im6o_mc_filter() is analogous to imo_multicast_filter().
646 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
647 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
648 * IN6_LOOKUP_MULTI() has been deprecated and removed.
649 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
650 for MLDv1 have an additional 'timer' argument which is used to
651 jitter the initial membership report for the solicited-node
652 multicast membership on-link.
653 * This is not strictly needed for MLDv2, which already jitters
654 its report transmissions. However, the 'timer' argument is
655 preserved in case MLDv1 is active on the interface.
657 * The KAME linked-list based IPv6 membership implementation has
658 been refactored to use a vector similar to that used by the IPv4
660 Code which maintains a list of its own multicast memberships
661 internally, e.g. carp, has been updated to reflect the new
664 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
665 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
666 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
667 implementation constraint which needs to be addressed in HEAD.
669 For application developers:
671 * The changes are broadly similar to those made for the IPv4
674 * The use of IPv4 and IPv6 multicast socket options on the same
675 socket, using mapped addresses, HAS NOT been tested or supported.
677 * There are a number of issues with the implementation of various
678 IPv6 multicast APIs which need to be resolved in the API surface
679 before the implementation is fully compatible with KAME userland
680 use, and these are mostly to do with interface index treatment.
682 * The literature available discusses the use of either the delta / ASM
683 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
684 using setsourcefilter(3)/getsourcefilter(3). For more information
685 please refer to RFC 3768, 'Socket Interface Extensions for
686 Multicast Source Filters'.
688 * Applications which use the published RFC 3678 APIs should be fine.
690 For systems administrators:
692 * The mtest(8) utility has been refactored to support IPv6, in
693 addition to IPv4. Interface addresses are no longer accepted
694 as arguments, their names must be used instead. The utility
695 will map the interface name to its first IPv4 address as
696 returned by getifaddrs(3).
698 * The ifmcstat(8) utility has also been updated to print the MLDv2
699 endpoint state and source filter lists via sysctl(3).
701 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
702 loopback of IPv6 multicast datagrams by default; it defaults to 1
703 to preserve the existing behaviour. Disabling multicast loopback is
704 recommended for optimal system performance.
706 * The IPv6 MROUTING code has been changed to examine this sysctl
707 instead of attempting to perform a group lookup before looping
708 back forwarded datagrams.
710 Bump __FreeBSD_version to 800084.
713 Implement low-level Bluetooth HCI API.
714 Bump __FreeBSD_version to 800083.
717 The layout of struct malloc_type, used by modules to register new
718 memory allocation types, has changed. Most modules will need to
719 be rebuilt or panics may be experienced.
720 Bump __FreeBSD_version to 800081.
723 Anticipate overflowing inp_flags - add inp_flags2.
724 This changes most offsets in inpcb, so checking v4 connection
725 state will require a world rebuild.
726 Bump __FreeBSD_version to 800080.
729 Add an llentry to struct route and struct route_in6. Modules
730 embedding a struct route will need to be recompiled.
731 Bump __FreeBSD_version to 800079.
734 The size of rt_metrics_lite and by extension rtentry has changed.
735 Networking administration apps will need to be recompiled.
736 The route command now supports show as an alias for get, weighting
737 of routes, sticky and nostick flags to alter the behavior of stateful
739 Bump __FreeBSD_version to 800078.
742 Do not use Giant for kbdmux(4) locking. This is wrong and
743 apparently causing more problems than it solves. This will
744 re-open the issue where interrupt handlers may race with
745 kbdmux(4) in polling mode. Typical symptoms include (but
746 not limited to) duplicated and/or missing characters when
747 low level console functions (such as gets) are used while
748 interrupts are enabled (for example geli password prompt,
749 mountroot prompt etc.). Disabling kbdmux(4) may help.
752 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
753 kernel modules referencing any of the above need to be recompiled.
754 Bump __FreeBSD_version to 800075.
757 GEOM_PART has become the default partition slicer for storage devices,
758 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
759 introduces some changes:
761 MSDOS/EBR: the devices created from MSDOS extended partition entries
762 (EBR) can be named differently than with GEOM_MBR and are now symlinks
763 to devices with offset-based names. fstabs may need to be modified.
765 BSD: the "geometry does not match label" warning is harmless in most
766 cases but it points to problems in file system misalignment with
767 disk geometry. The "c" partition is now implicit, covers the whole
768 top-level drive and cannot be (mis)used by users.
770 General: Kernel dumps are now not allowed to be written to devices
771 whose partition types indicate they are meant to be used for file
772 systems (or, in case of MSDOS partitions, as something else than
775 Most of these changes date approximately from 200812.
778 The uscanner(4) driver has been removed from the kernel. This follows
779 Linux removing theirs in 2.6 and making libusb the default interface
783 The multicast forwarding code has been cleaned up. netstat(1)
784 only relies on KVM now for printing bandwidth upcall meters.
785 The IPv4 and IPv6 modules are split into ip_mroute_mod and
786 ip6_mroute_mod respectively. The config(5) options for statically
787 compiling this code remain the same, i.e. 'options MROUTING'.
790 Support for the IFF_NEEDSGIANT network interface flag has been
791 removed, which means that non-MPSAFE network device drivers are no
792 longer supported. In particular, if_ar, if_sr, and network device
793 drivers from the old (legacy) USB stack can no longer be built or
797 POSIX.1 Native Language Support (NLS) has been enabled in libc and
798 a bunch of new language catalog files have also been added.
799 This means that some common libc messages are now localized and
800 they depend on the LC_MESSAGES environmental variable.
803 The k8temp(4) driver has been renamed to amdtemp(4) since
804 support for Family 10 and Family 11 CPU families was added.
807 IGMPv3 and Source-Specific Multicast (SSM) have been merged
808 to the IPv4 stack. VIMAGE hooks are in but not yet used.
810 For kernel developers, the most important changes are that the
811 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
812 and this lock has been downgraded to a non-recursive mutex.
814 Transport protocols (UDP, Raw IP) are now responsible for filtering
815 inbound multicast traffic according to group membership and source
816 filters. The imo_multicast_filter() KPI exists for this purpose.
817 Transports which do not use multicast (SCTP, TCP) already reject
818 multicast by default. Forwarding and receive performance may improve
819 as a mutex acquisition is no longer needed in the ip_input()
820 low-level input path. in_addmulti() and in_delmulti() are shimmed
821 to new KPIs which exist to support SSM in-kernel.
823 For application developers, it is recommended that loopback of
824 multicast datagrams be disabled for best performance, as this
825 will still cause the lock to be taken for each looped-back
826 datagram transmission. The net.inet.ip.mcast.loop sysctl may
827 be tuned to 0 to disable loopback by default; it defaults to 1
828 to preserve the existing behaviour.
830 For systems administrators, to obtain best performance with
831 multicast reception and multiple groups, it is always recommended
832 that a card with a suitably precise hash filter is used. Hash
833 collisions will still result in the lock being taken within the
834 transport protocol input path to check group membership.
836 If deploying FreeBSD in an environment with IGMP snooping switches,
837 it is recommended that the net.inet.igmp.sendlocal sysctl remain
838 enabled; this forces 224.0.0.0/24 group membership to be announced
841 The size of 'struct igmpstat' has changed; netstat needs to be
842 recompiled to reflect this.
843 Bump __FreeBSD_version to 800070.
846 libusb20.so.1 is now installed as libusb.so.1 and the ports system
847 updated to use it. This requires a buildworld/installworld in order to
848 update the library and dependencies (usbconfig, etc). Its advisable to
849 rebuild all ports which uses libusb. More specific directions are given
850 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
851 libusb are no longer required and can be removed.
854 A workaround is committed to allow the creation of System V shared
855 memory segment of size > 2 GB on the 64-bit architectures.
856 Due to a limitation of the existing ABI, the shm_segsz member
857 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
858 wrong for large segments. Note that limits must be explicitly
859 raised to allow such segments to be created.
862 The layout of struct ifnet has changed, requiring a rebuild of all
863 network device driver modules.
866 The /dev handling for the new USB stack has changed, a
867 buildworld/installworld is required for libusb20.
870 The new USB2 stack has now been permanently moved in and all kernel and
871 module names reverted to their previous values (eg, usb, ehci, ohci,
872 ums, ...). The old usb stack can be compiled in by prefixing the name
873 with the letter 'o', the old usb modules have been removed.
874 Updating entry 20090216 for xorg and 20090215 for libmap may still
878 The rc.conf(5) option if_up_delay has been renamed to
879 defaultroute_delay to better reflect its purpose. If you have
880 customized this setting in /etc/rc.conf you need to update it to
884 xorg 7.4 wants to configure its input devices via hald which does not
885 yet work with USB2. If the keyboard/mouse does not work in xorg then
887 Option "AllowEmptyInput" "off"
888 to your ServerLayout section. This will cause X to use the configured
889 kbd and mouse sections from your xorg.conf.
892 The GENERIC kernels for all architectures now default to the new USB2
893 stack. No kernel config options or code have been removed so if a
894 problem arises please report it and optionally revert to the old USB
895 stack. If you are loading USB kernel modules or have a custom kernel
896 that includes GENERIC then ensure that usb names are also changed over,
897 eg uftdi -> usb2_serial_ftdi.
899 Older programs linked against the ports libusb 0.1 need to be
900 redirected to the new stack's libusb20. /etc/libmap.conf can
902 # Map old usb library to new one for usb2 stack
903 libusb-0.1.so.8 libusb20.so.1
906 All USB ethernet devices now attach as interfaces under the name ueN
907 (eg. ue0). This is to provide a predictable name as vendors often
908 change usb chipsets in a product without notice.
911 The ichsmb(4) driver has been changed to require SMBus slave
912 addresses be left-justified (xxxxxxx0b) rather than right-justified.
913 All of the other SMBus controller drivers require left-justified
914 slave addresses, so this change makes all the drivers provide the
918 INET6 statistics (struct ip6stat) was updated.
919 netstat(1) needs to be recompiled.
922 NTFS has been removed from GENERIC kernel on amd64 to match
923 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
924 will load ntfs.ko module automatically when NTFS support is
925 actually needed, unless ntfs.ko is not installed or security
926 level prohibits loading kernel modules. If either is the case,
927 "options NTFS" has to be added into kernel config.
930 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
931 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
932 800061. User space tools that rely on the size of struct tcpcb in
933 tcp_var.h (e.g. sockstat) need to be recompiled.
936 ng_tty(4) module updated to match the new TTY subsystem.
937 Due to API change, user-level applications must be updated.
938 New API support added to mpd5 CVS and expected to be present
939 in next mpd5.3 release.
942 With __FreeBSD_version 800060 the makefs tool is part of
943 the base system (it was a port).
946 The afdata and ifnet locks have been changed from mutexes to
947 rwlocks, network modules will need to be re-compiled.
950 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
951 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
952 The new code reduced struct rtentry{} by 16 bytes on 32-bit
953 architecture and 40 bytes on 64-bit architecture. The userland
954 applications "arp" and "ndp" have been updated accordingly.
955 The output from "netstat -r" shows only routing entries and
956 none of the L2 information.
959 __FreeBSD_version 800057 marks the switchover from the
960 binary ath hal to source code. Users must add the line:
962 options AH_SUPPORT_AR5416
964 to their kernel config files when specifying:
968 The ath_hal module no longer exists; the code is now compiled
969 together with the driver in the ath module. It is now
970 possible to tailor chip support (i.e. reduce the set of chips
971 and thereby the code size); consult ath_hal(4) for details.
974 __FreeBSD_version 800054 adds memory barriers to
975 <machine/atomic.h>, new interfaces to ifnet to facilitate
976 multiple hardware transmit queues for cards that support
977 them, and a lock-less ring-buffer implementation to
978 enable drivers to more efficiently manage queueing of
982 A new version of ZFS (version 13) has been merged to -HEAD.
983 This version has zpool attribute "listsnapshots" off by
984 default, which means "zfs list" does not show snapshots,
985 and is the same as Solaris behavior.
988 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
991 The uhci, ohci, ehci and slhci USB Host controller drivers have
992 been put into separate modules. If you load the usb module
993 separately through loader.conf you will need to load the
994 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
995 controller add the following to loader.conf:
1001 The ABI used by the PMC toolset has changed. Please keep
1002 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1006 atapci kernel module now includes only generic PCI ATA
1007 driver. AHCI driver moved to ataahci kernel module.
1008 All vendor-specific code moved into separate kernel modules:
1009 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1010 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1011 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1012 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1015 The TTY subsystem of the kernel has been replaced by a new
1016 implementation, which provides better scalability and an
1017 improved driver model. Most common drivers have been migrated to
1018 the new TTY subsystem, while others have not. The following
1019 drivers have not yet been ported to the new TTY layer:
1022 cy, digi, rc, rp, sio
1028 ng_h4, ng_tty, ppp, sl, snp
1030 Adding these drivers to your kernel configuration file shall
1031 cause compilation to fail.
1034 ntpd has been upgraded to 4.2.4p5.
1037 OpenSSH has been upgraded to 5.1p1.
1039 For many years, FreeBSD's version of OpenSSH preferred DSA
1040 over RSA for host and user authentication keys. With this
1041 upgrade, we've switched to the vendor's default of RSA over
1042 DSA. This may cause upgraded clients to warn about unknown
1043 host keys even for previously known hosts. Users should
1044 follow the usual procedure for verifying host keys before
1045 accepting the RSA key.
1047 This can be circumvented by setting the "HostKeyAlgorithms"
1048 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1051 Please note that the sequence of keys offered for
1052 authentication has been changed as well. You may want to
1053 specify IdentityFile in a different order to revert this
1057 The sio(4) driver has been removed from the i386 and amd64
1058 kernel configuration files. This means uart(4) is now the
1059 default serial port driver on those platforms as well.
1061 To prevent collisions with the sio(4) driver, the uart(4) driver
1062 uses different names for its device nodes. This means the
1063 onboard serial port will now most likely be called "ttyu0"
1064 instead of "ttyd0". You may need to reconfigure applications to
1065 use the new device names.
1067 When using the serial port as a boot console, be sure to update
1068 /boot/device.hints and /etc/ttys before booting the new kernel.
1069 If you forget to do so, you can still manually specify the hints
1070 at the loader prompt:
1072 set hint.uart.0.at="isa"
1073 set hint.uart.0.port="0x3F8"
1074 set hint.uart.0.flags="0x10"
1075 set hint.uart.0.irq="4"
1079 The gpt(8) utility has been removed. Use gpart(8) to partition
1083 The version that Linuxulator emulates was changed from 2.4.2
1084 to 2.6.16. If you experience any problems with Linux binaries
1085 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1086 if it fixes the problem contact emulation mailing list.
1089 ISDN4BSD (I4B) was removed from the src tree. You may need to
1090 update a your kernel configuration and remove relevant entries.
1093 I have checked in code to support multiple routing tables.
1094 See the man pages setfib(1) and setfib(2).
1095 This is a hopefully backwards compatible version,
1096 but to make use of it you need to compile your kernel
1097 with options ROUTETABLES=2 (or more up to 16).
1100 The 802.11 wireless support was redone to enable multi-bss
1101 operation on devices that are capable. The underlying device
1102 is no longer used directly but instead wlanX devices are
1103 cloned with ifconfig. This requires changes to rc.conf files.
1104 For example, change:
1105 ifconfig_ath0="WPA DHCP"
1108 ifconfig_wlan0="WPA DHCP"
1109 see rc.conf(5) for more details. In addition, mergemaster of
1110 /etc/rc.d is highly recommended. Simultaneous update of userland
1111 and kernel wouldn't hurt either.
1113 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1114 modules were merged into the base wlan module. All references
1115 to these modules (e.g. in kernel config files) must be removed.
1118 psm(4) has gained write(2) support in native operation level.
1119 Arbitrary commands can be written to /dev/psm%d and status can
1120 be read back from it. Therefore, an application is responsible
1121 for status validation and error recovery. It is a no-op in
1122 other operation levels.
1125 Support for KSE threading has been removed from the kernel. To
1126 run legacy applications linked against KSE libmap.conf may
1127 be used. The following libmap.conf may be used to ensure
1128 compatibility with any prior release:
1130 libpthread.so.1 libthr.so.1
1131 libpthread.so.2 libthr.so.2
1132 libkse.so.3 libthr.so.3
1135 The layout of struct vmspace has changed. This affects libkvm
1136 and any executables that link against libkvm and use the
1137 kvm_getprocs() function. In particular, but not exclusively,
1138 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1139 The effects are minimal, but it's advisable to upgrade world
1143 The latest em driver no longer has support in it for the
1144 82575 adapter, this is now moved to the igb driver. The
1145 split was done to make new features that are incompatible
1146 with older hardware easier to do.
1149 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1150 likewise the kernel option is now GEOM_LINUX_LVM.
1153 The default NFS mount mode has changed from UDP to TCP for
1154 increased reliability. If you rely on (insecurely) NFS
1155 mounting across a firewall you may need to update your
1159 Belatedly note the addition of m_collapse for compacting
1163 The fts(3) structures have been changed to use adequate
1164 integer types for their members and so to be able to cope
1165 with huge file trees. The old fts(3) ABI is preserved
1166 through symbol versioning in libc, so third-party binaries
1167 using fts(3) should still work, although they will not take
1168 advantage of the extended types. At the same time, some
1169 third-party software might fail to build after this change
1170 due to unportable assumptions made in its source code about
1171 fts(3) structure members. Such software should be fixed
1172 by its vendor or, in the worst case, in the ports tree.
1173 FreeBSD_version 800015 marks this change for the unlikely
1174 case that a portable fix is impossible.
1177 To upgrade to -current after this date, you must be running
1178 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1179 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1182 The ADAPTIVE_GIANT kernel option has been retired because its
1183 functionality is the default now.
1186 The AT keyboard emulation of sunkbd(4) has been turned on
1187 by default. In order to make the special symbols of the Sun
1188 keyboards driven by sunkbd(4) work under X these now have
1189 to be configured the same way as Sun USB keyboards driven
1190 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1192 Option "XkbLayout" "us"
1193 Option "XkbRules" "xorg"
1194 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1197 It has been decided that it is desirable to provide ABI
1198 backwards compatibility to the FreeBSD 4/5/6 versions of the
1199 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1200 broken with the introduction of PCI domain support (see the
1201 20070930 entry). Unfortunately, this required the ABI of
1202 PCIOCGETCONF to be broken again in order to be able to
1203 provide backwards compatibility to the old version of that
1204 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1205 again. As for prominent ports this affects neither pciutils
1206 nor xorg-server this time, the hal port needs to be rebuilt
1210 The misnamed kthread_create() and friends have been renamed
1211 to kproc_create() etc. Many of the callers already
1212 used kproc_start()..
1213 I will return kthread_create() and friends in a while
1214 with implementations that actually create threads, not procs.
1215 Renaming corresponds with version 800002.
1224 Avoid using make -j when upgrading. While generally safe, there are
1225 sometimes problems using -j to upgrade. If your upgrade fails with
1226 -j, please try again without -j. From time to time in the past there
1227 have been problems using -j with buildworld and/or installworld. This
1228 is especially true when upgrading between "distant" versions (eg one
1229 that cross a major release boundary or several minor releases, or when
1230 several months have passed on the -current branch).
1232 Sometimes, obscure build problems are the result of environment
1233 poisoning. This can happen because the make utility reads its
1234 environment when searching for values for global variables. To run
1235 your build attempts in an "environmental clean room", prefix all make
1236 commands with 'env -i '. See the env(1) manual page for more details.
1238 When upgrading from one major version to another it is generally best
1239 to upgrade to the latest code in the currently installed branch first,
1240 then do an upgrade to the new branch. This is the best-tested upgrade
1241 path, and has the highest probability of being successful. Please try
1242 this approach before reporting problems with a major version upgrade.
1246 When upgrading the boot ZFS pool to a new version, always follow
1249 1.) recompile and reinstall the ZFS boot loader and boot block
1250 (this is part of "make buildworld" and "make installworld")
1252 2.) update the ZFS boot block on your boot drive
1254 The following example updates the ZFS boot block on the first
1255 partition (freebsd-boot) of a GPT partitioned drive ad0:
1256 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1258 Non-boot pools do not need these updates.
1262 If you are updating from a prior version of FreeBSD (even one just
1263 a few days old), you should follow this procedure. It is the most
1264 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1266 make kernel-toolchain
1267 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1268 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1270 To test a kernel once
1271 ---------------------
1272 If you just want to boot a kernel once (because you are not sure
1273 if it works, or if you want to boot a known bad kernel to provide
1274 debugging information) run
1275 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1276 nextboot -k testkernel
1278 To just build a kernel when you know that it won't mess you up
1279 --------------------------------------------------------------
1280 This assumes you are already running a CURRENT system. Replace
1281 ${arch} with the architecture of your machine (e.g. "i386",
1282 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1284 cd src/sys/${arch}/conf
1285 config KERNEL_NAME_HERE
1286 cd ../compile/KERNEL_NAME_HERE
1291 If this fails, go to the "To build a kernel" section.
1293 To rebuild everything and install it on the current system.
1294 -----------------------------------------------------------
1295 # Note: sometimes if you are running current you gotta do more than
1296 # is listed here if you are upgrading from a really old current.
1298 <make sure you have good level 0 dumps>
1300 make kernel KERNCONF=YOUR_KERNEL_HERE
1302 <reboot in single user> [3]
1310 To cross-install current onto a separate partition
1311 --------------------------------------------------
1312 # In this approach we use a separate partition to hold
1313 # current's root, 'usr', and 'var' directories. A partition
1314 # holding "/", "/usr" and "/var" should be about 2GB in
1317 <make sure you have good level 0 dumps>
1320 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1321 <maybe newfs current's root partition>
1322 <mount current's root partition on directory ${CURRENT_ROOT}>
1323 make installworld DESTDIR=${CURRENT_ROOT}
1324 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1325 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1326 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1327 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1328 <reboot into current>
1329 <do a "native" rebuild/install as described in the previous section>
1330 <maybe install compatibility libraries from ports/misc/compat*>
1334 To upgrade in-place from 8.x-stable to current
1335 ----------------------------------------------
1336 <make sure you have good level 0 dumps>
1338 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1340 <reboot in single user> [3]
1347 Make sure that you've read the UPDATING file to understand the
1348 tweaks to various things you need. At this point in the life
1349 cycle of current, things change often and you are on your own
1350 to cope. The defaults can also change, so please read ALL of
1351 the UPDATING entries.
1353 Also, if you are tracking -current, you must be subscribed to
1354 freebsd-current@freebsd.org. Make sure that before you update
1355 your sources that you have read and understood all the recent
1356 messages there. If in doubt, please track -stable which has
1357 much fewer pitfalls.
1359 [1] If you have third party modules, such as vmware, you
1360 should disable them at this point so they don't crash your
1363 [3] From the bootblocks, boot -s, and then do
1368 adjkerntz -i # if CMOS is wall time
1369 Also, when doing a major release upgrade, it is required that
1370 you boot into single user mode to do the installworld.
1372 [4] Note: This step is non-optional. Failure to do this step
1373 can result in a significant reduction in the functionality of the
1374 system. Attempting to do it by hand is not recommended and those
1375 that pursue this avenue should read this file carefully, as well
1376 as the archives of freebsd-current and freebsd-hackers mailing lists
1377 for potential gotchas. The -U option is also useful to consider.
1378 See mergemaster(8) for more information.
1380 [5] Usually this step is a noop. However, from time to time
1381 you may need to do this if you get unknown user in the following
1382 step. It never hurts to do it all the time. You may need to
1383 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1384 install) after the buildworld before this step if you last updated
1385 from current before 20020224 or from -stable before 20020408.
1387 [6] This only deletes old files and directories. Old libraries
1388 can be deleted by "make delete-old-libs", but you have to make
1389 sure that no program is using those libraries anymore.
1391 [8] In order to have a kernel that can run the 4.x binaries needed to
1392 do an installworld, you must include the COMPAT_FREEBSD4 option in
1393 your kernel. Failure to do so may leave you with a system that is
1394 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1395 required to run the 5.x binaries on more recent kernels. And so on
1396 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1398 Make sure that you merge any new devices from GENERIC since the
1399 last time you updated your kernel config file.
1401 [9] When checking out sources, you must include the -P flag to have
1402 cvs prune empty directories.
1404 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1405 "?=" instead of the "=" assignment operator, so that buildworld can
1406 override the CPUTYPE if it needs to.
1408 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1409 not on the command line, or in /etc/make.conf. buildworld will
1410 warn if it is improperly defined.
1413 This file contains a list, in reverse chronological order, of major
1414 breakages in tracking -current. Not all things will be listed here,
1415 and it only starts on October 16, 2004. Updating files can found in
1416 previous releases if your system is older than this.
1418 Copyright information:
1420 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1422 Redistribution, publication, translation and use, with or without
1423 modification, in full or in part, in any form or format of this
1424 document are permitted without further permission from the author.
1426 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1427 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1428 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1429 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1430 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1431 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1432 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1433 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1434 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1435 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1436 POSSIBILITY OF SUCH DAMAGE.
1438 Contact Warner Losh if you have any questions about your use of