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 The following sysctls and tunables are retired on x86 platforms:
28 machdep.hlt_logical_cpus
29 The following sysctl is retired:
30 machdep.hyperthreading_allowed
31 The sysctls were supposed to provide a way to dynamically offline and
32 online selected CPUs on x86 platforms, but the implementation has not
33 been reliable especially with SCHED_ULE scheduler.
34 machdep.hyperthreading_allowed tunable is still available to ignore
35 hyperthreading CPUs at OS level.
36 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
37 where X is an APIC ID of a CPU. Be advised, though, that disabling
38 CPUs in non-uniform fashion will result in non-uniform topology and
39 may lead to sub-optimal system performance with SCHED_ULE, which is
43 cpumask_t type is retired and cpuset_t is used in order to describe
47 Changes to ifconfig(8) for dynamic address family detection mandate
48 that you are running a kernel of 20110525 or later. Make sure to
49 follow the update procedure to boot a new kernel before installing
53 Support for sun4v architecture is officially dropped
56 Several KPI breaking changes have been committed to the mii(4) layer,
57 the PHY drivers and consequently some Ethernet drivers using mii(4).
58 This means that miibus.ko and the modules of the affected Ethernet
59 drivers need to be recompiled.
61 Note to kernel developers: Given that the OUI bit reversion problem
62 was fixed as part of these changes all mii(4) commits related to OUIs,
63 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
64 handling, no longer can be merged verbatim to stable/8 and previous
68 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
69 into their kernel configurations along with 'device pci'.
72 The default NFS client is now the new NFS client, so fstype "newnfs"
73 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
74 Although mounts via fstype "nfs" will usually work without userland
75 changes, it is recommended that the mount(8) and mount_nfs(8)
76 commands be rebuilt from sources and that a link to mount_nfs called
77 mount_oldnfs be created. The new client is compiled into the
78 kernel with "options NFSCL" and this is needed for diskless root
79 file systems. The GENERIC kernel configs have been changed to use
80 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
81 To use the regular/old client, you can "mount -t oldnfs ...". For
82 a diskless root file system, you must also include a line like:
84 vfs.root.mountfrom="oldnfs:"
86 in the boot/loader.conf on the root fs on the NFS server to make
87 a diskless root fs use the old client.
90 The GENERIC kernels for all architectures now default to the new
91 CAM-based ATA stack. It means that all legacy ATA drivers were
92 removed and replaced by respective CAM drivers. If you are using
93 ATA device names in /etc/fstab or other places, make sure to update
94 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
95 where 'Y's are the sequential numbers starting from zero for each type
96 in order of detection, unless configured otherwise with tunables,
97 see cam(4)). There will be symbolic links created in /dev/ to map
98 old adX devices to the respective adaY. They should provide basic
99 compatibility for file systems mounting in most cases, but they do
100 not support old user-level APIs and do not have respective providers
101 in GEOM. Consider using updated management tools with new device names.
103 It is possible to load devices ahci, ata, siis and mvs as modules,
104 but option ATA_CAM should remain in kernel configuration to make ata
105 module work as CAM driver supporting legacy ATA controllers. Device ata
106 still can be used in modular fashion (atacore + ...). Modules atadisk
107 and atapi* are not used and won't affect operation in ATA_CAM mode.
108 Note that to use CAM-based ATA kernel should include CAM devices
109 scbus, pass, da (or explicitly ada), cd and optionally others. All of
110 them are parts of the cam module.
112 ataraid(4) functionality is now supported by the RAID GEOM class.
113 To use it you can load geom_raid kernel module and use graid(8) tool
114 for management. Instead of /dev/arX device names, use /dev/raid/rX.
116 No kernel config options or code have been removed, so if a problem
117 arises, please report it and optionally revert to the old ATA stack.
118 In order to do it you can remove from the kernel config:
123 , and instead add back:
124 device atadisk # ATA disk drives
125 device ataraid # ATA RAID drives
126 device atapicd # ATAPI CDROM drives
127 device atapifd # ATAPI floppy drives
128 device atapist # ATAPI tape drives
131 The default NFS server has been changed to the new server, which
132 was referred to as the experimental server. If you need to switch
133 back to the old NFS server, you must now put the "-o" option on
134 both the mountd and nfsd commands. This can be done using the
135 mountd_flags and nfs_server_flags rc.conf variables until an
136 update to the rc scripts is committed, which is coming soon.
139 The GNU Objective-C runtime library (libobjc), and other Objective-C
140 related components have been removed from the base system. If you
141 require an Objective-C library, please use one of the available ports.
144 ath(4) has been split into bus- and device- modules. if_ath contains
145 the HAL, the TX rate control and the network device code. if_ath_pci
146 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
147 contains the AHB glue. Users need to load both if_ath_pci and if_ath
148 in order to use ath on everything else.
150 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
151 need to load if_ath and if_ath_pci for ath(4) operation.
154 As part of the replacement of sysinstall, the process of building
155 release media has changed significantly. For details, please re-read
156 release(7), which has been updated to reflect the new build process.
159 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
160 is the last available version under GPLv2. It brings a number of new
161 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
162 4.1 and SSE 4.2), better support for powerpc64, a number of new
163 directives, and lots of other small improvements. See the ChangeLog
164 file in contrib/binutils for the full details.
167 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
168 compliant, and will now use half of hash for authentication.
169 This will break interoperability with all stacks (including all
170 actual FreeBSD versions) who implement
171 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
173 The only workaround with such peers is to use another HMAC
174 algorithm for IPsec ("phase 2") authentication.
177 Remove the uio_yield prototype and symbol. This function has
178 been misnamed since it was introduced and should not be
179 globally exposed with this name. The equivalent functionality
180 is now available using kern_yield(curthread->td_user_pri).
181 The function remains undocumented.
184 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
185 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
186 for scalar sysctls is defined but disabled. Code that needs
187 UQUAD to pass the type checking that must compile on older
188 systems where the define is not present can check against
189 __FreeBSD_version >= 900030.
191 The system dialog(1) has been replaced with a new version previously
192 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
193 with the previous version, but the libdialog associated with it has
194 a largely incompatible API. As such, the original version of libdialog
195 will be kept temporarily as libodialog, until its base system consumers
196 are replaced or updated. Bump __FreeBSD_version to 900030.
199 If you are trying to run make universe on a -stable system, and you get
200 the following warning:
201 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
202 unknown. config(8) likely too old."
203 or something similar to it, then you must upgrade your -stable system
204 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
205 stable/8) or build the config from the latest stable/8 branch and
206 install it on your system.
208 Prior to this date, building a current universe on 8-stable system from
209 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
210 error in the first kernel build phase. A new config on those old
211 systems will fix that problem for older versions of -current.
214 The TCP stack has been modified to allow Khelp modules to interact with
215 it via helper hook points and store per-connection data in the TCP
216 control block. Bump __FreeBSD_version to 900029. User space tools that
217 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
221 Generic IEEE 802.3 annex 31B full duplex flow control support has been
222 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
223 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
224 to take advantage of it instead of using custom implementations. This
225 means that these drivers now no longer unconditionally advertise
226 support for flow control but only do so if flow control is a selected
227 media option. This was implemented in the generic support that way in
228 order to allow flow control to be switched on and off via ifconfig(8)
229 with the PHY specific default to typically off in order to protect
230 from unwanted effects. Consequently, if you used flow control with
231 one of the above mentioned drivers you now need to explicitly enable
233 ifconfig bge0 media auto mediaopt flowcontrol
235 Along with the above mentioned changes generic support for setting
236 1000baseT master mode also has been added and brgphy(4), ciphy(4),
237 e1000phy(4) as well as ip1000phy(4) have been converted to take
238 advantage of it. This means that these drivers now no longer take the
239 link0 parameter for selecting master mode but the master media option
240 has to be used instead, for example like in the following:
241 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
243 Selection of master mode now is also available with all other PHY
244 drivers supporting 1000baseT.
247 The TCP stack has received a significant update to add support for
248 modularised congestion control and generally improve the clarity of
249 congestion control decisions. Bump __FreeBSD_version to 900025. User
250 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
251 sockstat) need to be recompiled.
254 The man(1) utility has been replaced by a new version that no longer
255 uses /etc/manpath.config. Please consult man.conf(5) for how to
256 migrate local entries to the new format.
259 The copyright strings printed by login(1) and sshd(8) at the time of a
260 new connection have been removed to follow other operating systems and
264 A workaround for a fixed ld bug has been removed in kernel code,
265 so make sure that your system ld is built from sources after
266 revision 210245 from 2010-07-19 (r211583 if building head kernel
267 on stable/8, r211584 for stable/7; both from 2010-08-21).
268 A symptom of incorrect ld version is different addresses for
269 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
272 The $ipv6_prefer variable in rc.conf(5) has been split into
273 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
275 The $ip6addrctl_policy is a variable to choose a pre-defined
276 address selection policy set by ip6addrctl(8). A value
277 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
280 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
281 flag (see an entry of 20090926) is set on an interface with no
282 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
283 security reason. If you want IPv6 link-local address on all
284 interfaces by default, set this to "YES".
286 The old ipv6_prefer="YES" is equivalent to
287 ipv6_activate_all_interfaces="YES" and
288 ip6addrctl_policy="ipv6_prefer".
291 DTrace has grown support for userland tracing. Due to this, DTrace is
292 now i386 and amd64 only.
293 dtruss(1) is now installed by default on those systems and a new
294 kernel module is needed for userland tracing: fasttrap.
295 No changes to your kernel config file are necessary to enable
296 userland tracing, but you might consider adding 'STRIP=' and
297 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
298 to have informative userland stack traces in DTrace (ustack).
301 The acpi_aiboost(4) driver has been removed in favor of the new
302 aibs(4) driver. You should update your kernel configuration file.
305 BSD grep has been imported to the base system and it is built by
306 default. It is completely BSD licensed, highly GNU-compatible, uses
307 less memory than its GNU counterpart and has a small codebase.
308 However, it is slower than its GNU counterpart, which is mostly
309 noticeable for larger searches, for smaller ones it is measurable
310 but not significant. The reason is complex, the most important factor
311 is that we lack a modern and efficient regex library and GNU
312 overcomes this by optimizing the searches internally. Future work
313 on improving the regex performance is planned, for the meantime,
314 users that need better performance, can build GNU grep instead by
315 setting the WITH_GNU_GREP knob.
318 Due to the import of powerpc64 support, all existing powerpc kernel
319 configuration files must be updated with a machine directive like this:
320 machine powerpc powerpc
322 In addition, an updated config(8) is required to build powerpc kernels
326 A new version of ZFS (version 15) has been merged to -HEAD.
327 This version uses a python library for the following subcommands:
328 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
329 For full functionality of these commands the following port must
330 be installed: sysutils/py-zfs
333 'vm_page's are now hashed by physical address to an array of mutexes.
334 Currently this is only used to serialize access to hold_count. Over
335 time the page queue mutex will be peeled away. This changes the size
336 of pmap on every architecture. And requires all callers of vm_page_hold
337 and vm_page_unhold to be updated.
340 WITH_CTF can now be specified in src.conf (not recommended, there
341 are some problems with static executables), make.conf (would also
342 affect ports which do not use GNU make and do not override the
343 compile targets) or in the kernel config (via "makeoptions
345 When WITH_CTF was specified there before this was silently ignored,
346 so make sure that WITH_CTF is not used in places which could lead
347 to unwanted behavior.
350 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
351 to allow 32-bit compatibility on non-x86 platforms. All kernel
352 configurations on amd64 and ia64 platforms using these options must
353 be modified accordingly.
356 The utmp user accounting database has been replaced with utmpx,
357 the user accounting interface standardized by POSIX.
358 Unfortunately the semantics of utmp and utmpx don't match,
359 making it practically impossible to support both interfaces.
360 The user accounting database is used by tools like finger(1),
361 last(1), talk(1), w(1) and ac(8).
363 All applications in the base system use utmpx. This means only
364 local binaries (e.g. from the ports tree) may still use these
365 utmp database files. These applications must be rebuilt to make
368 After the system has been upgraded, it is safe to remove the old
369 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
370 assuming their contents is of no importance anymore. Old wtmp
371 databases can only be used by last(1) and ac(8) after they have
372 been converted to the new format using wtmpcvt(1).
375 Introduce the kernel thread "deadlock resolver" (which can be enabled
376 via the DEADLKRES option, see NOTES for more details) and the
377 sleepq_type() function for sleepqueues.
380 The rc.firewall and rc.firewall6 were unified, and
381 rc.firewall6 and rc.d/ip6fw were removed.
382 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
383 variables are obsoleted. Instead, the following new rc
384 variables are added to rc.d/ipfw:
386 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
387 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
388 firewall_simple_onet_ipv6, firewall_trusted_ipv6
390 The meanings correspond to the relevant IPv4 variables.
396 The default terminal emulation for syscons(4) has been changed
397 from cons25 to xterm on all platforms except pc98. This means
398 that the /etc/ttys file needs to be updated to ensure correct
399 operation of applications on the console.
401 The terminal emulation style can be toggled per window by using
402 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
403 options can be used to change the compile-time default back to
406 To prevent graphical artifacts, make sure the TERM environment
407 variable is set to match the terminal emulation that is being
408 performed by syscons(4).
411 The layout of the structure ieee80211req_scan_result has changed.
412 Applications that require wireless scan results (e.g. ifconfig(8))
413 from net80211 need to be recompiled.
415 Applications such as wpa_supplicant(8) may require a full world
416 build without using NO_CLEAN in order to get synchronized with the
420 The iwn(4) driver has been updated to support the 5000 and 5150 series.
421 There's one kernel module for each firmware. Adding "device iwnfw"
422 to the kernel configuration file means including all three firmware
423 images inside the kernel. If you want to include just the one for
424 your wireless card, use the the devices iwn4965fw, iwn5000fw or
428 The rc.d/network_ipv6, IPv6 configuration script has been integrated
429 into rc.d/netif. The changes are the following:
431 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
432 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
433 Note that both variables need the "inet6" keyword at the head.
435 Do not set $ipv6_network_interfaces manually if you do not
436 understand what you are doing. It is not needed in most cases.
438 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
441 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
442 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
444 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
445 all configured interfaces have "inet6 accept_rtadv" in the
446 $ifconfig_IF_ipv6. These are for backward compatibility.
448 3. A new variable $ipv6_prefer has been added. If NO, IPv6
449 functionality of interfaces with no corresponding
450 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
451 and the default address selection policy of ip6addrctl(8)
452 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
453 Note that if you want to configure IPv6 functionality on the
454 disabled interfaces after boot, first you need to clear the flag by
455 using ifconfig(8) like:
457 ifconfig em0 inet6 -ifdisabled
459 If YES, the default address selection policy is set as
462 The default value of $ipv6_prefer is NO.
464 4. If your system need to receive Router Advertisement messages,
465 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
466 scripts automatically invoke rtsol(8) when the interface becomes
467 UP. The Router Advertisement messages are used for SLAAC
468 (State-Less Address AutoConfiguration).
471 802.11s D3.03 support was committed. This is incompatible with the
472 previous code, which was based on D3.0.
475 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
476 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
477 control whether accepting Router Advertisement messages or not.
478 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
479 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
480 The ifconfig(8) utility now supports these flags.
483 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
484 mount(8) and -a option for df(1) to see them.
487 The old tunable hw.bus.devctl_disable has been superseded by
488 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
489 replaced by hw.bus.devctl_queue=0. The default for this new tunable
493 Remove the option STOP_NMI. The default action is now to use NMI only
494 for KDB via the newly introduced function stop_cpus_hard() and
495 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
498 The stable/8 branch created in subversion. This corresponds to the
499 RELENG_8 branch in CVS.
502 Bump the shared library version numbers for all libraries that do not
503 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
504 __FreeBSD_version to 800105.
507 Due to changes in the implementation of virtual network stack support,
508 all network-related kernel modules must be recompiled. As this change
509 breaks the ABI, bump __FreeBSD_version to 800104.
512 The TOE interface to the TCP syncache has been modified to remove
513 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
514 The cxgb driver is the only TOE consumer affected by this change, and
515 needs to be recompiled along with the kernel. As this change breaks
516 the ABI, bump __FreeBSD_version to 800103.
519 Padding has been added to struct tcpcb, sackhint and tcpstat in
520 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
521 maintaining the ABI. However, this change breaks the ABI, so bump
522 __FreeBSD_version to 800102. User space tools that rely on the size of
523 any of these structs (e.g. sockstat) need to be recompiled.
526 The NFS_LEGACYRPC option has been removed along with the old kernel
527 RPC implementation that this option selected. Kernel configurations
528 may need to be adjusted.
531 The network interface device nodes at /dev/net/<interface> have been
532 removed. All ioctl operations can be performed the normal way using
533 routing sockets. The kqueue functionality can generally be replaced
534 with routing sockets.
537 The documentation from the FreeBSD Documentation Project (Handbook,
538 FAQ, etc.) is now installed via packages by sysinstall(8) and under
539 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
542 The ABI of various structures related to the SYSV IPC API have been
543 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
544 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
548 Layout of struct vnet has changed as routing related variables were
549 moved to their own Vimage module. Modules need to be recompiled. Bump
550 __FreeBSD_version to 800099.
553 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
554 respectively. As long as no more than 16 groups per process are used,
555 no changes should be visible. When more than 16 groups are used, old
556 binaries may fail if they call getgroups() or getgrouplist() with
557 statically sized storage. Recompiling will work around this, but
558 applications should be modified to use dynamically allocated storage
559 for group arrays as POSIX.1-2008 does not cap an implementation's
560 number of supported groups at NGROUPS_MAX+1 as previous versions did.
562 NFS and portalfs mounts may also be affected as the list of groups is
563 truncated to 16. Users of NFS who use more than 16 groups, should
564 take care that negative group permissions are not used on the exported
565 file systems as they will not be reliable unless a GSSAPI based
566 authentication method is used.
569 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
570 option compiles in the support for adaptive spinning for lockmgrs
571 which want to enable it. The lockinit() function now accepts the flag
572 LK_ADAPTIVE in order to make the lock object subject to adaptive
573 spinning when both held in write and read mode.
576 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
577 changed. User applications that use this ioctl need to be rebuilt.
580 The layout of struct thread has changed. Kernel and modules need to
584 The layout of structs ifnet, domain, protosw and vnet_net has changed.
585 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
588 window(1) has been removed from the base system. It can now be
589 installed from ports. The port is called misc/window.
592 The way we are storing and accessing `routing table' entries has
593 changed. Programs reading the FIB, like netstat, need to be
597 A new netisr implementation has been added for FreeBSD 8. Network
598 file system modules, such as igmp, ipdivert, and others, should be
600 Bump __FreeBSD_version to 800096.
603 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
607 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
608 Bump __FreeBSD_version to 800094.
611 Add mnt_xflag field to 'struct mount'. File system modules need to be
613 Bump __FreeBSD_version to 800093.
616 The compiling option ADAPTIVE_SX has been retired while it has been
617 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
618 The KPI for sx_init_flags() changes as accepting flags:
619 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
620 been introduced in order to handle the reversed logic.
621 Bump __FreeBSD_version to 800092.
624 Add support for hierarchical jails. Remove global securelevel.
625 Bump __FreeBSD_version to 800091.
628 The layout of struct vnet_net has changed, therefore modules
630 Bump __FreeBSD_version to 800090.
633 The newly imported zic(8) produces a new format in the output. Please
634 run tzsetup(8) to install the newly created data to /etc/localtime.
637 The sysctl tree for the usb stack has renamed from hw.usb2.* to
638 hw.usb.* and is now consistent again with previous releases.
641 802.11 monitor mode support was revised and driver api's were changed.
642 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
643 of DLT_IEEE802_11. No user-visible data structures were changed but
644 applications that use DLT_IEEE802_11 may require changes.
645 Bump __FreeBSD_version to 800088.
648 The layout of the following structs has changed: sysctl_oid,
649 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
650 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
651 panics may be experienced. World rebuild is required for
652 correctly checking networking state from userland.
653 Bump __FreeBSD_version to 800085.
656 MLDv2 and Source-Specific Multicast (SSM) have been merged
657 to the IPv6 stack. VIMAGE hooks are in but not yet used.
658 The implementation of SSM within FreeBSD's IPv6 stack closely
659 follows the IPv4 implementation.
661 For kernel developers:
663 * The most important changes are that the ip6_output() and
664 ip6_input() paths no longer take the IN6_MULTI_LOCK,
665 and this lock has been downgraded to a non-recursive mutex.
667 * As with the changes to the IPv4 stack to support SSM, filtering
668 of inbound multicast traffic must now be performed by transport
669 protocols within the IPv6 stack. This does not apply to TCP and
670 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
672 * The KPIs used by IPv6 multicast are similar to those used by
673 the IPv4 stack, with the following differences:
674 * im6o_mc_filter() is analogous to imo_multicast_filter().
675 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
676 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
677 * IN6_LOOKUP_MULTI() has been deprecated and removed.
678 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
679 for MLDv1 have an additional 'timer' argument which is used to
680 jitter the initial membership report for the solicited-node
681 multicast membership on-link.
682 * This is not strictly needed for MLDv2, which already jitters
683 its report transmissions. However, the 'timer' argument is
684 preserved in case MLDv1 is active on the interface.
686 * The KAME linked-list based IPv6 membership implementation has
687 been refactored to use a vector similar to that used by the IPv4
689 Code which maintains a list of its own multicast memberships
690 internally, e.g. carp, has been updated to reflect the new
693 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
694 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
695 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
696 implementation constraint which needs to be addressed in HEAD.
698 For application developers:
700 * The changes are broadly similar to those made for the IPv4
703 * The use of IPv4 and IPv6 multicast socket options on the same
704 socket, using mapped addresses, HAS NOT been tested or supported.
706 * There are a number of issues with the implementation of various
707 IPv6 multicast APIs which need to be resolved in the API surface
708 before the implementation is fully compatible with KAME userland
709 use, and these are mostly to do with interface index treatment.
711 * The literature available discusses the use of either the delta / ASM
712 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
713 using setsourcefilter(3)/getsourcefilter(3). For more information
714 please refer to RFC 3768, 'Socket Interface Extensions for
715 Multicast Source Filters'.
717 * Applications which use the published RFC 3678 APIs should be fine.
719 For systems administrators:
721 * The mtest(8) utility has been refactored to support IPv6, in
722 addition to IPv4. Interface addresses are no longer accepted
723 as arguments, their names must be used instead. The utility
724 will map the interface name to its first IPv4 address as
725 returned by getifaddrs(3).
727 * The ifmcstat(8) utility has also been updated to print the MLDv2
728 endpoint state and source filter lists via sysctl(3).
730 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
731 loopback of IPv6 multicast datagrams by default; it defaults to 1
732 to preserve the existing behaviour. Disabling multicast loopback is
733 recommended for optimal system performance.
735 * The IPv6 MROUTING code has been changed to examine this sysctl
736 instead of attempting to perform a group lookup before looping
737 back forwarded datagrams.
739 Bump __FreeBSD_version to 800084.
742 Implement low-level Bluetooth HCI API.
743 Bump __FreeBSD_version to 800083.
746 The layout of struct malloc_type, used by modules to register new
747 memory allocation types, has changed. Most modules will need to
748 be rebuilt or panics may be experienced.
749 Bump __FreeBSD_version to 800081.
752 Anticipate overflowing inp_flags - add inp_flags2.
753 This changes most offsets in inpcb, so checking v4 connection
754 state will require a world rebuild.
755 Bump __FreeBSD_version to 800080.
758 Add an llentry to struct route and struct route_in6. Modules
759 embedding a struct route will need to be recompiled.
760 Bump __FreeBSD_version to 800079.
763 The size of rt_metrics_lite and by extension rtentry has changed.
764 Networking administration apps will need to be recompiled.
765 The route command now supports show as an alias for get, weighting
766 of routes, sticky and nostick flags to alter the behavior of stateful
768 Bump __FreeBSD_version to 800078.
771 Do not use Giant for kbdmux(4) locking. This is wrong and
772 apparently causing more problems than it solves. This will
773 re-open the issue where interrupt handlers may race with
774 kbdmux(4) in polling mode. Typical symptoms include (but
775 not limited to) duplicated and/or missing characters when
776 low level console functions (such as gets) are used while
777 interrupts are enabled (for example geli password prompt,
778 mountroot prompt etc.). Disabling kbdmux(4) may help.
781 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
782 kernel modules referencing any of the above need to be recompiled.
783 Bump __FreeBSD_version to 800075.
786 GEOM_PART has become the default partition slicer for storage devices,
787 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
788 introduces some changes:
790 MSDOS/EBR: the devices created from MSDOS extended partition entries
791 (EBR) can be named differently than with GEOM_MBR and are now symlinks
792 to devices with offset-based names. fstabs may need to be modified.
794 BSD: the "geometry does not match label" warning is harmless in most
795 cases but it points to problems in file system misalignment with
796 disk geometry. The "c" partition is now implicit, covers the whole
797 top-level drive and cannot be (mis)used by users.
799 General: Kernel dumps are now not allowed to be written to devices
800 whose partition types indicate they are meant to be used for file
801 systems (or, in case of MSDOS partitions, as something else than
804 Most of these changes date approximately from 200812.
807 The uscanner(4) driver has been removed from the kernel. This follows
808 Linux removing theirs in 2.6 and making libusb the default interface
812 The multicast forwarding code has been cleaned up. netstat(1)
813 only relies on KVM now for printing bandwidth upcall meters.
814 The IPv4 and IPv6 modules are split into ip_mroute_mod and
815 ip6_mroute_mod respectively. The config(5) options for statically
816 compiling this code remain the same, i.e. 'options MROUTING'.
819 Support for the IFF_NEEDSGIANT network interface flag has been
820 removed, which means that non-MPSAFE network device drivers are no
821 longer supported. In particular, if_ar, if_sr, and network device
822 drivers from the old (legacy) USB stack can no longer be built or
826 POSIX.1 Native Language Support (NLS) has been enabled in libc and
827 a bunch of new language catalog files have also been added.
828 This means that some common libc messages are now localized and
829 they depend on the LC_MESSAGES environmental variable.
832 The k8temp(4) driver has been renamed to amdtemp(4) since
833 support for Family 10 and Family 11 CPU families was added.
836 IGMPv3 and Source-Specific Multicast (SSM) have been merged
837 to the IPv4 stack. VIMAGE hooks are in but not yet used.
839 For kernel developers, the most important changes are that the
840 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
841 and this lock has been downgraded to a non-recursive mutex.
843 Transport protocols (UDP, Raw IP) are now responsible for filtering
844 inbound multicast traffic according to group membership and source
845 filters. The imo_multicast_filter() KPI exists for this purpose.
846 Transports which do not use multicast (SCTP, TCP) already reject
847 multicast by default. Forwarding and receive performance may improve
848 as a mutex acquisition is no longer needed in the ip_input()
849 low-level input path. in_addmulti() and in_delmulti() are shimmed
850 to new KPIs which exist to support SSM in-kernel.
852 For application developers, it is recommended that loopback of
853 multicast datagrams be disabled for best performance, as this
854 will still cause the lock to be taken for each looped-back
855 datagram transmission. The net.inet.ip.mcast.loop sysctl may
856 be tuned to 0 to disable loopback by default; it defaults to 1
857 to preserve the existing behaviour.
859 For systems administrators, to obtain best performance with
860 multicast reception and multiple groups, it is always recommended
861 that a card with a suitably precise hash filter is used. Hash
862 collisions will still result in the lock being taken within the
863 transport protocol input path to check group membership.
865 If deploying FreeBSD in an environment with IGMP snooping switches,
866 it is recommended that the net.inet.igmp.sendlocal sysctl remain
867 enabled; this forces 224.0.0.0/24 group membership to be announced
870 The size of 'struct igmpstat' has changed; netstat needs to be
871 recompiled to reflect this.
872 Bump __FreeBSD_version to 800070.
875 libusb20.so.1 is now installed as libusb.so.1 and the ports system
876 updated to use it. This requires a buildworld/installworld in order to
877 update the library and dependencies (usbconfig, etc). Its advisable to
878 rebuild all ports which uses libusb. More specific directions are given
879 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
880 libusb are no longer required and can be removed.
883 A workaround is committed to allow the creation of System V shared
884 memory segment of size > 2 GB on the 64-bit architectures.
885 Due to a limitation of the existing ABI, the shm_segsz member
886 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
887 wrong for large segments. Note that limits must be explicitly
888 raised to allow such segments to be created.
891 The layout of struct ifnet has changed, requiring a rebuild of all
892 network device driver modules.
895 The /dev handling for the new USB stack has changed, a
896 buildworld/installworld is required for libusb20.
899 The new USB2 stack has now been permanently moved in and all kernel and
900 module names reverted to their previous values (eg, usb, ehci, ohci,
901 ums, ...). The old usb stack can be compiled in by prefixing the name
902 with the letter 'o', the old usb modules have been removed.
903 Updating entry 20090216 for xorg and 20090215 for libmap may still
907 The rc.conf(5) option if_up_delay has been renamed to
908 defaultroute_delay to better reflect its purpose. If you have
909 customized this setting in /etc/rc.conf you need to update it to
913 xorg 7.4 wants to configure its input devices via hald which does not
914 yet work with USB2. If the keyboard/mouse does not work in xorg then
916 Option "AllowEmptyInput" "off"
917 to your ServerLayout section. This will cause X to use the configured
918 kbd and mouse sections from your xorg.conf.
921 The GENERIC kernels for all architectures now default to the new USB2
922 stack. No kernel config options or code have been removed so if a
923 problem arises please report it and optionally revert to the old USB
924 stack. If you are loading USB kernel modules or have a custom kernel
925 that includes GENERIC then ensure that usb names are also changed over,
926 eg uftdi -> usb2_serial_ftdi.
928 Older programs linked against the ports libusb 0.1 need to be
929 redirected to the new stack's libusb20. /etc/libmap.conf can
931 # Map old usb library to new one for usb2 stack
932 libusb-0.1.so.8 libusb20.so.1
935 All USB ethernet devices now attach as interfaces under the name ueN
936 (eg. ue0). This is to provide a predictable name as vendors often
937 change usb chipsets in a product without notice.
940 The ichsmb(4) driver has been changed to require SMBus slave
941 addresses be left-justified (xxxxxxx0b) rather than right-justified.
942 All of the other SMBus controller drivers require left-justified
943 slave addresses, so this change makes all the drivers provide the
947 INET6 statistics (struct ip6stat) was updated.
948 netstat(1) needs to be recompiled.
951 NTFS has been removed from GENERIC kernel on amd64 to match
952 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
953 will load ntfs.ko module automatically when NTFS support is
954 actually needed, unless ntfs.ko is not installed or security
955 level prohibits loading kernel modules. If either is the case,
956 "options NTFS" has to be added into kernel config.
959 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
960 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
961 800061. User space tools that rely on the size of struct tcpcb in
962 tcp_var.h (e.g. sockstat) need to be recompiled.
965 ng_tty(4) module updated to match the new TTY subsystem.
966 Due to API change, user-level applications must be updated.
967 New API support added to mpd5 CVS and expected to be present
968 in next mpd5.3 release.
971 With __FreeBSD_version 800060 the makefs tool is part of
972 the base system (it was a port).
975 The afdata and ifnet locks have been changed from mutexes to
976 rwlocks, network modules will need to be re-compiled.
979 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
980 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
981 The new code reduced struct rtentry{} by 16 bytes on 32-bit
982 architecture and 40 bytes on 64-bit architecture. The userland
983 applications "arp" and "ndp" have been updated accordingly.
984 The output from "netstat -r" shows only routing entries and
985 none of the L2 information.
988 __FreeBSD_version 800057 marks the switchover from the
989 binary ath hal to source code. Users must add the line:
991 options AH_SUPPORT_AR5416
993 to their kernel config files when specifying:
997 The ath_hal module no longer exists; the code is now compiled
998 together with the driver in the ath module. It is now
999 possible to tailor chip support (i.e. reduce the set of chips
1000 and thereby the code size); consult ath_hal(4) for details.
1003 __FreeBSD_version 800054 adds memory barriers to
1004 <machine/atomic.h>, new interfaces to ifnet to facilitate
1005 multiple hardware transmit queues for cards that support
1006 them, and a lock-less ring-buffer implementation to
1007 enable drivers to more efficiently manage queueing of
1011 A new version of ZFS (version 13) has been merged to -HEAD.
1012 This version has zpool attribute "listsnapshots" off by
1013 default, which means "zfs list" does not show snapshots,
1014 and is the same as Solaris behavior.
1017 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1020 The uhci, ohci, ehci and slhci USB Host controller drivers have
1021 been put into separate modules. If you load the usb module
1022 separately through loader.conf you will need to load the
1023 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1024 controller add the following to loader.conf:
1030 The ABI used by the PMC toolset has changed. Please keep
1031 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1035 atapci kernel module now includes only generic PCI ATA
1036 driver. AHCI driver moved to ataahci kernel module.
1037 All vendor-specific code moved into separate kernel modules:
1038 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1039 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1040 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1041 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1044 The TTY subsystem of the kernel has been replaced by a new
1045 implementation, which provides better scalability and an
1046 improved driver model. Most common drivers have been migrated to
1047 the new TTY subsystem, while others have not. The following
1048 drivers have not yet been ported to the new TTY layer:
1051 cy, digi, rc, rp, sio
1057 ng_h4, ng_tty, ppp, sl, snp
1059 Adding these drivers to your kernel configuration file shall
1060 cause compilation to fail.
1063 ntpd has been upgraded to 4.2.4p5.
1066 OpenSSH has been upgraded to 5.1p1.
1068 For many years, FreeBSD's version of OpenSSH preferred DSA
1069 over RSA for host and user authentication keys. With this
1070 upgrade, we've switched to the vendor's default of RSA over
1071 DSA. This may cause upgraded clients to warn about unknown
1072 host keys even for previously known hosts. Users should
1073 follow the usual procedure for verifying host keys before
1074 accepting the RSA key.
1076 This can be circumvented by setting the "HostKeyAlgorithms"
1077 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1080 Please note that the sequence of keys offered for
1081 authentication has been changed as well. You may want to
1082 specify IdentityFile in a different order to revert this
1086 The sio(4) driver has been removed from the i386 and amd64
1087 kernel configuration files. This means uart(4) is now the
1088 default serial port driver on those platforms as well.
1090 To prevent collisions with the sio(4) driver, the uart(4) driver
1091 uses different names for its device nodes. This means the
1092 onboard serial port will now most likely be called "ttyu0"
1093 instead of "ttyd0". You may need to reconfigure applications to
1094 use the new device names.
1096 When using the serial port as a boot console, be sure to update
1097 /boot/device.hints and /etc/ttys before booting the new kernel.
1098 If you forget to do so, you can still manually specify the hints
1099 at the loader prompt:
1101 set hint.uart.0.at="isa"
1102 set hint.uart.0.port="0x3F8"
1103 set hint.uart.0.flags="0x10"
1104 set hint.uart.0.irq="4"
1108 The gpt(8) utility has been removed. Use gpart(8) to partition
1112 The version that Linuxulator emulates was changed from 2.4.2
1113 to 2.6.16. If you experience any problems with Linux binaries
1114 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1115 if it fixes the problem contact emulation mailing list.
1118 ISDN4BSD (I4B) was removed from the src tree. You may need to
1119 update a your kernel configuration and remove relevant entries.
1122 I have checked in code to support multiple routing tables.
1123 See the man pages setfib(1) and setfib(2).
1124 This is a hopefully backwards compatible version,
1125 but to make use of it you need to compile your kernel
1126 with options ROUTETABLES=2 (or more up to 16).
1129 The 802.11 wireless support was redone to enable multi-bss
1130 operation on devices that are capable. The underlying device
1131 is no longer used directly but instead wlanX devices are
1132 cloned with ifconfig. This requires changes to rc.conf files.
1133 For example, change:
1134 ifconfig_ath0="WPA DHCP"
1137 ifconfig_wlan0="WPA DHCP"
1138 see rc.conf(5) for more details. In addition, mergemaster of
1139 /etc/rc.d is highly recommended. Simultaneous update of userland
1140 and kernel wouldn't hurt either.
1142 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1143 modules were merged into the base wlan module. All references
1144 to these modules (e.g. in kernel config files) must be removed.
1147 psm(4) has gained write(2) support in native operation level.
1148 Arbitrary commands can be written to /dev/psm%d and status can
1149 be read back from it. Therefore, an application is responsible
1150 for status validation and error recovery. It is a no-op in
1151 other operation levels.
1154 Support for KSE threading has been removed from the kernel. To
1155 run legacy applications linked against KSE libmap.conf may
1156 be used. The following libmap.conf may be used to ensure
1157 compatibility with any prior release:
1159 libpthread.so.1 libthr.so.1
1160 libpthread.so.2 libthr.so.2
1161 libkse.so.3 libthr.so.3
1164 The layout of struct vmspace has changed. This affects libkvm
1165 and any executables that link against libkvm and use the
1166 kvm_getprocs() function. In particular, but not exclusively,
1167 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1168 The effects are minimal, but it's advisable to upgrade world
1172 The latest em driver no longer has support in it for the
1173 82575 adapter, this is now moved to the igb driver. The
1174 split was done to make new features that are incompatible
1175 with older hardware easier to do.
1178 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1179 likewise the kernel option is now GEOM_LINUX_LVM.
1182 The default NFS mount mode has changed from UDP to TCP for
1183 increased reliability. If you rely on (insecurely) NFS
1184 mounting across a firewall you may need to update your
1188 Belatedly note the addition of m_collapse for compacting
1192 The fts(3) structures have been changed to use adequate
1193 integer types for their members and so to be able to cope
1194 with huge file trees. The old fts(3) ABI is preserved
1195 through symbol versioning in libc, so third-party binaries
1196 using fts(3) should still work, although they will not take
1197 advantage of the extended types. At the same time, some
1198 third-party software might fail to build after this change
1199 due to unportable assumptions made in its source code about
1200 fts(3) structure members. Such software should be fixed
1201 by its vendor or, in the worst case, in the ports tree.
1202 FreeBSD_version 800015 marks this change for the unlikely
1203 case that a portable fix is impossible.
1206 To upgrade to -current after this date, you must be running
1207 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1208 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1211 The ADAPTIVE_GIANT kernel option has been retired because its
1212 functionality is the default now.
1215 The AT keyboard emulation of sunkbd(4) has been turned on
1216 by default. In order to make the special symbols of the Sun
1217 keyboards driven by sunkbd(4) work under X these now have
1218 to be configured the same way as Sun USB keyboards driven
1219 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1221 Option "XkbLayout" "us"
1222 Option "XkbRules" "xorg"
1223 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1226 It has been decided that it is desirable to provide ABI
1227 backwards compatibility to the FreeBSD 4/5/6 versions of the
1228 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1229 broken with the introduction of PCI domain support (see the
1230 20070930 entry). Unfortunately, this required the ABI of
1231 PCIOCGETCONF to be broken again in order to be able to
1232 provide backwards compatibility to the old version of that
1233 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1234 again. As for prominent ports this affects neither pciutils
1235 nor xorg-server this time, the hal port needs to be rebuilt
1239 The misnamed kthread_create() and friends have been renamed
1240 to kproc_create() etc. Many of the callers already
1241 used kproc_start()..
1242 I will return kthread_create() and friends in a while
1243 with implementations that actually create threads, not procs.
1244 Renaming corresponds with version 800002.
1253 Avoid using make -j when upgrading. While generally safe, there are
1254 sometimes problems using -j to upgrade. If your upgrade fails with
1255 -j, please try again without -j. From time to time in the past there
1256 have been problems using -j with buildworld and/or installworld. This
1257 is especially true when upgrading between "distant" versions (eg one
1258 that cross a major release boundary or several minor releases, or when
1259 several months have passed on the -current branch).
1261 Sometimes, obscure build problems are the result of environment
1262 poisoning. This can happen because the make utility reads its
1263 environment when searching for values for global variables. To run
1264 your build attempts in an "environmental clean room", prefix all make
1265 commands with 'env -i '. See the env(1) manual page for more details.
1267 When upgrading from one major version to another it is generally best
1268 to upgrade to the latest code in the currently installed branch first,
1269 then do an upgrade to the new branch. This is the best-tested upgrade
1270 path, and has the highest probability of being successful. Please try
1271 this approach before reporting problems with a major version upgrade.
1275 When upgrading the boot ZFS pool to a new version, always follow
1278 1.) recompile and reinstall the ZFS boot loader and boot block
1279 (this is part of "make buildworld" and "make installworld")
1281 2.) update the ZFS boot block on your boot drive
1283 The following example updates the ZFS boot block on the first
1284 partition (freebsd-boot) of a GPT partitioned drive ad0:
1285 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1287 Non-boot pools do not need these updates.
1291 If you are updating from a prior version of FreeBSD (even one just
1292 a few days old), you should follow this procedure. It is the most
1293 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1295 make kernel-toolchain
1296 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1297 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1299 To test a kernel once
1300 ---------------------
1301 If you just want to boot a kernel once (because you are not sure
1302 if it works, or if you want to boot a known bad kernel to provide
1303 debugging information) run
1304 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1305 nextboot -k testkernel
1307 To just build a kernel when you know that it won't mess you up
1308 --------------------------------------------------------------
1309 This assumes you are already running a CURRENT system. Replace
1310 ${arch} with the architecture of your machine (e.g. "i386",
1311 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1313 cd src/sys/${arch}/conf
1314 config KERNEL_NAME_HERE
1315 cd ../compile/KERNEL_NAME_HERE
1320 If this fails, go to the "To build a kernel" section.
1322 To rebuild everything and install it on the current system.
1323 -----------------------------------------------------------
1324 # Note: sometimes if you are running current you gotta do more than
1325 # is listed here if you are upgrading from a really old current.
1327 <make sure you have good level 0 dumps>
1329 make kernel KERNCONF=YOUR_KERNEL_HERE
1331 <reboot in single user> [3]
1339 To cross-install current onto a separate partition
1340 --------------------------------------------------
1341 # In this approach we use a separate partition to hold
1342 # current's root, 'usr', and 'var' directories. A partition
1343 # holding "/", "/usr" and "/var" should be about 2GB in
1346 <make sure you have good level 0 dumps>
1349 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1350 <maybe newfs current's root partition>
1351 <mount current's root partition on directory ${CURRENT_ROOT}>
1352 make installworld DESTDIR=${CURRENT_ROOT}
1353 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1354 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1355 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1356 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1357 <reboot into current>
1358 <do a "native" rebuild/install as described in the previous section>
1359 <maybe install compatibility libraries from ports/misc/compat*>
1363 To upgrade in-place from 8.x-stable to current
1364 ----------------------------------------------
1365 <make sure you have good level 0 dumps>
1367 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1369 <reboot in single user> [3]
1376 Make sure that you've read the UPDATING file to understand the
1377 tweaks to various things you need. At this point in the life
1378 cycle of current, things change often and you are on your own
1379 to cope. The defaults can also change, so please read ALL of
1380 the UPDATING entries.
1382 Also, if you are tracking -current, you must be subscribed to
1383 freebsd-current@freebsd.org. Make sure that before you update
1384 your sources that you have read and understood all the recent
1385 messages there. If in doubt, please track -stable which has
1386 much fewer pitfalls.
1388 [1] If you have third party modules, such as vmware, you
1389 should disable them at this point so they don't crash your
1392 [3] From the bootblocks, boot -s, and then do
1397 adjkerntz -i # if CMOS is wall time
1398 Also, when doing a major release upgrade, it is required that
1399 you boot into single user mode to do the installworld.
1401 [4] Note: This step is non-optional. Failure to do this step
1402 can result in a significant reduction in the functionality of the
1403 system. Attempting to do it by hand is not recommended and those
1404 that pursue this avenue should read this file carefully, as well
1405 as the archives of freebsd-current and freebsd-hackers mailing lists
1406 for potential gotchas. The -U option is also useful to consider.
1407 See mergemaster(8) for more information.
1409 [5] Usually this step is a noop. However, from time to time
1410 you may need to do this if you get unknown user in the following
1411 step. It never hurts to do it all the time. You may need to
1412 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1413 install) after the buildworld before this step if you last updated
1414 from current before 20020224 or from -stable before 20020408.
1416 [6] This only deletes old files and directories. Old libraries
1417 can be deleted by "make delete-old-libs", but you have to make
1418 sure that no program is using those libraries anymore.
1420 [8] In order to have a kernel that can run the 4.x binaries needed to
1421 do an installworld, you must include the COMPAT_FREEBSD4 option in
1422 your kernel. Failure to do so may leave you with a system that is
1423 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1424 required to run the 5.x binaries on more recent kernels. And so on
1425 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1427 Make sure that you merge any new devices from GENERIC since the
1428 last time you updated your kernel config file.
1430 [9] When checking out sources, you must include the -P flag to have
1431 cvs prune empty directories.
1433 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1434 "?=" instead of the "=" assignment operator, so that buildworld can
1435 override the CPUTYPE if it needs to.
1437 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1438 not on the command line, or in /etc/make.conf. buildworld will
1439 warn if it is improperly defined.
1442 This file contains a list, in reverse chronological order, of major
1443 breakages in tracking -current. Not all things will be listed here,
1444 and it only starts on October 16, 2004. Updating files can found in
1445 previous releases if your system is older than this.
1447 Copyright information:
1449 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1451 Redistribution, publication, translation and use, with or without
1452 modification, in full or in part, in any form or format of this
1453 document are permitted without further permission from the author.
1455 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1456 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1457 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1458 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1459 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1460 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1461 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1462 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1463 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1464 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1465 POSSIBILITY OF SUCH DAMAGE.
1467 Contact Warner Losh if you have any questions about your use of