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 20120503: p1 FreeBSD-SA-12:01.openssl
13 Fix multiple OpenSSL vulnerabilities.
19 The broken amd(4) driver has been replaced with esp(4) in the amd64,
20 i386 and pc98 GENERIC kernel configuration files.
23 This commit modifies vfs_register() so that it uses a hash
24 calculation to set vfc_typenum, which is enabled by default.
25 The first time a system is booted after this change, the
26 vfc_typenum values will change for all file systems. The
27 main effect of this is a change to the NFS server file handles
28 for file systems that use vfc_typenum in their fsid, such as ZFS.
29 It will, however, prevent vfc_typenum from changing when file
30 systems are loaded in a different order for subsequent reboots.
31 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
32 until you are ready to remount all NFS clients after a reboot.
35 Bump the shared library version numbers for libraries that
36 do not use symbol versioning, have changed the ABI compared
37 to stable/8 and which shared library version was not bumped.
38 Done as part of 9.0-RELEASE cycle.
41 During the merge of Capsicum features, the fget(9) KPI was modified.
42 This may require the rebuilding of out-of-tree device drivers --
43 issues have been reported specifically with the nVidia device driver.
44 __FreeBSD_version is bumped to 900041.
46 Also, there is a period between 20110811 and 20110814 where the
47 special devices /dev/{stdin,stdout,stderr} did not work correctly.
48 Building world from a kernel during that window may not work.
51 The packet filter (pf) code has been updated to OpenBSD 4.5.
52 You need to update userland tools to be in sync with kernel.
53 This update breaks backward compatibility with earlier pfsync(4)
54 versions. Care must be taken when updating redundant firewall setups.
57 The following sysctls and tunables are retired on x86 platforms:
59 machdep.hlt_logical_cpus
60 The following sysctl is retired:
61 machdep.hyperthreading_allowed
62 The sysctls were supposed to provide a way to dynamically offline and
63 online selected CPUs on x86 platforms, but the implementation has not
64 been reliable especially with SCHED_ULE scheduler.
65 machdep.hyperthreading_allowed tunable is still available to ignore
66 hyperthreading CPUs at OS level.
67 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
68 where X is an APIC ID of a CPU. Be advised, though, that disabling
69 CPUs in non-uniform fashion will result in non-uniform topology and
70 may lead to sub-optimal system performance with SCHED_ULE, which is
74 cpumask_t type is retired and cpuset_t is used in order to describe
78 Changes to ifconfig(8) for dynamic address family detection mandate
79 that you are running a kernel of 20110525 or later. Make sure to
80 follow the update procedure to boot a new kernel before installing
84 Support for sun4v architecture is officially dropped
87 Several KPI breaking changes have been committed to the mii(4) layer,
88 the PHY drivers and consequently some Ethernet drivers using mii(4).
89 This means that miibus.ko and the modules of the affected Ethernet
90 drivers need to be recompiled.
92 Note to kernel developers: Given that the OUI bit reversion problem
93 was fixed as part of these changes all mii(4) commits related to OUIs,
94 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
95 handling, no longer can be merged verbatim to stable/8 and previous
99 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
100 into their kernel configurations along with 'device pci'.
103 The default NFS client is now the new NFS client, so fstype "newnfs"
104 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
105 Although mounts via fstype "nfs" will usually work without userland
106 changes, it is recommended that the mount(8) and mount_nfs(8)
107 commands be rebuilt from sources and that a link to mount_nfs called
108 mount_oldnfs be created. The new client is compiled into the
109 kernel with "options NFSCL" and this is needed for diskless root
110 file systems. The GENERIC kernel configs have been changed to use
111 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
112 To use the regular/old client, you can "mount -t oldnfs ...". For
113 a diskless root file system, you must also include a line like:
115 vfs.root.mountfrom="oldnfs:"
117 in the boot/loader.conf on the root fs on the NFS server to make
118 a diskless root fs use the old client.
121 The GENERIC kernels for all architectures now default to the new
122 CAM-based ATA stack. It means that all legacy ATA drivers were
123 removed and replaced by respective CAM drivers. If you are using
124 ATA device names in /etc/fstab or other places, make sure to update
125 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
126 where 'Y's are the sequential numbers starting from zero for each type
127 in order of detection, unless configured otherwise with tunables,
128 see cam(4)). There will be symbolic links created in /dev/ to map
129 old adX devices to the respective adaY. They should provide basic
130 compatibility for file systems mounting in most cases, but they do
131 not support old user-level APIs and do not have respective providers
132 in GEOM. Consider using updated management tools with new device names.
134 It is possible to load devices ahci, ata, siis and mvs as modules,
135 but option ATA_CAM should remain in kernel configuration to make ata
136 module work as CAM driver supporting legacy ATA controllers. Device ata
137 still can be used in modular fashion (atacore + ...). Modules atadisk
138 and atapi* are not used and won't affect operation in ATA_CAM mode.
139 Note that to use CAM-based ATA kernel should include CAM devices
140 scbus, pass, da (or explicitly ada), cd and optionally others. All of
141 them are parts of the cam module.
143 ataraid(4) functionality is now supported by the RAID GEOM class.
144 To use it you can load geom_raid kernel module and use graid(8) tool
145 for management. Instead of /dev/arX device names, use /dev/raid/rX.
147 No kernel config options or code have been removed, so if a problem
148 arises, please report it and optionally revert to the old ATA stack.
149 In order to do it you can remove from the kernel config:
154 , and instead add back:
155 device atadisk # ATA disk drives
156 device ataraid # ATA RAID drives
157 device atapicd # ATAPI CDROM drives
158 device atapifd # ATAPI floppy drives
159 device atapist # ATAPI tape drives
162 The default NFS server has been changed to the new server, which
163 was referred to as the experimental server. If you need to switch
164 back to the old NFS server, you must now put the "-o" option on
165 both the mountd and nfsd commands. This can be done using the
166 mountd_flags and nfs_server_flags rc.conf variables until an
167 update to the rc scripts is committed, which is coming soon.
170 The GNU Objective-C runtime library (libobjc), and other Objective-C
171 related components have been removed from the base system. If you
172 require an Objective-C library, please use one of the available ports.
175 ath(4) has been split into bus- and device- modules. if_ath contains
176 the HAL, the TX rate control and the network device code. if_ath_pci
177 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
178 contains the AHB glue. Users need to load both if_ath_pci and if_ath
179 in order to use ath on everything else.
181 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
182 need to load if_ath and if_ath_pci for ath(4) operation.
185 As part of the replacement of sysinstall, the process of building
186 release media has changed significantly. For details, please re-read
187 release(7), which has been updated to reflect the new build process.
190 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
191 is the last available version under GPLv2. It brings a number of new
192 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
193 4.1 and SSE 4.2), better support for powerpc64, a number of new
194 directives, and lots of other small improvements. See the ChangeLog
195 file in contrib/binutils for the full details.
198 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
199 compliant, and will now use half of hash for authentication.
200 This will break interoperability with all stacks (including all
201 actual FreeBSD versions) who implement
202 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
204 The only workaround with such peers is to use another HMAC
205 algorithm for IPsec ("phase 2") authentication.
208 Remove the uio_yield prototype and symbol. This function has
209 been misnamed since it was introduced and should not be
210 globally exposed with this name. The equivalent functionality
211 is now available using kern_yield(curthread->td_user_pri).
212 The function remains undocumented.
215 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
216 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
217 for scalar sysctls is defined but disabled. Code that needs
218 UQUAD to pass the type checking that must compile on older
219 systems where the define is not present can check against
220 __FreeBSD_version >= 900030.
222 The system dialog(1) has been replaced with a new version previously
223 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
224 with the previous version, but the libdialog associated with it has
225 a largely incompatible API. As such, the original version of libdialog
226 will be kept temporarily as libodialog, until its base system consumers
227 are replaced or updated. Bump __FreeBSD_version to 900030.
230 If you are trying to run make universe on a -stable system, and you get
231 the following warning:
232 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
233 unknown. config(8) likely too old."
234 or something similar to it, then you must upgrade your -stable system
235 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
236 stable/8) or build the config from the latest stable/8 branch and
237 install it on your system.
239 Prior to this date, building a current universe on 8-stable system from
240 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
241 error in the first kernel build phase. A new config on those old
242 systems will fix that problem for older versions of -current.
245 The TCP stack has been modified to allow Khelp modules to interact with
246 it via helper hook points and store per-connection data in the TCP
247 control block. Bump __FreeBSD_version to 900029. User space tools that
248 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
252 Generic IEEE 802.3 annex 31B full duplex flow control support has been
253 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
254 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
255 to take advantage of it instead of using custom implementations. This
256 means that these drivers now no longer unconditionally advertise
257 support for flow control but only do so if flow control is a selected
258 media option. This was implemented in the generic support that way in
259 order to allow flow control to be switched on and off via ifconfig(8)
260 with the PHY specific default to typically off in order to protect
261 from unwanted effects. Consequently, if you used flow control with
262 one of the above mentioned drivers you now need to explicitly enable
264 ifconfig bge0 media auto mediaopt flowcontrol
266 Along with the above mentioned changes generic support for setting
267 1000baseT master mode also has been added and brgphy(4), ciphy(4),
268 e1000phy(4) as well as ip1000phy(4) have been converted to take
269 advantage of it. This means that these drivers now no longer take the
270 link0 parameter for selecting master mode but the master media option
271 has to be used instead, for example like in the following:
272 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
274 Selection of master mode now is also available with all other PHY
275 drivers supporting 1000baseT.
278 The TCP stack has received a significant update to add support for
279 modularised congestion control and generally improve the clarity of
280 congestion control decisions. Bump __FreeBSD_version to 900025. User
281 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
282 sockstat) need to be recompiled.
285 The man(1) utility has been replaced by a new version that no longer
286 uses /etc/manpath.config. Please consult man.conf(5) for how to
287 migrate local entries to the new format.
290 The copyright strings printed by login(1) and sshd(8) at the time of a
291 new connection have been removed to follow other operating systems and
295 A workaround for a fixed ld bug has been removed in kernel code,
296 so make sure that your system ld is built from sources after
297 revision 210245 from 2010-07-19 (r211583 if building head kernel
298 on stable/8, r211584 for stable/7; both from 2010-08-21).
299 A symptom of incorrect ld version is different addresses for
300 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
303 The $ipv6_prefer variable in rc.conf(5) has been split into
304 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
306 The $ip6addrctl_policy is a variable to choose a pre-defined
307 address selection policy set by ip6addrctl(8). A value
308 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
311 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
312 flag (see an entry of 20090926) is set on an interface with no
313 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
314 security reason. If you want IPv6 link-local address on all
315 interfaces by default, set this to "YES".
317 The old ipv6_prefer="YES" is equivalent to
318 ipv6_activate_all_interfaces="YES" and
319 ip6addrctl_policy="ipv6_prefer".
322 DTrace has grown support for userland tracing. Due to this, DTrace is
323 now i386 and amd64 only.
324 dtruss(1) is now installed by default on those systems and a new
325 kernel module is needed for userland tracing: fasttrap.
326 No changes to your kernel config file are necessary to enable
327 userland tracing, but you might consider adding 'STRIP=' and
328 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
329 to have informative userland stack traces in DTrace (ustack).
332 The acpi_aiboost(4) driver has been removed in favor of the new
333 aibs(4) driver. You should update your kernel configuration file.
336 BSD grep has been imported to the base system and it is built by
337 default. It is completely BSD licensed, highly GNU-compatible, uses
338 less memory than its GNU counterpart and has a small codebase.
339 However, it is slower than its GNU counterpart, which is mostly
340 noticeable for larger searches, for smaller ones it is measurable
341 but not significant. The reason is complex, the most important factor
342 is that we lack a modern and efficient regex library and GNU
343 overcomes this by optimizing the searches internally. Future work
344 on improving the regex performance is planned, for the meantime,
345 users that need better performance, can build GNU grep instead by
346 setting the WITH_GNU_GREP knob.
349 Due to the import of powerpc64 support, all existing powerpc kernel
350 configuration files must be updated with a machine directive like this:
351 machine powerpc powerpc
353 In addition, an updated config(8) is required to build powerpc kernels
357 A new version of ZFS (version 15) has been merged to -HEAD.
358 This version uses a python library for the following subcommands:
359 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
360 For full functionality of these commands the following port must
361 be installed: sysutils/py-zfs
364 'vm_page's are now hashed by physical address to an array of mutexes.
365 Currently this is only used to serialize access to hold_count. Over
366 time the page queue mutex will be peeled away. This changes the size
367 of pmap on every architecture. And requires all callers of vm_page_hold
368 and vm_page_unhold to be updated.
371 WITH_CTF can now be specified in src.conf (not recommended, there
372 are some problems with static executables), make.conf (would also
373 affect ports which do not use GNU make and do not override the
374 compile targets) or in the kernel config (via "makeoptions
376 When WITH_CTF was specified there before this was silently ignored,
377 so make sure that WITH_CTF is not used in places which could lead
378 to unwanted behavior.
381 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
382 to allow 32-bit compatibility on non-x86 platforms. All kernel
383 configurations on amd64 and ia64 platforms using these options must
384 be modified accordingly.
387 The utmp user accounting database has been replaced with utmpx,
388 the user accounting interface standardized by POSIX.
389 Unfortunately the semantics of utmp and utmpx don't match,
390 making it practically impossible to support both interfaces.
391 The user accounting database is used by tools like finger(1),
392 last(1), talk(1), w(1) and ac(8).
394 All applications in the base system use utmpx. This means only
395 local binaries (e.g. from the ports tree) may still use these
396 utmp database files. These applications must be rebuilt to make
399 After the system has been upgraded, it is safe to remove the old
400 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
401 assuming their contents is of no importance anymore. Old wtmp
402 databases can only be used by last(1) and ac(8) after they have
403 been converted to the new format using wtmpcvt(1).
406 Introduce the kernel thread "deadlock resolver" (which can be enabled
407 via the DEADLKRES option, see NOTES for more details) and the
408 sleepq_type() function for sleepqueues.
411 The rc.firewall and rc.firewall6 were unified, and
412 rc.firewall6 and rc.d/ip6fw were removed.
413 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
414 variables are obsoleted. Instead, the following new rc
415 variables are added to rc.d/ipfw:
417 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
418 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
419 firewall_simple_onet_ipv6, firewall_trusted_ipv6
421 The meanings correspond to the relevant IPv4 variables.
427 The default terminal emulation for syscons(4) has been changed
428 from cons25 to xterm on all platforms except pc98. This means
429 that the /etc/ttys file needs to be updated to ensure correct
430 operation of applications on the console.
432 The terminal emulation style can be toggled per window by using
433 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
434 options can be used to change the compile-time default back to
437 To prevent graphical artifacts, make sure the TERM environment
438 variable is set to match the terminal emulation that is being
439 performed by syscons(4).
442 The layout of the structure ieee80211req_scan_result has changed.
443 Applications that require wireless scan results (e.g. ifconfig(8))
444 from net80211 need to be recompiled.
446 Applications such as wpa_supplicant(8) may require a full world
447 build without using NO_CLEAN in order to get synchronized with the
451 The iwn(4) driver has been updated to support the 5000 and 5150 series.
452 There's one kernel module for each firmware. Adding "device iwnfw"
453 to the kernel configuration file means including all three firmware
454 images inside the kernel. If you want to include just the one for
455 your wireless card, use the the devices iwn4965fw, iwn5000fw or
459 The rc.d/network_ipv6, IPv6 configuration script has been integrated
460 into rc.d/netif. The changes are the following:
462 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
463 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
464 Note that both variables need the "inet6" keyword at the head.
466 Do not set $ipv6_network_interfaces manually if you do not
467 understand what you are doing. It is not needed in most cases.
469 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
472 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
473 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
475 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
476 all configured interfaces have "inet6 accept_rtadv" in the
477 $ifconfig_IF_ipv6. These are for backward compatibility.
479 3. A new variable $ipv6_prefer has been added. If NO, IPv6
480 functionality of interfaces with no corresponding
481 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
482 and the default address selection policy of ip6addrctl(8)
483 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
484 Note that if you want to configure IPv6 functionality on the
485 disabled interfaces after boot, first you need to clear the flag by
486 using ifconfig(8) like:
488 ifconfig em0 inet6 -ifdisabled
490 If YES, the default address selection policy is set as
493 The default value of $ipv6_prefer is NO.
495 4. If your system need to receive Router Advertisement messages,
496 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
497 scripts automatically invoke rtsol(8) when the interface becomes
498 UP. The Router Advertisement messages are used for SLAAC
499 (State-Less Address AutoConfiguration).
502 802.11s D3.03 support was committed. This is incompatible with the
503 previous code, which was based on D3.0.
506 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
507 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
508 control whether accepting Router Advertisement messages or not.
509 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
510 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
511 The ifconfig(8) utility now supports these flags.
514 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
515 mount(8) and -a option for df(1) to see them.
518 The old tunable hw.bus.devctl_disable has been superseded by
519 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
520 replaced by hw.bus.devctl_queue=0. The default for this new tunable
524 Remove the option STOP_NMI. The default action is now to use NMI only
525 for KDB via the newly introduced function stop_cpus_hard() and
526 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
529 The stable/8 branch created in subversion. This corresponds to the
530 RELENG_8 branch in CVS.
533 Bump the shared library version numbers for all libraries that do not
534 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
535 __FreeBSD_version to 800105.
538 Due to changes in the implementation of virtual network stack support,
539 all network-related kernel modules must be recompiled. As this change
540 breaks the ABI, bump __FreeBSD_version to 800104.
543 The TOE interface to the TCP syncache has been modified to remove
544 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
545 The cxgb driver is the only TOE consumer affected by this change, and
546 needs to be recompiled along with the kernel. As this change breaks
547 the ABI, bump __FreeBSD_version to 800103.
550 Padding has been added to struct tcpcb, sackhint and tcpstat in
551 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
552 maintaining the ABI. However, this change breaks the ABI, so bump
553 __FreeBSD_version to 800102. User space tools that rely on the size of
554 any of these structs (e.g. sockstat) need to be recompiled.
557 The NFS_LEGACYRPC option has been removed along with the old kernel
558 RPC implementation that this option selected. Kernel configurations
559 may need to be adjusted.
562 The network interface device nodes at /dev/net/<interface> have been
563 removed. All ioctl operations can be performed the normal way using
564 routing sockets. The kqueue functionality can generally be replaced
565 with routing sockets.
568 The documentation from the FreeBSD Documentation Project (Handbook,
569 FAQ, etc.) is now installed via packages by sysinstall(8) and under
570 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
573 The ABI of various structures related to the SYSV IPC API have been
574 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
575 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
579 Layout of struct vnet has changed as routing related variables were
580 moved to their own Vimage module. Modules need to be recompiled. Bump
581 __FreeBSD_version to 800099.
584 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
585 respectively. As long as no more than 16 groups per process are used,
586 no changes should be visible. When more than 16 groups are used, old
587 binaries may fail if they call getgroups() or getgrouplist() with
588 statically sized storage. Recompiling will work around this, but
589 applications should be modified to use dynamically allocated storage
590 for group arrays as POSIX.1-2008 does not cap an implementation's
591 number of supported groups at NGROUPS_MAX+1 as previous versions did.
593 NFS and portalfs mounts may also be affected as the list of groups is
594 truncated to 16. Users of NFS who use more than 16 groups, should
595 take care that negative group permissions are not used on the exported
596 file systems as they will not be reliable unless a GSSAPI based
597 authentication method is used.
600 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
601 option compiles in the support for adaptive spinning for lockmgrs
602 which want to enable it. The lockinit() function now accepts the flag
603 LK_ADAPTIVE in order to make the lock object subject to adaptive
604 spinning when both held in write and read mode.
607 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
608 changed. User applications that use this ioctl need to be rebuilt.
611 The layout of struct thread has changed. Kernel and modules need to
615 The layout of structs ifnet, domain, protosw and vnet_net has changed.
616 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
619 window(1) has been removed from the base system. It can now be
620 installed from ports. The port is called misc/window.
623 The way we are storing and accessing `routing table' entries has
624 changed. Programs reading the FIB, like netstat, need to be
628 A new netisr implementation has been added for FreeBSD 8. Network
629 file system modules, such as igmp, ipdivert, and others, should be
631 Bump __FreeBSD_version to 800096.
634 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
638 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
639 Bump __FreeBSD_version to 800094.
642 Add mnt_xflag field to 'struct mount'. File system modules need to be
644 Bump __FreeBSD_version to 800093.
647 The compiling option ADAPTIVE_SX has been retired while it has been
648 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
649 The KPI for sx_init_flags() changes as accepting flags:
650 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
651 been introduced in order to handle the reversed logic.
652 Bump __FreeBSD_version to 800092.
655 Add support for hierarchical jails. Remove global securelevel.
656 Bump __FreeBSD_version to 800091.
659 The layout of struct vnet_net has changed, therefore modules
661 Bump __FreeBSD_version to 800090.
664 The newly imported zic(8) produces a new format in the output. Please
665 run tzsetup(8) to install the newly created data to /etc/localtime.
668 The sysctl tree for the usb stack has renamed from hw.usb2.* to
669 hw.usb.* and is now consistent again with previous releases.
672 802.11 monitor mode support was revised and driver api's were changed.
673 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
674 of DLT_IEEE802_11. No user-visible data structures were changed but
675 applications that use DLT_IEEE802_11 may require changes.
676 Bump __FreeBSD_version to 800088.
679 The layout of the following structs has changed: sysctl_oid,
680 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
681 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
682 panics may be experienced. World rebuild is required for
683 correctly checking networking state from userland.
684 Bump __FreeBSD_version to 800085.
687 MLDv2 and Source-Specific Multicast (SSM) have been merged
688 to the IPv6 stack. VIMAGE hooks are in but not yet used.
689 The implementation of SSM within FreeBSD's IPv6 stack closely
690 follows the IPv4 implementation.
692 For kernel developers:
694 * The most important changes are that the ip6_output() and
695 ip6_input() paths no longer take the IN6_MULTI_LOCK,
696 and this lock has been downgraded to a non-recursive mutex.
698 * As with the changes to the IPv4 stack to support SSM, filtering
699 of inbound multicast traffic must now be performed by transport
700 protocols within the IPv6 stack. This does not apply to TCP and
701 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
703 * The KPIs used by IPv6 multicast are similar to those used by
704 the IPv4 stack, with the following differences:
705 * im6o_mc_filter() is analogous to imo_multicast_filter().
706 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
707 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
708 * IN6_LOOKUP_MULTI() has been deprecated and removed.
709 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
710 for MLDv1 have an additional 'timer' argument which is used to
711 jitter the initial membership report for the solicited-node
712 multicast membership on-link.
713 * This is not strictly needed for MLDv2, which already jitters
714 its report transmissions. However, the 'timer' argument is
715 preserved in case MLDv1 is active on the interface.
717 * The KAME linked-list based IPv6 membership implementation has
718 been refactored to use a vector similar to that used by the IPv4
720 Code which maintains a list of its own multicast memberships
721 internally, e.g. carp, has been updated to reflect the new
724 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
725 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
726 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
727 implementation constraint which needs to be addressed in HEAD.
729 For application developers:
731 * The changes are broadly similar to those made for the IPv4
734 * The use of IPv4 and IPv6 multicast socket options on the same
735 socket, using mapped addresses, HAS NOT been tested or supported.
737 * There are a number of issues with the implementation of various
738 IPv6 multicast APIs which need to be resolved in the API surface
739 before the implementation is fully compatible with KAME userland
740 use, and these are mostly to do with interface index treatment.
742 * The literature available discusses the use of either the delta / ASM
743 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
744 using setsourcefilter(3)/getsourcefilter(3). For more information
745 please refer to RFC 3768, 'Socket Interface Extensions for
746 Multicast Source Filters'.
748 * Applications which use the published RFC 3678 APIs should be fine.
750 For systems administrators:
752 * The mtest(8) utility has been refactored to support IPv6, in
753 addition to IPv4. Interface addresses are no longer accepted
754 as arguments, their names must be used instead. The utility
755 will map the interface name to its first IPv4 address as
756 returned by getifaddrs(3).
758 * The ifmcstat(8) utility has also been updated to print the MLDv2
759 endpoint state and source filter lists via sysctl(3).
761 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
762 loopback of IPv6 multicast datagrams by default; it defaults to 1
763 to preserve the existing behaviour. Disabling multicast loopback is
764 recommended for optimal system performance.
766 * The IPv6 MROUTING code has been changed to examine this sysctl
767 instead of attempting to perform a group lookup before looping
768 back forwarded datagrams.
770 Bump __FreeBSD_version to 800084.
773 Implement low-level Bluetooth HCI API.
774 Bump __FreeBSD_version to 800083.
777 The layout of struct malloc_type, used by modules to register new
778 memory allocation types, has changed. Most modules will need to
779 be rebuilt or panics may be experienced.
780 Bump __FreeBSD_version to 800081.
783 Anticipate overflowing inp_flags - add inp_flags2.
784 This changes most offsets in inpcb, so checking v4 connection
785 state will require a world rebuild.
786 Bump __FreeBSD_version to 800080.
789 Add an llentry to struct route and struct route_in6. Modules
790 embedding a struct route will need to be recompiled.
791 Bump __FreeBSD_version to 800079.
794 The size of rt_metrics_lite and by extension rtentry has changed.
795 Networking administration apps will need to be recompiled.
796 The route command now supports show as an alias for get, weighting
797 of routes, sticky and nostick flags to alter the behavior of stateful
799 Bump __FreeBSD_version to 800078.
802 Do not use Giant for kbdmux(4) locking. This is wrong and
803 apparently causing more problems than it solves. This will
804 re-open the issue where interrupt handlers may race with
805 kbdmux(4) in polling mode. Typical symptoms include (but
806 not limited to) duplicated and/or missing characters when
807 low level console functions (such as gets) are used while
808 interrupts are enabled (for example geli password prompt,
809 mountroot prompt etc.). Disabling kbdmux(4) may help.
812 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
813 kernel modules referencing any of the above need to be recompiled.
814 Bump __FreeBSD_version to 800075.
817 GEOM_PART has become the default partition slicer for storage devices,
818 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
819 introduces some changes:
821 MSDOS/EBR: the devices created from MSDOS extended partition entries
822 (EBR) can be named differently than with GEOM_MBR and are now symlinks
823 to devices with offset-based names. fstabs may need to be modified.
825 BSD: the "geometry does not match label" warning is harmless in most
826 cases but it points to problems in file system misalignment with
827 disk geometry. The "c" partition is now implicit, covers the whole
828 top-level drive and cannot be (mis)used by users.
830 General: Kernel dumps are now not allowed to be written to devices
831 whose partition types indicate they are meant to be used for file
832 systems (or, in case of MSDOS partitions, as something else than
835 Most of these changes date approximately from 200812.
838 The uscanner(4) driver has been removed from the kernel. This follows
839 Linux removing theirs in 2.6 and making libusb the default interface
843 The multicast forwarding code has been cleaned up. netstat(1)
844 only relies on KVM now for printing bandwidth upcall meters.
845 The IPv4 and IPv6 modules are split into ip_mroute_mod and
846 ip6_mroute_mod respectively. The config(5) options for statically
847 compiling this code remain the same, i.e. 'options MROUTING'.
850 Support for the IFF_NEEDSGIANT network interface flag has been
851 removed, which means that non-MPSAFE network device drivers are no
852 longer supported. In particular, if_ar, if_sr, and network device
853 drivers from the old (legacy) USB stack can no longer be built or
857 POSIX.1 Native Language Support (NLS) has been enabled in libc and
858 a bunch of new language catalog files have also been added.
859 This means that some common libc messages are now localized and
860 they depend on the LC_MESSAGES environmental variable.
863 The k8temp(4) driver has been renamed to amdtemp(4) since
864 support for Family 10 and Family 11 CPU families was added.
867 IGMPv3 and Source-Specific Multicast (SSM) have been merged
868 to the IPv4 stack. VIMAGE hooks are in but not yet used.
870 For kernel developers, the most important changes are that the
871 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
872 and this lock has been downgraded to a non-recursive mutex.
874 Transport protocols (UDP, Raw IP) are now responsible for filtering
875 inbound multicast traffic according to group membership and source
876 filters. The imo_multicast_filter() KPI exists for this purpose.
877 Transports which do not use multicast (SCTP, TCP) already reject
878 multicast by default. Forwarding and receive performance may improve
879 as a mutex acquisition is no longer needed in the ip_input()
880 low-level input path. in_addmulti() and in_delmulti() are shimmed
881 to new KPIs which exist to support SSM in-kernel.
883 For application developers, it is recommended that loopback of
884 multicast datagrams be disabled for best performance, as this
885 will still cause the lock to be taken for each looped-back
886 datagram transmission. The net.inet.ip.mcast.loop sysctl may
887 be tuned to 0 to disable loopback by default; it defaults to 1
888 to preserve the existing behaviour.
890 For systems administrators, to obtain best performance with
891 multicast reception and multiple groups, it is always recommended
892 that a card with a suitably precise hash filter is used. Hash
893 collisions will still result in the lock being taken within the
894 transport protocol input path to check group membership.
896 If deploying FreeBSD in an environment with IGMP snooping switches,
897 it is recommended that the net.inet.igmp.sendlocal sysctl remain
898 enabled; this forces 224.0.0.0/24 group membership to be announced
901 The size of 'struct igmpstat' has changed; netstat needs to be
902 recompiled to reflect this.
903 Bump __FreeBSD_version to 800070.
906 libusb20.so.1 is now installed as libusb.so.1 and the ports system
907 updated to use it. This requires a buildworld/installworld in order to
908 update the library and dependencies (usbconfig, etc). Its advisable to
909 rebuild all ports which uses libusb. More specific directions are given
910 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
911 libusb are no longer required and can be removed.
914 A workaround is committed to allow the creation of System V shared
915 memory segment of size > 2 GB on the 64-bit architectures.
916 Due to a limitation of the existing ABI, the shm_segsz member
917 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
918 wrong for large segments. Note that limits must be explicitly
919 raised to allow such segments to be created.
922 The layout of struct ifnet has changed, requiring a rebuild of all
923 network device driver modules.
926 The /dev handling for the new USB stack has changed, a
927 buildworld/installworld is required for libusb20.
930 The new USB2 stack has now been permanently moved in and all kernel and
931 module names reverted to their previous values (eg, usb, ehci, ohci,
932 ums, ...). The old usb stack can be compiled in by prefixing the name
933 with the letter 'o', the old usb modules have been removed.
934 Updating entry 20090216 for xorg and 20090215 for libmap may still
938 The rc.conf(5) option if_up_delay has been renamed to
939 defaultroute_delay to better reflect its purpose. If you have
940 customized this setting in /etc/rc.conf you need to update it to
944 xorg 7.4 wants to configure its input devices via hald which does not
945 yet work with USB2. If the keyboard/mouse does not work in xorg then
947 Option "AllowEmptyInput" "off"
948 to your ServerLayout section. This will cause X to use the configured
949 kbd and mouse sections from your xorg.conf.
952 The GENERIC kernels for all architectures now default to the new USB2
953 stack. No kernel config options or code have been removed so if a
954 problem arises please report it and optionally revert to the old USB
955 stack. If you are loading USB kernel modules or have a custom kernel
956 that includes GENERIC then ensure that usb names are also changed over,
957 eg uftdi -> usb2_serial_ftdi.
959 Older programs linked against the ports libusb 0.1 need to be
960 redirected to the new stack's libusb20. /etc/libmap.conf can
962 # Map old usb library to new one for usb2 stack
963 libusb-0.1.so.8 libusb20.so.1
966 All USB ethernet devices now attach as interfaces under the name ueN
967 (eg. ue0). This is to provide a predictable name as vendors often
968 change usb chipsets in a product without notice.
971 The ichsmb(4) driver has been changed to require SMBus slave
972 addresses be left-justified (xxxxxxx0b) rather than right-justified.
973 All of the other SMBus controller drivers require left-justified
974 slave addresses, so this change makes all the drivers provide the
978 INET6 statistics (struct ip6stat) was updated.
979 netstat(1) needs to be recompiled.
982 NTFS has been removed from GENERIC kernel on amd64 to match
983 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
984 will load ntfs.ko module automatically when NTFS support is
985 actually needed, unless ntfs.ko is not installed or security
986 level prohibits loading kernel modules. If either is the case,
987 "options NTFS" has to be added into kernel config.
990 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
991 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
992 800061. User space tools that rely on the size of struct tcpcb in
993 tcp_var.h (e.g. sockstat) need to be recompiled.
996 ng_tty(4) module updated to match the new TTY subsystem.
997 Due to API change, user-level applications must be updated.
998 New API support added to mpd5 CVS and expected to be present
999 in next mpd5.3 release.
1002 With __FreeBSD_version 800060 the makefs tool is part of
1003 the base system (it was a port).
1006 The afdata and ifnet locks have been changed from mutexes to
1007 rwlocks, network modules will need to be re-compiled.
1010 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1011 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1012 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1013 architecture and 40 bytes on 64-bit architecture. The userland
1014 applications "arp" and "ndp" have been updated accordingly.
1015 The output from "netstat -r" shows only routing entries and
1016 none of the L2 information.
1019 __FreeBSD_version 800057 marks the switchover from the
1020 binary ath hal to source code. Users must add the line:
1022 options AH_SUPPORT_AR5416
1024 to their kernel config files when specifying:
1028 The ath_hal module no longer exists; the code is now compiled
1029 together with the driver in the ath module. It is now
1030 possible to tailor chip support (i.e. reduce the set of chips
1031 and thereby the code size); consult ath_hal(4) for details.
1034 __FreeBSD_version 800054 adds memory barriers to
1035 <machine/atomic.h>, new interfaces to ifnet to facilitate
1036 multiple hardware transmit queues for cards that support
1037 them, and a lock-less ring-buffer implementation to
1038 enable drivers to more efficiently manage queueing of
1042 A new version of ZFS (version 13) has been merged to -HEAD.
1043 This version has zpool attribute "listsnapshots" off by
1044 default, which means "zfs list" does not show snapshots,
1045 and is the same as Solaris behavior.
1048 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1051 The uhci, ohci, ehci and slhci USB Host controller drivers have
1052 been put into separate modules. If you load the usb module
1053 separately through loader.conf you will need to load the
1054 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1055 controller add the following to loader.conf:
1061 The ABI used by the PMC toolset has changed. Please keep
1062 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1066 atapci kernel module now includes only generic PCI ATA
1067 driver. AHCI driver moved to ataahci kernel module.
1068 All vendor-specific code moved into separate kernel modules:
1069 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1070 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1071 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1072 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1075 The TTY subsystem of the kernel has been replaced by a new
1076 implementation, which provides better scalability and an
1077 improved driver model. Most common drivers have been migrated to
1078 the new TTY subsystem, while others have not. The following
1079 drivers have not yet been ported to the new TTY layer:
1082 cy, digi, rc, rp, sio
1088 ng_h4, ng_tty, ppp, sl, snp
1090 Adding these drivers to your kernel configuration file shall
1091 cause compilation to fail.
1094 ntpd has been upgraded to 4.2.4p5.
1097 OpenSSH has been upgraded to 5.1p1.
1099 For many years, FreeBSD's version of OpenSSH preferred DSA
1100 over RSA for host and user authentication keys. With this
1101 upgrade, we've switched to the vendor's default of RSA over
1102 DSA. This may cause upgraded clients to warn about unknown
1103 host keys even for previously known hosts. Users should
1104 follow the usual procedure for verifying host keys before
1105 accepting the RSA key.
1107 This can be circumvented by setting the "HostKeyAlgorithms"
1108 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1111 Please note that the sequence of keys offered for
1112 authentication has been changed as well. You may want to
1113 specify IdentityFile in a different order to revert this
1117 The sio(4) driver has been removed from the i386 and amd64
1118 kernel configuration files. This means uart(4) is now the
1119 default serial port driver on those platforms as well.
1121 To prevent collisions with the sio(4) driver, the uart(4) driver
1122 uses different names for its device nodes. This means the
1123 onboard serial port will now most likely be called "ttyu0"
1124 instead of "ttyd0". You may need to reconfigure applications to
1125 use the new device names.
1127 When using the serial port as a boot console, be sure to update
1128 /boot/device.hints and /etc/ttys before booting the new kernel.
1129 If you forget to do so, you can still manually specify the hints
1130 at the loader prompt:
1132 set hint.uart.0.at="isa"
1133 set hint.uart.0.port="0x3F8"
1134 set hint.uart.0.flags="0x10"
1135 set hint.uart.0.irq="4"
1139 The gpt(8) utility has been removed. Use gpart(8) to partition
1143 The version that Linuxulator emulates was changed from 2.4.2
1144 to 2.6.16. If you experience any problems with Linux binaries
1145 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1146 if it fixes the problem contact emulation mailing list.
1149 ISDN4BSD (I4B) was removed from the src tree. You may need to
1150 update a your kernel configuration and remove relevant entries.
1153 I have checked in code to support multiple routing tables.
1154 See the man pages setfib(1) and setfib(2).
1155 This is a hopefully backwards compatible version,
1156 but to make use of it you need to compile your kernel
1157 with options ROUTETABLES=2 (or more up to 16).
1160 The 802.11 wireless support was redone to enable multi-bss
1161 operation on devices that are capable. The underlying device
1162 is no longer used directly but instead wlanX devices are
1163 cloned with ifconfig. This requires changes to rc.conf files.
1164 For example, change:
1165 ifconfig_ath0="WPA DHCP"
1168 ifconfig_wlan0="WPA DHCP"
1169 see rc.conf(5) for more details. In addition, mergemaster of
1170 /etc/rc.d is highly recommended. Simultaneous update of userland
1171 and kernel wouldn't hurt either.
1173 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1174 modules were merged into the base wlan module. All references
1175 to these modules (e.g. in kernel config files) must be removed.
1178 psm(4) has gained write(2) support in native operation level.
1179 Arbitrary commands can be written to /dev/psm%d and status can
1180 be read back from it. Therefore, an application is responsible
1181 for status validation and error recovery. It is a no-op in
1182 other operation levels.
1185 Support for KSE threading has been removed from the kernel. To
1186 run legacy applications linked against KSE libmap.conf may
1187 be used. The following libmap.conf may be used to ensure
1188 compatibility with any prior release:
1190 libpthread.so.1 libthr.so.1
1191 libpthread.so.2 libthr.so.2
1192 libkse.so.3 libthr.so.3
1195 The layout of struct vmspace has changed. This affects libkvm
1196 and any executables that link against libkvm and use the
1197 kvm_getprocs() function. In particular, but not exclusively,
1198 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1199 The effects are minimal, but it's advisable to upgrade world
1203 The latest em driver no longer has support in it for the
1204 82575 adapter, this is now moved to the igb driver. The
1205 split was done to make new features that are incompatible
1206 with older hardware easier to do.
1209 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1210 likewise the kernel option is now GEOM_LINUX_LVM.
1213 The default NFS mount mode has changed from UDP to TCP for
1214 increased reliability. If you rely on (insecurely) NFS
1215 mounting across a firewall you may need to update your
1219 Belatedly note the addition of m_collapse for compacting
1223 The fts(3) structures have been changed to use adequate
1224 integer types for their members and so to be able to cope
1225 with huge file trees. The old fts(3) ABI is preserved
1226 through symbol versioning in libc, so third-party binaries
1227 using fts(3) should still work, although they will not take
1228 advantage of the extended types. At the same time, some
1229 third-party software might fail to build after this change
1230 due to unportable assumptions made in its source code about
1231 fts(3) structure members. Such software should be fixed
1232 by its vendor or, in the worst case, in the ports tree.
1233 FreeBSD_version 800015 marks this change for the unlikely
1234 case that a portable fix is impossible.
1237 To upgrade to -current after this date, you must be running
1238 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1239 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1242 The ADAPTIVE_GIANT kernel option has been retired because its
1243 functionality is the default now.
1246 The AT keyboard emulation of sunkbd(4) has been turned on
1247 by default. In order to make the special symbols of the Sun
1248 keyboards driven by sunkbd(4) work under X these now have
1249 to be configured the same way as Sun USB keyboards driven
1250 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1252 Option "XkbLayout" "us"
1253 Option "XkbRules" "xorg"
1254 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1257 It has been decided that it is desirable to provide ABI
1258 backwards compatibility to the FreeBSD 4/5/6 versions of the
1259 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1260 broken with the introduction of PCI domain support (see the
1261 20070930 entry). Unfortunately, this required the ABI of
1262 PCIOCGETCONF to be broken again in order to be able to
1263 provide backwards compatibility to the old version of that
1264 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1265 again. As for prominent ports this affects neither pciutils
1266 nor xorg-server this time, the hal port needs to be rebuilt
1270 The misnamed kthread_create() and friends have been renamed
1271 to kproc_create() etc. Many of the callers already
1272 used kproc_start()..
1273 I will return kthread_create() and friends in a while
1274 with implementations that actually create threads, not procs.
1275 Renaming corresponds with version 800002.
1284 Avoid using make -j when upgrading. While generally safe, there are
1285 sometimes problems using -j to upgrade. If your upgrade fails with
1286 -j, please try again without -j. From time to time in the past there
1287 have been problems using -j with buildworld and/or installworld. This
1288 is especially true when upgrading between "distant" versions (eg one
1289 that cross a major release boundary or several minor releases, or when
1290 several months have passed on the -current branch).
1292 Sometimes, obscure build problems are the result of environment
1293 poisoning. This can happen because the make utility reads its
1294 environment when searching for values for global variables. To run
1295 your build attempts in an "environmental clean room", prefix all make
1296 commands with 'env -i '. See the env(1) manual page for more details.
1298 When upgrading from one major version to another it is generally best
1299 to upgrade to the latest code in the currently installed branch first,
1300 then do an upgrade to the new branch. This is the best-tested upgrade
1301 path, and has the highest probability of being successful. Please try
1302 this approach before reporting problems with a major version upgrade.
1306 When upgrading the boot ZFS pool to a new version, always follow
1309 1.) recompile and reinstall the ZFS boot loader and boot block
1310 (this is part of "make buildworld" and "make installworld")
1312 2.) update the ZFS boot block on your boot drive
1314 The following example updates the ZFS boot block on the first
1315 partition (freebsd-boot) of a GPT partitioned drive ad0:
1316 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1318 Non-boot pools do not need these updates.
1322 If you are updating from a prior version of FreeBSD (even one just
1323 a few days old), you should follow this procedure. It is the most
1324 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1326 make kernel-toolchain
1327 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1328 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1330 To test a kernel once
1331 ---------------------
1332 If you just want to boot a kernel once (because you are not sure
1333 if it works, or if you want to boot a known bad kernel to provide
1334 debugging information) run
1335 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1336 nextboot -k testkernel
1338 To just build a kernel when you know that it won't mess you up
1339 --------------------------------------------------------------
1340 This assumes you are already running a CURRENT system. Replace
1341 ${arch} with the architecture of your machine (e.g. "i386",
1342 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1344 cd src/sys/${arch}/conf
1345 config KERNEL_NAME_HERE
1346 cd ../compile/KERNEL_NAME_HERE
1351 If this fails, go to the "To build a kernel" section.
1353 To rebuild everything and install it on the current system.
1354 -----------------------------------------------------------
1355 # Note: sometimes if you are running current you gotta do more than
1356 # is listed here if you are upgrading from a really old current.
1358 <make sure you have good level 0 dumps>
1360 make kernel KERNCONF=YOUR_KERNEL_HERE
1362 <reboot in single user> [3]
1370 To cross-install current onto a separate partition
1371 --------------------------------------------------
1372 # In this approach we use a separate partition to hold
1373 # current's root, 'usr', and 'var' directories. A partition
1374 # holding "/", "/usr" and "/var" should be about 2GB in
1377 <make sure you have good level 0 dumps>
1380 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1381 <maybe newfs current's root partition>
1382 <mount current's root partition on directory ${CURRENT_ROOT}>
1383 make installworld DESTDIR=${CURRENT_ROOT}
1384 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1385 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1386 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1387 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1388 <reboot into current>
1389 <do a "native" rebuild/install as described in the previous section>
1390 <maybe install compatibility libraries from ports/misc/compat*>
1394 To upgrade in-place from 8.x-stable to current
1395 ----------------------------------------------
1396 <make sure you have good level 0 dumps>
1398 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1400 <reboot in single user> [3]
1407 Make sure that you've read the UPDATING file to understand the
1408 tweaks to various things you need. At this point in the life
1409 cycle of current, things change often and you are on your own
1410 to cope. The defaults can also change, so please read ALL of
1411 the UPDATING entries.
1413 Also, if you are tracking -current, you must be subscribed to
1414 freebsd-current@freebsd.org. Make sure that before you update
1415 your sources that you have read and understood all the recent
1416 messages there. If in doubt, please track -stable which has
1417 much fewer pitfalls.
1419 [1] If you have third party modules, such as vmware, you
1420 should disable them at this point so they don't crash your
1423 [3] From the bootblocks, boot -s, and then do
1428 adjkerntz -i # if CMOS is wall time
1429 Also, when doing a major release upgrade, it is required that
1430 you boot into single user mode to do the installworld.
1432 [4] Note: This step is non-optional. Failure to do this step
1433 can result in a significant reduction in the functionality of the
1434 system. Attempting to do it by hand is not recommended and those
1435 that pursue this avenue should read this file carefully, as well
1436 as the archives of freebsd-current and freebsd-hackers mailing lists
1437 for potential gotchas. The -U option is also useful to consider.
1438 See mergemaster(8) for more information.
1440 [5] Usually this step is a noop. However, from time to time
1441 you may need to do this if you get unknown user in the following
1442 step. It never hurts to do it all the time. You may need to
1443 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1444 install) after the buildworld before this step if you last updated
1445 from current before 20020224 or from -stable before 20020408.
1447 [6] This only deletes old files and directories. Old libraries
1448 can be deleted by "make delete-old-libs", but you have to make
1449 sure that no program is using those libraries anymore.
1451 [8] In order to have a kernel that can run the 4.x binaries needed to
1452 do an installworld, you must include the COMPAT_FREEBSD4 option in
1453 your kernel. Failure to do so may leave you with a system that is
1454 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1455 required to run the 5.x binaries on more recent kernels. And so on
1456 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1458 Make sure that you merge any new devices from GENERIC since the
1459 last time you updated your kernel config file.
1461 [9] When checking out sources, you must include the -P flag to have
1462 cvs prune empty directories.
1464 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1465 "?=" instead of the "=" assignment operator, so that buildworld can
1466 override the CPUTYPE if it needs to.
1468 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1469 not on the command line, or in /etc/make.conf. buildworld will
1470 warn if it is improperly defined.
1473 This file contains a list, in reverse chronological order, of major
1474 breakages in tracking -current. It is not guaranteed to be a complete
1475 list of such breakages, and only contains entries since October 10, 2007.
1476 If you need to see UPDATING entries from before that date, you will need
1477 to fetch an UPDATING file from an older FreeBSD release.
1479 Copyright information:
1481 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1483 Redistribution, publication, translation and use, with or without
1484 modification, in full or in part, in any form or format of this
1485 document are permitted without further permission from the author.
1487 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1488 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1489 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1490 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1491 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1492 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1493 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1494 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1495 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1496 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1497 POSSIBILITY OF SUCH DAMAGE.
1499 Contact Warner Losh if you have any questions about your use of