1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 Items affecting the ports and packages system can be found in
10 /usr/ports/UPDATING. Please read that file before running portupgrade.
12 NOTE TO PEOPLE WHO THINK THAT FreeBSD 10.x IS SLOW:
13 FreeBSD 10.x has many debugging features turned on, in both the kernel
14 and userland. These features attempt to detect incorrect use of
15 system primitives, and encourage loud failure through extra sanity
16 checking and fail stop semantics. They also substantially impact
17 system performance. If you want to do performance measurement,
18 benchmarking, and optimization, you'll want to turn them off. This
19 includes various WITNESS- related kernel options, INVARIANTS, malloc
20 debugging flags in userland, and various verbose features in the
21 kernel. Many developers choose to disable these features on build
22 machines to maximize performance. (To disable malloc debugging, run
23 ln -s aj /etc/malloc.conf.)
26 The carp(4) facility has been changed significantly. Configuration
27 of the CARP protocol via ifconfig(8) has changed, as well as format
28 of CARP events submitted to devd(8) has changed. See manual pages
29 for more information. The arpbalance feature of carp(4) is currently
30 not supported anymore.
32 Size of struct in_aliasreq, struct in6_aliasreq has changed. User
33 utilities using SIOCAIFADDR, SIOCAIFADDR_IN6, e.g. ifconfig(8),
34 need to be recompiled.
37 The acpi_wmi(4) status device /dev/wmistat has been renamed to
41 The option VFS_ALLOW_NONMPSAFE option has been added in order to
42 explicitely support non-MPSAFE filesystems.
43 It is on by default for all supported platform at this present
47 The broken amd(4) driver has been replaced with esp(4) in the amd64,
48 i386 and pc98 GENERIC kernel configuration files.
51 sysinstall has been removed
54 The stable/9 branch created in subversion. This corresponds to the
55 RELENG_9 branch in CVS.
58 This commit modifies vfs_register() so that it uses a hash
59 calculation to set vfc_typenum, which is enabled by default.
60 The first time a system is booted after this change, the
61 vfc_typenum values will change for all file systems. The
62 main effect of this is a change to the NFS server file handles
63 for file systems that use vfc_typenum in their fsid, such as ZFS.
64 It will, however, prevent vfc_typenum from changing when file
65 systems are loaded in a different order for subsequent reboots.
66 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
67 until you are ready to remount all NFS clients after a reboot.
70 Bump the shared library version numbers for libraries that
71 do not use symbol versioning, have changed the ABI compared
72 to stable/8 and which shared library version was not bumped.
73 Done as part of 9.0-RELEASE cycle.
76 During the merge of Capsicum features, the fget(9) KPI was modified.
77 This may require the rebuilding of out-of-tree device drivers --
78 issues have been reported specifically with the nVidia device driver.
79 __FreeBSD_version is bumped to 900041.
81 Also, there is a period between 20110811 and 20110814 where the
82 special devices /dev/{stdin,stdout,stderr} did not work correctly.
83 Building world from a kernel during that window may not work.
86 The packet filter (pf) code has been updated to OpenBSD 4.5.
87 You need to update userland tools to be in sync with kernel.
88 This update breaks backward compatibility with earlier pfsync(4)
89 versions. Care must be taken when updating redundant firewall setups.
92 The following sysctls and tunables are retired on x86 platforms:
94 machdep.hlt_logical_cpus
95 The following sysctl is retired:
96 machdep.hyperthreading_allowed
97 The sysctls were supposed to provide a way to dynamically offline and
98 online selected CPUs on x86 platforms, but the implementation has not
99 been reliable especially with SCHED_ULE scheduler.
100 machdep.hyperthreading_allowed tunable is still available to ignore
101 hyperthreading CPUs at OS level.
102 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
103 where X is an APIC ID of a CPU. Be advised, though, that disabling
104 CPUs in non-uniform fashion will result in non-uniform topology and
105 may lead to sub-optimal system performance with SCHED_ULE, which is
109 cpumask_t type is retired and cpuset_t is used in order to describe
113 Changes to ifconfig(8) for dynamic address family detection mandate
114 that you are running a kernel of 20110525 or later. Make sure to
115 follow the update procedure to boot a new kernel before installing
119 Support for sun4v architecture is officially dropped
122 Several KPI breaking changes have been committed to the mii(4) layer,
123 the PHY drivers and consequently some Ethernet drivers using mii(4).
124 This means that miibus.ko and the modules of the affected Ethernet
125 drivers need to be recompiled.
127 Note to kernel developers: Given that the OUI bit reversion problem
128 was fixed as part of these changes all mii(4) commits related to OUIs,
129 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
130 handling, no longer can be merged verbatim to stable/8 and previous
134 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
135 into their kernel configurations along with 'device pci'.
138 The default NFS client is now the new NFS client, so fstype "newnfs"
139 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
140 Although mounts via fstype "nfs" will usually work without userland
141 changes, it is recommended that the mount(8) and mount_nfs(8)
142 commands be rebuilt from sources and that a link to mount_nfs called
143 mount_oldnfs be created. The new client is compiled into the
144 kernel with "options NFSCL" and this is needed for diskless root
145 file systems. The GENERIC kernel configs have been changed to use
146 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
147 To use the regular/old client, you can "mount -t oldnfs ...". For
148 a diskless root file system, you must also include a line like:
150 vfs.root.mountfrom="oldnfs:"
152 in the boot/loader.conf on the root fs on the NFS server to make
153 a diskless root fs use the old client.
156 The GENERIC kernels for all architectures now default to the new
157 CAM-based ATA stack. It means that all legacy ATA drivers were
158 removed and replaced by respective CAM drivers. If you are using
159 ATA device names in /etc/fstab or other places, make sure to update
160 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
161 where 'Y's are the sequential numbers starting from zero for each type
162 in order of detection, unless configured otherwise with tunables,
163 see cam(4)). There will be symbolic links created in /dev/ to map
164 old adX devices to the respective adaY. They should provide basic
165 compatibility for file systems mounting in most cases, but they do
166 not support old user-level APIs and do not have respective providers
167 in GEOM. Consider using updated management tools with new device names.
169 It is possible to load devices ahci, ata, siis and mvs as modules,
170 but option ATA_CAM should remain in kernel configuration to make ata
171 module work as CAM driver supporting legacy ATA controllers. Device ata
172 still can be used in modular fashion (atacore + ...). Modules atadisk
173 and atapi* are not used and won't affect operation in ATA_CAM mode.
174 Note that to use CAM-based ATA kernel should include CAM devices
175 scbus, pass, da (or explicitly ada), cd and optionally others. All of
176 them are parts of the cam module.
178 ataraid(4) functionality is now supported by the RAID GEOM class.
179 To use it you can load geom_raid kernel module and use graid(8) tool
180 for management. Instead of /dev/arX device names, use /dev/raid/rX.
182 No kernel config options or code have been removed, so if a problem
183 arises, please report it and optionally revert to the old ATA stack.
184 In order to do it you can remove from the kernel config:
189 , and instead add back:
190 device atadisk # ATA disk drives
191 device ataraid # ATA RAID drives
192 device atapicd # ATAPI CDROM drives
193 device atapifd # ATAPI floppy drives
194 device atapist # ATAPI tape drives
197 The default NFS server has been changed to the new server, which
198 was referred to as the experimental server. If you need to switch
199 back to the old NFS server, you must now put the "-o" option on
200 both the mountd and nfsd commands. This can be done using the
201 mountd_flags and nfs_server_flags rc.conf variables until an
202 update to the rc scripts is committed, which is coming soon.
205 The GNU Objective-C runtime library (libobjc), and other Objective-C
206 related components have been removed from the base system. If you
207 require an Objective-C library, please use one of the available ports.
210 ath(4) has been split into bus- and device- modules. if_ath contains
211 the HAL, the TX rate control and the network device code. if_ath_pci
212 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
213 contains the AHB glue. Users need to load both if_ath_pci and if_ath
214 in order to use ath on everything else.
216 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
217 need to load if_ath and if_ath_pci for ath(4) operation.
220 As part of the replacement of sysinstall, the process of building
221 release media has changed significantly. For details, please re-read
222 release(7), which has been updated to reflect the new build process.
225 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
226 is the last available version under GPLv2. It brings a number of new
227 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
228 4.1 and SSE 4.2), better support for powerpc64, a number of new
229 directives, and lots of other small improvements. See the ChangeLog
230 file in contrib/binutils for the full details.
233 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
234 compliant, and will now use half of hash for authentication.
235 This will break interoperability with all stacks (including all
236 actual FreeBSD versions) who implement
237 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
239 The only workaround with such peers is to use another HMAC
240 algorithm for IPsec ("phase 2") authentication.
243 Remove the uio_yield prototype and symbol. This function has
244 been misnamed since it was introduced and should not be
245 globally exposed with this name. The equivalent functionality
246 is now available using kern_yield(curthread->td_user_pri).
247 The function remains undocumented.
250 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
251 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
252 for scalar sysctls is defined but disabled. Code that needs
253 UQUAD to pass the type checking that must compile on older
254 systems where the define is not present can check against
255 __FreeBSD_version >= 900030.
257 The system dialog(1) has been replaced with a new version previously
258 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
259 with the previous version, but the libdialog associated with it has
260 a largely incompatible API. As such, the original version of libdialog
261 will be kept temporarily as libodialog, until its base system consumers
262 are replaced or updated. Bump __FreeBSD_version to 900030.
265 If you are trying to run make universe on a -stable system, and you get
266 the following warning:
267 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
268 unknown. config(8) likely too old."
269 or something similar to it, then you must upgrade your -stable system
270 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
271 stable/8) or build the config from the latest stable/8 branch and
272 install it on your system.
274 Prior to this date, building a current universe on 8-stable system from
275 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
276 error in the first kernel build phase. A new config on those old
277 systems will fix that problem for older versions of -current.
280 The TCP stack has been modified to allow Khelp modules to interact with
281 it via helper hook points and store per-connection data in the TCP
282 control block. Bump __FreeBSD_version to 900029. User space tools that
283 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
287 Generic IEEE 802.3 annex 31B full duplex flow control support has been
288 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
289 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
290 to take advantage of it instead of using custom implementations. This
291 means that these drivers now no longer unconditionally advertise
292 support for flow control but only do so if flow control is a selected
293 media option. This was implemented in the generic support that way in
294 order to allow flow control to be switched on and off via ifconfig(8)
295 with the PHY specific default to typically off in order to protect
296 from unwanted effects. Consequently, if you used flow control with
297 one of the above mentioned drivers you now need to explicitly enable
299 ifconfig bge0 media auto mediaopt flowcontrol
301 Along with the above mentioned changes generic support for setting
302 1000baseT master mode also has been added and brgphy(4), ciphy(4),
303 e1000phy(4) as well as ip1000phy(4) have been converted to take
304 advantage of it. This means that these drivers now no longer take the
305 link0 parameter for selecting master mode but the master media option
306 has to be used instead, for example like in the following:
307 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
309 Selection of master mode now is also available with all other PHY
310 drivers supporting 1000baseT.
313 The TCP stack has received a significant update to add support for
314 modularised congestion control and generally improve the clarity of
315 congestion control decisions. Bump __FreeBSD_version to 900025. User
316 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
317 sockstat) need to be recompiled.
320 The man(1) utility has been replaced by a new version that no longer
321 uses /etc/manpath.config. Please consult man.conf(5) for how to
322 migrate local entries to the new format.
325 The copyright strings printed by login(1) and sshd(8) at the time of a
326 new connection have been removed to follow other operating systems and
330 A workaround for a fixed ld bug has been removed in kernel code,
331 so make sure that your system ld is built from sources after
332 revision 210245 from 2010-07-19 (r211583 if building head kernel
333 on stable/8, r211584 for stable/7; both from 2010-08-21).
334 A symptom of incorrect ld version is different addresses for
335 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
338 The $ipv6_prefer variable in rc.conf(5) has been split into
339 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
341 The $ip6addrctl_policy is a variable to choose a pre-defined
342 address selection policy set by ip6addrctl(8). A value
343 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
346 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
347 flag (see an entry of 20090926) is set on an interface with no
348 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
349 security reason. If you want IPv6 link-local address on all
350 interfaces by default, set this to "YES".
352 The old ipv6_prefer="YES" is equivalent to
353 ipv6_activate_all_interfaces="YES" and
354 ip6addrctl_policy="ipv6_prefer".
357 DTrace has grown support for userland tracing. Due to this, DTrace is
358 now i386 and amd64 only.
359 dtruss(1) is now installed by default on those systems and a new
360 kernel module is needed for userland tracing: fasttrap.
361 No changes to your kernel config file are necessary to enable
362 userland tracing, but you might consider adding 'STRIP=' and
363 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
364 to have informative userland stack traces in DTrace (ustack).
367 The acpi_aiboost(4) driver has been removed in favor of the new
368 aibs(4) driver. You should update your kernel configuration file.
371 BSD grep has been imported to the base system and it is built by
372 default. It is completely BSD licensed, highly GNU-compatible, uses
373 less memory than its GNU counterpart and has a small codebase.
374 However, it is slower than its GNU counterpart, which is mostly
375 noticeable for larger searches, for smaller ones it is measurable
376 but not significant. The reason is complex, the most important factor
377 is that we lack a modern and efficient regex library and GNU
378 overcomes this by optimizing the searches internally. Future work
379 on improving the regex performance is planned, for the meantime,
380 users that need better performance, can build GNU grep instead by
381 setting the WITH_GNU_GREP knob.
384 Due to the import of powerpc64 support, all existing powerpc kernel
385 configuration files must be updated with a machine directive like this:
386 machine powerpc powerpc
388 In addition, an updated config(8) is required to build powerpc kernels
392 A new version of ZFS (version 15) has been merged to -HEAD.
393 This version uses a python library for the following subcommands:
394 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
395 For full functionality of these commands the following port must
396 be installed: sysutils/py-zfs
399 'vm_page's are now hashed by physical address to an array of mutexes.
400 Currently this is only used to serialize access to hold_count. Over
401 time the page queue mutex will be peeled away. This changes the size
402 of pmap on every architecture. And requires all callers of vm_page_hold
403 and vm_page_unhold to be updated.
406 WITH_CTF can now be specified in src.conf (not recommended, there
407 are some problems with static executables), make.conf (would also
408 affect ports which do not use GNU make and do not override the
409 compile targets) or in the kernel config (via "makeoptions
411 When WITH_CTF was specified there before this was silently ignored,
412 so make sure that WITH_CTF is not used in places which could lead
413 to unwanted behavior.
416 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
417 to allow 32-bit compatibility on non-x86 platforms. All kernel
418 configurations on amd64 and ia64 platforms using these options must
419 be modified accordingly.
422 The utmp user accounting database has been replaced with utmpx,
423 the user accounting interface standardized by POSIX.
424 Unfortunately the semantics of utmp and utmpx don't match,
425 making it practically impossible to support both interfaces.
426 The user accounting database is used by tools like finger(1),
427 last(1), talk(1), w(1) and ac(8).
429 All applications in the base system use utmpx. This means only
430 local binaries (e.g. from the ports tree) may still use these
431 utmp database files. These applications must be rebuilt to make
434 After the system has been upgraded, it is safe to remove the old
435 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
436 assuming their contents is of no importance anymore. Old wtmp
437 databases can only be used by last(1) and ac(8) after they have
438 been converted to the new format using wtmpcvt(1).
441 Introduce the kernel thread "deadlock resolver" (which can be enabled
442 via the DEADLKRES option, see NOTES for more details) and the
443 sleepq_type() function for sleepqueues.
446 The rc.firewall and rc.firewall6 were unified, and
447 rc.firewall6 and rc.d/ip6fw were removed.
448 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
449 variables are obsoleted. Instead, the following new rc
450 variables are added to rc.d/ipfw:
452 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
453 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
454 firewall_simple_onet_ipv6, firewall_trusted_ipv6
456 The meanings correspond to the relevant IPv4 variables.
462 The default terminal emulation for syscons(4) has been changed
463 from cons25 to xterm on all platforms except pc98. This means
464 that the /etc/ttys file needs to be updated to ensure correct
465 operation of applications on the console.
467 The terminal emulation style can be toggled per window by using
468 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
469 options can be used to change the compile-time default back to
472 To prevent graphical artifacts, make sure the TERM environment
473 variable is set to match the terminal emulation that is being
474 performed by syscons(4).
477 The layout of the structure ieee80211req_scan_result has changed.
478 Applications that require wireless scan results (e.g. ifconfig(8))
479 from net80211 need to be recompiled.
481 Applications such as wpa_supplicant(8) may require a full world
482 build without using NO_CLEAN in order to get synchronized with the
486 The iwn(4) driver has been updated to support the 5000 and 5150 series.
487 There's one kernel module for each firmware. Adding "device iwnfw"
488 to the kernel configuration file means including all three firmware
489 images inside the kernel. If you want to include just the one for
490 your wireless card, use the devices iwn4965fw, iwn5000fw or
494 The rc.d/network_ipv6, IPv6 configuration script has been integrated
495 into rc.d/netif. The changes are the following:
497 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
498 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
499 Note that both variables need the "inet6" keyword at the head.
501 Do not set $ipv6_network_interfaces manually if you do not
502 understand what you are doing. It is not needed in most cases.
504 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
507 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
508 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
510 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
511 all configured interfaces have "inet6 accept_rtadv" in the
512 $ifconfig_IF_ipv6. These are for backward compatibility.
514 3. A new variable $ipv6_prefer has been added. If NO, IPv6
515 functionality of interfaces with no corresponding
516 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
517 and the default address selection policy of ip6addrctl(8)
518 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
519 Note that if you want to configure IPv6 functionality on the
520 disabled interfaces after boot, first you need to clear the flag by
521 using ifconfig(8) like:
523 ifconfig em0 inet6 -ifdisabled
525 If YES, the default address selection policy is set as
528 The default value of $ipv6_prefer is NO.
530 4. If your system need to receive Router Advertisement messages,
531 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
532 scripts automatically invoke rtsol(8) when the interface becomes
533 UP. The Router Advertisement messages are used for SLAAC
534 (State-Less Address AutoConfiguration).
537 802.11s D3.03 support was committed. This is incompatible with the
538 previous code, which was based on D3.0.
541 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
542 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
543 control whether accepting Router Advertisement messages or not.
544 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
545 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
546 The ifconfig(8) utility now supports these flags.
549 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
550 mount(8) and -a option for df(1) to see them.
553 The old tunable hw.bus.devctl_disable has been superseded by
554 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
555 replaced by hw.bus.devctl_queue=0. The default for this new tunable
559 Remove the option STOP_NMI. The default action is now to use NMI only
560 for KDB via the newly introduced function stop_cpus_hard() and
561 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
564 The stable/8 branch created in subversion. This corresponds to the
565 RELENG_8 branch in CVS.
568 Bump the shared library version numbers for all libraries that do not
569 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
570 __FreeBSD_version to 800105.
573 Due to changes in the implementation of virtual network stack support,
574 all network-related kernel modules must be recompiled. As this change
575 breaks the ABI, bump __FreeBSD_version to 800104.
578 The TOE interface to the TCP syncache has been modified to remove
579 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
580 The cxgb driver is the only TOE consumer affected by this change, and
581 needs to be recompiled along with the kernel. As this change breaks
582 the ABI, bump __FreeBSD_version to 800103.
585 Padding has been added to struct tcpcb, sackhint and tcpstat in
586 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
587 maintaining the ABI. However, this change breaks the ABI, so bump
588 __FreeBSD_version to 800102. User space tools that rely on the size of
589 any of these structs (e.g. sockstat) need to be recompiled.
592 The NFS_LEGACYRPC option has been removed along with the old kernel
593 RPC implementation that this option selected. Kernel configurations
594 may need to be adjusted.
597 The network interface device nodes at /dev/net/<interface> have been
598 removed. All ioctl operations can be performed the normal way using
599 routing sockets. The kqueue functionality can generally be replaced
600 with routing sockets.
603 The documentation from the FreeBSD Documentation Project (Handbook,
604 FAQ, etc.) is now installed via packages by sysinstall(8) and under
605 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
608 The ABI of various structures related to the SYSV IPC API have been
609 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
610 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
614 Layout of struct vnet has changed as routing related variables were
615 moved to their own Vimage module. Modules need to be recompiled. Bump
616 __FreeBSD_version to 800099.
619 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
620 respectively. As long as no more than 16 groups per process are used,
621 no changes should be visible. When more than 16 groups are used, old
622 binaries may fail if they call getgroups() or getgrouplist() with
623 statically sized storage. Recompiling will work around this, but
624 applications should be modified to use dynamically allocated storage
625 for group arrays as POSIX.1-2008 does not cap an implementation's
626 number of supported groups at NGROUPS_MAX+1 as previous versions did.
628 NFS and portalfs mounts may also be affected as the list of groups is
629 truncated to 16. Users of NFS who use more than 16 groups, should
630 take care that negative group permissions are not used on the exported
631 file systems as they will not be reliable unless a GSSAPI based
632 authentication method is used.
635 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
636 option compiles in the support for adaptive spinning for lockmgrs
637 which want to enable it. The lockinit() function now accepts the flag
638 LK_ADAPTIVE in order to make the lock object subject to adaptive
639 spinning when both held in write and read mode.
642 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
643 changed. User applications that use this ioctl need to be rebuilt.
646 The layout of struct thread has changed. Kernel and modules need to
650 The layout of structs ifnet, domain, protosw and vnet_net has changed.
651 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
654 window(1) has been removed from the base system. It can now be
655 installed from ports. The port is called misc/window.
658 The way we are storing and accessing `routing table' entries has
659 changed. Programs reading the FIB, like netstat, need to be
663 A new netisr implementation has been added for FreeBSD 8. Network
664 file system modules, such as igmp, ipdivert, and others, should be
666 Bump __FreeBSD_version to 800096.
669 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
673 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
674 Bump __FreeBSD_version to 800094.
677 Add mnt_xflag field to 'struct mount'. File system modules need to be
679 Bump __FreeBSD_version to 800093.
682 The compiling option ADAPTIVE_SX has been retired while it has been
683 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
684 The KPI for sx_init_flags() changes as accepting flags:
685 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
686 been introduced in order to handle the reversed logic.
687 Bump __FreeBSD_version to 800092.
690 Add support for hierarchical jails. Remove global securelevel.
691 Bump __FreeBSD_version to 800091.
694 The layout of struct vnet_net has changed, therefore modules
696 Bump __FreeBSD_version to 800090.
699 The newly imported zic(8) produces a new format in the output. Please
700 run tzsetup(8) to install the newly created data to /etc/localtime.
703 The sysctl tree for the usb stack has renamed from hw.usb2.* to
704 hw.usb.* and is now consistent again with previous releases.
707 802.11 monitor mode support was revised and driver api's were changed.
708 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
709 of DLT_IEEE802_11. No user-visible data structures were changed but
710 applications that use DLT_IEEE802_11 may require changes.
711 Bump __FreeBSD_version to 800088.
714 The layout of the following structs has changed: sysctl_oid,
715 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
716 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
717 panics may be experienced. World rebuild is required for
718 correctly checking networking state from userland.
719 Bump __FreeBSD_version to 800085.
722 MLDv2 and Source-Specific Multicast (SSM) have been merged
723 to the IPv6 stack. VIMAGE hooks are in but not yet used.
724 The implementation of SSM within FreeBSD's IPv6 stack closely
725 follows the IPv4 implementation.
727 For kernel developers:
729 * The most important changes are that the ip6_output() and
730 ip6_input() paths no longer take the IN6_MULTI_LOCK,
731 and this lock has been downgraded to a non-recursive mutex.
733 * As with the changes to the IPv4 stack to support SSM, filtering
734 of inbound multicast traffic must now be performed by transport
735 protocols within the IPv6 stack. This does not apply to TCP and
736 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
738 * The KPIs used by IPv6 multicast are similar to those used by
739 the IPv4 stack, with the following differences:
740 * im6o_mc_filter() is analogous to imo_multicast_filter().
741 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
742 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
743 * IN6_LOOKUP_MULTI() has been deprecated and removed.
744 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
745 for MLDv1 have an additional 'timer' argument which is used to
746 jitter the initial membership report for the solicited-node
747 multicast membership on-link.
748 * This is not strictly needed for MLDv2, which already jitters
749 its report transmissions. However, the 'timer' argument is
750 preserved in case MLDv1 is active on the interface.
752 * The KAME linked-list based IPv6 membership implementation has
753 been refactored to use a vector similar to that used by the IPv4
755 Code which maintains a list of its own multicast memberships
756 internally, e.g. carp, has been updated to reflect the new
759 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
760 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
761 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
762 implementation constraint which needs to be addressed in HEAD.
764 For application developers:
766 * The changes are broadly similar to those made for the IPv4
769 * The use of IPv4 and IPv6 multicast socket options on the same
770 socket, using mapped addresses, HAS NOT been tested or supported.
772 * There are a number of issues with the implementation of various
773 IPv6 multicast APIs which need to be resolved in the API surface
774 before the implementation is fully compatible with KAME userland
775 use, and these are mostly to do with interface index treatment.
777 * The literature available discusses the use of either the delta / ASM
778 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
779 using setsourcefilter(3)/getsourcefilter(3). For more information
780 please refer to RFC 3768, 'Socket Interface Extensions for
781 Multicast Source Filters'.
783 * Applications which use the published RFC 3678 APIs should be fine.
785 For systems administrators:
787 * The mtest(8) utility has been refactored to support IPv6, in
788 addition to IPv4. Interface addresses are no longer accepted
789 as arguments, their names must be used instead. The utility
790 will map the interface name to its first IPv4 address as
791 returned by getifaddrs(3).
793 * The ifmcstat(8) utility has also been updated to print the MLDv2
794 endpoint state and source filter lists via sysctl(3).
796 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
797 loopback of IPv6 multicast datagrams by default; it defaults to 1
798 to preserve the existing behaviour. Disabling multicast loopback is
799 recommended for optimal system performance.
801 * The IPv6 MROUTING code has been changed to examine this sysctl
802 instead of attempting to perform a group lookup before looping
803 back forwarded datagrams.
805 Bump __FreeBSD_version to 800084.
808 Implement low-level Bluetooth HCI API.
809 Bump __FreeBSD_version to 800083.
812 The layout of struct malloc_type, used by modules to register new
813 memory allocation types, has changed. Most modules will need to
814 be rebuilt or panics may be experienced.
815 Bump __FreeBSD_version to 800081.
818 Anticipate overflowing inp_flags - add inp_flags2.
819 This changes most offsets in inpcb, so checking v4 connection
820 state will require a world rebuild.
821 Bump __FreeBSD_version to 800080.
824 Add an llentry to struct route and struct route_in6. Modules
825 embedding a struct route will need to be recompiled.
826 Bump __FreeBSD_version to 800079.
829 The size of rt_metrics_lite and by extension rtentry has changed.
830 Networking administration apps will need to be recompiled.
831 The route command now supports show as an alias for get, weighting
832 of routes, sticky and nostick flags to alter the behavior of stateful
834 Bump __FreeBSD_version to 800078.
837 Do not use Giant for kbdmux(4) locking. This is wrong and
838 apparently causing more problems than it solves. This will
839 re-open the issue where interrupt handlers may race with
840 kbdmux(4) in polling mode. Typical symptoms include (but
841 not limited to) duplicated and/or missing characters when
842 low level console functions (such as gets) are used while
843 interrupts are enabled (for example geli password prompt,
844 mountroot prompt etc.). Disabling kbdmux(4) may help.
847 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
848 kernel modules referencing any of the above need to be recompiled.
849 Bump __FreeBSD_version to 800075.
852 GEOM_PART has become the default partition slicer for storage devices,
853 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
854 introduces some changes:
856 MSDOS/EBR: the devices created from MSDOS extended partition entries
857 (EBR) can be named differently than with GEOM_MBR and are now symlinks
858 to devices with offset-based names. fstabs may need to be modified.
860 BSD: the "geometry does not match label" warning is harmless in most
861 cases but it points to problems in file system misalignment with
862 disk geometry. The "c" partition is now implicit, covers the whole
863 top-level drive and cannot be (mis)used by users.
865 General: Kernel dumps are now not allowed to be written to devices
866 whose partition types indicate they are meant to be used for file
867 systems (or, in case of MSDOS partitions, as something else than
870 Most of these changes date approximately from 200812.
873 The uscanner(4) driver has been removed from the kernel. This follows
874 Linux removing theirs in 2.6 and making libusb the default interface
878 The multicast forwarding code has been cleaned up. netstat(1)
879 only relies on KVM now for printing bandwidth upcall meters.
880 The IPv4 and IPv6 modules are split into ip_mroute_mod and
881 ip6_mroute_mod respectively. The config(5) options for statically
882 compiling this code remain the same, i.e. 'options MROUTING'.
885 Support for the IFF_NEEDSGIANT network interface flag has been
886 removed, which means that non-MPSAFE network device drivers are no
887 longer supported. In particular, if_ar, if_sr, and network device
888 drivers from the old (legacy) USB stack can no longer be built or
892 POSIX.1 Native Language Support (NLS) has been enabled in libc and
893 a bunch of new language catalog files have also been added.
894 This means that some common libc messages are now localized and
895 they depend on the LC_MESSAGES environmental variable.
898 The k8temp(4) driver has been renamed to amdtemp(4) since
899 support for Family 10 and Family 11 CPU families was added.
902 IGMPv3 and Source-Specific Multicast (SSM) have been merged
903 to the IPv4 stack. VIMAGE hooks are in but not yet used.
905 For kernel developers, the most important changes are that the
906 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
907 and this lock has been downgraded to a non-recursive mutex.
909 Transport protocols (UDP, Raw IP) are now responsible for filtering
910 inbound multicast traffic according to group membership and source
911 filters. The imo_multicast_filter() KPI exists for this purpose.
912 Transports which do not use multicast (SCTP, TCP) already reject
913 multicast by default. Forwarding and receive performance may improve
914 as a mutex acquisition is no longer needed in the ip_input()
915 low-level input path. in_addmulti() and in_delmulti() are shimmed
916 to new KPIs which exist to support SSM in-kernel.
918 For application developers, it is recommended that loopback of
919 multicast datagrams be disabled for best performance, as this
920 will still cause the lock to be taken for each looped-back
921 datagram transmission. The net.inet.ip.mcast.loop sysctl may
922 be tuned to 0 to disable loopback by default; it defaults to 1
923 to preserve the existing behaviour.
925 For systems administrators, to obtain best performance with
926 multicast reception and multiple groups, it is always recommended
927 that a card with a suitably precise hash filter is used. Hash
928 collisions will still result in the lock being taken within the
929 transport protocol input path to check group membership.
931 If deploying FreeBSD in an environment with IGMP snooping switches,
932 it is recommended that the net.inet.igmp.sendlocal sysctl remain
933 enabled; this forces 224.0.0.0/24 group membership to be announced
936 The size of 'struct igmpstat' has changed; netstat needs to be
937 recompiled to reflect this.
938 Bump __FreeBSD_version to 800070.
941 libusb20.so.1 is now installed as libusb.so.1 and the ports system
942 updated to use it. This requires a buildworld/installworld in order to
943 update the library and dependencies (usbconfig, etc). Its advisable to
944 rebuild all ports which uses libusb. More specific directions are given
945 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
946 libusb are no longer required and can be removed.
949 A workaround is committed to allow the creation of System V shared
950 memory segment of size > 2 GB on the 64-bit architectures.
951 Due to a limitation of the existing ABI, the shm_segsz member
952 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
953 wrong for large segments. Note that limits must be explicitly
954 raised to allow such segments to be created.
957 The layout of struct ifnet has changed, requiring a rebuild of all
958 network device driver modules.
961 The /dev handling for the new USB stack has changed, a
962 buildworld/installworld is required for libusb20.
965 The new USB2 stack has now been permanently moved in and all kernel and
966 module names reverted to their previous values (eg, usb, ehci, ohci,
967 ums, ...). The old usb stack can be compiled in by prefixing the name
968 with the letter 'o', the old usb modules have been removed.
969 Updating entry 20090216 for xorg and 20090215 for libmap may still
973 The rc.conf(5) option if_up_delay has been renamed to
974 defaultroute_delay to better reflect its purpose. If you have
975 customized this setting in /etc/rc.conf you need to update it to
979 xorg 7.4 wants to configure its input devices via hald which does not
980 yet work with USB2. If the keyboard/mouse does not work in xorg then
982 Option "AllowEmptyInput" "off"
983 to your ServerLayout section. This will cause X to use the configured
984 kbd and mouse sections from your xorg.conf.
987 The GENERIC kernels for all architectures now default to the new USB2
988 stack. No kernel config options or code have been removed so if a
989 problem arises please report it and optionally revert to the old USB
990 stack. If you are loading USB kernel modules or have a custom kernel
991 that includes GENERIC then ensure that usb names are also changed over,
992 eg uftdi -> usb2_serial_ftdi.
994 Older programs linked against the ports libusb 0.1 need to be
995 redirected to the new stack's libusb20. /etc/libmap.conf can
997 # Map old usb library to new one for usb2 stack
998 libusb-0.1.so.8 libusb20.so.1
1001 All USB ethernet devices now attach as interfaces under the name ueN
1002 (eg. ue0). This is to provide a predictable name as vendors often
1003 change usb chipsets in a product without notice.
1006 The ichsmb(4) driver has been changed to require SMBus slave
1007 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1008 All of the other SMBus controller drivers require left-justified
1009 slave addresses, so this change makes all the drivers provide the
1013 INET6 statistics (struct ip6stat) was updated.
1014 netstat(1) needs to be recompiled.
1017 NTFS has been removed from GENERIC kernel on amd64 to match
1018 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1019 will load ntfs.ko module automatically when NTFS support is
1020 actually needed, unless ntfs.ko is not installed or security
1021 level prohibits loading kernel modules. If either is the case,
1022 "options NTFS" has to be added into kernel config.
1025 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1026 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1027 800061. User space tools that rely on the size of struct tcpcb in
1028 tcp_var.h (e.g. sockstat) need to be recompiled.
1031 ng_tty(4) module updated to match the new TTY subsystem.
1032 Due to API change, user-level applications must be updated.
1033 New API support added to mpd5 CVS and expected to be present
1034 in next mpd5.3 release.
1037 With __FreeBSD_version 800060 the makefs tool is part of
1038 the base system (it was a port).
1041 The afdata and ifnet locks have been changed from mutexes to
1042 rwlocks, network modules will need to be re-compiled.
1045 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1046 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1047 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1048 architecture and 40 bytes on 64-bit architecture. The userland
1049 applications "arp" and "ndp" have been updated accordingly.
1050 The output from "netstat -r" shows only routing entries and
1051 none of the L2 information.
1054 __FreeBSD_version 800057 marks the switchover from the
1055 binary ath hal to source code. Users must add the line:
1057 options AH_SUPPORT_AR5416
1059 to their kernel config files when specifying:
1063 The ath_hal module no longer exists; the code is now compiled
1064 together with the driver in the ath module. It is now
1065 possible to tailor chip support (i.e. reduce the set of chips
1066 and thereby the code size); consult ath_hal(4) for details.
1069 __FreeBSD_version 800054 adds memory barriers to
1070 <machine/atomic.h>, new interfaces to ifnet to facilitate
1071 multiple hardware transmit queues for cards that support
1072 them, and a lock-less ring-buffer implementation to
1073 enable drivers to more efficiently manage queueing of
1077 A new version of ZFS (version 13) has been merged to -HEAD.
1078 This version has zpool attribute "listsnapshots" off by
1079 default, which means "zfs list" does not show snapshots,
1080 and is the same as Solaris behavior.
1083 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1086 The uhci, ohci, ehci and slhci USB Host controller drivers have
1087 been put into separate modules. If you load the usb module
1088 separately through loader.conf you will need to load the
1089 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1090 controller add the following to loader.conf:
1096 The ABI used by the PMC toolset has changed. Please keep
1097 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1101 atapci kernel module now includes only generic PCI ATA
1102 driver. AHCI driver moved to ataahci kernel module.
1103 All vendor-specific code moved into separate kernel modules:
1104 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1105 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1106 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1107 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1110 The TTY subsystem of the kernel has been replaced by a new
1111 implementation, which provides better scalability and an
1112 improved driver model. Most common drivers have been migrated to
1113 the new TTY subsystem, while others have not. The following
1114 drivers have not yet been ported to the new TTY layer:
1117 cy, digi, rc, rp, sio
1123 ng_h4, ng_tty, ppp, sl, snp
1125 Adding these drivers to your kernel configuration file shall
1126 cause compilation to fail.
1129 ntpd has been upgraded to 4.2.4p5.
1132 OpenSSH has been upgraded to 5.1p1.
1134 For many years, FreeBSD's version of OpenSSH preferred DSA
1135 over RSA for host and user authentication keys. With this
1136 upgrade, we've switched to the vendor's default of RSA over
1137 DSA. This may cause upgraded clients to warn about unknown
1138 host keys even for previously known hosts. Users should
1139 follow the usual procedure for verifying host keys before
1140 accepting the RSA key.
1142 This can be circumvented by setting the "HostKeyAlgorithms"
1143 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1146 Please note that the sequence of keys offered for
1147 authentication has been changed as well. You may want to
1148 specify IdentityFile in a different order to revert this
1152 The sio(4) driver has been removed from the i386 and amd64
1153 kernel configuration files. This means uart(4) is now the
1154 default serial port driver on those platforms as well.
1156 To prevent collisions with the sio(4) driver, the uart(4) driver
1157 uses different names for its device nodes. This means the
1158 onboard serial port will now most likely be called "ttyu0"
1159 instead of "ttyd0". You may need to reconfigure applications to
1160 use the new device names.
1162 When using the serial port as a boot console, be sure to update
1163 /boot/device.hints and /etc/ttys before booting the new kernel.
1164 If you forget to do so, you can still manually specify the hints
1165 at the loader prompt:
1167 set hint.uart.0.at="isa"
1168 set hint.uart.0.port="0x3F8"
1169 set hint.uart.0.flags="0x10"
1170 set hint.uart.0.irq="4"
1174 The gpt(8) utility has been removed. Use gpart(8) to partition
1178 The version that Linuxulator emulates was changed from 2.4.2
1179 to 2.6.16. If you experience any problems with Linux binaries
1180 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1181 if it fixes the problem contact emulation mailing list.
1184 ISDN4BSD (I4B) was removed from the src tree. You may need to
1185 update a your kernel configuration and remove relevant entries.
1188 I have checked in code to support multiple routing tables.
1189 See the man pages setfib(1) and setfib(2).
1190 This is a hopefully backwards compatible version,
1191 but to make use of it you need to compile your kernel
1192 with options ROUTETABLES=2 (or more up to 16).
1195 The 802.11 wireless support was redone to enable multi-bss
1196 operation on devices that are capable. The underlying device
1197 is no longer used directly but instead wlanX devices are
1198 cloned with ifconfig. This requires changes to rc.conf files.
1199 For example, change:
1200 ifconfig_ath0="WPA DHCP"
1203 ifconfig_wlan0="WPA DHCP"
1204 see rc.conf(5) for more details. In addition, mergemaster of
1205 /etc/rc.d is highly recommended. Simultaneous update of userland
1206 and kernel wouldn't hurt either.
1208 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1209 modules were merged into the base wlan module. All references
1210 to these modules (e.g. in kernel config files) must be removed.
1213 psm(4) has gained write(2) support in native operation level.
1214 Arbitrary commands can be written to /dev/psm%d and status can
1215 be read back from it. Therefore, an application is responsible
1216 for status validation and error recovery. It is a no-op in
1217 other operation levels.
1220 Support for KSE threading has been removed from the kernel. To
1221 run legacy applications linked against KSE libmap.conf may
1222 be used. The following libmap.conf may be used to ensure
1223 compatibility with any prior release:
1225 libpthread.so.1 libthr.so.1
1226 libpthread.so.2 libthr.so.2
1227 libkse.so.3 libthr.so.3
1230 The layout of struct vmspace has changed. This affects libkvm
1231 and any executables that link against libkvm and use the
1232 kvm_getprocs() function. In particular, but not exclusively,
1233 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1234 The effects are minimal, but it's advisable to upgrade world
1238 The latest em driver no longer has support in it for the
1239 82575 adapter, this is now moved to the igb driver. The
1240 split was done to make new features that are incompatible
1241 with older hardware easier to do.
1244 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1245 likewise the kernel option is now GEOM_LINUX_LVM.
1248 The default NFS mount mode has changed from UDP to TCP for
1249 increased reliability. If you rely on (insecurely) NFS
1250 mounting across a firewall you may need to update your
1254 Belatedly note the addition of m_collapse for compacting
1258 The fts(3) structures have been changed to use adequate
1259 integer types for their members and so to be able to cope
1260 with huge file trees. The old fts(3) ABI is preserved
1261 through symbol versioning in libc, so third-party binaries
1262 using fts(3) should still work, although they will not take
1263 advantage of the extended types. At the same time, some
1264 third-party software might fail to build after this change
1265 due to unportable assumptions made in its source code about
1266 fts(3) structure members. Such software should be fixed
1267 by its vendor or, in the worst case, in the ports tree.
1268 FreeBSD_version 800015 marks this change for the unlikely
1269 case that a portable fix is impossible.
1272 To upgrade to -current after this date, you must be running
1273 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1274 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1277 The ADAPTIVE_GIANT kernel option has been retired because its
1278 functionality is the default now.
1281 The AT keyboard emulation of sunkbd(4) has been turned on
1282 by default. In order to make the special symbols of the Sun
1283 keyboards driven by sunkbd(4) work under X these now have
1284 to be configured the same way as Sun USB keyboards driven
1285 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1287 Option "XkbLayout" "us"
1288 Option "XkbRules" "xorg"
1289 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1292 It has been decided that it is desirable to provide ABI
1293 backwards compatibility to the FreeBSD 4/5/6 versions of the
1294 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1295 broken with the introduction of PCI domain support (see the
1296 20070930 entry). Unfortunately, this required the ABI of
1297 PCIOCGETCONF to be broken again in order to be able to
1298 provide backwards compatibility to the old version of that
1299 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1300 again. As for prominent ports this affects neither pciutils
1301 nor xorg-server this time, the hal port needs to be rebuilt
1305 The misnamed kthread_create() and friends have been renamed
1306 to kproc_create() etc. Many of the callers already
1307 used kproc_start()..
1308 I will return kthread_create() and friends in a while
1309 with implementations that actually create threads, not procs.
1310 Renaming corresponds with version 800002.
1319 Avoid using make -j when upgrading. While generally safe, there are
1320 sometimes problems using -j to upgrade. If your upgrade fails with
1321 -j, please try again without -j. From time to time in the past there
1322 have been problems using -j with buildworld and/or installworld. This
1323 is especially true when upgrading between "distant" versions (eg one
1324 that cross a major release boundary or several minor releases, or when
1325 several months have passed on the -current branch).
1327 Sometimes, obscure build problems are the result of environment
1328 poisoning. This can happen because the make utility reads its
1329 environment when searching for values for global variables. To run
1330 your build attempts in an "environmental clean room", prefix all make
1331 commands with 'env -i '. See the env(1) manual page for more details.
1333 When upgrading from one major version to another it is generally best
1334 to upgrade to the latest code in the currently installed branch first,
1335 then do an upgrade to the new branch. This is the best-tested upgrade
1336 path, and has the highest probability of being successful. Please try
1337 this approach before reporting problems with a major version upgrade.
1341 When upgrading the boot ZFS pool to a new version, always follow
1344 1.) recompile and reinstall the ZFS boot loader and boot block
1345 (this is part of "make buildworld" and "make installworld")
1347 2.) update the ZFS boot block on your boot drive
1349 The following example updates the ZFS boot block on the first
1350 partition (freebsd-boot) of a GPT partitioned drive ad0:
1351 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1353 Non-boot pools do not need these updates.
1357 If you are updating from a prior version of FreeBSD (even one just
1358 a few days old), you should follow this procedure. It is the most
1359 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1361 make kernel-toolchain
1362 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1363 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1365 To test a kernel once
1366 ---------------------
1367 If you just want to boot a kernel once (because you are not sure
1368 if it works, or if you want to boot a known bad kernel to provide
1369 debugging information) run
1370 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1371 nextboot -k testkernel
1373 To just build a kernel when you know that it won't mess you up
1374 --------------------------------------------------------------
1375 This assumes you are already running a CURRENT system. Replace
1376 ${arch} with the architecture of your machine (e.g. "i386",
1377 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1379 cd src/sys/${arch}/conf
1380 config KERNEL_NAME_HERE
1381 cd ../compile/KERNEL_NAME_HERE
1386 If this fails, go to the "To build a kernel" section.
1388 To rebuild everything and install it on the current system.
1389 -----------------------------------------------------------
1390 # Note: sometimes if you are running current you gotta do more than
1391 # is listed here if you are upgrading from a really old current.
1393 <make sure you have good level 0 dumps>
1395 make kernel KERNCONF=YOUR_KERNEL_HERE
1397 <reboot in single user> [3]
1405 To cross-install current onto a separate partition
1406 --------------------------------------------------
1407 # In this approach we use a separate partition to hold
1408 # current's root, 'usr', and 'var' directories. A partition
1409 # holding "/", "/usr" and "/var" should be about 2GB in
1412 <make sure you have good level 0 dumps>
1415 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1416 <maybe newfs current's root partition>
1417 <mount current's root partition on directory ${CURRENT_ROOT}>
1418 make installworld DESTDIR=${CURRENT_ROOT}
1419 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1420 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1421 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1422 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1423 <reboot into current>
1424 <do a "native" rebuild/install as described in the previous section>
1425 <maybe install compatibility libraries from ports/misc/compat*>
1429 To upgrade in-place from 8.x-stable to current
1430 ----------------------------------------------
1431 <make sure you have good level 0 dumps>
1433 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1435 <reboot in single user> [3]
1442 Make sure that you've read the UPDATING file to understand the
1443 tweaks to various things you need. At this point in the life
1444 cycle of current, things change often and you are on your own
1445 to cope. The defaults can also change, so please read ALL of
1446 the UPDATING entries.
1448 Also, if you are tracking -current, you must be subscribed to
1449 freebsd-current@freebsd.org. Make sure that before you update
1450 your sources that you have read and understood all the recent
1451 messages there. If in doubt, please track -stable which has
1452 much fewer pitfalls.
1454 [1] If you have third party modules, such as vmware, you
1455 should disable them at this point so they don't crash your
1458 [3] From the bootblocks, boot -s, and then do
1463 adjkerntz -i # if CMOS is wall time
1464 Also, when doing a major release upgrade, it is required that
1465 you boot into single user mode to do the installworld.
1467 [4] Note: This step is non-optional. Failure to do this step
1468 can result in a significant reduction in the functionality of the
1469 system. Attempting to do it by hand is not recommended and those
1470 that pursue this avenue should read this file carefully, as well
1471 as the archives of freebsd-current and freebsd-hackers mailing lists
1472 for potential gotchas. The -U option is also useful to consider.
1473 See mergemaster(8) for more information.
1475 [5] Usually this step is a noop. However, from time to time
1476 you may need to do this if you get unknown user in the following
1477 step. It never hurts to do it all the time. You may need to
1478 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1479 install) after the buildworld before this step if you last updated
1480 from current before 20020224 or from -stable before 20020408.
1482 [6] This only deletes old files and directories. Old libraries
1483 can be deleted by "make delete-old-libs", but you have to make
1484 sure that no program is using those libraries anymore.
1486 [8] In order to have a kernel that can run the 4.x binaries needed to
1487 do an installworld, you must include the COMPAT_FREEBSD4 option in
1488 your kernel. Failure to do so may leave you with a system that is
1489 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1490 required to run the 5.x binaries on more recent kernels. And so on
1491 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1493 Make sure that you merge any new devices from GENERIC since the
1494 last time you updated your kernel config file.
1496 [9] When checking out sources, you must include the -P flag to have
1497 cvs prune empty directories.
1499 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1500 "?=" instead of the "=" assignment operator, so that buildworld can
1501 override the CPUTYPE if it needs to.
1503 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1504 not on the command line, or in /etc/make.conf. buildworld will
1505 warn if it is improperly defined.
1508 This file contains a list, in reverse chronological order, of major
1509 breakages in tracking -current. It is not guaranteed to be a complete
1510 list of such breakages, and only contains entries since October 10, 2007.
1511 If you need to see UPDATING entries from before that date, you will need
1512 to fetch an UPDATING file from an older FreeBSD release.
1514 Copyright information:
1516 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1518 Redistribution, publication, translation and use, with or without
1519 modification, in full or in part, in any form or format of this
1520 document are permitted without further permission from the author.
1522 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1523 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1524 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1525 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1526 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1527 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1528 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1529 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1530 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1531 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1532 POSSIBILITY OF SUCH DAMAGE.
1534 Contact Warner Losh if you have any questions about your use of