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 20130726: p5 FreeBSD-SA-13:07.bind FreeBSD-SA-13:08.nfsserver
13 Fix Denial of Service vulnerability in named(8). [13:07]
15 Fix a bug that allows remote client bypass the normal
16 access checks when when -network or -host restrictions are
17 used at the same time with -mapall. [13:08]
19 20130618: p4 FreeBSD-SA-13:06.mmap
20 Fix a bug that allowed a tracing process (e.g. gdb) to write
21 to a memory-mapped file in the traced process's address space
22 even if neither the traced process nor the tracing process had
23 write access to that file.
25 20130429: p3 FreeBSD-SA-13:05.nfsserver
26 Fix a bug that allows NFS clients to issue READDIR on files.
28 20130402: p2 FreeBSD-SA-13:03.openssl FreeBSD-SA-13:04.bind
29 Fix multiple vulnerabilities in OpenSSL.
31 Fix Denial of Service vulnerability in named(8).
33 20130218: p1 FreeBSD-SA-13:01.bind FreeBSD-SA-13:02.libc
34 Fix Denial of Service vulnerability in named(8) with DNS64.
36 Fix Denial of Service vulnerability in libc's glob(3) functionality.
42 The sparc64 ZFS loader has been changed to no longer try to auto-
43 detect ZFS providers based on diskN aliases but now requires these
44 to be explicitly listed in the OFW boot-device environment variable.
47 Now unix domain sockets behave "as expected" on nullfs(5). Previously
48 nullfs(5) did not pass through all behaviours to the underlying layer,
49 as a result if we bound to a socket on the lower layer we could connect
50 only to the lower path; if we bound to the upper layer we could connect
51 only to the upper path. The new behavior is one can connect to both the
52 lower and the upper paths regardless what layer path one binds to.
55 The acpi_wmi(4) status device /dev/wmistat has been renamed to
59 A new VOP_ADVISE() was added to support posix_fadvise(2). All
60 filesystem modules must be recompiled.
63 The interface of the VOP_VPTOCNP(9) changed, now the returned
64 vnode shall be referenced, previously it was required to be
65 only held. All in-tree filesystems are converted.
71 The broken amd(4) driver has been replaced with esp(4) in the amd64,
72 i386 and pc98 GENERIC kernel configuration files.
75 This commit modifies vfs_register() so that it uses a hash
76 calculation to set vfc_typenum, which is enabled by default.
77 The first time a system is booted after this change, the
78 vfc_typenum values will change for all file systems. The
79 main effect of this is a change to the NFS server file handles
80 for file systems that use vfc_typenum in their fsid, such as ZFS.
81 It will, however, prevent vfc_typenum from changing when file
82 systems are loaded in a different order for subsequent reboots.
83 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
84 until you are ready to remount all NFS clients after a reboot.
87 Bump the shared library version numbers for libraries that
88 do not use symbol versioning, have changed the ABI compared
89 to stable/8 and which shared library version was not bumped.
90 Done as part of 9.0-RELEASE cycle.
93 During the merge of Capsicum features, the fget(9) KPI was modified.
94 This may require the rebuilding of out-of-tree device drivers --
95 issues have been reported specifically with the nVidia device driver.
96 __FreeBSD_version is bumped to 900041.
98 Also, there is a period between 20110811 and 20110814 where the
99 special devices /dev/{stdin,stdout,stderr} did not work correctly.
100 Building world from a kernel during that window may not work.
103 The packet filter (pf) code has been updated to OpenBSD 4.5.
104 You need to update userland tools to be in sync with kernel.
105 This update breaks backward compatibility with earlier pfsync(4)
106 versions. Care must be taken when updating redundant firewall setups.
109 The following sysctls and tunables are retired on x86 platforms:
111 machdep.hlt_logical_cpus
112 The following sysctl is retired:
113 machdep.hyperthreading_allowed
114 The sysctls were supposed to provide a way to dynamically offline and
115 online selected CPUs on x86 platforms, but the implementation has not
116 been reliable especially with SCHED_ULE scheduler.
117 machdep.hyperthreading_allowed tunable is still available to ignore
118 hyperthreading CPUs at OS level.
119 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
120 where X is an APIC ID of a CPU. Be advised, though, that disabling
121 CPUs in non-uniform fashion will result in non-uniform topology and
122 may lead to sub-optimal system performance with SCHED_ULE, which is
126 cpumask_t type is retired and cpuset_t is used in order to describe
130 Changes to ifconfig(8) for dynamic address family detection mandate
131 that you are running a kernel of 20110525 or later. Make sure to
132 follow the update procedure to boot a new kernel before installing
136 Support for sun4v architecture is officially dropped
139 Several KPI breaking changes have been committed to the mii(4) layer,
140 the PHY drivers and consequently some Ethernet drivers using mii(4).
141 This means that miibus.ko and the modules of the affected Ethernet
142 drivers need to be recompiled.
144 Note to kernel developers: Given that the OUI bit reversion problem
145 was fixed as part of these changes all mii(4) commits related to OUIs,
146 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
147 handling, no longer can be merged verbatim to stable/8 and previous
151 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
152 into their kernel configurations along with 'device pci'.
155 The default NFS client is now the new NFS client, so fstype "newnfs"
156 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
157 Although mounts via fstype "nfs" will usually work without userland
158 changes, it is recommended that the mount(8) and mount_nfs(8)
159 commands be rebuilt from sources and that a link to mount_nfs called
160 mount_oldnfs be created. The new client is compiled into the
161 kernel with "options NFSCL" and this is needed for diskless root
162 file systems. The GENERIC kernel configs have been changed to use
163 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
164 To use the regular/old client, you can "mount -t oldnfs ...". For
165 a diskless root file system, you must also include a line like:
167 vfs.root.mountfrom="oldnfs:"
169 in the boot/loader.conf on the root fs on the NFS server to make
170 a diskless root fs use the old client.
173 The GENERIC kernels for all architectures now default to the new
174 CAM-based ATA stack. It means that all legacy ATA drivers were
175 removed and replaced by respective CAM drivers. If you are using
176 ATA device names in /etc/fstab or other places, make sure to update
177 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
178 where 'Y's are the sequential numbers starting from zero for each type
179 in order of detection, unless configured otherwise with tunables,
180 see cam(4)). There will be symbolic links created in /dev/ to map
181 old adX devices to the respective adaY. They should provide basic
182 compatibility for file systems mounting in most cases, but they do
183 not support old user-level APIs and do not have respective providers
184 in GEOM. Consider using updated management tools with new device names.
186 It is possible to load devices ahci, ata, siis and mvs as modules,
187 but option ATA_CAM should remain in kernel configuration to make ata
188 module work as CAM driver supporting legacy ATA controllers. Device ata
189 still can be used in modular fashion (atacore + ...). Modules atadisk
190 and atapi* are not used and won't affect operation in ATA_CAM mode.
191 Note that to use CAM-based ATA kernel should include CAM devices
192 scbus, pass, da (or explicitly ada), cd and optionally others. All of
193 them are parts of the cam module.
195 ataraid(4) functionality is now supported by the RAID GEOM class.
196 To use it you can load geom_raid kernel module and use graid(8) tool
197 for management. Instead of /dev/arX device names, use /dev/raid/rX.
199 No kernel config options or code have been removed, so if a problem
200 arises, please report it and optionally revert to the old ATA stack.
201 In order to do it you can remove from the kernel config:
206 , and instead add back:
207 device atadisk # ATA disk drives
208 device ataraid # ATA RAID drives
209 device atapicd # ATAPI CDROM drives
210 device atapifd # ATAPI floppy drives
211 device atapist # ATAPI tape drives
214 The default NFS server has been changed to the new server, which
215 was referred to as the experimental server. If you need to switch
216 back to the old NFS server, you must now put the "-o" option on
217 both the mountd and nfsd commands. This can be done using the
218 mountd_flags and nfs_server_flags rc.conf variables until an
219 update to the rc scripts is committed, which is coming soon.
222 The GNU Objective-C runtime library (libobjc), and other Objective-C
223 related components have been removed from the base system. If you
224 require an Objective-C library, please use one of the available ports.
227 ath(4) has been split into bus- and device- modules. if_ath contains
228 the HAL, the TX rate control and the network device code. if_ath_pci
229 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
230 contains the AHB glue. Users need to load both if_ath_pci and if_ath
231 in order to use ath on everything else.
233 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
234 need to load if_ath and if_ath_pci for ath(4) operation.
237 As part of the replacement of sysinstall, the process of building
238 release media has changed significantly. For details, please re-read
239 release(7), which has been updated to reflect the new build process.
242 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
243 is the last available version under GPLv2. It brings a number of new
244 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
245 4.1 and SSE 4.2), better support for powerpc64, a number of new
246 directives, and lots of other small improvements. See the ChangeLog
247 file in contrib/binutils for the full details.
250 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
251 compliant, and will now use half of hash for authentication.
252 This will break interoperability with all stacks (including all
253 actual FreeBSD versions) who implement
254 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
256 The only workaround with such peers is to use another HMAC
257 algorithm for IPsec ("phase 2") authentication.
260 Remove the uio_yield prototype and symbol. This function has
261 been misnamed since it was introduced and should not be
262 globally exposed with this name. The equivalent functionality
263 is now available using kern_yield(curthread->td_user_pri).
264 The function remains undocumented.
267 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
268 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
269 for scalar sysctls is defined but disabled. Code that needs
270 UQUAD to pass the type checking that must compile on older
271 systems where the define is not present can check against
272 __FreeBSD_version >= 900030.
274 The system dialog(1) has been replaced with a new version previously
275 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
276 with the previous version, but the libdialog associated with it has
277 a largely incompatible API. As such, the original version of libdialog
278 will be kept temporarily as libodialog, until its base system consumers
279 are replaced or updated. Bump __FreeBSD_version to 900030.
282 If you are trying to run make universe on a -stable system, and you get
283 the following warning:
284 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
285 unknown. config(8) likely too old."
286 or something similar to it, then you must upgrade your -stable system
287 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
288 stable/8) or build the config from the latest stable/8 branch and
289 install it on your system.
291 Prior to this date, building a current universe on 8-stable system from
292 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
293 error in the first kernel build phase. A new config on those old
294 systems will fix that problem for older versions of -current.
297 The TCP stack has been modified to allow Khelp modules to interact with
298 it via helper hook points and store per-connection data in the TCP
299 control block. Bump __FreeBSD_version to 900029. User space tools that
300 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
304 Generic IEEE 802.3 annex 31B full duplex flow control support has been
305 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
306 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
307 to take advantage of it instead of using custom implementations. This
308 means that these drivers now no longer unconditionally advertise
309 support for flow control but only do so if flow control is a selected
310 media option. This was implemented in the generic support that way in
311 order to allow flow control to be switched on and off via ifconfig(8)
312 with the PHY specific default to typically off in order to protect
313 from unwanted effects. Consequently, if you used flow control with
314 one of the above mentioned drivers you now need to explicitly enable
316 ifconfig bge0 media auto mediaopt flowcontrol
318 Along with the above mentioned changes generic support for setting
319 1000baseT master mode also has been added and brgphy(4), ciphy(4),
320 e1000phy(4) as well as ip1000phy(4) have been converted to take
321 advantage of it. This means that these drivers now no longer take the
322 link0 parameter for selecting master mode but the master media option
323 has to be used instead, for example like in the following:
324 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
326 Selection of master mode now is also available with all other PHY
327 drivers supporting 1000baseT.
330 The TCP stack has received a significant update to add support for
331 modularised congestion control and generally improve the clarity of
332 congestion control decisions. Bump __FreeBSD_version to 900025. User
333 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
334 sockstat) need to be recompiled.
337 The man(1) utility has been replaced by a new version that no longer
338 uses /etc/manpath.config. Please consult man.conf(5) for how to
339 migrate local entries to the new format.
342 The copyright strings printed by login(1) and sshd(8) at the time of a
343 new connection have been removed to follow other operating systems and
347 A workaround for a fixed ld bug has been removed in kernel code,
348 so make sure that your system ld is built from sources after
349 revision 210245 from 2010-07-19 (r211583 if building head kernel
350 on stable/8, r211584 for stable/7; both from 2010-08-21).
351 A symptom of incorrect ld version is different addresses for
352 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
355 The $ipv6_prefer variable in rc.conf(5) has been split into
356 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
358 The $ip6addrctl_policy is a variable to choose a pre-defined
359 address selection policy set by ip6addrctl(8). A value
360 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
363 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
364 flag (see an entry of 20090926) is set on an interface with no
365 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
366 security reason. If you want IPv6 link-local address on all
367 interfaces by default, set this to "YES".
369 The old ipv6_prefer="YES" is equivalent to
370 ipv6_activate_all_interfaces="YES" and
371 ip6addrctl_policy="ipv6_prefer".
374 DTrace has grown support for userland tracing. Due to this, DTrace is
375 now i386 and amd64 only.
376 dtruss(1) is now installed by default on those systems and a new
377 kernel module is needed for userland tracing: fasttrap.
378 No changes to your kernel config file are necessary to enable
379 userland tracing, but you might consider adding 'STRIP=' and
380 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
381 to have informative userland stack traces in DTrace (ustack).
384 The acpi_aiboost(4) driver has been removed in favor of the new
385 aibs(4) driver. You should update your kernel configuration file.
388 BSD grep has been imported to the base system and it is built by
389 default. It is completely BSD licensed, highly GNU-compatible, uses
390 less memory than its GNU counterpart and has a small codebase.
391 However, it is slower than its GNU counterpart, which is mostly
392 noticeable for larger searches, for smaller ones it is measurable
393 but not significant. The reason is complex, the most important factor
394 is that we lack a modern and efficient regex library and GNU
395 overcomes this by optimizing the searches internally. Future work
396 on improving the regex performance is planned, for the meantime,
397 users that need better performance, can build GNU grep instead by
398 setting the WITH_GNU_GREP knob.
401 Due to the import of powerpc64 support, all existing powerpc kernel
402 configuration files must be updated with a machine directive like this:
403 machine powerpc powerpc
405 In addition, an updated config(8) is required to build powerpc kernels
409 A new version of ZFS (version 15) has been merged to -HEAD.
410 This version uses a python library for the following subcommands:
411 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
412 For full functionality of these commands the following port must
413 be installed: sysutils/py-zfs
416 'vm_page's are now hashed by physical address to an array of mutexes.
417 Currently this is only used to serialize access to hold_count. Over
418 time the page queue mutex will be peeled away. This changes the size
419 of pmap on every architecture. And requires all callers of vm_page_hold
420 and vm_page_unhold to be updated.
423 WITH_CTF can now be specified in src.conf (not recommended, there
424 are some problems with static executables), make.conf (would also
425 affect ports which do not use GNU make and do not override the
426 compile targets) or in the kernel config (via "makeoptions
428 When WITH_CTF was specified there before this was silently ignored,
429 so make sure that WITH_CTF is not used in places which could lead
430 to unwanted behavior.
433 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
434 to allow 32-bit compatibility on non-x86 platforms. All kernel
435 configurations on amd64 and ia64 platforms using these options must
436 be modified accordingly.
439 The utmp user accounting database has been replaced with utmpx,
440 the user accounting interface standardized by POSIX.
441 Unfortunately the semantics of utmp and utmpx don't match,
442 making it practically impossible to support both interfaces.
443 The user accounting database is used by tools like finger(1),
444 last(1), talk(1), w(1) and ac(8).
446 All applications in the base system use utmpx. This means only
447 local binaries (e.g. from the ports tree) may still use these
448 utmp database files. These applications must be rebuilt to make
451 After the system has been upgraded, it is safe to remove the old
452 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
453 assuming their contents is of no importance anymore. Old wtmp
454 databases can only be used by last(1) and ac(8) after they have
455 been converted to the new format using wtmpcvt(1).
458 Introduce the kernel thread "deadlock resolver" (which can be enabled
459 via the DEADLKRES option, see NOTES for more details) and the
460 sleepq_type() function for sleepqueues.
463 The rc.firewall and rc.firewall6 were unified, and
464 rc.firewall6 and rc.d/ip6fw were removed.
465 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
466 variables are obsoleted. Instead, the following new rc
467 variables are added to rc.d/ipfw:
469 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
470 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
471 firewall_simple_onet_ipv6, firewall_trusted_ipv6
473 The meanings correspond to the relevant IPv4 variables.
479 The default terminal emulation for syscons(4) has been changed
480 from cons25 to xterm on all platforms except pc98. This means
481 that the /etc/ttys file needs to be updated to ensure correct
482 operation of applications on the console.
484 The terminal emulation style can be toggled per window by using
485 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
486 options can be used to change the compile-time default back to
489 To prevent graphical artifacts, make sure the TERM environment
490 variable is set to match the terminal emulation that is being
491 performed by syscons(4).
494 The layout of the structure ieee80211req_scan_result has changed.
495 Applications that require wireless scan results (e.g. ifconfig(8))
496 from net80211 need to be recompiled.
498 Applications such as wpa_supplicant(8) may require a full world
499 build without using NO_CLEAN in order to get synchronized with the
503 The iwn(4) driver has been updated to support the 5000 and 5150 series.
504 There's one kernel module for each firmware. Adding "device iwnfw"
505 to the kernel configuration file means including all three firmware
506 images inside the kernel. If you want to include just the one for
507 your wireless card, use the devices iwn4965fw, iwn5000fw or
511 The rc.d/network_ipv6, IPv6 configuration script has been integrated
512 into rc.d/netif. The changes are the following:
514 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
515 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
516 Note that both variables need the "inet6" keyword at the head.
518 Do not set $ipv6_network_interfaces manually if you do not
519 understand what you are doing. It is not needed in most cases.
521 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
524 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
525 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
527 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
528 all configured interfaces have "inet6 accept_rtadv" in the
529 $ifconfig_IF_ipv6. These are for backward compatibility.
531 3. A new variable $ipv6_prefer has been added. If NO, IPv6
532 functionality of interfaces with no corresponding
533 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
534 and the default address selection policy of ip6addrctl(8)
535 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
536 Note that if you want to configure IPv6 functionality on the
537 disabled interfaces after boot, first you need to clear the flag by
538 using ifconfig(8) like:
540 ifconfig em0 inet6 -ifdisabled
542 If YES, the default address selection policy is set as
545 The default value of $ipv6_prefer is NO.
547 4. If your system need to receive Router Advertisement messages,
548 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
549 scripts automatically invoke rtsol(8) when the interface becomes
550 UP. The Router Advertisement messages are used for SLAAC
551 (State-Less Address AutoConfiguration).
554 802.11s D3.03 support was committed. This is incompatible with the
555 previous code, which was based on D3.0.
558 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
559 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
560 control whether accepting Router Advertisement messages or not.
561 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
562 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
563 The ifconfig(8) utility now supports these flags.
566 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
567 mount(8) and -a option for df(1) to see them.
570 The old tunable hw.bus.devctl_disable has been superseded by
571 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
572 replaced by hw.bus.devctl_queue=0. The default for this new tunable
576 Remove the option STOP_NMI. The default action is now to use NMI only
577 for KDB via the newly introduced function stop_cpus_hard() and
578 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
581 The stable/8 branch created in subversion. This corresponds to the
582 RELENG_8 branch in CVS.
585 Bump the shared library version numbers for all libraries that do not
586 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
587 __FreeBSD_version to 800105.
590 Due to changes in the implementation of virtual network stack support,
591 all network-related kernel modules must be recompiled. As this change
592 breaks the ABI, bump __FreeBSD_version to 800104.
595 The TOE interface to the TCP syncache has been modified to remove
596 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
597 The cxgb driver is the only TOE consumer affected by this change, and
598 needs to be recompiled along with the kernel. As this change breaks
599 the ABI, bump __FreeBSD_version to 800103.
602 Padding has been added to struct tcpcb, sackhint and tcpstat in
603 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
604 maintaining the ABI. However, this change breaks the ABI, so bump
605 __FreeBSD_version to 800102. User space tools that rely on the size of
606 any of these structs (e.g. sockstat) need to be recompiled.
609 The NFS_LEGACYRPC option has been removed along with the old kernel
610 RPC implementation that this option selected. Kernel configurations
611 may need to be adjusted.
614 The network interface device nodes at /dev/net/<interface> have been
615 removed. All ioctl operations can be performed the normal way using
616 routing sockets. The kqueue functionality can generally be replaced
617 with routing sockets.
620 The documentation from the FreeBSD Documentation Project (Handbook,
621 FAQ, etc.) is now installed via packages by sysinstall(8) and under
622 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
625 The ABI of various structures related to the SYSV IPC API have been
626 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
627 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
631 Layout of struct vnet has changed as routing related variables were
632 moved to their own Vimage module. Modules need to be recompiled. Bump
633 __FreeBSD_version to 800099.
636 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
637 respectively. As long as no more than 16 groups per process are used,
638 no changes should be visible. When more than 16 groups are used, old
639 binaries may fail if they call getgroups() or getgrouplist() with
640 statically sized storage. Recompiling will work around this, but
641 applications should be modified to use dynamically allocated storage
642 for group arrays as POSIX.1-2008 does not cap an implementation's
643 number of supported groups at NGROUPS_MAX+1 as previous versions did.
645 NFS and portalfs mounts may also be affected as the list of groups is
646 truncated to 16. Users of NFS who use more than 16 groups, should
647 take care that negative group permissions are not used on the exported
648 file systems as they will not be reliable unless a GSSAPI based
649 authentication method is used.
652 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
653 option compiles in the support for adaptive spinning for lockmgrs
654 which want to enable it. The lockinit() function now accepts the flag
655 LK_ADAPTIVE in order to make the lock object subject to adaptive
656 spinning when both held in write and read mode.
659 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
660 changed. User applications that use this ioctl need to be rebuilt.
663 The layout of struct thread has changed. Kernel and modules need to
667 The layout of structs ifnet, domain, protosw and vnet_net has changed.
668 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
671 window(1) has been removed from the base system. It can now be
672 installed from ports. The port is called misc/window.
675 The way we are storing and accessing `routing table' entries has
676 changed. Programs reading the FIB, like netstat, need to be
680 A new netisr implementation has been added for FreeBSD 8. Network
681 file system modules, such as igmp, ipdivert, and others, should be
683 Bump __FreeBSD_version to 800096.
686 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
690 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
691 Bump __FreeBSD_version to 800094.
694 Add mnt_xflag field to 'struct mount'. File system modules need to be
696 Bump __FreeBSD_version to 800093.
699 The compiling option ADAPTIVE_SX has been retired while it has been
700 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
701 The KPI for sx_init_flags() changes as accepting flags:
702 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
703 been introduced in order to handle the reversed logic.
704 Bump __FreeBSD_version to 800092.
707 Add support for hierarchical jails. Remove global securelevel.
708 Bump __FreeBSD_version to 800091.
711 The layout of struct vnet_net has changed, therefore modules
713 Bump __FreeBSD_version to 800090.
716 The newly imported zic(8) produces a new format in the output. Please
717 run tzsetup(8) to install the newly created data to /etc/localtime.
720 The sysctl tree for the usb stack has renamed from hw.usb2.* to
721 hw.usb.* and is now consistent again with previous releases.
724 802.11 monitor mode support was revised and driver api's were changed.
725 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
726 of DLT_IEEE802_11. No user-visible data structures were changed but
727 applications that use DLT_IEEE802_11 may require changes.
728 Bump __FreeBSD_version to 800088.
731 The layout of the following structs has changed: sysctl_oid,
732 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
733 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
734 panics may be experienced. World rebuild is required for
735 correctly checking networking state from userland.
736 Bump __FreeBSD_version to 800085.
739 MLDv2 and Source-Specific Multicast (SSM) have been merged
740 to the IPv6 stack. VIMAGE hooks are in but not yet used.
741 The implementation of SSM within FreeBSD's IPv6 stack closely
742 follows the IPv4 implementation.
744 For kernel developers:
746 * The most important changes are that the ip6_output() and
747 ip6_input() paths no longer take the IN6_MULTI_LOCK,
748 and this lock has been downgraded to a non-recursive mutex.
750 * As with the changes to the IPv4 stack to support SSM, filtering
751 of inbound multicast traffic must now be performed by transport
752 protocols within the IPv6 stack. This does not apply to TCP and
753 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
755 * The KPIs used by IPv6 multicast are similar to those used by
756 the IPv4 stack, with the following differences:
757 * im6o_mc_filter() is analogous to imo_multicast_filter().
758 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
759 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
760 * IN6_LOOKUP_MULTI() has been deprecated and removed.
761 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
762 for MLDv1 have an additional 'timer' argument which is used to
763 jitter the initial membership report for the solicited-node
764 multicast membership on-link.
765 * This is not strictly needed for MLDv2, which already jitters
766 its report transmissions. However, the 'timer' argument is
767 preserved in case MLDv1 is active on the interface.
769 * The KAME linked-list based IPv6 membership implementation has
770 been refactored to use a vector similar to that used by the IPv4
772 Code which maintains a list of its own multicast memberships
773 internally, e.g. carp, has been updated to reflect the new
776 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
777 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
778 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
779 implementation constraint which needs to be addressed in HEAD.
781 For application developers:
783 * The changes are broadly similar to those made for the IPv4
786 * The use of IPv4 and IPv6 multicast socket options on the same
787 socket, using mapped addresses, HAS NOT been tested or supported.
789 * There are a number of issues with the implementation of various
790 IPv6 multicast APIs which need to be resolved in the API surface
791 before the implementation is fully compatible with KAME userland
792 use, and these are mostly to do with interface index treatment.
794 * The literature available discusses the use of either the delta / ASM
795 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
796 using setsourcefilter(3)/getsourcefilter(3). For more information
797 please refer to RFC 3768, 'Socket Interface Extensions for
798 Multicast Source Filters'.
800 * Applications which use the published RFC 3678 APIs should be fine.
802 For systems administrators:
804 * The mtest(8) utility has been refactored to support IPv6, in
805 addition to IPv4. Interface addresses are no longer accepted
806 as arguments, their names must be used instead. The utility
807 will map the interface name to its first IPv4 address as
808 returned by getifaddrs(3).
810 * The ifmcstat(8) utility has also been updated to print the MLDv2
811 endpoint state and source filter lists via sysctl(3).
813 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
814 loopback of IPv6 multicast datagrams by default; it defaults to 1
815 to preserve the existing behaviour. Disabling multicast loopback is
816 recommended for optimal system performance.
818 * The IPv6 MROUTING code has been changed to examine this sysctl
819 instead of attempting to perform a group lookup before looping
820 back forwarded datagrams.
822 Bump __FreeBSD_version to 800084.
825 Implement low-level Bluetooth HCI API.
826 Bump __FreeBSD_version to 800083.
829 The layout of struct malloc_type, used by modules to register new
830 memory allocation types, has changed. Most modules will need to
831 be rebuilt or panics may be experienced.
832 Bump __FreeBSD_version to 800081.
835 Anticipate overflowing inp_flags - add inp_flags2.
836 This changes most offsets in inpcb, so checking v4 connection
837 state will require a world rebuild.
838 Bump __FreeBSD_version to 800080.
841 Add an llentry to struct route and struct route_in6. Modules
842 embedding a struct route will need to be recompiled.
843 Bump __FreeBSD_version to 800079.
846 The size of rt_metrics_lite and by extension rtentry has changed.
847 Networking administration apps will need to be recompiled.
848 The route command now supports show as an alias for get, weighting
849 of routes, sticky and nostick flags to alter the behavior of stateful
851 Bump __FreeBSD_version to 800078.
854 Do not use Giant for kbdmux(4) locking. This is wrong and
855 apparently causing more problems than it solves. This will
856 re-open the issue where interrupt handlers may race with
857 kbdmux(4) in polling mode. Typical symptoms include (but
858 not limited to) duplicated and/or missing characters when
859 low level console functions (such as gets) are used while
860 interrupts are enabled (for example geli password prompt,
861 mountroot prompt etc.). Disabling kbdmux(4) may help.
864 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
865 kernel modules referencing any of the above need to be recompiled.
866 Bump __FreeBSD_version to 800075.
869 GEOM_PART has become the default partition slicer for storage devices,
870 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
871 introduces some changes:
873 MSDOS/EBR: the devices created from MSDOS extended partition entries
874 (EBR) can be named differently than with GEOM_MBR and are now symlinks
875 to devices with offset-based names. fstabs may need to be modified.
877 BSD: the "geometry does not match label" warning is harmless in most
878 cases but it points to problems in file system misalignment with
879 disk geometry. The "c" partition is now implicit, covers the whole
880 top-level drive and cannot be (mis)used by users.
882 General: Kernel dumps are now not allowed to be written to devices
883 whose partition types indicate they are meant to be used for file
884 systems (or, in case of MSDOS partitions, as something else than
887 Most of these changes date approximately from 200812.
890 The uscanner(4) driver has been removed from the kernel. This follows
891 Linux removing theirs in 2.6 and making libusb the default interface
895 The multicast forwarding code has been cleaned up. netstat(1)
896 only relies on KVM now for printing bandwidth upcall meters.
897 The IPv4 and IPv6 modules are split into ip_mroute_mod and
898 ip6_mroute_mod respectively. The config(5) options for statically
899 compiling this code remain the same, i.e. 'options MROUTING'.
902 Support for the IFF_NEEDSGIANT network interface flag has been
903 removed, which means that non-MPSAFE network device drivers are no
904 longer supported. In particular, if_ar, if_sr, and network device
905 drivers from the old (legacy) USB stack can no longer be built or
909 POSIX.1 Native Language Support (NLS) has been enabled in libc and
910 a bunch of new language catalog files have also been added.
911 This means that some common libc messages are now localized and
912 they depend on the LC_MESSAGES environmental variable.
915 The k8temp(4) driver has been renamed to amdtemp(4) since
916 support for Family 10 and Family 11 CPU families was added.
919 IGMPv3 and Source-Specific Multicast (SSM) have been merged
920 to the IPv4 stack. VIMAGE hooks are in but not yet used.
922 For kernel developers, the most important changes are that the
923 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
924 and this lock has been downgraded to a non-recursive mutex.
926 Transport protocols (UDP, Raw IP) are now responsible for filtering
927 inbound multicast traffic according to group membership and source
928 filters. The imo_multicast_filter() KPI exists for this purpose.
929 Transports which do not use multicast (SCTP, TCP) already reject
930 multicast by default. Forwarding and receive performance may improve
931 as a mutex acquisition is no longer needed in the ip_input()
932 low-level input path. in_addmulti() and in_delmulti() are shimmed
933 to new KPIs which exist to support SSM in-kernel.
935 For application developers, it is recommended that loopback of
936 multicast datagrams be disabled for best performance, as this
937 will still cause the lock to be taken for each looped-back
938 datagram transmission. The net.inet.ip.mcast.loop sysctl may
939 be tuned to 0 to disable loopback by default; it defaults to 1
940 to preserve the existing behaviour.
942 For systems administrators, to obtain best performance with
943 multicast reception and multiple groups, it is always recommended
944 that a card with a suitably precise hash filter is used. Hash
945 collisions will still result in the lock being taken within the
946 transport protocol input path to check group membership.
948 If deploying FreeBSD in an environment with IGMP snooping switches,
949 it is recommended that the net.inet.igmp.sendlocal sysctl remain
950 enabled; this forces 224.0.0.0/24 group membership to be announced
953 The size of 'struct igmpstat' has changed; netstat needs to be
954 recompiled to reflect this.
955 Bump __FreeBSD_version to 800070.
958 libusb20.so.1 is now installed as libusb.so.1 and the ports system
959 updated to use it. This requires a buildworld/installworld in order to
960 update the library and dependencies (usbconfig, etc). Its advisable to
961 rebuild all ports which uses libusb. More specific directions are given
962 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
963 libusb are no longer required and can be removed.
966 A workaround is committed to allow the creation of System V shared
967 memory segment of size > 2 GB on the 64-bit architectures.
968 Due to a limitation of the existing ABI, the shm_segsz member
969 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
970 wrong for large segments. Note that limits must be explicitly
971 raised to allow such segments to be created.
974 The layout of struct ifnet has changed, requiring a rebuild of all
975 network device driver modules.
978 The /dev handling for the new USB stack has changed, a
979 buildworld/installworld is required for libusb20.
982 The new USB2 stack has now been permanently moved in and all kernel and
983 module names reverted to their previous values (eg, usb, ehci, ohci,
984 ums, ...). The old usb stack can be compiled in by prefixing the name
985 with the letter 'o', the old usb modules have been removed.
986 Updating entry 20090216 for xorg and 20090215 for libmap may still
990 The rc.conf(5) option if_up_delay has been renamed to
991 defaultroute_delay to better reflect its purpose. If you have
992 customized this setting in /etc/rc.conf you need to update it to
996 xorg 7.4 wants to configure its input devices via hald which does not
997 yet work with USB2. If the keyboard/mouse does not work in xorg then
999 Option "AllowEmptyInput" "off"
1000 to your ServerLayout section. This will cause X to use the configured
1001 kbd and mouse sections from your xorg.conf.
1004 The GENERIC kernels for all architectures now default to the new USB2
1005 stack. No kernel config options or code have been removed so if a
1006 problem arises please report it and optionally revert to the old USB
1007 stack. If you are loading USB kernel modules or have a custom kernel
1008 that includes GENERIC then ensure that usb names are also changed over,
1009 eg uftdi -> usb2_serial_ftdi.
1011 Older programs linked against the ports libusb 0.1 need to be
1012 redirected to the new stack's libusb20. /etc/libmap.conf can
1014 # Map old usb library to new one for usb2 stack
1015 libusb-0.1.so.8 libusb20.so.1
1018 All USB ethernet devices now attach as interfaces under the name ueN
1019 (eg. ue0). This is to provide a predictable name as vendors often
1020 change usb chipsets in a product without notice.
1023 The ichsmb(4) driver has been changed to require SMBus slave
1024 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1025 All of the other SMBus controller drivers require left-justified
1026 slave addresses, so this change makes all the drivers provide the
1030 INET6 statistics (struct ip6stat) was updated.
1031 netstat(1) needs to be recompiled.
1034 NTFS has been removed from GENERIC kernel on amd64 to match
1035 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1036 will load ntfs.ko module automatically when NTFS support is
1037 actually needed, unless ntfs.ko is not installed or security
1038 level prohibits loading kernel modules. If either is the case,
1039 "options NTFS" has to be added into kernel config.
1042 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1043 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1044 800061. User space tools that rely on the size of struct tcpcb in
1045 tcp_var.h (e.g. sockstat) need to be recompiled.
1048 ng_tty(4) module updated to match the new TTY subsystem.
1049 Due to API change, user-level applications must be updated.
1050 New API support added to mpd5 CVS and expected to be present
1051 in next mpd5.3 release.
1054 With __FreeBSD_version 800060 the makefs tool is part of
1055 the base system (it was a port).
1058 The afdata and ifnet locks have been changed from mutexes to
1059 rwlocks, network modules will need to be re-compiled.
1062 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1063 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1064 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1065 architecture and 40 bytes on 64-bit architecture. The userland
1066 applications "arp" and "ndp" have been updated accordingly.
1067 The output from "netstat -r" shows only routing entries and
1068 none of the L2 information.
1071 __FreeBSD_version 800057 marks the switchover from the
1072 binary ath hal to source code. Users must add the line:
1074 options AH_SUPPORT_AR5416
1076 to their kernel config files when specifying:
1080 The ath_hal module no longer exists; the code is now compiled
1081 together with the driver in the ath module. It is now
1082 possible to tailor chip support (i.e. reduce the set of chips
1083 and thereby the code size); consult ath_hal(4) for details.
1086 __FreeBSD_version 800054 adds memory barriers to
1087 <machine/atomic.h>, new interfaces to ifnet to facilitate
1088 multiple hardware transmit queues for cards that support
1089 them, and a lock-less ring-buffer implementation to
1090 enable drivers to more efficiently manage queueing of
1094 A new version of ZFS (version 13) has been merged to -HEAD.
1095 This version has zpool attribute "listsnapshots" off by
1096 default, which means "zfs list" does not show snapshots,
1097 and is the same as Solaris behavior.
1100 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1103 The uhci, ohci, ehci and slhci USB Host controller drivers have
1104 been put into separate modules. If you load the usb module
1105 separately through loader.conf you will need to load the
1106 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1107 controller add the following to loader.conf:
1113 The ABI used by the PMC toolset has changed. Please keep
1114 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1118 atapci kernel module now includes only generic PCI ATA
1119 driver. AHCI driver moved to ataahci kernel module.
1120 All vendor-specific code moved into separate kernel modules:
1121 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1122 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1123 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1124 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1127 The TTY subsystem of the kernel has been replaced by a new
1128 implementation, which provides better scalability and an
1129 improved driver model. Most common drivers have been migrated to
1130 the new TTY subsystem, while others have not. The following
1131 drivers have not yet been ported to the new TTY layer:
1134 cy, digi, rc, rp, sio
1140 ng_h4, ng_tty, ppp, sl, snp
1142 Adding these drivers to your kernel configuration file shall
1143 cause compilation to fail.
1146 ntpd has been upgraded to 4.2.4p5.
1149 OpenSSH has been upgraded to 5.1p1.
1151 For many years, FreeBSD's version of OpenSSH preferred DSA
1152 over RSA for host and user authentication keys. With this
1153 upgrade, we've switched to the vendor's default of RSA over
1154 DSA. This may cause upgraded clients to warn about unknown
1155 host keys even for previously known hosts. Users should
1156 follow the usual procedure for verifying host keys before
1157 accepting the RSA key.
1159 This can be circumvented by setting the "HostKeyAlgorithms"
1160 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1163 Please note that the sequence of keys offered for
1164 authentication has been changed as well. You may want to
1165 specify IdentityFile in a different order to revert this
1169 The sio(4) driver has been removed from the i386 and amd64
1170 kernel configuration files. This means uart(4) is now the
1171 default serial port driver on those platforms as well.
1173 To prevent collisions with the sio(4) driver, the uart(4) driver
1174 uses different names for its device nodes. This means the
1175 onboard serial port will now most likely be called "ttyu0"
1176 instead of "ttyd0". You may need to reconfigure applications to
1177 use the new device names.
1179 When using the serial port as a boot console, be sure to update
1180 /boot/device.hints and /etc/ttys before booting the new kernel.
1181 If you forget to do so, you can still manually specify the hints
1182 at the loader prompt:
1184 set hint.uart.0.at="isa"
1185 set hint.uart.0.port="0x3F8"
1186 set hint.uart.0.flags="0x10"
1187 set hint.uart.0.irq="4"
1191 The gpt(8) utility has been removed. Use gpart(8) to partition
1195 The version that Linuxulator emulates was changed from 2.4.2
1196 to 2.6.16. If you experience any problems with Linux binaries
1197 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1198 if it fixes the problem contact emulation mailing list.
1201 ISDN4BSD (I4B) was removed from the src tree. You may need to
1202 update a your kernel configuration and remove relevant entries.
1205 I have checked in code to support multiple routing tables.
1206 See the man pages setfib(1) and setfib(2).
1207 This is a hopefully backwards compatible version,
1208 but to make use of it you need to compile your kernel
1209 with options ROUTETABLES=2 (or more up to 16).
1212 The 802.11 wireless support was redone to enable multi-bss
1213 operation on devices that are capable. The underlying device
1214 is no longer used directly but instead wlanX devices are
1215 cloned with ifconfig. This requires changes to rc.conf files.
1216 For example, change:
1217 ifconfig_ath0="WPA DHCP"
1220 ifconfig_wlan0="WPA DHCP"
1221 see rc.conf(5) for more details. In addition, mergemaster of
1222 /etc/rc.d is highly recommended. Simultaneous update of userland
1223 and kernel wouldn't hurt either.
1225 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1226 modules were merged into the base wlan module. All references
1227 to these modules (e.g. in kernel config files) must be removed.
1230 psm(4) has gained write(2) support in native operation level.
1231 Arbitrary commands can be written to /dev/psm%d and status can
1232 be read back from it. Therefore, an application is responsible
1233 for status validation and error recovery. It is a no-op in
1234 other operation levels.
1237 Support for KSE threading has been removed from the kernel. To
1238 run legacy applications linked against KSE libmap.conf may
1239 be used. The following libmap.conf may be used to ensure
1240 compatibility with any prior release:
1242 libpthread.so.1 libthr.so.1
1243 libpthread.so.2 libthr.so.2
1244 libkse.so.3 libthr.so.3
1247 The layout of struct vmspace has changed. This affects libkvm
1248 and any executables that link against libkvm and use the
1249 kvm_getprocs() function. In particular, but not exclusively,
1250 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1251 The effects are minimal, but it's advisable to upgrade world
1255 The latest em driver no longer has support in it for the
1256 82575 adapter, this is now moved to the igb driver. The
1257 split was done to make new features that are incompatible
1258 with older hardware easier to do.
1261 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1262 likewise the kernel option is now GEOM_LINUX_LVM.
1265 The default NFS mount mode has changed from UDP to TCP for
1266 increased reliability. If you rely on (insecurely) NFS
1267 mounting across a firewall you may need to update your
1271 Belatedly note the addition of m_collapse for compacting
1275 The fts(3) structures have been changed to use adequate
1276 integer types for their members and so to be able to cope
1277 with huge file trees. The old fts(3) ABI is preserved
1278 through symbol versioning in libc, so third-party binaries
1279 using fts(3) should still work, although they will not take
1280 advantage of the extended types. At the same time, some
1281 third-party software might fail to build after this change
1282 due to unportable assumptions made in its source code about
1283 fts(3) structure members. Such software should be fixed
1284 by its vendor or, in the worst case, in the ports tree.
1285 FreeBSD_version 800015 marks this change for the unlikely
1286 case that a portable fix is impossible.
1289 To upgrade to -current after this date, you must be running
1290 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1291 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1294 The ADAPTIVE_GIANT kernel option has been retired because its
1295 functionality is the default now.
1298 The AT keyboard emulation of sunkbd(4) has been turned on
1299 by default. In order to make the special symbols of the Sun
1300 keyboards driven by sunkbd(4) work under X these now have
1301 to be configured the same way as Sun USB keyboards driven
1302 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1304 Option "XkbLayout" "us"
1305 Option "XkbRules" "xorg"
1306 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1309 It has been decided that it is desirable to provide ABI
1310 backwards compatibility to the FreeBSD 4/5/6 versions of the
1311 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1312 broken with the introduction of PCI domain support (see the
1313 20070930 entry). Unfortunately, this required the ABI of
1314 PCIOCGETCONF to be broken again in order to be able to
1315 provide backwards compatibility to the old version of that
1316 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1317 again. As for prominent ports this affects neither pciutils
1318 nor xorg-server this time, the hal port needs to be rebuilt
1322 The misnamed kthread_create() and friends have been renamed
1323 to kproc_create() etc. Many of the callers already
1324 used kproc_start()..
1325 I will return kthread_create() and friends in a while
1326 with implementations that actually create threads, not procs.
1327 Renaming corresponds with version 800002.
1336 Avoid using make -j when upgrading. While generally safe, there are
1337 sometimes problems using -j to upgrade. If your upgrade fails with
1338 -j, please try again without -j. From time to time in the past there
1339 have been problems using -j with buildworld and/or installworld. This
1340 is especially true when upgrading between "distant" versions (eg one
1341 that cross a major release boundary or several minor releases, or when
1342 several months have passed on the -current branch).
1344 Sometimes, obscure build problems are the result of environment
1345 poisoning. This can happen because the make utility reads its
1346 environment when searching for values for global variables. To run
1347 your build attempts in an "environmental clean room", prefix all make
1348 commands with 'env -i '. See the env(1) manual page for more details.
1350 When upgrading from one major version to another it is generally best
1351 to upgrade to the latest code in the currently installed branch first,
1352 then do an upgrade to the new branch. This is the best-tested upgrade
1353 path, and has the highest probability of being successful. Please try
1354 this approach before reporting problems with a major version upgrade.
1358 When upgrading the boot ZFS pool to a new version, always follow
1361 1.) recompile and reinstall the ZFS boot loader and boot block
1362 (this is part of "make buildworld" and "make installworld")
1364 2.) update the ZFS boot block on your boot drive
1366 The following example updates the ZFS boot block on the first
1367 partition (freebsd-boot) of a GPT partitioned drive ad0:
1368 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1370 Non-boot pools do not need these updates.
1374 If you are updating from a prior version of FreeBSD (even one just
1375 a few days old), you should follow this procedure. It is the most
1376 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1378 make kernel-toolchain
1379 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1380 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1382 To test a kernel once
1383 ---------------------
1384 If you just want to boot a kernel once (because you are not sure
1385 if it works, or if you want to boot a known bad kernel to provide
1386 debugging information) run
1387 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1388 nextboot -k testkernel
1390 To just build a kernel when you know that it won't mess you up
1391 --------------------------------------------------------------
1392 This assumes you are already running a CURRENT system. Replace
1393 ${arch} with the architecture of your machine (e.g. "i386",
1394 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1396 cd src/sys/${arch}/conf
1397 config KERNEL_NAME_HERE
1398 cd ../compile/KERNEL_NAME_HERE
1403 If this fails, go to the "To build a kernel" section.
1405 To rebuild everything and install it on the current system.
1406 -----------------------------------------------------------
1407 # Note: sometimes if you are running current you gotta do more than
1408 # is listed here if you are upgrading from a really old current.
1410 <make sure you have good level 0 dumps>
1412 make kernel KERNCONF=YOUR_KERNEL_HERE
1414 <reboot in single user> [3]
1422 To cross-install current onto a separate partition
1423 --------------------------------------------------
1424 # In this approach we use a separate partition to hold
1425 # current's root, 'usr', and 'var' directories. A partition
1426 # holding "/", "/usr" and "/var" should be about 2GB in
1429 <make sure you have good level 0 dumps>
1432 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1433 <maybe newfs current's root partition>
1434 <mount current's root partition on directory ${CURRENT_ROOT}>
1435 make installworld DESTDIR=${CURRENT_ROOT}
1436 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1437 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1438 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1439 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1440 <reboot into current>
1441 <do a "native" rebuild/install as described in the previous section>
1442 <maybe install compatibility libraries from ports/misc/compat*>
1446 To upgrade in-place from 8.x-stable to current
1447 ----------------------------------------------
1448 <make sure you have good level 0 dumps>
1450 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1452 <reboot in single user> [3]
1459 Make sure that you've read the UPDATING file to understand the
1460 tweaks to various things you need. At this point in the life
1461 cycle of current, things change often and you are on your own
1462 to cope. The defaults can also change, so please read ALL of
1463 the UPDATING entries.
1465 Also, if you are tracking -current, you must be subscribed to
1466 freebsd-current@freebsd.org. Make sure that before you update
1467 your sources that you have read and understood all the recent
1468 messages there. If in doubt, please track -stable which has
1469 much fewer pitfalls.
1471 [1] If you have third party modules, such as vmware, you
1472 should disable them at this point so they don't crash your
1475 [3] From the bootblocks, boot -s, and then do
1480 adjkerntz -i # if CMOS is wall time
1481 Also, when doing a major release upgrade, it is required that
1482 you boot into single user mode to do the installworld.
1484 [4] Note: This step is non-optional. Failure to do this step
1485 can result in a significant reduction in the functionality of the
1486 system. Attempting to do it by hand is not recommended and those
1487 that pursue this avenue should read this file carefully, as well
1488 as the archives of freebsd-current and freebsd-hackers mailing lists
1489 for potential gotchas. The -U option is also useful to consider.
1490 See mergemaster(8) for more information.
1492 [5] Usually this step is a noop. However, from time to time
1493 you may need to do this if you get unknown user in the following
1494 step. It never hurts to do it all the time. You may need to
1495 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1496 install) after the buildworld before this step if you last updated
1497 from current before 20020224 or from -stable before 20020408.
1499 [6] This only deletes old files and directories. Old libraries
1500 can be deleted by "make delete-old-libs", but you have to make
1501 sure that no program is using those libraries anymore.
1503 [8] In order to have a kernel that can run the 4.x binaries needed to
1504 do an installworld, you must include the COMPAT_FREEBSD4 option in
1505 your kernel. Failure to do so may leave you with a system that is
1506 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1507 required to run the 5.x binaries on more recent kernels. And so on
1508 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1510 Make sure that you merge any new devices from GENERIC since the
1511 last time you updated your kernel config file.
1513 [9] When checking out sources, you must include the -P flag to have
1514 cvs prune empty directories.
1516 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1517 "?=" instead of the "=" assignment operator, so that buildworld can
1518 override the CPUTYPE if it needs to.
1520 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1521 not on the command line, or in /etc/make.conf. buildworld will
1522 warn if it is improperly defined.
1525 This file contains a list, in reverse chronological order, of major
1526 breakages in tracking -current. It is not guaranteed to be a complete
1527 list of such breakages, and only contains entries since October 10, 2007.
1528 If you need to see UPDATING entries from before that date, you will need
1529 to fetch an UPDATING file from an older FreeBSD release.
1531 Copyright information:
1533 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1535 Redistribution, publication, translation and use, with or without
1536 modification, in full or in part, in any form or format of this
1537 document are permitted without further permission from the author.
1539 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1540 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1541 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1542 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1543 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1544 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1545 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1546 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1547 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1548 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1549 POSSIBILITY OF SUCH DAMAGE.
1551 Contact Warner Losh if you have any questions about your use of