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 http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/makeworld.html
11 Items affecting the ports and packages system can be found in
12 /usr/ports/UPDATING. Please read that file before running portupgrade.
15 The VFS KBI was changed with the merge of several nullfs
16 optimizations and fixes. All filesystem modules must be
20 With the addition of auditdistd(8), a new auditdistd user is now
21 depended on during installworld. "mergemaster -p" can be used to add
22 the user prior to installworld, as documented in the handbook.
28 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
29 Starting with this version the old system of ZFS pool versioning
30 is superseded by "feature flags". This concept enables forward
31 compatibility against certain future changes in functionality of ZFS
32 pools. The first two read-only compatible "feature flags" for ZFS
33 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
34 For more information read the new zpool-features(7) manual page.
35 Please refer to the "ZFS notes" section of this file for information
36 on upgrading boot ZFS pools.
39 The commit introducing bsd.compiler.mk breaks the traditional
40 building of kernels before this point. Add -m ${SRC}/share/mk
41 (for the right value of SRC) to your command lines to work
42 around; update your useland to a point after this; or use the
43 buildkernel/installkernel top-level targets. See also 20120829.
46 The IPFIREWALL_FORWARD kernel option has been removed. Its
47 functionality now turned on by default.
50 The random(4) support for the VIA hardware random number
51 generator (`PADLOCK') is no longer enabled unconditionally.
52 Add the PADLOCK_RNG option in the custom kernel config if
53 needed. The GENERIC kernels on i386 and amd64 do include the
54 option, so the change only affects the custom kernel
58 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
59 the traditional method, you must update your system with an installworld
60 before the kernel will build. The documented make buildkernel/installkernel
61 interfaces (coupled with fresh make kernel-toolchain) continue to work.
64 The sparc64 ZFS loader has been changed to no longer try to auto-
65 detect ZFS providers based on diskN aliases but now requires these
66 to be explicitly listed in the OFW boot-device environment variable.
69 Now unix domain sockets behave "as expected" on nullfs(5). Previously
70 nullfs(5) did not pass through all behaviours to the underlying layer,
71 as a result if we bound to a socket on the lower layer we could connect
72 only to the lower path; if we bound to the upper layer we could connect
73 only to the upper path. The new behavior is one can connect to both the
74 lower and the upper paths regardless what layer path one binds to.
77 The acpi_wmi(4) status device /dev/wmistat has been renamed to
81 A new VOP_ADVISE() was added to support posix_fadvise(2). All
82 filesystem modules must be recompiled.
85 The interface of the VOP_VPTOCNP(9) changed, now the returned
86 vnode shall be referenced, previously it was required to be
87 only held. All in-tree filesystems are converted.
93 The broken amd(4) driver has been replaced with esp(4) in the amd64,
94 i386 and pc98 GENERIC kernel configuration files.
97 This commit modifies vfs_register() so that it uses a hash
98 calculation to set vfc_typenum, which is enabled by default.
99 The first time a system is booted after this change, the
100 vfc_typenum values will change for all file systems. The
101 main effect of this is a change to the NFS server file handles
102 for file systems that use vfc_typenum in their fsid, such as ZFS.
103 It will, however, prevent vfc_typenum from changing when file
104 systems are loaded in a different order for subsequent reboots.
105 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
106 until you are ready to remount all NFS clients after a reboot.
109 Bump the shared library version numbers for libraries that
110 do not use symbol versioning, have changed the ABI compared
111 to stable/8 and which shared library version was not bumped.
112 Done as part of 9.0-RELEASE cycle.
115 During the merge of Capsicum features, the fget(9) KPI was modified.
116 This may require the rebuilding of out-of-tree device drivers --
117 issues have been reported specifically with the nVidia device driver.
118 __FreeBSD_version is bumped to 900041.
120 Also, there is a period between 20110811 and 20110814 where the
121 special devices /dev/{stdin,stdout,stderr} did not work correctly.
122 Building world from a kernel during that window may not work.
125 The packet filter (pf) code has been updated to OpenBSD 4.5.
126 You need to update userland tools to be in sync with kernel.
127 This update breaks backward compatibility with earlier pfsync(4)
128 versions. Care must be taken when updating redundant firewall setups.
131 The following sysctls and tunables are retired on x86 platforms:
133 machdep.hlt_logical_cpus
134 The following sysctl is retired:
135 machdep.hyperthreading_allowed
136 The sysctls were supposed to provide a way to dynamically offline and
137 online selected CPUs on x86 platforms, but the implementation has not
138 been reliable especially with SCHED_ULE scheduler.
139 machdep.hyperthreading_allowed tunable is still available to ignore
140 hyperthreading CPUs at OS level.
141 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
142 where X is an APIC ID of a CPU. Be advised, though, that disabling
143 CPUs in non-uniform fashion will result in non-uniform topology and
144 may lead to sub-optimal system performance with SCHED_ULE, which is
148 cpumask_t type is retired and cpuset_t is used in order to describe
152 Changes to ifconfig(8) for dynamic address family detection mandate
153 that you are running a kernel of 20110525 or later. Make sure to
154 follow the update procedure to boot a new kernel before installing
158 Support for sun4v architecture is officially dropped
161 Several KPI breaking changes have been committed to the mii(4) layer,
162 the PHY drivers and consequently some Ethernet drivers using mii(4).
163 This means that miibus.ko and the modules of the affected Ethernet
164 drivers need to be recompiled.
166 Note to kernel developers: Given that the OUI bit reversion problem
167 was fixed as part of these changes all mii(4) commits related to OUIs,
168 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
169 handling, no longer can be merged verbatim to stable/8 and previous
173 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
174 into their kernel configurations along with 'device pci'.
177 The default NFS client is now the new NFS client, so fstype "newnfs"
178 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
179 Although mounts via fstype "nfs" will usually work without userland
180 changes, it is recommended that the mount(8) and mount_nfs(8)
181 commands be rebuilt from sources and that a link to mount_nfs called
182 mount_oldnfs be created. The new client is compiled into the
183 kernel with "options NFSCL" and this is needed for diskless root
184 file systems. The GENERIC kernel configs have been changed to use
185 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
186 To use the regular/old client, you can "mount -t oldnfs ...". For
187 a diskless root file system, you must also include a line like:
189 vfs.root.mountfrom="oldnfs:"
191 in the boot/loader.conf on the root fs on the NFS server to make
192 a diskless root fs use the old client.
195 The GENERIC kernels for all architectures now default to the new
196 CAM-based ATA stack. It means that all legacy ATA drivers were
197 removed and replaced by respective CAM drivers. If you are using
198 ATA device names in /etc/fstab or other places, make sure to update
199 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
200 where 'Y's are the sequential numbers starting from zero for each type
201 in order of detection, unless configured otherwise with tunables,
202 see cam(4)). There will be symbolic links created in /dev/ to map
203 old adX devices to the respective adaY. They should provide basic
204 compatibility for file systems mounting in most cases, but they do
205 not support old user-level APIs and do not have respective providers
206 in GEOM. Consider using updated management tools with new device names.
208 It is possible to load devices ahci, ata, siis and mvs as modules,
209 but option ATA_CAM should remain in kernel configuration to make ata
210 module work as CAM driver supporting legacy ATA controllers. Device ata
211 still can be used in modular fashion (atacore + ...). Modules atadisk
212 and atapi* are not used and won't affect operation in ATA_CAM mode.
213 Note that to use CAM-based ATA kernel should include CAM devices
214 scbus, pass, da (or explicitly ada), cd and optionally others. All of
215 them are parts of the cam module.
217 ataraid(4) functionality is now supported by the RAID GEOM class.
218 To use it you can load geom_raid kernel module and use graid(8) tool
219 for management. Instead of /dev/arX device names, use /dev/raid/rX.
221 No kernel config options or code have been removed, so if a problem
222 arises, please report it and optionally revert to the old ATA stack.
223 In order to do it you can remove from the kernel config:
228 , and instead add back:
229 device atadisk # ATA disk drives
230 device ataraid # ATA RAID drives
231 device atapicd # ATAPI CDROM drives
232 device atapifd # ATAPI floppy drives
233 device atapist # ATAPI tape drives
236 The default NFS server has been changed to the new server, which
237 was referred to as the experimental server. If you need to switch
238 back to the old NFS server, you must now put the "-o" option on
239 both the mountd and nfsd commands. This can be done using the
240 mountd_flags and nfs_server_flags rc.conf variables until an
241 update to the rc scripts is committed, which is coming soon.
244 The GNU Objective-C runtime library (libobjc), and other Objective-C
245 related components have been removed from the base system. If you
246 require an Objective-C library, please use one of the available ports.
249 ath(4) has been split into bus- and device- modules. if_ath contains
250 the HAL, the TX rate control and the network device code. if_ath_pci
251 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
252 contains the AHB glue. Users need to load both if_ath_pci and if_ath
253 in order to use ath on everything else.
255 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
256 need to load if_ath and if_ath_pci for ath(4) operation.
259 As part of the replacement of sysinstall, the process of building
260 release media has changed significantly. For details, please re-read
261 release(7), which has been updated to reflect the new build process.
264 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
265 is the last available version under GPLv2. It brings a number of new
266 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
267 4.1 and SSE 4.2), better support for powerpc64, a number of new
268 directives, and lots of other small improvements. See the ChangeLog
269 file in contrib/binutils for the full details.
272 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
273 compliant, and will now use half of hash for authentication.
274 This will break interoperability with all stacks (including all
275 actual FreeBSD versions) who implement
276 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
278 The only workaround with such peers is to use another HMAC
279 algorithm for IPsec ("phase 2") authentication.
282 Remove the uio_yield prototype and symbol. This function has
283 been misnamed since it was introduced and should not be
284 globally exposed with this name. The equivalent functionality
285 is now available using kern_yield(curthread->td_user_pri).
286 The function remains undocumented.
289 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
290 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
291 for scalar sysctls is defined but disabled. Code that needs
292 UQUAD to pass the type checking that must compile on older
293 systems where the define is not present can check against
294 __FreeBSD_version >= 900030.
296 The system dialog(1) has been replaced with a new version previously
297 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
298 with the previous version, but the libdialog associated with it has
299 a largely incompatible API. As such, the original version of libdialog
300 will be kept temporarily as libodialog, until its base system consumers
301 are replaced or updated. Bump __FreeBSD_version to 900030.
304 If you are trying to run make universe on a -stable system, and you get
305 the following warning:
306 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
307 unknown. config(8) likely too old."
308 or something similar to it, then you must upgrade your -stable system
309 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
310 stable/8) or build the config from the latest stable/8 branch and
311 install it on your system.
313 Prior to this date, building a current universe on 8-stable system from
314 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
315 error in the first kernel build phase. A new config on those old
316 systems will fix that problem for older versions of -current.
319 The TCP stack has been modified to allow Khelp modules to interact with
320 it via helper hook points and store per-connection data in the TCP
321 control block. Bump __FreeBSD_version to 900029. User space tools that
322 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
326 Generic IEEE 802.3 annex 31B full duplex flow control support has been
327 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
328 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
329 to take advantage of it instead of using custom implementations. This
330 means that these drivers now no longer unconditionally advertise
331 support for flow control but only do so if flow control is a selected
332 media option. This was implemented in the generic support that way in
333 order to allow flow control to be switched on and off via ifconfig(8)
334 with the PHY specific default to typically off in order to protect
335 from unwanted effects. Consequently, if you used flow control with
336 one of the above mentioned drivers you now need to explicitly enable
338 ifconfig bge0 media auto mediaopt flowcontrol
340 Along with the above mentioned changes generic support for setting
341 1000baseT master mode also has been added and brgphy(4), ciphy(4),
342 e1000phy(4) as well as ip1000phy(4) have been converted to take
343 advantage of it. This means that these drivers now no longer take the
344 link0 parameter for selecting master mode but the master media option
345 has to be used instead, for example like in the following:
346 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
348 Selection of master mode now is also available with all other PHY
349 drivers supporting 1000baseT.
352 The TCP stack has received a significant update to add support for
353 modularised congestion control and generally improve the clarity of
354 congestion control decisions. Bump __FreeBSD_version to 900025. User
355 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
356 sockstat) need to be recompiled.
359 The man(1) utility has been replaced by a new version that no longer
360 uses /etc/manpath.config. Please consult man.conf(5) for how to
361 migrate local entries to the new format.
364 The copyright strings printed by login(1) and sshd(8) at the time of a
365 new connection have been removed to follow other operating systems and
369 A workaround for a fixed ld bug has been removed in kernel code,
370 so make sure that your system ld is built from sources after
371 revision 210245 from 2010-07-19 (r211583 if building head kernel
372 on stable/8, r211584 for stable/7; both from 2010-08-21).
373 A symptom of incorrect ld version is different addresses for
374 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
377 The $ipv6_prefer variable in rc.conf(5) has been split into
378 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
380 The $ip6addrctl_policy is a variable to choose a pre-defined
381 address selection policy set by ip6addrctl(8). A value
382 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
385 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
386 flag (see an entry of 20090926) is set on an interface with no
387 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
388 security reason. If you want IPv6 link-local address on all
389 interfaces by default, set this to "YES".
391 The old ipv6_prefer="YES" is equivalent to
392 ipv6_activate_all_interfaces="YES" and
393 ip6addrctl_policy="ipv6_prefer".
396 DTrace has grown support for userland tracing. Due to this, DTrace is
397 now i386 and amd64 only.
398 dtruss(1) is now installed by default on those systems and a new
399 kernel module is needed for userland tracing: fasttrap.
400 No changes to your kernel config file are necessary to enable
401 userland tracing, but you might consider adding 'STRIP=' and
402 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
403 to have informative userland stack traces in DTrace (ustack).
406 The acpi_aiboost(4) driver has been removed in favor of the new
407 aibs(4) driver. You should update your kernel configuration file.
410 BSD grep has been imported to the base system and it is built by
411 default. It is completely BSD licensed, highly GNU-compatible, uses
412 less memory than its GNU counterpart and has a small codebase.
413 However, it is slower than its GNU counterpart, which is mostly
414 noticeable for larger searches, for smaller ones it is measurable
415 but not significant. The reason is complex, the most important factor
416 is that we lack a modern and efficient regex library and GNU
417 overcomes this by optimizing the searches internally. Future work
418 on improving the regex performance is planned, for the meantime,
419 users that need better performance, can build GNU grep instead by
420 setting the WITH_GNU_GREP knob.
423 Due to the import of powerpc64 support, all existing powerpc kernel
424 configuration files must be updated with a machine directive like this:
425 machine powerpc powerpc
427 In addition, an updated config(8) is required to build powerpc kernels
431 A new version of ZFS (version 15) has been merged to -HEAD.
432 This version uses a python library for the following subcommands:
433 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
434 For full functionality of these commands the following port must
435 be installed: sysutils/py-zfs
438 'vm_page's are now hashed by physical address to an array of mutexes.
439 Currently this is only used to serialize access to hold_count. Over
440 time the page queue mutex will be peeled away. This changes the size
441 of pmap on every architecture. And requires all callers of vm_page_hold
442 and vm_page_unhold to be updated.
445 WITH_CTF can now be specified in src.conf (not recommended, there
446 are some problems with static executables), make.conf (would also
447 affect ports which do not use GNU make and do not override the
448 compile targets) or in the kernel config (via "makeoptions
450 When WITH_CTF was specified there before this was silently ignored,
451 so make sure that WITH_CTF is not used in places which could lead
452 to unwanted behavior.
455 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
456 to allow 32-bit compatibility on non-x86 platforms. All kernel
457 configurations on amd64 and ia64 platforms using these options must
458 be modified accordingly.
461 The utmp user accounting database has been replaced with utmpx,
462 the user accounting interface standardized by POSIX.
463 Unfortunately the semantics of utmp and utmpx don't match,
464 making it practically impossible to support both interfaces.
465 The user accounting database is used by tools like finger(1),
466 last(1), talk(1), w(1) and ac(8).
468 All applications in the base system use utmpx. This means only
469 local binaries (e.g. from the ports tree) may still use these
470 utmp database files. These applications must be rebuilt to make
473 After the system has been upgraded, it is safe to remove the old
474 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
475 assuming their contents is of no importance anymore. Old wtmp
476 databases can only be used by last(1) and ac(8) after they have
477 been converted to the new format using wtmpcvt(1).
480 Introduce the kernel thread "deadlock resolver" (which can be enabled
481 via the DEADLKRES option, see NOTES for more details) and the
482 sleepq_type() function for sleepqueues.
485 The rc.firewall and rc.firewall6 were unified, and
486 rc.firewall6 and rc.d/ip6fw were removed.
487 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
488 variables are obsoleted. Instead, the following new rc
489 variables are added to rc.d/ipfw:
491 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
492 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
493 firewall_simple_onet_ipv6, firewall_trusted_ipv6
495 The meanings correspond to the relevant IPv4 variables.
501 The default terminal emulation for syscons(4) has been changed
502 from cons25 to xterm on all platforms except pc98. This means
503 that the /etc/ttys file needs to be updated to ensure correct
504 operation of applications on the console.
506 The terminal emulation style can be toggled per window by using
507 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
508 options can be used to change the compile-time default back to
511 To prevent graphical artifacts, make sure the TERM environment
512 variable is set to match the terminal emulation that is being
513 performed by syscons(4).
516 The layout of the structure ieee80211req_scan_result has changed.
517 Applications that require wireless scan results (e.g. ifconfig(8))
518 from net80211 need to be recompiled.
520 Applications such as wpa_supplicant(8) may require a full world
521 build without using NO_CLEAN in order to get synchronized with the
525 The iwn(4) driver has been updated to support the 5000 and 5150 series.
526 There's one kernel module for each firmware. Adding "device iwnfw"
527 to the kernel configuration file means including all three firmware
528 images inside the kernel. If you want to include just the one for
529 your wireless card, use the devices iwn4965fw, iwn5000fw or
533 The rc.d/network_ipv6, IPv6 configuration script has been integrated
534 into rc.d/netif. The changes are the following:
536 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
537 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
538 Note that both variables need the "inet6" keyword at the head.
540 Do not set $ipv6_network_interfaces manually if you do not
541 understand what you are doing. It is not needed in most cases.
543 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
546 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
547 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
549 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
550 all configured interfaces have "inet6 accept_rtadv" in the
551 $ifconfig_IF_ipv6. These are for backward compatibility.
553 3. A new variable $ipv6_prefer has been added. If NO, IPv6
554 functionality of interfaces with no corresponding
555 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
556 and the default address selection policy of ip6addrctl(8)
557 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
558 Note that if you want to configure IPv6 functionality on the
559 disabled interfaces after boot, first you need to clear the flag by
560 using ifconfig(8) like:
562 ifconfig em0 inet6 -ifdisabled
564 If YES, the default address selection policy is set as
567 The default value of $ipv6_prefer is NO.
569 4. If your system need to receive Router Advertisement messages,
570 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
571 scripts automatically invoke rtsol(8) when the interface becomes
572 UP. The Router Advertisement messages are used for SLAAC
573 (State-Less Address AutoConfiguration).
576 802.11s D3.03 support was committed. This is incompatible with the
577 previous code, which was based on D3.0.
580 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
581 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
582 control whether accepting Router Advertisement messages or not.
583 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
584 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
585 The ifconfig(8) utility now supports these flags.
588 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
589 mount(8) and -a option for df(1) to see them.
592 The old tunable hw.bus.devctl_disable has been superseded by
593 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
594 replaced by hw.bus.devctl_queue=0. The default for this new tunable
598 Remove the option STOP_NMI. The default action is now to use NMI only
599 for KDB via the newly introduced function stop_cpus_hard() and
600 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
603 The stable/8 branch created in subversion. This corresponds to the
604 RELENG_8 branch in CVS.
607 Bump the shared library version numbers for all libraries that do not
608 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
609 __FreeBSD_version to 800105.
612 Due to changes in the implementation of virtual network stack support,
613 all network-related kernel modules must be recompiled. As this change
614 breaks the ABI, bump __FreeBSD_version to 800104.
617 The TOE interface to the TCP syncache has been modified to remove
618 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
619 The cxgb driver is the only TOE consumer affected by this change, and
620 needs to be recompiled along with the kernel. As this change breaks
621 the ABI, bump __FreeBSD_version to 800103.
624 Padding has been added to struct tcpcb, sackhint and tcpstat in
625 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
626 maintaining the ABI. However, this change breaks the ABI, so bump
627 __FreeBSD_version to 800102. User space tools that rely on the size of
628 any of these structs (e.g. sockstat) need to be recompiled.
631 The NFS_LEGACYRPC option has been removed along with the old kernel
632 RPC implementation that this option selected. Kernel configurations
633 may need to be adjusted.
636 The network interface device nodes at /dev/net/<interface> have been
637 removed. All ioctl operations can be performed the normal way using
638 routing sockets. The kqueue functionality can generally be replaced
639 with routing sockets.
642 The documentation from the FreeBSD Documentation Project (Handbook,
643 FAQ, etc.) is now installed via packages by sysinstall(8) and under
644 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
647 The ABI of various structures related to the SYSV IPC API have been
648 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
649 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
653 Layout of struct vnet has changed as routing related variables were
654 moved to their own Vimage module. Modules need to be recompiled. Bump
655 __FreeBSD_version to 800099.
658 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
659 respectively. As long as no more than 16 groups per process are used,
660 no changes should be visible. When more than 16 groups are used, old
661 binaries may fail if they call getgroups() or getgrouplist() with
662 statically sized storage. Recompiling will work around this, but
663 applications should be modified to use dynamically allocated storage
664 for group arrays as POSIX.1-2008 does not cap an implementation's
665 number of supported groups at NGROUPS_MAX+1 as previous versions did.
667 NFS and portalfs mounts may also be affected as the list of groups is
668 truncated to 16. Users of NFS who use more than 16 groups, should
669 take care that negative group permissions are not used on the exported
670 file systems as they will not be reliable unless a GSSAPI based
671 authentication method is used.
674 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
675 option compiles in the support for adaptive spinning for lockmgrs
676 which want to enable it. The lockinit() function now accepts the flag
677 LK_ADAPTIVE in order to make the lock object subject to adaptive
678 spinning when both held in write and read mode.
681 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
682 changed. User applications that use this ioctl need to be rebuilt.
685 The layout of struct thread has changed. Kernel and modules need to
689 The layout of structs ifnet, domain, protosw and vnet_net has changed.
690 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
693 window(1) has been removed from the base system. It can now be
694 installed from ports. The port is called misc/window.
697 The way we are storing and accessing `routing table' entries has
698 changed. Programs reading the FIB, like netstat, need to be
702 A new netisr implementation has been added for FreeBSD 8. Network
703 file system modules, such as igmp, ipdivert, and others, should be
705 Bump __FreeBSD_version to 800096.
708 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
712 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
713 Bump __FreeBSD_version to 800094.
716 Add mnt_xflag field to 'struct mount'. File system modules need to be
718 Bump __FreeBSD_version to 800093.
721 The compiling option ADAPTIVE_SX has been retired while it has been
722 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
723 The KPI for sx_init_flags() changes as accepting flags:
724 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
725 been introduced in order to handle the reversed logic.
726 Bump __FreeBSD_version to 800092.
729 Add support for hierarchical jails. Remove global securelevel.
730 Bump __FreeBSD_version to 800091.
733 The layout of struct vnet_net has changed, therefore modules
735 Bump __FreeBSD_version to 800090.
738 The newly imported zic(8) produces a new format in the output. Please
739 run tzsetup(8) to install the newly created data to /etc/localtime.
742 The sysctl tree for the usb stack has renamed from hw.usb2.* to
743 hw.usb.* and is now consistent again with previous releases.
746 802.11 monitor mode support was revised and driver api's were changed.
747 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
748 of DLT_IEEE802_11. No user-visible data structures were changed but
749 applications that use DLT_IEEE802_11 may require changes.
750 Bump __FreeBSD_version to 800088.
753 The layout of the following structs has changed: sysctl_oid,
754 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
755 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
756 panics may be experienced. World rebuild is required for
757 correctly checking networking state from userland.
758 Bump __FreeBSD_version to 800085.
761 MLDv2 and Source-Specific Multicast (SSM) have been merged
762 to the IPv6 stack. VIMAGE hooks are in but not yet used.
763 The implementation of SSM within FreeBSD's IPv6 stack closely
764 follows the IPv4 implementation.
766 For kernel developers:
768 * The most important changes are that the ip6_output() and
769 ip6_input() paths no longer take the IN6_MULTI_LOCK,
770 and this lock has been downgraded to a non-recursive mutex.
772 * As with the changes to the IPv4 stack to support SSM, filtering
773 of inbound multicast traffic must now be performed by transport
774 protocols within the IPv6 stack. This does not apply to TCP and
775 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
777 * The KPIs used by IPv6 multicast are similar to those used by
778 the IPv4 stack, with the following differences:
779 * im6o_mc_filter() is analogous to imo_multicast_filter().
780 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
781 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
782 * IN6_LOOKUP_MULTI() has been deprecated and removed.
783 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
784 for MLDv1 have an additional 'timer' argument which is used to
785 jitter the initial membership report for the solicited-node
786 multicast membership on-link.
787 * This is not strictly needed for MLDv2, which already jitters
788 its report transmissions. However, the 'timer' argument is
789 preserved in case MLDv1 is active on the interface.
791 * The KAME linked-list based IPv6 membership implementation has
792 been refactored to use a vector similar to that used by the IPv4
794 Code which maintains a list of its own multicast memberships
795 internally, e.g. carp, has been updated to reflect the new
798 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
799 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
800 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
801 implementation constraint which needs to be addressed in HEAD.
803 For application developers:
805 * The changes are broadly similar to those made for the IPv4
808 * The use of IPv4 and IPv6 multicast socket options on the same
809 socket, using mapped addresses, HAS NOT been tested or supported.
811 * There are a number of issues with the implementation of various
812 IPv6 multicast APIs which need to be resolved in the API surface
813 before the implementation is fully compatible with KAME userland
814 use, and these are mostly to do with interface index treatment.
816 * The literature available discusses the use of either the delta / ASM
817 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
818 using setsourcefilter(3)/getsourcefilter(3). For more information
819 please refer to RFC 3768, 'Socket Interface Extensions for
820 Multicast Source Filters'.
822 * Applications which use the published RFC 3678 APIs should be fine.
824 For systems administrators:
826 * The mtest(8) utility has been refactored to support IPv6, in
827 addition to IPv4. Interface addresses are no longer accepted
828 as arguments, their names must be used instead. The utility
829 will map the interface name to its first IPv4 address as
830 returned by getifaddrs(3).
832 * The ifmcstat(8) utility has also been updated to print the MLDv2
833 endpoint state and source filter lists via sysctl(3).
835 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
836 loopback of IPv6 multicast datagrams by default; it defaults to 1
837 to preserve the existing behaviour. Disabling multicast loopback is
838 recommended for optimal system performance.
840 * The IPv6 MROUTING code has been changed to examine this sysctl
841 instead of attempting to perform a group lookup before looping
842 back forwarded datagrams.
844 Bump __FreeBSD_version to 800084.
847 Implement low-level Bluetooth HCI API.
848 Bump __FreeBSD_version to 800083.
851 The layout of struct malloc_type, used by modules to register new
852 memory allocation types, has changed. Most modules will need to
853 be rebuilt or panics may be experienced.
854 Bump __FreeBSD_version to 800081.
857 Anticipate overflowing inp_flags - add inp_flags2.
858 This changes most offsets in inpcb, so checking v4 connection
859 state will require a world rebuild.
860 Bump __FreeBSD_version to 800080.
863 Add an llentry to struct route and struct route_in6. Modules
864 embedding a struct route will need to be recompiled.
865 Bump __FreeBSD_version to 800079.
868 The size of rt_metrics_lite and by extension rtentry has changed.
869 Networking administration apps will need to be recompiled.
870 The route command now supports show as an alias for get, weighting
871 of routes, sticky and nostick flags to alter the behavior of stateful
873 Bump __FreeBSD_version to 800078.
876 Do not use Giant for kbdmux(4) locking. This is wrong and
877 apparently causing more problems than it solves. This will
878 re-open the issue where interrupt handlers may race with
879 kbdmux(4) in polling mode. Typical symptoms include (but
880 not limited to) duplicated and/or missing characters when
881 low level console functions (such as gets) are used while
882 interrupts are enabled (for example geli password prompt,
883 mountroot prompt etc.). Disabling kbdmux(4) may help.
886 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
887 kernel modules referencing any of the above need to be recompiled.
888 Bump __FreeBSD_version to 800075.
891 GEOM_PART has become the default partition slicer for storage devices,
892 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
893 introduces some changes:
895 MSDOS/EBR: the devices created from MSDOS extended partition entries
896 (EBR) can be named differently than with GEOM_MBR and are now symlinks
897 to devices with offset-based names. fstabs may need to be modified.
899 BSD: the "geometry does not match label" warning is harmless in most
900 cases but it points to problems in file system misalignment with
901 disk geometry. The "c" partition is now implicit, covers the whole
902 top-level drive and cannot be (mis)used by users.
904 General: Kernel dumps are now not allowed to be written to devices
905 whose partition types indicate they are meant to be used for file
906 systems (or, in case of MSDOS partitions, as something else than
909 Most of these changes date approximately from 200812.
912 The uscanner(4) driver has been removed from the kernel. This follows
913 Linux removing theirs in 2.6 and making libusb the default interface
917 The multicast forwarding code has been cleaned up. netstat(1)
918 only relies on KVM now for printing bandwidth upcall meters.
919 The IPv4 and IPv6 modules are split into ip_mroute_mod and
920 ip6_mroute_mod respectively. The config(5) options for statically
921 compiling this code remain the same, i.e. 'options MROUTING'.
924 Support for the IFF_NEEDSGIANT network interface flag has been
925 removed, which means that non-MPSAFE network device drivers are no
926 longer supported. In particular, if_ar, if_sr, and network device
927 drivers from the old (legacy) USB stack can no longer be built or
931 POSIX.1 Native Language Support (NLS) has been enabled in libc and
932 a bunch of new language catalog files have also been added.
933 This means that some common libc messages are now localized and
934 they depend on the LC_MESSAGES environmental variable.
937 The k8temp(4) driver has been renamed to amdtemp(4) since
938 support for Family 10 and Family 11 CPU families was added.
941 IGMPv3 and Source-Specific Multicast (SSM) have been merged
942 to the IPv4 stack. VIMAGE hooks are in but not yet used.
944 For kernel developers, the most important changes are that the
945 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
946 and this lock has been downgraded to a non-recursive mutex.
948 Transport protocols (UDP, Raw IP) are now responsible for filtering
949 inbound multicast traffic according to group membership and source
950 filters. The imo_multicast_filter() KPI exists for this purpose.
951 Transports which do not use multicast (SCTP, TCP) already reject
952 multicast by default. Forwarding and receive performance may improve
953 as a mutex acquisition is no longer needed in the ip_input()
954 low-level input path. in_addmulti() and in_delmulti() are shimmed
955 to new KPIs which exist to support SSM in-kernel.
957 For application developers, it is recommended that loopback of
958 multicast datagrams be disabled for best performance, as this
959 will still cause the lock to be taken for each looped-back
960 datagram transmission. The net.inet.ip.mcast.loop sysctl may
961 be tuned to 0 to disable loopback by default; it defaults to 1
962 to preserve the existing behaviour.
964 For systems administrators, to obtain best performance with
965 multicast reception and multiple groups, it is always recommended
966 that a card with a suitably precise hash filter is used. Hash
967 collisions will still result in the lock being taken within the
968 transport protocol input path to check group membership.
970 If deploying FreeBSD in an environment with IGMP snooping switches,
971 it is recommended that the net.inet.igmp.sendlocal sysctl remain
972 enabled; this forces 224.0.0.0/24 group membership to be announced
975 The size of 'struct igmpstat' has changed; netstat needs to be
976 recompiled to reflect this.
977 Bump __FreeBSD_version to 800070.
980 libusb20.so.1 is now installed as libusb.so.1 and the ports system
981 updated to use it. This requires a buildworld/installworld in order to
982 update the library and dependencies (usbconfig, etc). Its advisable to
983 rebuild all ports which uses libusb. More specific directions are given
984 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
985 libusb are no longer required and can be removed.
988 A workaround is committed to allow the creation of System V shared
989 memory segment of size > 2 GB on the 64-bit architectures.
990 Due to a limitation of the existing ABI, the shm_segsz member
991 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
992 wrong for large segments. Note that limits must be explicitly
993 raised to allow such segments to be created.
996 The layout of struct ifnet has changed, requiring a rebuild of all
997 network device driver modules.
1000 The /dev handling for the new USB stack has changed, a
1001 buildworld/installworld is required for libusb20.
1004 The new USB2 stack has now been permanently moved in and all kernel and
1005 module names reverted to their previous values (eg, usb, ehci, ohci,
1006 ums, ...). The old usb stack can be compiled in by prefixing the name
1007 with the letter 'o', the old usb modules have been removed.
1008 Updating entry 20090216 for xorg and 20090215 for libmap may still
1012 The rc.conf(5) option if_up_delay has been renamed to
1013 defaultroute_delay to better reflect its purpose. If you have
1014 customized this setting in /etc/rc.conf you need to update it to
1018 xorg 7.4 wants to configure its input devices via hald which does not
1019 yet work with USB2. If the keyboard/mouse does not work in xorg then
1021 Option "AllowEmptyInput" "off"
1022 to your ServerLayout section. This will cause X to use the configured
1023 kbd and mouse sections from your xorg.conf.
1026 The GENERIC kernels for all architectures now default to the new USB2
1027 stack. No kernel config options or code have been removed so if a
1028 problem arises please report it and optionally revert to the old USB
1029 stack. If you are loading USB kernel modules or have a custom kernel
1030 that includes GENERIC then ensure that usb names are also changed over,
1031 eg uftdi -> usb2_serial_ftdi.
1033 Older programs linked against the ports libusb 0.1 need to be
1034 redirected to the new stack's libusb20. /etc/libmap.conf can
1036 # Map old usb library to new one for usb2 stack
1037 libusb-0.1.so.8 libusb20.so.1
1040 All USB ethernet devices now attach as interfaces under the name ueN
1041 (eg. ue0). This is to provide a predictable name as vendors often
1042 change usb chipsets in a product without notice.
1045 The ichsmb(4) driver has been changed to require SMBus slave
1046 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1047 All of the other SMBus controller drivers require left-justified
1048 slave addresses, so this change makes all the drivers provide the
1052 INET6 statistics (struct ip6stat) was updated.
1053 netstat(1) needs to be recompiled.
1056 NTFS has been removed from GENERIC kernel on amd64 to match
1057 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1058 will load ntfs.ko module automatically when NTFS support is
1059 actually needed, unless ntfs.ko is not installed or security
1060 level prohibits loading kernel modules. If either is the case,
1061 "options NTFS" has to be added into kernel config.
1064 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1065 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1066 800061. User space tools that rely on the size of struct tcpcb in
1067 tcp_var.h (e.g. sockstat) need to be recompiled.
1070 ng_tty(4) module updated to match the new TTY subsystem.
1071 Due to API change, user-level applications must be updated.
1072 New API support added to mpd5 CVS and expected to be present
1073 in next mpd5.3 release.
1076 With __FreeBSD_version 800060 the makefs tool is part of
1077 the base system (it was a port).
1080 The afdata and ifnet locks have been changed from mutexes to
1081 rwlocks, network modules will need to be re-compiled.
1084 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1085 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1086 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1087 architecture and 40 bytes on 64-bit architecture. The userland
1088 applications "arp" and "ndp" have been updated accordingly.
1089 The output from "netstat -r" shows only routing entries and
1090 none of the L2 information.
1093 __FreeBSD_version 800057 marks the switchover from the
1094 binary ath hal to source code. Users must add the line:
1096 options AH_SUPPORT_AR5416
1098 to their kernel config files when specifying:
1102 The ath_hal module no longer exists; the code is now compiled
1103 together with the driver in the ath module. It is now
1104 possible to tailor chip support (i.e. reduce the set of chips
1105 and thereby the code size); consult ath_hal(4) for details.
1108 __FreeBSD_version 800054 adds memory barriers to
1109 <machine/atomic.h>, new interfaces to ifnet to facilitate
1110 multiple hardware transmit queues for cards that support
1111 them, and a lock-less ring-buffer implementation to
1112 enable drivers to more efficiently manage queueing of
1116 A new version of ZFS (version 13) has been merged to -HEAD.
1117 This version has zpool attribute "listsnapshots" off by
1118 default, which means "zfs list" does not show snapshots,
1119 and is the same as Solaris behavior.
1122 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1125 The uhci, ohci, ehci and slhci USB Host controller drivers have
1126 been put into separate modules. If you load the usb module
1127 separately through loader.conf you will need to load the
1128 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1129 controller add the following to loader.conf:
1135 The ABI used by the PMC toolset has changed. Please keep
1136 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1140 atapci kernel module now includes only generic PCI ATA
1141 driver. AHCI driver moved to ataahci kernel module.
1142 All vendor-specific code moved into separate kernel modules:
1143 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1144 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1145 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1146 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1149 The TTY subsystem of the kernel has been replaced by a new
1150 implementation, which provides better scalability and an
1151 improved driver model. Most common drivers have been migrated to
1152 the new TTY subsystem, while others have not. The following
1153 drivers have not yet been ported to the new TTY layer:
1156 cy, digi, rc, rp, sio
1162 ng_h4, ng_tty, ppp, sl, snp
1164 Adding these drivers to your kernel configuration file shall
1165 cause compilation to fail.
1168 ntpd has been upgraded to 4.2.4p5.
1171 OpenSSH has been upgraded to 5.1p1.
1173 For many years, FreeBSD's version of OpenSSH preferred DSA
1174 over RSA for host and user authentication keys. With this
1175 upgrade, we've switched to the vendor's default of RSA over
1176 DSA. This may cause upgraded clients to warn about unknown
1177 host keys even for previously known hosts. Users should
1178 follow the usual procedure for verifying host keys before
1179 accepting the RSA key.
1181 This can be circumvented by setting the "HostKeyAlgorithms"
1182 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1185 Please note that the sequence of keys offered for
1186 authentication has been changed as well. You may want to
1187 specify IdentityFile in a different order to revert this
1191 The sio(4) driver has been removed from the i386 and amd64
1192 kernel configuration files. This means uart(4) is now the
1193 default serial port driver on those platforms as well.
1195 To prevent collisions with the sio(4) driver, the uart(4) driver
1196 uses different names for its device nodes. This means the
1197 onboard serial port will now most likely be called "ttyu0"
1198 instead of "ttyd0". You may need to reconfigure applications to
1199 use the new device names.
1201 When using the serial port as a boot console, be sure to update
1202 /boot/device.hints and /etc/ttys before booting the new kernel.
1203 If you forget to do so, you can still manually specify the hints
1204 at the loader prompt:
1206 set hint.uart.0.at="isa"
1207 set hint.uart.0.port="0x3F8"
1208 set hint.uart.0.flags="0x10"
1209 set hint.uart.0.irq="4"
1213 The gpt(8) utility has been removed. Use gpart(8) to partition
1217 The version that Linuxulator emulates was changed from 2.4.2
1218 to 2.6.16. If you experience any problems with Linux binaries
1219 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1220 if it fixes the problem contact emulation mailing list.
1223 ISDN4BSD (I4B) was removed from the src tree. You may need to
1224 update a your kernel configuration and remove relevant entries.
1227 I have checked in code to support multiple routing tables.
1228 See the man pages setfib(1) and setfib(2).
1229 This is a hopefully backwards compatible version,
1230 but to make use of it you need to compile your kernel
1231 with options ROUTETABLES=2 (or more up to 16).
1234 The 802.11 wireless support was redone to enable multi-bss
1235 operation on devices that are capable. The underlying device
1236 is no longer used directly but instead wlanX devices are
1237 cloned with ifconfig. This requires changes to rc.conf files.
1238 For example, change:
1239 ifconfig_ath0="WPA DHCP"
1242 ifconfig_wlan0="WPA DHCP"
1243 see rc.conf(5) for more details. In addition, mergemaster of
1244 /etc/rc.d is highly recommended. Simultaneous update of userland
1245 and kernel wouldn't hurt either.
1247 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1248 modules were merged into the base wlan module. All references
1249 to these modules (e.g. in kernel config files) must be removed.
1252 psm(4) has gained write(2) support in native operation level.
1253 Arbitrary commands can be written to /dev/psm%d and status can
1254 be read back from it. Therefore, an application is responsible
1255 for status validation and error recovery. It is a no-op in
1256 other operation levels.
1259 Support for KSE threading has been removed from the kernel. To
1260 run legacy applications linked against KSE libmap.conf may
1261 be used. The following libmap.conf may be used to ensure
1262 compatibility with any prior release:
1264 libpthread.so.1 libthr.so.1
1265 libpthread.so.2 libthr.so.2
1266 libkse.so.3 libthr.so.3
1269 The layout of struct vmspace has changed. This affects libkvm
1270 and any executables that link against libkvm and use the
1271 kvm_getprocs() function. In particular, but not exclusively,
1272 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1273 The effects are minimal, but it's advisable to upgrade world
1277 The latest em driver no longer has support in it for the
1278 82575 adapter, this is now moved to the igb driver. The
1279 split was done to make new features that are incompatible
1280 with older hardware easier to do.
1283 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1284 likewise the kernel option is now GEOM_LINUX_LVM.
1287 The default NFS mount mode has changed from UDP to TCP for
1288 increased reliability. If you rely on (insecurely) NFS
1289 mounting across a firewall you may need to update your
1293 Belatedly note the addition of m_collapse for compacting
1297 The fts(3) structures have been changed to use adequate
1298 integer types for their members and so to be able to cope
1299 with huge file trees. The old fts(3) ABI is preserved
1300 through symbol versioning in libc, so third-party binaries
1301 using fts(3) should still work, although they will not take
1302 advantage of the extended types. At the same time, some
1303 third-party software might fail to build after this change
1304 due to unportable assumptions made in its source code about
1305 fts(3) structure members. Such software should be fixed
1306 by its vendor or, in the worst case, in the ports tree.
1307 FreeBSD_version 800015 marks this change for the unlikely
1308 case that a portable fix is impossible.
1311 To upgrade to -current after this date, you must be running
1312 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1313 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1316 The ADAPTIVE_GIANT kernel option has been retired because its
1317 functionality is the default now.
1320 The AT keyboard emulation of sunkbd(4) has been turned on
1321 by default. In order to make the special symbols of the Sun
1322 keyboards driven by sunkbd(4) work under X these now have
1323 to be configured the same way as Sun USB keyboards driven
1324 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1326 Option "XkbLayout" "us"
1327 Option "XkbRules" "xorg"
1328 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1331 It has been decided that it is desirable to provide ABI
1332 backwards compatibility to the FreeBSD 4/5/6 versions of the
1333 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1334 broken with the introduction of PCI domain support (see the
1335 20070930 entry). Unfortunately, this required the ABI of
1336 PCIOCGETCONF to be broken again in order to be able to
1337 provide backwards compatibility to the old version of that
1338 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1339 again. As for prominent ports this affects neither pciutils
1340 nor xorg-server this time, the hal port needs to be rebuilt
1344 The misnamed kthread_create() and friends have been renamed
1345 to kproc_create() etc. Many of the callers already
1346 used kproc_start()..
1347 I will return kthread_create() and friends in a while
1348 with implementations that actually create threads, not procs.
1349 Renaming corresponds with version 800002.
1358 Avoid using make -j when upgrading. While generally safe, there are
1359 sometimes problems using -j to upgrade. If your upgrade fails with
1360 -j, please try again without -j. From time to time in the past there
1361 have been problems using -j with buildworld and/or installworld. This
1362 is especially true when upgrading between "distant" versions (eg one
1363 that cross a major release boundary or several minor releases, or when
1364 several months have passed on the -current branch).
1366 Sometimes, obscure build problems are the result of environment
1367 poisoning. This can happen because the make utility reads its
1368 environment when searching for values for global variables. To run
1369 your build attempts in an "environmental clean room", prefix all make
1370 commands with 'env -i '. See the env(1) manual page for more details.
1372 When upgrading from one major version to another it is generally best
1373 to upgrade to the latest code in the currently installed branch first,
1374 then do an upgrade to the new branch. This is the best-tested upgrade
1375 path, and has the highest probability of being successful. Please try
1376 this approach before reporting problems with a major version upgrade.
1380 When upgrading the boot ZFS pool to a new version, always follow
1383 1.) recompile and reinstall the ZFS boot loader and boot block
1384 (this is part of "make buildworld" and "make installworld")
1386 2.) update the ZFS boot block on your boot drive
1388 The following example updates the ZFS boot block on the first
1389 partition (freebsd-boot) of a GPT partitioned drive ad0:
1390 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1392 Non-boot pools do not need these updates.
1396 If you are updating from a prior version of FreeBSD (even one just
1397 a few days old), you should follow this procedure. It is the most
1398 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1400 make kernel-toolchain
1401 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1402 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1404 To test a kernel once
1405 ---------------------
1406 If you just want to boot a kernel once (because you are not sure
1407 if it works, or if you want to boot a known bad kernel to provide
1408 debugging information) run
1409 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1410 nextboot -k testkernel
1412 To just build a kernel when you know that it won't mess you up
1413 --------------------------------------------------------------
1414 This assumes you are already running a CURRENT system. Replace
1415 ${arch} with the architecture of your machine (e.g. "i386",
1416 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1418 cd src/sys/${arch}/conf
1419 config KERNEL_NAME_HERE
1420 cd ../compile/KERNEL_NAME_HERE
1425 If this fails, go to the "To build a kernel" section.
1427 To rebuild everything and install it on the current system.
1428 -----------------------------------------------------------
1429 # Note: sometimes if you are running current you gotta do more than
1430 # is listed here if you are upgrading from a really old current.
1432 <make sure you have good level 0 dumps>
1434 make kernel KERNCONF=YOUR_KERNEL_HERE
1436 <reboot in single user> [3]
1444 To cross-install current onto a separate partition
1445 --------------------------------------------------
1446 # In this approach we use a separate partition to hold
1447 # current's root, 'usr', and 'var' directories. A partition
1448 # holding "/", "/usr" and "/var" should be about 2GB in
1451 <make sure you have good level 0 dumps>
1454 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1455 <maybe newfs current's root partition>
1456 <mount current's root partition on directory ${CURRENT_ROOT}>
1457 make installworld DESTDIR=${CURRENT_ROOT}
1458 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1459 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1460 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1461 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1462 <reboot into current>
1463 <do a "native" rebuild/install as described in the previous section>
1464 <maybe install compatibility libraries from ports/misc/compat*>
1468 To upgrade in-place from stable to current
1469 ----------------------------------------------
1470 <make sure you have good level 0 dumps>
1472 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1474 <reboot in single user> [3]
1481 Make sure that you've read the UPDATING file to understand the
1482 tweaks to various things you need. At this point in the life
1483 cycle of current, things change often and you are on your own
1484 to cope. The defaults can also change, so please read ALL of
1485 the UPDATING entries.
1487 Also, if you are tracking -current, you must be subscribed to
1488 freebsd-current@freebsd.org. Make sure that before you update
1489 your sources that you have read and understood all the recent
1490 messages there. If in doubt, please track -stable which has
1491 much fewer pitfalls.
1493 [1] If you have third party modules, such as vmware, you
1494 should disable them at this point so they don't crash your
1497 [3] From the bootblocks, boot -s, and then do
1502 adjkerntz -i # if CMOS is wall time
1503 Also, when doing a major release upgrade, it is required that
1504 you boot into single user mode to do the installworld.
1506 [4] Note: This step is non-optional. Failure to do this step
1507 can result in a significant reduction in the functionality of the
1508 system. Attempting to do it by hand is not recommended and those
1509 that pursue this avenue should read this file carefully, as well
1510 as the archives of freebsd-current and freebsd-hackers mailing lists
1511 for potential gotchas. The -U option is also useful to consider.
1512 See mergemaster(8) for more information.
1514 [5] Usually this step is a noop. However, from time to time
1515 you may need to do this if you get unknown user in the following
1516 step. It never hurts to do it all the time. You may need to
1517 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1518 install) after the buildworld before this step if you last updated
1519 from current before 20020224 or from -stable before 20020408.
1521 [6] This only deletes old files and directories. Old libraries
1522 can be deleted by "make delete-old-libs", but you have to make
1523 sure that no program is using those libraries anymore.
1525 [8] In order to have a kernel that can run the 4.x binaries needed to
1526 do an installworld, you must include the COMPAT_FREEBSD4 option in
1527 your kernel. Failure to do so may leave you with a system that is
1528 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1529 required to run the 5.x binaries on more recent kernels. And so on
1530 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1532 Make sure that you merge any new devices from GENERIC since the
1533 last time you updated your kernel config file.
1535 [9] When checking out sources, you must include the -P flag to have
1536 cvs prune empty directories.
1538 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1539 "?=" instead of the "=" assignment operator, so that buildworld can
1540 override the CPUTYPE if it needs to.
1542 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1543 not on the command line, or in /etc/make.conf. buildworld will
1544 warn if it is improperly defined.
1547 This file contains a list, in reverse chronological order, of major
1548 breakages in tracking -current. It is not guaranteed to be a complete
1549 list of such breakages, and only contains entries since October 10, 2007.
1550 If you need to see UPDATING entries from before that date, you will need
1551 to fetch an UPDATING file from an older FreeBSD release.
1553 Copyright information:
1555 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1557 Redistribution, publication, translation and use, with or without
1558 modification, in full or in part, in any form or format of this
1559 document are permitted without further permission from the author.
1561 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1562 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1563 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1564 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1565 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1566 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1567 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1568 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1569 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1570 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1571 POSSIBILITY OF SUCH DAMAGE.
1573 Contact Warner Losh if you have any questions about your use of