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 20131026: p8 FreeBSD-EN-13:04.freebsd-update
13 Fix multiple freebsd-update bugs that break upgrading to
16 20130910: p7 FreeBSD-SA-13:12.ifioctl
17 FreeBSD-SA-13:13.nullfs
19 In IPv6 and NetATM, stop SIOCSIFADDR, SIOCSIFBRDADDR,
20 SIOCSIFDSTADDR and SIOCSIFNETMASK at the socket layer rather
21 than pass them on to the link layer without validation or
22 credential checks. [SA-13:12]
24 Prevent cross-mount hardlinks between different nullfs mounts
25 of the same underlying filesystem. [SA-13:13]
27 20130822: p6 FreeBSD-SA-13:09.ip_multicast
30 Fix an integer overflow in computing the size of a temporary buffer
31 can result in a buffer which is too small for the requested
34 Fix a bug that could lead to kernel memory disclosure with
35 SCTP state cookie. [13:10]
37 Fix a data corruption problem with mfi(4) operating on > 2TB
38 disks in a JBOD. [EN-13:03]
40 20130726: p5 FreeBSD-SA-13:07.bind FreeBSD-SA-13:08.nfsserver
41 Fix Denial of Service vulnerability in named(8). [13:07]
43 Fix a bug that allows remote client bypass the normal
44 access checks when when -network or -host restrictions are
45 used at the same time with -mapall. [13:08]
47 20130618: p4 FreeBSD-SA-13:06.mmap
48 Fix a bug that allowed a tracing process (e.g. gdb) to write
49 to a memory-mapped file in the traced process's address space
50 even if neither the traced process nor the tracing process had
51 write access to that file.
53 20130429: p3 FreeBSD-SA-13:05.nfsserver
54 Fix a bug that allows NFS clients to issue READDIR on files.
56 20130402: p2 FreeBSD-SA-13:03.openssl FreeBSD-SA-13:04.bind
57 Fix multiple vulnerabilities in OpenSSL.
59 Fix Denial of Service vulnerability in named(8).
61 20130218: p1 FreeBSD-SA-13:01.bind FreeBSD-SA-13:02.libc
62 Fix Denial of Service vulnerability in named(8) with DNS64.
64 Fix Denial of Service vulnerability in libc's glob(3) functionality.
70 The sparc64 ZFS loader has been changed to no longer try to auto-
71 detect ZFS providers based on diskN aliases but now requires these
72 to be explicitly listed in the OFW boot-device environment variable.
75 Now unix domain sockets behave "as expected" on nullfs(5). Previously
76 nullfs(5) did not pass through all behaviours to the underlying layer,
77 as a result if we bound to a socket on the lower layer we could connect
78 only to the lower path; if we bound to the upper layer we could connect
79 only to the upper path. The new behavior is one can connect to both the
80 lower and the upper paths regardless what layer path one binds to.
83 The acpi_wmi(4) status device /dev/wmistat has been renamed to
87 A new VOP_ADVISE() was added to support posix_fadvise(2). All
88 filesystem modules must be recompiled.
91 The interface of the VOP_VPTOCNP(9) changed, now the returned
92 vnode shall be referenced, previously it was required to be
93 only held. All in-tree filesystems are converted.
99 The broken amd(4) driver has been replaced with esp(4) in the amd64,
100 i386 and pc98 GENERIC kernel configuration files.
103 This commit modifies vfs_register() so that it uses a hash
104 calculation to set vfc_typenum, which is enabled by default.
105 The first time a system is booted after this change, the
106 vfc_typenum values will change for all file systems. The
107 main effect of this is a change to the NFS server file handles
108 for file systems that use vfc_typenum in their fsid, such as ZFS.
109 It will, however, prevent vfc_typenum from changing when file
110 systems are loaded in a different order for subsequent reboots.
111 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
112 until you are ready to remount all NFS clients after a reboot.
115 Bump the shared library version numbers for libraries that
116 do not use symbol versioning, have changed the ABI compared
117 to stable/8 and which shared library version was not bumped.
118 Done as part of 9.0-RELEASE cycle.
121 During the merge of Capsicum features, the fget(9) KPI was modified.
122 This may require the rebuilding of out-of-tree device drivers --
123 issues have been reported specifically with the nVidia device driver.
124 __FreeBSD_version is bumped to 900041.
126 Also, there is a period between 20110811 and 20110814 where the
127 special devices /dev/{stdin,stdout,stderr} did not work correctly.
128 Building world from a kernel during that window may not work.
131 The packet filter (pf) code has been updated to OpenBSD 4.5.
132 You need to update userland tools to be in sync with kernel.
133 This update breaks backward compatibility with earlier pfsync(4)
134 versions. Care must be taken when updating redundant firewall setups.
137 The following sysctls and tunables are retired on x86 platforms:
139 machdep.hlt_logical_cpus
140 The following sysctl is retired:
141 machdep.hyperthreading_allowed
142 The sysctls were supposed to provide a way to dynamically offline and
143 online selected CPUs on x86 platforms, but the implementation has not
144 been reliable especially with SCHED_ULE scheduler.
145 machdep.hyperthreading_allowed tunable is still available to ignore
146 hyperthreading CPUs at OS level.
147 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
148 where X is an APIC ID of a CPU. Be advised, though, that disabling
149 CPUs in non-uniform fashion will result in non-uniform topology and
150 may lead to sub-optimal system performance with SCHED_ULE, which is
154 cpumask_t type is retired and cpuset_t is used in order to describe
158 Changes to ifconfig(8) for dynamic address family detection mandate
159 that you are running a kernel of 20110525 or later. Make sure to
160 follow the update procedure to boot a new kernel before installing
164 Support for sun4v architecture is officially dropped
167 Several KPI breaking changes have been committed to the mii(4) layer,
168 the PHY drivers and consequently some Ethernet drivers using mii(4).
169 This means that miibus.ko and the modules of the affected Ethernet
170 drivers need to be recompiled.
172 Note to kernel developers: Given that the OUI bit reversion problem
173 was fixed as part of these changes all mii(4) commits related to OUIs,
174 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
175 handling, no longer can be merged verbatim to stable/8 and previous
179 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
180 into their kernel configurations along with 'device pci'.
183 The default NFS client is now the new NFS client, so fstype "newnfs"
184 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
185 Although mounts via fstype "nfs" will usually work without userland
186 changes, it is recommended that the mount(8) and mount_nfs(8)
187 commands be rebuilt from sources and that a link to mount_nfs called
188 mount_oldnfs be created. The new client is compiled into the
189 kernel with "options NFSCL" and this is needed for diskless root
190 file systems. The GENERIC kernel configs have been changed to use
191 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
192 To use the regular/old client, you can "mount -t oldnfs ...". For
193 a diskless root file system, you must also include a line like:
195 vfs.root.mountfrom="oldnfs:"
197 in the boot/loader.conf on the root fs on the NFS server to make
198 a diskless root fs use the old client.
201 The GENERIC kernels for all architectures now default to the new
202 CAM-based ATA stack. It means that all legacy ATA drivers were
203 removed and replaced by respective CAM drivers. If you are using
204 ATA device names in /etc/fstab or other places, make sure to update
205 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
206 where 'Y's are the sequential numbers starting from zero for each type
207 in order of detection, unless configured otherwise with tunables,
208 see cam(4)). There will be symbolic links created in /dev/ to map
209 old adX devices to the respective adaY. They should provide basic
210 compatibility for file systems mounting in most cases, but they do
211 not support old user-level APIs and do not have respective providers
212 in GEOM. Consider using updated management tools with new device names.
214 It is possible to load devices ahci, ata, siis and mvs as modules,
215 but option ATA_CAM should remain in kernel configuration to make ata
216 module work as CAM driver supporting legacy ATA controllers. Device ata
217 still can be used in modular fashion (atacore + ...). Modules atadisk
218 and atapi* are not used and won't affect operation in ATA_CAM mode.
219 Note that to use CAM-based ATA kernel should include CAM devices
220 scbus, pass, da (or explicitly ada), cd and optionally others. All of
221 them are parts of the cam module.
223 ataraid(4) functionality is now supported by the RAID GEOM class.
224 To use it you can load geom_raid kernel module and use graid(8) tool
225 for management. Instead of /dev/arX device names, use /dev/raid/rX.
227 No kernel config options or code have been removed, so if a problem
228 arises, please report it and optionally revert to the old ATA stack.
229 In order to do it you can remove from the kernel config:
234 , and instead add back:
235 device atadisk # ATA disk drives
236 device ataraid # ATA RAID drives
237 device atapicd # ATAPI CDROM drives
238 device atapifd # ATAPI floppy drives
239 device atapist # ATAPI tape drives
242 The default NFS server has been changed to the new server, which
243 was referred to as the experimental server. If you need to switch
244 back to the old NFS server, you must now put the "-o" option on
245 both the mountd and nfsd commands. This can be done using the
246 mountd_flags and nfs_server_flags rc.conf variables until an
247 update to the rc scripts is committed, which is coming soon.
250 The GNU Objective-C runtime library (libobjc), and other Objective-C
251 related components have been removed from the base system. If you
252 require an Objective-C library, please use one of the available ports.
255 ath(4) has been split into bus- and device- modules. if_ath contains
256 the HAL, the TX rate control and the network device code. if_ath_pci
257 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
258 contains the AHB glue. Users need to load both if_ath_pci and if_ath
259 in order to use ath on everything else.
261 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
262 need to load if_ath and if_ath_pci for ath(4) operation.
265 As part of the replacement of sysinstall, the process of building
266 release media has changed significantly. For details, please re-read
267 release(7), which has been updated to reflect the new build process.
270 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
271 is the last available version under GPLv2. It brings a number of new
272 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
273 4.1 and SSE 4.2), better support for powerpc64, a number of new
274 directives, and lots of other small improvements. See the ChangeLog
275 file in contrib/binutils for the full details.
278 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
279 compliant, and will now use half of hash for authentication.
280 This will break interoperability with all stacks (including all
281 actual FreeBSD versions) who implement
282 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
284 The only workaround with such peers is to use another HMAC
285 algorithm for IPsec ("phase 2") authentication.
288 Remove the uio_yield prototype and symbol. This function has
289 been misnamed since it was introduced and should not be
290 globally exposed with this name. The equivalent functionality
291 is now available using kern_yield(curthread->td_user_pri).
292 The function remains undocumented.
295 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
296 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
297 for scalar sysctls is defined but disabled. Code that needs
298 UQUAD to pass the type checking that must compile on older
299 systems where the define is not present can check against
300 __FreeBSD_version >= 900030.
302 The system dialog(1) has been replaced with a new version previously
303 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
304 with the previous version, but the libdialog associated with it has
305 a largely incompatible API. As such, the original version of libdialog
306 will be kept temporarily as libodialog, until its base system consumers
307 are replaced or updated. Bump __FreeBSD_version to 900030.
310 If you are trying to run make universe on a -stable system, and you get
311 the following warning:
312 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
313 unknown. config(8) likely too old."
314 or something similar to it, then you must upgrade your -stable system
315 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
316 stable/8) or build the config from the latest stable/8 branch and
317 install it on your system.
319 Prior to this date, building a current universe on 8-stable system from
320 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
321 error in the first kernel build phase. A new config on those old
322 systems will fix that problem for older versions of -current.
325 The TCP stack has been modified to allow Khelp modules to interact with
326 it via helper hook points and store per-connection data in the TCP
327 control block. Bump __FreeBSD_version to 900029. User space tools that
328 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
332 Generic IEEE 802.3 annex 31B full duplex flow control support has been
333 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
334 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
335 to take advantage of it instead of using custom implementations. This
336 means that these drivers now no longer unconditionally advertise
337 support for flow control but only do so if flow control is a selected
338 media option. This was implemented in the generic support that way in
339 order to allow flow control to be switched on and off via ifconfig(8)
340 with the PHY specific default to typically off in order to protect
341 from unwanted effects. Consequently, if you used flow control with
342 one of the above mentioned drivers you now need to explicitly enable
344 ifconfig bge0 media auto mediaopt flowcontrol
346 Along with the above mentioned changes generic support for setting
347 1000baseT master mode also has been added and brgphy(4), ciphy(4),
348 e1000phy(4) as well as ip1000phy(4) have been converted to take
349 advantage of it. This means that these drivers now no longer take the
350 link0 parameter for selecting master mode but the master media option
351 has to be used instead, for example like in the following:
352 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
354 Selection of master mode now is also available with all other PHY
355 drivers supporting 1000baseT.
358 The TCP stack has received a significant update to add support for
359 modularised congestion control and generally improve the clarity of
360 congestion control decisions. Bump __FreeBSD_version to 900025. User
361 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
362 sockstat) need to be recompiled.
365 The man(1) utility has been replaced by a new version that no longer
366 uses /etc/manpath.config. Please consult man.conf(5) for how to
367 migrate local entries to the new format.
370 The copyright strings printed by login(1) and sshd(8) at the time of a
371 new connection have been removed to follow other operating systems and
375 A workaround for a fixed ld bug has been removed in kernel code,
376 so make sure that your system ld is built from sources after
377 revision 210245 from 2010-07-19 (r211583 if building head kernel
378 on stable/8, r211584 for stable/7; both from 2010-08-21).
379 A symptom of incorrect ld version is different addresses for
380 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
383 The $ipv6_prefer variable in rc.conf(5) has been split into
384 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
386 The $ip6addrctl_policy is a variable to choose a pre-defined
387 address selection policy set by ip6addrctl(8). A value
388 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
391 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
392 flag (see an entry of 20090926) is set on an interface with no
393 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
394 security reason. If you want IPv6 link-local address on all
395 interfaces by default, set this to "YES".
397 The old ipv6_prefer="YES" is equivalent to
398 ipv6_activate_all_interfaces="YES" and
399 ip6addrctl_policy="ipv6_prefer".
402 DTrace has grown support for userland tracing. Due to this, DTrace is
403 now i386 and amd64 only.
404 dtruss(1) is now installed by default on those systems and a new
405 kernel module is needed for userland tracing: fasttrap.
406 No changes to your kernel config file are necessary to enable
407 userland tracing, but you might consider adding 'STRIP=' and
408 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
409 to have informative userland stack traces in DTrace (ustack).
412 The acpi_aiboost(4) driver has been removed in favor of the new
413 aibs(4) driver. You should update your kernel configuration file.
416 BSD grep has been imported to the base system and it is built by
417 default. It is completely BSD licensed, highly GNU-compatible, uses
418 less memory than its GNU counterpart and has a small codebase.
419 However, it is slower than its GNU counterpart, which is mostly
420 noticeable for larger searches, for smaller ones it is measurable
421 but not significant. The reason is complex, the most important factor
422 is that we lack a modern and efficient regex library and GNU
423 overcomes this by optimizing the searches internally. Future work
424 on improving the regex performance is planned, for the meantime,
425 users that need better performance, can build GNU grep instead by
426 setting the WITH_GNU_GREP knob.
429 Due to the import of powerpc64 support, all existing powerpc kernel
430 configuration files must be updated with a machine directive like this:
431 machine powerpc powerpc
433 In addition, an updated config(8) is required to build powerpc kernels
437 A new version of ZFS (version 15) has been merged to -HEAD.
438 This version uses a python library for the following subcommands:
439 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
440 For full functionality of these commands the following port must
441 be installed: sysutils/py-zfs
444 'vm_page's are now hashed by physical address to an array of mutexes.
445 Currently this is only used to serialize access to hold_count. Over
446 time the page queue mutex will be peeled away. This changes the size
447 of pmap on every architecture. And requires all callers of vm_page_hold
448 and vm_page_unhold to be updated.
451 WITH_CTF can now be specified in src.conf (not recommended, there
452 are some problems with static executables), make.conf (would also
453 affect ports which do not use GNU make and do not override the
454 compile targets) or in the kernel config (via "makeoptions
456 When WITH_CTF was specified there before this was silently ignored,
457 so make sure that WITH_CTF is not used in places which could lead
458 to unwanted behavior.
461 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
462 to allow 32-bit compatibility on non-x86 platforms. All kernel
463 configurations on amd64 and ia64 platforms using these options must
464 be modified accordingly.
467 The utmp user accounting database has been replaced with utmpx,
468 the user accounting interface standardized by POSIX.
469 Unfortunately the semantics of utmp and utmpx don't match,
470 making it practically impossible to support both interfaces.
471 The user accounting database is used by tools like finger(1),
472 last(1), talk(1), w(1) and ac(8).
474 All applications in the base system use utmpx. This means only
475 local binaries (e.g. from the ports tree) may still use these
476 utmp database files. These applications must be rebuilt to make
479 After the system has been upgraded, it is safe to remove the old
480 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
481 assuming their contents is of no importance anymore. Old wtmp
482 databases can only be used by last(1) and ac(8) after they have
483 been converted to the new format using wtmpcvt(1).
486 Introduce the kernel thread "deadlock resolver" (which can be enabled
487 via the DEADLKRES option, see NOTES for more details) and the
488 sleepq_type() function for sleepqueues.
491 The rc.firewall and rc.firewall6 were unified, and
492 rc.firewall6 and rc.d/ip6fw were removed.
493 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
494 variables are obsoleted. Instead, the following new rc
495 variables are added to rc.d/ipfw:
497 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
498 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
499 firewall_simple_onet_ipv6, firewall_trusted_ipv6
501 The meanings correspond to the relevant IPv4 variables.
507 The default terminal emulation for syscons(4) has been changed
508 from cons25 to xterm on all platforms except pc98. This means
509 that the /etc/ttys file needs to be updated to ensure correct
510 operation of applications on the console.
512 The terminal emulation style can be toggled per window by using
513 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
514 options can be used to change the compile-time default back to
517 To prevent graphical artifacts, make sure the TERM environment
518 variable is set to match the terminal emulation that is being
519 performed by syscons(4).
522 The layout of the structure ieee80211req_scan_result has changed.
523 Applications that require wireless scan results (e.g. ifconfig(8))
524 from net80211 need to be recompiled.
526 Applications such as wpa_supplicant(8) may require a full world
527 build without using NO_CLEAN in order to get synchronized with the
531 The iwn(4) driver has been updated to support the 5000 and 5150 series.
532 There's one kernel module for each firmware. Adding "device iwnfw"
533 to the kernel configuration file means including all three firmware
534 images inside the kernel. If you want to include just the one for
535 your wireless card, use the devices iwn4965fw, iwn5000fw or
539 The rc.d/network_ipv6, IPv6 configuration script has been integrated
540 into rc.d/netif. The changes are the following:
542 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
543 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
544 Note that both variables need the "inet6" keyword at the head.
546 Do not set $ipv6_network_interfaces manually if you do not
547 understand what you are doing. It is not needed in most cases.
549 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
552 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
553 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
555 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
556 all configured interfaces have "inet6 accept_rtadv" in the
557 $ifconfig_IF_ipv6. These are for backward compatibility.
559 3. A new variable $ipv6_prefer has been added. If NO, IPv6
560 functionality of interfaces with no corresponding
561 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
562 and the default address selection policy of ip6addrctl(8)
563 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
564 Note that if you want to configure IPv6 functionality on the
565 disabled interfaces after boot, first you need to clear the flag by
566 using ifconfig(8) like:
568 ifconfig em0 inet6 -ifdisabled
570 If YES, the default address selection policy is set as
573 The default value of $ipv6_prefer is NO.
575 4. If your system need to receive Router Advertisement messages,
576 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
577 scripts automatically invoke rtsol(8) when the interface becomes
578 UP. The Router Advertisement messages are used for SLAAC
579 (State-Less Address AutoConfiguration).
582 802.11s D3.03 support was committed. This is incompatible with the
583 previous code, which was based on D3.0.
586 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
587 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
588 control whether accepting Router Advertisement messages or not.
589 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
590 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
591 The ifconfig(8) utility now supports these flags.
594 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
595 mount(8) and -a option for df(1) to see them.
598 The old tunable hw.bus.devctl_disable has been superseded by
599 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
600 replaced by hw.bus.devctl_queue=0. The default for this new tunable
604 Remove the option STOP_NMI. The default action is now to use NMI only
605 for KDB via the newly introduced function stop_cpus_hard() and
606 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
609 The stable/8 branch created in subversion. This corresponds to the
610 RELENG_8 branch in CVS.
613 Bump the shared library version numbers for all libraries that do not
614 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
615 __FreeBSD_version to 800105.
618 Due to changes in the implementation of virtual network stack support,
619 all network-related kernel modules must be recompiled. As this change
620 breaks the ABI, bump __FreeBSD_version to 800104.
623 The TOE interface to the TCP syncache has been modified to remove
624 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
625 The cxgb driver is the only TOE consumer affected by this change, and
626 needs to be recompiled along with the kernel. As this change breaks
627 the ABI, bump __FreeBSD_version to 800103.
630 Padding has been added to struct tcpcb, sackhint and tcpstat in
631 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
632 maintaining the ABI. However, this change breaks the ABI, so bump
633 __FreeBSD_version to 800102. User space tools that rely on the size of
634 any of these structs (e.g. sockstat) need to be recompiled.
637 The NFS_LEGACYRPC option has been removed along with the old kernel
638 RPC implementation that this option selected. Kernel configurations
639 may need to be adjusted.
642 The network interface device nodes at /dev/net/<interface> have been
643 removed. All ioctl operations can be performed the normal way using
644 routing sockets. The kqueue functionality can generally be replaced
645 with routing sockets.
648 The documentation from the FreeBSD Documentation Project (Handbook,
649 FAQ, etc.) is now installed via packages by sysinstall(8) and under
650 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
653 The ABI of various structures related to the SYSV IPC API have been
654 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
655 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
659 Layout of struct vnet has changed as routing related variables were
660 moved to their own Vimage module. Modules need to be recompiled. Bump
661 __FreeBSD_version to 800099.
664 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
665 respectively. As long as no more than 16 groups per process are used,
666 no changes should be visible. When more than 16 groups are used, old
667 binaries may fail if they call getgroups() or getgrouplist() with
668 statically sized storage. Recompiling will work around this, but
669 applications should be modified to use dynamically allocated storage
670 for group arrays as POSIX.1-2008 does not cap an implementation's
671 number of supported groups at NGROUPS_MAX+1 as previous versions did.
673 NFS and portalfs mounts may also be affected as the list of groups is
674 truncated to 16. Users of NFS who use more than 16 groups, should
675 take care that negative group permissions are not used on the exported
676 file systems as they will not be reliable unless a GSSAPI based
677 authentication method is used.
680 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
681 option compiles in the support for adaptive spinning for lockmgrs
682 which want to enable it. The lockinit() function now accepts the flag
683 LK_ADAPTIVE in order to make the lock object subject to adaptive
684 spinning when both held in write and read mode.
687 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
688 changed. User applications that use this ioctl need to be rebuilt.
691 The layout of struct thread has changed. Kernel and modules need to
695 The layout of structs ifnet, domain, protosw and vnet_net has changed.
696 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
699 window(1) has been removed from the base system. It can now be
700 installed from ports. The port is called misc/window.
703 The way we are storing and accessing `routing table' entries has
704 changed. Programs reading the FIB, like netstat, need to be
708 A new netisr implementation has been added for FreeBSD 8. Network
709 file system modules, such as igmp, ipdivert, and others, should be
711 Bump __FreeBSD_version to 800096.
714 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
718 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
719 Bump __FreeBSD_version to 800094.
722 Add mnt_xflag field to 'struct mount'. File system modules need to be
724 Bump __FreeBSD_version to 800093.
727 The compiling option ADAPTIVE_SX has been retired while it has been
728 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
729 The KPI for sx_init_flags() changes as accepting flags:
730 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
731 been introduced in order to handle the reversed logic.
732 Bump __FreeBSD_version to 800092.
735 Add support for hierarchical jails. Remove global securelevel.
736 Bump __FreeBSD_version to 800091.
739 The layout of struct vnet_net has changed, therefore modules
741 Bump __FreeBSD_version to 800090.
744 The newly imported zic(8) produces a new format in the output. Please
745 run tzsetup(8) to install the newly created data to /etc/localtime.
748 The sysctl tree for the usb stack has renamed from hw.usb2.* to
749 hw.usb.* and is now consistent again with previous releases.
752 802.11 monitor mode support was revised and driver api's were changed.
753 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
754 of DLT_IEEE802_11. No user-visible data structures were changed but
755 applications that use DLT_IEEE802_11 may require changes.
756 Bump __FreeBSD_version to 800088.
759 The layout of the following structs has changed: sysctl_oid,
760 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
761 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
762 panics may be experienced. World rebuild is required for
763 correctly checking networking state from userland.
764 Bump __FreeBSD_version to 800085.
767 MLDv2 and Source-Specific Multicast (SSM) have been merged
768 to the IPv6 stack. VIMAGE hooks are in but not yet used.
769 The implementation of SSM within FreeBSD's IPv6 stack closely
770 follows the IPv4 implementation.
772 For kernel developers:
774 * The most important changes are that the ip6_output() and
775 ip6_input() paths no longer take the IN6_MULTI_LOCK,
776 and this lock has been downgraded to a non-recursive mutex.
778 * As with the changes to the IPv4 stack to support SSM, filtering
779 of inbound multicast traffic must now be performed by transport
780 protocols within the IPv6 stack. This does not apply to TCP and
781 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
783 * The KPIs used by IPv6 multicast are similar to those used by
784 the IPv4 stack, with the following differences:
785 * im6o_mc_filter() is analogous to imo_multicast_filter().
786 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
787 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
788 * IN6_LOOKUP_MULTI() has been deprecated and removed.
789 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
790 for MLDv1 have an additional 'timer' argument which is used to
791 jitter the initial membership report for the solicited-node
792 multicast membership on-link.
793 * This is not strictly needed for MLDv2, which already jitters
794 its report transmissions. However, the 'timer' argument is
795 preserved in case MLDv1 is active on the interface.
797 * The KAME linked-list based IPv6 membership implementation has
798 been refactored to use a vector similar to that used by the IPv4
800 Code which maintains a list of its own multicast memberships
801 internally, e.g. carp, has been updated to reflect the new
804 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
805 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
806 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
807 implementation constraint which needs to be addressed in HEAD.
809 For application developers:
811 * The changes are broadly similar to those made for the IPv4
814 * The use of IPv4 and IPv6 multicast socket options on the same
815 socket, using mapped addresses, HAS NOT been tested or supported.
817 * There are a number of issues with the implementation of various
818 IPv6 multicast APIs which need to be resolved in the API surface
819 before the implementation is fully compatible with KAME userland
820 use, and these are mostly to do with interface index treatment.
822 * The literature available discusses the use of either the delta / ASM
823 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
824 using setsourcefilter(3)/getsourcefilter(3). For more information
825 please refer to RFC 3768, 'Socket Interface Extensions for
826 Multicast Source Filters'.
828 * Applications which use the published RFC 3678 APIs should be fine.
830 For systems administrators:
832 * The mtest(8) utility has been refactored to support IPv6, in
833 addition to IPv4. Interface addresses are no longer accepted
834 as arguments, their names must be used instead. The utility
835 will map the interface name to its first IPv4 address as
836 returned by getifaddrs(3).
838 * The ifmcstat(8) utility has also been updated to print the MLDv2
839 endpoint state and source filter lists via sysctl(3).
841 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
842 loopback of IPv6 multicast datagrams by default; it defaults to 1
843 to preserve the existing behaviour. Disabling multicast loopback is
844 recommended for optimal system performance.
846 * The IPv6 MROUTING code has been changed to examine this sysctl
847 instead of attempting to perform a group lookup before looping
848 back forwarded datagrams.
850 Bump __FreeBSD_version to 800084.
853 Implement low-level Bluetooth HCI API.
854 Bump __FreeBSD_version to 800083.
857 The layout of struct malloc_type, used by modules to register new
858 memory allocation types, has changed. Most modules will need to
859 be rebuilt or panics may be experienced.
860 Bump __FreeBSD_version to 800081.
863 Anticipate overflowing inp_flags - add inp_flags2.
864 This changes most offsets in inpcb, so checking v4 connection
865 state will require a world rebuild.
866 Bump __FreeBSD_version to 800080.
869 Add an llentry to struct route and struct route_in6. Modules
870 embedding a struct route will need to be recompiled.
871 Bump __FreeBSD_version to 800079.
874 The size of rt_metrics_lite and by extension rtentry has changed.
875 Networking administration apps will need to be recompiled.
876 The route command now supports show as an alias for get, weighting
877 of routes, sticky and nostick flags to alter the behavior of stateful
879 Bump __FreeBSD_version to 800078.
882 Do not use Giant for kbdmux(4) locking. This is wrong and
883 apparently causing more problems than it solves. This will
884 re-open the issue where interrupt handlers may race with
885 kbdmux(4) in polling mode. Typical symptoms include (but
886 not limited to) duplicated and/or missing characters when
887 low level console functions (such as gets) are used while
888 interrupts are enabled (for example geli password prompt,
889 mountroot prompt etc.). Disabling kbdmux(4) may help.
892 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
893 kernel modules referencing any of the above need to be recompiled.
894 Bump __FreeBSD_version to 800075.
897 GEOM_PART has become the default partition slicer for storage devices,
898 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
899 introduces some changes:
901 MSDOS/EBR: the devices created from MSDOS extended partition entries
902 (EBR) can be named differently than with GEOM_MBR and are now symlinks
903 to devices with offset-based names. fstabs may need to be modified.
905 BSD: the "geometry does not match label" warning is harmless in most
906 cases but it points to problems in file system misalignment with
907 disk geometry. The "c" partition is now implicit, covers the whole
908 top-level drive and cannot be (mis)used by users.
910 General: Kernel dumps are now not allowed to be written to devices
911 whose partition types indicate they are meant to be used for file
912 systems (or, in case of MSDOS partitions, as something else than
915 Most of these changes date approximately from 200812.
918 The uscanner(4) driver has been removed from the kernel. This follows
919 Linux removing theirs in 2.6 and making libusb the default interface
923 The multicast forwarding code has been cleaned up. netstat(1)
924 only relies on KVM now for printing bandwidth upcall meters.
925 The IPv4 and IPv6 modules are split into ip_mroute_mod and
926 ip6_mroute_mod respectively. The config(5) options for statically
927 compiling this code remain the same, i.e. 'options MROUTING'.
930 Support for the IFF_NEEDSGIANT network interface flag has been
931 removed, which means that non-MPSAFE network device drivers are no
932 longer supported. In particular, if_ar, if_sr, and network device
933 drivers from the old (legacy) USB stack can no longer be built or
937 POSIX.1 Native Language Support (NLS) has been enabled in libc and
938 a bunch of new language catalog files have also been added.
939 This means that some common libc messages are now localized and
940 they depend on the LC_MESSAGES environmental variable.
943 The k8temp(4) driver has been renamed to amdtemp(4) since
944 support for Family 10 and Family 11 CPU families was added.
947 IGMPv3 and Source-Specific Multicast (SSM) have been merged
948 to the IPv4 stack. VIMAGE hooks are in but not yet used.
950 For kernel developers, the most important changes are that the
951 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
952 and this lock has been downgraded to a non-recursive mutex.
954 Transport protocols (UDP, Raw IP) are now responsible for filtering
955 inbound multicast traffic according to group membership and source
956 filters. The imo_multicast_filter() KPI exists for this purpose.
957 Transports which do not use multicast (SCTP, TCP) already reject
958 multicast by default. Forwarding and receive performance may improve
959 as a mutex acquisition is no longer needed in the ip_input()
960 low-level input path. in_addmulti() and in_delmulti() are shimmed
961 to new KPIs which exist to support SSM in-kernel.
963 For application developers, it is recommended that loopback of
964 multicast datagrams be disabled for best performance, as this
965 will still cause the lock to be taken for each looped-back
966 datagram transmission. The net.inet.ip.mcast.loop sysctl may
967 be tuned to 0 to disable loopback by default; it defaults to 1
968 to preserve the existing behaviour.
970 For systems administrators, to obtain best performance with
971 multicast reception and multiple groups, it is always recommended
972 that a card with a suitably precise hash filter is used. Hash
973 collisions will still result in the lock being taken within the
974 transport protocol input path to check group membership.
976 If deploying FreeBSD in an environment with IGMP snooping switches,
977 it is recommended that the net.inet.igmp.sendlocal sysctl remain
978 enabled; this forces 224.0.0.0/24 group membership to be announced
981 The size of 'struct igmpstat' has changed; netstat needs to be
982 recompiled to reflect this.
983 Bump __FreeBSD_version to 800070.
986 libusb20.so.1 is now installed as libusb.so.1 and the ports system
987 updated to use it. This requires a buildworld/installworld in order to
988 update the library and dependencies (usbconfig, etc). Its advisable to
989 rebuild all ports which uses libusb. More specific directions are given
990 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
991 libusb are no longer required and can be removed.
994 A workaround is committed to allow the creation of System V shared
995 memory segment of size > 2 GB on the 64-bit architectures.
996 Due to a limitation of the existing ABI, the shm_segsz member
997 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
998 wrong for large segments. Note that limits must be explicitly
999 raised to allow such segments to be created.
1002 The layout of struct ifnet has changed, requiring a rebuild of all
1003 network device driver modules.
1006 The /dev handling for the new USB stack has changed, a
1007 buildworld/installworld is required for libusb20.
1010 The new USB2 stack has now been permanently moved in and all kernel and
1011 module names reverted to their previous values (eg, usb, ehci, ohci,
1012 ums, ...). The old usb stack can be compiled in by prefixing the name
1013 with the letter 'o', the old usb modules have been removed.
1014 Updating entry 20090216 for xorg and 20090215 for libmap may still
1018 The rc.conf(5) option if_up_delay has been renamed to
1019 defaultroute_delay to better reflect its purpose. If you have
1020 customized this setting in /etc/rc.conf you need to update it to
1024 xorg 7.4 wants to configure its input devices via hald which does not
1025 yet work with USB2. If the keyboard/mouse does not work in xorg then
1027 Option "AllowEmptyInput" "off"
1028 to your ServerLayout section. This will cause X to use the configured
1029 kbd and mouse sections from your xorg.conf.
1032 The GENERIC kernels for all architectures now default to the new USB2
1033 stack. No kernel config options or code have been removed so if a
1034 problem arises please report it and optionally revert to the old USB
1035 stack. If you are loading USB kernel modules or have a custom kernel
1036 that includes GENERIC then ensure that usb names are also changed over,
1037 eg uftdi -> usb2_serial_ftdi.
1039 Older programs linked against the ports libusb 0.1 need to be
1040 redirected to the new stack's libusb20. /etc/libmap.conf can
1042 # Map old usb library to new one for usb2 stack
1043 libusb-0.1.so.8 libusb20.so.1
1046 All USB ethernet devices now attach as interfaces under the name ueN
1047 (eg. ue0). This is to provide a predictable name as vendors often
1048 change usb chipsets in a product without notice.
1051 The ichsmb(4) driver has been changed to require SMBus slave
1052 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1053 All of the other SMBus controller drivers require left-justified
1054 slave addresses, so this change makes all the drivers provide the
1058 INET6 statistics (struct ip6stat) was updated.
1059 netstat(1) needs to be recompiled.
1062 NTFS has been removed from GENERIC kernel on amd64 to match
1063 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1064 will load ntfs.ko module automatically when NTFS support is
1065 actually needed, unless ntfs.ko is not installed or security
1066 level prohibits loading kernel modules. If either is the case,
1067 "options NTFS" has to be added into kernel config.
1070 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1071 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1072 800061. User space tools that rely on the size of struct tcpcb in
1073 tcp_var.h (e.g. sockstat) need to be recompiled.
1076 ng_tty(4) module updated to match the new TTY subsystem.
1077 Due to API change, user-level applications must be updated.
1078 New API support added to mpd5 CVS and expected to be present
1079 in next mpd5.3 release.
1082 With __FreeBSD_version 800060 the makefs tool is part of
1083 the base system (it was a port).
1086 The afdata and ifnet locks have been changed from mutexes to
1087 rwlocks, network modules will need to be re-compiled.
1090 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1091 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1092 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1093 architecture and 40 bytes on 64-bit architecture. The userland
1094 applications "arp" and "ndp" have been updated accordingly.
1095 The output from "netstat -r" shows only routing entries and
1096 none of the L2 information.
1099 __FreeBSD_version 800057 marks the switchover from the
1100 binary ath hal to source code. Users must add the line:
1102 options AH_SUPPORT_AR5416
1104 to their kernel config files when specifying:
1108 The ath_hal module no longer exists; the code is now compiled
1109 together with the driver in the ath module. It is now
1110 possible to tailor chip support (i.e. reduce the set of chips
1111 and thereby the code size); consult ath_hal(4) for details.
1114 __FreeBSD_version 800054 adds memory barriers to
1115 <machine/atomic.h>, new interfaces to ifnet to facilitate
1116 multiple hardware transmit queues for cards that support
1117 them, and a lock-less ring-buffer implementation to
1118 enable drivers to more efficiently manage queueing of
1122 A new version of ZFS (version 13) has been merged to -HEAD.
1123 This version has zpool attribute "listsnapshots" off by
1124 default, which means "zfs list" does not show snapshots,
1125 and is the same as Solaris behavior.
1128 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1131 The uhci, ohci, ehci and slhci USB Host controller drivers have
1132 been put into separate modules. If you load the usb module
1133 separately through loader.conf you will need to load the
1134 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1135 controller add the following to loader.conf:
1141 The ABI used by the PMC toolset has changed. Please keep
1142 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1146 atapci kernel module now includes only generic PCI ATA
1147 driver. AHCI driver moved to ataahci kernel module.
1148 All vendor-specific code moved into separate kernel modules:
1149 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1150 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1151 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1152 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1155 The TTY subsystem of the kernel has been replaced by a new
1156 implementation, which provides better scalability and an
1157 improved driver model. Most common drivers have been migrated to
1158 the new TTY subsystem, while others have not. The following
1159 drivers have not yet been ported to the new TTY layer:
1162 cy, digi, rc, rp, sio
1168 ng_h4, ng_tty, ppp, sl, snp
1170 Adding these drivers to your kernel configuration file shall
1171 cause compilation to fail.
1174 ntpd has been upgraded to 4.2.4p5.
1177 OpenSSH has been upgraded to 5.1p1.
1179 For many years, FreeBSD's version of OpenSSH preferred DSA
1180 over RSA for host and user authentication keys. With this
1181 upgrade, we've switched to the vendor's default of RSA over
1182 DSA. This may cause upgraded clients to warn about unknown
1183 host keys even for previously known hosts. Users should
1184 follow the usual procedure for verifying host keys before
1185 accepting the RSA key.
1187 This can be circumvented by setting the "HostKeyAlgorithms"
1188 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1191 Please note that the sequence of keys offered for
1192 authentication has been changed as well. You may want to
1193 specify IdentityFile in a different order to revert this
1197 The sio(4) driver has been removed from the i386 and amd64
1198 kernel configuration files. This means uart(4) is now the
1199 default serial port driver on those platforms as well.
1201 To prevent collisions with the sio(4) driver, the uart(4) driver
1202 uses different names for its device nodes. This means the
1203 onboard serial port will now most likely be called "ttyu0"
1204 instead of "ttyd0". You may need to reconfigure applications to
1205 use the new device names.
1207 When using the serial port as a boot console, be sure to update
1208 /boot/device.hints and /etc/ttys before booting the new kernel.
1209 If you forget to do so, you can still manually specify the hints
1210 at the loader prompt:
1212 set hint.uart.0.at="isa"
1213 set hint.uart.0.port="0x3F8"
1214 set hint.uart.0.flags="0x10"
1215 set hint.uart.0.irq="4"
1219 The gpt(8) utility has been removed. Use gpart(8) to partition
1223 The version that Linuxulator emulates was changed from 2.4.2
1224 to 2.6.16. If you experience any problems with Linux binaries
1225 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1226 if it fixes the problem contact emulation mailing list.
1229 ISDN4BSD (I4B) was removed from the src tree. You may need to
1230 update a your kernel configuration and remove relevant entries.
1233 I have checked in code to support multiple routing tables.
1234 See the man pages setfib(1) and setfib(2).
1235 This is a hopefully backwards compatible version,
1236 but to make use of it you need to compile your kernel
1237 with options ROUTETABLES=2 (or more up to 16).
1240 The 802.11 wireless support was redone to enable multi-bss
1241 operation on devices that are capable. The underlying device
1242 is no longer used directly but instead wlanX devices are
1243 cloned with ifconfig. This requires changes to rc.conf files.
1244 For example, change:
1245 ifconfig_ath0="WPA DHCP"
1248 ifconfig_wlan0="WPA DHCP"
1249 see rc.conf(5) for more details. In addition, mergemaster of
1250 /etc/rc.d is highly recommended. Simultaneous update of userland
1251 and kernel wouldn't hurt either.
1253 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1254 modules were merged into the base wlan module. All references
1255 to these modules (e.g. in kernel config files) must be removed.
1258 psm(4) has gained write(2) support in native operation level.
1259 Arbitrary commands can be written to /dev/psm%d and status can
1260 be read back from it. Therefore, an application is responsible
1261 for status validation and error recovery. It is a no-op in
1262 other operation levels.
1265 Support for KSE threading has been removed from the kernel. To
1266 run legacy applications linked against KSE libmap.conf may
1267 be used. The following libmap.conf may be used to ensure
1268 compatibility with any prior release:
1270 libpthread.so.1 libthr.so.1
1271 libpthread.so.2 libthr.so.2
1272 libkse.so.3 libthr.so.3
1275 The layout of struct vmspace has changed. This affects libkvm
1276 and any executables that link against libkvm and use the
1277 kvm_getprocs() function. In particular, but not exclusively,
1278 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1279 The effects are minimal, but it's advisable to upgrade world
1283 The latest em driver no longer has support in it for the
1284 82575 adapter, this is now moved to the igb driver. The
1285 split was done to make new features that are incompatible
1286 with older hardware easier to do.
1289 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1290 likewise the kernel option is now GEOM_LINUX_LVM.
1293 The default NFS mount mode has changed from UDP to TCP for
1294 increased reliability. If you rely on (insecurely) NFS
1295 mounting across a firewall you may need to update your
1299 Belatedly note the addition of m_collapse for compacting
1303 The fts(3) structures have been changed to use adequate
1304 integer types for their members and so to be able to cope
1305 with huge file trees. The old fts(3) ABI is preserved
1306 through symbol versioning in libc, so third-party binaries
1307 using fts(3) should still work, although they will not take
1308 advantage of the extended types. At the same time, some
1309 third-party software might fail to build after this change
1310 due to unportable assumptions made in its source code about
1311 fts(3) structure members. Such software should be fixed
1312 by its vendor or, in the worst case, in the ports tree.
1313 FreeBSD_version 800015 marks this change for the unlikely
1314 case that a portable fix is impossible.
1317 To upgrade to -current after this date, you must be running
1318 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1319 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1322 The ADAPTIVE_GIANT kernel option has been retired because its
1323 functionality is the default now.
1326 The AT keyboard emulation of sunkbd(4) has been turned on
1327 by default. In order to make the special symbols of the Sun
1328 keyboards driven by sunkbd(4) work under X these now have
1329 to be configured the same way as Sun USB keyboards driven
1330 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1332 Option "XkbLayout" "us"
1333 Option "XkbRules" "xorg"
1334 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1337 It has been decided that it is desirable to provide ABI
1338 backwards compatibility to the FreeBSD 4/5/6 versions of the
1339 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1340 broken with the introduction of PCI domain support (see the
1341 20070930 entry). Unfortunately, this required the ABI of
1342 PCIOCGETCONF to be broken again in order to be able to
1343 provide backwards compatibility to the old version of that
1344 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1345 again. As for prominent ports this affects neither pciutils
1346 nor xorg-server this time, the hal port needs to be rebuilt
1350 The misnamed kthread_create() and friends have been renamed
1351 to kproc_create() etc. Many of the callers already
1352 used kproc_start()..
1353 I will return kthread_create() and friends in a while
1354 with implementations that actually create threads, not procs.
1355 Renaming corresponds with version 800002.
1364 Avoid using make -j when upgrading. While generally safe, there are
1365 sometimes problems using -j to upgrade. If your upgrade fails with
1366 -j, please try again without -j. From time to time in the past there
1367 have been problems using -j with buildworld and/or installworld. This
1368 is especially true when upgrading between "distant" versions (eg one
1369 that cross a major release boundary or several minor releases, or when
1370 several months have passed on the -current branch).
1372 Sometimes, obscure build problems are the result of environment
1373 poisoning. This can happen because the make utility reads its
1374 environment when searching for values for global variables. To run
1375 your build attempts in an "environmental clean room", prefix all make
1376 commands with 'env -i '. See the env(1) manual page for more details.
1378 When upgrading from one major version to another it is generally best
1379 to upgrade to the latest code in the currently installed branch first,
1380 then do an upgrade to the new branch. This is the best-tested upgrade
1381 path, and has the highest probability of being successful. Please try
1382 this approach before reporting problems with a major version upgrade.
1386 When upgrading the boot ZFS pool to a new version, always follow
1389 1.) recompile and reinstall the ZFS boot loader and boot block
1390 (this is part of "make buildworld" and "make installworld")
1392 2.) update the ZFS boot block on your boot drive
1394 The following example updates the ZFS boot block on the first
1395 partition (freebsd-boot) of a GPT partitioned drive ad0:
1396 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1398 Non-boot pools do not need these updates.
1402 If you are updating from a prior version of FreeBSD (even one just
1403 a few days old), you should follow this procedure. It is the most
1404 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1406 make kernel-toolchain
1407 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1408 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1410 To test a kernel once
1411 ---------------------
1412 If you just want to boot a kernel once (because you are not sure
1413 if it works, or if you want to boot a known bad kernel to provide
1414 debugging information) run
1415 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1416 nextboot -k testkernel
1418 To just build a kernel when you know that it won't mess you up
1419 --------------------------------------------------------------
1420 This assumes you are already running a CURRENT system. Replace
1421 ${arch} with the architecture of your machine (e.g. "i386",
1422 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1424 cd src/sys/${arch}/conf
1425 config KERNEL_NAME_HERE
1426 cd ../compile/KERNEL_NAME_HERE
1431 If this fails, go to the "To build a kernel" section.
1433 To rebuild everything and install it on the current system.
1434 -----------------------------------------------------------
1435 # Note: sometimes if you are running current you gotta do more than
1436 # is listed here if you are upgrading from a really old current.
1438 <make sure you have good level 0 dumps>
1440 make kernel KERNCONF=YOUR_KERNEL_HERE
1442 <reboot in single user> [3]
1450 To cross-install current onto a separate partition
1451 --------------------------------------------------
1452 # In this approach we use a separate partition to hold
1453 # current's root, 'usr', and 'var' directories. A partition
1454 # holding "/", "/usr" and "/var" should be about 2GB in
1457 <make sure you have good level 0 dumps>
1460 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1461 <maybe newfs current's root partition>
1462 <mount current's root partition on directory ${CURRENT_ROOT}>
1463 make installworld DESTDIR=${CURRENT_ROOT}
1464 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1465 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1466 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1467 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1468 <reboot into current>
1469 <do a "native" rebuild/install as described in the previous section>
1470 <maybe install compatibility libraries from ports/misc/compat*>
1474 To upgrade in-place from 8.x-stable to current
1475 ----------------------------------------------
1476 <make sure you have good level 0 dumps>
1478 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1480 <reboot in single user> [3]
1487 Make sure that you've read the UPDATING file to understand the
1488 tweaks to various things you need. At this point in the life
1489 cycle of current, things change often and you are on your own
1490 to cope. The defaults can also change, so please read ALL of
1491 the UPDATING entries.
1493 Also, if you are tracking -current, you must be subscribed to
1494 freebsd-current@freebsd.org. Make sure that before you update
1495 your sources that you have read and understood all the recent
1496 messages there. If in doubt, please track -stable which has
1497 much fewer pitfalls.
1499 [1] If you have third party modules, such as vmware, you
1500 should disable them at this point so they don't crash your
1503 [3] From the bootblocks, boot -s, and then do
1508 adjkerntz -i # if CMOS is wall time
1509 Also, when doing a major release upgrade, it is required that
1510 you boot into single user mode to do the installworld.
1512 [4] Note: This step is non-optional. Failure to do this step
1513 can result in a significant reduction in the functionality of the
1514 system. Attempting to do it by hand is not recommended and those
1515 that pursue this avenue should read this file carefully, as well
1516 as the archives of freebsd-current and freebsd-hackers mailing lists
1517 for potential gotchas. The -U option is also useful to consider.
1518 See mergemaster(8) for more information.
1520 [5] Usually this step is a noop. However, from time to time
1521 you may need to do this if you get unknown user in the following
1522 step. It never hurts to do it all the time. You may need to
1523 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1524 install) after the buildworld before this step if you last updated
1525 from current before 20020224 or from -stable before 20020408.
1527 [6] This only deletes old files and directories. Old libraries
1528 can be deleted by "make delete-old-libs", but you have to make
1529 sure that no program is using those libraries anymore.
1531 [8] In order to have a kernel that can run the 4.x binaries needed to
1532 do an installworld, you must include the COMPAT_FREEBSD4 option in
1533 your kernel. Failure to do so may leave you with a system that is
1534 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1535 required to run the 5.x binaries on more recent kernels. And so on
1536 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1538 Make sure that you merge any new devices from GENERIC since the
1539 last time you updated your kernel config file.
1541 [9] When checking out sources, you must include the -P flag to have
1542 cvs prune empty directories.
1544 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1545 "?=" instead of the "=" assignment operator, so that buildworld can
1546 override the CPUTYPE if it needs to.
1548 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1549 not on the command line, or in /etc/make.conf. buildworld will
1550 warn if it is improperly defined.
1553 This file contains a list, in reverse chronological order, of major
1554 breakages in tracking -current. It is not guaranteed to be a complete
1555 list of such breakages, and only contains entries since October 10, 2007.
1556 If you need to see UPDATING entries from before that date, you will need
1557 to fetch an UPDATING file from an older FreeBSD release.
1559 Copyright information:
1561 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1563 Redistribution, publication, translation and use, with or without
1564 modification, in full or in part, in any form or format of this
1565 document are permitted without further permission from the author.
1567 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1568 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1569 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1570 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1571 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1572 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1573 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1574 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1575 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1576 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1577 POSSIBILITY OF SUCH DAMAGE.
1579 Contact Warner Losh if you have any questions about your use of