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 20130618: p4 FreeBSD-SA-13:06.mmap
13 Fix a bug that allowed a tracing process (e.g. gdb) to write
14 to a memory-mapped file in the traced process's address space
15 even if neither the traced process nor the tracing process had
16 write access to that file.
18 20130429: p3 FreeBSD-SA-13:05.nfsserver
19 Fix a bug that allows NFS clients to issue READDIR on files.
21 20130402: p2 FreeBSD-SA-13:03.openssl FreeBSD-SA-13:04.bind
22 Fix multiple vulnerabilities in OpenSSL.
24 Fix Denial of Service vulnerability in named(8).
26 20130218: p1 FreeBSD-SA-13:01.bind FreeBSD-SA-13:02.libc
27 Fix Denial of Service vulnerability in named(8) with DNS64.
29 Fix Denial of Service vulnerability in libc's glob(3) functionality.
35 The sparc64 ZFS loader has been changed to no longer try to auto-
36 detect ZFS providers based on diskN aliases but now requires these
37 to be explicitly listed in the OFW boot-device environment variable.
40 Now unix domain sockets behave "as expected" on nullfs(5). Previously
41 nullfs(5) did not pass through all behaviours to the underlying layer,
42 as a result if we bound to a socket on the lower layer we could connect
43 only to the lower path; if we bound to the upper layer we could connect
44 only to the upper path. The new behavior is one can connect to both the
45 lower and the upper paths regardless what layer path one binds to.
48 The acpi_wmi(4) status device /dev/wmistat has been renamed to
52 A new VOP_ADVISE() was added to support posix_fadvise(2). All
53 filesystem modules must be recompiled.
56 The interface of the VOP_VPTOCNP(9) changed, now the returned
57 vnode shall be referenced, previously it was required to be
58 only held. All in-tree filesystems are converted.
64 The broken amd(4) driver has been replaced with esp(4) in the amd64,
65 i386 and pc98 GENERIC kernel configuration files.
68 This commit modifies vfs_register() so that it uses a hash
69 calculation to set vfc_typenum, which is enabled by default.
70 The first time a system is booted after this change, the
71 vfc_typenum values will change for all file systems. The
72 main effect of this is a change to the NFS server file handles
73 for file systems that use vfc_typenum in their fsid, such as ZFS.
74 It will, however, prevent vfc_typenum from changing when file
75 systems are loaded in a different order for subsequent reboots.
76 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
77 until you are ready to remount all NFS clients after a reboot.
80 Bump the shared library version numbers for libraries that
81 do not use symbol versioning, have changed the ABI compared
82 to stable/8 and which shared library version was not bumped.
83 Done as part of 9.0-RELEASE cycle.
86 During the merge of Capsicum features, the fget(9) KPI was modified.
87 This may require the rebuilding of out-of-tree device drivers --
88 issues have been reported specifically with the nVidia device driver.
89 __FreeBSD_version is bumped to 900041.
91 Also, there is a period between 20110811 and 20110814 where the
92 special devices /dev/{stdin,stdout,stderr} did not work correctly.
93 Building world from a kernel during that window may not work.
96 The packet filter (pf) code has been updated to OpenBSD 4.5.
97 You need to update userland tools to be in sync with kernel.
98 This update breaks backward compatibility with earlier pfsync(4)
99 versions. Care must be taken when updating redundant firewall setups.
102 The following sysctls and tunables are retired on x86 platforms:
104 machdep.hlt_logical_cpus
105 The following sysctl is retired:
106 machdep.hyperthreading_allowed
107 The sysctls were supposed to provide a way to dynamically offline and
108 online selected CPUs on x86 platforms, but the implementation has not
109 been reliable especially with SCHED_ULE scheduler.
110 machdep.hyperthreading_allowed tunable is still available to ignore
111 hyperthreading CPUs at OS level.
112 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
113 where X is an APIC ID of a CPU. Be advised, though, that disabling
114 CPUs in non-uniform fashion will result in non-uniform topology and
115 may lead to sub-optimal system performance with SCHED_ULE, which is
119 cpumask_t type is retired and cpuset_t is used in order to describe
123 Changes to ifconfig(8) for dynamic address family detection mandate
124 that you are running a kernel of 20110525 or later. Make sure to
125 follow the update procedure to boot a new kernel before installing
129 Support for sun4v architecture is officially dropped
132 Several KPI breaking changes have been committed to the mii(4) layer,
133 the PHY drivers and consequently some Ethernet drivers using mii(4).
134 This means that miibus.ko and the modules of the affected Ethernet
135 drivers need to be recompiled.
137 Note to kernel developers: Given that the OUI bit reversion problem
138 was fixed as part of these changes all mii(4) commits related to OUIs,
139 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
140 handling, no longer can be merged verbatim to stable/8 and previous
144 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
145 into their kernel configurations along with 'device pci'.
148 The default NFS client is now the new NFS client, so fstype "newnfs"
149 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
150 Although mounts via fstype "nfs" will usually work without userland
151 changes, it is recommended that the mount(8) and mount_nfs(8)
152 commands be rebuilt from sources and that a link to mount_nfs called
153 mount_oldnfs be created. The new client is compiled into the
154 kernel with "options NFSCL" and this is needed for diskless root
155 file systems. The GENERIC kernel configs have been changed to use
156 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
157 To use the regular/old client, you can "mount -t oldnfs ...". For
158 a diskless root file system, you must also include a line like:
160 vfs.root.mountfrom="oldnfs:"
162 in the boot/loader.conf on the root fs on the NFS server to make
163 a diskless root fs use the old client.
166 The GENERIC kernels for all architectures now default to the new
167 CAM-based ATA stack. It means that all legacy ATA drivers were
168 removed and replaced by respective CAM drivers. If you are using
169 ATA device names in /etc/fstab or other places, make sure to update
170 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
171 where 'Y's are the sequential numbers starting from zero for each type
172 in order of detection, unless configured otherwise with tunables,
173 see cam(4)). There will be symbolic links created in /dev/ to map
174 old adX devices to the respective adaY. They should provide basic
175 compatibility for file systems mounting in most cases, but they do
176 not support old user-level APIs and do not have respective providers
177 in GEOM. Consider using updated management tools with new device names.
179 It is possible to load devices ahci, ata, siis and mvs as modules,
180 but option ATA_CAM should remain in kernel configuration to make ata
181 module work as CAM driver supporting legacy ATA controllers. Device ata
182 still can be used in modular fashion (atacore + ...). Modules atadisk
183 and atapi* are not used and won't affect operation in ATA_CAM mode.
184 Note that to use CAM-based ATA kernel should include CAM devices
185 scbus, pass, da (or explicitly ada), cd and optionally others. All of
186 them are parts of the cam module.
188 ataraid(4) functionality is now supported by the RAID GEOM class.
189 To use it you can load geom_raid kernel module and use graid(8) tool
190 for management. Instead of /dev/arX device names, use /dev/raid/rX.
192 No kernel config options or code have been removed, so if a problem
193 arises, please report it and optionally revert to the old ATA stack.
194 In order to do it you can remove from the kernel config:
199 , and instead add back:
200 device atadisk # ATA disk drives
201 device ataraid # ATA RAID drives
202 device atapicd # ATAPI CDROM drives
203 device atapifd # ATAPI floppy drives
204 device atapist # ATAPI tape drives
207 The default NFS server has been changed to the new server, which
208 was referred to as the experimental server. If you need to switch
209 back to the old NFS server, you must now put the "-o" option on
210 both the mountd and nfsd commands. This can be done using the
211 mountd_flags and nfs_server_flags rc.conf variables until an
212 update to the rc scripts is committed, which is coming soon.
215 The GNU Objective-C runtime library (libobjc), and other Objective-C
216 related components have been removed from the base system. If you
217 require an Objective-C library, please use one of the available ports.
220 ath(4) has been split into bus- and device- modules. if_ath contains
221 the HAL, the TX rate control and the network device code. if_ath_pci
222 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
223 contains the AHB glue. Users need to load both if_ath_pci and if_ath
224 in order to use ath on everything else.
226 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
227 need to load if_ath and if_ath_pci for ath(4) operation.
230 As part of the replacement of sysinstall, the process of building
231 release media has changed significantly. For details, please re-read
232 release(7), which has been updated to reflect the new build process.
235 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
236 is the last available version under GPLv2. It brings a number of new
237 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
238 4.1 and SSE 4.2), better support for powerpc64, a number of new
239 directives, and lots of other small improvements. See the ChangeLog
240 file in contrib/binutils for the full details.
243 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
244 compliant, and will now use half of hash for authentication.
245 This will break interoperability with all stacks (including all
246 actual FreeBSD versions) who implement
247 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
249 The only workaround with such peers is to use another HMAC
250 algorithm for IPsec ("phase 2") authentication.
253 Remove the uio_yield prototype and symbol. This function has
254 been misnamed since it was introduced and should not be
255 globally exposed with this name. The equivalent functionality
256 is now available using kern_yield(curthread->td_user_pri).
257 The function remains undocumented.
260 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
261 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
262 for scalar sysctls is defined but disabled. Code that needs
263 UQUAD to pass the type checking that must compile on older
264 systems where the define is not present can check against
265 __FreeBSD_version >= 900030.
267 The system dialog(1) has been replaced with a new version previously
268 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
269 with the previous version, but the libdialog associated with it has
270 a largely incompatible API. As such, the original version of libdialog
271 will be kept temporarily as libodialog, until its base system consumers
272 are replaced or updated. Bump __FreeBSD_version to 900030.
275 If you are trying to run make universe on a -stable system, and you get
276 the following warning:
277 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
278 unknown. config(8) likely too old."
279 or something similar to it, then you must upgrade your -stable system
280 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
281 stable/8) or build the config from the latest stable/8 branch and
282 install it on your system.
284 Prior to this date, building a current universe on 8-stable system from
285 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
286 error in the first kernel build phase. A new config on those old
287 systems will fix that problem for older versions of -current.
290 The TCP stack has been modified to allow Khelp modules to interact with
291 it via helper hook points and store per-connection data in the TCP
292 control block. Bump __FreeBSD_version to 900029. User space tools that
293 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
297 Generic IEEE 802.3 annex 31B full duplex flow control support has been
298 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
299 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
300 to take advantage of it instead of using custom implementations. This
301 means that these drivers now no longer unconditionally advertise
302 support for flow control but only do so if flow control is a selected
303 media option. This was implemented in the generic support that way in
304 order to allow flow control to be switched on and off via ifconfig(8)
305 with the PHY specific default to typically off in order to protect
306 from unwanted effects. Consequently, if you used flow control with
307 one of the above mentioned drivers you now need to explicitly enable
309 ifconfig bge0 media auto mediaopt flowcontrol
311 Along with the above mentioned changes generic support for setting
312 1000baseT master mode also has been added and brgphy(4), ciphy(4),
313 e1000phy(4) as well as ip1000phy(4) have been converted to take
314 advantage of it. This means that these drivers now no longer take the
315 link0 parameter for selecting master mode but the master media option
316 has to be used instead, for example like in the following:
317 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
319 Selection of master mode now is also available with all other PHY
320 drivers supporting 1000baseT.
323 The TCP stack has received a significant update to add support for
324 modularised congestion control and generally improve the clarity of
325 congestion control decisions. Bump __FreeBSD_version to 900025. User
326 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
327 sockstat) need to be recompiled.
330 The man(1) utility has been replaced by a new version that no longer
331 uses /etc/manpath.config. Please consult man.conf(5) for how to
332 migrate local entries to the new format.
335 The copyright strings printed by login(1) and sshd(8) at the time of a
336 new connection have been removed to follow other operating systems and
340 A workaround for a fixed ld bug has been removed in kernel code,
341 so make sure that your system ld is built from sources after
342 revision 210245 from 2010-07-19 (r211583 if building head kernel
343 on stable/8, r211584 for stable/7; both from 2010-08-21).
344 A symptom of incorrect ld version is different addresses for
345 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
348 The $ipv6_prefer variable in rc.conf(5) has been split into
349 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
351 The $ip6addrctl_policy is a variable to choose a pre-defined
352 address selection policy set by ip6addrctl(8). A value
353 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
356 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
357 flag (see an entry of 20090926) is set on an interface with no
358 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
359 security reason. If you want IPv6 link-local address on all
360 interfaces by default, set this to "YES".
362 The old ipv6_prefer="YES" is equivalent to
363 ipv6_activate_all_interfaces="YES" and
364 ip6addrctl_policy="ipv6_prefer".
367 DTrace has grown support for userland tracing. Due to this, DTrace is
368 now i386 and amd64 only.
369 dtruss(1) is now installed by default on those systems and a new
370 kernel module is needed for userland tracing: fasttrap.
371 No changes to your kernel config file are necessary to enable
372 userland tracing, but you might consider adding 'STRIP=' and
373 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
374 to have informative userland stack traces in DTrace (ustack).
377 The acpi_aiboost(4) driver has been removed in favor of the new
378 aibs(4) driver. You should update your kernel configuration file.
381 BSD grep has been imported to the base system and it is built by
382 default. It is completely BSD licensed, highly GNU-compatible, uses
383 less memory than its GNU counterpart and has a small codebase.
384 However, it is slower than its GNU counterpart, which is mostly
385 noticeable for larger searches, for smaller ones it is measurable
386 but not significant. The reason is complex, the most important factor
387 is that we lack a modern and efficient regex library and GNU
388 overcomes this by optimizing the searches internally. Future work
389 on improving the regex performance is planned, for the meantime,
390 users that need better performance, can build GNU grep instead by
391 setting the WITH_GNU_GREP knob.
394 Due to the import of powerpc64 support, all existing powerpc kernel
395 configuration files must be updated with a machine directive like this:
396 machine powerpc powerpc
398 In addition, an updated config(8) is required to build powerpc kernels
402 A new version of ZFS (version 15) has been merged to -HEAD.
403 This version uses a python library for the following subcommands:
404 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
405 For full functionality of these commands the following port must
406 be installed: sysutils/py-zfs
409 'vm_page's are now hashed by physical address to an array of mutexes.
410 Currently this is only used to serialize access to hold_count. Over
411 time the page queue mutex will be peeled away. This changes the size
412 of pmap on every architecture. And requires all callers of vm_page_hold
413 and vm_page_unhold to be updated.
416 WITH_CTF can now be specified in src.conf (not recommended, there
417 are some problems with static executables), make.conf (would also
418 affect ports which do not use GNU make and do not override the
419 compile targets) or in the kernel config (via "makeoptions
421 When WITH_CTF was specified there before this was silently ignored,
422 so make sure that WITH_CTF is not used in places which could lead
423 to unwanted behavior.
426 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
427 to allow 32-bit compatibility on non-x86 platforms. All kernel
428 configurations on amd64 and ia64 platforms using these options must
429 be modified accordingly.
432 The utmp user accounting database has been replaced with utmpx,
433 the user accounting interface standardized by POSIX.
434 Unfortunately the semantics of utmp and utmpx don't match,
435 making it practically impossible to support both interfaces.
436 The user accounting database is used by tools like finger(1),
437 last(1), talk(1), w(1) and ac(8).
439 All applications in the base system use utmpx. This means only
440 local binaries (e.g. from the ports tree) may still use these
441 utmp database files. These applications must be rebuilt to make
444 After the system has been upgraded, it is safe to remove the old
445 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
446 assuming their contents is of no importance anymore. Old wtmp
447 databases can only be used by last(1) and ac(8) after they have
448 been converted to the new format using wtmpcvt(1).
451 Introduce the kernel thread "deadlock resolver" (which can be enabled
452 via the DEADLKRES option, see NOTES for more details) and the
453 sleepq_type() function for sleepqueues.
456 The rc.firewall and rc.firewall6 were unified, and
457 rc.firewall6 and rc.d/ip6fw were removed.
458 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
459 variables are obsoleted. Instead, the following new rc
460 variables are added to rc.d/ipfw:
462 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
463 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
464 firewall_simple_onet_ipv6, firewall_trusted_ipv6
466 The meanings correspond to the relevant IPv4 variables.
472 The default terminal emulation for syscons(4) has been changed
473 from cons25 to xterm on all platforms except pc98. This means
474 that the /etc/ttys file needs to be updated to ensure correct
475 operation of applications on the console.
477 The terminal emulation style can be toggled per window by using
478 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
479 options can be used to change the compile-time default back to
482 To prevent graphical artifacts, make sure the TERM environment
483 variable is set to match the terminal emulation that is being
484 performed by syscons(4).
487 The layout of the structure ieee80211req_scan_result has changed.
488 Applications that require wireless scan results (e.g. ifconfig(8))
489 from net80211 need to be recompiled.
491 Applications such as wpa_supplicant(8) may require a full world
492 build without using NO_CLEAN in order to get synchronized with the
496 The iwn(4) driver has been updated to support the 5000 and 5150 series.
497 There's one kernel module for each firmware. Adding "device iwnfw"
498 to the kernel configuration file means including all three firmware
499 images inside the kernel. If you want to include just the one for
500 your wireless card, use the devices iwn4965fw, iwn5000fw or
504 The rc.d/network_ipv6, IPv6 configuration script has been integrated
505 into rc.d/netif. The changes are the following:
507 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
508 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
509 Note that both variables need the "inet6" keyword at the head.
511 Do not set $ipv6_network_interfaces manually if you do not
512 understand what you are doing. It is not needed in most cases.
514 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
517 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
518 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
520 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
521 all configured interfaces have "inet6 accept_rtadv" in the
522 $ifconfig_IF_ipv6. These are for backward compatibility.
524 3. A new variable $ipv6_prefer has been added. If NO, IPv6
525 functionality of interfaces with no corresponding
526 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
527 and the default address selection policy of ip6addrctl(8)
528 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
529 Note that if you want to configure IPv6 functionality on the
530 disabled interfaces after boot, first you need to clear the flag by
531 using ifconfig(8) like:
533 ifconfig em0 inet6 -ifdisabled
535 If YES, the default address selection policy is set as
538 The default value of $ipv6_prefer is NO.
540 4. If your system need to receive Router Advertisement messages,
541 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
542 scripts automatically invoke rtsol(8) when the interface becomes
543 UP. The Router Advertisement messages are used for SLAAC
544 (State-Less Address AutoConfiguration).
547 802.11s D3.03 support was committed. This is incompatible with the
548 previous code, which was based on D3.0.
551 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
552 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
553 control whether accepting Router Advertisement messages or not.
554 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
555 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
556 The ifconfig(8) utility now supports these flags.
559 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
560 mount(8) and -a option for df(1) to see them.
563 The old tunable hw.bus.devctl_disable has been superseded by
564 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
565 replaced by hw.bus.devctl_queue=0. The default for this new tunable
569 Remove the option STOP_NMI. The default action is now to use NMI only
570 for KDB via the newly introduced function stop_cpus_hard() and
571 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
574 The stable/8 branch created in subversion. This corresponds to the
575 RELENG_8 branch in CVS.
578 Bump the shared library version numbers for all libraries that do not
579 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
580 __FreeBSD_version to 800105.
583 Due to changes in the implementation of virtual network stack support,
584 all network-related kernel modules must be recompiled. As this change
585 breaks the ABI, bump __FreeBSD_version to 800104.
588 The TOE interface to the TCP syncache has been modified to remove
589 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
590 The cxgb driver is the only TOE consumer affected by this change, and
591 needs to be recompiled along with the kernel. As this change breaks
592 the ABI, bump __FreeBSD_version to 800103.
595 Padding has been added to struct tcpcb, sackhint and tcpstat in
596 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
597 maintaining the ABI. However, this change breaks the ABI, so bump
598 __FreeBSD_version to 800102. User space tools that rely on the size of
599 any of these structs (e.g. sockstat) need to be recompiled.
602 The NFS_LEGACYRPC option has been removed along with the old kernel
603 RPC implementation that this option selected. Kernel configurations
604 may need to be adjusted.
607 The network interface device nodes at /dev/net/<interface> have been
608 removed. All ioctl operations can be performed the normal way using
609 routing sockets. The kqueue functionality can generally be replaced
610 with routing sockets.
613 The documentation from the FreeBSD Documentation Project (Handbook,
614 FAQ, etc.) is now installed via packages by sysinstall(8) and under
615 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
618 The ABI of various structures related to the SYSV IPC API have been
619 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
620 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
624 Layout of struct vnet has changed as routing related variables were
625 moved to their own Vimage module. Modules need to be recompiled. Bump
626 __FreeBSD_version to 800099.
629 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
630 respectively. As long as no more than 16 groups per process are used,
631 no changes should be visible. When more than 16 groups are used, old
632 binaries may fail if they call getgroups() or getgrouplist() with
633 statically sized storage. Recompiling will work around this, but
634 applications should be modified to use dynamically allocated storage
635 for group arrays as POSIX.1-2008 does not cap an implementation's
636 number of supported groups at NGROUPS_MAX+1 as previous versions did.
638 NFS and portalfs mounts may also be affected as the list of groups is
639 truncated to 16. Users of NFS who use more than 16 groups, should
640 take care that negative group permissions are not used on the exported
641 file systems as they will not be reliable unless a GSSAPI based
642 authentication method is used.
645 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
646 option compiles in the support for adaptive spinning for lockmgrs
647 which want to enable it. The lockinit() function now accepts the flag
648 LK_ADAPTIVE in order to make the lock object subject to adaptive
649 spinning when both held in write and read mode.
652 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
653 changed. User applications that use this ioctl need to be rebuilt.
656 The layout of struct thread has changed. Kernel and modules need to
660 The layout of structs ifnet, domain, protosw and vnet_net has changed.
661 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
664 window(1) has been removed from the base system. It can now be
665 installed from ports. The port is called misc/window.
668 The way we are storing and accessing `routing table' entries has
669 changed. Programs reading the FIB, like netstat, need to be
673 A new netisr implementation has been added for FreeBSD 8. Network
674 file system modules, such as igmp, ipdivert, and others, should be
676 Bump __FreeBSD_version to 800096.
679 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
683 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
684 Bump __FreeBSD_version to 800094.
687 Add mnt_xflag field to 'struct mount'. File system modules need to be
689 Bump __FreeBSD_version to 800093.
692 The compiling option ADAPTIVE_SX has been retired while it has been
693 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
694 The KPI for sx_init_flags() changes as accepting flags:
695 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
696 been introduced in order to handle the reversed logic.
697 Bump __FreeBSD_version to 800092.
700 Add support for hierarchical jails. Remove global securelevel.
701 Bump __FreeBSD_version to 800091.
704 The layout of struct vnet_net has changed, therefore modules
706 Bump __FreeBSD_version to 800090.
709 The newly imported zic(8) produces a new format in the output. Please
710 run tzsetup(8) to install the newly created data to /etc/localtime.
713 The sysctl tree for the usb stack has renamed from hw.usb2.* to
714 hw.usb.* and is now consistent again with previous releases.
717 802.11 monitor mode support was revised and driver api's were changed.
718 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
719 of DLT_IEEE802_11. No user-visible data structures were changed but
720 applications that use DLT_IEEE802_11 may require changes.
721 Bump __FreeBSD_version to 800088.
724 The layout of the following structs has changed: sysctl_oid,
725 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
726 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
727 panics may be experienced. World rebuild is required for
728 correctly checking networking state from userland.
729 Bump __FreeBSD_version to 800085.
732 MLDv2 and Source-Specific Multicast (SSM) have been merged
733 to the IPv6 stack. VIMAGE hooks are in but not yet used.
734 The implementation of SSM within FreeBSD's IPv6 stack closely
735 follows the IPv4 implementation.
737 For kernel developers:
739 * The most important changes are that the ip6_output() and
740 ip6_input() paths no longer take the IN6_MULTI_LOCK,
741 and this lock has been downgraded to a non-recursive mutex.
743 * As with the changes to the IPv4 stack to support SSM, filtering
744 of inbound multicast traffic must now be performed by transport
745 protocols within the IPv6 stack. This does not apply to TCP and
746 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
748 * The KPIs used by IPv6 multicast are similar to those used by
749 the IPv4 stack, with the following differences:
750 * im6o_mc_filter() is analogous to imo_multicast_filter().
751 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
752 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
753 * IN6_LOOKUP_MULTI() has been deprecated and removed.
754 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
755 for MLDv1 have an additional 'timer' argument which is used to
756 jitter the initial membership report for the solicited-node
757 multicast membership on-link.
758 * This is not strictly needed for MLDv2, which already jitters
759 its report transmissions. However, the 'timer' argument is
760 preserved in case MLDv1 is active on the interface.
762 * The KAME linked-list based IPv6 membership implementation has
763 been refactored to use a vector similar to that used by the IPv4
765 Code which maintains a list of its own multicast memberships
766 internally, e.g. carp, has been updated to reflect the new
769 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
770 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
771 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
772 implementation constraint which needs to be addressed in HEAD.
774 For application developers:
776 * The changes are broadly similar to those made for the IPv4
779 * The use of IPv4 and IPv6 multicast socket options on the same
780 socket, using mapped addresses, HAS NOT been tested or supported.
782 * There are a number of issues with the implementation of various
783 IPv6 multicast APIs which need to be resolved in the API surface
784 before the implementation is fully compatible with KAME userland
785 use, and these are mostly to do with interface index treatment.
787 * The literature available discusses the use of either the delta / ASM
788 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
789 using setsourcefilter(3)/getsourcefilter(3). For more information
790 please refer to RFC 3768, 'Socket Interface Extensions for
791 Multicast Source Filters'.
793 * Applications which use the published RFC 3678 APIs should be fine.
795 For systems administrators:
797 * The mtest(8) utility has been refactored to support IPv6, in
798 addition to IPv4. Interface addresses are no longer accepted
799 as arguments, their names must be used instead. The utility
800 will map the interface name to its first IPv4 address as
801 returned by getifaddrs(3).
803 * The ifmcstat(8) utility has also been updated to print the MLDv2
804 endpoint state and source filter lists via sysctl(3).
806 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
807 loopback of IPv6 multicast datagrams by default; it defaults to 1
808 to preserve the existing behaviour. Disabling multicast loopback is
809 recommended for optimal system performance.
811 * The IPv6 MROUTING code has been changed to examine this sysctl
812 instead of attempting to perform a group lookup before looping
813 back forwarded datagrams.
815 Bump __FreeBSD_version to 800084.
818 Implement low-level Bluetooth HCI API.
819 Bump __FreeBSD_version to 800083.
822 The layout of struct malloc_type, used by modules to register new
823 memory allocation types, has changed. Most modules will need to
824 be rebuilt or panics may be experienced.
825 Bump __FreeBSD_version to 800081.
828 Anticipate overflowing inp_flags - add inp_flags2.
829 This changes most offsets in inpcb, so checking v4 connection
830 state will require a world rebuild.
831 Bump __FreeBSD_version to 800080.
834 Add an llentry to struct route and struct route_in6. Modules
835 embedding a struct route will need to be recompiled.
836 Bump __FreeBSD_version to 800079.
839 The size of rt_metrics_lite and by extension rtentry has changed.
840 Networking administration apps will need to be recompiled.
841 The route command now supports show as an alias for get, weighting
842 of routes, sticky and nostick flags to alter the behavior of stateful
844 Bump __FreeBSD_version to 800078.
847 Do not use Giant for kbdmux(4) locking. This is wrong and
848 apparently causing more problems than it solves. This will
849 re-open the issue where interrupt handlers may race with
850 kbdmux(4) in polling mode. Typical symptoms include (but
851 not limited to) duplicated and/or missing characters when
852 low level console functions (such as gets) are used while
853 interrupts are enabled (for example geli password prompt,
854 mountroot prompt etc.). Disabling kbdmux(4) may help.
857 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
858 kernel modules referencing any of the above need to be recompiled.
859 Bump __FreeBSD_version to 800075.
862 GEOM_PART has become the default partition slicer for storage devices,
863 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
864 introduces some changes:
866 MSDOS/EBR: the devices created from MSDOS extended partition entries
867 (EBR) can be named differently than with GEOM_MBR and are now symlinks
868 to devices with offset-based names. fstabs may need to be modified.
870 BSD: the "geometry does not match label" warning is harmless in most
871 cases but it points to problems in file system misalignment with
872 disk geometry. The "c" partition is now implicit, covers the whole
873 top-level drive and cannot be (mis)used by users.
875 General: Kernel dumps are now not allowed to be written to devices
876 whose partition types indicate they are meant to be used for file
877 systems (or, in case of MSDOS partitions, as something else than
880 Most of these changes date approximately from 200812.
883 The uscanner(4) driver has been removed from the kernel. This follows
884 Linux removing theirs in 2.6 and making libusb the default interface
888 The multicast forwarding code has been cleaned up. netstat(1)
889 only relies on KVM now for printing bandwidth upcall meters.
890 The IPv4 and IPv6 modules are split into ip_mroute_mod and
891 ip6_mroute_mod respectively. The config(5) options for statically
892 compiling this code remain the same, i.e. 'options MROUTING'.
895 Support for the IFF_NEEDSGIANT network interface flag has been
896 removed, which means that non-MPSAFE network device drivers are no
897 longer supported. In particular, if_ar, if_sr, and network device
898 drivers from the old (legacy) USB stack can no longer be built or
902 POSIX.1 Native Language Support (NLS) has been enabled in libc and
903 a bunch of new language catalog files have also been added.
904 This means that some common libc messages are now localized and
905 they depend on the LC_MESSAGES environmental variable.
908 The k8temp(4) driver has been renamed to amdtemp(4) since
909 support for Family 10 and Family 11 CPU families was added.
912 IGMPv3 and Source-Specific Multicast (SSM) have been merged
913 to the IPv4 stack. VIMAGE hooks are in but not yet used.
915 For kernel developers, the most important changes are that the
916 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
917 and this lock has been downgraded to a non-recursive mutex.
919 Transport protocols (UDP, Raw IP) are now responsible for filtering
920 inbound multicast traffic according to group membership and source
921 filters. The imo_multicast_filter() KPI exists for this purpose.
922 Transports which do not use multicast (SCTP, TCP) already reject
923 multicast by default. Forwarding and receive performance may improve
924 as a mutex acquisition is no longer needed in the ip_input()
925 low-level input path. in_addmulti() and in_delmulti() are shimmed
926 to new KPIs which exist to support SSM in-kernel.
928 For application developers, it is recommended that loopback of
929 multicast datagrams be disabled for best performance, as this
930 will still cause the lock to be taken for each looped-back
931 datagram transmission. The net.inet.ip.mcast.loop sysctl may
932 be tuned to 0 to disable loopback by default; it defaults to 1
933 to preserve the existing behaviour.
935 For systems administrators, to obtain best performance with
936 multicast reception and multiple groups, it is always recommended
937 that a card with a suitably precise hash filter is used. Hash
938 collisions will still result in the lock being taken within the
939 transport protocol input path to check group membership.
941 If deploying FreeBSD in an environment with IGMP snooping switches,
942 it is recommended that the net.inet.igmp.sendlocal sysctl remain
943 enabled; this forces 224.0.0.0/24 group membership to be announced
946 The size of 'struct igmpstat' has changed; netstat needs to be
947 recompiled to reflect this.
948 Bump __FreeBSD_version to 800070.
951 libusb20.so.1 is now installed as libusb.so.1 and the ports system
952 updated to use it. This requires a buildworld/installworld in order to
953 update the library and dependencies (usbconfig, etc). Its advisable to
954 rebuild all ports which uses libusb. More specific directions are given
955 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
956 libusb are no longer required and can be removed.
959 A workaround is committed to allow the creation of System V shared
960 memory segment of size > 2 GB on the 64-bit architectures.
961 Due to a limitation of the existing ABI, the shm_segsz member
962 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
963 wrong for large segments. Note that limits must be explicitly
964 raised to allow such segments to be created.
967 The layout of struct ifnet has changed, requiring a rebuild of all
968 network device driver modules.
971 The /dev handling for the new USB stack has changed, a
972 buildworld/installworld is required for libusb20.
975 The new USB2 stack has now been permanently moved in and all kernel and
976 module names reverted to their previous values (eg, usb, ehci, ohci,
977 ums, ...). The old usb stack can be compiled in by prefixing the name
978 with the letter 'o', the old usb modules have been removed.
979 Updating entry 20090216 for xorg and 20090215 for libmap may still
983 The rc.conf(5) option if_up_delay has been renamed to
984 defaultroute_delay to better reflect its purpose. If you have
985 customized this setting in /etc/rc.conf you need to update it to
989 xorg 7.4 wants to configure its input devices via hald which does not
990 yet work with USB2. If the keyboard/mouse does not work in xorg then
992 Option "AllowEmptyInput" "off"
993 to your ServerLayout section. This will cause X to use the configured
994 kbd and mouse sections from your xorg.conf.
997 The GENERIC kernels for all architectures now default to the new USB2
998 stack. No kernel config options or code have been removed so if a
999 problem arises please report it and optionally revert to the old USB
1000 stack. If you are loading USB kernel modules or have a custom kernel
1001 that includes GENERIC then ensure that usb names are also changed over,
1002 eg uftdi -> usb2_serial_ftdi.
1004 Older programs linked against the ports libusb 0.1 need to be
1005 redirected to the new stack's libusb20. /etc/libmap.conf can
1007 # Map old usb library to new one for usb2 stack
1008 libusb-0.1.so.8 libusb20.so.1
1011 All USB ethernet devices now attach as interfaces under the name ueN
1012 (eg. ue0). This is to provide a predictable name as vendors often
1013 change usb chipsets in a product without notice.
1016 The ichsmb(4) driver has been changed to require SMBus slave
1017 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1018 All of the other SMBus controller drivers require left-justified
1019 slave addresses, so this change makes all the drivers provide the
1023 INET6 statistics (struct ip6stat) was updated.
1024 netstat(1) needs to be recompiled.
1027 NTFS has been removed from GENERIC kernel on amd64 to match
1028 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1029 will load ntfs.ko module automatically when NTFS support is
1030 actually needed, unless ntfs.ko is not installed or security
1031 level prohibits loading kernel modules. If either is the case,
1032 "options NTFS" has to be added into kernel config.
1035 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1036 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1037 800061. User space tools that rely on the size of struct tcpcb in
1038 tcp_var.h (e.g. sockstat) need to be recompiled.
1041 ng_tty(4) module updated to match the new TTY subsystem.
1042 Due to API change, user-level applications must be updated.
1043 New API support added to mpd5 CVS and expected to be present
1044 in next mpd5.3 release.
1047 With __FreeBSD_version 800060 the makefs tool is part of
1048 the base system (it was a port).
1051 The afdata and ifnet locks have been changed from mutexes to
1052 rwlocks, network modules will need to be re-compiled.
1055 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1056 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1057 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1058 architecture and 40 bytes on 64-bit architecture. The userland
1059 applications "arp" and "ndp" have been updated accordingly.
1060 The output from "netstat -r" shows only routing entries and
1061 none of the L2 information.
1064 __FreeBSD_version 800057 marks the switchover from the
1065 binary ath hal to source code. Users must add the line:
1067 options AH_SUPPORT_AR5416
1069 to their kernel config files when specifying:
1073 The ath_hal module no longer exists; the code is now compiled
1074 together with the driver in the ath module. It is now
1075 possible to tailor chip support (i.e. reduce the set of chips
1076 and thereby the code size); consult ath_hal(4) for details.
1079 __FreeBSD_version 800054 adds memory barriers to
1080 <machine/atomic.h>, new interfaces to ifnet to facilitate
1081 multiple hardware transmit queues for cards that support
1082 them, and a lock-less ring-buffer implementation to
1083 enable drivers to more efficiently manage queueing of
1087 A new version of ZFS (version 13) has been merged to -HEAD.
1088 This version has zpool attribute "listsnapshots" off by
1089 default, which means "zfs list" does not show snapshots,
1090 and is the same as Solaris behavior.
1093 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1096 The uhci, ohci, ehci and slhci USB Host controller drivers have
1097 been put into separate modules. If you load the usb module
1098 separately through loader.conf you will need to load the
1099 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1100 controller add the following to loader.conf:
1106 The ABI used by the PMC toolset has changed. Please keep
1107 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1111 atapci kernel module now includes only generic PCI ATA
1112 driver. AHCI driver moved to ataahci kernel module.
1113 All vendor-specific code moved into separate kernel modules:
1114 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1115 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1116 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1117 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1120 The TTY subsystem of the kernel has been replaced by a new
1121 implementation, which provides better scalability and an
1122 improved driver model. Most common drivers have been migrated to
1123 the new TTY subsystem, while others have not. The following
1124 drivers have not yet been ported to the new TTY layer:
1127 cy, digi, rc, rp, sio
1133 ng_h4, ng_tty, ppp, sl, snp
1135 Adding these drivers to your kernel configuration file shall
1136 cause compilation to fail.
1139 ntpd has been upgraded to 4.2.4p5.
1142 OpenSSH has been upgraded to 5.1p1.
1144 For many years, FreeBSD's version of OpenSSH preferred DSA
1145 over RSA for host and user authentication keys. With this
1146 upgrade, we've switched to the vendor's default of RSA over
1147 DSA. This may cause upgraded clients to warn about unknown
1148 host keys even for previously known hosts. Users should
1149 follow the usual procedure for verifying host keys before
1150 accepting the RSA key.
1152 This can be circumvented by setting the "HostKeyAlgorithms"
1153 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1156 Please note that the sequence of keys offered for
1157 authentication has been changed as well. You may want to
1158 specify IdentityFile in a different order to revert this
1162 The sio(4) driver has been removed from the i386 and amd64
1163 kernel configuration files. This means uart(4) is now the
1164 default serial port driver on those platforms as well.
1166 To prevent collisions with the sio(4) driver, the uart(4) driver
1167 uses different names for its device nodes. This means the
1168 onboard serial port will now most likely be called "ttyu0"
1169 instead of "ttyd0". You may need to reconfigure applications to
1170 use the new device names.
1172 When using the serial port as a boot console, be sure to update
1173 /boot/device.hints and /etc/ttys before booting the new kernel.
1174 If you forget to do so, you can still manually specify the hints
1175 at the loader prompt:
1177 set hint.uart.0.at="isa"
1178 set hint.uart.0.port="0x3F8"
1179 set hint.uart.0.flags="0x10"
1180 set hint.uart.0.irq="4"
1184 The gpt(8) utility has been removed. Use gpart(8) to partition
1188 The version that Linuxulator emulates was changed from 2.4.2
1189 to 2.6.16. If you experience any problems with Linux binaries
1190 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1191 if it fixes the problem contact emulation mailing list.
1194 ISDN4BSD (I4B) was removed from the src tree. You may need to
1195 update a your kernel configuration and remove relevant entries.
1198 I have checked in code to support multiple routing tables.
1199 See the man pages setfib(1) and setfib(2).
1200 This is a hopefully backwards compatible version,
1201 but to make use of it you need to compile your kernel
1202 with options ROUTETABLES=2 (or more up to 16).
1205 The 802.11 wireless support was redone to enable multi-bss
1206 operation on devices that are capable. The underlying device
1207 is no longer used directly but instead wlanX devices are
1208 cloned with ifconfig. This requires changes to rc.conf files.
1209 For example, change:
1210 ifconfig_ath0="WPA DHCP"
1213 ifconfig_wlan0="WPA DHCP"
1214 see rc.conf(5) for more details. In addition, mergemaster of
1215 /etc/rc.d is highly recommended. Simultaneous update of userland
1216 and kernel wouldn't hurt either.
1218 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1219 modules were merged into the base wlan module. All references
1220 to these modules (e.g. in kernel config files) must be removed.
1223 psm(4) has gained write(2) support in native operation level.
1224 Arbitrary commands can be written to /dev/psm%d and status can
1225 be read back from it. Therefore, an application is responsible
1226 for status validation and error recovery. It is a no-op in
1227 other operation levels.
1230 Support for KSE threading has been removed from the kernel. To
1231 run legacy applications linked against KSE libmap.conf may
1232 be used. The following libmap.conf may be used to ensure
1233 compatibility with any prior release:
1235 libpthread.so.1 libthr.so.1
1236 libpthread.so.2 libthr.so.2
1237 libkse.so.3 libthr.so.3
1240 The layout of struct vmspace has changed. This affects libkvm
1241 and any executables that link against libkvm and use the
1242 kvm_getprocs() function. In particular, but not exclusively,
1243 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1244 The effects are minimal, but it's advisable to upgrade world
1248 The latest em driver no longer has support in it for the
1249 82575 adapter, this is now moved to the igb driver. The
1250 split was done to make new features that are incompatible
1251 with older hardware easier to do.
1254 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1255 likewise the kernel option is now GEOM_LINUX_LVM.
1258 The default NFS mount mode has changed from UDP to TCP for
1259 increased reliability. If you rely on (insecurely) NFS
1260 mounting across a firewall you may need to update your
1264 Belatedly note the addition of m_collapse for compacting
1268 The fts(3) structures have been changed to use adequate
1269 integer types for their members and so to be able to cope
1270 with huge file trees. The old fts(3) ABI is preserved
1271 through symbol versioning in libc, so third-party binaries
1272 using fts(3) should still work, although they will not take
1273 advantage of the extended types. At the same time, some
1274 third-party software might fail to build after this change
1275 due to unportable assumptions made in its source code about
1276 fts(3) structure members. Such software should be fixed
1277 by its vendor or, in the worst case, in the ports tree.
1278 FreeBSD_version 800015 marks this change for the unlikely
1279 case that a portable fix is impossible.
1282 To upgrade to -current after this date, you must be running
1283 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1284 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1287 The ADAPTIVE_GIANT kernel option has been retired because its
1288 functionality is the default now.
1291 The AT keyboard emulation of sunkbd(4) has been turned on
1292 by default. In order to make the special symbols of the Sun
1293 keyboards driven by sunkbd(4) work under X these now have
1294 to be configured the same way as Sun USB keyboards driven
1295 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1297 Option "XkbLayout" "us"
1298 Option "XkbRules" "xorg"
1299 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1302 It has been decided that it is desirable to provide ABI
1303 backwards compatibility to the FreeBSD 4/5/6 versions of the
1304 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1305 broken with the introduction of PCI domain support (see the
1306 20070930 entry). Unfortunately, this required the ABI of
1307 PCIOCGETCONF to be broken again in order to be able to
1308 provide backwards compatibility to the old version of that
1309 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1310 again. As for prominent ports this affects neither pciutils
1311 nor xorg-server this time, the hal port needs to be rebuilt
1315 The misnamed kthread_create() and friends have been renamed
1316 to kproc_create() etc. Many of the callers already
1317 used kproc_start()..
1318 I will return kthread_create() and friends in a while
1319 with implementations that actually create threads, not procs.
1320 Renaming corresponds with version 800002.
1329 Avoid using make -j when upgrading. While generally safe, there are
1330 sometimes problems using -j to upgrade. If your upgrade fails with
1331 -j, please try again without -j. From time to time in the past there
1332 have been problems using -j with buildworld and/or installworld. This
1333 is especially true when upgrading between "distant" versions (eg one
1334 that cross a major release boundary or several minor releases, or when
1335 several months have passed on the -current branch).
1337 Sometimes, obscure build problems are the result of environment
1338 poisoning. This can happen because the make utility reads its
1339 environment when searching for values for global variables. To run
1340 your build attempts in an "environmental clean room", prefix all make
1341 commands with 'env -i '. See the env(1) manual page for more details.
1343 When upgrading from one major version to another it is generally best
1344 to upgrade to the latest code in the currently installed branch first,
1345 then do an upgrade to the new branch. This is the best-tested upgrade
1346 path, and has the highest probability of being successful. Please try
1347 this approach before reporting problems with a major version upgrade.
1351 When upgrading the boot ZFS pool to a new version, always follow
1354 1.) recompile and reinstall the ZFS boot loader and boot block
1355 (this is part of "make buildworld" and "make installworld")
1357 2.) update the ZFS boot block on your boot drive
1359 The following example updates the ZFS boot block on the first
1360 partition (freebsd-boot) of a GPT partitioned drive ad0:
1361 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1363 Non-boot pools do not need these updates.
1367 If you are updating from a prior version of FreeBSD (even one just
1368 a few days old), you should follow this procedure. It is the most
1369 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1371 make kernel-toolchain
1372 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1373 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1375 To test a kernel once
1376 ---------------------
1377 If you just want to boot a kernel once (because you are not sure
1378 if it works, or if you want to boot a known bad kernel to provide
1379 debugging information) run
1380 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1381 nextboot -k testkernel
1383 To just build a kernel when you know that it won't mess you up
1384 --------------------------------------------------------------
1385 This assumes you are already running a CURRENT system. Replace
1386 ${arch} with the architecture of your machine (e.g. "i386",
1387 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1389 cd src/sys/${arch}/conf
1390 config KERNEL_NAME_HERE
1391 cd ../compile/KERNEL_NAME_HERE
1396 If this fails, go to the "To build a kernel" section.
1398 To rebuild everything and install it on the current system.
1399 -----------------------------------------------------------
1400 # Note: sometimes if you are running current you gotta do more than
1401 # is listed here if you are upgrading from a really old current.
1403 <make sure you have good level 0 dumps>
1405 make kernel KERNCONF=YOUR_KERNEL_HERE
1407 <reboot in single user> [3]
1415 To cross-install current onto a separate partition
1416 --------------------------------------------------
1417 # In this approach we use a separate partition to hold
1418 # current's root, 'usr', and 'var' directories. A partition
1419 # holding "/", "/usr" and "/var" should be about 2GB in
1422 <make sure you have good level 0 dumps>
1425 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1426 <maybe newfs current's root partition>
1427 <mount current's root partition on directory ${CURRENT_ROOT}>
1428 make installworld DESTDIR=${CURRENT_ROOT}
1429 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1430 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1431 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1432 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1433 <reboot into current>
1434 <do a "native" rebuild/install as described in the previous section>
1435 <maybe install compatibility libraries from ports/misc/compat*>
1439 To upgrade in-place from 8.x-stable to current
1440 ----------------------------------------------
1441 <make sure you have good level 0 dumps>
1443 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1445 <reboot in single user> [3]
1452 Make sure that you've read the UPDATING file to understand the
1453 tweaks to various things you need. At this point in the life
1454 cycle of current, things change often and you are on your own
1455 to cope. The defaults can also change, so please read ALL of
1456 the UPDATING entries.
1458 Also, if you are tracking -current, you must be subscribed to
1459 freebsd-current@freebsd.org. Make sure that before you update
1460 your sources that you have read and understood all the recent
1461 messages there. If in doubt, please track -stable which has
1462 much fewer pitfalls.
1464 [1] If you have third party modules, such as vmware, you
1465 should disable them at this point so they don't crash your
1468 [3] From the bootblocks, boot -s, and then do
1473 adjkerntz -i # if CMOS is wall time
1474 Also, when doing a major release upgrade, it is required that
1475 you boot into single user mode to do the installworld.
1477 [4] Note: This step is non-optional. Failure to do this step
1478 can result in a significant reduction in the functionality of the
1479 system. Attempting to do it by hand is not recommended and those
1480 that pursue this avenue should read this file carefully, as well
1481 as the archives of freebsd-current and freebsd-hackers mailing lists
1482 for potential gotchas. The -U option is also useful to consider.
1483 See mergemaster(8) for more information.
1485 [5] Usually this step is a noop. However, from time to time
1486 you may need to do this if you get unknown user in the following
1487 step. It never hurts to do it all the time. You may need to
1488 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1489 install) after the buildworld before this step if you last updated
1490 from current before 20020224 or from -stable before 20020408.
1492 [6] This only deletes old files and directories. Old libraries
1493 can be deleted by "make delete-old-libs", but you have to make
1494 sure that no program is using those libraries anymore.
1496 [8] In order to have a kernel that can run the 4.x binaries needed to
1497 do an installworld, you must include the COMPAT_FREEBSD4 option in
1498 your kernel. Failure to do so may leave you with a system that is
1499 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1500 required to run the 5.x binaries on more recent kernels. And so on
1501 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1503 Make sure that you merge any new devices from GENERIC since the
1504 last time you updated your kernel config file.
1506 [9] When checking out sources, you must include the -P flag to have
1507 cvs prune empty directories.
1509 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1510 "?=" instead of the "=" assignment operator, so that buildworld can
1511 override the CPUTYPE if it needs to.
1513 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1514 not on the command line, or in /etc/make.conf. buildworld will
1515 warn if it is improperly defined.
1518 This file contains a list, in reverse chronological order, of major
1519 breakages in tracking -current. It is not guaranteed to be a complete
1520 list of such breakages, and only contains entries since October 10, 2007.
1521 If you need to see UPDATING entries from before that date, you will need
1522 to fetch an UPDATING file from an older FreeBSD release.
1524 Copyright information:
1526 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1528 Redistribution, publication, translation and use, with or without
1529 modification, in full or in part, in any form or format of this
1530 document are permitted without further permission from the author.
1532 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1533 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1534 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1535 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1536 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1537 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1538 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1539 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1540 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1541 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1542 POSSIBILITY OF SUCH DAMAGE.
1544 Contact Warner Losh if you have any questions about your use of