1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 Items affecting the ports and packages system can be found in
10 /usr/ports/UPDATING. Please read that file before running portupgrade.
12 NOTE TO PEOPLE WHO THINK THAT FreeBSD 9.x IS SLOW:
13 FreeBSD 9.x has many debugging features turned on, in both the kernel
14 and userland. These features attempt to detect incorrect use of
15 system primitives, and encourage loud failure through extra sanity
16 checking and fail stop semantics. They also substantially impact
17 system performance. If you want to do performance measurement,
18 benchmarking, and optimization, you'll want to turn them off. This
19 includes various WITNESS- related kernel options, INVARIANTS, malloc
20 debugging flags in userland, and various verbose features in the
21 kernel. Many developers choose to disable these features on build
22 machines to maximize performance. (To disable malloc debugging, run
23 ln -s aj /etc/malloc.conf.)
26 The following sysctls and tunables are retired on x86 platforms:
28 machdep.hlt_logical_cpus
29 The following sysctl is retired:
30 machdep.hyperthreading_allowed
31 The sysctls were supposed to provide a way to dynamically offline and
32 online selected CPUs on x86 platforms, but the implementation has not
33 been reliable especially with SCHED_ULE scheduler.
34 machdep.hyperthreading_allowed tunable is still available to ignore
35 hyperthreading CPUs at OS level.
36 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
37 where X is an APIC ID of a CPU. Be advised, though, that disabling
38 CPUs in non-uniform fashion will result in non-uniform topology and
39 may lead to sub-optimal system performance with SCHED_ULE, which is
43 cpumask_t type is retired and cpuset_t is used in order to describe
47 Changes to ifconfig(8) for dynamic address family detection mandate
48 that you are running a kernel of 20110525 or later. Make sure to
49 follow the update procedure to boot a new kernel before installing
53 Support for sun4v architecture is officially dropped
56 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
57 into their kernel configurations along with 'device pci'.
60 The default NFS client is now the new NFS client, so fstype "newnfs"
61 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
62 Although mounts via fstype "nfs" will usually work without userland
63 changes, it is recommended that the mount(8) and mount_nfs(8)
64 commands be rebuilt from sources and that a link to mount_nfs called
65 mount_oldnfs be created. The new client is compiled into the
66 kernel with "options NFSCL" and this is needed for diskless root
67 file systems. The GENERIC kernel configs have been changed to use
68 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
69 To use the regular/old client, you can "mount -t oldnfs ...". For
70 a diskless root file system, you must also include a line like:
72 vfs.root.mountfrom="oldnfs:"
74 in the boot/loader.conf on the root fs on the NFS server to make
75 a diskless root fs use the old client.
78 The GENERIC kernels for all architectures now default to the new
79 CAM-based ATA stack. It means that all legacy ATA drivers were
80 removed and replaced by respective CAM drivers. If you are using
81 ATA device names in /etc/fstab or other places, make sure to update
82 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
83 where 'Y's are the sequential numbers starting from zero for each type
84 in order of detection, unless configured otherwise with tunables,
85 see cam(4)). There will be symbolic links created in /dev/ to map
86 old adX devices to the respective adaY. They should provide basic
87 compatibility for file systems mounting in most cases, but they do
88 not support old user-level APIs and do not have respective providers
89 in GEOM. Consider using updated management tools with new device names.
91 It is possible to load devices ahci, ata, siis and mvs as modules,
92 but option ATA_CAM should remain in kernel configuration to make ata
93 module work as CAM driver supporting legacy ATA controllers. Device ata
94 still can be used in modular fashion (atacore + ...). Modules atadisk
95 and atapi* are not used and won't affect operation in ATA_CAM mode.
96 Note that to use CAM-based ATA kernel should include CAM devices
97 scbus, pass, da (or explicitly ada), cd and optionally others. All of
98 them are parts of the cam module.
100 ataraid(4) functionality is now supported by the RAID GEOM class.
101 To use it you can load geom_raid kernel module and use graid(8) tool
102 for management. Instead of /dev/arX device names, use /dev/raid/rX.
104 No kernel config options or code have been removed, so if a problem
105 arises, please report it and optionally revert to the old ATA stack.
106 In order to do it you can remove from the kernel config:
111 , and instead add back:
112 device atadisk # ATA disk drives
113 device ataraid # ATA RAID drives
114 device atapicd # ATAPI CDROM drives
115 device atapifd # ATAPI floppy drives
116 device atapist # ATAPI tape drives
119 The default NFS server has been changed to the new server, which
120 was referred to as the experimental server. If you need to switch
121 back to the old NFS server, you must now put the "-o" option on
122 both the mountd and nfsd commands. This can be done using the
123 mountd_flags and nfs_server_flags rc.conf variables until an
124 update to the rc scripts is committed, which is coming soon.
127 The GNU Objective-C runtime library (libobjc), and other Objective-C
128 related components have been removed from the base system. If you
129 require an Objective-C library, please use one of the available ports.
132 ath(4) has been split into bus- and device- modules. if_ath contains
133 the HAL, the TX rate control and the network device code. if_ath_pci
134 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
135 contains the AHB glue. Users need to load both if_ath_pci and if_ath
136 in order to use ath on everything else.
138 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
139 need to load if_ath and if_ath_pci for ath(4) operation.
142 As part of the replacement of sysinstall, the process of building
143 release media has changed significantly. For details, please re-read
144 release(7), which has been updated to reflect the new build process.
147 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
148 is the last available version under GPLv2. It brings a number of new
149 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
150 4.1 and SSE 4.2), better support for powerpc64, a number of new
151 directives, and lots of other small improvements. See the ChangeLog
152 file in contrib/binutils for the full details.
155 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
156 compliant, and will now use half of hash for authentication.
157 This will break interoperability with all stacks (including all
158 actual FreeBSD versions) who implement
159 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
161 The only workaround with such peers is to use another HMAC
162 algorithm for IPsec ("phase 2") authentication.
165 Remove the uio_yield prototype and symbol. This function has
166 been misnamed since it was introduced and should not be
167 globally exposed with this name. The equivalent functionality
168 is now available using kern_yield(curthread->td_user_pri).
169 The function remains undocumented.
172 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
173 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
174 for scalar sysctls is defined but disabled. Code that needs
175 UQUAD to pass the type checking that must compile on older
176 systems where the define is not present can check against
177 __FreeBSD_version >= 900030.
179 The system dialog(1) has been replaced with a new version previously
180 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
181 with the previous version, but the libdialog associated with it has
182 a largely incompatible API. As such, the original version of libdialog
183 will be kept temporarily as libodialog, until its base system consumers
184 are replaced or updated. Bump __FreeBSD_version to 900030.
187 If you are trying to run make universe on a -stable system, and you get
188 the following warning:
189 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
190 unknown. config(8) likely too old."
191 or something similar to it, then you must upgrade your -stable system
192 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
193 stable/8) or build the config from the latest stable/8 branch and
194 install it on your system.
196 Prior to this date, building a current universe on 8-stable system from
197 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
198 error in the first kernel build phase. A new config on those old
199 systems will fix that problem for older versions of -current.
202 The TCP stack has been modified to allow Khelp modules to interact with
203 it via helper hook points and store per-connection data in the TCP
204 control block. Bump __FreeBSD_version to 900029. User space tools that
205 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
209 Generic IEEE 802.3 annex 31B full duplex flow control support has been
210 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
211 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
212 to take advantage of it instead of using custom implementations. This
213 means that these drivers now no longer unconditionally advertise
214 support for flow control but only do so if flow control is a selected
215 media option. This was implemented in the generic support that way in
216 order to allow flow control to be switched on and off via ifconfig(8)
217 with the PHY specific default to typically off in order to protect
218 from unwanted effects. Consequently, if you used flow control with
219 one of the above mentioned drivers you now need to explicitly enable
221 ifconfig bge0 media auto mediaopt flowcontrol
223 Along with the above mentioned changes generic support for setting
224 1000baseT master mode also has been added and brgphy(4), ciphy(4),
225 e1000phy(4) as well as ip1000phy(4) have been converted to take
226 advantage of it. This means that these drivers now no longer take the
227 link0 parameter for selecting master mode but the master media option
228 has to be used instead, for example like in the following:
229 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
231 Selection of master mode now is also available with all other PHY
232 drivers supporting 1000baseT.
235 The TCP stack has received a significant update to add support for
236 modularised congestion control and generally improve the clarity of
237 congestion control decisions. Bump __FreeBSD_version to 900025. User
238 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
239 sockstat) need to be recompiled.
242 The man(1) utility has been replaced by a new version that no longer
243 uses /etc/manpath.config. Please consult man.conf(5) for how to
244 migrate local entries to the new format.
247 The copyright strings printed by login(1) and sshd(8) at the time of a
248 new connection have been removed to follow other operating systems and
252 A workaround for a fixed ld bug has been removed in kernel code,
253 so make sure that your system ld is built from sources after
254 revision 210245 from 2010-07-19 (r211583 if building head kernel
255 on stable/8, r211584 for stable/7; both from 2010-08-21).
256 A symptom of incorrect ld version is different addresses for
257 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
260 The $ipv6_prefer variable in rc.conf(5) has been split into
261 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
263 The $ip6addrctl_policy is a variable to choose a pre-defined
264 address selection policy set by ip6addrctl(8). A value
265 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
268 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
269 flag (see an entry of 20090926) is set on an interface with no
270 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
271 security reason. If you want IPv6 link-local address on all
272 interfaces by default, set this to "YES".
274 The old ipv6_prefer="YES" is equivalent to
275 ipv6_activate_all_interfaces="YES" and
276 ip6addrctl_policy="ipv6_prefer".
279 DTrace has grown support for userland tracing. Due to this, DTrace is
280 now i386 and amd64 only.
281 dtruss(1) is now installed by default on those systems and a new
282 kernel module is needed for userland tracing: fasttrap.
283 No changes to your kernel config file are necessary to enable
284 userland tracing, but you might consider adding 'STRIP=' and
285 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
286 to have informative userland stack traces in DTrace (ustack).
289 The acpi_aiboost(4) driver has been removed in favor of the new
290 aibs(4) driver. You should update your kernel configuration file.
293 BSD grep has been imported to the base system and it is built by
294 default. It is completely BSD licensed, highly GNU-compatible, uses
295 less memory than its GNU counterpart and has a small codebase.
296 However, it is slower than its GNU counterpart, which is mostly
297 noticeable for larger searches, for smaller ones it is measurable
298 but not significant. The reason is complex, the most important factor
299 is that we lack a modern and efficient regex library and GNU
300 overcomes this by optimizing the searches internally. Future work
301 on improving the regex performance is planned, for the meantime,
302 users that need better performance, can build GNU grep instead by
303 setting the WITH_GNU_GREP knob.
306 Due to the import of powerpc64 support, all existing powerpc kernel
307 configuration files must be updated with a machine directive like this:
308 machine powerpc powerpc
310 In addition, an updated config(8) is required to build powerpc kernels
314 A new version of ZFS (version 15) has been merged to -HEAD.
315 This version uses a python library for the following subcommands:
316 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
317 For full functionality of these commands the following port must
318 be installed: sysutils/py-zfs
321 'vm_page's are now hashed by physical address to an array of mutexes.
322 Currently this is only used to serialize access to hold_count. Over
323 time the page queue mutex will be peeled away. This changes the size
324 of pmap on every architecture. And requires all callers of vm_page_hold
325 and vm_page_unhold to be updated.
328 WITH_CTF can now be specified in src.conf (not recommended, there
329 are some problems with static executables), make.conf (would also
330 affect ports which do not use GNU make and do not override the
331 compile targets) or in the kernel config (via "makeoptions
333 When WITH_CTF was specified there before this was silently ignored,
334 so make sure that WITH_CTF is not used in places which could lead
335 to unwanted behavior.
338 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
339 to allow 32-bit compatibility on non-x86 platforms. All kernel
340 configurations on amd64 and ia64 platforms using these options must
341 be modified accordingly.
344 The utmp user accounting database has been replaced with utmpx,
345 the user accounting interface standardized by POSIX.
346 Unfortunately the semantics of utmp and utmpx don't match,
347 making it practically impossible to support both interfaces.
348 The user accounting database is used by tools like finger(1),
349 last(1), talk(1), w(1) and ac(8).
351 All applications in the base system use utmpx. This means only
352 local binaries (e.g. from the ports tree) may still use these
353 utmp database files. These applications must be rebuilt to make
356 After the system has been upgraded, it is safe to remove the old
357 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
358 assuming their contents is of no importance anymore. Old wtmp
359 databases can only be used by last(1) and ac(8) after they have
360 been converted to the new format using wtmpcvt(1).
363 Introduce the kernel thread "deadlock resolver" (which can be enabled
364 via the DEADLKRES option, see NOTES for more details) and the
365 sleepq_type() function for sleepqueues.
368 The rc.firewall and rc.firewall6 were unified, and
369 rc.firewall6 and rc.d/ip6fw were removed.
370 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
371 variables are obsoleted. Instead, the following new rc
372 variables are added to rc.d/ipfw:
374 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
375 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
376 firewall_simple_onet_ipv6, firewall_trusted_ipv6
378 The meanings correspond to the relevant IPv4 variables.
384 The default terminal emulation for syscons(4) has been changed
385 from cons25 to xterm on all platforms except pc98. This means
386 that the /etc/ttys file needs to be updated to ensure correct
387 operation of applications on the console.
389 The terminal emulation style can be toggled per window by using
390 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
391 options can be used to change the compile-time default back to
394 To prevent graphical artifacts, make sure the TERM environment
395 variable is set to match the terminal emulation that is being
396 performed by syscons(4).
399 The layout of the structure ieee80211req_scan_result has changed.
400 Applications that require wireless scan results (e.g. ifconfig(8))
401 from net80211 need to be recompiled.
403 Applications such as wpa_supplicant(8) may require a full world
404 build without using NO_CLEAN in order to get synchronized with the
408 The iwn(4) driver has been updated to support the 5000 and 5150 series.
409 There's one kernel module for each firmware. Adding "device iwnfw"
410 to the kernel configuration file means including all three firmware
411 images inside the kernel. If you want to include just the one for
412 your wireless card, use the the devices iwn4965fw, iwn5000fw or
416 The rc.d/network_ipv6, IPv6 configuration script has been integrated
417 into rc.d/netif. The changes are the following:
419 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
420 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
421 Note that both variables need the "inet6" keyword at the head.
423 Do not set $ipv6_network_interfaces manually if you do not
424 understand what you are doing. It is not needed in most cases.
426 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
429 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
430 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
432 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
433 all configured interfaces have "inet6 accept_rtadv" in the
434 $ifconfig_IF_ipv6. These are for backward compatibility.
436 3. A new variable $ipv6_prefer has been added. If NO, IPv6
437 functionality of interfaces with no corresponding
438 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
439 and the default address selection policy of ip6addrctl(8)
440 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
441 Note that if you want to configure IPv6 functionality on the
442 disabled interfaces after boot, first you need to clear the flag by
443 using ifconfig(8) like:
445 ifconfig em0 inet6 -ifdisabled
447 If YES, the default address selection policy is set as
450 The default value of $ipv6_prefer is NO.
452 4. If your system need to receive Router Advertisement messages,
453 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
454 scripts automatically invoke rtsol(8) when the interface becomes
455 UP. The Router Advertisement messages are used for SLAAC
456 (State-Less Address AutoConfiguration).
459 802.11s D3.03 support was committed. This is incompatible with the
460 previous code, which was based on D3.0.
463 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
464 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
465 control whether accepting Router Advertisement messages or not.
466 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
467 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
468 The ifconfig(8) utility now supports these flags.
471 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
472 mount(8) and -a option for df(1) to see them.
475 The old tunable hw.bus.devctl_disable has been superseded by
476 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
477 replaced by hw.bus.devctl_queue=0. The default for this new tunable
481 Remove the option STOP_NMI. The default action is now to use NMI only
482 for KDB via the newly introduced function stop_cpus_hard() and
483 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
486 The stable/8 branch created in subversion. This corresponds to the
487 RELENG_8 branch in CVS.
490 Bump the shared library version numbers for all libraries that do not
491 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
492 __FreeBSD_version to 800105.
495 Due to changes in the implementation of virtual network stack support,
496 all network-related kernel modules must be recompiled. As this change
497 breaks the ABI, bump __FreeBSD_version to 800104.
500 The TOE interface to the TCP syncache has been modified to remove
501 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
502 The cxgb driver is the only TOE consumer affected by this change, and
503 needs to be recompiled along with the kernel. As this change breaks
504 the ABI, bump __FreeBSD_version to 800103.
507 Padding has been added to struct tcpcb, sackhint and tcpstat in
508 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
509 maintaining the ABI. However, this change breaks the ABI, so bump
510 __FreeBSD_version to 800102. User space tools that rely on the size of
511 any of these structs (e.g. sockstat) need to be recompiled.
514 The NFS_LEGACYRPC option has been removed along with the old kernel
515 RPC implementation that this option selected. Kernel configurations
516 may need to be adjusted.
519 The network interface device nodes at /dev/net/<interface> have been
520 removed. All ioctl operations can be performed the normal way using
521 routing sockets. The kqueue functionality can generally be replaced
522 with routing sockets.
525 The documentation from the FreeBSD Documentation Project (Handbook,
526 FAQ, etc.) is now installed via packages by sysinstall(8) and under
527 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
530 The ABI of various structures related to the SYSV IPC API have been
531 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
532 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
536 Layout of struct vnet has changed as routing related variables were
537 moved to their own Vimage module. Modules need to be recompiled. Bump
538 __FreeBSD_version to 800099.
541 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
542 respectively. As long as no more than 16 groups per process are used,
543 no changes should be visible. When more than 16 groups are used, old
544 binaries may fail if they call getgroups() or getgrouplist() with
545 statically sized storage. Recompiling will work around this, but
546 applications should be modified to use dynamically allocated storage
547 for group arrays as POSIX.1-2008 does not cap an implementation's
548 number of supported groups at NGROUPS_MAX+1 as previous versions did.
550 NFS and portalfs mounts may also be affected as the list of groups is
551 truncated to 16. Users of NFS who use more than 16 groups, should
552 take care that negative group permissions are not used on the exported
553 file systems as they will not be reliable unless a GSSAPI based
554 authentication method is used.
557 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
558 option compiles in the support for adaptive spinning for lockmgrs
559 which want to enable it. The lockinit() function now accepts the flag
560 LK_ADAPTIVE in order to make the lock object subject to adaptive
561 spinning when both held in write and read mode.
564 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
565 changed. User applications that use this ioctl need to be rebuilt.
568 The layout of struct thread has changed. Kernel and modules need to
572 The layout of structs ifnet, domain, protosw and vnet_net has changed.
573 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
576 window(1) has been removed from the base system. It can now be
577 installed from ports. The port is called misc/window.
580 The way we are storing and accessing `routing table' entries has
581 changed. Programs reading the FIB, like netstat, need to be
585 A new netisr implementation has been added for FreeBSD 8. Network
586 file system modules, such as igmp, ipdivert, and others, should be
588 Bump __FreeBSD_version to 800096.
591 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
595 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
596 Bump __FreeBSD_version to 800094.
599 Add mnt_xflag field to 'struct mount'. File system modules need to be
601 Bump __FreeBSD_version to 800093.
604 The compiling option ADAPTIVE_SX has been retired while it has been
605 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
606 The KPI for sx_init_flags() changes as accepting flags:
607 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
608 been introduced in order to handle the reversed logic.
609 Bump __FreeBSD_version to 800092.
612 Add support for hierarchical jails. Remove global securelevel.
613 Bump __FreeBSD_version to 800091.
616 The layout of struct vnet_net has changed, therefore modules
618 Bump __FreeBSD_version to 800090.
621 The newly imported zic(8) produces a new format in the output. Please
622 run tzsetup(8) to install the newly created data to /etc/localtime.
625 The sysctl tree for the usb stack has renamed from hw.usb2.* to
626 hw.usb.* and is now consistent again with previous releases.
629 802.11 monitor mode support was revised and driver api's were changed.
630 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
631 of DLT_IEEE802_11. No user-visible data structures were changed but
632 applications that use DLT_IEEE802_11 may require changes.
633 Bump __FreeBSD_version to 800088.
636 The layout of the following structs has changed: sysctl_oid,
637 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
638 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
639 panics may be experienced. World rebuild is required for
640 correctly checking networking state from userland.
641 Bump __FreeBSD_version to 800085.
644 MLDv2 and Source-Specific Multicast (SSM) have been merged
645 to the IPv6 stack. VIMAGE hooks are in but not yet used.
646 The implementation of SSM within FreeBSD's IPv6 stack closely
647 follows the IPv4 implementation.
649 For kernel developers:
651 * The most important changes are that the ip6_output() and
652 ip6_input() paths no longer take the IN6_MULTI_LOCK,
653 and this lock has been downgraded to a non-recursive mutex.
655 * As with the changes to the IPv4 stack to support SSM, filtering
656 of inbound multicast traffic must now be performed by transport
657 protocols within the IPv6 stack. This does not apply to TCP and
658 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
660 * The KPIs used by IPv6 multicast are similar to those used by
661 the IPv4 stack, with the following differences:
662 * im6o_mc_filter() is analogous to imo_multicast_filter().
663 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
664 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
665 * IN6_LOOKUP_MULTI() has been deprecated and removed.
666 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
667 for MLDv1 have an additional 'timer' argument which is used to
668 jitter the initial membership report for the solicited-node
669 multicast membership on-link.
670 * This is not strictly needed for MLDv2, which already jitters
671 its report transmissions. However, the 'timer' argument is
672 preserved in case MLDv1 is active on the interface.
674 * The KAME linked-list based IPv6 membership implementation has
675 been refactored to use a vector similar to that used by the IPv4
677 Code which maintains a list of its own multicast memberships
678 internally, e.g. carp, has been updated to reflect the new
681 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
682 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
683 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
684 implementation constraint which needs to be addressed in HEAD.
686 For application developers:
688 * The changes are broadly similar to those made for the IPv4
691 * The use of IPv4 and IPv6 multicast socket options on the same
692 socket, using mapped addresses, HAS NOT been tested or supported.
694 * There are a number of issues with the implementation of various
695 IPv6 multicast APIs which need to be resolved in the API surface
696 before the implementation is fully compatible with KAME userland
697 use, and these are mostly to do with interface index treatment.
699 * The literature available discusses the use of either the delta / ASM
700 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
701 using setsourcefilter(3)/getsourcefilter(3). For more information
702 please refer to RFC 3768, 'Socket Interface Extensions for
703 Multicast Source Filters'.
705 * Applications which use the published RFC 3678 APIs should be fine.
707 For systems administrators:
709 * The mtest(8) utility has been refactored to support IPv6, in
710 addition to IPv4. Interface addresses are no longer accepted
711 as arguments, their names must be used instead. The utility
712 will map the interface name to its first IPv4 address as
713 returned by getifaddrs(3).
715 * The ifmcstat(8) utility has also been updated to print the MLDv2
716 endpoint state and source filter lists via sysctl(3).
718 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
719 loopback of IPv6 multicast datagrams by default; it defaults to 1
720 to preserve the existing behaviour. Disabling multicast loopback is
721 recommended for optimal system performance.
723 * The IPv6 MROUTING code has been changed to examine this sysctl
724 instead of attempting to perform a group lookup before looping
725 back forwarded datagrams.
727 Bump __FreeBSD_version to 800084.
730 Implement low-level Bluetooth HCI API.
731 Bump __FreeBSD_version to 800083.
734 The layout of struct malloc_type, used by modules to register new
735 memory allocation types, has changed. Most modules will need to
736 be rebuilt or panics may be experienced.
737 Bump __FreeBSD_version to 800081.
740 Anticipate overflowing inp_flags - add inp_flags2.
741 This changes most offsets in inpcb, so checking v4 connection
742 state will require a world rebuild.
743 Bump __FreeBSD_version to 800080.
746 Add an llentry to struct route and struct route_in6. Modules
747 embedding a struct route will need to be recompiled.
748 Bump __FreeBSD_version to 800079.
751 The size of rt_metrics_lite and by extension rtentry has changed.
752 Networking administration apps will need to be recompiled.
753 The route command now supports show as an alias for get, weighting
754 of routes, sticky and nostick flags to alter the behavior of stateful
756 Bump __FreeBSD_version to 800078.
759 Do not use Giant for kbdmux(4) locking. This is wrong and
760 apparently causing more problems than it solves. This will
761 re-open the issue where interrupt handlers may race with
762 kbdmux(4) in polling mode. Typical symptoms include (but
763 not limited to) duplicated and/or missing characters when
764 low level console functions (such as gets) are used while
765 interrupts are enabled (for example geli password prompt,
766 mountroot prompt etc.). Disabling kbdmux(4) may help.
769 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
770 kernel modules referencing any of the above need to be recompiled.
771 Bump __FreeBSD_version to 800075.
774 GEOM_PART has become the default partition slicer for storage devices,
775 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
776 introduces some changes:
778 MSDOS/EBR: the devices created from MSDOS extended partition entries
779 (EBR) can be named differently than with GEOM_MBR and are now symlinks
780 to devices with offset-based names. fstabs may need to be modified.
782 BSD: the "geometry does not match label" warning is harmless in most
783 cases but it points to problems in file system misalignment with
784 disk geometry. The "c" partition is now implicit, covers the whole
785 top-level drive and cannot be (mis)used by users.
787 General: Kernel dumps are now not allowed to be written to devices
788 whose partition types indicate they are meant to be used for file
789 systems (or, in case of MSDOS partitions, as something else than
792 Most of these changes date approximately from 200812.
795 The uscanner(4) driver has been removed from the kernel. This follows
796 Linux removing theirs in 2.6 and making libusb the default interface
800 The multicast forwarding code has been cleaned up. netstat(1)
801 only relies on KVM now for printing bandwidth upcall meters.
802 The IPv4 and IPv6 modules are split into ip_mroute_mod and
803 ip6_mroute_mod respectively. The config(5) options for statically
804 compiling this code remain the same, i.e. 'options MROUTING'.
807 Support for the IFF_NEEDSGIANT network interface flag has been
808 removed, which means that non-MPSAFE network device drivers are no
809 longer supported. In particular, if_ar, if_sr, and network device
810 drivers from the old (legacy) USB stack can no longer be built or
814 POSIX.1 Native Language Support (NLS) has been enabled in libc and
815 a bunch of new language catalog files have also been added.
816 This means that some common libc messages are now localized and
817 they depend on the LC_MESSAGES environmental variable.
820 The k8temp(4) driver has been renamed to amdtemp(4) since
821 support for Family 10 and Family 11 CPU families was added.
824 IGMPv3 and Source-Specific Multicast (SSM) have been merged
825 to the IPv4 stack. VIMAGE hooks are in but not yet used.
827 For kernel developers, the most important changes are that the
828 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
829 and this lock has been downgraded to a non-recursive mutex.
831 Transport protocols (UDP, Raw IP) are now responsible for filtering
832 inbound multicast traffic according to group membership and source
833 filters. The imo_multicast_filter() KPI exists for this purpose.
834 Transports which do not use multicast (SCTP, TCP) already reject
835 multicast by default. Forwarding and receive performance may improve
836 as a mutex acquisition is no longer needed in the ip_input()
837 low-level input path. in_addmulti() and in_delmulti() are shimmed
838 to new KPIs which exist to support SSM in-kernel.
840 For application developers, it is recommended that loopback of
841 multicast datagrams be disabled for best performance, as this
842 will still cause the lock to be taken for each looped-back
843 datagram transmission. The net.inet.ip.mcast.loop sysctl may
844 be tuned to 0 to disable loopback by default; it defaults to 1
845 to preserve the existing behaviour.
847 For systems administrators, to obtain best performance with
848 multicast reception and multiple groups, it is always recommended
849 that a card with a suitably precise hash filter is used. Hash
850 collisions will still result in the lock being taken within the
851 transport protocol input path to check group membership.
853 If deploying FreeBSD in an environment with IGMP snooping switches,
854 it is recommended that the net.inet.igmp.sendlocal sysctl remain
855 enabled; this forces 224.0.0.0/24 group membership to be announced
858 The size of 'struct igmpstat' has changed; netstat needs to be
859 recompiled to reflect this.
860 Bump __FreeBSD_version to 800070.
863 libusb20.so.1 is now installed as libusb.so.1 and the ports system
864 updated to use it. This requires a buildworld/installworld in order to
865 update the library and dependencies (usbconfig, etc). Its advisable to
866 rebuild all ports which uses libusb. More specific directions are given
867 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
868 libusb are no longer required and can be removed.
871 A workaround is committed to allow the creation of System V shared
872 memory segment of size > 2 GB on the 64-bit architectures.
873 Due to a limitation of the existing ABI, the shm_segsz member
874 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
875 wrong for large segments. Note that limits must be explicitly
876 raised to allow such segments to be created.
879 The layout of struct ifnet has changed, requiring a rebuild of all
880 network device driver modules.
883 The /dev handling for the new USB stack has changed, a
884 buildworld/installworld is required for libusb20.
887 The new USB2 stack has now been permanently moved in and all kernel and
888 module names reverted to their previous values (eg, usb, ehci, ohci,
889 ums, ...). The old usb stack can be compiled in by prefixing the name
890 with the letter 'o', the old usb modules have been removed.
891 Updating entry 20090216 for xorg and 20090215 for libmap may still
895 The rc.conf(5) option if_up_delay has been renamed to
896 defaultroute_delay to better reflect its purpose. If you have
897 customized this setting in /etc/rc.conf you need to update it to
901 xorg 7.4 wants to configure its input devices via hald which does not
902 yet work with USB2. If the keyboard/mouse does not work in xorg then
904 Option "AllowEmptyInput" "off"
905 to your ServerLayout section. This will cause X to use the configured
906 kbd and mouse sections from your xorg.conf.
909 The GENERIC kernels for all architectures now default to the new USB2
910 stack. No kernel config options or code have been removed so if a
911 problem arises please report it and optionally revert to the old USB
912 stack. If you are loading USB kernel modules or have a custom kernel
913 that includes GENERIC then ensure that usb names are also changed over,
914 eg uftdi -> usb2_serial_ftdi.
916 Older programs linked against the ports libusb 0.1 need to be
917 redirected to the new stack's libusb20. /etc/libmap.conf can
919 # Map old usb library to new one for usb2 stack
920 libusb-0.1.so.8 libusb20.so.1
923 All USB ethernet devices now attach as interfaces under the name ueN
924 (eg. ue0). This is to provide a predictable name as vendors often
925 change usb chipsets in a product without notice.
928 The ichsmb(4) driver has been changed to require SMBus slave
929 addresses be left-justified (xxxxxxx0b) rather than right-justified.
930 All of the other SMBus controller drivers require left-justified
931 slave addresses, so this change makes all the drivers provide the
935 INET6 statistics (struct ip6stat) was updated.
936 netstat(1) needs to be recompiled.
939 NTFS has been removed from GENERIC kernel on amd64 to match
940 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
941 will load ntfs.ko module automatically when NTFS support is
942 actually needed, unless ntfs.ko is not installed or security
943 level prohibits loading kernel modules. If either is the case,
944 "options NTFS" has to be added into kernel config.
947 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
948 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
949 800061. User space tools that rely on the size of struct tcpcb in
950 tcp_var.h (e.g. sockstat) need to be recompiled.
953 ng_tty(4) module updated to match the new TTY subsystem.
954 Due to API change, user-level applications must be updated.
955 New API support added to mpd5 CVS and expected to be present
956 in next mpd5.3 release.
959 With __FreeBSD_version 800060 the makefs tool is part of
960 the base system (it was a port).
963 The afdata and ifnet locks have been changed from mutexes to
964 rwlocks, network modules will need to be re-compiled.
967 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
968 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
969 The new code reduced struct rtentry{} by 16 bytes on 32-bit
970 architecture and 40 bytes on 64-bit architecture. The userland
971 applications "arp" and "ndp" have been updated accordingly.
972 The output from "netstat -r" shows only routing entries and
973 none of the L2 information.
976 __FreeBSD_version 800057 marks the switchover from the
977 binary ath hal to source code. Users must add the line:
979 options AH_SUPPORT_AR5416
981 to their kernel config files when specifying:
985 The ath_hal module no longer exists; the code is now compiled
986 together with the driver in the ath module. It is now
987 possible to tailor chip support (i.e. reduce the set of chips
988 and thereby the code size); consult ath_hal(4) for details.
991 __FreeBSD_version 800054 adds memory barriers to
992 <machine/atomic.h>, new interfaces to ifnet to facilitate
993 multiple hardware transmit queues for cards that support
994 them, and a lock-less ring-buffer implementation to
995 enable drivers to more efficiently manage queueing of
999 A new version of ZFS (version 13) has been merged to -HEAD.
1000 This version has zpool attribute "listsnapshots" off by
1001 default, which means "zfs list" does not show snapshots,
1002 and is the same as Solaris behavior.
1005 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1008 The uhci, ohci, ehci and slhci USB Host controller drivers have
1009 been put into separate modules. If you load the usb module
1010 separately through loader.conf you will need to load the
1011 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1012 controller add the following to loader.conf:
1018 The ABI used by the PMC toolset has changed. Please keep
1019 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1023 atapci kernel module now includes only generic PCI ATA
1024 driver. AHCI driver moved to ataahci kernel module.
1025 All vendor-specific code moved into separate kernel modules:
1026 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1027 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1028 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1029 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1032 The TTY subsystem of the kernel has been replaced by a new
1033 implementation, which provides better scalability and an
1034 improved driver model. Most common drivers have been migrated to
1035 the new TTY subsystem, while others have not. The following
1036 drivers have not yet been ported to the new TTY layer:
1039 cy, digi, rc, rp, sio
1045 ng_h4, ng_tty, ppp, sl, snp
1047 Adding these drivers to your kernel configuration file shall
1048 cause compilation to fail.
1051 ntpd has been upgraded to 4.2.4p5.
1054 OpenSSH has been upgraded to 5.1p1.
1056 For many years, FreeBSD's version of OpenSSH preferred DSA
1057 over RSA for host and user authentication keys. With this
1058 upgrade, we've switched to the vendor's default of RSA over
1059 DSA. This may cause upgraded clients to warn about unknown
1060 host keys even for previously known hosts. Users should
1061 follow the usual procedure for verifying host keys before
1062 accepting the RSA key.
1064 This can be circumvented by setting the "HostKeyAlgorithms"
1065 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1068 Please note that the sequence of keys offered for
1069 authentication has been changed as well. You may want to
1070 specify IdentityFile in a different order to revert this
1074 The sio(4) driver has been removed from the i386 and amd64
1075 kernel configuration files. This means uart(4) is now the
1076 default serial port driver on those platforms as well.
1078 To prevent collisions with the sio(4) driver, the uart(4) driver
1079 uses different names for its device nodes. This means the
1080 onboard serial port will now most likely be called "ttyu0"
1081 instead of "ttyd0". You may need to reconfigure applications to
1082 use the new device names.
1084 When using the serial port as a boot console, be sure to update
1085 /boot/device.hints and /etc/ttys before booting the new kernel.
1086 If you forget to do so, you can still manually specify the hints
1087 at the loader prompt:
1089 set hint.uart.0.at="isa"
1090 set hint.uart.0.port="0x3F8"
1091 set hint.uart.0.flags="0x10"
1092 set hint.uart.0.irq="4"
1096 The gpt(8) utility has been removed. Use gpart(8) to partition
1100 The version that Linuxulator emulates was changed from 2.4.2
1101 to 2.6.16. If you experience any problems with Linux binaries
1102 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1103 if it fixes the problem contact emulation mailing list.
1106 ISDN4BSD (I4B) was removed from the src tree. You may need to
1107 update a your kernel configuration and remove relevant entries.
1110 I have checked in code to support multiple routing tables.
1111 See the man pages setfib(1) and setfib(2).
1112 This is a hopefully backwards compatible version,
1113 but to make use of it you need to compile your kernel
1114 with options ROUTETABLES=2 (or more up to 16).
1117 The 802.11 wireless support was redone to enable multi-bss
1118 operation on devices that are capable. The underlying device
1119 is no longer used directly but instead wlanX devices are
1120 cloned with ifconfig. This requires changes to rc.conf files.
1121 For example, change:
1122 ifconfig_ath0="WPA DHCP"
1125 ifconfig_wlan0="WPA DHCP"
1126 see rc.conf(5) for more details. In addition, mergemaster of
1127 /etc/rc.d is highly recommended. Simultaneous update of userland
1128 and kernel wouldn't hurt either.
1130 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1131 modules were merged into the base wlan module. All references
1132 to these modules (e.g. in kernel config files) must be removed.
1135 psm(4) has gained write(2) support in native operation level.
1136 Arbitrary commands can be written to /dev/psm%d and status can
1137 be read back from it. Therefore, an application is responsible
1138 for status validation and error recovery. It is a no-op in
1139 other operation levels.
1142 Support for KSE threading has been removed from the kernel. To
1143 run legacy applications linked against KSE libmap.conf may
1144 be used. The following libmap.conf may be used to ensure
1145 compatibility with any prior release:
1147 libpthread.so.1 libthr.so.1
1148 libpthread.so.2 libthr.so.2
1149 libkse.so.3 libthr.so.3
1152 The layout of struct vmspace has changed. This affects libkvm
1153 and any executables that link against libkvm and use the
1154 kvm_getprocs() function. In particular, but not exclusively,
1155 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1156 The effects are minimal, but it's advisable to upgrade world
1160 The latest em driver no longer has support in it for the
1161 82575 adapter, this is now moved to the igb driver. The
1162 split was done to make new features that are incompatible
1163 with older hardware easier to do.
1166 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1167 likewise the kernel option is now GEOM_LINUX_LVM.
1170 The default NFS mount mode has changed from UDP to TCP for
1171 increased reliability. If you rely on (insecurely) NFS
1172 mounting across a firewall you may need to update your
1176 Belatedly note the addition of m_collapse for compacting
1180 The fts(3) structures have been changed to use adequate
1181 integer types for their members and so to be able to cope
1182 with huge file trees. The old fts(3) ABI is preserved
1183 through symbol versioning in libc, so third-party binaries
1184 using fts(3) should still work, although they will not take
1185 advantage of the extended types. At the same time, some
1186 third-party software might fail to build after this change
1187 due to unportable assumptions made in its source code about
1188 fts(3) structure members. Such software should be fixed
1189 by its vendor or, in the worst case, in the ports tree.
1190 FreeBSD_version 800015 marks this change for the unlikely
1191 case that a portable fix is impossible.
1194 To upgrade to -current after this date, you must be running
1195 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1196 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1199 The ADAPTIVE_GIANT kernel option has been retired because its
1200 functionality is the default now.
1203 The AT keyboard emulation of sunkbd(4) has been turned on
1204 by default. In order to make the special symbols of the Sun
1205 keyboards driven by sunkbd(4) work under X these now have
1206 to be configured the same way as Sun USB keyboards driven
1207 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1209 Option "XkbLayout" "us"
1210 Option "XkbRules" "xorg"
1211 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1214 It has been decided that it is desirable to provide ABI
1215 backwards compatibility to the FreeBSD 4/5/6 versions of the
1216 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1217 broken with the introduction of PCI domain support (see the
1218 20070930 entry). Unfortunately, this required the ABI of
1219 PCIOCGETCONF to be broken again in order to be able to
1220 provide backwards compatibility to the old version of that
1221 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1222 again. As for prominent ports this affects neither pciutils
1223 nor xorg-server this time, the hal port needs to be rebuilt
1227 The misnamed kthread_create() and friends have been renamed
1228 to kproc_create() etc. Many of the callers already
1229 used kproc_start()..
1230 I will return kthread_create() and friends in a while
1231 with implementations that actually create threads, not procs.
1232 Renaming corresponds with version 800002.
1241 Avoid using make -j when upgrading. While generally safe, there are
1242 sometimes problems using -j to upgrade. If your upgrade fails with
1243 -j, please try again without -j. From time to time in the past there
1244 have been problems using -j with buildworld and/or installworld. This
1245 is especially true when upgrading between "distant" versions (eg one
1246 that cross a major release boundary or several minor releases, or when
1247 several months have passed on the -current branch).
1249 Sometimes, obscure build problems are the result of environment
1250 poisoning. This can happen because the make utility reads its
1251 environment when searching for values for global variables. To run
1252 your build attempts in an "environmental clean room", prefix all make
1253 commands with 'env -i '. See the env(1) manual page for more details.
1255 When upgrading from one major version to another it is generally best
1256 to upgrade to the latest code in the currently installed branch first,
1257 then do an upgrade to the new branch. This is the best-tested upgrade
1258 path, and has the highest probability of being successful. Please try
1259 this approach before reporting problems with a major version upgrade.
1263 When upgrading the boot ZFS pool to a new version, always follow
1266 1.) recompile and reinstall the ZFS boot loader and boot block
1267 (this is part of "make buildworld" and "make installworld")
1269 2.) update the ZFS boot block on your boot drive
1271 The following example updates the ZFS boot block on the first
1272 partition (freebsd-boot) of a GPT partitioned drive ad0:
1273 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1275 Non-boot pools do not need these updates.
1279 If you are updating from a prior version of FreeBSD (even one just
1280 a few days old), you should follow this procedure. It is the most
1281 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1283 make kernel-toolchain
1284 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1285 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1287 To test a kernel once
1288 ---------------------
1289 If you just want to boot a kernel once (because you are not sure
1290 if it works, or if you want to boot a known bad kernel to provide
1291 debugging information) run
1292 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1293 nextboot -k testkernel
1295 To just build a kernel when you know that it won't mess you up
1296 --------------------------------------------------------------
1297 This assumes you are already running a CURRENT system. Replace
1298 ${arch} with the architecture of your machine (e.g. "i386",
1299 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1301 cd src/sys/${arch}/conf
1302 config KERNEL_NAME_HERE
1303 cd ../compile/KERNEL_NAME_HERE
1308 If this fails, go to the "To build a kernel" section.
1310 To rebuild everything and install it on the current system.
1311 -----------------------------------------------------------
1312 # Note: sometimes if you are running current you gotta do more than
1313 # is listed here if you are upgrading from a really old current.
1315 <make sure you have good level 0 dumps>
1317 make kernel KERNCONF=YOUR_KERNEL_HERE
1319 <reboot in single user> [3]
1327 To cross-install current onto a separate partition
1328 --------------------------------------------------
1329 # In this approach we use a separate partition to hold
1330 # current's root, 'usr', and 'var' directories. A partition
1331 # holding "/", "/usr" and "/var" should be about 2GB in
1334 <make sure you have good level 0 dumps>
1337 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1338 <maybe newfs current's root partition>
1339 <mount current's root partition on directory ${CURRENT_ROOT}>
1340 make installworld DESTDIR=${CURRENT_ROOT}
1341 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1342 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1343 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1344 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1345 <reboot into current>
1346 <do a "native" rebuild/install as described in the previous section>
1347 <maybe install compatibility libraries from ports/misc/compat*>
1351 To upgrade in-place from 8.x-stable to current
1352 ----------------------------------------------
1353 <make sure you have good level 0 dumps>
1355 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1357 <reboot in single user> [3]
1364 Make sure that you've read the UPDATING file to understand the
1365 tweaks to various things you need. At this point in the life
1366 cycle of current, things change often and you are on your own
1367 to cope. The defaults can also change, so please read ALL of
1368 the UPDATING entries.
1370 Also, if you are tracking -current, you must be subscribed to
1371 freebsd-current@freebsd.org. Make sure that before you update
1372 your sources that you have read and understood all the recent
1373 messages there. If in doubt, please track -stable which has
1374 much fewer pitfalls.
1376 [1] If you have third party modules, such as vmware, you
1377 should disable them at this point so they don't crash your
1380 [3] From the bootblocks, boot -s, and then do
1385 adjkerntz -i # if CMOS is wall time
1386 Also, when doing a major release upgrade, it is required that
1387 you boot into single user mode to do the installworld.
1389 [4] Note: This step is non-optional. Failure to do this step
1390 can result in a significant reduction in the functionality of the
1391 system. Attempting to do it by hand is not recommended and those
1392 that pursue this avenue should read this file carefully, as well
1393 as the archives of freebsd-current and freebsd-hackers mailing lists
1394 for potential gotchas. The -U option is also useful to consider.
1395 See mergemaster(8) for more information.
1397 [5] Usually this step is a noop. However, from time to time
1398 you may need to do this if you get unknown user in the following
1399 step. It never hurts to do it all the time. You may need to
1400 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1401 install) after the buildworld before this step if you last updated
1402 from current before 20020224 or from -stable before 20020408.
1404 [6] This only deletes old files and directories. Old libraries
1405 can be deleted by "make delete-old-libs", but you have to make
1406 sure that no program is using those libraries anymore.
1408 [8] In order to have a kernel that can run the 4.x binaries needed to
1409 do an installworld, you must include the COMPAT_FREEBSD4 option in
1410 your kernel. Failure to do so may leave you with a system that is
1411 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1412 required to run the 5.x binaries on more recent kernels. And so on
1413 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1415 Make sure that you merge any new devices from GENERIC since the
1416 last time you updated your kernel config file.
1418 [9] When checking out sources, you must include the -P flag to have
1419 cvs prune empty directories.
1421 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1422 "?=" instead of the "=" assignment operator, so that buildworld can
1423 override the CPUTYPE if it needs to.
1425 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1426 not on the command line, or in /etc/make.conf. buildworld will
1427 warn if it is improperly defined.
1430 This file contains a list, in reverse chronological order, of major
1431 breakages in tracking -current. Not all things will be listed here,
1432 and it only starts on October 16, 2004. Updating files can found in
1433 previous releases if your system is older than this.
1435 Copyright information:
1437 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1439 Redistribution, publication, translation and use, with or without
1440 modification, in full or in part, in any form or format of this
1441 document are permitted without further permission from the author.
1443 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1444 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1445 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1446 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1447 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1448 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1449 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1450 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1451 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1452 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1453 POSSIBILITY OF SUCH DAMAGE.
1455 Contact Warner Losh if you have any questions about your use of