diff options
Diffstat (limited to 'contrib/amd/NEWS')
| -rw-r--r-- | contrib/amd/NEWS | 700 |
1 files changed, 608 insertions, 92 deletions
diff --git a/contrib/amd/NEWS b/contrib/amd/NEWS index 60d58a9..5cde984 100644 --- a/contrib/amd/NEWS +++ b/contrib/amd/NEWS @@ -1,147 +1,663 @@ -*** Notes specific to am-utils version 6.0.10-pre: +*** Notes specific to am-utils version 6.1.5 + +New amd.conf global parameter: nfs_allow_any_interface. By default it is +set to 'no' which means that Amd accepts local NFS packets only from +127.0.0.1. If set to 'yes' then Amd will accept local NFS packets from any +local interface; this is useful on hosts that may have multiple interfaces +where the system is forced to send all outgoing packets (even those bound to +the same host) via an address other than 127.0.0.1. + +Add support for specifying the host to match in the mount selectors netgrp +and netgrpd. Now one can use either netgrp(<group-name>) or +netgrp(<group-name>,<host-name>). + +- Bugs fixed: + * handle old-style filehandles correctly (for mount points longer + than 28 chars) + * don't turn off attribute cache for regular NFS mounts (improves + performance) + * detect G/DBM support via gdbm_compat library (Debian) + * detect NDBM support in libc (FreeBSD 6) - minor new ports: - ia64-unknown-linux-rh2.1AS (Red Hat Itanium Advanced Server) - i386-unknown-freebsd5.0 (5.0-RELEASE) - sparc64-unknown-linux-suse7.3 + i386-unknown-freebsd6.1 (RELEASE) + i386-unknown-openbsd3.9 + powerpc-apple-darwin8.6.0 + +*** Notes specific to am-utils version 6.1.4 + +Support new mount options for type:=pcfs mounts: longname, nowin95, +shortname, user=N, group=N, mask=N, and dirmask=N. + +Two new amd.conf [global] parameters: nfs_retry_interval_toplvl and +nfs_retransmit_counter_toplvl. They are similar to nfs_retry_interval and +nfs_retransmit_counter, and allow you to set the "timeo" and "retrans" NFS +mount parameters, respectively, but ONLY for Amd's top-level mounts (which +are NFSv2/UDP currently). This is useful because on some systems you may +wish to set these parameters differently than the OS default, so as to +better tune Amd's responsiveness under heavy scheduler loads. + +- minor new ports: + i386-pc-linux-fc5 (Fedora Core 5) + i386-pc-linux-suse10.1 (beta 8) + i386-unknown-freebsd6.0 (RELEASE) + i386-unknown-netbsdelf2.1 + i386-unknown-netbsdelf3.0 (RELEASE) + i386-unknown-openbsd3.8 + powerpc-apple-darwin8.5.0 + +- Bugs fixed: + * one serious memory leak in amfs_generic (caught by Coverity) + * assorted potential (but rare) NULL pointer dereferences (Coverity) + * correctly print nfs_args->addr info (sin_family/port/addr) + * pawd should resolve path repeatedly until no more to do + * use-after-free bug in amfs_lookup_mntfs (Coverity) + +*** Notes specific to am-utils version 6.1.3 + +- Bugs fixed: + * amq should de-register properly on exit + * convert all sprintf to safer xsnprintf + * convert all strcat to safer xstrlcat + * convert all strcpy to safer xstrlcpy + * fix three buffer overruns in expand_op (amd/opts.c) + * pawd was trying UDP only, now try TCP if UDP failed + +Moved pawd's path-matching functionality into Amd, where it can be done a +lot more efficiently (we no longer need to construct and send the whole +mounted tree, only to match small parts of it). This will lessen the CPU +and network load on systems that use pawd heavily, and also minimize the +chance that we exceed default or hard-coded UDP/TCP RPC packet sizes. + +*** Notes specific to am-utils version 6.1.2.1 + +- Bugs fixed: + * properly turn off the attrcache in freebsd and openbsd + * can turn off attrcache on netbsd, but need kernel patch, see + README.attrcache + * pawd goes into an infinite loop on type:=auto + * consistent search for file system mnttab/mount names + +*** Notes specific to am-utils version 6.1.2 + +MAJOR BUG FIXES: Synchronize Amd's view of its file systems with the +kernel's NFS client-side DNLC/dcache. Amd changes its view when it reloads +maps (automatically or via "amq -f") because new map entries could be added, +old one removed, or existing ones changed. Amd also changes its view when a +simple entry has expired and was flushed, or was forced out via "amq -u". +Amd was not updating the mtime of its parent directory (often the amd +automount point): this resulted in the kernel re-using cached entries, which +are now possibly stale. Many users had seen this problem in the form of +occasional ESTALE errors, or dangling/broken automounted symlinks, +especially on systems under heavy use. To tell the kernel to ignore (flush) +its old entries for an directory, the mtime of the directory must be updated +(monotonically incremented). Amd was indeed doing so in several places, but +unfortunately it was using time(2) which only provides a one-second +resolution clock: this was fine a decade ago, but not good enough on today's +fast systems; using only a one-second resolution clock meant that on busy +systems that invoke Amd many times a second, some rapidly changing entries +do not get flushed from the kernel, and the kernel thus uses stale entries. +The solution to all of these was to rework the whole clock timer updates to +use gettimeofday(), using a micro-second resolution timer, and to use that +time whenever Amd needs to update an mtime/atime/ctime of any node. +Finally, we now update the mtime in places that were never updated before +(when a whole map is flushed or a single entry times out). + +Warning: some OSs, we discovered, are incapable of turning off their +NFS attribute cache entirely. This means that Amd cannot work fully +reliability on these systems, not under heavy load. This is documented in +detail in the README.attrcache file included with this distribution. + +Include test-attrcache script to test the NFS attribute cache behavior using +Amd. + +Tell syslog not to log automatically to /dev/console; it's unfriendly. If +user really wants to, they can set it in /etc/syslog.conf. + +- minor new ports: + i386-pc-linux-deb3.1 + i386-unknown-netbsdelf3.0 + powerpc-apple-darwin8.2.0 + +- bugs fixed: + * minor documentation corrections + +*** Notes specific to am-utils version 6.1.1 + +New amd.conf global parameter: forced_unmounts (default to "no"). If set to +"yes," and the client OS supports forced or lazy unmounts, then Amd will +attempt to use them if it gets any of three serious error conditions when +trying to unmount an existing mount point or mount on top of one: EIO, +ESTALE, or EBUSY. This could be useful to recover from serious conditions +such as hardware failure of mounted disks, or NFS servers which are down +permanently, were migrated, or changed their IP address. Only +"type:=toplvl" mounts hung with EBUSY are forcibly unmounted using this +option: this is useful to ensure that a new Amd can mount itself even if a +previous Amd died and left its mount points hung, or to force Amd to +shutdown cleanly, even if some processes (i.e., user shells) have their CWD +on Amd's own mount point. This functionality is available for Linux, BSD44 +systems, Solaris, OSF/1, and partially for AIX. + +New amd.conf global parameter: truncate_log (default to "no"). If set to +"yes", then Amd will truncate the log file (if it's a regular file) on +startup. This could be useful when conducting extensive testing on Amd maps +(or Amd itself) and you don't want to see log data from a previous run in +the same file. + +- minor new ports: + i386-pc-linux-fc4 + i386-pc-linux-suse9.3 + i386-pc-linuxoldld-deb3.1 + +- bugs fixed: + * safer mtab handling for Linux (locks + handles /proc/mounts) + * small compile problems on Solaris 6 (rpcvers_t) + * small compile problems on HPUX 10 (h_errno) + * possibly missing definition of INADDR_NONE in wire.c + * extern for sleep(3) may be missing on older gcc systems + * updated nfs_args structure on aix4. + * possible running off end of exported_ap[] array. + * buffer overflow in pawd. + * aix4 clean build. + * use strlcat/snprintf in a few places for safety. + * recover from IP address change of a down NFS server + * don't discard restarted mntfs that was used. + +*** Notes specific to am-utils version 6.1 + +- bugs fixed: + * set timeo/retrans for type:=nfs only if user asked + +Also, if you want to know what's new in 6.1 compared to 6.0.x, it's +EVERYTHING below this line, up to "6.0.4." Yes, that's a lot of stuff. + +*** Notes specific to am-utils version 6.1-rc7 + +Remove alloca from am-utils, and rewrite code that used it. + +Assorted minor code cleanups. + +- minor new ports: + alpha-unknown-linux-gentoo1.4.16 + alphaev56-dec-osf4.0f (using both cc and gcc) + mips-sgi-irix6.5 (using both cc and gcc) + mips-unknown-linux-gentoo1.4.16 + sparc64-unknown-linux-gentoo1.4.16 + +- bugs fixed: + * getwire() detects networks correctly on OSF/1 + +*** Notes specific to am-utils version 6.1-rc6 + +- minor new ports: + i386-pc-linux-deb3.0 + i386-pc-linux-gentoo1.4.16 + i386-pc-linux-suse9.2 + i386-unknown-freebsd5.4 + i386-unknown-netbsdelf2.0.2 + i386-unknown-openbsd3.7 + powerpc-unknown-linux-yellowdog2.3 + +- bugs fixed: + * minor compile error of nfs_subr.c on some systems. + * AIX 5.2/5.3 PPC compile fixes. + +*** Notes specific to am-utils version 6.1-rc5 + +For NetBSD systems, $os used to say "netbsdelf1" or "netbsdelf2." Now it +just says "netbsd." + +- minor new ports: + powerpc-apple-darwin7.9.0 + +- bugs fixed: + * silly (but nasty) null pointer dereferencing + * improved fix for '-opts' syntax for resetting map options + +*** Notes specific to am-utils version 6.1-rc4 + +- bugs fixed: + * check for "macosx" in M4 macros properly + * minor memory leaks (thanks to Valgrind) + +*** Notes specific to am-utils version 6.1-rc3 + +Minor fix to configure.in. Reran bootstrap to get a working configure +script that indeed checks for certain Linux nfs/autofs headers. + +*** Notes specific to am-utils version 6.1-rc2 + +New amd.conf global parameter: normalize_slashes (default to "yes"). If set +to "no," then Amd will not condense repeated slashes or remove trailing ones +from strings representing pathnames. This is sometimes useful with SMB +mounts, which often require multiple slash characters in pathnames. + +Using a custom version of strlcpy instead of strncpy (but only where it +makes sense), to minimize string overflow changes. Audited all use of +strncpy/strlcpy to ensure safety. + +On Apple machines, use "powerpc" for $arch, instead of "Power Macintosh". +Also, use sw_vers to find out more appropriate OS name (macosx) and OS +version (10.3.x) than uname(3) reports. + +- minor new ports: + powerpc64-unknown-linux-rhel4 + powerpc64-unknown-linux-sles9 + +- bugs fixed: + * pawd handles all file systems + * fix double-free in type:=nfsx + * timeo and retrans shouldn't be set for type:=toplvl + * fix inconsistency in handling filehandle generation number + * document proper use of hosts.allow (don't spawn) + * single dash '-' map entry now resets the defaults + +*** Notes specific to am-utils version 6.1-rc1 + +- minor new ports: + i386-pc-linux-fc2 (Fedora Core 2). + i386-pc-linux-fc3 (Fedora Core 3). + i386-pc-linux-rhel3 + i386-pc-linux-rhel4 + i386-pc-linux-suse8.2 + i386-pc-linux-suse9.1 + i386-pc-solaris2.10 + i386-unknown-freebsd4.8 i386-unknown-freebsd4.9 + i386-unknown-freebsd4.10 + i386-unknown-freebsd4.11 + i386-unknown-freebsd5.1 (5.1-RELEASE) i386-unknown-freebsd5.2 (5.2-RELEASE) - -- new amd.conf directive "nfs_allow_insecure_port". Used to work around + i386-unknown-freebsd5.2.1 (5.2.1-RELEASE) + i386-unknown-freebsd5.3 (5.3-RELEASE) + i386-unknown-freebsd6.0 (6.0-CURRENT-SNAP001) + i386-unknown-netbsdelf1.6.1 + i386-unknown-netbsdelf2.0 + i386-unknown-openbsd3.6 + ia64-hp-hpux11.20 + ia64-unknown-linux-rhel4 + mipsel-unknown-linux-rhPS2 (Linux on Sony PlayStation 2) + powerpc-apple-darwin7.6.0 + powerpc-apple-darwin7.7.0 + powerpc-apple-darwin7.8.0 + powerpc-ibm-aix5.2.0.0 + powerpc-ibm-aix5.3.0.0 + sparc-sun-solaris2.10 + sparc64-unknown-linux-deb3.0 + x86_64-unknown-linux-rh2.9.5AS + +- support for executable maps ala Sun automounter. Set map_type=exec in + amd.conf, and map_name to a program/script that takes a key as argv[1], + and returns key-value pair on stdout. See also exec_map_timeout [global] + parameter which defines how many seconds (default 10 sec) Amd will wait + for an executable map program to return output before timing out. See + am-utils manual for full details. + +- new amd.conf parameter "nfs_allow_insecure_port". Used to work around bugs in certain kernels, which cause them to try and talk to amd from unprivileged ports. -- bug fixes: - rename log() in fsinfo to avoid glibc/gcc-3.3 conflict - am_pref free NULL pointer - compiles with modern bison/flex versions - handles sites with various combos of db, n/dbm, and gdbm +- new amd.conf parameter: localhost_address. Used to override the localhost + (often 127.0.0.1) address Amd uses to connect to for the local NFS server + and RPC server. + +- new amd.conf [global] parameter: domain_strip (default "yes"). If set to + "no," Amd won't strip domain names from host names, which is useful if + your Amd maps are served by multiple domains and you want to tell from the + logs which exact host did what. + +- new amd.conf [global] parameter: auto_attrcache (default to 0). Sets + Amd's own NFS attribute-cache timeout in seconds. A value of 0 turns off + attribute caching, meaning that Amd will be consulted via a kernel-RPC + each time someone stat's the mount point (which could be abused as a + denial-of-service attack). If you're concerned, set this to something + greater than zero (a value of 1 second is currently recommended). + Warning: if you set this option to any non-zero value, especially a large + value, and you get ESTALE errors on your particular OS, then set this + value back to 0 seconds. + +- four new amd.conf [global] parameters, similar to nfs_retry_interval and + nfs_retransmit_counter, which allow you to set the "timeo" and "retrans" + NFS mount parameters, respectively. Now you can set those parameters + globally and separately for UDP vs. TCP, using any of these: + nfs_retry_interval_udp, nfs_retransmit_counter_udp nfs_retry_interval_tcp, + and nfs_retransmit_counter_tcp. + +- new amd.conf [global] parameter: preferred_amq_port. Allows you to select + the UDP+TCP port that Amd's amq service will use with the RPC portmapper. + Useful with firewalls and NAT'ed environments. + +- new amd.conf option "debug_mtab_file". Allows user to define the mtab + file during debug-mtab mode. The default path is "/tmp/mnttab". + +- new function selector xhost(ARG) which will match ARG against the current + host name. This works even if ARG is a CNAME (unlike the host==ARG + selector). + +- support restarting the automounter's own mount points (only over NFS, + for now). + +- fully support WebNFS as per RFC 2054. It now tries v3/TCP first, falling + back to v2/UDP if this doesn't work. The "webnfs" pseudo-mount options + has been renamed (again) to "public" to match Solaris 2. + +- restructured the restarting of already-mounted filesystems, in the process + also fixing a problem with restarting nfsx components. + +- support escaped slashes, needed for SMB mounts. Use '\\\/\\\/' in a + string to get a double slash. + +- amd -v now prints domain, host, and hostd values: foo, example.com, and + foo.example.com, respectively. + +- On Linux, if umount(2) failed with EIO or ESTALE, try the new umount2(2) + system call with MNT_FORCE+MNT_DETACH. This could be quite helpful to + unmounting hung mount points that otherwise cannot be fixed without a + reboot. + +- The ping=N mount option now works. N defaults to 30 seconds for all NFS + servers. It can now be set to any value for each server separately. + Setting it to a large value can reduce the amount of NFS_NULL chatter on + your network considerably, especially in large sites. Setting this to -1 + will turn off pings for that server (useful in NFS-HA setups). Setting N + to 0 will pick the default ping value in Amd (currently 30 seconds). Note + that if you have multiple Amd entries using the same file server, and each + entry sets a different value of N, then each time Amd mounts a new entry, + the ping value will be re-evaluated (and updated, turned off, or turned + back on as needed). Note that NFS_NULL pings are sent for both UDP and + TCP mounts, because even a hung TCP mount can cause user processes to + hang. + +- file system inheritance code restructured, so it's no longer a pseudo file + system, but actually integrated into Amd (as it should have been). + +- for type:=program, the "umount" program doesn't have to be defined; it'll + default to "unmount ${fs}". + +- "amd -v" now prints the distribution name if it's known (e.g., rh9, fc3, + suse8, etc.). -*** Notes specific to am-utils version 6.0.9: +- bugs fixed: + * various memory management problems (leaks, etc) + * fixed nfsx support + * fixed a race involving late replies to network queries which + arrive after the file system has already been mounted + * recognize pcfs_args_t fields in FreeBSD 5 + * recognize other mount types in pawd: host, linkx, and nfsx + * allow exactly one of umount and unmount in type:=program + * race condition between calls to mntctl() on AIX + * plock/mlockall wasn't inherited by fork(); moved after + daemonizing. + * fix inconsistency between Socket and TLI RPC timeouts. + * don't warn when couldn't rmdir a dir with a readonly ancestor. + * avoid hangs of amd in ctl-amd (must chdir to /) + * workaround occasional daemonizing problems (parent won't die) + * don't hang on exit if debug_options=mtab was used + * utimeout=N mount option works with non-nfs types (ufs, pcfs, etc.) + * SEGV (null pointer deref) in type:=program and type:=cachefs + * unmount_on_exit of type:=program caused amd to hang + * match amd2ldif output with ldap.schema + +*** Notes specific to am-utils version 6.1b4 + +- minor new ports: + i686-apple-darwin6.6 + +- speed up the recovery of inherited (restarted) filesystems by using the + proper waiting channels + +- added support for mounting webnfs filesystems, see entry below. It doesn't + do any probing currently, so it will default to v2/UDP unless another + version and/or protocol are explicitly specified. + +- pseudo-mount option "ignore_portmapper" renamed to "webnfs" + +- bugs fixed: + * properly time out autofs filesystems on Linux + * link mounts with relative targets weren't working on autofs + * the link side of the nfsl file system wasn't working on autofs + * umount code was accidentally turning all symlinks into directories + during attempted umounts, causing stale filehandles + * various minor build fixes for "impossible" configurations + * prevent ldap code from dereferencing a null pointer + +*** Notes specific to am-utils version 6.1b3 + +- new amd.conf option autofs_use_lofs, set by default to "yes". "yes" means + using in-place mounts (lofs, bind mounts, etc.), thus utilizing one of + Autofs's main advantages. "no" means using symlinks instead, which has + the "/bin/pwd" problem and certain efficiency issues on Solaris 2.6+ and + is also not supported on Solaris Autofs v1 and derivatives; however, the + autofs code that uses symlinks is simpler and more thoroughly tested. + +- new amd.conf option map_default (can be used in [global] and overwritten + in the per-map section). This will overwrite the /defaults entry of the + map itself, to allow people to set defaults in amd.conf (useful when you + cannot control your amd maps, or you'd rather not modify them globally). + +- for type:=program, you can use either unmount:=XXX or umount:=XXX (but not + both). This new 'unmount' name is an alias for convenience. + +- fixed the "multiple matching sub-entries in a map entry" semantics to try + mounting those sub-entries one by one, until either one succeeds or all + fail. The old semantics of trying to mount everything in parallel and use + the one that mounted fastest hasn't worked in a long time; in fact, 6.0 + currently simply ignores all but the first matching sub-entry. + +- made amd fail much faster (instantly, in fact) if the remote server + doesn't have a functional portmapper or NFS service. Also reduced the + total timeout to 3 seconds for a completely downed server. + +- new pseudo-mount option "ignore_portmapper"; not very useful currently, + will make more sense when we also accept hard-coded ports for mountd and + nfsd. + +- amd will no longer query the portmapper for all possible NFS versions and + protocols if the user requested to use specific ones. + +- increased the major number for the library, so that 6.0 and 6.1 can't + share libraries anymore. + +- support tcpd/libwrap tcpwrappers. If your system supports libwrap, then + you can use /etc/hosts.allow and /etc/hosts.deny to control remote Amq + access to Amd. The new amd.conf parameter use_tcpwrappers is set to "yes" + by default. + +- support NULL entries in Hesiod maps, if they start with a ".". + +- code reorganization + +- documentation cleanup, corrections, and general updates. Better + references to all man pages. Support newer texi2html. Proper building of + DVI and PSI files. Allow building of am-utils manual in one long Web + page. + +- minor new ports: + + ia64-unknown-linux-rh2.1AS (Red Hat Itanium Advanced Server) + i386-unknown-freebsd5.0 (5.0-RELEASE) + sparc64-unknown-linux-suse7.3 + i386-unknown-netbsdelf1.6.1 + i386-unknown-openbsd3.3 + i386-pc-solaris2.9 + +- bugs fixed: + + * autofs mode on Linux was segfaulting on a silly error (and noone + complained, which proves that I'm probably the only one testing + these beta releases, tsk tsk). + * fixed handling of host entries over autofs. + * fixed handling of nfsl entries over autofs. + * the matching in find_mntfs() was causing problems for inherited + filesystems, so make an exception for them. Tighten the + matching even more, to take into account the f/s type as well. + * recognize xlatecookie mnttab option on netbsd + * document Solaris lex bug (use flex) + * document AIX 5.x NFS bug (need patch) + * document Solaris 8 autofs version change (need to fix system + header file) + * ensure lex doesn't run out of output slots + * support GNU flex-2.5.31+ + * force version.texi to be rebuilt unconditionally + * mk-amd-map open db file exclusively (security) + * turn off maintainer-only rules in distros + * don't core dump if log_file is NULL (Solaris) + * don't include malloc.h if stdlib.h exists + * recognize file system failures (EIO) upon reading file maps + +*** Notes specific to am-utils version 6.1b2 + +- new mount flag "softlookup", which determines how amd will respond to + lookups of NFS shares already mounted (return a valid symlink or return + EIO). The default, if "softlookup" is not specified, depends on whether the + mount is "soft" or "hard". + +- return EIO instead of ENOENT if amd thinks the server is down; this allows + well-written applications to sleep and retry the operation. + +- minor new ports: -- Minor new ports: i386-apple-darwin6.0 + i386-pc-linux-rh8.0 + ia64-unknown-linux-rh2.1AW sparc-sun-solaris2.9 +- automatic support for loop mounts on Linux (deprecates the "loop" mount + option) + +- new amd.conf parameter ldap_proto_version (default 2) for setting the LDAP + protocol version to use. + - bugs fixed: + * redundancy mode (multiple servers for the same share) wasn't working + * non-autofs mode had some rather nasty hangs on downed file servers + * double-free'ing problem in assign_error_mntfs and free_continuation + * free'ing non-malloc'ed memory in amfs_auto_mount + * late server ping replies were not ignored + * amfs_auto_lookup_mntfs wasn't propagating errors up to callers + * autofs-v4 on Solaris 9 works * handle std{in,out,err} correctly when releasing controlling tty (for real this time) * don't cast pointers between enum_t and u_long, it doesn't work on 64-bit big-endian platforms + * fix compile problem with mlockall() on Darwin -*** Notes specific to am-utils version 6.0.8: +*** Notes specific to am-utils version 6.1b1 + +- Major Autofs work + Partial support for Sun Autofs v1 + Documented known problems with Sun Autofs v1 (possible deadlocks) + Fixes for Sun Autofs v2/v3 + Preliminary support for Sun Autofs v4 (Solaris 9) + Kernel-based expirations for Linux Autofs - Minor new ports: - i386-pc-linux-rh7.2 - i386-pc-linux-rh7.3 - i386-unknown-freebsd4.4 - i386-unknown-freebsd5.0 - ia64-hp-hpux11.20 - ia64-unknown-linux-rh7.1 powerpc-ibm-aix5.1.0.0 + i386-unknown-netbsd1.6A - Work around IBM's NFSv3 ABI change in aix4.3 -- Support network/netmask and network/masklen syntax in in_network() +- trivial regression test suite started: run "make check" on a built + am-utils to execute tests. Currently only one test which checks to see if + "amd -v" executes correctly. + +- new command line option "amd -A arch" to overwrite the value of $arch. + +- bugs fixed: + + * Linux loop mounts of ISO images + * assorted LDAP fixes + * strerror not found on some systems + * small fixes for hpux9 and aix43 + * exclude ldap/hesiod support unless both libraries+headers exist + * fully support "xlatecookie" mount option + * security: if -D noamq option, don't listen on socket. + +*** Notes specific to am-utils version 6.1a5: -- Support disabling LDAP and Hesiod support using configure +- browsable_dirs support for Solaris autofs, *without* mount storms! -- Support xfs on Linux +- new amd.conf global parameter: map_reload_interval (default 1 hour). + Determines how often Amd checks to see if maps have changed at the source + (and then reloading only those that have changed). -- Red Hat specific Amd startup script included +- "amd -v" now lists bug-reporting address. -- Remove (non-functional) autofs code and detection support. If you want - autofs support, use am-utils-6.1. +- assorted code cleanups and porting to use latest versions of GNU + Autotools. + +- opts:=loop works for type:=cdfs, for mounting ISO-9660 files on Linux. - bugs fixed: + * fixed sublink support in Linux autofs (broken in a4) * hlfsd takes uid 0's home from root's passwd entry instead of defaulting to '/' * (not really our bug) Linux ignores the microseconds field in mtime, so hlfsd and amd need to increment the seconds field all the time to prevent symlink caching - * generic map parsing bug which was rejecting a numerical mount option - if it was the last option in the string - * MacOS X compile bug - * minor fix for GNU getopt - * Linux has no "dev" mtab option - * "nolock" is an NFS mount option, not a generic one - * Irix N32 ABI fixes with cc - * security: use mkstemp instead of mktemp - * correct timeo values for Linux tcp/udp NFS mounts - * hlfsd use of setuid() not seteuid() - * AIX ABI changes to nfs_args fixes - -*** Notes specific to am-utils version 6.0.7: - -- minor new ports: i386-unknown-freebsd5.0 + * generic map parsing bug which was rejecting a numerical mount + option if it was the last option in the string. + * file descriptor leak in Linux autofs. + * "nolock" is an NFS mount option, not a generic one. + * use mlockall(2) on systems that have it, for plock=yes. Now + pinning Amd's pages in memory works on Linux. + * ctl-amd/ctl-hlfsd correctly refer to @sysconfdir@ for alternate + location of configuration files. -- bug fixes: - type:=host, don't fail mount if one share is already mounted - don't report NFS NE_IO error as success (Linux) +*** Notes specific to am-utils version 6.1a4: -*** Notes specific to am-utils version 6.0.6: +- full autofs support for Solaris 2.[67], including symlinks, sublinks and + direct mounts -- new ports: - Mac OS X support (Darwin, Rhapsody) +- fixed mount/umount deadlock in Linux autofs -- reworked autoconf scripts so that newer versions of known OSs will default - their configuration to the last known version (better chance that it will - pass autoconf and work). +- fixed sublinks in Linux autofs -- linux support for lofs and "bind" mounts, so type:=lofs can work (whether - you're using autofs or not) +- support for network/netmask pairs in the in_network() selector -- renamed amd.conf option "selectors_on_default" to "selectors_in_defaults" - (on -> in, and added "s"). Old name remains in place for compatibility. +- support disabling LDAP and Hesiod support using configure -- lots of cleanups of debugging and logging messages +- forward-ported all the fixes from the stable branch (MacOS X support, + minor Linux fixes) -- Bugs fixed: - lots of Linux-related ones, so it compiles for recent kernels - better checking on various mount options of the form foo=N or foo=STR - fixed NFS errno mapping bug which mapped ENOENT to success on Linux +- bind-mount support for type==link and type==lofs with Linux 2.4+ + +- FiST lofs support under Linux (also in 6.0.6s2) + +*** Notes specific to am-utils version 6.1a3: + +- various things from the 6.0 branch: + compile fixes for Linux 2.4-ac and 2.2.19pre+ + Darwin/Rhapsody/OS X support + much reduced configure script (works around a bug in Darwin's cpp) -*** Notes specific to am-utils version 6.0.5: +*** Notes specific to am-utils version 6.1a2: -- Minor ports: - alphaev6-dec-osf5.1 - i386-unknown-freebsd4.2 - support for Linux 2.4 kernels and newer GLIBC versions - cleanup of AIX and IRIX6 ports +- working autofs support for Solaris 2.[67], but incomplete -- Support new mount options: - proplist (ACLs over NFS, DU-4.0) - kerb, rdirplus, readdirsize, and xlatecookie (NetBSD-1.5K+) - nonlm, lock, nolock (Linux) +- forward-ported all the changes up to 6.0.5s2 -- Don't force rsize/wsize on Linux 2.2.18+ and 2.4.x. Improves performance - a lot. +- removed support for amq -M -- Lots of documentation updates: texinfo, man pages, scripts, and more. +- known bugs + nfsx support is broken + linux NFS codes fixes + NFS cache aliasing fixes + lots of stuff ported from 6.0 branch -- LDAP: included proposed Schema and Internet Draft for LDAP Schema. See - README.ldap for more information. +*** Notes specific to am-utils version 6.1a1: -- Removed all remains of the dangerous amq -M code. +- working autofs (v3 and v4) support for Linux! -- Major bug fix: initialize NFS fileid field correctly, to avoid cache - aliasing problems, esp. on Linux. +- forward-ported all the changes in 6.0.4s4 -- Major Linux bug fix: map errnos to NFS errors (mistakenly turned off a - while back). Also support unused errno 41. +- bugs fixed + client-side fail-over to NFSv2/UDP -- Other bugs fixes: - default MAXHOSTNAMELEN (if undefined) is 256, not 64 - truncate hostnames to MAXHOSTNAMELEN (security fix) - alignment problem in getwire() on ALPHA - other buf overflow problems (security fixes) - ctl-amd supports "condrestart" (Red Hat) - bell char is \g not \a - lostaltmail verify MAILDIR is defined - am-eject accepts "floppy" and "cdrom" as args - other assorted small bug fixes +- known bugs + autofs v3 will probably break with host maps *** Notes specific to am-utils version 6.0.4: @@ -514,7 +1030,7 @@ the functions compute_nfs_args() and compute_automounter_nfs_args(). - bugs fixed: mnttab name ufs/cdfs/pcfs/etc filesystems corrected use pmap_ping for amq (a must for secure portmappers, bsdi2/3) - test for xfs (irix) as a disk-based filesystem + test for xfs (irix) as a disk-based file system set correct nfs_prot headers for Solaris 2.5 removed stale code from lostaltmail.in lostaltmail will look for conf file in multiple locations @@ -931,11 +1447,11 @@ used anywhere were removed. --enable-libs[=ARG] configure/compile with ARG (-L/-l) library flags -- filesystem, mount table entries, and mount type tests can now look in +- file system, mount table entries, and mount type tests can now look in /lib/modules and /proc/filesystems for statically/dyadically loadable kernel modules (linux) -- prefer vfat over msdos/pc/etc filesystem for PCFS. +- prefer vfat over msdos/pc/etc file system for PCFS. - moved all fixed headers to include/am_defs. Left only #define/#undef entries in aux/acconfig.h. |
