associated changes that had to happen to make this possible as well as
bugs fixed along the way.
Bring in required TLI library routines to support this.
Since we don't support TLI we've essentially copied what NetBSD
has done, adding a thin layer to emulate direct the TLI calls
into BSD socket calls.
This is mostly from Sun's tirpc release that was made in 1994,
however some fixes were backported from the 1999 release (supposedly
only made available after this porting effort was underway).
The submitter has agreed to continue on and bring us up to the
1999 release.
Several key features are introduced with this update:
Client calls are thread safe. (1999 code has server side thread
safe)
Updated, a more modern interface.
Many userland updates were done to bring the code up to par with
the recent RPC API.
There is an update to the pthreads library, a function
pthread_main_np() was added to emulate a function of Sun's threads
library.
While we're at it, bring in NetBSD's lockd, it's been far too
long of a wait.
New rpcbind(8) replaces portmap(8) (supporting communication over
an authenticated Unix-domain socket, and by default only allowing
set and unset requests over that channel). It's much more secure
than the old portmapper.
Umount(8), mountd(8), mount_nfs(8), nfsd(8) have also been upgraded
to support TI-RPC and to support IPV6.
Umount(8) is also fixed to unmount pathnames longer than 80 chars,
which are currently truncated by the Kernel statfs structure.
Submitted by: Martin Blapp <mb@imp.ch>
Manpage review: ru
Secure RPC implemented by: wpaul
happened as it was working around problems elsewhere (ie: binutils/ld
not doing the right thing according to the ELF design). libcrypt has
been adjusted to not need the runtime -lmd. It's still not quite right
(ld is supposed to work damnit) but at least it doesn't impact all the
users of libcrypt in Marcel's cross-build model.
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.
Boy, I'm glad we're not using sup anymore. This update would have been
insane otherwise.
find two users with the same UID (i.e. root and toor), but yp_mkdb(8)
forbits duplicate keys, so only one of them will end up in the *.byuid
maps (probably toor, since it comes after root in the template file).
If I asked rpc.yppasswdd(8) to change toor's password, it would update
the *.byname maps correctly, but incorrectly modify root's entry in
the *.byuid maps since the only matching record with UID=0 in those
maps belongs to root.
To fix this, we check that both the name and UID are correct before trying
to write new entries to the maps.
/var/yp/master.passwd template file and it uses the same kind of code
as chpass(1), it may also be vulnerable to the bug from PR #1519.
May as well deal with it since I'm in the area. (yppasswdd in -stable
doesn't do additions, therefore it shouldn't be have this problem.)
by Peter Wemm:
- In yppasswdproc_update_1_svc(), I wasn't paying attention and put
a couple of lines of code _after_ a return() instead of before.
(*blush*)
- The removal of certain temp files didn't always work (this showed
up mostly if you were using /etc/master.passwd as your NIS passwd
template instead of /var/yp/master.passwd). This is because the
whole temp file creation mechanism I was using was tragically
broken (you can't rename across filesystems).
This problem I found myself:
- If you have a very large password database (30,000 or more entries),
there can be a delay of several seconds while pw_copy() copies the
ASCII template file and subsitutes in the modified/new entry. During
this time, the clnt_udp() code in the RPC library may get impatient
and retry its request. This will get queued at the server and be
treated as a second request. By then the password change will have
been completed and the second request will fail (the old password is
no longer valid). To attempt to fix this, we save the IP address and
port of each request and ignore any subsequent requests from the
same IP and same port that arrive within five minutes of each other.
option to pwd_mkdb and adding this option to utilities invoking it.
Further, the filling of both the secure and insecure databases has been
merged into one loop giving also a performance improvemnet.
Note that I did *not* change the adduser command. I don't read perl
(it is a write only language anyway).
The change will drastically improve performance for passwd and
friends with large passwd files. Vipw's performance won't change.
In order to do that some kind of diff should be made between the
old and new master.passwd and depending the amount of changes, an
incremental or complete update of the databases should be agreed
upon.
If rpc.yppasswdd is invoked with the -i flag, password changes will
be made to the master.passwd template file and the hash map files
in-place, which means it won't have to run a complete map update.
Instead, it calls /var/yp/Makefile with the 'pushpw' target, which
just pushes the maps to the slaves and runs yp_mkdb -c to tell the
local ypserv to flush its database cache.
The server will check the passwd.byname and passwd.byuid maps to see
if they were built in 'insecure' or 'secure' mode (i.e. with real
encrypted passwords in them or without) and update them accordingly.
This combined with rpc.ypxfrd greatly reduces the amount of time it
takes to complete an NIS password change, especially with very large
passwd databases.
really own (and which can end up being mangled later). The manifestation
of this bug is that the first attempt by a user to change their NIS password
succeeds, but all subsequent attempts fail. rpc.yppasswdd also logs
a message about not being able to find a file called
'/var/yp/<some garbage string>/master.passwd.' (Note that for some
bizarre reason, this doesn't happen with the malloc() from FreeBSD 2.1.0.
I suppose this means we can chalk up another victory for phkmalloc. :)
This bug only occurs if you use the -m flag with rpc.yppasswdd.
Fix this by copying the domain name to a static buffer and returning
a pointer to that instead.
Reported by: Jian-Da Li (jdli@csie.nctu.edu.tw)
man pages up to mdoc guidelines and fix some minor formatting glitches.
Also fixed a number of man pages to not abuse the .Xr macro to
display functions and path names and a lot of other junk.
also controlled by /var/yp/securenets).
Add -u flag to turn off the privileged port check done by yp_access();
some commercial systems (IRIX, Solaris 2.x, HP-UX, and probably others)
don't use a reserved port for submitting yppasswd updates. If we always
enforce the check, these client systems will be unable to submit updates
to us.
Document securenets support and -u flag in man page.
Like ypserv, you can compile rpc.yppasswdd to use the tcpwrapper package
instead of securenets if you want to.
chpass(1) are on the way too.) This version supports all the features
of the old one and adds several new ones:
- Supports real multi-domain operation (optional, can be turned
on with a command-line flag). This means you can actually have
several different domains all served from one NIS server and
allow users in any of the supported domains to change their passwords.
The old yppasswdd only allowed changing passwords in the domain
that was set as the system default domain name on the NIS master
server. The new one can change passwords in any domain by trying
to match the user information passed to it against all the passwd
maps it can find. This is something of a hack, but the yppasswd.x
protocol definiton does not allow for a domain to be passwd as an
argument to rpc.yppasswdd, so the server has no choice but to
grope around for a likely match. Since this method can fail if
the same user exists in two domains, this feature is off by default.
If the feature is turned on and the server becomes confused by
duplicate entries, it will abort the update.
- Does not require NIS client services to be available. NIS servers do
_NOT_ necessarily have to be configured as NIS clients in order to
function: the ypserv, ypxfr and yppush programs I've written recently
will operate fine even if the system domain name isn't set, ypbind isn't
running and there are no magic '+' entries in any of the /etc files.
Now rpc.yppasswdd is the same way. The old yppasswdd would not work
like this because it depended on getpwent(3) and friends to look up
users: this will obviously only work if the system where yppasswdd is
running is configured as an NIS client. The new rpc.yppasswdd doesn't
use getpwent(3) at all: instead it searches through the master.passwd
map databases directly. This also makes it easier for it to handle
multiple domains.
- Allows the superuser on the NIS master server to change any user's
password without requiring password authentication. rpc.yppasswdd
creates a UNIX domain socket (/var/run/ypsock) which it monitors
using the same svc_run() loop used to handle incoming RPC requests.
It also clears all the permission bits for /var/run/ypsock; since
this socket is owned by root, this prevents anyone except root from
successfully connect()ing to it. (Using a UNIX domain socket also
prevents IP spoofing attacks.) By building code into passwd(1) and
chpass(1) to take advantage of this 'trusted' channel, the superuser
can use them to send private requests to rpc.yppasswdd.
- Allows the superuser on the NIS master to use chpass(1) to update _all_
of a user's master.passwd information. The UNIX domain access point
accepts a full master.passwd style structure (along with a domain
name and other information), which allows the superuser to update all
of a user's master.passwd information in the NIS master.passwd maps.
Normal users on NIS clients are still only allowed to change their full
name and shell information with chpass.
- Allows the superuser on the NIS master to _add_ records to the NIS
master.passwd maps using chpass(1). This feature is also switchable
with a command-line flag and is off by default.
