You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Rasmus Lerdorf <ra...@lerdorf.on.ca> on 1997/06/17 04:31:00 UTC
CVS Access?
I'd really like to play with the new NT code in a bit more of a timely
fashion than what the from-cvs directory allows me. I don't need a
HyperReal account, just an entry in the CVSROOT passwd file. If you are
running a semi-recent version of CVS, create a "writers" file that lists
the users with write access, and anybody else in the passwd file will be
read-only CVS users.
-Rasmus
Re: CVS Access?
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 17 Jun 1997, Dean Gaudet wrote:
> Brian (and me) really want to get cvsup going ... but hyperreal is BSDI.
> If someone has already built this stuff for bsdi (and linux, because I
> want to mirror the full repository) and knows how to set it up could you
> step forward? :) Otherwise I'll try to get to it later this week. I
> sort of dread having to build modula-3 first.
I'm (unfortunately) not sure that cvsup is the way to go for this. It is
not accessable enough for people using systems that don't have prebuild
binaries.
I'm wondering if either an anoncvs setup, or even just using sup would be
a better general choice. anoncvs has the disadvantage that it sucks. sup
has the disadvantage that it downloads all of a new file, not just the
changes.
cvsup is the coolest thing since sliced bread, but it isn't nice to build
on many systems.
>
> I know that I had read-only access to the repository long before I had
> write access. But Brian had to create me an accout. So you'll probably
> have to wait until he's back.
>
> Dean
>
> On Mon, 16 Jun 1997, Rasmus Lerdorf wrote:
>
> > I'd really like to play with the new NT code in a bit more of a timely
> > fashion than what the from-cvs directory allows me. I don't need a
> > HyperReal account, just an entry in the CVSROOT passwd file. If you are
> > running a semi-recent version of CVS, create a "writers" file that lists
> > the users with write access, and anybody else in the passwd file will be
> > read-only CVS users.
> >
> > -Rasmus
> >
> >
>
Re: CVS Access?
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 17 Jun 1997 rasmus@bellglobal.com wrote:
> > I don't trust it. I guess the main thing I don't trust is CVS's remote
> > server that is started as root before switching to whatever uid. Since
> > that is running on taz anyway, it shouldn't hurt too much to provide this
> > type of read only access.
>
> But, do you have any basis for this lack of trust?
Yes.
>
> It simply does a setuid (pw->pw_uid); along with setting USER and LOGNAME
> env variables. This is exactly how most people run Apache. And Apache
> simply does a setuid() as well. Why trust one less than the other?
I don't trust Apache either. Guess that just says I'm paranoid.
I have gone through Apache and fixed most of the blatent security holes.
I have not begun to look at the CVS server. If someone who knows what
they are doing gets a shell on taz, then they can get root.
There is more code before the setuid() in CVS, and more possible code
paths. After the setuid(), there are various bits of code that look
insecure just by looking. I don't think CVS was really designed with the
idea of an untrusted user in mind and I don't think it is necessarily
secure in that role.
>
> Have a look at the CVS server.c file. It isn't very complicated, and I don't
> see any blatant problems in it.
>
> I have had read-only CVS access to the mod_php sources for well over a year
> now. I have never had any problems with it.
>
> -Rasmus
>
Re: CVS Access?
Posted by ra...@bellglobal.com.
> I don't trust it. I guess the main thing I don't trust is CVS's remote
> server that is started as root before switching to whatever uid. Since
> that is running on taz anyway, it shouldn't hurt too much to provide this
> type of read only access.
But, do you have any basis for this lack of trust?
It simply does a setuid (pw->pw_uid); along with setting USER and LOGNAME
env variables. This is exactly how most people run Apache. And Apache
simply does a setuid() as well. Why trust one less than the other?
Have a look at the CVS server.c file. It isn't very complicated, and I don't
see any blatant problems in it.
I have had read-only CVS access to the mod_php sources for well over a year
now. I have never had any problems with it.
-Rasmus
Re: CVS Access?
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 17 Jun 1997 rasmus@bellglobal.com wrote:
> > I know that I had read-only access to the repository long before I had
> > write access. But Brian had to create me an accout. So you'll probably
> > have to wait until he's back.
>
> Well, there is no requirement for an actual account to exist in order to
> have CVS read access. I have included the appropriate bit of CVS
> documentation explaining this. I am not sure we necessarily need cvsup.
> Straight remote CVS may not be super-efficient, but it would certainly
> be more efficient than me downloading the entire snapshot every six hours.
> >From a bandwidth point of view I'd only be fetching patches.
I don't trust it. I guess the main thing I don't trust is CVS's remote
server that is started as root before switching to whatever uid. Since
that is running on taz anyway, it shouldn't hurt too much to provide this
type of read only access.
<RANT TYPE="OpenBSD">
For those not entirely cvs-literate, this is similar to the anoncvs
service that OpenBSD goes on and on and on about to the point of making
their sources nearly impossible to access any other way, while continually
whining about how other *BSDs aren't as open because they don't use
anoncvs but only have 103 other ways of getting the source tree, most of
which OpenBSD doesn't have.
</RANT>
I still don't trust it, but wouldn't object. I would suggest this is
really Brian's decision though, so needs to wait for him to rejoin the
world of the dead...
>
>
> When authenticating a password, the server first checks for the user
> in the CVS `passwd' file. If it finds the user, it compares against
> that password. If it does not find the user, or if the CVS `passwd'
> file does not exist, then the server tries to match the password using
> the system's user-lookup routine. When using the CVS `passwd' file,
> the server runs under as the username specified in the the third
> argument in the entry, or as the first argument if there is no third
> argument (in this way CVS allows imaginary usernames provided the CVS
> `passwd' file indicates corresponding valid system usernames). In any
> case, CVS will have no privileges which the (valid) user would not have.
>
> It is possible to "map" cvs-specific usernames onto system usernames
> (i.e., onto system login names) in the `$CVSROOT/CVSROOT/passwd' file
> by appending a colon and the system username after the password. For
> example:
>
> cvs:ULtgRLXo7NRxs:kfogel
> generic:1sOp854gDF3DY:spwang
> anyone:1sOp854gDF3DY:spwang
>
>
Re: CVS Access?
Posted by ra...@bellglobal.com.
> I know that I had read-only access to the repository long before I had
> write access. But Brian had to create me an accout. So you'll probably
> have to wait until he's back.
Well, there is no requirement for an actual account to exist in order to
have CVS read access. I have included the appropriate bit of CVS
documentation explaining this. I am not sure we necessarily need cvsup.
Straight remote CVS may not be super-efficient, but it would certainly
be more efficient than me downloading the entire snapshot every six hours.
>From a bandwidth point of view I'd only be fetching patches.
When authenticating a password, the server first checks for the user
in the CVS `passwd' file. If it finds the user, it compares against
that password. If it does not find the user, or if the CVS `passwd'
file does not exist, then the server tries to match the password using
the system's user-lookup routine. When using the CVS `passwd' file,
the server runs under as the username specified in the the third
argument in the entry, or as the first argument if there is no third
argument (in this way CVS allows imaginary usernames provided the CVS
`passwd' file indicates corresponding valid system usernames). In any
case, CVS will have no privileges which the (valid) user would not have.
It is possible to "map" cvs-specific usernames onto system usernames
(i.e., onto system login names) in the `$CVSROOT/CVSROOT/passwd' file
by appending a colon and the system username after the password. For
example:
cvs:ULtgRLXo7NRxs:kfogel
generic:1sOp854gDF3DY:spwang
anyone:1sOp854gDF3DY:spwang
Re: CVS Access?
Posted by Dean Gaudet <dg...@arctic.org>.
Brian (and me) really want to get cvsup going ... but hyperreal is BSDI.
If someone has already built this stuff for bsdi (and linux, because I
want to mirror the full repository) and knows how to set it up could you
step forward? :) Otherwise I'll try to get to it later this week. I
sort of dread having to build modula-3 first.
I know that I had read-only access to the repository long before I had
write access. But Brian had to create me an accout. So you'll probably
have to wait until he's back.
Dean
On Mon, 16 Jun 1997, Rasmus Lerdorf wrote:
> I'd really like to play with the new NT code in a bit more of a timely
> fashion than what the from-cvs directory allows me. I don't need a
> HyperReal account, just an entry in the CVSROOT passwd file. If you are
> running a semi-recent version of CVS, create a "writers" file that lists
> the users with write access, and anybody else in the passwd file will be
> read-only CVS users.
>
> -Rasmus
>
>