Skip to contentSkip to conversation

Making crossmnt a default export option

DISCUSSION Steve Dickson 5 participants 5 messages Mar 25, 2009 – Apr 1, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Hello,

What would be the ramifications of turning on the crossmnt option on by default? Currently the the default is nocrossmnt.

In my recent work with pseudo exports, it has become very apparent that having the crossmnt always set, especially with multi file system exports, allows mount to "just work"!

So historically what is the reason for not allowing the clients to always cross file systems? Is it a security issue? And again, what would be the ramifications always allowing, by default, clients to cross file systems?

steved.

#1 lore raw
Quoted 5 lines · Steve Dickson

Hello,

What would be the ramifications of turning on the crossmnt option on by default? Currently the the default is nocrossmnt.

Quoted 8 lines

In my recent work with pseudo exports, it has become very apparent that having the crossmnt always set, especially with multi file system exports, allows mount to "just work"!

So historically what is the reason for not allowing the clients to always cross file systems? Is it a security issue? And again, what would be the ramifications always allowing, by default, clients to cross file systems?

http://www.ietf.org/rfc/rfc1813.txt:
3.3.3 Procedure 3: LOOKUP - Lookup filename
IMPLEMENTATION

At first glance, in the case where what.name refers to a mount point on the server, two different replies seem possible. The server can return either the file handle for the underlying directory that is mounted on or the file handle of the root of the mounted directory. This ambiguity is simply resolved. A server will not allow a LOOKUP operation to cross a mountpoint to the root of a different filesystem, even if the filesystem is exported.

Quoted 7 lines

steved.

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
In reply to Steve Dickson
Quoted 5 lines · Steve Dickson

Hello,

What would be the ramifications of turning on the crossmnt option on by default? Currently the the default is nocrossmnt.

Quoted 8 lines

In my recent work with pseudo exports, it has become very apparent that having the crossmnt always set, especially with multi file system exports, allows mount to "just work"!

So historically what is the reason for not allowing the clients to always cross file systems? Is it a security issue? And again, what would be the ramifications always allowing, by default, clients to cross file systems?

I think the only real problem would be that Solaris tends to get very confused when you do. As Benny pointed out, the original NFSv2 and NFSv3 specs didn't allow you to cross mount points, and so Solaris never got round to implementing it.
I don't know what the current status is, but the original Solaris 10 NFSv4 client was also incapable of crossing mount points despite the fact that rfc3530 does support it.

Cheers
Trond

#3 lore raw
In reply to Steve Dickson
Quoted 5 lines · Steve Dickson

Hello,

What would be the ramifications of turning on the crossmnt option on by default? Currently the the default is nocrossmnt.

Quoted 8 lines

In my recent work with pseudo exports, it has become very apparent that having the crossmnt always set, especially with multi file system exports, allows mount to "just work"!

So historically what is the reason for not allowing the clients to always cross file systems? Is it a security issue? And again, what would be the ramifications always allowing, by default, clients to cross file systems?

As Trond mentioned, the OpenSolaris client doesn't automatically create mount points when crossing the server's filesystem boundaries (for NFSv3).
NFSv4 support for that behavior is now present for OpenSolaris.

So the risk is running into clients that don't behave like the Linux client and then "mv" will not be able to determine the appropriate behavior when mv'ing files between those filesystems.

Spencer

Quoted 8 lines

steved.

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

#4 lore raw
Quoted 15 lines · spencer shepler
Quoted 5 lines · Steve Dickson

Hello,

What would be the ramifications of turning on the crossmnt option on by default? Currently the the default is nocrossmnt.

Quoted 8 lines

In my recent work with pseudo exports, it has become very apparent that having the crossmnt always set, especially with multi file system exports, allows mount to "just work"!

So historically what is the reason for not allowing the clients to always cross file systems? Is it a security issue? And again, what would be the ramifications always allowing, by default, clients to cross file systems?

Quoted 8 lines

As Trond mentioned, the OpenSolaris client doesn't automatically create mount points when crossing the server's filesystem boundaries (for NFSv3).
NFSv4 support for that behavior is now present for OpenSolaris.

So the risk is running into clients that don't behave like the Linux client and then "mv" will not be able to determine the appropriate behavior when mv'ing files between those filesystems.

And some applications may be confused when they see different files that appear to be on the same filesystem but that have different inodes. Etc.

And this requires people to change their exports. We've done this before (with sync and subtree_check), but it's something to avoid if at all possible.

So preserving the current behavior for v2/v3 is important.

--b.