2.6.22 [PATCH 0/3] NLM: fix source address in server callbacks

DISCUSSION Frank van Maarseveen 3 participants 3 messages Jul 9, 2007

Subsystems (2)

NFS client NFS server

lore

Share & tools

lore

Frank van Maarseveen

author

2007-07-09 20:19 UTC #0 lore raw

Here's the patchset again but this time for 2.6.22 (verified) and based on the earlier cleanup patch for creating a SUNRPC transport. The xs_bind (formerly xs_bindresvport) function slightly been recoded to make more clear what it does.

The first patch changes the TCP server socket code to save the socket name right after an accept(). This local address of a connection (destination address of a request) is needed for connecting back when doing the callback later. For UDP the address was already saved for RPC replies because UDP is connectionless. This patch is IPv6 clean.

The second patch changes the NFS client code to be able to bind not only to a privileged port but also to bind to a specific local address. The client code is used by the server for callbacks.

The third patch is the actual NLM fix.

Signature · 10 lines

Frank

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
NFS maillist  -  [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

Trond Myklebust

2007-07-09 22:23 UTC #1 lore raw

Quoted 11 lines · Frank van Maarseveen

Here's the patchset again but this time for 2.6.22 (verified) and based on the earlier cleanup patch for creating a SUNRPC transport. The xs_bind (formerly xs_bindresvport) function slightly been recoded to make more clear what it does.
The first patch changes the TCP server socket code to save the socket name right after an accept(). This local address of a connection (destination address of a request) is needed for connecting back when doing the callback later. For UDP the address was already saved for RPC replies because UDP is connectionless. This patch is IPv6 clean.

Quoted 5 lines

The second patch changes the NFS client code to be able to bind not only to a privileged port but also to bind to a specific local address. The client code is used by the server for callbacks.
The third patch is the actual NLM fix.

I've appended them to the NFS_ALL and the git tree. Unless Neil objects, I propose that we just shove them in together with the existing NFS merge.

Cheers
Trond

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

Neil Brown

2007-07-09 22:48 UTC #2 lore raw

Quoted 16 lines · On Monday July 9, [email protected] wrote:

Quoted 15 lines · Frank van Maarseveen
Here's the patchset again but this time for 2.6.22 (verified) and based on the earlier cleanup patch for creating a SUNRPC transport. The xs_bind (formerly xs_bindresvport) function slightly been recoded to make more clear what it does.
The first patch changes the TCP server socket code to save the socket name right after an accept(). This local address of a connection (destination address of a request) is needed for connecting back when doing the callback later. For UDP the address was already saved for RPC replies because UDP is connectionless. This patch is IPv6 clean.
The second patch changes the NFS client code to be able to bind not only to a privileged port but also to bind to a specific local address. The client code is used by the server for callbacks.

Quoted 5 lines

Quoted 1 line
The third patch is the actual NLM fix.
I've appended them to the NFS_ALL and the git tree. Unless Neil objects, I propose that we just shove them in together with the existing NFS merge.

No objection from me - I support your proposal!

Thanks,
NeilBrown