The Server Side Transport switch implements a pluggable transport
provider layer for RPC servers. This allows network transport drivers
to be dynamically loaded and registered, thereby extending the set of
transports that can carry RPC. This patchset includes a transport
provider driver for TCP and UDP sockets.
This version of the patchset is functionally equivalent to the
preceding patchset and includes the following improvements:
- Fixed a bug in svc_create_socket that resulted in svc_xprt_received
being called without holding the XPT_BUSY bit.
- Added a BUG_ON in svc_xprt_received to catch providers calling
svc_xprt_received without holding the XPT_BUSY bit.
- Removed code setting the XPT_BUSY bit in the create path and moved
this logic to common code.
- Cleaned up the SMP locking strategy comment and moved it to
svc_xprt.c
- Changed return type of svc_unreg_xprt_class to void
- Simplified transport class removal logic
- Fixed a white-space abuse issue
- Made some style changes to the patch comments
- Created a separate patch that moves the connection limit code to
generic logic
- Created a separate patch that moves the call to svc_xprt_received
out of tcp_accept and into common logic
- Created a separate patch for removal of unnecessary call to
svc_sock_enqueue in svc_tcp_accept function
- Removed silly BUG_ON from kref svc_xprt_free callback function
- Moved unlock of svc_xprt_class list to after the try_module_lock
call.
- Refactored svc_create_xprt logic
- Removed unnecessary call to svc_find_xprt in portlist code
--
Signed-off-by: Tom Tucker <[email protected]>
On Sun, Dec 30, 2007 at 09:07:12PM -0600, Tom Tucker wrote:
>
> The Server Side Transport switch implements a pluggable transport
> provider layer for RPC servers. This allows network transport drivers
> to be dynamically loaded and registered, thereby extending the set of
> transports that can carry RPC. This patchset includes a transport
> provider driver for TCP and UDP sockets.
>
>
> This version of the patchset is functionally equivalent to the
> preceding patchset and includes the following improvements:
Thanks! Got it.
--b.
>
> - Fixed a bug in svc_create_socket that resulted in svc_xprt_received
> being called without holding the XPT_BUSY bit.
>
> - Added a BUG_ON in svc_xprt_received to catch providers calling
> svc_xprt_received without holding the XPT_BUSY bit.
>
> - Removed code setting the XPT_BUSY bit in the create path and moved
> this logic to common code.
>
> - Cleaned up the SMP locking strategy comment and moved it to
> svc_xprt.c
>
> - Changed return type of svc_unreg_xprt_class to void
>
> - Simplified transport class removal logic
>
> - Fixed a white-space abuse issue
>
> - Made some style changes to the patch comments
>
> - Created a separate patch that moves the connection limit code to
> generic logic
>
> - Created a separate patch that moves the call to svc_xprt_received
> out of tcp_accept and into common logic
>
> - Created a separate patch for removal of unnecessary call to
> svc_sock_enqueue in svc_tcp_accept function
>
> - Removed silly BUG_ON from kref svc_xprt_free callback function
>
> - Moved unlock of svc_xprt_class list to after the try_module_lock
> call.
>
> - Refactored svc_create_xprt logic
>
> - Removed unnecessary call to svc_find_xprt in portlist code
>
> --
> Signed-off-by: Tom Tucker <[email protected]>