Re: [PATCH 00/19] Proposed server-side patches for 2.6.31

["J. Bruce Fields" <bfields@fieldses.org>]

On Thu, Apr 23, 2009 at 07:31:16PM -0400, Chuck Lever wrote:
> Hi Bruce -
> 
> Here are some patches I'd like to propose for 2.6.31.  Please review
> and consider them for linux-next.  They've seen a little testing here
> on x86_64, but they should get some public exercise.

I'm a little behind, apologies for any delay--I'll try to give these a
good reading soon.  Meanwhile--could you put these in a public git tree
and give me the url?  That'll make it trivial to pull them into -next,
at least.

--b.

> 
> The first 17 are updated versions of patches I submitted for 2.6.30,
> but were dropped at the last minute to simplify things.  They provide
> some clean-up for the nfsctl API, and a little more IPv6 support in
> that area as well, in preparation for full-on IPv6-related changes to
> come.  I think I've addressed earlier review comments.
> 
> As usual, it would be most helpful if you would drop the whole series
> if you have issues with any of these patches.
> 
> Patch 18 addresses a long-standing problem with our NLM/NSM
> implementation.  If lockd.ko isn't loaded before statd starts up
> (which is the typical case for both our server and client) then
> lockd's NSM state number is never set.  We then end up sending a zero
> state number in all NLM requests.  This is also the case if lockd.ko
> is unloaded and reloaded during normal operation -- the NSM state
> number is reset to zero.  See the patch description for further
> detail.
> 
> Patch 19 provides a little more 64-bit alignment clean up in
> nsm_init_private() .
> 
> These two are independent of all the others and either or both can be
> applied as you see fit.
> 
> ---
> 
> Chuck Lever (19):
>       lockd: clean up 64-bit alignment fix in nsm_init_private()
>       lockd: Update NSM state from SM_MON replies
>       NFSD: Stricter buffer size checking in fs/nfsd/nfsctl.c
>       NFSD: Stricter buffer size checking in write_versions()
>       NFSD: Stricter buffer size checking in write_recoverydir()
>       SUNRPC: Clean up one_sock_name()
>       SUNRPC: Support PF_INET6 in one_sock_name()
>       SUNRPC: Switch one_sock_name() to use snprintf()
>       SUNRPC: pass buffer size to svc_sock_names()
>       SUNRPC: pass buffer size to svc_addsock()
>       NFSD: Prevent a buffer overflow in svc_xprt_names()
>       NFSD: move lockd_up() before svc_addsock()
>       NFSD: Finish refactoring __write_ports()
>       NFSD: Note an additional requirement when passing TCP sockets to portlist
>       NFSD: Refactor socket creation out of __write_ports()
>       NFSD: Refactor portlist socket closing into a helper
>       NFSD: Refactor transport addition out of __write_ports()
>       NFSD: Refactor transport removal out of __write_ports()
>       SUNRPC: Fix error return value of svc_addr_len()
> 
> 
>  fs/lockd/clntproc.c             |    2 
>  fs/lockd/mon.c                  |   18 +-
>  fs/nfsd/nfsctl.c                |  284 ++++++++++++++++++++++++---------------
>  include/linux/sunrpc/svc_xprt.h |    7 +
>  include/linux/sunrpc/svcsock.h  |    7 +
>  net/sunrpc/svc_xprt.c           |   56 +++++---
>  net/sunrpc/svcsock.c            |   87 +++++++++---
>  7 files changed, 299 insertions(+), 162 deletions(-)
> 
> -- 
> Chuck Lever <chuck.lever@oracle.com>