Re: [PATCH 3/3] nfsd4: implement secinfo_no_name

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

On Thu, Jan 06, 2011 at 11:54:02AM +0800, Mi Jinlong wrote:
> 
> 
> J. Bruce Fields 写道:
> > On Wed, Jan 05, 2011 at 10:37:13AM -0500, J. Bruce Fields wrote:
> >> On Wed, Jan 05, 2011 at 10:15:11AM -0500, Fred Isaman wrote:
> >>> On Tue, Jan 4, 2011 at 8:05 PM, J. Bruce Fields <bfields@fieldses.org> wrote:
> >>>> On Thu, Dec 30, 2010 at 12:13:27PM +0800, Mi Jinlong wrote:
> >>>>>
> >>>>> J. Bruce Fields:
> >>>>>> On Mon, Dec 27, 2010 at 02:29:57PM +0800, Mi Jinlong wrote:
> >>>>>>> When testing this patch, oops appears.
> >>>>>>>
> >>>>>>> We should implement a nfsd4_encode_secinfo_no_name() instead using
> >>>>>>> nfsd4_encode_secinfo().
> >>>>>>>
> >>>>>>> With the following patch, kernel will run correctly.
> >>>>>> Whoops, yes, you're correct.  I've applied your patch.  Thanks!
> >>>>>>
> >>>>>> (What are you using for testing?)
> >>>>> Hi Bruce:
> >>>>>
> >>>>>   I test it by the pynfs41 at your tree, and add some simple
> >>>>>   test case as following.
> >>>>>
> >>>>>   I have a question about the op.secinfo_no_name(), should it take
> >>>>>   a argument?
> >>>>>   Right now, maybe it always request with NFS4_SECINFO_STYLE4_CURRENT_FH,
> >>>>>   or I can't find how to set the argument.
> >>> There was a bug in how the op.secinfo_no_name() function was auto
> >>> constructed.  I've pushed a fix to
> >>> my tree at git://linux-nfs.org/~iisaman/newpynfs.git.
> >> Thanks, Fred.  I've pulled that into my repository too.
> > 
> > But then these secinfo_no_name tests don't work ("TypeError:
> > data.opsecinfo_no_name == None"); Mi Jinlong, would you mind looking
> > into that?
> 
>   After Fred's patch, we should rebuild the test site at pynfs41 dir 
>   with "python setup.py build".
> 
>   And update test case, call op.secinfo_no_name() with 0 as following.

Thanks!

Could I ask for a few changes?:
	- The second half of testSupported2 should go into a separate
	  test in a separate st_secinfo file.  (It will be a very small
	  file for now!  But we can port some of the 4.0 tests to it
	  some day.)
	- That test should create a file in the same way that e.g. the
	  st_open tests do instead of assuming a file named "/tree"
	  exists.

--b.

> 
> thanks,
> Mi Jinlong
> 
> >From 344f0ee9a625d43b918be5dfc4fc9df8e8cbac28 Mon Sep 17 00:00:00 2001
> From: Mi Jinlong <mijinlong@cn.fujitsu.com>
> Date: Thu, 30 Dec 2010 12:13:27 +0800
> Subject: [PATCH] CLNT: Add some simple secinfo_no_name tests
> 
> Add two simple secinfo_no_name tests as Bruce said:
>   - send PUTROOTFH+SECINFO_NO_NAME, check that you get back a
>     legal result.
>   - send PUTROOTFH+SECINFO+GETFH, and/or
>     PUTROOTFH+SECINFO_NO_NAME+GETFH, check that the GETFH returns
>     NOFILEHANDLE.
> 
> Signed-off-by: Mi Jinlong <mijinlong@cn.fujitsu.com>
> ---
>  nfs4.1/server41tests/__init__.py           |    1 +
>  nfs4.1/server41tests/st_secinfo_no_name.py |   35 ++++++++++++++++++++++++++++
>  2 files changed, 36 insertions(+), 0 deletions(-)
>  create mode 100644 nfs4.1/server41tests/st_secinfo_no_name.py
> 
> diff --git a/nfs4.1/server41tests/__init__.py b/nfs4.1/server41tests/__init__.py
> index 22cd664..6aee88e 100644
> --- a/nfs4.1/server41tests/__init__.py
> +++ b/nfs4.1/server41tests/__init__.py
> @@ -2,6 +2,7 @@ __all__ = ["st_exchange_id.py", # draft 21
>             "st_compound.py",
>             "st_create_session.py",
>             "st_destroy_session.py",
> +           "st_secinfo_no_name.py",
>             "st_sequence.py",
>  	   "st_trunking.py",
>             "st_open.py",
> diff --git a/nfs4.1/server41tests/st_secinfo_no_name.py b/nfs4.1/server41tests/st_secinfo_no_name.py
> new file mode 100644
> index 0000000..b66fdb6
> --- /dev/null
> +++ b/nfs4.1/server41tests/st_secinfo_no_name.py
> @@ -0,0 +1,35 @@
> +from st_create_session import create_session
> +from nfs4_const import *
> +from environment import check, fail, bad_sessionid, create_file
> +from nfs4_type import channel_attrs4
> +import nfs4_ops as op
> +import nfs4lib
> +
> +def testSupported(t, env):
> +    """Do a simple SECINFO_NO_NAME
> +       send PUTROOTFH+SECINFO_NO_NAME, check is result legal
> +
> +    FLAGS: all
> +    CODE: SECNN1
> +    """
> +    c = env.c1.new_client(env.testname(t))
> +    sess = c.create_session()
> +    res = sess.compound([op.putrootfh(), op.secinfo_no_name(0)])
> +    check(res);
> +
> +def testSupported2(t, env):
> +    """GETFH after do a SECINFO_NO_NAME or SECINFO
> +       result in a NOFILEHANDLE error, See rfc 5661 section 2.6.3.1.1.8
> +
> +    FLAGS: all
> +    CODE: SECNN2
> +    """
> +    c = env.c1.new_client(env.testname(t))
> +    sess = c.create_session()
> +    res = sess.compound([op.putrootfh(), op.secinfo_no_name(0), op.getfh()])
> +    print res
> +    check(res, NFS4ERR_NOFILEHANDLE);
> +
> +    res = sess.compound([op.putrootfh(), op.secinfo("tree"), op.getfh()])
> +    print res
> +    check(res, NFS4ERR_NOFILEHANDLE);
> -- 
> 1.7.3.4
>