2017-01-09 15:25:54

by Mkrtchyan, Tigran

Subject: [PATCH] setattr: test how v4.0 server handles setting of zero length principals

my interpretation of rfc3530 expects NFS4ERR_INVAL is owner and
owner_group are empty strings.

Signed-off-by: Tigran Mkrtchyan <[email protected]>
---
nfs4.0/servertests/st_setattr.py | 31 +++++++++++++++++++++++++++++++
1 file changed, 31 insertions(+)

diff --git a/nfs4.0/servertests/st_setattr.py b/nfs4.0/servertests/st_setattr.py
index b6a398b..b192ece 100644
--- a/nfs4.0/servertests/st_setattr.py
+++ b/nfs4.0/servertests/st_setattr.py
@@ -701,3 +701,34 @@ def testChangeGranularity(t, env):
chattr4 = res.resarray[7].obj_attributes
if chattr1 == chattr2 or chattr2 == chattr3 or chattr3 == chattr4:
t.fail("consecutive SETATTR(mode)'s don't all change change attribute")
+
+def testEmptyPrincipal(t, env):
+ """Setting owner with zero length principal must fail
+
+ FLAGS: setattr all
+ DEPEND: MKFILE
+ CODE: SATT16
+ """
+ c = env.c1
+ path = c.homedir + [t.code]
+ res = c.create_obj(path, NF4SOCK)
+ check(res)
+ ops = c.use_obj(path) + [c.setattr({FATTR4_OWNER: ''})]
+ res = c.compound(ops)
+ check(res, NFS4ERR_INVAL, "Setting empty owner")
+
+
+def testEmptyGroupPrincipal(t, env):
+ """Setting owner group with zero length principal must fail
+
+ FLAGS: setattr all
+ DEPEND: MKFILE
+ CODE: SATT17
+ """
+ c = env.c1
+ path = c.homedir + [t.code]
+ res = c.create_obj(path, NF4SOCK)
+ check(res)
+ ops = c.use_obj(path) + [c.setattr({FATTR4_OWNER_GROUP: ''})]
+ res = c.compound(ops)
+ check(res, NFS4ERR_INVAL, "Setting empty owner_group")
--
2.9.3

2017-01-13 17:03:36

by J. Bruce Fields

Subject: Re: [PATCH] setattr: test how v4.0 server handles setting of zero length principals

On Mon, Jan 09, 2017 at 04:15:56PM +0100, Tigran Mkrtchyan wrote:
> my interpretation of rfc3530 expects NFS4ERR_INVAL is owner and
> owner_group are empty strings.

Thanks. This hangs the Linux server. Possibly an idmapd bug, but I
think nfsd should be checking for this case before it calls idmapd.

Anyway, that's not your problem.

Any reason you're using NF4SOCK instead of NF4REG?

--b.

>
> Signed-off-by: Tigran Mkrtchyan <[email protected]>
> ---
> nfs4.0/servertests/st_setattr.py | 31 +++++++++++++++++++++++++++++++
> 1 file changed, 31 insertions(+)
>
> diff --git a/nfs4.0/servertests/st_setattr.py b/nfs4.0/servertests/st_setattr.py
> index b6a398b..b192ece 100644
> --- a/nfs4.0/servertests/st_setattr.py
> +++ b/nfs4.0/servertests/st_setattr.py
> @@ -701,3 +701,34 @@ def testChangeGranularity(t, env):
> chattr4 = res.resarray[7].obj_attributes
> if chattr1 == chattr2 or chattr2 == chattr3 or chattr3 == chattr4:
> t.fail("consecutive SETATTR(mode)'s don't all change change attribute")
> +
> +def testEmptyPrincipal(t, env):
> + """Setting owner with zero length principal must fail
> +
> + FLAGS: setattr all
> + DEPEND: MKFILE
> + CODE: SATT16
> + """
> + c = env.c1
> + path = c.homedir + [t.code]
> + res = c.create_obj(path, NF4SOCK)
> + check(res)
> + ops = c.use_obj(path) + [c.setattr({FATTR4_OWNER: ''})]
> + res = c.compound(ops)
> + check(res, NFS4ERR_INVAL, "Setting empty owner")
> +
> +
> +def testEmptyGroupPrincipal(t, env):
> + """Setting owner group with zero length principal must fail
> +
> + FLAGS: setattr all
> + DEPEND: MKFILE
> + CODE: SATT17
> + """
> + c = env.c1
> + path = c.homedir + [t.code]
> + res = c.create_obj(path, NF4SOCK)
> + check(res)
> + ops = c.use_obj(path) + [c.setattr({FATTR4_OWNER_GROUP: ''})]
> + res = c.compound(ops)
> + check(res, NFS4ERR_INVAL, "Setting empty owner_group")
> --
> 2.9.3