register NFS_ACL with rpcbind
Subsystems (2)
PATCH 1/1 register NFS_ACL with rpcbind
Hi.
Here is a patch to modify the NFS server to register the NFS_ACL services with the rpcbind daemon. This allows the client to ping for the existence of the NFS_ACL support via commands such as "rpcinfo -t <server> nfs_acl".
The changelog for the patch which turned off this functionality mentioned something about not registering the NFS_ACL as being part of some tradition. I can't find this tradition and the only other implementation which supports NFS_ACL does register them with the rpcbind daemon.
Thanx... ps
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +1 -1
@@ -346,5 +346,5 @@ struct svc_version nfsd_acl_version2 = { .vs_proc = nfsd_acl_procedures2, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
fs/nfsd/nfs3acl.c renamed from fs/nfsd/nfs3acl.c.org +1 -1
@@ -264,6 +264,6 @@ struct svc_version nfsd_acl_version3 = { .vs_proc = nfsd_acl_procedures3, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
PATCH v2 1/1 register NFS_ACL with rpcbind
Hi.
Here is a patch to modify the NFS server to register the NFS_ACL services with the rpcbind daemon. This allows the client to ping for the existence of the NFS_ACL support via commands such as "rpcinfo -t <server> nfs_acl".
This patch also modifies the NFS_ACL support so that responses to version 2 NULLPROC requests can be made.
The changelog for the patch which turned off this functionality mentioned something about not registering the NFS_ACL as being part of some tradition. I can't find this tradition and the only other implementation which supports NFS_ACL does register them with the rpcbind daemon.
Thanx... ps
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +11 -2
@@ -217,6 +217,16 @@ static int nfsaclsvc_decode_accessargs(s* XDR encode functions */ +/* +* There must be an encoding function for void results so svc_process +* will work properly. +*/ +int +nfsaclsvc_encode_voidres (struct svc_rqst *rqstp,__be32 *p,void *dummy) +{ +return xdr_ressize_check (rqstp, p); +} +/* GETACL */ static int nfsaclsvc_encode_getaclres (struct svc_rqst *rqstp,__be32 *p,struct nfsd3_getaclres *resp)
@@ -308,7 +318,6 @@ static int nfsaclsvc_release_access(stru }#define nfsaclsvc_decode_voidargs NULL -#define nfsaclsvc_encode_voidres NULL#define nfsaclsvc_release_void NULL#define nfsd3_fhandleargs nfsd_fhandle#define nfsd3_attrstatres nfsd_attrstat
@@ -346,5 +355,5 @@ struct svc_version nfsd_acl_version2 = { .vs_proc = nfsd_acl_procedures2, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
fs/nfsd/nfs3acl.c renamed from fs/nfsd/nfs3acl.c.org +1 -1
@@ -264,6 +264,6 @@ struct svc_version nfsd_acl_version3 = { .vs_proc = nfsd_acl_procedures3, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
Quoted 7 lines · Peter Staubach
Hi.
Here is a patch to modify the NFS server to register the
NFS_ACLservices with the rpcbind daemon. This allows the client to ping for the existence of theNFS_ACLsupport via commands such as "rpcinfo -t <server>nfs_acl".
Quoted 8 lines
This patch also modifies the
NFS_ACLsupport so that responses to version 2 NULLPROC requests can be made.The changelog for the patch which turned off this functionality mentioned something about not registering the
NFS_ACLas being part of some tradition. I can't find this tradition and the only other implementation which supportsNFS_ACLdoes register them with the rpcbind daemon.
I don't understand the reasoning behind .vs_hidden for NFS_ACL, hopefully Olaf can clarify. NFS_ACL is the only user of .vs_hidden as far as I can see though, so if this is changeg, shouldn't the entire commit bc5fea4 which introduced the flag be reverted?
Thanks,
Andreas
Quoted 1 line · On Tuesday 03 November 2009 10:13:27 Andreas Gruenbacher wrote:
I don't understand the reasoning behind .vs_hidden for
NFS_ACL, hopef=
ully
Quoted 1 line
Olaf can clarify.
NFS_ACLis the only user of .vs_hidden as far as I =
can
Quoted 1 line
see though, so if this is changeg, shouldn't the entire commit bc5fea=
4
Quoted 1 line
which introduced the flag be reverted?
I can't remember the details of that one. I do remember that this is based on someone's request who told me that we shouldn't register nfsac= l
with portmap. I didn't check myself whether Solaris did or did not do it at that time.
I have no issue with reverting that change, and removing the whole =2Evs_hidden kludge too.
Thanks,
Olaf
--=20
Neo didn't bring down the Matrix. SOA did. (soafacts.com)
--------------------------------------------
Olaf Kirch - Director Server ([email protected]) SUSE LINUX Products GmbH, Maxfeldstr. 5, D-90409 N=FCrnberg
GF: Markus Rex, HRB 16746 (AG N=FCrnberg)
Quoted 6 lines · Olaf Kirch
Quoted 4 lines · On Tuesday 03 November 2009 10:13:27 Andreas Gruenbacher wrote:
I don't understand the reasoning behind .vs_hidden for
NFS_ACL, hopefully Olaf can clarify.NFS_ACLis the only user of .vs_hidden as far as I can see though, so if this is changeg, shouldn't the entire commit bc5fea4 which introduced the flag be reverted?
Quoted 8 lines
I can't remember the details of that one. I do remember that this is based on someone's request who told me that we shouldn't register nfsacl with portmap. I didn't check myself whether Solaris did or did not do it at that time.
I have no issue with reverting that change, and removing the whole .vs_hidden kludge too.
It seems that vs_hidden is used in 1 place outside of the NFS_ACL server code. It is used in the NFSv4 callback code.
I will look to see how difficult that might be to fix this spot as well and then get rid of vs_hidden.
Thanx... ps
Quoted 16 lines · Peter Staubach
Quoted 6 lines · Olaf Kirch
Quoted 4 lines · On Tuesday 03 November 2009 10:13:27 Andreas Gruenbacher wrote:
I don't understand the reasoning behind .vs_hidden for
NFS_ACL, hopefully Olaf can clarify.NFS_ACLis the only user of .vs_hidden as far as I can see though, so if this is changeg, shouldn't the entire commit bc5fea4 which introduced the flag be reverted?Quoted 8 lines
I can't remember the details of that one. I do remember that this is based on someone's request who told me that we shouldn't register nfsacl with portmap. I didn't check myself whether Solaris did or did not do it at that time.
I have no issue with reverting that change, and removing the whole .vs_hidden kludge too.
Quoted 5 lines
It seems that
vs_hiddenis used in 1 place outside of theNFS_ACLserver code. It is used in the NFSv4 callback code.I will look to see how difficult that might be to fix this spot as well and then get rid of
vs_hidden.
See archive of this mailing list from earlier in October. This change was added because it's hard to get rid of the svc_unregister() call done by svc_create().
I have another solution for that problem that I'm preparing for 2.6.33.
--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com
Quoted 22 lines · Chuck Lever
Quoted 21 lines · Peter Staubach
Quoted 8 lines · Olaf Kirch
Quoted 6 lines · On Tuesday 03 November 2009 10:13:27 Andreas Gruenbacher wrote:
I don't understand the reasoning behind .vs_hidden for
NFS_ACL, hopefully
Olaf can clarify.NFS_ACLis the only user of .vs_hidden as far as I can
see though, so if this is changeg, shouldn't the entire commit bc5fea4 which introduced the flag be reverted?Quoted 8 lines
I can't remember the details of that one. I do remember that this is based on someone's request who told me that we shouldn't register nfsacl with portmap. I didn't check myself whether Solaris did or did not do it at that time.
I have no issue with reverting that change, and removing the whole .vs_hidden kludge too.
It seems that
vs_hiddenis used in 1 place outside of theNFS_ACLserver code. It is used in the NFSv4 callback code.
Quoted 9 lines
Quoted 2 lines
I will look to see how difficult that might be to fix this spot as well and then get rid of
vs_hidden.See archive of this mailing list from earlier in October. This change was added because it's hard to get rid of the
svc_unregister()call done bysvc_create().I have another solution for that problem that I'm preparing for 2.6.33.
Cool.
In the meantime, can we get this one in, Bruce?
Thanx... ps
Quoted 27 lines · Peter Staubach
Hi.
Here is a patch to modify the NFS server to register the
NFS_ACLservices with the rpcbind daemon. This allows the client to ping for the existence of theNFS_ACLsupport via commands such as "rpcinfo -t <server>nfs_acl".This patch also modifies the
NFS_ACLsupport so that responses to version 2 NULLPROC requests can be made.The changelog for the patch which turned off this functionality mentioned something about not registering the
NFS_ACLas being part of some tradition. I can't find this tradition and the only other implementation which supportsNFS_ACLdoes register them with the rpcbind daemon.Thanx... psSigned-off-by Peter Staubach <[email protected]>fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +0
-0@@ -217,6 +217,16 @@ static int nfsaclsvc_decode_accessargs(s* XDR encode functions */
fs/nfsd/nfs2acl.c static int nfsaclsvc_decode_accessargs(s · line 228
Quoted 10 lines
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +9
-0+/* +* There must be an encoding function for void results so svc_process +* will work properly. +*/ +int +nfsaclsvc_encode_voidres (struct svc_rqst *rqstp,__be32 *p,void *dummy) +{ +return xdr_ressize_check (rqstp, p); +}
Out of curiosity: have you tested a null rpc?
Also, doesn't v3 have the same problem?
--b.
Quoted 29 lines
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +2
-2+/* GETACL */ static int nfsaclsvc_encode_getaclres (struct svc_rqst *rqstp,__be32 *p,struct nfsd3_getaclres *resp)@@ -308,7 +318,6 @@ static int nfsaclsvc_release_access(stru }#define nfsaclsvc_decode_voidargs NULL -#define nfsaclsvc_encode_voidres NULL#define nfsaclsvc_release_void NULL#define nfsd3_fhandleargs nfsd_fhandle#define nfsd3_attrstatres nfsd_attrstat@@ -346,5 +355,5 @@ struct svc_version nfsd_acl_version2 = { .vs_proc = nfsd_acl_procedures2, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };fs/nfsd/nfs3acl.c renamed from fs/nfsd/nfs3acl.c.org +1
-1@@ -264,6 +264,6 @@ struct svc_version nfsd_acl_version3 = { .vs_proc = nfsd_acl_procedures3, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
Quoted 30 lines · Peter Staubach
Quoted 29 lines · Chuck Lever
Quoted 18 lines · Peter Staubach
Quoted 8 lines · Olaf Kirch
Quoted 6 lines · On Tuesday 03 November 2009 10:13:27 Andreas Gruenbacher wrote:
I don't understand the reasoning behind .vs_hidden for
NFS_ACL, hopefully
Olaf can clarify.NFS_ACLis the only user of .vs_hidden as far as I can
see though, so if this is changeg, shouldn't the entire commit bc5fea4 which introduced the flag be reverted?Quoted 8 lines
I can't remember the details of that one. I do remember that this is based on someone's request who told me that we shouldn't register nfsacl with portmap. I didn't check myself whether Solaris did or did not do it at that time.
I have no issue with reverting that change, and removing the whole .vs_hidden kludge too.
Quoted 5 lines
It seems that
vs_hiddenis used in 1 place outside of theNFS_ACLserver code. It is used in the NFSv4 callback code.I will look to see how difficult that might be to fix this spot as well and then get rid of
vs_hidden.See archive of this mailing list from earlier in October. This change was added because it's hard to get rid of the
svc_unregister()call done bysvc_create().
Quoted 6 lines
Quoted 2 lines
I have another solution for that problem that I'm preparing for 2.6.33.
Cool.
In the meantime, can we get this one in, Bruce?
As far as I know, there is no "meantime" in this case. 2.6.33 is the next merge window.
I don't have a problem with getting rid of .vs_hidden anyway... all I'm saying is maybe you don't have to work too hard at it. :-)
--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com
Quoted 29 lines · J. Bruce Fields
Quoted 28 lines · Peter Staubach
Hi.
Here is a patch to modify the NFS server to register the
NFS_ACLservices with the rpcbind daemon. This allows the client to ping for the existence of theNFS_ACLsupport via commands such as "rpcinfo -t <server>nfs_acl".This patch also modifies the
NFS_ACLsupport so that responses to version 2 NULLPROC requests can be made.The changelog for the patch which turned off this functionality mentioned something about not registering the
NFS_ACLas being part of some tradition. I can't find this tradition and the only other implementation which supportsNFS_ACLdoes register them with the rpcbind daemon.Thanx... psSigned-off-by Peter Staubach <[email protected]>fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +0
-0@@ -217,6 +217,16 @@ static int nfsaclsvc_decode_accessargs(s* XDR encode functions */
Quoted 12 lines
fs/nfsd/nfs2acl.c static int nfsaclsvc_decode_accessargs(s · line 228
Quoted 9 lines
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +9
-0+/* +* There must be an encoding function for void results so svc_process +* will work properly. +*/ +int +nfsaclsvc_encode_voidres (struct svc_rqst *rqstp,__be32 *p,void *dummy) +{ +return xdr_ressize_check (rqstp, p); +}Out of curiosity: have you tested a null rpc?
Yup. "rpcinfo -t <server> nfs_acl" works like a charm with this patch applied for both versions 2 and 3.
Quoted 2 lines
Also, doesn't v3 have the same problem?
Nope. It already had a similar XDR routine. Which is similar to the NFS XDR routine to do the same sort of thing. I suppose that someday, we could clean them up, but these routines aren't doing any harm for the moment.
Thanx... ps
Quoted 31 lines
--b.
Quoted 29 lines
fs/nfsd/nfs2acl.c renamed from fs/nfsd/nfs2acl.c.org +2
-2+/* GETACL */ static int nfsaclsvc_encode_getaclres (struct svc_rqst *rqstp,__be32 *p,struct nfsd3_getaclres *resp)@@ -308,7 +318,6 @@ static int nfsaclsvc_release_access(stru }#define nfsaclsvc_decode_voidargs NULL -#define nfsaclsvc_encode_voidres NULL#define nfsaclsvc_release_void NULL#define nfsd3_fhandleargs nfsd_fhandle#define nfsd3_attrstatres nfsd_attrstat@@ -346,5 +355,5 @@ struct svc_version nfsd_acl_version2 = { .vs_proc = nfsd_acl_procedures2, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };fs/nfsd/nfs3acl.c renamed from fs/nfsd/nfs3acl.c.org +1
-1@@ -264,6 +264,6 @@ struct svc_version nfsd_acl_version3 = { .vs_proc = nfsd_acl_procedures3, .vs_dispatch = nfsd_dispatch, .vs_xdrsize = NFS3_SVC_XDRSIZE, - .vs_hidden = 1, + .vs_hidden = 0, };
Quoted 7 lines · Peter Staubach
Quoted 2 lines · J. Bruce Fields
Out of curiosity: have you tested a null rpc?
Yup. "rpcinfo -t <server>
nfs_acl" works like a charm with this patch applied for both versions 2 and 3.
Quoted 6 lines
Quoted 2 lines
Also, doesn't v3 have the same problem?
Nope. It already had a similar XDR routine. Which is similar to the NFS XDR routine to do the same sort of thing. I suppose that someday, we could clean them up,
Yes.
Quoted 1 line
but these routines aren't doing any harm for the moment.
OK. Applied.
--b.