[PATCH 000 of 14] knfsd: Introduction

[PATCH 000 of 14] knfsd: Introduction

[NeilBrown]

Following are 14 assorted patches for nfsd.
Patches are against 2.6.17-mm2 and are suitable for inclusion in 2.6.18.

Primarily minor bug fixes.

Last three patches are initial support for RPC privacy (i.e. decrypt
the RPC request and encrypt the reply).  This code is sub-optimal in
several ways (memcopys ...) which will probably get tidied up once we
determine the best way to do so.

Thanks,
NeilBrown

 [PATCH 001 of 14] knfsd: Improve the test for cross-device-rename in nfsd
 [PATCH 002 of 14] knfsd: Fixing missing 'expkey' support for fsid type 3
 [PATCH 003 of 14] knfsd: Remove noise about filehandle being uptodate.
 [PATCH 004 of 14] knfsd: Ignore ref_fh when crossing a mountpoint.
 [PATCH 005 of 14] knfsd: nfsd4: fix open_confirm locking
 [PATCH 006 of 14] knfsd: nfsd: call nfsd_setuser() on fh_compose(), fix nfsd4 permissions problem
 [PATCH 007 of 14] knfsd: nfsd4: remove superfluous grace period checks
 [PATCH 008 of 14] knfsd: nfsd: fix misplaced fh_unlock() in nfsd_link()
 [PATCH 009 of 14] knfsd: svcrpc: gss: simplify rsc_parse()
 [PATCH 010 of 14] knfsd: nfsd4: fix some open argument tests
 [PATCH 011 of 14] knfsd: nfsd4: fix open flag passing
 [PATCH 012 of 14] knfsd: svcrpc: Simplify nfsd rpcsec_gss integrity code
 [PATCH 013 of 14] knfsd: nfsd: mark rqstp to prevent use of sendfile in privacy case
 [PATCH 014 of 14] knfsd: svcrpc: gss: server-side implementation of rpcsec_gss privacy.

[PATCH 001 of 14] knfsd: Improve the test for cross-device-rename in nfsd

[NeilBrown]

Just testing the i_sb isn't really enough, at least
the vfsmnt must be the same.  Thanks Al.

Cc: Al Viro <viro@ftp.linux.org.uk>

### Diffstat output
 ./fs/nfsd/vfs.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff .prev/fs/nfsd/vfs.c ./fs/nfsd/vfs.c
--- .prev/fs/nfsd/vfs.c	2006-06-27 12:15:18.000000000 +1000
+++ ./fs/nfsd/vfs.c	2006-06-27 12:15:21.000000000 +1000
@@ -1556,7 +1556,7 @@ nfsd_rename(struct svc_rqst *rqstp, stru
 	tdir = tdentry->d_inode;
 
 	err = (rqstp->rq_vers == 2) ? nfserr_acces : nfserr_xdev;
-	if (fdir->i_sb != tdir->i_sb)
+	if (ffhp->fh_export != tfhp->fh_export)
 		goto out;
 
 	err = nfserr_perm;

[PATCH 002 of 14] knfsd: Fixing missing 'expkey' support for fsid type 3

[NeilBrown]

From: Frank Filz <ffilzlnx@us.ibm.com>

Type '3' is used for the fsid in filehandles when the device number
of the device holding the filesystem has more than 8 bits in either
major or minor.
Unfortunately expkey_parse doesn't recognise type 3.  Fix this.

(Slighty modified from Frank's original)

Signed-Off-By: Frank Filz <ffilzlnx@us.ibm.com>
Signed-off-by: Neil Brown <neilb@suse.de>

### Diffstat output
 ./fs/nfsd/export.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff .prev/fs/nfsd/export.c ./fs/nfsd/export.c
--- .prev/fs/nfsd/export.c	2006-06-27 12:15:18.000000000 +1000
+++ ./fs/nfsd/export.c	2006-06-27 12:15:55.000000000 +1000
@@ -126,7 +126,7 @@ static int expkey_parse(struct cache_det
 	if (*ep)
 		goto out;
 	dprintk("found fsidtype %d\n", fsidtype);
-	if (fsidtype > 2)
+	if (key_len(fsidtype)==0) /* invalid type */
 		goto out;
 	if ((len=qword_get(&mesg, buf, PAGE_SIZE)) <= 0)
 		goto out;

[PATCH 003 of 14] knfsd: Remove noise about filehandle being uptodate.

[NeilBrown]

There is a perfectly valid situation where fh_update gets called on an
already uptodate filehandle - in nfsd_create_v3 where a
CREATE_UNCHECKED finds an existing file and wants to just set the
size.

We could possible optimise out the call in that case, but the only
harm involved is that fh_update prints a warning, so it is easier to
remove the warning.

Signed-off-by: Neil Brown <neilb@suse.de>

### Diffstat output
 ./fs/nfsd/nfsfh.c |    6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff .prev/fs/nfsd/nfsfh.c ./fs/nfsd/nfsfh.c
--- .prev/fs/nfsd/nfsfh.c	2006-06-27 12:15:18.000000000 +1000
+++ ./fs/nfsd/nfsfh.c	2006-06-27 12:16:05.000000000 +1000
@@ -461,7 +461,7 @@ fh_update(struct svc_fh *fhp)
 	} else {
 		int size;
 		if (fhp->fh_handle.fh_fileid_type != 0)
-			goto out_uptodate;
+			goto out;
 		datap = fhp->fh_handle.fh_auth+
 			fhp->fh_handle.fh_size/4 -1;
 		size = (fhp->fh_maxsize - fhp->fh_handle.fh_size)/4;
@@ -481,10 +481,6 @@ out_negative:
 	printk(KERN_ERR "fh_update: %s/%s still negative!\n",
 		dentry->d_parent->d_name.name, dentry->d_name.name);
 	goto out;
-out_uptodate:
-	printk(KERN_ERR "fh_update: %s/%s already up-to-date!\n",
-		dentry->d_parent->d_name.name, dentry->d_name.name);
-	goto out;
 }
 
 /*

[PATCH 004 of 14] knfsd: Ignore ref_fh when crossing a mountpoint.

[NeilBrown]

nfsd tries to return to a client the same sort of filehandle as was
used by the client.  This removes some filehandle aliasing issues and
means that a server upgrade followed by a downgrade will not confused
clients not restarted during that time.

However when crossing a mountpoint, the filehandle used for one
filesystem doesn't provide any useful information on what sort of
filehandle should be used on the other, and can provide misleading
information.  So if the reference filehandle is on a different
filesystem to the one being generated, ignore it.

Signed-off-by: Neil Brown <neilb@suse.de>

### Diffstat output
 ./fs/nfsd/nfsfh.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff .prev/fs/nfsd/nfsfh.c ./fs/nfsd/nfsfh.c
--- .prev/fs/nfsd/nfsfh.c	2006-06-27 12:16:05.000000000 +1000
+++ ./fs/nfsd/nfsfh.c	2006-06-27 12:16:27.000000000 +1000
@@ -312,8 +312,8 @@ int
 fh_compose(struct svc_fh *fhp, struct svc_export *exp, struct dentry *dentry, struct svc_fh *ref_fh)
 {
 	/* ref_fh is a reference file handle.
-	 * if it is non-null, then we should compose a filehandle which is
-	 * of the same version, where possible.
+	 * if it is non-null and for the same filesystem, then we should compose
+	 * a filehandle which is of the same version, where possible.
 	 * Currently, that means that if ref_fh->fh_handle.fh_version == 0xca
 	 * Then create a 32byte filehandle using nfs_fhbase_old
 	 *
@@ -332,7 +332,7 @@ fh_compose(struct svc_fh *fhp, struct sv
 		parent->d_name.name, dentry->d_name.name,
 		(inode ? inode->i_ino : 0));
 
-	if (ref_fh) {
+	if (ref_fh && ref_fh->fh_export == exp) {
 		ref_fh_version = ref_fh->fh_handle.fh_version;
 		if (ref_fh_version == 0xca)
 			ref_fh_fsid_type = 0;

[PATCH 005 of 14] knfsd: nfsd4: fix open_confirm locking

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>

Fix an improper unlock in an error path.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/nfs4state.c |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff .prev/fs/nfsd/nfs4state.c ./fs/nfsd/nfs4state.c
--- .prev/fs/nfsd/nfs4state.c	2006-06-27 14:36:03.000000000 +1000
+++ ./fs/nfsd/nfs4state.c	2006-06-27 14:36:03.000000000 +1000
@@ -2252,8 +2252,9 @@ nfsd4_open_confirm(struct svc_rqst *rqst
 			(int)current_fh->fh_dentry->d_name.len,
 			current_fh->fh_dentry->d_name.name);
 
-	if ((status = fh_verify(rqstp, current_fh, S_IFREG, 0)))
-		goto out;
+	status = fh_verify(rqstp, current_fh, S_IFREG, 0);
+	if (status)
+		return status;
 
 	nfs4_lock_state();
 

[PATCH 006 of 14] knfsd: nfsd: call nfsd_setuser() on fh_compose(), fix nfsd4 permissions problem

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


In the typical v2/v3 case the only new filehandles used as arguments to
operations are filehandles taken directly off the wire, which don't get
dentries until fh_verify() is called.

But in v4 the filehandles that are arguments to operations were often
created by previous operations (putrootfh, lookup, etc.) using
fh_compose, which sets the dentry in the filehandle without calling
nfsd_setuser().

This also means that, for example, if filesystem B is mounted on filesystem
A, and filesystem A is exported without root-squashing, then a client can
bypass the rootsquashing on B using a compound that starts at a filehandle
in A, crosses into B using lookups, and then does stuff in B.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/nfsfh.c |   15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff .prev/fs/nfsd/nfsfh.c ./fs/nfsd/nfsfh.c
--- .prev/fs/nfsd/nfsfh.c	2006-06-27 12:16:27.000000000 +1000
+++ ./fs/nfsd/nfsfh.c	2006-06-27 14:41:52.000000000 +1000
@@ -187,13 +187,6 @@ fh_verify(struct svc_rqst *rqstp, struct
 			goto out;
 		}
 
-		/* Set user creds for this exportpoint */
-		error = nfsd_setuser(rqstp, exp);
-		if (error) {
-			error = nfserrno(error);
-			goto out;
-		}
-
 		/*
 		 * Look up the dentry using the NFS file handle.
 		 */
@@ -251,6 +244,14 @@ fh_verify(struct svc_rqst *rqstp, struct
 	}
 	cache_get(&exp->h);
 
+	/* Set user creds for this exportpoint; necessary even in the "just
+	 * checking" case because this may be a filehandle that was created by
+	 * fh_compose, and that is about to be used in another nfsv4 compound
+	 * operation */
+	error = nfserrno(nfsd_setuser(rqstp, exp));
+	if (error)
+		goto out;
+
 	error = nfsd_mode_check(rqstp, dentry->d_inode->i_mode, type);
 	if (error)
 		goto out;

[PATCH 007 of 14] knfsd: nfsd4: remove superfluous grace period checks

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


We're checking nfs_in_grace here a few times when there isn't really any
reason to--bad_stateid is probably the more sensible return value anyway.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/nfs4state.c |    4 ----
 1 file changed, 4 deletions(-)

diff .prev/fs/nfsd/nfs4state.c ./fs/nfsd/nfs4state.c
--- .prev/fs/nfsd/nfs4state.c	2006-06-27 14:36:03.000000000 +1000
+++ ./fs/nfsd/nfs4state.c	2006-06-27 14:44:03.000000000 +1000
@@ -2070,16 +2070,12 @@ nfs4_preprocess_stateid_op(struct svc_fh
 	if (!stateid->si_fileid) { /* delegation stateid */
 		if(!(dp = find_delegation_stateid(ino, stateid))) {
 			dprintk("NFSD: delegation stateid not found\n");
-			if (nfs4_in_grace())
-				status = nfserr_grace;
 			goto out;
 		}
 		stidp = &dp->dl_stateid;
 	} else { /* open or lock stateid */
 		if (!(stp = find_stateid(stateid, flags))) {
 			dprintk("NFSD: open or lock stateid not found\n");
-			if (nfs4_in_grace())
-				status = nfserr_grace;
 			goto out;
 		}
 		if ((flags & CHECK_FH) && nfs4_check_fh(current_fh, stp))

[PATCH 008 of 14] knfsd: nfsd: fix misplaced fh_unlock() in nfsd_link()

[NeilBrown]

From: David M. Richter <richterd@citi.umich.edu>

In the event that lookup_one_len() fails in nfsd_link(), fh_unlock() is
skipped and locks are held overlong.

Patch was tested on 2.6.17-rc2 by causing lookup_one_len() to fail and
verifying that fh_unlock() gets called appropriately.

Signed-off-by: David M. Richter <richterd@citi.umich.edu>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/vfs.c |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff .prev/fs/nfsd/vfs.c ./fs/nfsd/vfs.c
--- .prev/fs/nfsd/vfs.c	2006-06-27 12:15:21.000000000 +1000
+++ ./fs/nfsd/vfs.c	2006-06-27 14:44:27.000000000 +1000
@@ -1520,14 +1520,15 @@ nfsd_link(struct svc_rqst *rqstp, struct
 			err = nfserrno(err);
 	}
 
-	fh_unlock(ffhp);
 	dput(dnew);
+out_unlock:
+	fh_unlock(ffhp);
 out:
 	return err;
 
 out_nfserr:
 	err = nfserrno(err);
-	goto out;
+	goto out_unlock;
 }
 
 /*

[PATCH 009 of 14] knfsd: svcrpc: gss: simplify rsc_parse()

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


Adopt a simpler convention for gss_mech_put(), to simplify rsc_parse().

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./net/sunrpc/auth_gss/gss_mech_switch.c |    6 +++---
 ./net/sunrpc/auth_gss/svcauth_gss.c     |   12 ++++--------
 2 files changed, 7 insertions(+), 11 deletions(-)

diff .prev/net/sunrpc/auth_gss/gss_mech_switch.c ./net/sunrpc/auth_gss/gss_mech_switch.c
--- .prev/net/sunrpc/auth_gss/gss_mech_switch.c	2006-06-27 14:59:44.000000000 +1000
+++ ./net/sunrpc/auth_gss/gss_mech_switch.c	2006-06-27 14:59:44.000000000 +1000
@@ -224,7 +224,8 @@ EXPORT_SYMBOL(gss_service_to_auth_domain
 void
 gss_mech_put(struct gss_api_mech * gm)
 {
-	module_put(gm->gm_owner);
+	if (gm)
+		module_put(gm->gm_owner);
 }
 
 EXPORT_SYMBOL(gss_mech_put);
@@ -307,8 +308,7 @@ gss_delete_sec_context(struct gss_ctx	**
 		(*context_handle)->mech_type->gm_ops
 			->gss_delete_sec_context((*context_handle)
 							->internal_ctx_id);
-	if ((*context_handle)->mech_type)
-		gss_mech_put((*context_handle)->mech_type);
+	gss_mech_put((*context_handle)->mech_type);
 	kfree(*context_handle);
 	*context_handle=NULL;
 	return GSS_S_COMPLETE;

diff .prev/net/sunrpc/auth_gss/svcauth_gss.c ./net/sunrpc/auth_gss/svcauth_gss.c
--- .prev/net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 14:59:44.000000000 +1000
+++ ./net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 14:59:44.000000000 +1000
@@ -425,6 +425,7 @@ static int rsc_parse(struct cache_detail
 	struct rsc rsci, *rscp = NULL;
 	time_t expiry;
 	int status = -EINVAL;
+	struct gss_api_mech *gm = NULL;
 
 	memset(&rsci, 0, sizeof(rsci));
 	/* context handle */
@@ -453,7 +454,6 @@ static int rsc_parse(struct cache_detail
 		set_bit(CACHE_NEGATIVE, &rsci.h.flags);
 	else {
 		int N, i;
-		struct gss_api_mech *gm;
 
 		/* gid */
 		if (get_int(&mesg, &rsci.cred.cr_gid))
@@ -488,21 +488,17 @@ static int rsc_parse(struct cache_detail
 		status = -EINVAL;
 		/* mech-specific data: */
 		len = qword_get(&mesg, buf, mlen);
-		if (len < 0) {
-			gss_mech_put(gm);
+		if (len < 0)
 			goto out;
-		}
 		status = gss_import_sec_context(buf, len, gm, &rsci.mechctx);
-		if (status) {
-			gss_mech_put(gm);
+		if (status)
 			goto out;
-		}
-		gss_mech_put(gm);
 	}
 	rsci.h.expiry_time = expiry;
 	rscp = rsc_update(&rsci, rscp);
 	status = 0;
 out:
+	gss_mech_put(gm);
 	rsc_free(&rsci);
 	if (rscp)
 		cache_put(&rscp->h, &rsc_cache);

[PATCH 010 of 14] knfsd: nfsd4: fix some open argument tests

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


These tests always returned true; clearly that wasn't what was intended.

In keeping with kernel style, make them functions instead of macros while
we're at it.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/nfs4state.c |   17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff .prev/fs/nfsd/nfs4state.c ./fs/nfsd/nfs4state.c
--- .prev/fs/nfsd/nfs4state.c	2006-06-27 14:44:03.000000000 +1000
+++ ./fs/nfsd/nfs4state.c	2006-06-27 15:00:26.000000000 +1000
@@ -1237,8 +1237,15 @@ find_file(struct inode *ino)
 	return NULL;
 }
 
-#define TEST_ACCESS(x) ((x > 0 || x < 4)?1:0)
-#define TEST_DENY(x) ((x >= 0 || x < 5)?1:0)
+static int access_valid(u32 x)
+{
+	return (x > 0 && x < 4);
+}
+
+static int deny_valid(u32 x)
+{
+	return (x >= 0 && x < 5);
+}
 
 static void
 set_access(unsigned int *access, unsigned long bmap) {
@@ -1745,7 +1752,8 @@ nfsd4_process_open2(struct svc_rqst *rqs
 	int status;
 
 	status = nfserr_inval;
-	if (!TEST_ACCESS(open->op_share_access) || !TEST_DENY(open->op_share_deny))
+	if (!access_valid(open->op_share_access)
+			|| !deny_valid(open->op_share_deny))
 		goto out;
 	/*
 	 * Lookup file; if found, lookup stateid and check open request,
@@ -2317,7 +2325,8 @@ nfsd4_open_downgrade(struct svc_rqst *rq
 			(int)current_fh->fh_dentry->d_name.len,
 			current_fh->fh_dentry->d_name.name);
 
-	if (!TEST_ACCESS(od->od_share_access) || !TEST_DENY(od->od_share_deny))
+	if (!access_valid(od->od_share_access)
+			|| !deny_valid(od->od_share_deny))
 		return nfserr_inval;
 
 	nfs4_lock_state();

[PATCH 011 of 14] knfsd: nfsd4: fix open flag passing

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


Since nfsv4 actually keeps around the file descriptors it gets from open
(instead of just using them for a single read or write operation), we need
to make sure that we can do RDWR opens and not just RDONLY/WRONLY.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/nfs4state.c |    6 +++---
 ./fs/nfsd/vfs.c       |    5 ++++-
 2 files changed, 7 insertions(+), 4 deletions(-)

diff .prev/fs/nfsd/nfs4state.c ./fs/nfsd/nfs4state.c
--- .prev/fs/nfsd/nfs4state.c	2006-06-27 15:00:26.000000000 +1000
+++ ./fs/nfsd/nfs4state.c	2006-06-27 15:04:03.000000000 +1000
@@ -1790,10 +1790,10 @@ nfsd4_process_open2(struct svc_rqst *rqs
 	} else {
 		/* Stateid was not found, this is a new OPEN */
 		int flags = 0;
+		if (open->op_share_access & NFS4_SHARE_ACCESS_READ)
+			flags |= MAY_READ;
 		if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
-			flags = MAY_WRITE;
-		else
-			flags = MAY_READ;
+			flags |= MAY_WRITE;
 		status = nfs4_new_open(rqstp, &stp, dp, current_fh, flags);
 		if (status)
 			goto out;

diff .prev/fs/nfsd/vfs.c ./fs/nfsd/vfs.c
--- .prev/fs/nfsd/vfs.c	2006-06-27 14:59:40.000000000 +1000
+++ ./fs/nfsd/vfs.c	2006-06-27 15:04:03.000000000 +1000
@@ -673,7 +673,10 @@ nfsd_open(struct svc_rqst *rqstp, struct
 		goto out_nfserr;
 
 	if (access & MAY_WRITE) {
-		flags = O_WRONLY|O_LARGEFILE;
+		if (access & MAY_READ)
+			flags = O_RDWR|O_LARGEFILE;
+		else
+			flags = O_WRONLY|O_LARGEFILE;
 
 		DQUOT_INIT(inode);
 	}

[PATCH 012 of 14] knfsd: svcrpc: Simplify nfsd rpcsec_gss integrity code

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


Pull out some of the integrity code into its own function, otherwise
svcauth_gss_release() is going to become very ungainly after the addition
of privacy code.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./net/sunrpc/auth_gss/svcauth_gss.c |  115 ++++++++++++++++++++----------------
 1 file changed, 64 insertions(+), 51 deletions(-)

diff .prev/net/sunrpc/auth_gss/svcauth_gss.c ./net/sunrpc/auth_gss/svcauth_gss.c
--- .prev/net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 14:59:44.000000000 +1000
+++ ./net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 15:05:11.000000000 +1000
@@ -1072,8 +1072,8 @@ out:
 	return ret;
 }
 
-static int
-svcauth_gss_release(struct svc_rqst *rqstp)
+static inline int
+svcauth_gss_wrap_resp_integ(struct svc_rqst *rqstp)
 {
 	struct gss_svc_data *gsd = (struct gss_svc_data *)rqstp->rq_auth_data;
 	struct rpc_gss_wire_cred *gc = &gsd->clcred;
@@ -1085,6 +1085,67 @@ svcauth_gss_release(struct svc_rqst *rqs
 	int integ_offset, integ_len;
 	int stat = -EINVAL;
 
+	p = gsd->body_start;
+	gsd->body_start = NULL;
+	/* move accept_stat to right place: */
+	memcpy(p, p + 2, 4);
+	/* don't wrap in failure case: */
+	/* Note: counting on not getting here if call was not even
+	 * accepted! */
+	if (*p != rpc_success) {
+		resbuf->head[0].iov_len -= 2 * 4;
+		goto out;
+	}
+	p++;
+	integ_offset = (u8 *)(p + 1) - (u8 *)resbuf->head[0].iov_base;
+	integ_len = resbuf->len - integ_offset;
+	BUG_ON(integ_len % 4);
+	*p++ = htonl(integ_len);
+	*p++ = htonl(gc->gc_seq);
+	if (xdr_buf_subsegment(resbuf, &integ_buf, integ_offset,
+				integ_len))
+		BUG();
+	if (resbuf->page_len == 0
+			&& resbuf->head[0].iov_len + RPC_MAX_AUTH_SIZE
+			< PAGE_SIZE) {
+		BUG_ON(resbuf->tail[0].iov_len);
+		/* Use head for everything */
+		resv = &resbuf->head[0];
+	} else if (resbuf->tail[0].iov_base == NULL) {
+		if (resbuf->head[0].iov_len + RPC_MAX_AUTH_SIZE > PAGE_SIZE)
+			goto out_err;
+		resbuf->tail[0].iov_base = resbuf->head[0].iov_base
+						+ resbuf->head[0].iov_len;
+		resbuf->tail[0].iov_len = 0;
+		rqstp->rq_restailpage = 0;
+		resv = &resbuf->tail[0];
+	} else {
+		resv = &resbuf->tail[0];
+	}
+	mic.data = (u8 *)resv->iov_base + resv->iov_len + 4;
+	if (gss_get_mic(gsd->rsci->mechctx, &integ_buf, &mic))
+		goto out_err;
+	svc_putu32(resv, htonl(mic.len));
+	memset(mic.data + mic.len, 0,
+			round_up_to_quad(mic.len) - mic.len);
+	resv->iov_len += XDR_QUADLEN(mic.len) << 2;
+	/* not strictly required: */
+	resbuf->len += XDR_QUADLEN(mic.len) << 2;
+	BUG_ON(resv->iov_len > PAGE_SIZE);
+out:
+	stat = 0;
+out_err:
+	return stat;
+}
+
+static int
+svcauth_gss_release(struct svc_rqst *rqstp)
+{
+	struct gss_svc_data *gsd = (struct gss_svc_data *)rqstp->rq_auth_data;
+	struct rpc_gss_wire_cred *gc = &gsd->clcred;
+	struct xdr_buf *resbuf = &rqstp->rq_res;
+	int stat = -EINVAL;
+
 	if (gc->gc_proc != RPC_GSS_PROC_DATA)
 		goto out;
 	/* Release can be called twice, but we only wrap once. */
@@ -1097,55 +1158,7 @@ svcauth_gss_release(struct svc_rqst *rqs
 	case RPC_GSS_SVC_NONE:
 		break;
 	case RPC_GSS_SVC_INTEGRITY:
-		p = gsd->body_start;
-		gsd->body_start = NULL;
-		/* move accept_stat to right place: */
-		memcpy(p, p + 2, 4);
-		/* don't wrap in failure case: */
-		/* Note: counting on not getting here if call was not even
-		 * accepted! */
-		if (*p != rpc_success) {
-			resbuf->head[0].iov_len -= 2 * 4;
-			goto out;
-		}
-		p++;
-		integ_offset = (u8 *)(p + 1) - (u8 *)resbuf->head[0].iov_base;
-		integ_len = resbuf->len - integ_offset;
-		BUG_ON(integ_len % 4);
-		*p++ = htonl(integ_len);
-		*p++ = htonl(gc->gc_seq);
-		if (xdr_buf_subsegment(resbuf, &integ_buf, integ_offset,
-					integ_len))
-			BUG();
-		if (resbuf->page_len == 0
-			&& resbuf->head[0].iov_len + RPC_MAX_AUTH_SIZE
-				< PAGE_SIZE) {
-			BUG_ON(resbuf->tail[0].iov_len);
-			/* Use head for everything */
-			resv = &resbuf->head[0];
-		} else if (resbuf->tail[0].iov_base == NULL) {
-			if (resbuf->head[0].iov_len + RPC_MAX_AUTH_SIZE
-					> PAGE_SIZE)
-				goto out_err;
-			resbuf->tail[0].iov_base =
-				resbuf->head[0].iov_base
-				+ resbuf->head[0].iov_len;
-			resbuf->tail[0].iov_len = 0;
-			rqstp->rq_restailpage = 0;
-			resv = &resbuf->tail[0];
-		} else {
-			resv = &resbuf->tail[0];
-		}
-		mic.data = (u8 *)resv->iov_base + resv->iov_len + 4;
-		if (gss_get_mic(gsd->rsci->mechctx, &integ_buf, &mic))
-			goto out_err;
-		svc_putu32(resv, htonl(mic.len));
-		memset(mic.data + mic.len, 0,
-				round_up_to_quad(mic.len) - mic.len);
-		resv->iov_len += XDR_QUADLEN(mic.len) << 2;
-		/* not strictly required: */
-		resbuf->len += XDR_QUADLEN(mic.len) << 2;
-		BUG_ON(resv->iov_len > PAGE_SIZE);
+		svcauth_gss_wrap_resp_integ(rqstp);
 		break;
 	case RPC_GSS_SVC_PRIVACY:
 	default:

[PATCH 013 of 14] knfsd: nfsd: mark rqstp to prevent use of sendfile in privacy case

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


Add a rq_sendfile_ok flag to svc_rqst which will be cleared in the privacy
case so that the wrapping code will get copies of the read data instead of
real page cache pages.  This makes life simpler when we encrypt the
response.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./fs/nfsd/vfs.c              |    2 +-
 ./include/linux/sunrpc/svc.h |    4 +++-
 ./net/sunrpc/svc.c           |    2 ++
 3 files changed, 6 insertions(+), 2 deletions(-)

diff .prev/fs/nfsd/vfs.c ./fs/nfsd/vfs.c
--- .prev/fs/nfsd/vfs.c	2006-06-27 15:04:03.000000000 +1000
+++ ./fs/nfsd/vfs.c	2006-06-27 15:07:01.000000000 +1000
@@ -840,7 +840,7 @@ nfsd_vfs_read(struct svc_rqst *rqstp, st
 	if (ra && ra->p_set)
 		file->f_ra = ra->p_ra;
 
-	if (file->f_op->sendfile) {
+	if (file->f_op->sendfile && rqstp->rq_sendfile_ok) {
 		svc_pushback_unused_pages(rqstp);
 		err = file->f_op->sendfile(file, &offset, *count,
 						 nfsd_read_actor, rqstp);

diff .prev/include/linux/sunrpc/svc.h ./include/linux/sunrpc/svc.h
--- .prev/include/linux/sunrpc/svc.h	2006-06-27 15:07:01.000000000 +1000
+++ ./include/linux/sunrpc/svc.h	2006-06-27 15:07:01.000000000 +1000
@@ -159,7 +159,9 @@ struct svc_rqst {
 						 * determine what device number
 						 * to report (real or virtual)
 						 */
-
+	int			rq_sendfile_ok; /* turned off in gss privacy
+						 * to prevent encrypting page
+						 * cache pages */
 	wait_queue_head_t	rq_wait;	/* synchronization */
 };
 

diff .prev/net/sunrpc/svc.c ./net/sunrpc/svc.c
--- .prev/net/sunrpc/svc.c	2006-06-27 15:07:01.000000000 +1000
+++ ./net/sunrpc/svc.c	2006-06-27 15:07:01.000000000 +1000
@@ -281,6 +281,8 @@ svc_process(struct svc_serv *serv, struc
 	rqstp->rq_res.page_len = 0;
 	rqstp->rq_res.buflen = PAGE_SIZE;
 	rqstp->rq_res.tail[0].iov_len = 0;
+	/* Will be turned off only in gss privacy case: */
+	rqstp->rq_sendfile_ok = 1;
 	/* tcp needs a space for the record length... */
 	if (rqstp->rq_prot == IPPROTO_TCP)
 		svc_putu32(resv, 0);

[PATCH 014 of 14] knfsd: svcrpc: gss: server-side implementation of rpcsec_gss privacy.

[NeilBrown]

From: J. Bruce Fields <bfields@citi.umich.edu>


Server-side implementation of rpcsec_gss privacy, which enables encryption
of the payload of every rpc request and response.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

### Diffstat output
 ./net/sunrpc/auth_gss/svcauth_gss.c |  154 ++++++++++++++++++++++++++++++++++--
 ./net/sunrpc/svc.c                  |    1 
 2 files changed, 148 insertions(+), 7 deletions(-)

diff .prev/net/sunrpc/auth_gss/svcauth_gss.c ./net/sunrpc/auth_gss/svcauth_gss.c
--- .prev/net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 15:05:11.000000000 +1000
+++ ./net/sunrpc/auth_gss/svcauth_gss.c	2006-06-27 15:07:54.000000000 +1000
@@ -832,6 +832,74 @@ out:
 	return stat;
 }
 
+static inline int
+total_buf_len(struct xdr_buf *buf)
+{
+	return buf->head[0].iov_len + buf->page_len + buf->tail[0].iov_len;
+}
+
+static void
+fix_priv_head(struct xdr_buf *buf, int pad)
+{
+	if (buf->page_len == 0) {
+		/* We need to adjust head and buf->len in tandem in this
+		 * case to make svc_defer() work--it finds the original
+		 * buffer start using buf->len - buf->head[0].iov_len. */
+		buf->head[0].iov_len -= pad;
+	}
+}
+
+static int
+unwrap_priv_data(struct svc_rqst *rqstp, struct xdr_buf *buf, u32 seq, struct gss_ctx *ctx)
+{
+	u32 priv_len, maj_stat;
+	int pad, saved_len, remaining_len, offset;
+
+	rqstp->rq_sendfile_ok = 0;
+
+	priv_len = ntohl(svc_getu32(&buf->head[0]));
+	if (rqstp->rq_deferred) {
+		/* Already decrypted last time through! The sequence number
+		 * check at out_seq is unnecessary but harmless: */
+		goto out_seq;
+	}
+	/* buf->len is the number of bytes from the original start of the
+	 * request to the end, where head[0].iov_len is just the bytes
+	 * not yet read from the head, so these two values are different: */
+	remaining_len = total_buf_len(buf);
+	if (priv_len > remaining_len)
+		return -EINVAL;
+	pad = remaining_len - priv_len;
+	buf->len -= pad;
+	fix_priv_head(buf, pad);
+
+	/* Maybe it would be better to give gss_unwrap a length parameter: */
+	saved_len = buf->len;
+	buf->len = priv_len;
+	maj_stat = gss_unwrap(ctx, 0, buf);
+	pad = priv_len - buf->len;
+	buf->len = saved_len;
+	buf->len -= pad;
+	/* The upper layers assume the buffer is aligned on 4-byte boundaries.
+	 * In the krb5p case, at least, the data ends up offset, so we need to
+	 * move it around. */
+	/* XXX: This is very inefficient.  It would be better to either do
+	 * this while we encrypt, or maybe in the receive code, if we can peak
+	 * ahead and work out the service and mechanism there. */
+	offset = buf->head[0].iov_len % 4;
+	if (offset) {
+		buf->buflen = RPCSVC_MAXPAYLOAD;
+		xdr_shift_buf(buf, offset);
+		fix_priv_head(buf, pad);
+	}
+	if (maj_stat != GSS_S_COMPLETE)
+		return -EINVAL;
+out_seq:
+	if (ntohl(svc_getu32(&buf->head[0])) != seq)
+		return -EINVAL;
+	return 0;
+}
+
 struct gss_svc_data {
 	/* decoded gss client cred: */
 	struct rpc_gss_wire_cred	clcred;
@@ -1047,7 +1115,14 @@ svcauth_gss_accept(struct svc_rqst *rqst
 			svc_putu32(resv, 0);
 			break;
 		case RPC_GSS_SVC_PRIVACY:
-			/* currently unsupported */
+			if (unwrap_priv_data(rqstp, &rqstp->rq_arg,
+					gc->gc_seq, rsci->mechctx))
+				goto auth_err;
+			/* placeholders for length and seq. number: */
+			svcdata->body_start = resv->iov_base + resv->iov_len;
+			svc_putu32(resv, 0);
+			svc_putu32(resv, 0);
+			break;
 		default:
 			goto auth_err;
 		}
@@ -1089,9 +1164,8 @@ svcauth_gss_wrap_resp_integ(struct svc_r
 	gsd->body_start = NULL;
 	/* move accept_stat to right place: */
 	memcpy(p, p + 2, 4);
-	/* don't wrap in failure case: */
-	/* Note: counting on not getting here if call was not even
-	 * accepted! */
+	/* Don't wrap in failure case: */
+	/* Counting on not getting here if call was not even accepted! */
 	if (*p != rpc_success) {
 		resbuf->head[0].iov_len -= 2 * 4;
 		goto out;
@@ -1138,6 +1212,65 @@ out_err:
 	return stat;
 }
 
+static inline int
+svcauth_gss_wrap_resp_priv(struct svc_rqst *rqstp)
+{
+	struct gss_svc_data *gsd = (struct gss_svc_data *)rqstp->rq_auth_data;
+	struct rpc_gss_wire_cred *gc = &gsd->clcred;
+	struct xdr_buf *resbuf = &rqstp->rq_res;
+	struct page **inpages = NULL;
+	u32 *p;
+	int offset, *len;
+	int pad;
+
+	p = gsd->body_start;
+	gsd->body_start = NULL;
+	/* move accept_stat to right place: */
+	memcpy(p, p + 2, 4);
+	/* Don't wrap in failure case: */
+	/* Counting on not getting here if call was not even accepted! */
+	if (*p != rpc_success) {
+		resbuf->head[0].iov_len -= 2 * 4;
+		return 0;
+	}
+	p++;
+	len = p++;
+	offset = (u8 *)p - (u8 *)resbuf->head[0].iov_base;
+	*p++ = htonl(gc->gc_seq);
+	inpages = resbuf->pages;
+	/* XXX: Would be better to write some xdr helper functions for
+	 * nfs{2,3,4}xdr.c that place the data right, instead of copying: */
+	if (resbuf->tail[0].iov_base && rqstp->rq_restailpage == 0) {
+		BUG_ON(resbuf->tail[0].iov_base >= resbuf->head[0].iov_base
+							+ PAGE_SIZE);
+		BUG_ON(resbuf->tail[0].iov_base < resbuf->head[0].iov_base);
+		if (resbuf->tail[0].iov_len + resbuf->head[0].iov_len
+				+ 2 * RPC_MAX_AUTH_SIZE > PAGE_SIZE)
+			return -ENOMEM;
+		memmove(resbuf->tail[0].iov_base + RPC_MAX_AUTH_SIZE,
+			resbuf->tail[0].iov_base,
+			resbuf->tail[0].iov_len);
+		resbuf->tail[0].iov_base += RPC_MAX_AUTH_SIZE;
+	}
+	if (resbuf->tail[0].iov_base == NULL) {
+		if (resbuf->head[0].iov_len + 2*RPC_MAX_AUTH_SIZE > PAGE_SIZE)
+			return -ENOMEM;
+		resbuf->tail[0].iov_base = resbuf->head[0].iov_base
+			+ resbuf->head[0].iov_len + RPC_MAX_AUTH_SIZE;
+		resbuf->tail[0].iov_len = 0;
+		rqstp->rq_restailpage = 0;
+	}
+	if (gss_wrap(gsd->rsci->mechctx, offset, resbuf, inpages))
+		return -ENOMEM;
+	*len = htonl(resbuf->len - offset);
+	pad = 3 - ((resbuf->len - offset - 1)&3);
+	p = (u32 *)(resbuf->tail[0].iov_base + resbuf->tail[0].iov_len);
+	memset(p, 0, pad);
+	resbuf->tail[0].iov_len += pad;
+	resbuf->len += pad;
+	return 0;
+}
+
 static int
 svcauth_gss_release(struct svc_rqst *rqstp)
 {
@@ -1152,15 +1285,22 @@ svcauth_gss_release(struct svc_rqst *rqs
 	if (gsd->body_start == NULL)
 		goto out;
 	/* normally not set till svc_send, but we need it here: */
-	resbuf->len = resbuf->head[0].iov_len
-		+ resbuf->page_len + resbuf->tail[0].iov_len;
+	/* XXX: what for?  Do we mess it up the moment we call svc_putu32
+	 * or whatever? */
+	resbuf->len = total_buf_len(resbuf);
 	switch (gc->gc_svc) {
 	case RPC_GSS_SVC_NONE:
 		break;
 	case RPC_GSS_SVC_INTEGRITY:
-		svcauth_gss_wrap_resp_integ(rqstp);
+		stat = svcauth_gss_wrap_resp_integ(rqstp);
+		if (stat)
+			goto out_err;
 		break;
 	case RPC_GSS_SVC_PRIVACY:
+		stat = svcauth_gss_wrap_resp_priv(rqstp);
+		if (stat)
+			goto out_err;
+		break;
 	default:
 		goto out_err;
 	}

diff .prev/net/sunrpc/svc.c ./net/sunrpc/svc.c
--- .prev/net/sunrpc/svc.c	2006-06-27 15:07:01.000000000 +1000
+++ ./net/sunrpc/svc.c	2006-06-27 15:07:54.000000000 +1000
@@ -280,6 +280,7 @@ svc_process(struct svc_serv *serv, struc
 	rqstp->rq_res.page_base = 0;
 	rqstp->rq_res.page_len = 0;
 	rqstp->rq_res.buflen = PAGE_SIZE;
+	rqstp->rq_res.tail[0].iov_base = NULL;
 	rqstp->rq_res.tail[0].iov_len = 0;
 	/* Will be turned off only in gss privacy case: */
 	rqstp->rq_sendfile_ok = 1;

[PATCH 000 of 14] knfsd: Introduction

[NeilBrown]

The following series of patches changes the code for lookup up
authentication/authorisation caches in sunrpc as used by nfsd.  That
than having a delightful macro that defines a function of a particular
type of cache, the functions are coded on a more traditional manner
using library routines.

This will hopefully make the code more maintainable.

This series should not effect the functionality at all -- except maybe
to fix some very minor bugs

This is against 2.6.16-rc5-mm2 and is NOT suitable for 2.6.16, but maybe for .17
(though holding it back to .18 wouldn't be a big problem if any issues came up).

Thanks,
NeilBrown


 [PATCH 000 of 14] knfsd: Introduction
 [PATCH 001 of 14] knfsd: Change the store of auth_domains to not be a 'cache'.
 [PATCH 002 of 14] knfsd: Break the hard linkage from svc_expkey to svc_export
 [PATCH 003 of 14] knfsd: Get rid of 'inplace' sunrpc caches
 [PATCH 004 of 14] knfsd: Create cache_lookup function instead of using a macro to declare one.
 [PATCH 005 of 14] knfsd: Convert ip_map cache to use the new lookup routine.
 [PATCH 006 of 14] knfsd: Use new cache_lookup for svc_export
 [PATCH 007 of 14] knfsd: Use new cache_lookup for svc_expkey cache.
 [PATCH 008 of 14] knfsd: Use new sunrpc cache for rsi cache
 [PATCH 009 of 14] knfsd: Use new cache code for rsc cache
 [PATCH 010 of 14] knfsd: Use new cache code for name/id lookup caches
 [PATCH 011 of 14] knfsd: An assortment of little fixes to the sunrpc cache code.
 [PATCH 012 of 14] knfsd: Remove DefineCacheLookup
 [PATCH 013 of 14] knfsd: Unexport cache_fresh and fix a small race.
 [PATCH 014 of 14] knfsd: Convert sunrpc_cache to use krefs

Re: [PATCH 000 of 14] knfsd: Introduction

[Trond Myklebust]

On Thu, 2006-03-09 at 17:51 +1100, NeilBrown wrote:
>  [PATCH 000 of 14] knfsd: Introduction
>  [PATCH 001 of 14] knfsd: Change the store of auth_domains to not be a 'cache'.
>  [PATCH 002 of 14] knfsd: Break the hard linkage from svc_expkey to svc_export
>  [PATCH 003 of 14] knfsd: Get rid of 'inplace' sunrpc caches
>  [PATCH 004 of 14] knfsd: Create cache_lookup function instead of using a macro to declare one.
>  [PATCH 005 of 14] knfsd: Convert ip_map cache to use the new lookup routine.
>  [PATCH 006 of 14] knfsd: Use new cache_lookup for svc_export
>  [PATCH 007 of 14] knfsd: Use new cache_lookup for svc_expkey cache.
>  [PATCH 008 of 14] knfsd: Use new sunrpc cache for rsi cache
>  [PATCH 009 of 14] knfsd: Use new cache code for rsc cache
>  [PATCH 010 of 14] knfsd: Use new cache code for name/id lookup caches
>  [PATCH 011 of 14] knfsd: An assortment of little fixes to the sunrpc cache code.
>  [PATCH 012 of 14] knfsd: Remove DefineCacheLookup
>  [PATCH 013 of 14] knfsd: Unexport cache_fresh and fix a small race.
>  [PATCH 014 of 14] knfsd: Convert sunrpc_cache to use krefs

Any plans to update Documentation/rpc-cache.txt?

Cheers,
  Trond

Re: [PATCH 000 of 14] knfsd: Introduction

[Neil Brown]

On Thursday March 9, trond.myklebust@fys.uio.no wrote:
> On Thu, 2006-03-09 at 17:51 +1100, NeilBrown wrote:
> >  [PATCH 000 of 14] knfsd: Introduction
> >  [PATCH 001 of 14] knfsd: Change the store of auth_domains to not be a 'cache'.
> >  [PATCH 002 of 14] knfsd: Break the hard linkage from svc_expkey to svc_export
> >  [PATCH 003 of 14] knfsd: Get rid of 'inplace' sunrpc caches
> >  [PATCH 004 of 14] knfsd: Create cache_lookup function instead of using a macro to declare one.
> >  [PATCH 005 of 14] knfsd: Convert ip_map cache to use the new lookup routine.
> >  [PATCH 006 of 14] knfsd: Use new cache_lookup for svc_export
> >  [PATCH 007 of 14] knfsd: Use new cache_lookup for svc_expkey cache.
> >  [PATCH 008 of 14] knfsd: Use new sunrpc cache for rsi cache
> >  [PATCH 009 of 14] knfsd: Use new cache code for rsc cache
> >  [PATCH 010 of 14] knfsd: Use new cache code for name/id lookup caches
> >  [PATCH 011 of 14] knfsd: An assortment of little fixes to the sunrpc cache code.
> >  [PATCH 012 of 14] knfsd: Remove DefineCacheLookup
> >  [PATCH 013 of 14] knfsd: Unexport cache_fresh and fix a small race.
> >  [PATCH 014 of 14] knfsd: Convert sunrpc_cache to use krefs
> 
> Any plans to update Documentation/rpc-cache.txt?

What a good idea!  I'll put it on my todo list, thanks.

NeilBrown