Merge branch 'for-2.6.34' of git://linux-nfs.org/~bfields/linux
[linux-2.6.git] / fs / nfsd / nfs4xdr.c
index e15f2cf..78c7e24 100644 (file)
@@ -1,6 +1,4 @@
 /*
- *  fs/nfs/nfs4xdr.c
- *
  *  Server-side XDR for NFSv4
  *
  *  Copyright (c) 2002 The Regents of the University of Michigan.
  * at the end of nfs4svc_decode_compoundargs.
  */
 
-#include <linux/param.h>
-#include <linux/smp.h>
-#include <linux/fs.h>
 #include <linux/namei.h>
-#include <linux/vfs.h>
-#include <linux/sunrpc/xdr.h>
-#include <linux/sunrpc/svc.h>
-#include <linux/sunrpc/clnt.h>
-#include <linux/nfsd/nfsd.h>
-#include <linux/nfsd/state.h>
-#include <linux/nfsd/xdr4.h>
+#include <linux/statfs.h>
+#include <linux/utsname.h>
 #include <linux/nfsd_idmap.h>
-#include <linux/nfs4.h>
 #include <linux/nfs4_acl.h>
-#include <linux/sunrpc/gss_api.h>
 #include <linux/sunrpc/svcauth_gss.h>
 
+#include "xdr4.h"
+#include "vfs.h"
+
 #define NFSDDBG_FACILITY               NFSDDBG_XDR
 
 /*
@@ -84,16 +75,6 @@ check_filename(char *str, int len, __be32 err)
        return 0;
 }
 
-/*
- * START OF "GENERIC" DECODE ROUTINES.
- *   These may look a little ugly since they are imported from a "generic"
- * set of XDR encode/decode routines which are intended to be shared by
- * all of our NFSv4 implementations (OpenBSD, MacOS X...).
- *
- * If the pain of reading these is too great, it should be a straightforward
- * task to translate them into Linux-specific versions which are more
- * consistent with the style used in NFSv2/v3...
- */
 #define DECODE_HEAD                            \
        __be32 *p;                              \
        __be32 status
@@ -102,7 +83,8 @@ check_filename(char *str, int len, __be32 err)
 out:                                           \
        return status;                          \
 xdr_error:                                     \
-       printk(KERN_NOTICE "xdr error! (%s:%d)\n", __FILE__, __LINE__); \
+       dprintk("NFSD: xdr error (%s:%d)\n",    \
+                       __FILE__, __LINE__);    \
        status = nfserr_bad_xdr;                \
        goto out
 
@@ -124,7 +106,8 @@ xdr_error:                                  \
        if (!(x = (p==argp->tmp || p == argp->tmpp) ? \
                savemem(argp, p, nbytes) :      \
                (char *)p)) {                   \
-               printk(KERN_NOTICE "xdr error! (%s:%d)\n", __FILE__, __LINE__); \
+               dprintk("NFSD: xdr error (%s:%d)\n", \
+                               __FILE__, __LINE__); \
                goto xdr_error;                 \
                }                               \
        p += XDR_QUADLEN(nbytes);               \
@@ -140,17 +123,18 @@ xdr_error:                                        \
                p = argp->p;                    \
                argp->p += XDR_QUADLEN(nbytes); \
        } else if (!(p = read_buf(argp, nbytes))) { \
-               printk(KERN_NOTICE "xdr error! (%s:%d)\n", __FILE__, __LINE__); \
+               dprintk("NFSD: xdr error (%s:%d)\n", \
+                               __FILE__, __LINE__); \
                goto xdr_error;                 \
        }                                       \
 } while (0)
 
-static __be32 *read_buf(struct nfsd4_compoundargs *argp, int nbytes)
+static __be32 *read_buf(struct nfsd4_compoundargs *argp, u32 nbytes)
 {
        /* We want more bytes than seem to be available.
         * Maybe we need a new page, maybe we have just run out
         */
-       int avail = (char*)argp->end - (char*)argp->p;
+       unsigned int avail = (char *)argp->end - (char *)argp->p;
        __be32 *p;
        if (avail + argp->pagelen < nbytes)
                return NULL;
@@ -166,6 +150,11 @@ static __be32 *read_buf(struct nfsd4_compoundargs *argp, int nbytes)
                        return NULL;
                
        }
+       /*
+        * The following memcpy is safe because read_buf is always
+        * called with nbytes > avail, and the two cases above both
+        * guarantee p points to at least nbytes bytes.
+        */
        memcpy(p, argp->p, avail);
        /* step to next page */
        argp->p = page_address(argp->pagelist[0]);
@@ -182,6 +171,11 @@ static __be32 *read_buf(struct nfsd4_compoundargs *argp, int nbytes)
        return p;
 }
 
+static int zero_clientid(clientid_t *clid)
+{
+       return (clid->cl_boot == 0) && (clid->cl_id == 0);
+}
+
 static int
 defer_free(struct nfsd4_compoundargs *argp,
                void (*release)(const void *), void *p)
@@ -224,6 +218,7 @@ nfsd4_decode_bitmap(struct nfsd4_compoundargs *argp, u32 *bmval)
 
        bmval[0] = 0;
        bmval[1] = 0;
+       bmval[2] = 0;
 
        READ_BUF(4);
        READ32(bmlen);
@@ -235,13 +230,15 @@ nfsd4_decode_bitmap(struct nfsd4_compoundargs *argp, u32 *bmval)
                READ32(bmval[0]);
        if (bmlen > 1)
                READ32(bmval[1]);
+       if (bmlen > 2)
+               READ32(bmval[2]);
 
        DECODE_TAIL;
 }
 
 static __be32
-nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval, struct iattr *iattr,
-    struct nfs4_acl **acl)
+nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval,
+                  struct iattr *iattr, struct nfs4_acl **acl)
 {
        int expected_len, len = 0;
        u32 dummy32;
@@ -253,15 +250,6 @@ nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval, struct iattr *ia
        if ((status = nfsd4_decode_bitmap(argp, bmval)))
                return status;
 
-       /*
-        * According to spec, unsupported attributes return ERR_ATTRNOTSUPP;
-        * read-only attributes return ERR_INVAL.
-        */
-       if ((bmval[0] & ~NFSD_SUPPORTED_ATTRS_WORD0) || (bmval[1] & ~NFSD_SUPPORTED_ATTRS_WORD1))
-               return nfserr_attrnotsupp;
-       if ((bmval[0] & ~NFSD_WRITEABLE_ATTRS_WORD0) || (bmval[1] & ~NFSD_WRITEABLE_ATTRS_WORD1))
-               return nfserr_inval;
-
        READ_BUF(4);
        READ32(expected_len);
 
@@ -368,20 +356,6 @@ nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval, struct iattr *ia
                        goto xdr_error;
                }
        }
-       if (bmval[1] & FATTR4_WORD1_TIME_METADATA) {
-               /* We require the high 32 bits of 'seconds' to be 0, and we ignore
-                  all 32 bits of 'nseconds'. */
-               READ_BUF(12);
-               len += 12;
-               READ32(dummy32);
-               if (dummy32)
-                       return nfserr_inval;
-               READ32(iattr->ia_ctime.tv_sec);
-               READ32(iattr->ia_ctime.tv_nsec);
-               if (iattr->ia_ctime.tv_nsec >= (u32)1000000000)
-                       return nfserr_inval;
-               iattr->ia_valid |= ATTR_CTIME;
-       }
        if (bmval[1] & FATTR4_WORD1_TIME_MODIFY_SET) {
                READ_BUF(4);
                len += 4;
@@ -408,7 +382,11 @@ nfsd4_decode_fattr(struct nfsd4_compoundargs *argp, u32 *bmval, struct iattr *ia
                        goto xdr_error;
                }
        }
-       if (len != expected_len)
+       if (bmval[0] & ~NFSD_WRITEABLE_ATTRS_WORD0
+           || bmval[1] & ~NFSD_WRITEABLE_ATTRS_WORD1
+           || bmval[2] & ~NFSD_WRITEABLE_ATTRS_WORD2)
+               READ_BUF(expected_len - len);
+       else if (len != expected_len)
                goto xdr_error;
 
        DECODE_TAIL;
@@ -419,6 +397,18 @@ out_nfserr:
 }
 
 static __be32
+nfsd4_decode_stateid(struct nfsd4_compoundargs *argp, stateid_t *sid)
+{
+       DECODE_HEAD;
+
+       READ_BUF(sizeof(stateid_t));
+       READ32(sid->si_generation);
+       COPYMEM(&sid->si_opaque, sizeof(stateid_opaque_t));
+
+       DECODE_TAIL;
+}
+
+static __be32
 nfsd4_decode_access(struct nfsd4_compoundargs *argp, struct nfsd4_access *access)
 {
        DECODE_HEAD;
@@ -435,10 +425,9 @@ nfsd4_decode_close(struct nfsd4_compoundargs *argp, struct nfsd4_close *close)
        DECODE_HEAD;
 
        close->cl_stateowner = NULL;
-       READ_BUF(4 + sizeof(stateid_t));
+       READ_BUF(4);
        READ32(close->cl_seqid);
-       READ32(close->cl_stateid.si_generation);
-       COPYMEM(&close->cl_stateid.si_opaque, sizeof(stateid_opaque_t));
+       return nfsd4_decode_stateid(argp, &close->cl_stateid);
 
        DECODE_TAIL;
 }
@@ -490,7 +479,9 @@ nfsd4_decode_create(struct nfsd4_compoundargs *argp, struct nfsd4_create *create
        if ((status = check_filename(create->cr_name, create->cr_namelen, nfserr_inval)))
                return status;
 
-       if ((status = nfsd4_decode_fattr(argp, create->cr_bmval, &create->cr_iattr, &create->cr_acl)))
+       status = nfsd4_decode_fattr(argp, create->cr_bmval, &create->cr_iattr,
+                                   &create->cr_acl);
+       if (status)
                goto out;
 
        DECODE_TAIL;
@@ -499,13 +490,7 @@ nfsd4_decode_create(struct nfsd4_compoundargs *argp, struct nfsd4_create *create
 static inline __be32
 nfsd4_decode_delegreturn(struct nfsd4_compoundargs *argp, struct nfsd4_delegreturn *dr)
 {
-       DECODE_HEAD;
-
-       READ_BUF(sizeof(stateid_t));
-       READ32(dr->dr_stateid.si_generation);
-       COPYMEM(&dr->dr_stateid.si_opaque, sizeof(stateid_opaque_t));
-
-       DECODE_TAIL;
+       return nfsd4_decode_stateid(argp, &dr->dr_stateid);
 }
 
 static inline __be32
@@ -548,20 +533,22 @@ nfsd4_decode_lock(struct nfsd4_compoundargs *argp, struct nfsd4_lock *lock)
        READ32(lock->lk_is_new);
 
        if (lock->lk_is_new) {
-               READ_BUF(36);
+               READ_BUF(4);
                READ32(lock->lk_new_open_seqid);
-               READ32(lock->lk_new_open_stateid.si_generation);
-
-               COPYMEM(&lock->lk_new_open_stateid.si_opaque, sizeof(stateid_opaque_t));
+               status = nfsd4_decode_stateid(argp, &lock->lk_new_open_stateid);
+               if (status)
+                       return status;
+               READ_BUF(8 + sizeof(clientid_t));
                READ32(lock->lk_new_lock_seqid);
                COPYMEM(&lock->lk_new_clientid, sizeof(clientid_t));
                READ32(lock->lk_new_owner.len);
                READ_BUF(lock->lk_new_owner.len);
                READMEM(lock->lk_new_owner.data, lock->lk_new_owner.len);
        } else {
-               READ_BUF(20);
-               READ32(lock->lk_old_lock_stateid.si_generation);
-               COPYMEM(&lock->lk_old_lock_stateid.si_opaque, sizeof(stateid_opaque_t));
+               status = nfsd4_decode_stateid(argp, &lock->lk_old_lock_stateid);
+               if (status)
+                       return status;
+               READ_BUF(4);
                READ32(lock->lk_old_lock_seqid);
        }
 
@@ -584,6 +571,8 @@ nfsd4_decode_lockt(struct nfsd4_compoundargs *argp, struct nfsd4_lockt *lockt)
        READ_BUF(lockt->lt_owner.len);
        READMEM(lockt->lt_owner.data, lockt->lt_owner.len);
 
+       if (argp->minorversion && !zero_clientid(&lockt->lt_clientid))
+               return nfserr_inval;
        DECODE_TAIL;
 }
 
@@ -593,13 +582,15 @@ nfsd4_decode_locku(struct nfsd4_compoundargs *argp, struct nfsd4_locku *locku)
        DECODE_HEAD;
 
        locku->lu_stateowner = NULL;
-       READ_BUF(24 + sizeof(stateid_t));
+       READ_BUF(8);
        READ32(locku->lu_type);
        if ((locku->lu_type < NFS4_READ_LT) || (locku->lu_type > NFS4_WRITEW_LT))
                goto xdr_error;
        READ32(locku->lu_seqid);
-       READ32(locku->lu_stateid.si_generation);
-       COPYMEM(&locku->lu_stateid.si_opaque, sizeof(stateid_opaque_t));
+       status = nfsd4_decode_stateid(argp, &locku->lu_stateid);
+       if (status)
+               return status;
+       READ_BUF(16);
        READ64(locku->lu_offset);
        READ64(locku->lu_length);
 
@@ -651,13 +642,25 @@ nfsd4_decode_open(struct nfsd4_compoundargs *argp, struct nfsd4_open *open)
                switch (open->op_createmode) {
                case NFS4_CREATE_UNCHECKED:
                case NFS4_CREATE_GUARDED:
-                       if ((status = nfsd4_decode_fattr(argp, open->op_bmval, &open->op_iattr, &open->op_acl)))
+                       status = nfsd4_decode_fattr(argp, open->op_bmval,
+                               &open->op_iattr, &open->op_acl);
+                       if (status)
                                goto out;
                        break;
                case NFS4_CREATE_EXCLUSIVE:
                        READ_BUF(8);
                        COPYMEM(open->op_verf.data, 8);
                        break;
+               case NFS4_CREATE_EXCLUSIVE4_1:
+                       if (argp->minorversion < 1)
+                               goto xdr_error;
+                       READ_BUF(8);
+                       COPYMEM(open->op_verf.data, 8);
+                       status = nfsd4_decode_fattr(argp, open->op_bmval,
+                               &open->op_iattr, &open->op_acl);
+                       if (status)
+                               goto out;
+                       break;
                default:
                        goto xdr_error;
                }
@@ -684,8 +687,10 @@ nfsd4_decode_open(struct nfsd4_compoundargs *argp, struct nfsd4_open *open)
                READ32(open->op_delegate_type);
                break;
        case NFS4_OPEN_CLAIM_DELEGATE_CUR:
-               READ_BUF(sizeof(stateid_t) + 4);
-               COPYMEM(&open->op_delegate_stateid, sizeof(stateid_t));
+               status = nfsd4_decode_stateid(argp, &open->op_delegate_stateid);
+               if (status)
+                       return status;
+               READ_BUF(4);
                READ32(open->op_fname.len);
                READ_BUF(open->op_fname.len);
                SAVEMEM(open->op_fname.data, open->op_fname.len);
@@ -705,9 +710,10 @@ nfsd4_decode_open_confirm(struct nfsd4_compoundargs *argp, struct nfsd4_open_con
        DECODE_HEAD;
                    
        open_conf->oc_stateowner = NULL;
-       READ_BUF(4 + sizeof(stateid_t));
-       READ32(open_conf->oc_req_stateid.si_generation);
-       COPYMEM(&open_conf->oc_req_stateid.si_opaque, sizeof(stateid_opaque_t));
+       status = nfsd4_decode_stateid(argp, &open_conf->oc_req_stateid);
+       if (status)
+               return status;
+       READ_BUF(4);
        READ32(open_conf->oc_seqid);
                                                        
        DECODE_TAIL;
@@ -719,9 +725,10 @@ nfsd4_decode_open_downgrade(struct nfsd4_compoundargs *argp, struct nfsd4_open_d
        DECODE_HEAD;
                    
        open_down->od_stateowner = NULL;
-       READ_BUF(12 + sizeof(stateid_t));
-       READ32(open_down->od_stateid.si_generation);
-       COPYMEM(&open_down->od_stateid.si_opaque, sizeof(stateid_opaque_t));
+       status = nfsd4_decode_stateid(argp, &open_down->od_stateid);
+       if (status)
+               return status;
+       READ_BUF(12);
        READ32(open_down->od_seqid);
        READ32(open_down->od_share_access);
        READ32(open_down->od_share_deny);
@@ -749,9 +756,10 @@ nfsd4_decode_read(struct nfsd4_compoundargs *argp, struct nfsd4_read *read)
 {
        DECODE_HEAD;
 
-       READ_BUF(sizeof(stateid_t) + 12);
-       READ32(read->rd_stateid.si_generation);
-       COPYMEM(&read->rd_stateid.si_opaque, sizeof(stateid_opaque_t));
+       status = nfsd4_decode_stateid(argp, &read->rd_stateid);
+       if (status)
+               return status;
+       READ_BUF(12);
        READ64(read->rd_offset);
        READ32(read->rd_length);
 
@@ -840,15 +848,13 @@ nfsd4_decode_secinfo(struct nfsd4_compoundargs *argp,
 static __be32
 nfsd4_decode_setattr(struct nfsd4_compoundargs *argp, struct nfsd4_setattr *setattr)
 {
-       DECODE_HEAD;
-
-       READ_BUF(sizeof(stateid_t));
-       READ32(setattr->sa_stateid.si_generation);
-       COPYMEM(&setattr->sa_stateid.si_opaque, sizeof(stateid_opaque_t));
-       if ((status = nfsd4_decode_fattr(argp, setattr->sa_bmval, &setattr->sa_iattr, &setattr->sa_acl)))
-               goto out;
+       __be32 status;
 
-       DECODE_TAIL;
+       status = nfsd4_decode_stateid(argp, &setattr->sa_stateid);
+       if (status)
+               return status;
+       return nfsd4_decode_fattr(argp, setattr->sa_bmval, &setattr->sa_iattr,
+                                 &setattr->sa_acl);
 }
 
 static __be32
@@ -933,9 +939,10 @@ nfsd4_decode_write(struct nfsd4_compoundargs *argp, struct nfsd4_write *write)
        int len;
        DECODE_HEAD;
 
-       READ_BUF(sizeof(stateid_opaque_t) + 20);
-       READ32(write->wr_stateid.si_generation);
-       COPYMEM(&write->wr_stateid.si_opaque, sizeof(stateid_opaque_t));
+       status = nfsd4_decode_stateid(argp, &write->wr_stateid);
+       if (status)
+               return status;
+       READ_BUF(16);
        READ64(write->wr_offset);
        READ32(write->wr_stable_how);
        if (write->wr_stable_how > 2)
@@ -948,7 +955,8 @@ nfsd4_decode_write(struct nfsd4_compoundargs *argp, struct nfsd4_write *write)
         */
        avail = (char*)argp->end - (char*)argp->p;
        if (avail + argp->pagelen < write->wr_buflen) {
-               printk(KERN_NOTICE "xdr error! (%s:%d)\n", __FILE__, __LINE__); 
+               dprintk("NFSD: xdr error (%s:%d)\n",
+                               __FILE__, __LINE__);
                goto xdr_error;
        }
        argp->rqstp->rq_vec[0].iov_base = p;
@@ -987,14 +995,375 @@ nfsd4_decode_release_lockowner(struct nfsd4_compoundargs *argp, struct nfsd4_rel
        READ_BUF(rlockowner->rl_owner.len);
        READMEM(rlockowner->rl_owner.data, rlockowner->rl_owner.len);
 
+       if (argp->minorversion && !zero_clientid(&rlockowner->rl_clientid))
+               return nfserr_inval;
        DECODE_TAIL;
 }
 
 static __be32
+nfsd4_decode_exchange_id(struct nfsd4_compoundargs *argp,
+                        struct nfsd4_exchange_id *exid)
+{
+       int dummy;
+       DECODE_HEAD;
+
+       READ_BUF(NFS4_VERIFIER_SIZE);
+       COPYMEM(exid->verifier.data, NFS4_VERIFIER_SIZE);
+
+       READ_BUF(4);
+       READ32(exid->clname.len);
+
+       READ_BUF(exid->clname.len);
+       SAVEMEM(exid->clname.data, exid->clname.len);
+
+       READ_BUF(4);
+       READ32(exid->flags);
+
+       /* Ignore state_protect4_a */
+       READ_BUF(4);
+       READ32(exid->spa_how);
+       switch (exid->spa_how) {
+       case SP4_NONE:
+               break;
+       case SP4_MACH_CRED:
+               /* spo_must_enforce */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy * 4);
+               p += dummy;
+
+               /* spo_must_allow */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy * 4);
+               p += dummy;
+               break;
+       case SP4_SSV:
+               /* ssp_ops */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy * 4);
+               p += dummy;
+
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy * 4);
+               p += dummy;
+
+               /* ssp_hash_algs<> */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy);
+               p += XDR_QUADLEN(dummy);
+
+               /* ssp_encr_algs<> */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy);
+               p += XDR_QUADLEN(dummy);
+
+               /* ssp_window and ssp_num_gss_handles */
+               READ_BUF(8);
+               READ32(dummy);
+               READ32(dummy);
+               break;
+       default:
+               goto xdr_error;
+       }
+
+       /* Ignore Implementation ID */
+       READ_BUF(4);    /* nfs_impl_id4 array length */
+       READ32(dummy);
+
+       if (dummy > 1)
+               goto xdr_error;
+
+       if (dummy == 1) {
+               /* nii_domain */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy);
+               p += XDR_QUADLEN(dummy);
+
+               /* nii_name */
+               READ_BUF(4);
+               READ32(dummy);
+               READ_BUF(dummy);
+               p += XDR_QUADLEN(dummy);
+
+               /* nii_date */
+               READ_BUF(12);
+               p += 3;
+       }
+       DECODE_TAIL;
+}
+
+static __be32
+nfsd4_decode_create_session(struct nfsd4_compoundargs *argp,
+                           struct nfsd4_create_session *sess)
+{
+       DECODE_HEAD;
+
+       u32 dummy;
+       char *machine_name;
+       int i;
+       int nr_secflavs;
+
+       READ_BUF(16);
+       COPYMEM(&sess->clientid, 8);
+       READ32(sess->seqid);
+       READ32(sess->flags);
+
+       /* Fore channel attrs */
+       READ_BUF(28);
+       READ32(dummy); /* headerpadsz is always 0 */
+       READ32(sess->fore_channel.maxreq_sz);
+       READ32(sess->fore_channel.maxresp_sz);
+       READ32(sess->fore_channel.maxresp_cached);
+       READ32(sess->fore_channel.maxops);
+       READ32(sess->fore_channel.maxreqs);
+       READ32(sess->fore_channel.nr_rdma_attrs);
+       if (sess->fore_channel.nr_rdma_attrs == 1) {
+               READ_BUF(4);
+               READ32(sess->fore_channel.rdma_attrs);
+       } else if (sess->fore_channel.nr_rdma_attrs > 1) {
+               dprintk("Too many fore channel attr bitmaps!\n");
+               goto xdr_error;
+       }
+
+       /* Back channel attrs */
+       READ_BUF(28);
+       READ32(dummy); /* headerpadsz is always 0 */
+       READ32(sess->back_channel.maxreq_sz);
+       READ32(sess->back_channel.maxresp_sz);
+       READ32(sess->back_channel.maxresp_cached);
+       READ32(sess->back_channel.maxops);
+       READ32(sess->back_channel.maxreqs);
+       READ32(sess->back_channel.nr_rdma_attrs);
+       if (sess->back_channel.nr_rdma_attrs == 1) {
+               READ_BUF(4);
+               READ32(sess->back_channel.rdma_attrs);
+       } else if (sess->back_channel.nr_rdma_attrs > 1) {
+               dprintk("Too many back channel attr bitmaps!\n");
+               goto xdr_error;
+       }
+
+       READ_BUF(8);
+       READ32(sess->callback_prog);
+
+       /* callback_sec_params4 */
+       READ32(nr_secflavs);
+       for (i = 0; i < nr_secflavs; ++i) {
+               READ_BUF(4);
+               READ32(dummy);
+               switch (dummy) {
+               case RPC_AUTH_NULL:
+                       /* Nothing to read */
+                       break;
+               case RPC_AUTH_UNIX:
+                       READ_BUF(8);
+                       /* stamp */
+                       READ32(dummy);
+
+                       /* machine name */
+                       READ32(dummy);
+                       READ_BUF(dummy);
+                       SAVEMEM(machine_name, dummy);
+
+                       /* uid, gid */
+                       READ_BUF(8);
+                       READ32(sess->uid);
+                       READ32(sess->gid);
+
+                       /* more gids */
+                       READ_BUF(4);
+                       READ32(dummy);
+                       READ_BUF(dummy * 4);
+                       for (i = 0; i < dummy; ++i)
+                               READ32(dummy);
+                       break;
+               case RPC_AUTH_GSS:
+                       dprintk("RPC_AUTH_GSS callback secflavor "
+                               "not supported!\n");
+                       READ_BUF(8);
+                       /* gcbp_service */
+                       READ32(dummy);
+                       /* gcbp_handle_from_server */
+                       READ32(dummy);
+                       READ_BUF(dummy);
+                       p += XDR_QUADLEN(dummy);
+                       /* gcbp_handle_from_client */
+                       READ_BUF(4);
+                       READ32(dummy);
+                       READ_BUF(dummy);
+                       p += XDR_QUADLEN(dummy);
+                       break;
+               default:
+                       dprintk("Illegal callback secflavor\n");
+                       return nfserr_inval;
+               }
+       }
+       DECODE_TAIL;
+}
+
+static __be32
+nfsd4_decode_destroy_session(struct nfsd4_compoundargs *argp,
+                            struct nfsd4_destroy_session *destroy_session)
+{
+       DECODE_HEAD;
+       READ_BUF(NFS4_MAX_SESSIONID_LEN);
+       COPYMEM(destroy_session->sessionid.data, NFS4_MAX_SESSIONID_LEN);
+
+       DECODE_TAIL;
+}
+
+static __be32
+nfsd4_decode_sequence(struct nfsd4_compoundargs *argp,
+                     struct nfsd4_sequence *seq)
+{
+       DECODE_HEAD;
+
+       READ_BUF(NFS4_MAX_SESSIONID_LEN + 16);
+       COPYMEM(seq->sessionid.data, NFS4_MAX_SESSIONID_LEN);
+       READ32(seq->seqid);
+       READ32(seq->slotid);
+       READ32(seq->maxslots);
+       READ32(seq->cachethis);
+
+       DECODE_TAIL;
+}
+
+static __be32
+nfsd4_decode_noop(struct nfsd4_compoundargs *argp, void *p)
+{
+       return nfs_ok;
+}
+
+static __be32
+nfsd4_decode_notsupp(struct nfsd4_compoundargs *argp, void *p)
+{
+       return nfserr_notsupp;
+}
+
+typedef __be32(*nfsd4_dec)(struct nfsd4_compoundargs *argp, void *);
+
+static nfsd4_dec nfsd4_dec_ops[] = {
+       [OP_ACCESS]             = (nfsd4_dec)nfsd4_decode_access,
+       [OP_CLOSE]              = (nfsd4_dec)nfsd4_decode_close,
+       [OP_COMMIT]             = (nfsd4_dec)nfsd4_decode_commit,
+       [OP_CREATE]             = (nfsd4_dec)nfsd4_decode_create,
+       [OP_DELEGPURGE]         = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_DELEGRETURN]        = (nfsd4_dec)nfsd4_decode_delegreturn,
+       [OP_GETATTR]            = (nfsd4_dec)nfsd4_decode_getattr,
+       [OP_GETFH]              = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_LINK]               = (nfsd4_dec)nfsd4_decode_link,
+       [OP_LOCK]               = (nfsd4_dec)nfsd4_decode_lock,
+       [OP_LOCKT]              = (nfsd4_dec)nfsd4_decode_lockt,
+       [OP_LOCKU]              = (nfsd4_dec)nfsd4_decode_locku,
+       [OP_LOOKUP]             = (nfsd4_dec)nfsd4_decode_lookup,
+       [OP_LOOKUPP]            = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_NVERIFY]            = (nfsd4_dec)nfsd4_decode_verify,
+       [OP_OPEN]               = (nfsd4_dec)nfsd4_decode_open,
+       [OP_OPENATTR]           = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_OPEN_CONFIRM]       = (nfsd4_dec)nfsd4_decode_open_confirm,
+       [OP_OPEN_DOWNGRADE]     = (nfsd4_dec)nfsd4_decode_open_downgrade,
+       [OP_PUTFH]              = (nfsd4_dec)nfsd4_decode_putfh,
+       [OP_PUTPUBFH]           = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_PUTROOTFH]          = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_READ]               = (nfsd4_dec)nfsd4_decode_read,
+       [OP_READDIR]            = (nfsd4_dec)nfsd4_decode_readdir,
+       [OP_READLINK]           = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_REMOVE]             = (nfsd4_dec)nfsd4_decode_remove,
+       [OP_RENAME]             = (nfsd4_dec)nfsd4_decode_rename,
+       [OP_RENEW]              = (nfsd4_dec)nfsd4_decode_renew,
+       [OP_RESTOREFH]          = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_SAVEFH]             = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_SECINFO]            = (nfsd4_dec)nfsd4_decode_secinfo,
+       [OP_SETATTR]            = (nfsd4_dec)nfsd4_decode_setattr,
+       [OP_SETCLIENTID]        = (nfsd4_dec)nfsd4_decode_setclientid,
+       [OP_SETCLIENTID_CONFIRM] = (nfsd4_dec)nfsd4_decode_setclientid_confirm,
+       [OP_VERIFY]             = (nfsd4_dec)nfsd4_decode_verify,
+       [OP_WRITE]              = (nfsd4_dec)nfsd4_decode_write,
+       [OP_RELEASE_LOCKOWNER]  = (nfsd4_dec)nfsd4_decode_release_lockowner,
+};
+
+static nfsd4_dec nfsd41_dec_ops[] = {
+       [OP_ACCESS]             = (nfsd4_dec)nfsd4_decode_access,
+       [OP_CLOSE]              = (nfsd4_dec)nfsd4_decode_close,
+       [OP_COMMIT]             = (nfsd4_dec)nfsd4_decode_commit,
+       [OP_CREATE]             = (nfsd4_dec)nfsd4_decode_create,
+       [OP_DELEGPURGE]         = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_DELEGRETURN]        = (nfsd4_dec)nfsd4_decode_delegreturn,
+       [OP_GETATTR]            = (nfsd4_dec)nfsd4_decode_getattr,
+       [OP_GETFH]              = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_LINK]               = (nfsd4_dec)nfsd4_decode_link,
+       [OP_LOCK]               = (nfsd4_dec)nfsd4_decode_lock,
+       [OP_LOCKT]              = (nfsd4_dec)nfsd4_decode_lockt,
+       [OP_LOCKU]              = (nfsd4_dec)nfsd4_decode_locku,
+       [OP_LOOKUP]             = (nfsd4_dec)nfsd4_decode_lookup,
+       [OP_LOOKUPP]            = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_NVERIFY]            = (nfsd4_dec)nfsd4_decode_verify,
+       [OP_OPEN]               = (nfsd4_dec)nfsd4_decode_open,
+       [OP_OPENATTR]           = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_OPEN_CONFIRM]       = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_OPEN_DOWNGRADE]     = (nfsd4_dec)nfsd4_decode_open_downgrade,
+       [OP_PUTFH]              = (nfsd4_dec)nfsd4_decode_putfh,
+       [OP_PUTPUBFH]           = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_PUTROOTFH]          = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_READ]               = (nfsd4_dec)nfsd4_decode_read,
+       [OP_READDIR]            = (nfsd4_dec)nfsd4_decode_readdir,
+       [OP_READLINK]           = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_REMOVE]             = (nfsd4_dec)nfsd4_decode_remove,
+       [OP_RENAME]             = (nfsd4_dec)nfsd4_decode_rename,
+       [OP_RENEW]              = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_RESTOREFH]          = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_SAVEFH]             = (nfsd4_dec)nfsd4_decode_noop,
+       [OP_SECINFO]            = (nfsd4_dec)nfsd4_decode_secinfo,
+       [OP_SETATTR]            = (nfsd4_dec)nfsd4_decode_setattr,
+       [OP_SETCLIENTID]        = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_SETCLIENTID_CONFIRM]= (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_VERIFY]             = (nfsd4_dec)nfsd4_decode_verify,
+       [OP_WRITE]              = (nfsd4_dec)nfsd4_decode_write,
+       [OP_RELEASE_LOCKOWNER]  = (nfsd4_dec)nfsd4_decode_notsupp,
+
+       /* new operations for NFSv4.1 */
+       [OP_BACKCHANNEL_CTL]    = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_BIND_CONN_TO_SESSION]= (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_EXCHANGE_ID]        = (nfsd4_dec)nfsd4_decode_exchange_id,
+       [OP_CREATE_SESSION]     = (nfsd4_dec)nfsd4_decode_create_session,
+       [OP_DESTROY_SESSION]    = (nfsd4_dec)nfsd4_decode_destroy_session,
+       [OP_FREE_STATEID]       = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_GET_DIR_DELEGATION] = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_GETDEVICEINFO]      = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_GETDEVICELIST]      = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_LAYOUTCOMMIT]       = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_LAYOUTGET]          = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_LAYOUTRETURN]       = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_SECINFO_NO_NAME]    = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_SEQUENCE]           = (nfsd4_dec)nfsd4_decode_sequence,
+       [OP_SET_SSV]            = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_TEST_STATEID]       = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_WANT_DELEGATION]    = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_DESTROY_CLIENTID]   = (nfsd4_dec)nfsd4_decode_notsupp,
+       [OP_RECLAIM_COMPLETE]   = (nfsd4_dec)nfsd4_decode_notsupp,
+};
+
+struct nfsd4_minorversion_ops {
+       nfsd4_dec *decoders;
+       int nops;
+};
+
+static struct nfsd4_minorversion_ops nfsd4_minorversion[] = {
+       [0] = { nfsd4_dec_ops, ARRAY_SIZE(nfsd4_dec_ops) },
+       [1] = { nfsd41_dec_ops, ARRAY_SIZE(nfsd41_dec_ops) },
+};
+
+static __be32
 nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
 {
        DECODE_HEAD;
        struct nfsd4_op *op;
+       struct nfsd4_minorversion_ops *ops;
        int i;
 
        /*
@@ -1019,11 +1388,15 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
                argp->ops = kmalloc(argp->opcnt * sizeof(*argp->ops), GFP_KERNEL);
                if (!argp->ops) {
                        argp->ops = argp->iops;
-                       printk(KERN_INFO "nfsd: couldn't allocate room for COMPOUND\n");
+                       dprintk("nfsd: couldn't allocate room for COMPOUND\n");
                        goto xdr_error;
                }
        }
 
+       if (argp->minorversion >= ARRAY_SIZE(nfsd4_minorversion))
+               argp->opcnt = 0;
+
+       ops = &nfsd4_minorversion[argp->minorversion];
        for (i = 0; i < argp->opcnt; i++) {
                op = &argp->ops[i];
                op->replay = NULL;
@@ -1061,120 +1434,11 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
                }
                op->opnum = ntohl(*argp->p++);
 
-               switch (op->opnum) {
-               case 2: /* Reserved operation */
-                       op->opnum = OP_ILLEGAL;
-                       if (argp->minorversion == 0)
-                               op->status = nfserr_op_illegal;
-                       else
-                               op->status = nfserr_minor_vers_mismatch;
-                       break;
-               case OP_ACCESS:
-                       op->status = nfsd4_decode_access(argp, &op->u.access);
-                       break;
-               case OP_CLOSE:
-                       op->status = nfsd4_decode_close(argp, &op->u.close);
-                       break;
-               case OP_COMMIT:
-                       op->status = nfsd4_decode_commit(argp, &op->u.commit);
-                       break;
-               case OP_CREATE:
-                       op->status = nfsd4_decode_create(argp, &op->u.create);
-                       break;
-               case OP_DELEGRETURN:
-                       op->status = nfsd4_decode_delegreturn(argp, &op->u.delegreturn);
-                       break;
-               case OP_GETATTR:
-                       op->status = nfsd4_decode_getattr(argp, &op->u.getattr);
-                       break;
-               case OP_GETFH:
-                       op->status = nfs_ok;
-                       break;
-               case OP_LINK:
-                       op->status = nfsd4_decode_link(argp, &op->u.link);
-                       break;
-               case OP_LOCK:
-                       op->status = nfsd4_decode_lock(argp, &op->u.lock);
-                       break;
-               case OP_LOCKT:
-                       op->status = nfsd4_decode_lockt(argp, &op->u.lockt);
-                       break;
-               case OP_LOCKU:
-                       op->status = nfsd4_decode_locku(argp, &op->u.locku);
-                       break;
-               case OP_LOOKUP:
-                       op->status = nfsd4_decode_lookup(argp, &op->u.lookup);
-                       break;
-               case OP_LOOKUPP:
-                       op->status = nfs_ok;
-                       break;
-               case OP_NVERIFY:
-                       op->status = nfsd4_decode_verify(argp, &op->u.nverify);
-                       break;
-               case OP_OPEN:
-                       op->status = nfsd4_decode_open(argp, &op->u.open);
-                       break;
-               case OP_OPEN_CONFIRM:
-                       op->status = nfsd4_decode_open_confirm(argp, &op->u.open_confirm);
-                       break;
-               case OP_OPEN_DOWNGRADE:
-                       op->status = nfsd4_decode_open_downgrade(argp, &op->u.open_downgrade);
-                       break;
-               case OP_PUTFH:
-                       op->status = nfsd4_decode_putfh(argp, &op->u.putfh);
-                       break;
-               case OP_PUTROOTFH:
-                       op->status = nfs_ok;
-                       break;
-               case OP_READ:
-                       op->status = nfsd4_decode_read(argp, &op->u.read);
-                       break;
-               case OP_READDIR:
-                       op->status = nfsd4_decode_readdir(argp, &op->u.readdir);
-                       break;
-               case OP_READLINK:
-                       op->status = nfs_ok;
-                       break;
-               case OP_REMOVE:
-                       op->status = nfsd4_decode_remove(argp, &op->u.remove);
-                       break;
-               case OP_RENAME:
-                       op->status = nfsd4_decode_rename(argp, &op->u.rename);
-                       break;
-               case OP_RESTOREFH:
-                       op->status = nfs_ok;
-                       break;
-               case OP_RENEW:
-                       op->status = nfsd4_decode_renew(argp, &op->u.renew);
-                       break;
-               case OP_SAVEFH:
-                       op->status = nfs_ok;
-                       break;
-               case OP_SECINFO:
-                       op->status = nfsd4_decode_secinfo(argp, &op->u.secinfo);
-                       break;
-               case OP_SETATTR:
-                       op->status = nfsd4_decode_setattr(argp, &op->u.setattr);
-                       break;
-               case OP_SETCLIENTID:
-                       op->status = nfsd4_decode_setclientid(argp, &op->u.setclientid);
-                       break;
-               case OP_SETCLIENTID_CONFIRM:
-                       op->status = nfsd4_decode_setclientid_confirm(argp, &op->u.setclientid_confirm);
-                       break;
-               case OP_VERIFY:
-                       op->status = nfsd4_decode_verify(argp, &op->u.verify);
-                       break;
-               case OP_WRITE:
-                       op->status = nfsd4_decode_write(argp, &op->u.write);
-                       break;
-               case OP_RELEASE_LOCKOWNER:
-                       op->status = nfsd4_decode_release_lockowner(argp, &op->u.release_lockowner);
-                       break;
-               default:
+               if (op->opnum >= FIRST_NFS4_OP && op->opnum <= LAST_NFS4_OP)
+                       op->status = ops->decoders[op->opnum](argp, &op->u);
+               else {
                        op->opnum = OP_ILLEGAL;
                        op->status = nfserr_op_illegal;
-                       break;
                }
 
                if (op->status) {
@@ -1185,39 +1449,52 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
 
        DECODE_TAIL;
 }
-/*
- * END OF "GENERIC" DECODE ROUTINES.
- */
-
-/*
- * START OF "GENERIC" ENCODE ROUTINES.
- *   These may look a little ugly since they are imported from a "generic"
- * set of XDR encode/decode routines which are intended to be shared by
- * all of our NFSv4 implementations (OpenBSD, MacOS X...).
- *
- * If the pain of reading these is too great, it should be a straightforward
- * task to translate them into Linux-specific versions which are more
- * consistent with the style used in NFSv2/v3...
- */
-#define ENCODE_HEAD              __be32 *p
 
 #define WRITE32(n)               *p++ = htonl(n)
 #define WRITE64(n)               do {                          \
        *p++ = htonl((u32)((n) >> 32));                         \
        *p++ = htonl((u32)(n));                                 \
 } while (0)
-#define WRITEMEM(ptr,nbytes)     do {                          \
+#define WRITEMEM(ptr,nbytes)     do { if (nbytes > 0) {                \
        *(p + XDR_QUADLEN(nbytes) -1) = 0;                      \
        memcpy(p, ptr, nbytes);                                 \
        p += XDR_QUADLEN(nbytes);                               \
-} while (0)
-#define WRITECINFO(c)          do {                            \
-       *p++ = htonl(c.atomic);                                 \
-       *p++ = htonl(c.before_ctime_sec);                               \
-       *p++ = htonl(c.before_ctime_nsec);                              \
-       *p++ = htonl(c.after_ctime_sec);                                \
-       *p++ = htonl(c.after_ctime_nsec);                               \
-} while (0)
+}} while (0)
+
+static void write32(__be32 **p, u32 n)
+{
+       *(*p)++ = n;
+}
+
+static void write64(__be32 **p, u64 n)
+{
+       write32(p, (u32)(n >> 32));
+       write32(p, (u32)n);
+}
+
+static void write_change(__be32 **p, struct kstat *stat, struct inode *inode)
+{
+       if (IS_I_VERSION(inode)) {
+               write64(p, inode->i_version);
+       } else {
+               write32(p, stat->ctime.tv_sec);
+               write32(p, stat->ctime.tv_nsec);
+       }
+}
+
+static void write_cinfo(__be32 **p, struct nfsd4_change_info *c)
+{
+       write32(p, c->atomic);
+       if (c->change_supported) {
+               write64(p, c->before_change);
+               write64(p, c->after_change);
+       } else {
+               write32(p, c->before_ctime_sec);
+               write32(p, c->before_ctime_nsec);
+               write32(p, c->after_ctime_sec);
+               write32(p, c->after_ctime_nsec);
+       }
+}
 
 #define RESERVE_SPACE(nbytes)  do {                            \
        p = resp->p;                                            \
@@ -1229,7 +1506,6 @@ nfsd4_decode_compound(struct nfsd4_compoundargs *argp)
  * Header routine to setup seqid operation replay cache
  */
 #define ENCODE_SEQID_OP_HEAD                                   \
-       __be32 *p;                                              \
        __be32 *save;                                           \
                                                                \
        save = resp->p;
@@ -1315,24 +1591,29 @@ static __be32 nfsd4_encode_fs_location4(struct nfsd4_fs_location *location,
 static char *nfsd4_path(struct svc_rqst *rqstp, struct svc_export *exp, __be32 *stat)
 {
        struct svc_fh tmp_fh;
-       char *path, *rootpath;
+       char *path = NULL, *rootpath;
+       size_t rootlen;
 
        fh_init(&tmp_fh, NFS4_FHSIZE);
        *stat = exp_pseudoroot(rqstp, &tmp_fh);
        if (*stat)
                return NULL;
-       rootpath = tmp_fh.fh_export->ex_path;
+       rootpath = tmp_fh.fh_export->ex_pathname;
 
-       path = exp->ex_path;
+       path = exp->ex_pathname;
 
-       if (strncmp(path, rootpath, strlen(rootpath))) {
-               printk("nfsd: fs_locations failed;"
+       rootlen = strlen(rootpath);
+       if (strncmp(path, rootpath, rootlen)) {
+               dprintk("nfsd: fs_locations failed;"
                        "%s is not contained in %s\n", path, rootpath);
                *stat = nfserr_notsupp;
-               return NULL;
+               path = NULL;
+               goto out;
        }
-
-       return path + strlen(rootpath);
+       path += rootlen;
+out:
+       fh_put(&tmp_fh);
+       return path;
 }
 
 /*
@@ -1444,10 +1725,11 @@ static __be32 fattr_handle_absent_fs(u32 *bmval0, u32 *bmval1, u32 *rdattr_err)
 __be32
 nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
                struct dentry *dentry, __be32 *buffer, int *countp, u32 *bmval,
-               struct svc_rqst *rqstp)
+               struct svc_rqst *rqstp, int ignore_crossmnt)
 {
        u32 bmval0 = bmval[0];
        u32 bmval1 = bmval[1];
+       u32 bmval2 = bmval[2];
        struct kstat stat;
        struct svc_fh tempfh;
        struct kstatfs statfs;
@@ -1461,18 +1743,22 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
        int err;
        int aclsupport = 0;
        struct nfs4_acl *acl = NULL;
+       struct nfsd4_compoundres *resp = rqstp->rq_resp;
+       u32 minorversion = resp->cstate.minorversion;
 
        BUG_ON(bmval1 & NFSD_WRITEONLY_ATTRS_WORD1);
-       BUG_ON(bmval0 & ~NFSD_SUPPORTED_ATTRS_WORD0);
-       BUG_ON(bmval1 & ~NFSD_SUPPORTED_ATTRS_WORD1);
+       BUG_ON(bmval0 & ~nfsd_suppattrs0(minorversion));
+       BUG_ON(bmval1 & ~nfsd_suppattrs1(minorversion));
+       BUG_ON(bmval2 & ~nfsd_suppattrs2(minorversion));
 
        if (exp->ex_fslocs.migrated) {
+               BUG_ON(bmval[2]);
                status = fattr_handle_absent_fs(&bmval0, &bmval1, &rdattr_err);
                if (status)
                        goto out;
        }
 
-       err = vfs_getattr(exp->ex_mnt, dentry, &stat);
+       err = vfs_getattr(exp->ex_path.mnt, dentry, &stat);
        if (err)
                goto out_nfserr;
        if ((bmval0 & (FATTR4_WORD0_FILES_FREE | FATTR4_WORD0_FILES_TOTAL |
@@ -1504,30 +1790,43 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
                                goto out_nfserr;
                }
        }
-       if (bmval0 & FATTR4_WORD0_FS_LOCATIONS) {
-               if (exp->ex_fslocs.locations == NULL) {
-                       bmval0 &= ~FATTR4_WORD0_FS_LOCATIONS;
-               }
-       }
        if ((buflen -= 16) < 0)
                goto out_resource;
 
-       WRITE32(2);
-       WRITE32(bmval0);
-       WRITE32(bmval1);
+       if (unlikely(bmval2)) {
+               WRITE32(3);
+               WRITE32(bmval0);
+               WRITE32(bmval1);
+               WRITE32(bmval2);
+       } else if (likely(bmval1)) {
+               WRITE32(2);
+               WRITE32(bmval0);
+               WRITE32(bmval1);
+       } else {
+               WRITE32(1);
+               WRITE32(bmval0);
+       }
        attrlenp = p++;                /* to be backfilled later */
 
        if (bmval0 & FATTR4_WORD0_SUPPORTED_ATTRS) {
-               u32 word0 = NFSD_SUPPORTED_ATTRS_WORD0;
+               u32 word0 = nfsd_suppattrs0(minorversion);
+               u32 word1 = nfsd_suppattrs1(minorversion);
+               u32 word2 = nfsd_suppattrs2(minorversion);
+
                if ((buflen -= 12) < 0)
                        goto out_resource;
                if (!aclsupport)
                        word0 &= ~FATTR4_WORD0_ACL;
-               if (!exp->ex_fslocs.locations)
-                       word0 &= ~FATTR4_WORD0_FS_LOCATIONS;
-               WRITE32(2);
-               WRITE32(word0);
-               WRITE32(NFSD_SUPPORTED_ATTRS_WORD1);
+               if (!word2) {
+                       WRITE32(2);
+                       WRITE32(word0);
+                       WRITE32(word1);
+               } else {
+                       WRITE32(3);
+                       WRITE32(word0);
+                       WRITE32(word1);
+                       WRITE32(word2);
+               }
        }
        if (bmval0 & FATTR4_WORD0_TYPE) {
                if ((buflen -= 4) < 0)
@@ -1546,16 +1845,9 @@ nfsd4_encode_fattr(struct svc_fh *fhp, struct svc_export *exp,
                        WRITE32(NFS4_FH_PERSISTENT|NFS4_FH_VOL_RENAME);
        }
        if (bmval0 & FATTR4_WORD0_CHANGE) {
-               /*
-                * Note: This _must_ be consistent with the scheme for writing
-                * change_info, so any changes made here must be reflected there
-                * as well.  (See xdr4.h:set_change_info() and the WRITECINFO()
-                * macro above.)
-                */
                if ((buflen -= 8) < 0)
                        goto out_resource;
-               WRITE32(stat.ctime.tv_sec);
-               WRITE32(stat.ctime.tv_nsec);
+               write_change(&p, &stat, dentry->d_inode);
        }
        if (bmval0 & FATTR4_WORD0_SIZE) {
                if ((buflen -= 8) < 0)
@@ -1824,14 +2116,32 @@ out_acl:
        if (bmval1 & FATTR4_WORD1_MOUNTED_ON_FILEID) {
                if ((buflen -= 8) < 0)
                        goto out_resource;
-               if (exp->ex_mnt->mnt_root->d_inode == dentry->d_inode) {
-                       err = vfs_getattr(exp->ex_mnt->mnt_parent,
-                               exp->ex_mnt->mnt_mountpoint, &stat);
+               /*
+                * Get parent's attributes if not ignoring crossmount
+                * and this is the root of a cross-mounted filesystem.
+                */
+               if (ignore_crossmnt == 0 &&
+                   dentry == exp->ex_path.mnt->mnt_root) {
+                       struct path path = exp->ex_path;
+                       path_get(&path);
+                       while (follow_up(&path)) {
+                               if (path.dentry != path.mnt->mnt_root)
+                                       break;
+                       }
+                       err = vfs_getattr(path.mnt, path.dentry, &stat);
+                       path_put(&path);
                        if (err)
                                goto out_nfserr;
                }
                WRITE64(stat.ino);
        }
+       if (bmval2 & FATTR4_WORD2_SUPPATTR_EXCLCREAT) {
+               WRITE32(3);
+               WRITE32(NFSD_SUPPATTR_EXCLCREAT_WORD0);
+               WRITE32(NFSD_SUPPATTR_EXCLCREAT_WORD1);
+               WRITE32(NFSD_SUPPATTR_EXCLCREAT_WORD2);
+       }
+
        *attrlenp = htonl((char *)p - (char *)attrlenp - 4);
        *countp = p - buffer;
        status = nfs_ok;
@@ -1853,6 +2163,15 @@ out_serverfault:
        goto out;
 }
 
+static inline int attributes_need_mount(u32 *bmval)
+{
+       if (bmval[0] & ~(FATTR4_WORD0_RDATTR_ERROR | FATTR4_WORD0_LEASE_TIME))
+               return 1;
+       if (bmval[1] & ~FATTR4_WORD1_MOUNTED_ON_FILEID)
+               return 1;
+       return 0;
+}
+
 static __be32
 nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,
                const char *name, int namlen, __be32 *p, int *buflen)
@@ -1860,15 +2179,37 @@ nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,
        struct svc_export *exp = cd->rd_fhp->fh_export;
        struct dentry *dentry;
        __be32 nfserr;
+       int ignore_crossmnt = 0;
 
        dentry = lookup_one_len(name, cd->rd_fhp->fh_dentry, namlen);
        if (IS_ERR(dentry))
                return nfserrno(PTR_ERR(dentry));
+       if (!dentry->d_inode) {
+               /*
+                * nfsd_buffered_readdir drops the i_mutex between
+                * readdir and calling this callback, leaving a window
+                * where this directory entry could have gone away.
+                */
+               dput(dentry);
+               return nfserr_noent;
+       }
 
        exp_get(exp);
-       if (d_mountpoint(dentry)) {
+       /*
+        * In the case of a mountpoint, the client may be asking for
+        * attributes that are only properties of the underlying filesystem
+        * as opposed to the cross-mounted file system. In such a case,
+        * we will not follow the cross mount and will fill the attribtutes
+        * directly from the mountpoint dentry.
+        */
+       if (nfsd_mountpoint(dentry, exp)) {
                int err;
 
+               if (!(exp->ex_flags & NFSEXP_V4ROOT)
+                               && !attributes_need_mount(cd->rd_bmval)) {
+                       ignore_crossmnt = 1;
+                       goto out_encode;
+               }
                /*
                 * Why the heck aren't we just using nfsd_lookup??
                 * Different "."/".." handling?  Something else?
@@ -1884,8 +2225,9 @@ nfsd4_encode_dirent_fattr(struct nfsd4_readdir *cd,
                        goto out_put;
 
        }
+out_encode:
        nfserr = nfsd4_encode_fattr(NULL, exp, dentry, p, buflen, cd->rd_bmval,
-                                       cd->rd_rqstp);
+                                       cd->rd_rqstp, ignore_crossmnt);
 out_put:
        dput(dentry);
        exp_put(exp);
@@ -1917,6 +2259,7 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
        struct nfsd4_readdir *cd = container_of(ccd, struct nfsd4_readdir, common);
        int buflen;
        __be32 *p = cd->buffer;
+       __be32 *cookiep;
        __be32 nfserr = nfserr_toosmall;
 
        /* In nfsv4, "." and ".." never make it onto the wire.. */
@@ -1933,7 +2276,7 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
                goto fail;
 
        *p++ = xdr_one;                             /* mark entry present */
-       cd->offset = p;                             /* remember pointer */
+       cookiep = p;
        p = xdr_encode_hyper(p, NFS_OFFSET_MAX);    /* offset of next entry */
        p = xdr_encode_array(p, name, namlen);      /* name length & name */
 
@@ -1947,6 +2290,8 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
                goto fail;
        case nfserr_dropit:
                goto fail;
+       case nfserr_noent:
+               goto skip_entry;
        default:
                /*
                 * If the client requested the RDATTR_ERROR attribute,
@@ -1965,6 +2310,8 @@ nfsd4_encode_dirent(void *ccdv, const char *name, int namlen,
        }
        cd->buflen -= (p - cd->buffer);
        cd->buffer = p;
+       cd->offset = cookiep;
+skip_entry:
        cd->common.err = nfs_ok;
        return 0;
 fail:
@@ -1973,9 +2320,20 @@ fail:
 }
 
 static void
+nfsd4_encode_stateid(struct nfsd4_compoundres *resp, stateid_t *sid)
+{
+       __be32 *p;
+
+       RESERVE_SPACE(sizeof(stateid_t));
+       WRITE32(sid->si_generation);
+       WRITEMEM(&sid->si_opaque, sizeof(stateid_opaque_t));
+       ADJUST_ARGS();
+}
+
+static __be32
 nfsd4_encode_access(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_access *access)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(8);
@@ -1983,48 +2341,49 @@ nfsd4_encode_access(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_
                WRITE32(access->ac_resp_access);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_close(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_close *close)
 {
        ENCODE_SEQID_OP_HEAD;
 
-       if (!nfserr) {
-               RESERVE_SPACE(sizeof(stateid_t));
-               WRITE32(close->cl_stateid.si_generation);
-               WRITEMEM(&close->cl_stateid.si_opaque, sizeof(stateid_opaque_t));
-               ADJUST_ARGS();
-       }
+       if (!nfserr)
+               nfsd4_encode_stateid(resp, &close->cl_stateid);
+
        ENCODE_SEQID_OP_TAIL(close->cl_stateowner);
+       return nfserr;
 }
 
 
-static void
+static __be32
 nfsd4_encode_commit(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_commit *commit)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(8);
                WRITEMEM(commit->co_verf.data, 8);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_create(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_create *create)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(32);
-               WRITECINFO(create->cr_cinfo);
+               write_cinfo(&p, &create->cr_cinfo);
                WRITE32(2);
                WRITE32(create->cr_bmval[0]);
                WRITE32(create->cr_bmval[1]);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
 static __be32
@@ -2039,17 +2398,18 @@ nfsd4_encode_getattr(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4
        buflen = resp->end - resp->p - (COMPOUND_ERR_SLACK_SPACE >> 2);
        nfserr = nfsd4_encode_fattr(fhp, fhp->fh_export, fhp->fh_dentry,
                                    resp->p, &buflen, getattr->ga_bmval,
-                                   resp->rqstp);
+                                   resp->rqstp, 0);
        if (!nfserr)
                resp->p += buflen;
        return nfserr;
 }
 
-static void
-nfsd4_encode_getfh(struct nfsd4_compoundres *resp, __be32 nfserr, struct svc_fh *fhp)
+static __be32
+nfsd4_encode_getfh(struct nfsd4_compoundres *resp, __be32 nfserr, struct svc_fh **fhpp)
 {
+       struct svc_fh *fhp = *fhpp;
        unsigned int len;
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                len = fhp->fh_handle.fh_size;
@@ -2058,6 +2418,7 @@ nfsd4_encode_getfh(struct nfsd4_compoundres *resp, __be32 nfserr, struct svc_fh
                WRITEMEM(&fhp->fh_handle.fh_base, len);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
 /*
@@ -2067,7 +2428,7 @@ nfsd4_encode_getfh(struct nfsd4_compoundres *resp, __be32 nfserr, struct svc_fh
 static void
 nfsd4_encode_lock_denied(struct nfsd4_compoundres *resp, struct nfsd4_lock_denied *ld)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        RESERVE_SPACE(32 + XDR_LEN(ld->ld_sop ? ld->ld_sop->so_owner.len : 0));
        WRITE64(ld->ld_start);
@@ -2085,70 +2446,67 @@ nfsd4_encode_lock_denied(struct nfsd4_compoundres *resp, struct nfsd4_lock_denie
        ADJUST_ARGS();
 }
 
-static void
+static __be32
 nfsd4_encode_lock(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_lock *lock)
 {
        ENCODE_SEQID_OP_HEAD;
 
-       if (!nfserr) {
-               RESERVE_SPACE(4 + sizeof(stateid_t));
-               WRITE32(lock->lk_resp_stateid.si_generation);
-               WRITEMEM(&lock->lk_resp_stateid.si_opaque, sizeof(stateid_opaque_t));
-               ADJUST_ARGS();
-       } else if (nfserr == nfserr_denied)
+       if (!nfserr)
+               nfsd4_encode_stateid(resp, &lock->lk_resp_stateid);
+       else if (nfserr == nfserr_denied)
                nfsd4_encode_lock_denied(resp, &lock->lk_denied);
 
        ENCODE_SEQID_OP_TAIL(lock->lk_replay_owner);
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_lockt(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_lockt *lockt)
 {
        if (nfserr == nfserr_denied)
                nfsd4_encode_lock_denied(resp, &lockt->lt_denied);
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_locku(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_locku *locku)
 {
        ENCODE_SEQID_OP_HEAD;
 
-       if (!nfserr) {
-               RESERVE_SPACE(sizeof(stateid_t));
-               WRITE32(locku->lu_stateid.si_generation);
-               WRITEMEM(&locku->lu_stateid.si_opaque, sizeof(stateid_opaque_t));
-               ADJUST_ARGS();
-       }
-                                       
+       if (!nfserr)
+               nfsd4_encode_stateid(resp, &locku->lu_stateid);
+
        ENCODE_SEQID_OP_TAIL(locku->lu_stateowner);
+       return nfserr;
 }
 
 
-static void
+static __be32
 nfsd4_encode_link(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_link *link)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(20);
-               WRITECINFO(link->li_cinfo);
+               write_cinfo(&p, &link->li_cinfo);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
 
-static void
+static __be32
 nfsd4_encode_open(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_open *open)
 {
+       __be32 *p;
        ENCODE_SEQID_OP_HEAD;
 
        if (nfserr)
                goto out;
 
-       RESERVE_SPACE(36 + sizeof(stateid_t));
-       WRITE32(open->op_stateid.si_generation);
-       WRITEMEM(&open->op_stateid.si_opaque, sizeof(stateid_opaque_t));
-       WRITECINFO(open->op_cinfo);
+       nfsd4_encode_stateid(resp, &open->op_stateid);
+       RESERVE_SPACE(40);
+       write_cinfo(&p, &open->op_cinfo);
        WRITE32(open->op_rflags);
        WRITE32(2);
        WRITE32(open->op_bmval[0]);
@@ -2160,8 +2518,8 @@ nfsd4_encode_open(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_op
        case NFS4_OPEN_DELEGATE_NONE:
                break;
        case NFS4_OPEN_DELEGATE_READ:
-               RESERVE_SPACE(20 + sizeof(stateid_t));
-               WRITEMEM(&open->op_delegate_stateid, sizeof(stateid_t));
+               nfsd4_encode_stateid(resp, &open->op_delegate_stateid);
+               RESERVE_SPACE(20);
                WRITE32(open->op_recall);
 
                /*
@@ -2174,8 +2532,8 @@ nfsd4_encode_open(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_op
                ADJUST_ARGS();
                break;
        case NFS4_OPEN_DELEGATE_WRITE:
-               RESERVE_SPACE(32 + sizeof(stateid_t));
-               WRITEMEM(&open->op_delegate_stateid, sizeof(stateid_t));
+               nfsd4_encode_stateid(resp, &open->op_delegate_stateid);
+               RESERVE_SPACE(32);
                WRITE32(0);
 
                /*
@@ -2200,36 +2558,31 @@ nfsd4_encode_open(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_op
        /* XXX save filehandle here */
 out:
        ENCODE_SEQID_OP_TAIL(open->op_stateowner);
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_open_confirm(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_open_confirm *oc)
 {
        ENCODE_SEQID_OP_HEAD;
-                                       
-       if (!nfserr) {
-               RESERVE_SPACE(sizeof(stateid_t));
-               WRITE32(oc->oc_resp_stateid.si_generation);
-               WRITEMEM(&oc->oc_resp_stateid.si_opaque, sizeof(stateid_opaque_t));
-               ADJUST_ARGS();
-       }
+
+       if (!nfserr)
+               nfsd4_encode_stateid(resp, &oc->oc_resp_stateid);
 
        ENCODE_SEQID_OP_TAIL(oc->oc_stateowner);
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_open_downgrade(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_open_downgrade *od)
 {
        ENCODE_SEQID_OP_HEAD;
-                                       
-       if (!nfserr) {
-               RESERVE_SPACE(sizeof(stateid_t));
-               WRITE32(od->od_stateid.si_generation);
-               WRITEMEM(&od->od_stateid.si_opaque, sizeof(stateid_opaque_t));
-               ADJUST_ARGS();
-       }
+
+       if (!nfserr)
+               nfsd4_encode_stateid(resp, &od->od_stateid);
 
        ENCODE_SEQID_OP_TAIL(od->od_stateowner);
+       return nfserr;
 }
 
 static __be32
@@ -2240,7 +2593,7 @@ nfsd4_encode_read(struct nfsd4_compoundres *resp, __be32 nfserr,
        int v, pn;
        unsigned long maxcount; 
        long len;
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (nfserr)
                return nfserr;
@@ -2302,7 +2655,7 @@ nfsd4_encode_readlink(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd
 {
        int maxcount;
        char *page;
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (nfserr)
                return nfserr;
@@ -2351,7 +2704,7 @@ nfsd4_encode_readdir(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4
        int maxcount;
        loff_t offset;
        __be32 *page, *savep, *tailbase;
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (nfserr)
                return nfserr;
@@ -2424,32 +2777,34 @@ err_no_verf:
        return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_remove(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_remove *remove)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(20);
-               WRITECINFO(remove->rm_cinfo);
+               write_cinfo(&p, &remove->rm_cinfo);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_rename(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_rename *rename)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(40);
-               WRITECINFO(rename->rn_sinfo);
-               WRITECINFO(rename->rn_tinfo);
+               write_cinfo(&p, &rename->rn_sinfo);
+               write_cinfo(&p, &rename->rn_tinfo);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_secinfo(struct nfsd4_compoundres *resp, __be32 nfserr,
                     struct nfsd4_secinfo *secinfo)
 {
@@ -2458,7 +2813,7 @@ nfsd4_encode_secinfo(struct nfsd4_compoundres *resp, __be32 nfserr,
        u32 nflavs;
        struct exp_flavor_info *flavs;
        struct exp_flavor_info def_flavs[2];
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (nfserr)
                goto out;
@@ -2513,16 +2868,17 @@ nfsd4_encode_secinfo(struct nfsd4_compoundres *resp, __be32 nfserr,
 out:
        if (exp)
                exp_put(exp);
+       return nfserr;
 }
 
 /*
  * The SETATTR encode routine is special -- it always encodes a bitmap,
  * regardless of the error status.
  */
-static void
+static __be32
 nfsd4_encode_setattr(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_setattr *setattr)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        RESERVE_SPACE(12);
        if (nfserr) {
@@ -2536,12 +2892,13 @@ nfsd4_encode_setattr(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4
                WRITE32(setattr->sa_bmval[1]);
        }
        ADJUST_ARGS();
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_setclientid(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_setclientid *scd)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(8 + sizeof(nfs4_verifier));
@@ -2555,12 +2912,13 @@ nfsd4_encode_setclientid(struct nfsd4_compoundres *resp, __be32 nfserr, struct n
                WRITE32(0);
                ADJUST_ARGS();
        }
+       return nfserr;
 }
 
-static void
+static __be32
 nfsd4_encode_write(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_write *write)
 {
-       ENCODE_HEAD;
+       __be32 *p;
 
        if (!nfserr) {
                RESERVE_SPACE(16);
@@ -2569,114 +2927,289 @@ nfsd4_encode_write(struct nfsd4_compoundres *resp, __be32 nfserr, struct nfsd4_w
                WRITEMEM(write->wr_verifier.data, 8);
                ADJUST_ARGS();
        }
+       return nfserr;
+}
+
+static __be32
+nfsd4_encode_exchange_id(struct nfsd4_compoundres *resp, int nfserr,
+                        struct nfsd4_exchange_id *exid)
+{
+       __be32 *p;
+       char *major_id;
+       char *server_scope;
+       int major_id_sz;
+       int server_scope_sz;
+       uint64_t minor_id = 0;
+
+       if (nfserr)
+               return nfserr;
+
+       major_id = utsname()->nodename;
+       major_id_sz = strlen(major_id);
+       server_scope = utsname()->nodename;
+       server_scope_sz = strlen(server_scope);
+
+       RESERVE_SPACE(
+               8 /* eir_clientid */ +
+               4 /* eir_sequenceid */ +
+               4 /* eir_flags */ +
+               4 /* spr_how (SP4_NONE) */ +
+               8 /* so_minor_id */ +
+               4 /* so_major_id.len */ +
+               (XDR_QUADLEN(major_id_sz) * 4) +
+               4 /* eir_server_scope.len */ +
+               (XDR_QUADLEN(server_scope_sz) * 4) +
+               4 /* eir_server_impl_id.count (0) */);
+
+       WRITEMEM(&exid->clientid, 8);
+       WRITE32(exid->seqid);
+       WRITE32(exid->flags);
+
+       /* state_protect4_r. Currently only support SP4_NONE */
+       BUG_ON(exid->spa_how != SP4_NONE);
+       WRITE32(exid->spa_how);
+
+       /* The server_owner struct */
+       WRITE64(minor_id);      /* Minor id */
+       /* major id */
+       WRITE32(major_id_sz);
+       WRITEMEM(major_id, major_id_sz);
+
+       /* Server scope */
+       WRITE32(server_scope_sz);
+       WRITEMEM(server_scope, server_scope_sz);
+
+       /* Implementation id */
+       WRITE32(0);     /* zero length nfs_impl_id4 array */
+       ADJUST_ARGS();
+       return 0;
+}
+
+static __be32
+nfsd4_encode_create_session(struct nfsd4_compoundres *resp, int nfserr,
+                           struct nfsd4_create_session *sess)
+{
+       __be32 *p;
+
+       if (nfserr)
+               return nfserr;
+
+       RESERVE_SPACE(24);
+       WRITEMEM(sess->sessionid.data, NFS4_MAX_SESSIONID_LEN);
+       WRITE32(sess->seqid);
+       WRITE32(sess->flags);
+       ADJUST_ARGS();
+
+       RESERVE_SPACE(28);
+       WRITE32(0); /* headerpadsz */
+       WRITE32(sess->fore_channel.maxreq_sz);
+       WRITE32(sess->fore_channel.maxresp_sz);
+       WRITE32(sess->fore_channel.maxresp_cached);
+       WRITE32(sess->fore_channel.maxops);
+       WRITE32(sess->fore_channel.maxreqs);
+       WRITE32(sess->fore_channel.nr_rdma_attrs);
+       ADJUST_ARGS();
+
+       if (sess->fore_channel.nr_rdma_attrs) {
+               RESERVE_SPACE(4);
+               WRITE32(sess->fore_channel.rdma_attrs);
+               ADJUST_ARGS();
+       }
+
+       RESERVE_SPACE(28);
+       WRITE32(0); /* headerpadsz */
+       WRITE32(sess->back_channel.maxreq_sz);
+       WRITE32(sess->back_channel.maxresp_sz);
+       WRITE32(sess->back_channel.maxresp_cached);
+       WRITE32(sess->back_channel.maxops);
+       WRITE32(sess->back_channel.maxreqs);
+       WRITE32(sess->back_channel.nr_rdma_attrs);
+       ADJUST_ARGS();
+
+       if (sess->back_channel.nr_rdma_attrs) {
+               RESERVE_SPACE(4);
+               WRITE32(sess->back_channel.rdma_attrs);
+               ADJUST_ARGS();
+       }
+       return 0;
+}
+
+static __be32
+nfsd4_encode_destroy_session(struct nfsd4_compoundres *resp, int nfserr,
+                            struct nfsd4_destroy_session *destroy_session)
+{
+       return nfserr;
+}
+
+__be32
+nfsd4_encode_sequence(struct nfsd4_compoundres *resp, int nfserr,
+                     struct nfsd4_sequence *seq)
+{
+       __be32 *p;
+
+       if (nfserr)
+               return nfserr;
+
+       RESERVE_SPACE(NFS4_MAX_SESSIONID_LEN + 20);
+       WRITEMEM(seq->sessionid.data, NFS4_MAX_SESSIONID_LEN);
+       WRITE32(seq->seqid);
+       WRITE32(seq->slotid);
+       WRITE32(seq->maxslots);
+       /*
+        * FIXME: for now:
+        *   target_maxslots = maxslots
+        *   status_flags = 0
+        */
+       WRITE32(seq->maxslots);
+       WRITE32(0);
+
+       ADJUST_ARGS();
+       resp->cstate.datap = p; /* DRC cache data pointer */
+       return 0;
+}
+
+static __be32
+nfsd4_encode_noop(struct nfsd4_compoundres *resp, __be32 nfserr, void *p)
+{
+       return nfserr;
+}
+
+typedef __be32(* nfsd4_enc)(struct nfsd4_compoundres *, __be32, void *);
+
+/*
+ * Note: nfsd4_enc_ops vector is shared for v4.0 and v4.1
+ * since we don't need to filter out obsolete ops as this is
+ * done in the decoding phase.
+ */
+static nfsd4_enc nfsd4_enc_ops[] = {
+       [OP_ACCESS]             = (nfsd4_enc)nfsd4_encode_access,
+       [OP_CLOSE]              = (nfsd4_enc)nfsd4_encode_close,
+       [OP_COMMIT]             = (nfsd4_enc)nfsd4_encode_commit,
+       [OP_CREATE]             = (nfsd4_enc)nfsd4_encode_create,
+       [OP_DELEGPURGE]         = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_DELEGRETURN]        = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_GETATTR]            = (nfsd4_enc)nfsd4_encode_getattr,
+       [OP_GETFH]              = (nfsd4_enc)nfsd4_encode_getfh,
+       [OP_LINK]               = (nfsd4_enc)nfsd4_encode_link,
+       [OP_LOCK]               = (nfsd4_enc)nfsd4_encode_lock,
+       [OP_LOCKT]              = (nfsd4_enc)nfsd4_encode_lockt,
+       [OP_LOCKU]              = (nfsd4_enc)nfsd4_encode_locku,
+       [OP_LOOKUP]             = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_LOOKUPP]            = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_NVERIFY]            = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_OPEN]               = (nfsd4_enc)nfsd4_encode_open,
+       [OP_OPENATTR]           = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_OPEN_CONFIRM]       = (nfsd4_enc)nfsd4_encode_open_confirm,
+       [OP_OPEN_DOWNGRADE]     = (nfsd4_enc)nfsd4_encode_open_downgrade,
+       [OP_PUTFH]              = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_PUTPUBFH]           = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_PUTROOTFH]          = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_READ]               = (nfsd4_enc)nfsd4_encode_read,
+       [OP_READDIR]            = (nfsd4_enc)nfsd4_encode_readdir,
+       [OP_READLINK]           = (nfsd4_enc)nfsd4_encode_readlink,
+       [OP_REMOVE]             = (nfsd4_enc)nfsd4_encode_remove,
+       [OP_RENAME]             = (nfsd4_enc)nfsd4_encode_rename,
+       [OP_RENEW]              = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_RESTOREFH]          = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_SAVEFH]             = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_SECINFO]            = (nfsd4_enc)nfsd4_encode_secinfo,
+       [OP_SETATTR]            = (nfsd4_enc)nfsd4_encode_setattr,
+       [OP_SETCLIENTID]        = (nfsd4_enc)nfsd4_encode_setclientid,
+       [OP_SETCLIENTID_CONFIRM] = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_VERIFY]             = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_WRITE]              = (nfsd4_enc)nfsd4_encode_write,
+       [OP_RELEASE_LOCKOWNER]  = (nfsd4_enc)nfsd4_encode_noop,
+
+       /* NFSv4.1 operations */
+       [OP_BACKCHANNEL_CTL]    = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_BIND_CONN_TO_SESSION] = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_EXCHANGE_ID]        = (nfsd4_enc)nfsd4_encode_exchange_id,
+       [OP_CREATE_SESSION]     = (nfsd4_enc)nfsd4_encode_create_session,
+       [OP_DESTROY_SESSION]    = (nfsd4_enc)nfsd4_encode_destroy_session,
+       [OP_FREE_STATEID]       = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_GET_DIR_DELEGATION] = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_GETDEVICEINFO]      = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_GETDEVICELIST]      = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_LAYOUTCOMMIT]       = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_LAYOUTGET]          = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_LAYOUTRETURN]       = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_SECINFO_NO_NAME]    = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_SEQUENCE]           = (nfsd4_enc)nfsd4_encode_sequence,
+       [OP_SET_SSV]            = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_TEST_STATEID]       = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_WANT_DELEGATION]    = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_DESTROY_CLIENTID]   = (nfsd4_enc)nfsd4_encode_noop,
+       [OP_RECLAIM_COMPLETE]   = (nfsd4_enc)nfsd4_encode_noop,
+};
+
+/*
+ * Calculate the total amount of memory that the compound response has taken
+ * after encoding the current operation.
+ *
+ * pad: add on 8 bytes for the next operation's op_code and status so that
+ * there is room to cache a failure on the next operation.
+ *
+ * Compare this length to the session se_fmaxresp_cached.
+ *
+ * Our se_fmaxresp_cached will always be a multiple of PAGE_SIZE, and so
+ * will be at least a page and will therefore hold the xdr_buf head.
+ */
+static int nfsd4_check_drc_limit(struct nfsd4_compoundres *resp)
+{
+       int status = 0;
+       struct xdr_buf *xb = &resp->rqstp->rq_res;
+       struct nfsd4_compoundargs *args = resp->rqstp->rq_argp;
+       struct nfsd4_session *session = NULL;
+       struct nfsd4_slot *slot = resp->cstate.slot;
+       u32 length, tlen = 0, pad = 8;
+
+       if (!nfsd4_has_session(&resp->cstate))
+               return status;
+
+       session = resp->cstate.session;
+       if (session == NULL || slot->sl_cachethis == 0)
+               return status;
+
+       if (resp->opcnt >= args->opcnt)
+               pad = 0; /* this is the last operation */
+
+       if (xb->page_len == 0) {
+               length = (char *)resp->p - (char *)xb->head[0].iov_base + pad;
+       } else {
+               if (xb->tail[0].iov_base && xb->tail[0].iov_len > 0)
+                       tlen = (char *)resp->p - (char *)xb->tail[0].iov_base;
+
+               length = xb->head[0].iov_len + xb->page_len + tlen + pad;
+       }
+       dprintk("%s length %u, xb->page_len %u tlen %u pad %u\n", __func__,
+               length, xb->page_len, tlen, pad);
+
+       if (length <= session->se_fchannel.maxresp_cached)
+               return status;
+       else
+               return nfserr_rep_too_big_to_cache;
 }
 
 void
 nfsd4_encode_operation(struct nfsd4_compoundres *resp, struct nfsd4_op *op)
 {
        __be32 *statp;
-       ENCODE_HEAD;
+       __be32 *p;
 
        RESERVE_SPACE(8);
        WRITE32(op->opnum);
        statp = p++;    /* to be backfilled at the end */
        ADJUST_ARGS();
 
-       switch (op->opnum) {
-       case OP_ACCESS:
-               nfsd4_encode_access(resp, op->status, &op->u.access);
-               break;
-       case OP_CLOSE:
-               nfsd4_encode_close(resp, op->status, &op->u.close);
-               break;
-       case OP_COMMIT:
-               nfsd4_encode_commit(resp, op->status, &op->u.commit);
-               break;
-       case OP_CREATE:
-               nfsd4_encode_create(resp, op->status, &op->u.create);
-               break;
-       case OP_DELEGRETURN:
-               break;
-       case OP_GETATTR:
-               op->status = nfsd4_encode_getattr(resp, op->status, &op->u.getattr);
-               break;
-       case OP_GETFH:
-               nfsd4_encode_getfh(resp, op->status, op->u.getfh);
-               break;
-       case OP_LINK:
-               nfsd4_encode_link(resp, op->status, &op->u.link);
-               break;
-       case OP_LOCK:
-               nfsd4_encode_lock(resp, op->status, &op->u.lock);
-               break;
-       case OP_LOCKT:
-               nfsd4_encode_lockt(resp, op->status, &op->u.lockt);
-               break;
-       case OP_LOCKU:
-               nfsd4_encode_locku(resp, op->status, &op->u.locku);
-               break;
-       case OP_LOOKUP:
-               break;
-       case OP_LOOKUPP:
-               break;
-       case OP_NVERIFY:
-               break;
-       case OP_OPEN:
-               nfsd4_encode_open(resp, op->status, &op->u.open);
-               break;
-       case OP_OPEN_CONFIRM:
-               nfsd4_encode_open_confirm(resp, op->status, &op->u.open_confirm);
-               break;
-       case OP_OPEN_DOWNGRADE:
-               nfsd4_encode_open_downgrade(resp, op->status, &op->u.open_downgrade);
-               break;
-       case OP_PUTFH:
-               break;
-       case OP_PUTROOTFH:
-               break;
-       case OP_READ:
-               op->status = nfsd4_encode_read(resp, op->status, &op->u.read);
-               break;
-       case OP_READDIR:
-               op->status = nfsd4_encode_readdir(resp, op->status, &op->u.readdir);
-               break;
-       case OP_READLINK:
-               op->status = nfsd4_encode_readlink(resp, op->status, &op->u.readlink);
-               break;
-       case OP_REMOVE:
-               nfsd4_encode_remove(resp, op->status, &op->u.remove);
-               break;
-       case OP_RENAME:
-               nfsd4_encode_rename(resp, op->status, &op->u.rename);
-               break;
-       case OP_RENEW:
-               break;
-       case OP_RESTOREFH:
-               break;
-       case OP_SAVEFH:
-               break;
-       case OP_SECINFO:
-               nfsd4_encode_secinfo(resp, op->status, &op->u.secinfo);
-               break;
-       case OP_SETATTR:
-               nfsd4_encode_setattr(resp, op->status, &op->u.setattr);
-               break;
-       case OP_SETCLIENTID:
-               nfsd4_encode_setclientid(resp, op->status, &op->u.setclientid);
-               break;
-       case OP_SETCLIENTID_CONFIRM:
-               break;
-       case OP_VERIFY:
-               break;
-       case OP_WRITE:
-               nfsd4_encode_write(resp, op->status, &op->u.write);
-               break;
-       case OP_RELEASE_LOCKOWNER:
-               break;
-       default:
-               break;
-       }
-
+       if (op->opnum == OP_ILLEGAL)
+               goto status;
+       BUG_ON(op->opnum < 0 || op->opnum >= ARRAY_SIZE(nfsd4_enc_ops) ||
+              !nfsd4_enc_ops[op->opnum]);
+       op->status = nfsd4_enc_ops[op->opnum](resp, op->status, &op->u);
+       /* nfsd4_check_drc_limit guarantees enough room for error status */
+       if (!op->status && nfsd4_check_drc_limit(resp))
+               op->status = nfserr_rep_too_big_to_cache;
+status:
        /*
         * Note: We write the status directly, instead of using WRITE32(),
         * since it is already in network byte order.
@@ -2695,7 +3228,7 @@ nfsd4_encode_operation(struct nfsd4_compoundres *resp, struct nfsd4_op *op)
 void
 nfsd4_encode_replay(struct nfsd4_compoundres *resp, struct nfsd4_op *op)
 {
-       ENCODE_HEAD;
+       __be32 *p;
        struct nfs4_replay *rp = op->replay;
 
        BUG_ON(!rp);
@@ -2710,10 +3243,6 @@ nfsd4_encode_replay(struct nfsd4_compoundres *resp, struct nfsd4_op *op)
        ADJUST_ARGS();
 }
 
-/*
- * END OF "GENERIC" ENCODE ROUTINES.
- */
-
 int
 nfs4svc_encode_voidres(struct svc_rqst *rqstp, __be32 *p, void *dummy)
 {
@@ -2763,6 +3292,7 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
        /*
         * All that remains is to write the tag and operation count...
         */
+       struct nfsd4_compound_state *cs = &resp->cstate;
        struct kvec *iov;
        p = resp->tagp;
        *p++ = htonl(resp->taglen);
@@ -2776,6 +3306,12 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
                iov = &rqstp->rq_res.head[0];
        iov->iov_len = ((char*)resp->p) - (char*)iov->iov_base;
        BUG_ON(iov->iov_len > PAGE_SIZE);
+       if (nfsd4_has_session(cs) && cs->status != nfserr_replay_cache) {
+               nfsd4_store_cache_entry(resp);
+               dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
+               resp->cstate.slot->sl_inuse = false;
+               nfsd4_put_session(resp->cstate.session);
+       }
        return 1;
 }