]> nv-tegra.nvidia Code Review - linux-3.10.git/blobdiff - net/sunrpc/svcauth_unix.c
nfsd4: move principal name into svc_cred
[linux-3.10.git] / net / sunrpc / svcauth_unix.c
index d6baf6fdf8a97d43f8569d9fef0bc762ebc616b9..12e4897d0bf3d8da150289a00cc37b734ab77d96 100644 (file)
@@ -5,13 +5,20 @@
 #include <linux/sunrpc/xdr.h>
 #include <linux/sunrpc/svcsock.h>
 #include <linux/sunrpc/svcauth.h>
+#include <linux/sunrpc/gss_api.h>
 #include <linux/err.h>
 #include <linux/seq_file.h>
 #include <linux/hash.h>
 #include <linux/string.h>
-
+#include <linux/slab.h>
+#include <net/sock.h>
+#include <net/ipv6.h>
+#include <linux/kernel.h>
 #define RPCDBG_FACILITY        RPCDBG_AUTH
 
+#include <linux/sunrpc/clnt.h>
+
+#include "netns.h"
 
 /*
  * AUTHUNIX and AUTHNULL credentials are both handled here.
 
 struct unix_domain {
        struct auth_domain      h;
-       int     addr_changes;
        /* other stuff later */
 };
 
-struct auth_domain *unix_domain_find(char *name)
-{
-       struct auth_domain *rv, ud;
-       struct unix_domain *new;
-
-       ud.name = name;
-       
-       rv = auth_domain_lookup(&ud, 0);
-
- foundit:
-       if (rv && rv->flavour != RPC_AUTH_UNIX) {
-               auth_domain_put(rv);
-               return NULL;
-       }
-       if (rv)
-               return rv;
-
-       new = kmalloc(sizeof(*new), GFP_KERNEL);
-       if (new == NULL)
-               return NULL;
-       cache_init(&new->h.h);
-       new->h.name = kstrdup(name, GFP_KERNEL);
-       new->h.flavour = RPC_AUTH_UNIX;
-       new->addr_changes = 0;
-       new->h.h.expiry_time = NEVER;
-
-       rv = auth_domain_lookup(&new->h, 2);
-       if (rv == &new->h) {
-               if (atomic_dec_and_test(&new->h.h.refcnt)) BUG();
-       } else {
-               auth_domain_put(&new->h);
-               goto foundit;
-       }
-
-       return rv;
-}
+extern struct auth_ops svcauth_null;
+extern struct auth_ops svcauth_unix;
 
 static void svcauth_unix_domain_release(struct auth_domain *dom)
 {
@@ -72,6 +44,39 @@ static void svcauth_unix_domain_release(struct auth_domain *dom)
        kfree(ud);
 }
 
+struct auth_domain *unix_domain_find(char *name)
+{
+       struct auth_domain *rv;
+       struct unix_domain *new = NULL;
+
+       rv = auth_domain_lookup(name, NULL);
+       while(1) {
+               if (rv) {
+                       if (new && rv != &new->h)
+                               svcauth_unix_domain_release(&new->h);
+
+                       if (rv->flavour != &svcauth_unix) {
+                               auth_domain_put(rv);
+                               return NULL;
+                       }
+                       return rv;
+               }
+
+               new = kmalloc(sizeof(*new), GFP_KERNEL);
+               if (new == NULL)
+                       return NULL;
+               kref_init(&new->h.ref);
+               new->h.name = kstrdup(name, GFP_KERNEL);
+               if (new->h.name == NULL) {
+                       kfree(new);
+                       return NULL;
+               }
+               new->h.flavour = &svcauth_unix;
+               rv = auth_domain_lookup(name, &new->h);
+       }
+}
+EXPORT_SYMBOL_GPL(unix_domain_find);
+
 
 /**************************************************
  * cache for IP address to unix_domain
@@ -79,70 +84,99 @@ static void svcauth_unix_domain_release(struct auth_domain *dom)
  */
 #define        IP_HASHBITS     8
 #define        IP_HASHMAX      (1<<IP_HASHBITS)
-#define        IP_HASHMASK     (IP_HASHMAX-1)
 
 struct ip_map {
        struct cache_head       h;
        char                    m_class[8]; /* e.g. "nfsd" */
-       struct in_addr          m_addr;
+       struct in6_addr         m_addr;
        struct unix_domain      *m_client;
-       int                     m_add_change;
 };
-static struct cache_head       *ip_table[IP_HASHMAX];
 
-static void ip_map_put(struct cache_head *item, struct cache_detail *cd)
+static void ip_map_put(struct kref *kref)
 {
+       struct cache_head *item = container_of(kref, struct cache_head, ref);
        struct ip_map *im = container_of(item, struct ip_map,h);
-       if (cache_put(item, cd)) {
-               if (test_bit(CACHE_VALID, &item->flags) &&
-                   !test_bit(CACHE_NEGATIVE, &item->flags))
-                       auth_domain_put(&im->m_client->h);
-               kfree(im);
-       }
+
+       if (test_bit(CACHE_VALID, &item->flags) &&
+           !test_bit(CACHE_NEGATIVE, &item->flags))
+               auth_domain_put(&im->m_client->h);
+       kfree(im);
 }
 
-static inline int ip_map_hash(struct ip_map *item)
+#if IP_HASHBITS == 8
+/* hash_long on a 64 bit machine is currently REALLY BAD for
+ * IP addresses in reverse-endian (i.e. on a little-endian machine).
+ * So use a trivial but reliable hash instead
+ */
+static inline int hash_ip(__be32 ip)
+{
+       int hash = (__force u32)ip ^ ((__force u32)ip>>16);
+       return (hash ^ (hash>>8)) & 0xff;
+}
+#endif
+static inline int hash_ip6(struct in6_addr ip)
 {
-       return hash_str(item->m_class, IP_HASHBITS) ^ 
-               hash_long((unsigned long)item->m_addr.s_addr, IP_HASHBITS);
+       return (hash_ip(ip.s6_addr32[0]) ^
+               hash_ip(ip.s6_addr32[1]) ^
+               hash_ip(ip.s6_addr32[2]) ^
+               hash_ip(ip.s6_addr32[3]));
 }
-static inline int ip_map_match(struct ip_map *item, struct ip_map *tmp)
+static int ip_map_match(struct cache_head *corig, struct cache_head *cnew)
 {
-       return strcmp(tmp->m_class, item->m_class) == 0
-               && tmp->m_addr.s_addr == item->m_addr.s_addr;
+       struct ip_map *orig = container_of(corig, struct ip_map, h);
+       struct ip_map *new = container_of(cnew, struct ip_map, h);
+       return strcmp(orig->m_class, new->m_class) == 0 &&
+              ipv6_addr_equal(&orig->m_addr, &new->m_addr);
 }
-static inline void ip_map_init(struct ip_map *new, struct ip_map *item)
+static void ip_map_init(struct cache_head *cnew, struct cache_head *citem)
 {
+       struct ip_map *new = container_of(cnew, struct ip_map, h);
+       struct ip_map *item = container_of(citem, struct ip_map, h);
+
        strcpy(new->m_class, item->m_class);
-       new->m_addr.s_addr = item->m_addr.s_addr;
+       new->m_addr = item->m_addr;
 }
-static inline void ip_map_update(struct ip_map *new, struct ip_map *item)
+static void update(struct cache_head *cnew, struct cache_head *citem)
 {
-       cache_get(&item->m_client->h.h);
+       struct ip_map *new = container_of(cnew, struct ip_map, h);
+       struct ip_map *item = container_of(citem, struct ip_map, h);
+
+       kref_get(&item->m_client->h.ref);
        new->m_client = item->m_client;
-       new->m_add_change = item->m_add_change;
+}
+static struct cache_head *ip_map_alloc(void)
+{
+       struct ip_map *i = kmalloc(sizeof(*i), GFP_KERNEL);
+       if (i)
+               return &i->h;
+       else
+               return NULL;
 }
 
 static void ip_map_request(struct cache_detail *cd,
                                  struct cache_head *h,
                                  char **bpp, int *blen)
 {
-       char text_addr[20];
+       char text_addr[40];
        struct ip_map *im = container_of(h, struct ip_map, h);
-       __u32 addr = im->m_addr.s_addr;
-       
-       snprintf(text_addr, 20, "%u.%u.%u.%u",
-                ntohl(addr) >> 24 & 0xff,
-                ntohl(addr) >> 16 & 0xff,
-                ntohl(addr) >>  8 & 0xff,
-                ntohl(addr) >>  0 & 0xff);
 
+       if (ipv6_addr_v4mapped(&(im->m_addr))) {
+               snprintf(text_addr, 20, "%pI4", &im->m_addr.s6_addr32[3]);
+       } else {
+               snprintf(text_addr, 40, "%pI6", &im->m_addr);
+       }
        qword_add(bpp, blen, im->m_class);
        qword_add(bpp, blen, text_addr);
        (*bpp)[-1] = '\n';
 }
 
-static struct ip_map *ip_map_lookup(struct ip_map *, int);
+static int ip_map_upcall(struct cache_detail *cd, struct cache_head *h)
+{
+       return sunrpc_cache_pipe_upcall(cd, h, ip_map_request);
+}
+
+static struct ip_map *__ip_map_lookup(struct cache_detail *cd, char *class, struct in6_addr *addr);
+static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm, struct unix_domain *udom, time_t expiry);
 
 static int ip_map_parse(struct cache_detail *cd,
                          char *mesg, int mlen)
@@ -152,9 +186,16 @@ static int ip_map_parse(struct cache_detail *cd,
         * for scratch: */
        char *buf = mesg;
        int len;
-       int b1,b2,b3,b4;
-       char c;
-       struct ip_map ipm, *ipmp;
+       char class[8];
+       union {
+               struct sockaddr         sa;
+               struct sockaddr_in      s4;
+               struct sockaddr_in6     s6;
+       } address;
+       struct sockaddr_in6 sin6;
+       int err;
+
+       struct ip_map *ipmp;
        struct auth_domain *dom;
        time_t expiry;
 
@@ -163,16 +204,31 @@ static int ip_map_parse(struct cache_detail *cd,
        mesg[mlen-1] = 0;
 
        /* class */
-       len = qword_get(&mesg, ipm.m_class, sizeof(ipm.m_class));
+       len = qword_get(&mesg, class, sizeof(class));
        if (len <= 0) return -EINVAL;
 
        /* ip address */
        len = qword_get(&mesg, buf, mlen);
        if (len <= 0) return -EINVAL;
 
-       if (sscanf(buf, "%u.%u.%u.%u%c", &b1, &b2, &b3, &b4, &c) != 4)
+       if (rpc_pton(cd->net, buf, len, &address.sa, sizeof(address)) == 0)
+               return -EINVAL;
+       switch (address.sa.sa_family) {
+       case AF_INET:
+               /* Form a mapped IPv4 address in sin6 */
+               sin6.sin6_family = AF_INET6;
+               ipv6_addr_set_v4mapped(address.s4.sin_addr.s_addr,
+                               &sin6.sin6_addr);
+               break;
+#if IS_ENABLED(CONFIG_IPV6)
+       case AF_INET6:
+               memcpy(&sin6, &address.s6, sizeof(sin6));
+               break;
+#endif
+       default:
                return -EINVAL;
-       
+       }
+
        expiry = get_expiry(&mesg);
        if (expiry ==0)
                return -EINVAL;
@@ -188,25 +244,20 @@ static int ip_map_parse(struct cache_detail *cd,
        } else
                dom = NULL;
 
-       ipm.m_addr.s_addr =
-               htonl((((((b1<<8)|b2)<<8)|b3)<<8)|b4);
-       ipm.h.flags = 0;
-       if (dom) {
-               ipm.m_client = container_of(dom, struct unix_domain, h);
-               ipm.m_add_change = ipm.m_client->addr_changes;
+       /* IPv6 scope IDs are ignored for now */
+       ipmp = __ip_map_lookup(cd, class, &sin6.sin6_addr);
+       if (ipmp) {
+               err = __ip_map_update(cd, ipmp,
+                            container_of(dom, struct unix_domain, h),
+                            expiry);
        } else
-               set_bit(CACHE_NEGATIVE, &ipm.h.flags);
-       ipm.h.expiry_time = expiry;
+               err = -ENOMEM;
 
-       ipmp = ip_map_lookup(&ipm, 1);
-       if (ipmp)
-               ip_map_put(&ipmp->h, &ip_map_cache);
        if (dom)
                auth_domain_put(dom);
-       if (!ipmp)
-               return -ENOMEM;
+
        cache_flush();
-       return 0;
+       return err;
 }
 
 static int ip_map_show(struct seq_file *m,
@@ -214,7 +265,7 @@ static int ip_map_show(struct seq_file *m,
                       struct cache_head *h)
 {
        struct ip_map *im;
-       struct in_addr addr;
+       struct in6_addr addr;
        char *dom = "-no-domain-";
 
        if (h == NULL) {
@@ -225,155 +276,482 @@ static int ip_map_show(struct seq_file *m,
        /* class addr domain */
        addr = im->m_addr;
 
-       if (test_bit(CACHE_VALID, &h->flags) && 
+       if (test_bit(CACHE_VALID, &h->flags) &&
            !test_bit(CACHE_NEGATIVE, &h->flags))
                dom = im->m_client->h.name;
 
-       seq_printf(m, "%s %d.%d.%d.%d %s\n",
-                  im->m_class,
-                  htonl(addr.s_addr) >> 24 & 0xff,
-                  htonl(addr.s_addr) >> 16 & 0xff,
-                  htonl(addr.s_addr) >>  8 & 0xff,
-                  htonl(addr.s_addr) >>  0 & 0xff,
-                  dom
-                  );
+       if (ipv6_addr_v4mapped(&addr)) {
+               seq_printf(m, "%s %pI4 %s\n",
+                       im->m_class, &addr.s6_addr32[3], dom);
+       } else {
+               seq_printf(m, "%s %pI6 %s\n", im->m_class, &addr, dom);
+       }
        return 0;
 }
-       
 
-struct cache_detail ip_map_cache = {
-       .hash_size      = IP_HASHMAX,
-       .hash_table     = ip_table,
-       .name           = "auth.unix.ip",
-       .cache_put      = ip_map_put,
-       .cache_request  = ip_map_request,
-       .cache_parse    = ip_map_parse,
-       .cache_show     = ip_map_show,
-};
 
-static DefineSimpleCacheLookup(ip_map, 0)
+static struct ip_map *__ip_map_lookup(struct cache_detail *cd, char *class,
+               struct in6_addr *addr)
+{
+       struct ip_map ip;
+       struct cache_head *ch;
+
+       strcpy(ip.m_class, class);
+       ip.m_addr = *addr;
+       ch = sunrpc_cache_lookup(cd, &ip.h,
+                                hash_str(class, IP_HASHBITS) ^
+                                hash_ip6(*addr));
+
+       if (ch)
+               return container_of(ch, struct ip_map, h);
+       else
+               return NULL;
+}
+
+static inline struct ip_map *ip_map_lookup(struct net *net, char *class,
+               struct in6_addr *addr)
+{
+       struct sunrpc_net *sn;
 
+       sn = net_generic(net, sunrpc_net_id);
+       return __ip_map_lookup(sn->ip_map_cache, class, addr);
+}
 
-int auth_unix_add_addr(struct in_addr addr, struct auth_domain *dom)
+static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm,
+               struct unix_domain *udom, time_t expiry)
 {
-       struct unix_domain *udom;
-       struct ip_map ip, *ipmp;
+       struct ip_map ip;
+       struct cache_head *ch;
 
-       if (dom->flavour != RPC_AUTH_UNIX)
-               return -EINVAL;
-       udom = container_of(dom, struct unix_domain, h);
-       strcpy(ip.m_class, "nfsd");
-       ip.m_addr = addr;
        ip.m_client = udom;
-       ip.m_add_change = udom->addr_changes+1;
        ip.h.flags = 0;
-       ip.h.expiry_time = NEVER;
-       
-       ipmp = ip_map_lookup(&ip, 1);
-
-       if (ipmp) {
-               ip_map_put(&ipmp->h, &ip_map_cache);
-               return 0;
-       } else
+       if (!udom)
+               set_bit(CACHE_NEGATIVE, &ip.h.flags);
+       ip.h.expiry_time = expiry;
+       ch = sunrpc_cache_update(cd, &ip.h, &ipm->h,
+                                hash_str(ipm->m_class, IP_HASHBITS) ^
+                                hash_ip6(ipm->m_addr));
+       if (!ch)
                return -ENOMEM;
+       cache_put(ch, cd);
+       return 0;
 }
 
-int auth_unix_forget_old(struct auth_domain *dom)
+static inline int ip_map_update(struct net *net, struct ip_map *ipm,
+               struct unix_domain *udom, time_t expiry)
 {
-       struct unix_domain *udom;
-       
-       if (dom->flavour != RPC_AUTH_UNIX)
-               return -EINVAL;
-       udom = container_of(dom, struct unix_domain, h);
-       udom->addr_changes++;
-       return 0;
+       struct sunrpc_net *sn;
+
+       sn = net_generic(net, sunrpc_net_id);
+       return __ip_map_update(sn->ip_map_cache, ipm, udom, expiry);
 }
 
-struct auth_domain *auth_unix_lookup(struct in_addr addr)
+void svcauth_unix_purge(struct net *net)
 {
-       struct ip_map key, *ipm;
-       struct auth_domain *rv;
+       struct sunrpc_net *sn;
 
-       strcpy(key.m_class, "nfsd");
-       key.m_addr = addr;
+       sn = net_generic(net, sunrpc_net_id);
+       cache_purge(sn->ip_map_cache);
+}
+EXPORT_SYMBOL_GPL(svcauth_unix_purge);
 
-       ipm = ip_map_lookup(&key, 0);
+static inline struct ip_map *
+ip_map_cached_get(struct svc_xprt *xprt)
+{
+       struct ip_map *ipm = NULL;
+       struct sunrpc_net *sn;
+
+       if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) {
+               spin_lock(&xprt->xpt_lock);
+               ipm = xprt->xpt_auth_cache;
+               if (ipm != NULL) {
+                       if (!cache_valid(&ipm->h)) {
+                               /*
+                                * The entry has been invalidated since it was
+                                * remembered, e.g. by a second mount from the
+                                * same IP address.
+                                */
+                               sn = net_generic(xprt->xpt_net, sunrpc_net_id);
+                               xprt->xpt_auth_cache = NULL;
+                               spin_unlock(&xprt->xpt_lock);
+                               cache_put(&ipm->h, sn->ip_map_cache);
+                               return NULL;
+                       }
+                       cache_get(&ipm->h);
+               }
+               spin_unlock(&xprt->xpt_lock);
+       }
+       return ipm;
+}
 
-       if (!ipm)
-               return NULL;
-       if (cache_check(&ip_map_cache, &ipm->h, NULL))
+static inline void
+ip_map_cached_put(struct svc_xprt *xprt, struct ip_map *ipm)
+{
+       if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags)) {
+               spin_lock(&xprt->xpt_lock);
+               if (xprt->xpt_auth_cache == NULL) {
+                       /* newly cached, keep the reference */
+                       xprt->xpt_auth_cache = ipm;
+                       ipm = NULL;
+               }
+               spin_unlock(&xprt->xpt_lock);
+       }
+       if (ipm) {
+               struct sunrpc_net *sn;
+
+               sn = net_generic(xprt->xpt_net, sunrpc_net_id);
+               cache_put(&ipm->h, sn->ip_map_cache);
+       }
+}
+
+void
+svcauth_unix_info_release(struct svc_xprt *xpt)
+{
+       struct ip_map *ipm;
+
+       ipm = xpt->xpt_auth_cache;
+       if (ipm != NULL) {
+               struct sunrpc_net *sn;
+
+               sn = net_generic(xpt->xpt_net, sunrpc_net_id);
+               cache_put(&ipm->h, sn->ip_map_cache);
+       }
+}
+
+/****************************************************************************
+ * auth.unix.gid cache
+ * simple cache to map a UID to a list of GIDs
+ * because AUTH_UNIX aka AUTH_SYS has a max of 16
+ */
+#define        GID_HASHBITS    8
+#define        GID_HASHMAX     (1<<GID_HASHBITS)
+
+struct unix_gid {
+       struct cache_head       h;
+       uid_t                   uid;
+       struct group_info       *gi;
+};
+
+static void unix_gid_put(struct kref *kref)
+{
+       struct cache_head *item = container_of(kref, struct cache_head, ref);
+       struct unix_gid *ug = container_of(item, struct unix_gid, h);
+       if (test_bit(CACHE_VALID, &item->flags) &&
+           !test_bit(CACHE_NEGATIVE, &item->flags))
+               put_group_info(ug->gi);
+       kfree(ug);
+}
+
+static int unix_gid_match(struct cache_head *corig, struct cache_head *cnew)
+{
+       struct unix_gid *orig = container_of(corig, struct unix_gid, h);
+       struct unix_gid *new = container_of(cnew, struct unix_gid, h);
+       return orig->uid == new->uid;
+}
+static void unix_gid_init(struct cache_head *cnew, struct cache_head *citem)
+{
+       struct unix_gid *new = container_of(cnew, struct unix_gid, h);
+       struct unix_gid *item = container_of(citem, struct unix_gid, h);
+       new->uid = item->uid;
+}
+static void unix_gid_update(struct cache_head *cnew, struct cache_head *citem)
+{
+       struct unix_gid *new = container_of(cnew, struct unix_gid, h);
+       struct unix_gid *item = container_of(citem, struct unix_gid, h);
+
+       get_group_info(item->gi);
+       new->gi = item->gi;
+}
+static struct cache_head *unix_gid_alloc(void)
+{
+       struct unix_gid *g = kmalloc(sizeof(*g), GFP_KERNEL);
+       if (g)
+               return &g->h;
+       else
                return NULL;
+}
 
-       if ((ipm->m_client->addr_changes - ipm->m_add_change) >0) {
-               if (test_and_set_bit(CACHE_NEGATIVE, &ipm->h.flags) == 0)
-                       auth_domain_put(&ipm->m_client->h);
-               rv = NULL;
-       } else {
-               rv = &ipm->m_client->h;
-               cache_get(&rv->h);
+static void unix_gid_request(struct cache_detail *cd,
+                            struct cache_head *h,
+                            char **bpp, int *blen)
+{
+       char tuid[20];
+       struct unix_gid *ug = container_of(h, struct unix_gid, h);
+
+       snprintf(tuid, 20, "%u", ug->uid);
+       qword_add(bpp, blen, tuid);
+       (*bpp)[-1] = '\n';
+}
+
+static int unix_gid_upcall(struct cache_detail *cd, struct cache_head *h)
+{
+       return sunrpc_cache_pipe_upcall(cd, h, unix_gid_request);
+}
+
+static struct unix_gid *unix_gid_lookup(struct cache_detail *cd, uid_t uid);
+
+static int unix_gid_parse(struct cache_detail *cd,
+                       char *mesg, int mlen)
+{
+       /* uid expiry Ngid gid0 gid1 ... gidN-1 */
+       int uid;
+       int gids;
+       int rv;
+       int i;
+       int err;
+       time_t expiry;
+       struct unix_gid ug, *ugp;
+
+       if (mesg[mlen - 1] != '\n')
+               return -EINVAL;
+       mesg[mlen-1] = 0;
+
+       rv = get_int(&mesg, &uid);
+       if (rv)
+               return -EINVAL;
+       ug.uid = uid;
+
+       expiry = get_expiry(&mesg);
+       if (expiry == 0)
+               return -EINVAL;
+
+       rv = get_int(&mesg, &gids);
+       if (rv || gids < 0 || gids > 8192)
+               return -EINVAL;
+
+       ug.gi = groups_alloc(gids);
+       if (!ug.gi)
+               return -ENOMEM;
+
+       for (i = 0 ; i < gids ; i++) {
+               int gid;
+               rv = get_int(&mesg, &gid);
+               err = -EINVAL;
+               if (rv)
+                       goto out;
+               GROUP_AT(ug.gi, i) = gid;
        }
-       ip_map_put(&ipm->h, &ip_map_cache);
-       return rv;
+
+       ugp = unix_gid_lookup(cd, uid);
+       if (ugp) {
+               struct cache_head *ch;
+               ug.h.flags = 0;
+               ug.h.expiry_time = expiry;
+               ch = sunrpc_cache_update(cd,
+                                        &ug.h, &ugp->h,
+                                        hash_long(uid, GID_HASHBITS));
+               if (!ch)
+                       err = -ENOMEM;
+               else {
+                       err = 0;
+                       cache_put(ch, cd);
+               }
+       } else
+               err = -ENOMEM;
+ out:
+       if (ug.gi)
+               put_group_info(ug.gi);
+       return err;
 }
 
-void svcauth_unix_purge(void)
+static int unix_gid_show(struct seq_file *m,
+                        struct cache_detail *cd,
+                        struct cache_head *h)
 {
-       cache_purge(&ip_map_cache);
-       cache_purge(&auth_domain_cache);
+       struct unix_gid *ug;
+       int i;
+       int glen;
+
+       if (h == NULL) {
+               seq_puts(m, "#uid cnt: gids...\n");
+               return 0;
+       }
+       ug = container_of(h, struct unix_gid, h);
+       if (test_bit(CACHE_VALID, &h->flags) &&
+           !test_bit(CACHE_NEGATIVE, &h->flags))
+               glen = ug->gi->ngroups;
+       else
+               glen = 0;
+
+       seq_printf(m, "%u %d:", ug->uid, glen);
+       for (i = 0; i < glen; i++)
+               seq_printf(m, " %d", GROUP_AT(ug->gi, i));
+       seq_printf(m, "\n");
+       return 0;
 }
 
-static int
+static struct cache_detail unix_gid_cache_template = {
+       .owner          = THIS_MODULE,
+       .hash_size      = GID_HASHMAX,
+       .name           = "auth.unix.gid",
+       .cache_put      = unix_gid_put,
+       .cache_upcall   = unix_gid_upcall,
+       .cache_parse    = unix_gid_parse,
+       .cache_show     = unix_gid_show,
+       .match          = unix_gid_match,
+       .init           = unix_gid_init,
+       .update         = unix_gid_update,
+       .alloc          = unix_gid_alloc,
+};
+
+int unix_gid_cache_create(struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+       struct cache_detail *cd;
+       int err;
+
+       cd = cache_create_net(&unix_gid_cache_template, net);
+       if (IS_ERR(cd))
+               return PTR_ERR(cd);
+       err = cache_register_net(cd, net);
+       if (err) {
+               cache_destroy_net(cd, net);
+               return err;
+       }
+       sn->unix_gid_cache = cd;
+       return 0;
+}
+
+void unix_gid_cache_destroy(struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+       struct cache_detail *cd = sn->unix_gid_cache;
+
+       sn->unix_gid_cache = NULL;
+       cache_purge(cd);
+       cache_unregister_net(cd, net);
+       cache_destroy_net(cd, net);
+}
+
+static struct unix_gid *unix_gid_lookup(struct cache_detail *cd, uid_t uid)
+{
+       struct unix_gid ug;
+       struct cache_head *ch;
+
+       ug.uid = uid;
+       ch = sunrpc_cache_lookup(cd, &ug.h, hash_long(uid, GID_HASHBITS));
+       if (ch)
+               return container_of(ch, struct unix_gid, h);
+       else
+               return NULL;
+}
+
+static struct group_info *unix_gid_find(uid_t uid, struct svc_rqst *rqstp)
+{
+       struct unix_gid *ug;
+       struct group_info *gi;
+       int ret;
+       struct sunrpc_net *sn = net_generic(rqstp->rq_xprt->xpt_net,
+                                           sunrpc_net_id);
+
+       ug = unix_gid_lookup(sn->unix_gid_cache, uid);
+       if (!ug)
+               return ERR_PTR(-EAGAIN);
+       ret = cache_check(sn->unix_gid_cache, &ug->h, &rqstp->rq_chandle);
+       switch (ret) {
+       case -ENOENT:
+               return ERR_PTR(-ENOENT);
+       case -ETIMEDOUT:
+               return ERR_PTR(-ESHUTDOWN);
+       case 0:
+               gi = get_group_info(ug->gi);
+               cache_put(&ug->h, sn->unix_gid_cache);
+               return gi;
+       default:
+               return ERR_PTR(-EAGAIN);
+       }
+}
+
+int
 svcauth_unix_set_client(struct svc_rqst *rqstp)
 {
-       struct ip_map key, *ipm;
+       struct sockaddr_in *sin;
+       struct sockaddr_in6 *sin6, sin6_storage;
+       struct ip_map *ipm;
+       struct group_info *gi;
+       struct svc_cred *cred = &rqstp->rq_cred;
+       struct svc_xprt *xprt = rqstp->rq_xprt;
+       struct net *net = xprt->xpt_net;
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+
+       switch (rqstp->rq_addr.ss_family) {
+       case AF_INET:
+               sin = svc_addr_in(rqstp);
+               sin6 = &sin6_storage;
+               ipv6_addr_set_v4mapped(sin->sin_addr.s_addr, &sin6->sin6_addr);
+               break;
+       case AF_INET6:
+               sin6 = svc_addr_in6(rqstp);
+               break;
+       default:
+               BUG();
+       }
 
        rqstp->rq_client = NULL;
        if (rqstp->rq_proc == 0)
                return SVC_OK;
 
-       strcpy(key.m_class, rqstp->rq_server->sv_program->pg_class);
-       key.m_addr = rqstp->rq_addr.sin_addr;
-
-       ipm = ip_map_lookup(&key, 0);
+       ipm = ip_map_cached_get(xprt);
+       if (ipm == NULL)
+               ipm = __ip_map_lookup(sn->ip_map_cache, rqstp->rq_server->sv_program->pg_class,
+                                   &sin6->sin6_addr);
 
        if (ipm == NULL)
                return SVC_DENIED;
 
-       switch (cache_check(&ip_map_cache, &ipm->h, &rqstp->rq_chandle)) {
+       switch (cache_check(sn->ip_map_cache, &ipm->h, &rqstp->rq_chandle)) {
                default:
                        BUG();
+               case -ETIMEDOUT:
+                       return SVC_CLOSE;
                case -EAGAIN:
                        return SVC_DROP;
                case -ENOENT:
                        return SVC_DENIED;
                case 0:
                        rqstp->rq_client = &ipm->m_client->h;
-                       cache_get(&rqstp->rq_client->h);
-                       ip_map_put(&ipm->h, &ip_map_cache);
+                       kref_get(&rqstp->rq_client->ref);
+                       ip_map_cached_put(xprt, ipm);
                        break;
        }
+
+       gi = unix_gid_find(cred->cr_uid, rqstp);
+       switch (PTR_ERR(gi)) {
+       case -EAGAIN:
+               return SVC_DROP;
+       case -ESHUTDOWN:
+               return SVC_CLOSE;
+       case -ENOENT:
+               break;
+       default:
+               put_group_info(cred->cr_group_info);
+               cred->cr_group_info = gi;
+       }
        return SVC_OK;
 }
 
+EXPORT_SYMBOL_GPL(svcauth_unix_set_client);
+
 static int
-svcauth_null_accept(struct svc_rqst *rqstp, u32 *authp)
+svcauth_null_accept(struct svc_rqst *rqstp, __be32 *authp)
 {
        struct kvec     *argv = &rqstp->rq_arg.head[0];
        struct kvec     *resv = &rqstp->rq_res.head[0];
        struct svc_cred *cred = &rqstp->rq_cred;
 
        cred->cr_group_info = NULL;
+       cred->cr_principal = NULL;
        rqstp->rq_client = NULL;
 
        if (argv->iov_len < 3*4)
                return SVC_GARBAGE;
 
-       if (svc_getu32(argv) != 0) { 
+       if (svc_getu32(argv) != 0) {
                dprintk("svc: bad null cred\n");
                *authp = rpc_autherr_badcred;
                return SVC_DENIED;
        }
-       if (svc_getu32(argv) != RPC_AUTH_NULL || svc_getu32(argv) != 0) {
+       if (svc_getu32(argv) != htonl(RPC_AUTH_NULL) || svc_getu32(argv) != 0) {
                dprintk("svc: bad null verf\n");
                *authp = rpc_autherr_badverf;
                return SVC_DENIED;
@@ -384,12 +762,13 @@ svcauth_null_accept(struct svc_rqst *rqstp, u32 *authp)
        cred->cr_gid = (gid_t) -1;
        cred->cr_group_info = groups_alloc(0);
        if (cred->cr_group_info == NULL)
-               return SVC_DROP; /* kmalloc failure - client must retry */
+               return SVC_CLOSE; /* kmalloc failure - client must retry */
 
        /* Put NULL verifier */
-       svc_putu32(resv, RPC_AUTH_NULL);
-       svc_putu32(resv, 0);
+       svc_putnl(resv, RPC_AUTH_NULL);
+       svc_putnl(resv, 0);
 
+       rqstp->rq_flavor = RPC_AUTH_NULL;
        return SVC_OK;
 }
 
@@ -418,7 +797,7 @@ struct auth_ops svcauth_null = {
 
 
 static int
-svcauth_unix_accept(struct svc_rqst *rqstp, u32 *authp)
+svcauth_unix_accept(struct svc_rqst *rqstp, __be32 *authp)
 {
        struct kvec     *argv = &rqstp->rq_arg.head[0];
        struct kvec     *resv = &rqstp->rq_res.head[0];
@@ -427,6 +806,7 @@ svcauth_unix_accept(struct svc_rqst *rqstp, u32 *authp)
        int             len   = argv->iov_len;
 
        cred->cr_group_info = NULL;
+       cred->cr_principal = NULL;
        rqstp->rq_client = NULL;
 
        if ((len -= 3*4) < 0)
@@ -434,32 +814,32 @@ svcauth_unix_accept(struct svc_rqst *rqstp, u32 *authp)
 
        svc_getu32(argv);                       /* length */
        svc_getu32(argv);                       /* time stamp */
-       slen = XDR_QUADLEN(ntohl(svc_getu32(argv)));    /* machname length */
+       slen = XDR_QUADLEN(svc_getnl(argv));    /* machname length */
        if (slen > 64 || (len -= (slen + 3)*4) < 0)
                goto badcred;
-       argv->iov_base = (void*)((u32*)argv->iov_base + slen);  /* skip machname */
+       argv->iov_base = (void*)((__be32*)argv->iov_base + slen);       /* skip machname */
        argv->iov_len -= slen*4;
 
-       cred->cr_uid = ntohl(svc_getu32(argv));         /* uid */
-       cred->cr_gid = ntohl(svc_getu32(argv));         /* gid */
-       slen = ntohl(svc_getu32(argv));                 /* gids length */
+       cred->cr_uid = svc_getnl(argv);         /* uid */
+       cred->cr_gid = svc_getnl(argv);         /* gid */
+       slen = svc_getnl(argv);                 /* gids length */
        if (slen > 16 || (len -= (slen + 2)*4) < 0)
                goto badcred;
        cred->cr_group_info = groups_alloc(slen);
        if (cred->cr_group_info == NULL)
-               return SVC_DROP;
+               return SVC_CLOSE;
        for (i = 0; i < slen; i++)
-               GROUP_AT(cred->cr_group_info, i) = ntohl(svc_getu32(argv));
-
-       if (svc_getu32(argv) != RPC_AUTH_NULL || svc_getu32(argv) != 0) {
+               GROUP_AT(cred->cr_group_info, i) = svc_getnl(argv);
+       if (svc_getu32(argv) != htonl(RPC_AUTH_NULL) || svc_getu32(argv) != 0) {
                *authp = rpc_autherr_badverf;
                return SVC_DENIED;
        }
 
        /* Put NULL verifier */
-       svc_putu32(resv, RPC_AUTH_NULL);
-       svc_putu32(resv, 0);
+       svc_putnl(resv, RPC_AUTH_NULL);
+       svc_putnl(resv, 0);
 
+       rqstp->rq_flavor = RPC_AUTH_UNIX;
        return SVC_OK;
 
 badcred:
@@ -493,3 +873,45 @@ struct auth_ops svcauth_unix = {
        .set_client     = svcauth_unix_set_client,
 };
 
+static struct cache_detail ip_map_cache_template = {
+       .owner          = THIS_MODULE,
+       .hash_size      = IP_HASHMAX,
+       .name           = "auth.unix.ip",
+       .cache_put      = ip_map_put,
+       .cache_upcall   = ip_map_upcall,
+       .cache_parse    = ip_map_parse,
+       .cache_show     = ip_map_show,
+       .match          = ip_map_match,
+       .init           = ip_map_init,
+       .update         = update,
+       .alloc          = ip_map_alloc,
+};
+
+int ip_map_cache_create(struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+       struct cache_detail *cd;
+       int err;
+
+       cd = cache_create_net(&ip_map_cache_template, net);
+       if (IS_ERR(cd))
+               return PTR_ERR(cd);
+       err = cache_register_net(cd, net);
+       if (err) {
+               cache_destroy_net(cd, net);
+               return err;
+       }
+       sn->ip_map_cache = cd;
+       return 0;
+}
+
+void ip_map_cache_destroy(struct net *net)
+{
+       struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
+       struct cache_detail *cd = sn->ip_map_cache;
+
+       sn->ip_map_cache = NULL;
+       cache_purge(cd);
+       cache_unregister_net(cd, net);
+       cache_destroy_net(cd, net);
+}