* 5602 fix improves checks that NID is set correctly and causes incorrect

   configurations to fail rather than trip a server assertion.

*  5606 fix ensures both peers agree on maximum "immediate" get.

*  ranal updates

diff --git a/lnet/autoconf/lustre-lnet.m4 b/lnet/autoconf/lustre-lnet.m4
index 7023714..06bdf39 100644 (file)
--- a/lnet/autoconf/lustre-lnet.m4
+++ b/lnet/autoconf/lustre-lnet.m4
@@ -278,7 +278,7 @@ AC_DEFUN([LP_CONFIG_RANAL],
 [#### Rapid Array
 AC_MSG_CHECKING([if RapidArray kernel headers are present])
 # placeholder
 [#### Rapid Array
 AC_MSG_CHECKING([if RapidArray kernel headers are present])
 # placeholder

-RACPPFLAGS="-I/tmp"
+RACPPFLAGS="-I${LINUX}/drivers/xd1/include"

 EXTRA_KCFLAGS_save="$EXTRA_KCFLAGS"
 EXTRA_KCFLAGS="$EXTRA_KCFLAGS $RACPPFLAGS"
 LB_LINUX_TRY_COMPILE([
 EXTRA_KCFLAGS_save="$EXTRA_KCFLAGS"
 EXTRA_KCFLAGS="$EXTRA_KCFLAGS $RACPPFLAGS"
 LB_LINUX_TRY_COMPILE([

diff --git a/lnet/include/lnet/lnetctl.h b/lnet/include/lnet/lnetctl.h
index b771690..6a7df9a 100644 (file)
--- a/lnet/include/lnet/lnetctl.h
+++ b/lnet/include/lnet/lnetctl.h
@@ -34,6 +34,7 @@
 
 int ptl_name2nal(char *str);
 int ptl_parse_ipaddr (__u32 *ipaddrp, char *str);
 
 int ptl_name2nal(char *str);
 int ptl_parse_ipaddr (__u32 *ipaddrp, char *str);

+int ptl_parse_anynid (ptl_nid_t *nidp, char *str);

 int ptl_parse_nid (ptl_nid_t *nidp, char *str);
 char * ptl_nid2str (char *buffer, ptl_nid_t nid);
 
 int ptl_parse_nid (ptl_nid_t *nidp, char *str);
 char * ptl_nid2str (char *buffer, ptl_nid_t nid);
 

diff --git a/lnet/include/lnet/ptlctl.h b/lnet/include/lnet/ptlctl.h
index b771690..6a7df9a 100644 (file)
--- a/lnet/include/lnet/ptlctl.h
+++ b/lnet/include/lnet/ptlctl.h
@@ -34,6 +34,7 @@
 
 int ptl_name2nal(char *str);
 int ptl_parse_ipaddr (__u32 *ipaddrp, char *str);
 
 int ptl_name2nal(char *str);
 int ptl_parse_ipaddr (__u32 *ipaddrp, char *str);

+int ptl_parse_anynid (ptl_nid_t *nidp, char *str);

 int ptl_parse_nid (ptl_nid_t *nidp, char *str);
 char * ptl_nid2str (char *buffer, ptl_nid_t nid);
 
 int ptl_parse_nid (ptl_nid_t *nidp, char *str);
 char * ptl_nid2str (char *buffer, ptl_nid_t nid);
 

diff --git a/lnet/klnds/openiblnd/openiblnd.c b/lnet/klnds/openiblnd/openiblnd.c
index 4289210..adc5ba8 100644 (file)
--- a/lnet/klnds/openiblnd/openiblnd.c
+++ b/lnet/klnds/openiblnd/openiblnd.c
@@ -161,6 +161,11 @@ kibnal_unpack_msg(kib_msg_t *msg, int nob)
                 __swab64s(&msg->ibm_dststamp);
         }
         
                 __swab64s(&msg->ibm_dststamp);
         }
         

+        if (msg->ibm_srcnid == PTL_NID_ANY) {
+                CERROR("Bad src nid: "LPX64"\n", msg->ibm_srcnid);
+                return -EPROTO;
+        }
+

         switch (msg->ibm_type) {
         default:
                 CERROR("Unknown message type %x\n", msg->ibm_type);
         switch (msg->ibm_type) {
         default:
                 CERROR("Unknown message type %x\n", msg->ibm_type);

diff --git a/lnet/klnds/openiblnd/openiblnd_cb.c b/lnet/klnds/openiblnd/openiblnd_cb.c
index 34d465f..62b64a4 100644 (file)
--- a/lnet/klnds/openiblnd/openiblnd_cb.c
+++ b/lnet/klnds/openiblnd/openiblnd_cb.c
@@ -1272,7 +1272,7 @@ kibnal_sendmsg(lib_nal_t    *nal,
 
                 /* Will it fit in a message? */
                 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob]);
 
                 /* Will it fit in a message? */
                 nob = offsetof(kib_msg_t, ibm_u.immediate.ibim_payload[payload_nob]);

-                if (nob >= IBNAL_MSG_SIZE) {
+                if (nob > IBNAL_MSG_SIZE) {

                         CERROR("REPLY for "LPX64" too big (RDMA not requested): %d\n", 
                                nid, payload_nob);
                         return (PTL_FAIL);
                         CERROR("REPLY for "LPX64" too big (RDMA not requested): %d\n", 
                                nid, payload_nob);
                         return (PTL_FAIL);

diff --git a/lnet/klnds/ralnd/ralnd.c b/lnet/klnds/ralnd/ralnd.c
index 6333bcf..f272c1e 100644 (file)
--- a/lnet/klnds/ralnd/ralnd.c
+++ b/lnet/klnds/ralnd/ralnd.c
@@ -877,7 +877,7 @@ kranal_conn_handshake (struct socket *sock, kra_peer_t *peer)
 
         /* Schedule all packets blocking for a connection */
         while (!list_empty(&peer->rap_tx_queue)) {
 
         /* Schedule all packets blocking for a connection */
         while (!list_empty(&peer->rap_tx_queue)) {

-                tx = list_entry(&peer->rap_tx_queue.next,
+                tx = list_entry(peer->rap_tx_queue.next,

                                 kra_tx_t, tx_list);
 
                 list_del(&tx->tx_list);
                                 kra_tx_t, tx_list);
 
                 list_del(&tx->tx_list);


@@ -929,14 +929,14 @@ kranal_connect (kra_peer_t *peer)
 
                 /* reset reconnection timeouts */
                 peer->rap_reconnect_interval = RANAL_MIN_RECONNECT_INTERVAL;
 
                 /* reset reconnection timeouts */
                 peer->rap_reconnect_interval = RANAL_MIN_RECONNECT_INTERVAL;

-                peer->rap_reconnect_time = CURRENT_TIME;
+                peer->rap_reconnect_time = CURRENT_SECONDS;

 
                 write_unlock_irqrestore(&kranal_data.kra_global_lock, flags);
                 return;
         }
 
         LASSERT (peer->rap_reconnect_interval != 0);
 
                 write_unlock_irqrestore(&kranal_data.kra_global_lock, flags);
                 return;
         }
 
         LASSERT (peer->rap_reconnect_interval != 0);

-        peer->rap_reconnect_time = CURRENT_TIME + peer->rap_reconnect_interval;
+        peer->rap_reconnect_time = CURRENT_SECONDS + peer->rap_reconnect_interval;

         peer->rap_reconnect_interval = MAX(RANAL_MAX_RECONNECT_INTERVAL,
                                            1 * peer->rap_reconnect_interval);
 
         peer->rap_reconnect_interval = MAX(RANAL_MAX_RECONNECT_INTERVAL,
                                            1 * peer->rap_reconnect_interval);
 


@@ -1268,7 +1268,7 @@ kranal_create_peer (ptl_nid_t nid)
         INIT_LIST_HEAD(&peer->rap_conns);
         INIT_LIST_HEAD(&peer->rap_tx_queue);
 
         INIT_LIST_HEAD(&peer->rap_conns);
         INIT_LIST_HEAD(&peer->rap_tx_queue);
 

-        peer->rap_reconnect_time = CURRENT_TIME;
+        peer->rap_reconnect_time = CURRENT_SECONDS;

         peer->rap_reconnect_interval = RANAL_MIN_RECONNECT_INTERVAL;
 
         atomic_inc(&kranal_data.kra_npeers);
         peer->rap_reconnect_interval = RANAL_MIN_RECONNECT_INTERVAL;
 
         atomic_inc(&kranal_data.kra_npeers);


@@ -1998,7 +1998,7 @@ kranal_api_startup (nal_t *nal, ptl_pid_t requested_pid,
         }
 
         for (i = 0; i < RANAL_N_CONND; i++) {
         }
 
         for (i = 0; i < RANAL_N_CONND; i++) {

-                rc = kranal_thread_start(kranal_connd, (void *)i);
+                rc = kranal_thread_start(kranal_connd, (void *)(unsigned long)i);

                 if (rc != 0) {
                         CERROR("Can't spawn ranal connd[%d]: %d\n",
                                i, rc);
                 if (rc != 0) {
                         CERROR("Can't spawn ranal connd[%d]: %d\n",
                                i, rc);

diff --git a/lnet/klnds/ralnd/ralnd.h b/lnet/klnds/ralnd/ralnd.h
index d904c72..7e43705 100644 (file)
--- a/lnet/klnds/ralnd/ralnd.h
+++ b/lnet/klnds/ralnd/ralnd.h
@@ -49,10 +49,12 @@
 #include <linux/sysctl.h>
 
 #include <net/sock.h>
 #include <linux/sysctl.h>
 
 #include <net/sock.h>

+#include <linux/in.h>

 
 #define DEBUG_SUBSYSTEM S_NAL
 
 #include <linux/kp30.h>
 
 #define DEBUG_SUBSYSTEM S_NAL
 
 #include <linux/kp30.h>

+#include <linux/portals_compat25.h>

 #include <portals/p30.h>
 #include <portals/lib-p30.h>
 #include <portals/nal.h>
 #include <portals/p30.h>
 #include <portals/lib-p30.h>
 #include <portals/nal.h>


@@ -350,7 +352,7 @@ typedef struct kra_peer
         atomic_t            rap_refcount;       /* # users */
         int                 rap_persistence;    /* "known" peer refs */
         int                 rap_connecting;     /* connection forming */
         atomic_t            rap_refcount;       /* # users */
         int                 rap_persistence;    /* "known" peer refs */
         int                 rap_connecting;     /* connection forming */

-        unsigned long       rap_reconnect_time; /* CURRENT_TIME when reconnect OK */
+        unsigned long       rap_reconnect_time; /* CURRENT_SECONDS when reconnect OK */

         unsigned long       rap_reconnect_interval; /* exponential backoff */
 } kra_peer_t;
 
         unsigned long       rap_reconnect_interval; /* exponential backoff */
 } kra_peer_t;
 


@@ -455,17 +457,11 @@ kranal_tx_mapped (kra_tx_t *tx)
                 tx->tx_buftype == RANAL_BUF_PHYS_MAPPED);
 }
 
                 tx->tx_buftype == RANAL_BUF_PHYS_MAPPED);
 }
 

-#if CONFIG_X86

 static inline __u64
 kranal_page2phys (struct page *p)
 {
 static inline __u64
 kranal_page2phys (struct page *p)
 {

-        __u64 page_number = p - mem_map;
-        
-        return (page_number << PAGE_SHIFT);
+        return page_to_phys(p);

 }
 }

-#else
-# error "no page->phys"
-#endif

 
 extern void kranal_free_acceptsock (kra_acceptsock_t *ras);
 extern int kranal_listener_procint (ctl_table *table, 
 
 extern void kranal_free_acceptsock (kra_acceptsock_t *ras);
 extern int kranal_listener_procint (ctl_table *table, 


@@ -479,7 +475,7 @@ extern kra_peer_t *kranal_create_peer (ptl_nid_t nid);
 extern kra_peer_t *kranal_find_peer_locked (ptl_nid_t nid);
 extern void kranal_post_fma (kra_conn_t *conn, kra_tx_t *tx);
 extern int kranal_del_peer (ptl_nid_t nid, int single_share);
 extern kra_peer_t *kranal_find_peer_locked (ptl_nid_t nid);
 extern void kranal_post_fma (kra_conn_t *conn, kra_tx_t *tx);
 extern int kranal_del_peer (ptl_nid_t nid, int single_share);

-extern void kranal_device_callback (RAP_INT32 devid);
+extern void kranal_device_callback (RAP_INT32 devid, RAP_PVOID arg);

 extern int kranal_thread_start (int(*fn)(void *arg), void *arg);
 extern int kranal_connd (void *arg);
 extern int kranal_reaper (void *arg);
 extern int kranal_thread_start (int(*fn)(void *arg), void *arg);
 extern int kranal_connd (void *arg);
 extern int kranal_reaper (void *arg);

diff --git a/lnet/klnds/ralnd/ralnd_cb.c b/lnet/klnds/ralnd/ralnd_cb.c
index d4bacdf..ff080f4 100644 (file)
--- a/lnet/klnds/ralnd/ralnd_cb.c
+++ b/lnet/klnds/ralnd/ralnd_cb.c
@@ -38,7 +38,7 @@ kranal_dist(lib_nal_t *nal, ptl_nid_t nid, unsigned long *dist)
 }
 
 void
 }
 
 void

-kranal_device_callback(RAP_INT32 devid)
+kranal_device_callback(RAP_INT32 devid, RAP_PVOID arg)

 {
         kra_device_t *dev;
         int           i;
 {
         kra_device_t *dev;
         int           i;


@@ -255,13 +255,13 @@ kranal_setup_phys_buffer (kra_tx_t *tx, int nkiov, ptl_kiov_t *kiov,
                         /* Can't have gaps */
                         CERROR("Can't make payload contiguous in I/O VM:"
                                "page %d, offset %d, len %d \n", 
                         /* Can't have gaps */
                         CERROR("Can't make payload contiguous in I/O VM:"
                                "page %d, offset %d, len %d \n", 

-                               phys - tx->tx_phys, 
+                               (int)(phys - tx->tx_phys), 

                                kiov->kiov_offset, kiov->kiov_len);                        
                         return -EINVAL;
                 }
 
                 if ((phys - tx->tx_phys) == PTL_MD_MAX_IOV) {
                                kiov->kiov_offset, kiov->kiov_len);                        
                         return -EINVAL;
                 }
 
                 if ((phys - tx->tx_phys) == PTL_MD_MAX_IOV) {

-                        CERROR ("payload too big (%d)\n", phys - tx->tx_phys);
+                        CERROR ("payload too big (%d)\n", (int)(phys - tx->tx_phys));

                         return -EMSGSIZE;
                 }
 
                         return -EMSGSIZE;
                 }
 


@@ -481,7 +481,7 @@ kranal_launch_tx (kra_tx_t *tx, ptl_nid_t nid)
         if (!peer->rap_connecting) {
                 LASSERT (list_empty(&peer->rap_tx_queue));
                 
         if (!peer->rap_connecting) {
                 LASSERT (list_empty(&peer->rap_tx_queue));
                 

-                now = CURRENT_TIME;
+                now = CURRENT_SECONDS;

                 if (now < peer->rap_reconnect_time) {
                         write_unlock_irqrestore(g_lock, flags);
                         kranal_tx_done(tx, -EHOSTUNREACH);
                 if (now < peer->rap_reconnect_time) {
                         write_unlock_irqrestore(g_lock, flags);
                         kranal_tx_done(tx, -EHOSTUNREACH);


@@ -583,8 +583,8 @@ kranal_do_send (lib_nal_t    *nal,
                 unsigned int  niov, 
                 struct iovec *iov, 
                 ptl_kiov_t   *kiov,
                 unsigned int  niov, 
                 struct iovec *iov, 
                 ptl_kiov_t   *kiov,

-                size_t        offset,
-                size_t        nob)
+                int           offset,
+                int           nob)

 {
         kra_conn_t *conn;
         kra_tx_t   *tx;
 {
         kra_conn_t *conn;
         kra_tx_t   *tx;


@@ -592,8 +592,8 @@ kranal_do_send (lib_nal_t    *nal,
 
         /* NB 'private' is different depending on what we're sending.... */
 
 
         /* NB 'private' is different depending on what we're sending.... */
 

-        CDEBUG(D_NET, "sending "LPSZ" bytes in %d frags to nid:"LPX64
-               " pid %d\n", nob, niov, nid , pid);
+        CDEBUG(D_NET, "sending %d bytes in %d frags to nid:"LPX64" pid %d\n",
+               nob, niov, nid, pid);

 
         LASSERT (nob == 0 || niov > 0);
         LASSERT (niov <= PTL_MD_MAX_IOV);
 
         LASSERT (nob == 0 || niov > 0);
         LASSERT (niov <= PTL_MD_MAX_IOV);


@@ -765,9 +765,9 @@ kranal_send_pages (lib_nal_t *nal, void *private, lib_msg_t *cookie,
 }
 
 ptl_err_t
 }
 
 ptl_err_t

-kranal_recvmsg (lib_nal_t *nal, void *private, lib_msg_t *libmsg,
+kranal_do_recv (lib_nal_t *nal, void *private, lib_msg_t *libmsg,

                 unsigned int niov, struct iovec *iov, ptl_kiov_t *kiov,
                 unsigned int niov, struct iovec *iov, ptl_kiov_t *kiov,

-                size_t offset, size_t mlen, size_t rlen)
+                int offset, int mlen, int rlen)

 {
         kra_conn_t  *conn = private;
         kra_msg_t   *rxmsg = conn->rac_rxmsg;
 {
         kra_conn_t  *conn = private;
         kra_msg_t   *rxmsg = conn->rac_rxmsg;


@@ -860,7 +860,7 @@ kranal_recv (lib_nal_t *nal, void *private, lib_msg_t *msg,
              unsigned int niov, struct iovec *iov, 
              size_t offset, size_t mlen, size_t rlen)
 {
              unsigned int niov, struct iovec *iov, 
              size_t offset, size_t mlen, size_t rlen)
 {

-        return kranal_recvmsg(nal, private, msg, niov, iov, NULL,
+        return kranal_do_recv(nal, private, msg, niov, iov, NULL,

                               offset, mlen, rlen);
 }
 
                               offset, mlen, rlen);
 }
 


@@ -869,7 +869,7 @@ kranal_recv_pages (lib_nal_t *nal, void *private, lib_msg_t *msg,
                    unsigned int niov, ptl_kiov_t *kiov, 
                    size_t offset, size_t mlen, size_t rlen)
 {
                    unsigned int niov, ptl_kiov_t *kiov, 
                    size_t offset, size_t mlen, size_t rlen)
 {

-        return kranal_recvmsg(nal, private, msg, niov, NULL, kiov,
+        return kranal_do_recv(nal, private, msg, niov, NULL, kiov,

                               offset, mlen, rlen);
 }
 
                               offset, mlen, rlen);
 }
 

diff --git a/lnet/utils/portals.c b/lnet/utils/portals.c
index 70440e0..4efca98 100644 (file)
--- a/lnet/utils/portals.c
+++ b/lnet/utils/portals.c
@@ -369,11 +369,6 @@ ptl_parse_nid (ptl_nid_t *nidp, char *str)
         char               *end;
         unsigned long long  ullval;
         
         char               *end;
         unsigned long long  ullval;
         

-        if (!strcmp (str, "_all_")) {
-                *nidp = PTL_NID_ANY;
-                return (0);
-        }
-

         if (ptl_parse_ipaddr (&ipaddr, str) == 0) {
 #if !CRAY_PORTALS
                 *nidp = (ptl_nid_t)ipaddr;
         if (ptl_parse_ipaddr (&ipaddr, str) == 0) {
 #if !CRAY_PORTALS
                 *nidp = (ptl_nid_t)ipaddr;


@@ -384,8 +379,8 @@ ptl_parse_nid (ptl_nid_t *nidp, char *str)
         }
 
         ullval = strtoull(str, &end, 0);
         }
 
         ullval = strtoull(str, &end, 0);

-        if (*end == 0) {
-                /* parsed whole string */
+        if (end != str && *end == 0) {
+                /* parsed whole non-empty string */

                 *nidp = (ptl_nid_t)ullval;
                 return (0);
         }
                 *nidp = (ptl_nid_t)ullval;
                 return (0);
         }


@@ -393,6 +388,17 @@ ptl_parse_nid (ptl_nid_t *nidp, char *str)
         return (-1);
 }
 
         return (-1);
 }
 

+int
+ptl_parse_anynid (ptl_nid_t *nidp, char *str)
+{
+        if (!strcmp (str, "_all_")) {
+                *nidp = PTL_NID_ANY;
+                return 0;
+        }
+
+        return ptl_parse_nid(nidp, str);
+}
+

 __u64 ptl_nid2u64(ptl_nid_t nid)
 {
         switch (sizeof (nid)) {
 __u64 ptl_nid2u64(ptl_nid_t nid)
 {
         switch (sizeof (nid)) {


@@ -791,7 +797,7 @@ jt_ptl_del_peer (int argc, char **argv)
         }
                 
         if (argc > 1 &&
         }
                 
         if (argc > 1 &&

-            ptl_parse_nid (&nid, argv[1]) != 0) {
+            ptl_parse_anynid (&nid, argv[1]) != 0) {

                 fprintf (stderr, "Can't parse nid: %s\n", argv[1]);
                 return -1;
         }
                 fprintf (stderr, "Can't parse nid: %s\n", argv[1]);
                 return -1;
         }


@@ -1040,7 +1046,7 @@ int jt_ptl_disconnect(int argc, char **argv)
                 return 0;
 
         if (argc >= 2 &&
                 return 0;
 
         if (argc >= 2 &&

-            ptl_parse_nid (&nid, argv[1]) != 0) {
+            ptl_parse_anynid (&nid, argv[1]) != 0) {

                 fprintf (stderr, "Can't parse nid %s\n", argv[1]);
                 return -1;
         }
                 fprintf (stderr, "Can't parse nid %s\n", argv[1]);
                 return -1;
         }


@@ -1082,7 +1088,7 @@ int jt_ptl_push_connection (int argc, char **argv)
                 return -1;
         
         if (argc > 1 &&
                 return -1;
         
         if (argc > 1 &&

-            ptl_parse_nid (&nid, argv[1]) != 0) {
+            ptl_parse_anynid (&nid, argv[1]) != 0) {

                 fprintf(stderr, "Can't parse nid: %s\n", argv[1]);
                 return -1;
         }
                 fprintf(stderr, "Can't parse nid: %s\n", argv[1]);
                 return -1;
         }


@@ -1289,7 +1295,7 @@ jt_ptl_fail_nid (int argc, char **argv)
 
         if (!strcmp (argv[1], "_all_"))
                 nid = PTL_NID_ANY;
 
         if (!strcmp (argv[1], "_all_"))
                 nid = PTL_NID_ANY;

-        else if (ptl_parse_nid (&nid, argv[1]) != 0)
+        else if (ptl_parse_anynid (&nid, argv[1]) != 0)

         {
                 fprintf (stderr, "Can't parse nid \"%s\"\n", argv[1]);
                 return (-1);
         {
                 fprintf (stderr, "Can't parse nid \"%s\"\n", argv[1]);
                 return (-1);