Whamcloud - gitweb
Land b_smallfix onto HEAD (20040428_2142)
[fs/lustre-release.git] / lnet / include / linux / kp30.h
index 6d7f3f3..8a56b55 100644 (file)
 #ifndef _KP30_INCLUDED
 #define _KP30_INCLUDED
 
-
+#include <linux/libcfs.h>
 #define PORTAL_DEBUG
 
 #ifndef offsetof
-# define offsetof(typ,memb)    ((int)((char *)&(((typ *)0)->memb)))
-#endif
-
-#define LOWEST_BIT_SET(x)      ((x) & ~((x) - 1))
-
-#ifndef CONFIG_SMP
-# define smp_processor_id() 0
-#endif
-
-/*
- *  Debugging
- */
-extern unsigned int portal_subsystem_debug;
-extern unsigned int portal_stack;
-extern unsigned int portal_debug;
-extern unsigned int portal_printk;
-/* Debugging subsystems  (8 bit ID)
- *
- * If you add debug subsystem #32, you need to send email to phil, because
- * you're going to break kernel subsystem debug filtering. */
-#define S_UNDEFINED    (0 << 24)
-#define S_MDC          (1 << 24)
-#define S_MDS          (2 << 24)
-#define S_OSC          (3 << 24)
-#define S_OST          (4 << 24)
-#define S_CLASS        (5 << 24)
-#define S_OBDFS        (6 << 24) /* obsolete */
-#define S_LLITE        (7 << 24)
-#define S_RPC          (8 << 24)
-#define S_EXT2OBD      (9 << 24) /* obsolete */
-#define S_PORTALS     (10 << 24)
-#define S_SOCKNAL     (11 << 24)
-#define S_QSWNAL      (12 << 24)
-#define S_PINGER      (13 << 24)
-#define S_FILTER      (14 << 24)
-#define S_TRACE       (15 << 24) /* obsolete */
-#define S_ECHO        (16 << 24)
-#define S_LDLM        (17 << 24)
-#define S_LOV         (18 << 24)
-#define S_GMNAL       (19 << 24)
-#define S_PTLROUTER   (20 << 24)
-#define S_COBD        (21 << 24)
-#define S_PTLBD       (22 << 24)
-#define S_LOG         (23 << 24)
-
-/* If you change these values, please keep portals/linux/utils/debug.c
- * up to date! */
-
-/* Debugging masks (24 bits, non-overlapping) */
-#define D_TRACE     (1 << 0) /* ENTRY/EXIT markers */
-#define D_INODE     (1 << 1)
-#define D_SUPER     (1 << 2)
-#define D_EXT2      (1 << 3) /* anything from ext2_debug */
-#define D_MALLOC    (1 << 4) /* print malloc, free information */
-#define D_CACHE     (1 << 5) /* cache-related items */
-#define D_INFO      (1 << 6) /* general information */
-#define D_IOCTL     (1 << 7) /* ioctl related information */
-#define D_BLOCKS    (1 << 8) /* ext2 block allocation */
-#define D_NET       (1 << 9) /* network communications */
-#define D_WARNING   (1 << 10)
-#define D_BUFFS     (1 << 11)
-#define D_OTHER     (1 << 12)
-#define D_DENTRY    (1 << 13)
-#define D_PORTALS   (1 << 14) /* ENTRY/EXIT markers */
-#define D_PAGE      (1 << 15) /* bulk page handling */
-#define D_DLMTRACE  (1 << 16)
-#define D_ERROR     (1 << 17) /* CERROR(...) == CDEBUG (D_ERROR, ...) */
-#define D_EMERG     (1 << 18) /* CEMERG(...) == CDEBUG (D_EMERG, ...) */
-#define D_HA        (1 << 19) /* recovery and failover */
-#define D_RPCTRACE  (1 << 20) /* for distributed debugging */
-#define D_VFSTRACE  (1 << 21)
-
-#ifndef __KERNEL__
-#define THREAD_SIZE 8192
-#endif
-#ifdef  __ia64__
-#define CDEBUG_STACK() (THREAD_SIZE -                                      \
-                        ((unsigned long)__builtin_dwarf_cfa() &            \
-                         (THREAD_SIZE - 1)))
-#else
-#define CDEBUG_STACK() (THREAD_SIZE -                                      \
-                        ((unsigned long)__builtin_frame_address(0) &       \
-                         (THREAD_SIZE - 1)))
-#endif
-
-#ifdef __KERNEL__
-#define CHECK_STACK(stack)                                                    \
-        do {                                                                  \
-                if ((stack) > 3*THREAD_SIZE/4 && (stack) > portal_stack) {    \
-                        portals_debug_msg(DEBUG_SUBSYSTEM, D_ERROR,           \
-                                          __FILE__, __FUNCTION__, __LINE__,   \
-                                          (stack),                            \
-                                          "maximum lustre stack %u\n",        \
-                                          portal_stack = (stack));            \
-                      /*panic("LBUG");*/                                      \
-                }                                                             \
-        } while (0)
-#else
-#define CHECK_STACK(stack) do { } while(0)
-#endif
-
-#if 1
-#define CDEBUG(mask, format, a...)                                            \
-do {                                                                          \
-        CHECK_STACK(CDEBUG_STACK());                                          \
-        if (!(mask) || ((mask) & (D_ERROR | D_EMERG)) ||                      \
-            (portal_debug & (mask) &&                                         \
-             portal_subsystem_debug & (1 << (DEBUG_SUBSYSTEM >> 24))))        \
-                portals_debug_msg(DEBUG_SUBSYSTEM, mask,                      \
-                                  __FILE__, __FUNCTION__, __LINE__,           \
-                                  CDEBUG_STACK(), format , ## a);             \
-} while (0)
-
-#define CWARN(format, a...) CDEBUG(D_WARNING, format, ## a)
-#define CERROR(format, a...) CDEBUG(D_ERROR, format, ## a)
-#define CEMERG(format, a...) CDEBUG(D_EMERG, format, ## a)
-
-#define GOTO(label, rc)                                                 \
-do {                                                                    \
-        long GOTO__ret = (long)(rc);                                    \
-        CDEBUG(D_TRACE,"Process leaving via %s (rc=%lu : %ld : %lx)\n", \
-               #label, (unsigned long)GOTO__ret, (signed long)GOTO__ret,\
-               (signed long)GOTO__ret);                                 \
-        goto label;                                                     \
-} while (0)
-
-#define RETURN(rc)                                                      \
-do {                                                                    \
-        typeof(rc) RETURN__ret = (rc);                                  \
-        CDEBUG(D_TRACE, "Process leaving (rc=%lu : %ld : %lx)\n",       \
-               (long)RETURN__ret, (long)RETURN__ret, (long)RETURN__ret);\
-        return RETURN__ret;                                             \
-} while (0)
-
-#define ENTRY                                                           \
-do {                                                                    \
-        CDEBUG(D_TRACE, "Process entered\n");                           \
-} while (0)
-
-#define EXIT                                                            \
-do {                                                                    \
-        CDEBUG(D_TRACE, "Process leaving\n");                           \
-} while(0)
-#else
-#define CDEBUG(mask, format, a...)      do { } while (0)
-#define CWARN(format, a...)             do { } while (0)
-#define CERROR(format, a...)            printk("<3>" format, ## a)
-#define CEMERG(format, a...)            printk("<0>" format, ## a)
-#define GOTO(label, rc)                 do { (void)(rc); goto label; } while (0)
-#define RETURN(rc)                      return (rc)
-#define ENTRY                           do { } while (0)
-#define EXIT                            do { } while (0)
+# define offsetof(typ,memb)     ((unsigned long)((char *)&(((typ *)0)->memb)))
 #endif
 
+#define LOWEST_BIT_SET(x)       ((x) & ~((x) - 1))
 
 #ifdef __KERNEL__
 # include <linux/vmalloc.h>
@@ -171,7 +21,7 @@ do {                                                                    \
 # include <linux/highmem.h>
 # include <linux/module.h>
 # include <linux/version.h>
-# include <portals/lib-nal.h>
+# include <portals/p30.h>
 # include <linux/smp_lock.h>
 # include <asm/atomic.h>
 
@@ -183,7 +33,6 @@ do {                                                                          \
         PREPARE_TQUEUE((wq), (cb), (cbdata));                                 \
 } while (0)
 
-#define ll_invalidate_inode_pages invalidate_inode_pages
 #define PageUptodate Page_Uptodate
 #define our_recalc_sigpending(current) recalc_sigpending(current)
 #define num_online_cpus() smp_num_cpus
@@ -192,6 +41,7 @@ static inline void our_cond_resched(void)
         if (current->need_resched)
                schedule ();
 }
+#define work_struct_t       struct tq_struct
 
 #else
 
@@ -199,7 +49,6 @@ static inline void our_cond_resched(void)
 do {                                                                          \
         INIT_WORK((wq), (void *)(cb), (void *)(cbdata));                      \
 } while (0)
-#define ll_invalidate_inode_pages(inode) invalidate_inode_pages((inode)->i_mapping)
 #define wait_on_page wait_on_page_locked
 #define our_recalc_sigpending(current) recalc_sigpending()
 #define strtok(a,b) strpbrk(a, b)
@@ -207,14 +56,37 @@ static inline void our_cond_resched(void)
 {
         cond_resched();
 }
+#define work_struct_t      struct work_struct
+
 #endif /* LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0) */
 
 #ifdef PORTAL_DEBUG
-extern void kportal_assertion_failed(char *expr,char *file,char *func,int line);
+extern void kportal_assertion_failed(char *expr, char *file, const char *func,
+                                     const int line);
 #define LASSERT(e) ((e) ? 0 : kportal_assertion_failed( #e , __FILE__,  \
                                                         __FUNCTION__, __LINE__))
+/* it would be great to dump_stack() here, but some kernels
+ * export it as show_stack() and I can't be bothered to
+ * proprely engage in that dance right now */
+#define LASSERTF(cond, fmt...)                                                \
+        do {                                                                  \
+                if (unlikely(!(cond))) {                                      \
+                        portals_debug_msg(0, D_EMERG,  __FILE__, __FUNCTION__,\
+                                          __LINE__,  CDEBUG_STACK,            \
+                                          "ASSERTION(" #cond ") failed:" fmt);\
+                        LBUG();                                               \
+                }                                                             \
+        } while (0)
+
 #else
 #define LASSERT(e)
+#define LASSERTF(cond, fmt...) do { } while (0)
+#endif
+
+#ifdef CONFIG_SMP
+#define LASSERT_SPIN_LOCKED(lock) LASSERT(spin_is_locked(lock))
+#else
+#define LASSERT_SPIN_LOCKED(lock) do {} while(0)
 #endif
 
 #ifdef __arch_um__
@@ -260,30 +132,38 @@ do {                                                                          \
 
 #define PORTAL_VMALLOC_SIZE        16384
 
-#define PORTAL_ALLOC(ptr, size)                                           \
+#define PORTAL_ALLOC_GFP(ptr, size, mask)                                 \
 do {                                                                      \
-        long s = size;                                                    \
-        LASSERT (!in_interrupt());                                        \
-        if (s > PORTAL_VMALLOC_SIZE)                                      \
-                (ptr) = vmalloc(s);                                       \
+        LASSERT(!in_interrupt() ||                                        \
+               (size <= PORTAL_VMALLOC_SIZE && mask == GFP_ATOMIC));      \
+        if ((size) > PORTAL_VMALLOC_SIZE)                                 \
+                (ptr) = vmalloc(size);                                    \
         else                                                              \
-                (ptr) = kmalloc(s, GFP_NOFS);                             \
-        if ((ptr) == NULL)                                                \
-                CERROR("PORTALS: out of memory at %s:%d (tried to alloc"  \
-                       " '" #ptr "' = %ld)\n", __FILE__, __LINE__, s);    \
-        else {                                                            \
-                portal_kmem_inc((ptr), s);                                \
-                memset((ptr), 0, s);                                      \
+                (ptr) = kmalloc((size), (mask));                          \
+        if ((ptr) == NULL) {                                              \
+                CERROR("PORTALS: out of memory at %s:%d (tried to alloc '"\
+                       #ptr "' = %d)\n", __FILE__, __LINE__, (int)(size));\
+                CERROR("PORTALS: %d total bytes allocated by portals\n",  \
+                       atomic_read(&portal_kmemory));                     \
+        } else {                                                          \
+                portal_kmem_inc((ptr), (size));                           \
+                memset((ptr), 0, (size));                                 \
         }                                                                 \
-        CDEBUG(D_MALLOC, "kmalloced '" #ptr "': %ld at %p (tot %d).\n",   \
-               s, (ptr), atomic_read (&portal_kmemory));                  \
+        CDEBUG(D_MALLOC, "kmalloced '" #ptr "': %d at %p (tot %d).\n",    \
+               (int)(size), (ptr), atomic_read (&portal_kmemory));        \
 } while (0)
 
+#define PORTAL_ALLOC(ptr, size) \
+        PORTAL_ALLOC_GFP(ptr, size, GFP_NOFS)
+
+#define PORTAL_ALLOC_ATOMIC(ptr, size) \
+        PORTAL_ALLOC_GFP(ptr, size, GFP_ATOMIC)
+
 #define PORTAL_FREE(ptr, size)                                          \
 do {                                                                    \
-        long s = (size);                                                \
+        int s = (size);                                                 \
         if ((ptr) == NULL) {                                            \
-                CERROR("PORTALS: free NULL '" #ptr "' (%ld bytes) at "  \
+                CERROR("PORTALS: free NULL '" #ptr "' (%d bytes) at "   \
                        "%s:%d\n", s, __FILE__, __LINE__);               \
                 break;                                                  \
         }                                                               \
@@ -292,40 +172,8 @@ do {                                                                    \
         else                                                            \
                 kfree(ptr);                                             \
         portal_kmem_dec((ptr), s);                                      \
-        CDEBUG(D_MALLOC, "kfreed '" #ptr "': %ld at %p (tot %d).\n",    \
-               s, (ptr), atomic_read (&portal_kmemory));                \
-} while (0)
-
-#define PORTAL_SLAB_ALLOC(ptr, slab, size)                                \
-do {                                                                      \
-        long s = (size);                                                  \
-        LASSERT (!in_interrupt());                                        \
-        (ptr) = kmem_cache_alloc((slab), SLAB_KERNEL);                    \
-        if ((ptr) == NULL) {                                              \
-                CERROR("PORTALS: out of memory at %s:%d (tried to alloc"  \
-                       " '" #ptr "' from slab '" #slab "')\n", __FILE__,  \
-                       __LINE__);                                         \
-        } else {                                                          \
-                portal_kmem_inc((ptr), s);                                \
-                memset((ptr), 0, s);                                      \
-        }                                                                 \
-        CDEBUG(D_MALLOC, "kmalloced '" #ptr "': %ld at %p (tot %d).\n",   \
-               s, (ptr), atomic_read (&portal_kmemory));                  \
-} while (0)
-
-#define PORTAL_SLAB_FREE(ptr, slab, size)                               \
-do {                                                                    \
-        long s = (size);                                                \
-        if ((ptr) == NULL) {                                            \
-                CERROR("PORTALS: free NULL '" #ptr "' (%ld bytes) at "  \
-                       "%s:%d\n", s, __FILE__, __LINE__);               \
-                break;                                                  \
-        }                                                               \
-        memset((ptr), 0x5a, s);                                         \
-        kmem_cache_free((slab), ptr);                                   \
-        portal_kmem_dec((ptr), s);                                      \
-        CDEBUG(D_MALLOC, "kfreed '" #ptr "': %ld at %p (tot %d).\n",    \
-               s, (ptr), atomic_read (&portal_kmemory));                \
+        CDEBUG(D_MALLOC, "kfreed '" #ptr "': %d at %p (tot %d).\n",     \
+               s, (ptr), atomic_read(&portal_kmemory));                 \
 } while (0)
 
 /* ------------------------------------------------------------------- */
@@ -354,165 +202,6 @@ do {                                                                    \
 #endif
 
 /******************************************************************************/
-/* Kernel Portals Router interface */
-
-typedef void (*kpr_fwd_callback_t)(void *arg, int error); // completion callback
-
-/* space for routing targets to stash "stuff" in a forwarded packet */
-typedef union {
-        long long        _alignment;
-        void            *_space[16];            /* scale with CPU arch */
-} kprfd_scratch_t;
-
-/* Kernel Portals Routing Forwarded message Descriptor */
-typedef struct {
-        struct list_head     kprfd_list;        /* stash in queues (routing target can use) */
-        ptl_nid_t            kprfd_target_nid;  /* final destination NID */
-        ptl_nid_t            kprfd_gateway_nid; /* gateway NID */
-        int                  kprfd_nob;         /* # message bytes (including header) */
-        int                  kprfd_niov;        /* # message frags (including header) */
-        struct iovec        *kprfd_iov;         /* message fragments */
-        void                *kprfd_router_arg;  // originating NAL's router arg
-        kpr_fwd_callback_t   kprfd_callback;    /* completion callback */
-        void                *kprfd_callback_arg; /* completion callback arg */
-        kprfd_scratch_t      kprfd_scratch;    // scratchpad for routing targets
-} kpr_fwd_desc_t;
-
-typedef void  (*kpr_fwd_t)(void *arg, kpr_fwd_desc_t *fwd);
-
-/* NAL's routing interface (Kernel Portals Routing Nal Interface) */
-typedef const struct {
-        int             kprni_nalid;    /* NAL's id */
-        void           *kprni_arg;      /* Arg to pass when calling into NAL */
-        kpr_fwd_t       kprni_fwd;      /* NAL's forwarding entrypoint */
-} kpr_nal_interface_t;
-
-/* Router's routing interface (Kernel Portals Routing Router Interface) */
-typedef const struct {
-        /* register the calling NAL with the router and get back the handle for
-         * subsequent calls */
-        int     (*kprri_register) (kpr_nal_interface_t *nal_interface,
-                                   void **router_arg);
-
-        /* ask the router to find a gateway that forwards to 'nid' and is a peer
-         * of the calling NAL */
-        int     (*kprri_lookup) (void *router_arg, ptl_nid_t nid,
-                                 ptl_nid_t *gateway_nid);
-
-        /* hand a packet over to the router for forwarding */
-        kpr_fwd_t kprri_fwd_start;
-
-        /* hand a packet back to the router for completion */
-        void    (*kprri_fwd_done) (void *router_arg, kpr_fwd_desc_t *fwd,
-                                   int error);
-
-        /* the calling NAL is shutting down */
-        void    (*kprri_shutdown) (void *router_arg);
-
-        /* deregister the calling NAL with the router */
-        void    (*kprri_deregister) (void *router_arg);
-
-} kpr_router_interface_t;
-
-/* Convenient struct for NAL to stash router interface/args */
-typedef struct {
-        kpr_router_interface_t  *kpr_interface;
-        void                    *kpr_arg;
-} kpr_router_t;
-
-/* Router's control interface (Kernel Portals Routing Control Interface) */
-typedef const struct {
-        int     (*kprci_add_route)(int gateway_nal, ptl_nid_t gateway_nid,
-                                   ptl_nid_t lo_nid, ptl_nid_t hi_nid);
-        int     (*kprci_del_route)(ptl_nid_t nid);
-        int     (*kprci_get_route)(int index, int *gateway_nal,
-                                   ptl_nid_t *gateway, ptl_nid_t *lo_nid,
-                                   ptl_nid_t *hi_nid);
-} kpr_control_interface_t;
-
-extern kpr_control_interface_t  kpr_control_interface;
-extern kpr_router_interface_t   kpr_router_interface;
-
-static inline int
-kpr_register (kpr_router_t *router, kpr_nal_interface_t *nalif)
-{
-        int    rc;
-
-        router->kpr_interface = PORTAL_SYMBOL_GET (kpr_router_interface);
-        if (router->kpr_interface == NULL)
-                return (-ENOENT);
-
-        rc = (router->kpr_interface)->kprri_register (nalif, &router->kpr_arg);
-        if (rc != 0)
-                router->kpr_interface = NULL;
-
-        PORTAL_SYMBOL_PUT (kpr_router_interface);
-        return (rc);
-}
-
-static inline int
-kpr_routing (kpr_router_t *router)
-{
-        return (router->kpr_interface != NULL);
-}
-
-static inline int
-kpr_lookup (kpr_router_t *router, ptl_nid_t nid, ptl_nid_t *gateway_nid)
-{
-        if (!kpr_routing (router))
-                return (-EHOSTUNREACH);
-
-        return (router->kpr_interface->kprri_lookup(router->kpr_arg, nid,
-                                                    gateway_nid));
-}
-
-static inline void
-kpr_fwd_init (kpr_fwd_desc_t *fwd, ptl_nid_t nid,
-              int nob, int niov, struct iovec *iov,
-              kpr_fwd_callback_t callback, void *callback_arg)
-{
-        fwd->kprfd_target_nid   = nid;
-        fwd->kprfd_gateway_nid  = nid;
-        fwd->kprfd_nob          = nob;
-        fwd->kprfd_niov         = niov;
-        fwd->kprfd_iov          = iov;
-        fwd->kprfd_callback     = callback;
-        fwd->kprfd_callback_arg = callback_arg;
-}
-
-static inline void
-kpr_fwd_start (kpr_router_t *router, kpr_fwd_desc_t *fwd)
-{
-        if (!kpr_routing (router))
-                fwd->kprfd_callback (fwd->kprfd_callback_arg, -EHOSTUNREACH);
-        else
-                router->kpr_interface->kprri_fwd_start (router->kpr_arg, fwd);
-}
-
-static inline void
-kpr_fwd_done (kpr_router_t *router, kpr_fwd_desc_t *fwd, int error)
-{
-        LASSERT (kpr_routing (router));
-        router->kpr_interface->kprri_fwd_done (router->kpr_arg, fwd, error);
-}
-
-static inline void
-kpr_shutdown (kpr_router_t *router)
-{
-        if (kpr_routing (router))
-                router->kpr_interface->kprri_shutdown (router->kpr_arg);
-}
-
-static inline void
-kpr_deregister (kpr_router_t *router)
-{
-        if (!kpr_routing (router))
-                return;
-        router->kpr_interface->kprri_deregister (router->kpr_arg);
-        router->kpr_interface = NULL;
-}
-
-/******************************************************************************/
 
 #ifdef PORTALS_PROFILING
 #define prof_enum(FOO) PROF__##FOO
@@ -562,14 +251,18 @@ extern struct prof_ent prof_ents[MAX_PROFS];
 #endif /* PORTALS_PROFILING */
 
 /* debug.c */
-void portals_run_lbug_upcall(char * file, char *fn, int line);
+extern spinlock_t stack_backtrace_lock;
+
+char *portals_debug_dumpstack(void);
+void portals_run_upcall(char **argv);
+void portals_run_lbug_upcall(char * file, const char *fn, const int line);
 void portals_debug_dumplog(void);
 int portals_debug_init(unsigned long bufsize);
 int portals_debug_cleanup(void);
 int portals_debug_clear_buffer(void);
 int portals_debug_mark_buffer(char *text);
 int portals_debug_set_daemon(unsigned int cmd, unsigned int length,
-                char *file, unsigned int size);
+                             char *file, unsigned int size);
 __s32 portals_debug_copy_to_user(char *buf, unsigned long len);
 #if (__GNUC__)
 /* Use the special GNU C __attribute__ hack to have the compiler check the
@@ -580,13 +273,14 @@ __s32 portals_debug_copy_to_user(char *buf, unsigned long len);
 # warning printf has been defined as a macro...
 # undef printf
 #endif
-void portals_debug_msg (int subsys, int mask, char *file, char *fn, int line,
-                        unsigned long stack, const char *format, ...)
+void portals_debug_msg(int subsys, int mask, char *file, const char *fn,
+                       const int line, unsigned long stack,
+                       char *format, ...)
         __attribute__ ((format (printf, 7, 8)));
 #else
-void portals_debug_msg (int subsys, int mask, char *file, char *fn,
-                        int line, unsigned long stack,
-                        const char *format, ...);
+void portals_debug_msg(int subsys, int mask, char *file, const char *fn,
+                       const int line, unsigned long stack,
+                       const char *format, ...);
 #endif /* __GNUC__ */
 void portals_debug_set_level(unsigned int debug_level);
 
@@ -602,9 +296,12 @@ extern void kportal_blockallsigs (void);
 # include <stdlib.h>
 #ifndef __CYGWIN__
 # include <stdint.h>
+#else
+# include <cygwin-ioctl.h>
 #endif
 # include <unistd.h>
 # include <time.h>
+# include <limits.h>
 # include <asm/types.h>
 # ifndef DEBUG_SUBSYSTEM
 #  define DEBUG_SUBSYSTEM S_UNDEFINED
@@ -613,22 +310,125 @@ extern void kportal_blockallsigs (void);
 #  undef NDEBUG
 #  include <assert.h>
 #  define LASSERT(e)     assert(e)
+#  define LASSERTF(cond, args...)     assert(cond)
 # else
 #  define LASSERT(e)
+#  define LASSERTF(cond, args...) do { } while (0)
 # endif
 # define printk(format, args...) printf (format, ## args)
 # define PORTAL_ALLOC(ptr, size) do { (ptr) = malloc(size); } while (0);
 # define PORTAL_FREE(a, b) do { free(a); } while (0);
+void portals_debug_dumplog(void);
 # define portals_debug_msg(subsys, mask, file, fn, line, stack, format, a...) \
-    printf ("%02x:%06x (@%lu %s:%s,l. %d %d %lu): " format,                    \
-            (subsys) >> 24, (mask), (long)time(0), file, fn, line,            \
-            getpid() , stack, ## a);
+    printf("%02x:%06x (@%lu %s:%s,l. %d %d %lu): " format,                    \
+           (subsys), (mask), (long)time(0), file, fn, line,                   \
+           getpid() , stack, ## a);
 #endif
 
+/* support decl needed both by kernel and liblustre */
+char *portals_nid2str(int nal, ptl_nid_t nid, char *str);
+
 #ifndef CURRENT_TIME
 # define CURRENT_TIME time(0)
 #endif
 
+/******************************************************************************/
+/* Light-weight trace
+ * Support for temporary event tracing with minimal Heisenberg effect. */
+#define LWT_SUPPORT  0
+
+#define LWT_MEMORY   (64<<20)
+#define LWT_MAX_CPUS 4
+
+typedef struct {
+        cycles_t    lwte_when;
+        char       *lwte_where;
+        void       *lwte_task;
+        long        lwte_p1;
+        long        lwte_p2;
+        long        lwte_p3;
+        long        lwte_p4;
+#if BITS_PER_LONG > 32
+        long        lwte_pad;
+#endif
+} lwt_event_t;
+
+#if LWT_SUPPORT
+#ifdef __KERNEL__
+#define LWT_EVENTS_PER_PAGE (PAGE_SIZE / sizeof (lwt_event_t))
+
+typedef struct _lwt_page {
+        struct list_head     lwtp_list;
+        struct page         *lwtp_page;
+        lwt_event_t         *lwtp_events;
+} lwt_page_t;
+
+typedef struct {
+        int                lwtc_current_index;
+        lwt_page_t        *lwtc_current_page;
+} lwt_cpu_t;
+
+extern int       lwt_enabled;
+extern lwt_cpu_t lwt_cpus[];
+
+extern int  lwt_init (void);
+extern void lwt_fini (void);
+extern int  lwt_lookup_string (int *size, char *knlptr,
+                               char *usrptr, int usrsize);
+extern int  lwt_control (int enable, int clear);
+extern int  lwt_snapshot (cycles_t *now, int *ncpu, int *total_size,
+                          void *user_ptr, int user_size);
+
+/* Note that we _don't_ define LWT_EVENT at all if LWT_SUPPORT isn't set.
+ * This stuff is meant for finding specific problems; it never stays in
+ * production code... */
+
+#define LWTSTR(n)       #n
+#define LWTWHERE(f,l)   f ":" LWTSTR(l)
+
+#define LWT_EVENT(p1, p2, p3, p4)                                       \
+do {                                                                    \
+        unsigned long    flags;                                         \
+        lwt_cpu_t       *cpu;                                           \
+        lwt_page_t      *p;                                             \
+        lwt_event_t     *e;                                             \
+                                                                        \
+        local_irq_save (flags);                                         \
+                                                                        \
+        if (lwt_enabled) {                                              \
+                cpu = &lwt_cpus[smp_processor_id()];                    \
+                p = cpu->lwtc_current_page;                             \
+                e = &p->lwtp_events[cpu->lwtc_current_index++];         \
+                                                                        \
+                if (cpu->lwtc_current_index >= LWT_EVENTS_PER_PAGE) {   \
+                        cpu->lwtc_current_page =                        \
+                                list_entry (p->lwtp_list.next,          \
+                                            lwt_page_t, lwtp_list);     \
+                        cpu->lwtc_current_index = 0;                    \
+                }                                                       \
+                                                                        \
+                e->lwte_when  = get_cycles();                           \
+                e->lwte_where = LWTWHERE(__FILE__,__LINE__);            \
+                e->lwte_task  = current;                                \
+                e->lwte_p1    = (long)(p1);                             \
+                e->lwte_p2    = (long)(p2);                             \
+                e->lwte_p3    = (long)(p3);                             \
+                e->lwte_p4    = (long)(p4);                             \
+        }                                                               \
+                                                                        \
+        local_irq_restore (flags);                                      \
+} while (0)
+#else  /* __KERNEL__ */
+#define LWT_EVENT(p1,p2,p3,p4)     /* no userland implementation yet */
+#endif /* __KERNEL__ */
+#endif /* LWT_SUPPORT */
+
+struct portals_device_userstate
+{
+        int          pdu_memhog_pages;
+        struct page *pdu_memhog_root_page;
+};
+
 #include <linux/portals_lib.h>
 
 /*
@@ -639,38 +439,6 @@ extern void kportal_blockallsigs (void);
 #define PING_SYNC       0
 #define PING_ASYNC      1
 
-struct portal_ioctl_data {
-        __u32 ioc_len;
-        __u32 ioc_version;
-        __u64 ioc_nid;
-        __u64 ioc_nid2;
-        __u64 ioc_nid3;
-        __u32 ioc_count;
-        __u32 ioc_nal;
-        __u32 ioc_nal_cmd;
-        __u32 ioc_fd;
-        __u32 ioc_id;
-
-        __u32 ioc_flags;
-        __u32 ioc_size;
-
-        __u32 ioc_wait;
-        __u32 ioc_timeout;
-        __u32 ioc_misc;
-
-        __u32 ioc_inllen1;
-        char *ioc_inlbuf1;
-        __u32 ioc_inllen2;
-        char *ioc_inlbuf2;
-
-        __u32 ioc_plen1; /* buffers in userspace */
-        char *ioc_pbuf1;
-        __u32 ioc_plen2; /* buffers in userspace */
-        char *ioc_pbuf2;
-
-        char ioc_bulk[0];
-};
-
 struct portal_ioctl_hdr {
         __u32 ioc_len;
         __u32 ioc_version;
@@ -856,33 +624,29 @@ static inline int portal_ioctl_getdata(char *buf, char *end, void *arg)
 #define IOC_PORTAL_CLEAR_DEBUG             _IOWR('e', 32, long)
 #define IOC_PORTAL_MARK_DEBUG              _IOWR('e', 33, long)
 #define IOC_PORTAL_PANIC                   _IOWR('e', 34, long)
-#define IOC_PORTAL_ADD_ROUTE               _IOWR('e', 35, long)
-#define IOC_PORTAL_DEL_ROUTE               _IOWR('e', 36, long)
-#define IOC_PORTAL_GET_ROUTE               _IOWR('e', 37, long)
-#define IOC_PORTAL_NAL_CMD                _IOWR('e', 38, long)
-#define IOC_PORTAL_GET_NID                 _IOWR('e', 39, long)
-#define IOC_PORTAL_FAIL_NID                _IOWR('e', 40, long)
-#define IOC_PORTAL_SET_DAEMON              _IOWR('e', 41, long)
-
-#define IOC_PORTAL_MAX_NR               41
+#define IOC_PORTAL_NAL_CMD                 _IOWR('e', 35, long)
+#define IOC_PORTAL_GET_NID                 _IOWR('e', 36, long)
+#define IOC_PORTAL_FAIL_NID                _IOWR('e', 37, long)
+#define IOC_PORTAL_SET_DAEMON              _IOWR('e', 38, long)
+#define IOC_PORTAL_LWT_CONTROL             _IOWR('e', 39, long)
+#define IOC_PORTAL_LWT_SNAPSHOT            _IOWR('e', 40, long)
+#define IOC_PORTAL_LWT_LOOKUP_STRING       _IOWR('e', 41, long)
+#define IOC_PORTAL_MEMHOG                  _IOWR('e', 42, long)
+#define IOC_PORTAL_MAX_NR                             42
 
 enum {
-        QSWNAL  =  1,
-        SOCKNAL,
-        GMNAL,
-        TOENAL,
-        TCPNAL,
-        SCIMACNAL,
+        QSWNAL    = 1,
+        SOCKNAL   = 2,
+        GMNAL     = 3,
+        /*          4 unused */
+        TCPNAL    = 5,
+        ROUTER    = 6,
+        IBNAL     = 7,
+        CRAY_KB_ERNAL = 8,
         NAL_ENUM_END_MARKER
 };
 
-#ifdef __KERNEL__
-extern ptl_handle_ni_t  kqswnal_ni;
-extern ptl_handle_ni_t  ksocknal_ni;
-extern ptl_handle_ni_t  ktoenal_ni;
-extern ptl_handle_ni_t  kgmnal_ni;
-extern ptl_handle_ni_t  kscimacnal_ni;
-#endif
+#define PTL_NALFMT_SIZE               26 /* %u:%u.%u.%u.%u (10+4+4+4+3+1) */
 
 #define NAL_MAX_NR (NAL_ENUM_END_MARKER - 1)
 
@@ -890,6 +654,15 @@ extern ptl_handle_ni_t  kscimacnal_ni;
 #define NAL_CMD_CLOSE_CONNECTION     101
 #define NAL_CMD_REGISTER_MYNID       102
 #define NAL_CMD_PUSH_CONNECTION      103
+#define NAL_CMD_GET_CONN             104
+#define NAL_CMD_DEL_AUTOCONN         105
+#define NAL_CMD_ADD_AUTOCONN         106
+#define NAL_CMD_GET_AUTOCONN         107
+#define NAL_CMD_GET_TXDESC           108
+#define NAL_CMD_ADD_ROUTE            109
+#define NAL_CMD_DEL_ROUTE            110
+#define NAL_CMD_GET_ROUTE            111
+#define NAL_CMD_NOTIFY_ROUTER        112
 
 enum {
         DEBUG_DAEMON_START       =  1,
@@ -898,38 +671,43 @@ enum {
         DEBUG_DAEMON_CONTINUE    =  4,
 };
 
-/* XXX remove to lustre ASAP */
-struct lustre_peer {
-        ptl_nid_t       peer_nid;
-        ptl_handle_ni_t peer_ni;
-};
 
-/* module.c */
-typedef int (*nal_cmd_handler_t)(struct portal_ioctl_data *, void * private);
-int kportal_nal_register(int nal, nal_cmd_handler_t handler, void * private);
-int kportal_nal_unregister(int nal);
+enum cfg_record_type {
+        PORTALS_CFG_TYPE = 1,
+        LUSTRE_CFG_TYPE = 123,
+};
 
-ptl_handle_ni_t *kportal_get_ni (int nal);
-void kportal_put_ni (int nal);
+typedef int (*cfg_record_cb_t)(enum cfg_record_type, int len, void *data);
 
 #ifdef __CYGWIN__
-#ifndef BITS_PER_LONG
-#if (~0UL) == 0xffffffffUL
-#define BITS_PER_LONG 32
-#else
-#define BITS_PER_LONG 64
-#endif
+# ifndef BITS_PER_LONG
+#  if (~0UL) == 0xffffffffUL
+#   define BITS_PER_LONG 32
+#  else
+#   define BITS_PER_LONG 64
+#  endif
+# endif
 #endif
+
+#ifndef LP_POISON
+# define LI_POISON ((int)0x5a5a5a5a5a5a5a5a)
+# define LL_POISON ((long)0x5a5a5a5a5a5a5a5a)
+# define LP_POISON ((void *)(long)0x5a5a5a5a5a5a5a5a)
 #endif
 
-#if (BITS_PER_LONG == 32 || __WORDSIZE == 32)
+#if defined(__x86_64__)
+# define LPU64 "%Lu"
+# define LPD64 "%Ld"
+# define LPX64 "%#Lx"
+# define LPSZ  "%lu"
+# define LPSSZ "%ld"
+#elif (BITS_PER_LONG == 32 || __WORDSIZE == 32)
 # define LPU64 "%Lu"
 # define LPD64 "%Ld"
 # define LPX64 "%#Lx"
 # define LPSZ  "%u"
 # define LPSSZ "%d"
-#endif
-#if (BITS_PER_LONG == 64 || __WORDSIZE == 64)
+#elif (BITS_PER_LONG == 64 || __WORDSIZE == 64)
 # define LPU64 "%lu"
 # define LPD64 "%ld"
 # define LPX64 "%#lx"