tdb2: make tdb1 use same low-level lock functions.

This means they will use the TDB_ATTRIBUTE_FLOCK functions, and get automatic
reopen support.

diff --git a/ccan/tdb2/lock.c b/ccan/tdb2/lock.c
index 957adf83358057d458d09320d1ce546feb62bffe..173130b2d57eefaf315b2a00ad55e4a5e90e1536 100644 (file)
--- a/ccan/tdb2/lock.c
+++ b/ccan/tdb2/lock.c
@@ -184,9 +184,9 @@ static int unlock(struct tdb_context *tdb, int rw, off_t off, off_t len)
 
    note that a len of zero means lock to end of file
 */
 
    note that a len of zero means lock to end of file
 */

-static enum TDB_ERROR tdb_brlock(struct tdb_context *tdb,
-                                int rw_type, tdb_off_t offset, tdb_off_t len,
-                                enum tdb_lock_flags flags)
+enum TDB_ERROR tdb_brlock(struct tdb_context *tdb,
+                         int rw_type, tdb_off_t offset, tdb_off_t len,
+                         enum tdb_lock_flags flags)

 {
        int ret;
 
 {
        int ret;
 


@@ -226,8 +226,8 @@ static enum TDB_ERROR tdb_brlock(struct tdb_context *tdb,
        return TDB_SUCCESS;
 }
 
        return TDB_SUCCESS;
 }
 

-static enum TDB_ERROR tdb_brunlock(struct tdb_context *tdb,
-                                  int rw_type, tdb_off_t offset, size_t len)
+enum TDB_ERROR tdb_brunlock(struct tdb_context *tdb,
+                           int rw_type, tdb_off_t offset, size_t len)

 {
        if (tdb->flags & TDB_NOLOCK) {
                return TDB_SUCCESS;
 {
        if (tdb->flags & TDB_NOLOCK) {
                return TDB_SUCCESS;


@@ -252,7 +252,7 @@ static enum TDB_ERROR tdb_brunlock(struct tdb_context *tdb,
   deadlock detection and claim a deadlock when progress can be
   made. For those OSes we may loop for a while.
 */
   deadlock detection and claim a deadlock when progress can be
   made. For those OSes we may loop for a while.
 */

-enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb)
+enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb, off_t start)

 {
        int count = 1000;
 
 {
        int count = 1000;
 


@@ -278,8 +278,7 @@ enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb)
 
        while (count--) {
                struct timeval tv;
 
        while (count--) {
                struct timeval tv;

-               if (tdb_brlock(tdb, F_WRLCK,
-                              TDB_HASH_LOCK_START, 0,
+               if (tdb_brlock(tdb, F_WRLCK, start, 0,

                               TDB_LOCK_WAIT|TDB_LOCK_PROBE) == TDB_SUCCESS) {
                        tdb->file->allrecord_lock.ltype = F_WRLCK;
                        tdb->file->allrecord_lock.off = 0;
                               TDB_LOCK_WAIT|TDB_LOCK_PROBE) == TDB_SUCCESS) {
                        tdb->file->allrecord_lock.ltype = F_WRLCK;
                        tdb->file->allrecord_lock.off = 0;


@@ -341,15 +340,16 @@ enum TDB_ERROR tdb_lock_and_recover(struct tdb_context *tdb)
 }
 
 /* lock an offset in the database. */
 }
 
 /* lock an offset in the database. */

-static enum TDB_ERROR tdb_nest_lock(struct tdb_context *tdb,
-                                   tdb_off_t offset, int ltype,
-                                   enum tdb_lock_flags flags)
+enum TDB_ERROR tdb_nest_lock(struct tdb_context *tdb,
+                            tdb_off_t offset, int ltype,
+                            enum tdb_lock_flags flags)

 {
        struct tdb_lock *new_lck;
        enum TDB_ERROR ecode;
 
 {
        struct tdb_lock *new_lck;
        enum TDB_ERROR ecode;
 

-       if (offset > (TDB_HASH_LOCK_START + TDB_HASH_LOCK_RANGE
-                     + tdb->file->map_size / 8)) {
+       if (!(tdb->flags & TDB_VERSION1)
+           && offset > (TDB_HASH_LOCK_START + TDB_HASH_LOCK_RANGE
+                        + tdb->file->map_size / 8)) {

                return tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
                                  "tdb_nest_lock: invalid offset %zu ltype=%d",
                                  (size_t)offset, ltype);
                return tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
                                  "tdb_nest_lock: invalid offset %zu ltype=%d",
                                  (size_t)offset, ltype);


@@ -437,8 +437,8 @@ static enum TDB_ERROR tdb_nest_lock(struct tdb_context *tdb,
        return TDB_SUCCESS;
 }
 
        return TDB_SUCCESS;
 }
 

-static enum TDB_ERROR tdb_nest_unlock(struct tdb_context *tdb,
-                                     tdb_off_t off, int ltype)
+enum TDB_ERROR tdb_nest_unlock(struct tdb_context *tdb,
+                              tdb_off_t off, int ltype)

 {
        struct tdb_lock *lck;
        enum TDB_ERROR ecode;
 {
        struct tdb_lock *lck;
        enum TDB_ERROR ecode;


@@ -493,9 +493,9 @@ void tdb_transaction_unlock(struct tdb_context *tdb, int ltype)
 
 /* We only need to lock individual bytes, but Linux merges consecutive locks
  * so we lock in contiguous ranges. */
 
 /* We only need to lock individual bytes, but Linux merges consecutive locks
  * so we lock in contiguous ranges. */

-static enum TDB_ERROR tdb_lock_gradual(struct tdb_context *tdb,
-                                      int ltype, enum tdb_lock_flags flags,
-                                      tdb_off_t off, tdb_off_t len)
+enum TDB_ERROR tdb_lock_gradual(struct tdb_context *tdb,
+                               int ltype, enum tdb_lock_flags flags,
+                               tdb_off_t off, tdb_off_t len)

 {
        enum TDB_ERROR ecode;
        enum tdb_lock_flags nb_flags = (flags & ~TDB_LOCK_WAIT);
 {
        enum TDB_ERROR ecode;
        enum tdb_lock_flags nb_flags = (flags & ~TDB_LOCK_WAIT);

diff --git a/ccan/tdb2/private.h b/ccan/tdb2/private.h
index 89e69610d428fe82edf0fc71d160f69e89489225..fdd505357433fadf607c98d96f385530b509ba6f 100644 (file)
--- a/ccan/tdb2/private.h
+++ b/ccan/tdb2/private.h
@@ -545,7 +545,7 @@ bool tdb_has_hash_locks(struct tdb_context *tdb);
 enum TDB_ERROR tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
                                  enum tdb_lock_flags flags, bool upgradable);
 void tdb_allrecord_unlock(struct tdb_context *tdb, int ltype);
 enum TDB_ERROR tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
                                  enum tdb_lock_flags flags, bool upgradable);
 void tdb_allrecord_unlock(struct tdb_context *tdb, int ltype);

-enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb);
+enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb, off_t start);

 
 /* Serialize db open. */
 enum TDB_ERROR tdb_lock_open(struct tdb_context *tdb,
 
 /* Serialize db open. */
 enum TDB_ERROR tdb_lock_open(struct tdb_context *tdb,


@@ -561,6 +561,25 @@ bool tdb_has_expansion_lock(struct tdb_context *tdb);
 /* If it needs recovery, grab all the locks and do it. */
 enum TDB_ERROR tdb_lock_and_recover(struct tdb_context *tdb);
 
 /* If it needs recovery, grab all the locks and do it. */
 enum TDB_ERROR tdb_lock_and_recover(struct tdb_context *tdb);
 

+/* Byte-range lock wrappers for TDB1 to access. */
+enum TDB_ERROR tdb_brlock(struct tdb_context *tdb,
+                         int rw_type, tdb_off_t offset, tdb_off_t len,
+                         enum tdb_lock_flags flags);
+
+enum TDB_ERROR tdb_brunlock(struct tdb_context *tdb,
+                           int rw_type, tdb_off_t offset, size_t len);
+
+enum TDB_ERROR tdb_nest_lock(struct tdb_context *tdb,
+                            tdb_off_t offset, int ltype,
+                            enum tdb_lock_flags flags);
+
+enum TDB_ERROR tdb_nest_unlock(struct tdb_context *tdb,
+                              tdb_off_t off, int ltype);
+
+enum TDB_ERROR tdb_lock_gradual(struct tdb_context *tdb,
+                               int ltype, enum tdb_lock_flags flags,
+                               tdb_off_t off, tdb_off_t len);
+

 /* Default lock and unlock functions. */
 int tdb_fcntl_lock(int fd, int rw, off_t off, off_t len, bool waitflag, void *);
 int tdb_fcntl_unlock(int fd, int rw, off_t off, off_t len, void *);
 /* Default lock and unlock functions. */
 int tdb_fcntl_lock(int fd, int rw, off_t off, off_t len, bool waitflag, void *);
 int tdb_fcntl_unlock(int fd, int rw, off_t off, off_t len, void *);

diff --git a/ccan/tdb2/tdb1_lock.c b/ccan/tdb2/tdb1_lock.c
index 85348df04dc67c0a4e2504127bf4b7ccd1e2a2b1..2252e4791b91c9d8d2afb392b9fd3c87f86d9056 100644 (file)
--- a/ccan/tdb2/tdb1_lock.c
+++ b/ccan/tdb2/tdb1_lock.c
@@ -27,93 +27,6 @@
 
 #include "tdb1_private.h"
 
 
 #include "tdb1_private.h"
 

-static int fcntl_lock(struct tdb1_context *tdb,
-                     int rw, off_t off, off_t len, bool waitflag)
-{
-       struct flock fl;
-
-       fl.l_type = rw;
-       fl.l_whence = SEEK_SET;
-       fl.l_start = off;
-       fl.l_len = len;
-       fl.l_pid = 0;
-
-       if (waitflag)
-               return fcntl(tdb->file->fd, F_SETLKW, &fl);
-       else
-               return fcntl(tdb->file->fd, F_SETLK, &fl);
-}
-
-static int fcntl_unlock(struct tdb1_context *tdb, int rw, off_t off, off_t len)
-{
-       struct flock fl;
-#if 0 /* Check they matched up locks and unlocks correctly. */
-       char line[80];
-       FILE *locks;
-       bool found = false;
-
-       locks = fopen("/proc/locks", "r");
-
-       while (fgets(line, 80, locks)) {
-               char *p;
-               int type, start, l;
-
-               /* eg. 1: FLOCK  ADVISORY  WRITE 2440 08:01:2180826 0 EOF */
-               p = strchr(line, ':') + 1;
-               if (strncmp(p, " POSIX  ADVISORY  ", strlen(" POSIX  ADVISORY  ")))
-                       continue;
-               p += strlen(" FLOCK  ADVISORY  ");
-               if (strncmp(p, "READ  ", strlen("READ  ")) == 0)
-                       type = F_RDLCK;
-               else if (strncmp(p, "WRITE ", strlen("WRITE ")) == 0)
-                       type = F_WRLCK;
-               else
-                       abort();
-               p += 6;
-               if (atoi(p) != getpid())
-                       continue;
-               p = strchr(strchr(p, ' ') + 1, ' ') + 1;
-               start = atoi(p);
-               p = strchr(p, ' ') + 1;
-               if (strncmp(p, "EOF", 3) == 0)
-                       l = 0;
-               else
-                       l = atoi(p) - start + 1;
-
-               if (off == start) {
-                       if (len != l) {
-                               fprintf(stderr, "Len %u should be %u: %s",
-                                       (int)len, l, line);
-                               abort();
-                       }
-                       if (type != rw) {
-                               fprintf(stderr, "Type %s wrong: %s",
-                                       rw == F_RDLCK ? "READ" : "WRITE", line);
-                               abort();
-                       }
-                       found = true;
-                       break;
-               }
-       }
-
-       if (!found) {
-               fprintf(stderr, "Unlock on %u@%u not found!\n",
-                       (int)off, (int)len);
-               abort();
-       }
-
-       fclose(locks);
-#endif
-
-       fl.l_type = F_UNLCK;
-       fl.l_whence = SEEK_SET;
-       fl.l_start = off;
-       fl.l_len = len;
-       fl.l_pid = 0;
-
-       return fcntl(tdb->file->fd, F_SETLKW, &fl);
-}
-

 /* list -1 is the alloc list, otherwise a hash chain. */
 static tdb1_off_t lock_offset(int list)
 {
 /* list -1 is the alloc list, otherwise a hash chain. */
 static tdb1_off_t lock_offset(int list)
 {


@@ -132,102 +45,30 @@ int tdb1_brlock(struct tdb1_context *tdb,
               int rw_type, tdb1_off_t offset, size_t len,
               enum tdb_lock_flags flags)
 {
               int rw_type, tdb1_off_t offset, size_t len,
               enum tdb_lock_flags flags)
 {

-       int ret;
-
-       if (tdb->flags & TDB_NOLOCK) {
+       enum TDB_ERROR ecode = tdb_brlock(tdb, rw_type, offset, len, flags
+                                         | TDB_LOCK_NOCHECK);
+       if (ecode == TDB_SUCCESS)

                return 0;
                return 0;

-       }
-
-       if ((rw_type == F_WRLCK) && (tdb->read_only || tdb->traverse_read)) {
-               tdb->last_error = TDB_ERR_RDONLY;
-               return -1;
-       }
-
-       do {
-               ret = fcntl_lock(tdb, rw_type, offset, len,
-                                flags & TDB_LOCK_WAIT);
-       } while (ret == -1 && errno == EINTR);
-
-       if (ret == -1) {
-               tdb->last_error = TDB_ERR_LOCK;
-               /* Generic lock error. errno set by fcntl.
-                * EAGAIN is an expected return from non-blocking
-                * locks. */
-               if (!(flags & TDB_LOCK_PROBE) && errno != EAGAIN) {
-                       tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                                  "tdb1_brlock failed (fd=%d) at offset %d rw_type=%d flags=%d len=%d",
-                                  tdb->file->fd, offset, rw_type, flags, (int)len);
-               }
-               return -1;
-       }
-       return 0;
+       tdb->last_error = ecode;
+       return -1;

 }
 
 int tdb1_brunlock(struct tdb1_context *tdb,
                 int rw_type, tdb1_off_t offset, size_t len)
 {
 }
 
 int tdb1_brunlock(struct tdb1_context *tdb,
                 int rw_type, tdb1_off_t offset, size_t len)
 {

-       int ret;
-
-       if (tdb->flags & TDB_NOLOCK) {
+       enum TDB_ERROR ecode = tdb_brunlock(tdb, rw_type, offset, len);
+       if (ecode == TDB_SUCCESS)

                return 0;
                return 0;

-       }
-
-       do {
-               ret = fcntl_unlock(tdb, rw_type, offset, len);
-       } while (ret == -1 && errno == EINTR);
-
-       if (ret == -1) {
-               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                                       "tdb1_brunlock failed (fd=%d) at offset"
-                                       " %d rw_type=%d len=%d",
-                                       tdb->file->fd, offset, rw_type, (int)len);
-       }
-       return ret;
+       tdb->last_error = ecode;
+       return -1;

 }
 
 }
 

-/*
-  upgrade a read lock to a write lock. This needs to be handled in a
-  special way as some OSes (such as solaris) have too conservative
-  deadlock detection and claim a deadlock when progress can be
-  made. For those OSes we may loop for a while.
-*/

 int tdb1_allrecord_upgrade(struct tdb1_context *tdb)
 {
 int tdb1_allrecord_upgrade(struct tdb1_context *tdb)
 {

-       int count = 1000;
-
-       if (tdb->file->allrecord_lock.count != 1) {
-               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                                       "tdb1_allrecord_upgrade failed: "
-                                       "count %u too high",
-                                       tdb->file->allrecord_lock.count);
-               return -1;
-       }
-
-       if (tdb->file->allrecord_lock.off != 1) {
-               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                                       "tdb1_allrecord_upgrade failed:"
-                                       " already upgraded?");
-               return -1;
-       }
-
-       while (count--) {
-               struct timeval tv;
-               if (tdb1_brlock(tdb, F_WRLCK, TDB1_FREELIST_TOP, 0,
-                              TDB_LOCK_WAIT|TDB_LOCK_PROBE) == 0) {
-                       tdb->file->allrecord_lock.ltype = F_WRLCK;
-                       tdb->file->allrecord_lock.off = 0;
-                       return 0;
-               }
-               if (errno != EDEADLK) {
-                       break;
-               }
-               /* sleep for as short a time as we can - more portable than usleep() */
-               tv.tv_sec = 0;
-               tv.tv_usec = 1;
-               select(0, NULL, NULL, NULL, &tv);
-       }
-       tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                               "tdb1_allrecord_upgrade failed");
+       enum TDB_ERROR ecode = tdb_allrecord_upgrade(tdb, TDB1_FREELIST_TOP);
+       if (ecode == TDB_SUCCESS)
+               return 0;
+       tdb->last_error = ecode;

        return -1;
 }
 
        return -1;
 }
 


@@ -248,7 +89,7 @@ static struct tdb_lock *tdb1_find_nestlock(struct tdb1_context *tdb,
 int tdb1_nest_lock(struct tdb1_context *tdb, uint32_t offset, int ltype,
                  enum tdb_lock_flags flags)
 {
 int tdb1_nest_lock(struct tdb1_context *tdb, uint32_t offset, int ltype,
                  enum tdb_lock_flags flags)
 {

-       struct tdb_lock *new_lck;
+       enum TDB_ERROR ecode;

 
        if (offset >= lock_offset(tdb->header.hash_size)) {
                tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
 
        if (offset >= lock_offset(tdb->header.hash_size)) {
                tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,


@@ -257,39 +98,12 @@ int tdb1_nest_lock(struct tdb1_context *tdb, uint32_t offset, int ltype,
                                        offset, ltype);
                return -1;
        }
                                        offset, ltype);
                return -1;
        }

-       if (tdb->flags & TDB_NOLOCK)
-               return 0;

 
 

-       new_lck = tdb1_find_nestlock(tdb, offset);
-       if (new_lck) {
-               /*
-                * Just increment the in-memory struct, posix locks
-                * don't stack.
-                */
-               new_lck->count++;
-               return 0;
-       }
-
-       new_lck = (struct tdb_lock *)realloc(
-               tdb->file->lockrecs,
-               sizeof(*tdb->file->lockrecs) * (tdb->file->num_lockrecs+1));
-       if (new_lck == NULL) {
-               errno = ENOMEM;
+       ecode = tdb_nest_lock(tdb, offset, ltype, flags | TDB_LOCK_NOCHECK);
+       if (unlikely(ecode != TDB_SUCCESS)) {
+               tdb->last_error = ecode;

                return -1;
        }
                return -1;
        }

-       tdb->file->lockrecs = new_lck;
-
-       /* Since fcntl locks don't nest, we do a lock for the first one,
-          and simply bump the count for future ones */
-       if (tdb1_brlock(tdb, ltype, offset, 1, flags)) {
-               return -1;
-       }
-
-       tdb->file->lockrecs[tdb->file->num_lockrecs].off = offset;
-       tdb->file->lockrecs[tdb->file->num_lockrecs].count = 1;
-       tdb->file->lockrecs[tdb->file->num_lockrecs].ltype = ltype;
-       tdb->file->num_lockrecs++;
-

        return 0;
 }
 
        return 0;
 }
 


@@ -298,11 +112,13 @@ static int tdb1_lock_and_recover(struct tdb1_context *tdb)
        int ret;
 
        /* We need to match locking order in transaction commit. */
        int ret;
 
        /* We need to match locking order in transaction commit. */

-       if (tdb1_brlock(tdb, F_WRLCK, TDB1_FREELIST_TOP, 0, TDB_LOCK_WAIT)) {
+       if (tdb1_brlock(tdb, F_WRLCK, TDB1_FREELIST_TOP, 0,
+                       TDB_LOCK_WAIT|TDB_LOCK_NOCHECK)) {

                return -1;
        }
 
                return -1;
        }
 

-       if (tdb1_brlock(tdb, F_WRLCK, TDB1_OPEN_LOCK, 1, TDB_LOCK_WAIT)) {
+       if (tdb1_brlock(tdb, F_WRLCK, TDB1_OPEN_LOCK, 1,
+                       TDB_LOCK_WAIT|TDB_LOCK_NOCHECK)) {

                tdb1_brunlock(tdb, F_WRLCK, TDB1_FREELIST_TOP, 0);
                return -1;
        }
                tdb1_brunlock(tdb, F_WRLCK, TDB1_FREELIST_TOP, 0);
                return -1;
        }


@@ -374,57 +190,22 @@ int tdb1_lock(struct tdb1_context *tdb, int list, int ltype)
 
 int tdb1_nest_unlock(struct tdb1_context *tdb, uint32_t offset, int ltype)
 {
 
 int tdb1_nest_unlock(struct tdb1_context *tdb, uint32_t offset, int ltype)
 {

-       int ret = -1;
-       struct tdb_lock *lck;
-
-       if (tdb->flags & TDB_NOLOCK)
-               return 0;
+       enum TDB_ERROR ecode;

 
        /* Sanity checks */
        if (offset >= lock_offset(tdb->header.hash_size)) {
                tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
                                        "tdb1_unlock: offset %u invalid (%d)",
                                        offset, tdb->header.hash_size);
 
        /* Sanity checks */
        if (offset >= lock_offset(tdb->header.hash_size)) {
                tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
                                        "tdb1_unlock: offset %u invalid (%d)",
                                        offset, tdb->header.hash_size);

-               return ret;
-       }
-
-       lck = tdb1_find_nestlock(tdb, offset);
-       if ((lck == NULL) || (lck->count == 0)) {
-               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
-                                       "tdb1_unlock: count is 0");

                return -1;
        }
 
                return -1;
        }
 

-       if (lck->count > 1) {
-               lck->count--;
-               return 0;
-       }
-
-       /*
-        * This lock has count==1 left, so we need to unlock it in the
-        * kernel. We don't bother with decrementing the in-memory array
-        * element, we're about to overwrite it with the last array element
-        * anyway.
-        */
-
-       ret = tdb1_brunlock(tdb, ltype, offset, 1);
-
-       /*
-        * Shrink the array by overwriting the element just unlocked with the
-        * last array element.
-        */
-       *lck = tdb->file->lockrecs[--tdb->file->num_lockrecs];
-
-       /*
-        * We don't bother with realloc when the array shrinks, but if we have
-        * a completely idle tdb we should get rid of the locked array.
-        */
-
-       if (tdb->file->num_lockrecs == 0) {
-               SAFE_FREE(tdb->file->lockrecs);
+       ecode = tdb_nest_unlock(tdb, offset, ltype);
+       if (unlikely(ecode != TDB_SUCCESS)) {
+               tdb->last_error = ecode;
+               return -1;

        }
        }

-
-       return ret;
+       return 0;

 }
 
 int tdb1_unlock(struct tdb1_context *tdb, int list, int ltype)
 }
 
 int tdb1_unlock(struct tdb1_context *tdb, int list, int ltype)


@@ -460,87 +241,46 @@ int tdb1_transaction_unlock(struct tdb1_context *tdb, int ltype)
        return tdb1_nest_unlock(tdb, TDB1_TRANSACTION_LOCK, ltype);
 }
 
        return tdb1_nest_unlock(tdb, TDB1_TRANSACTION_LOCK, ltype);
 }
 

-/* Returns 0 if all done, -1 if error, 1 if ok. */
-static int tdb1_allrecord_check(struct tdb1_context *tdb, int ltype,
-                              enum tdb_lock_flags flags, bool upgradable)
+/* lock/unlock entire database.  It can only be upgradable if you have some
+ * other way of guaranteeing exclusivity (ie. transaction write lock).
+ * We do the locking gradually to avoid being starved by smaller locks. */
+int tdb1_allrecord_lock(struct tdb1_context *tdb, int ltype,
+                      enum tdb_lock_flags flags, bool upgradable)

 {
 {

-       /* There are no locks on read-only dbs */
-       if (tdb->read_only || tdb->traverse_read) {
-               tdb->last_error = TDB_ERR_LOCK;
+       enum TDB_ERROR ecode;
+
+       /* tdb_lock_gradual() doesn't know about tdb->traverse_read. */
+       if (tdb->traverse_read && !(tdb->flags & TDB_NOLOCK)) {
+               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK,
+                                            TDB_LOG_USE_ERROR,
+                                            "tdb1_allrecord_lock during"
+                                            " tdb1_read_traverse");

                return -1;
        }
 
                return -1;
        }
 

-       if (tdb->file->allrecord_lock.count && tdb->file->allrecord_lock.ltype == ltype) {
+       if (tdb->file->allrecord_lock.count
+           && tdb->file->allrecord_lock.ltype == ltype) {

                tdb->file->allrecord_lock.count++;
                return 0;
        }
 
                tdb->file->allrecord_lock.count++;
                return 0;
        }
 

-       if (tdb->file->allrecord_lock.count) {
-               /* a global lock of a different type exists */
-               tdb->last_error = TDB_ERR_LOCK;
-               return -1;
-       }
-

        if (tdb1_have_extra_locks(tdb)) {
                /* can't combine global and chain locks */
        if (tdb1_have_extra_locks(tdb)) {
                /* can't combine global and chain locks */

-               tdb->last_error = TDB_ERR_LOCK;
+               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK,
+                                            TDB_LOG_USE_ERROR,
+                                            "tdb1_allrecord_lock holding"
+                                            " other locks");

                return -1;
        }
 
        if (upgradable && ltype != F_RDLCK) {
                /* tdb error: you can't upgrade a write lock! */
                return -1;
        }
 
        if (upgradable && ltype != F_RDLCK) {
                /* tdb error: you can't upgrade a write lock! */

-               tdb->last_error = TDB_ERR_LOCK;
-               return -1;
-       }
-       return 1;
-}
-
-/* We only need to lock individual bytes, but Linux merges consecutive locks
- * so we lock in contiguous ranges. */
-static int tdb1_chainlock_gradual(struct tdb1_context *tdb,
-                                int ltype, enum tdb_lock_flags flags,
-                                size_t off, size_t len)
-{
-       int ret;
-       enum tdb_lock_flags nb_flags = (flags & ~TDB_LOCK_WAIT);
-
-       if (len <= 4) {
-               /* Single record.  Just do blocking lock. */
-               return tdb1_brlock(tdb, ltype, off, len, flags);
-       }
-
-       /* First we try non-blocking. */
-       ret = tdb1_brlock(tdb, ltype, off, len, nb_flags);
-       if (ret == 0) {
-               return 0;
-       }
-
-       /* Try locking first half, then second. */
-       ret = tdb1_chainlock_gradual(tdb, ltype, flags, off, len / 2);
-       if (ret == -1)
-               return -1;
-
-       ret = tdb1_chainlock_gradual(tdb, ltype, flags,
-                                   off + len / 2, len - len / 2);
-       if (ret == -1) {
-               tdb1_brunlock(tdb, ltype, off, len / 2);
+               tdb->last_error = tdb_logerr(tdb, TDB_ERR_LOCK,
+                                            TDB_LOG_ERROR,
+                                            "tdb1_allrecord_lock cannot"
+                                            " have upgradable write lock");

                return -1;
        }
                return -1;
        }

-       return 0;
-}
-
-/* lock/unlock entire database.  It can only be upgradable if you have some
- * other way of guaranteeing exclusivity (ie. transaction write lock).
- * We do the locking gradually to avoid being starved by smaller locks. */
-int tdb1_allrecord_lock(struct tdb1_context *tdb, int ltype,
-                      enum tdb_lock_flags flags, bool upgradable)
-{
-       switch (tdb1_allrecord_check(tdb, ltype, flags, upgradable)) {
-       case -1:
-               return -1;
-       case 0:
-               return 0;
-       }

 
        /* We cover two kinds of locks:
         * 1) Normal chain locks.  Taken for almost all operations.
 
        /* We cover two kinds of locks:
         * 1) Normal chain locks.  Taken for almost all operations.


@@ -549,8 +289,10 @@ int tdb1_allrecord_lock(struct tdb1_context *tdb, int ltype,
         *
         * It is (1) which cause the starvation problem, so we're only
         * gradual for that. */
         *
         * It is (1) which cause the starvation problem, so we're only
         * gradual for that. */

-       if (tdb1_chainlock_gradual(tdb, ltype, flags, TDB1_FREELIST_TOP,
-                                 tdb->header.hash_size * 4) == -1) {
+       ecode = tdb_lock_gradual(tdb, ltype, flags | TDB_LOCK_NOCHECK,
+                                TDB1_FREELIST_TOP, tdb->header.hash_size * 4);
+       if (ecode != TDB_SUCCESS) {
+               tdb->last_error = ecode;

                return -1;
        }
 
                return -1;
        }
 


@@ -562,6 +304,8 @@ int tdb1_allrecord_lock(struct tdb1_context *tdb, int ltype,
                return -1;
        }
 
                return -1;
        }
 

+       /* FIXME: Temporary cast. */
+       tdb->file->allrecord_lock.owner = (void *)(struct tdb1_context *)tdb;

        tdb->file->allrecord_lock.count = 1;
        /* If it's upgradable, it's actually exclusive so we can treat
         * it as a write lock. */
        tdb->file->allrecord_lock.count = 1;
        /* If it's upgradable, it's actually exclusive so we can treat
         * it as a write lock. */

diff --git a/ccan/tdb2/tdb1_open.c b/ccan/tdb2/tdb1_open.c
index 5c4f62f67c8c3f3d572a8576e049344d5e1206e1..1fe981ac76b2a7f688ad6c4c70f534235f4e3001 100644 (file)
--- a/ccan/tdb2/tdb1_open.c
+++ b/ccan/tdb2/tdb1_open.c
@@ -189,6 +189,8 @@ struct tdb1_context *tdb1_open_ex(const char *name, int hash_size, int tdb1_flag
        tdb->file->map_ptr = NULL;
        tdb->flags = tdb1_flags|TDB_VERSION1;
        tdb->open_flags = open_flags;
        tdb->file->map_ptr = NULL;
        tdb->flags = tdb1_flags|TDB_VERSION1;
        tdb->open_flags = open_flags;

+       tdb->lock_fn = tdb_fcntl_lock;
+       tdb->unlock_fn = tdb_fcntl_unlock;

        if (log_ctx) {
                tdb->log_fn = log_ctx->log_fn;
                tdb->log_data = log_ctx->log_private;
        if (log_ctx) {
                tdb->log_fn = log_ctx->log_fn;
                tdb->log_data = log_ctx->log_private;

diff --git a/ccan/tdb2/tdb1_private.h b/ccan/tdb2/tdb1_private.h
index 5470955a54d9bf10e21a60bf2757b09bd74a1d59..9350076f27d08a76e45be3de46a117c90af678cd 100644 (file)
--- a/ccan/tdb2/tdb1_private.h
+++ b/ccan/tdb2/tdb1_private.h
@@ -35,6 +35,32 @@
 #define tdb_error(tdb) \
        tdb_error((struct tdb_context *)(tdb))
 
 #define tdb_error(tdb) \
        tdb_error((struct tdb_context *)(tdb))
 

+#define tdb_brlock(tdb1, rw_type, offset, len, flags)  \
+       tdb_brlock((struct tdb_context *)(tdb1),        \
+                  (rw_type), (offset), (len), (flags))
+
+#define tdb_brunlock(tdb1, rw_type, offset, len) \
+       tdb_brunlock((struct tdb_context *)(tdb1), (rw_type), (offset), (len))
+
+#define tdb_nest_lock(tdb1, offset, ltype, flags)      \
+       tdb_nest_lock((struct tdb_context *)(tdb1), (offset), (ltype), (flags))
+
+#define tdb_nest_unlock(tdb1, offset, ltype)   \
+       tdb_nest_unlock((struct tdb_context *)(tdb1), (offset), (ltype))
+
+#define tdb_allrecord_lock(tdb1, offset, flags, upgradable)            \
+       tdb_allrecord_lock((struct tdb_context *)(tdb1),                \
+                          (offset), (flags), (upgradable))
+
+#define tdb_allrecord_unlock(tdb1, ltype)                              \
+       tdb_allrecord_unlock((struct tdb_context *)(tdb1), (ltype))
+
+#define tdb_allrecord_upgrade(tdb1, start)                             \
+       tdb_allrecord_upgrade((struct tdb_context *)(tdb1), (start))
+
+#define tdb_lock_gradual(tdb1, ltype, flags, off, len) \
+       tdb_lock_gradual((struct tdb_context *)(tdb1),  \
+                        (ltype), (flags), (off), (len))

 /***** END FIXME ***/
 
 #include <limits.h>
 /***** END FIXME ***/
 
 #include <limits.h>


@@ -171,14 +197,24 @@ struct tdb1_context {
        /* The actual file information */
        struct tdb_file *file;
 
        /* The actual file information */
        struct tdb_file *file;
 

-       int read_only; /* opened read-only */
+       int open_flags; /* flags used in the open - needed by reopen */
+
+       /* low level (fnctl) lock functions. */
+       int (*lock_fn)(int fd, int rw, off_t off, off_t len, bool w, void *);
+       int (*unlock_fn)(int fd, int rw, off_t off, off_t len, void *);
+       void *lock_data;
+
+       uint32_t flags; /* the flags passed to tdb1_open */
+
+       /* Our statistics. */
+       struct tdb_attribute_stats stats;
+
+       bool read_only; /* opened read-only */

        int traverse_read; /* read-only traversal */
        int traverse_write; /* read-write traversal */
        struct tdb1_header header; /* a cached copy of the header */
        int traverse_read; /* read-only traversal */
        int traverse_write; /* read-write traversal */
        struct tdb1_header header; /* a cached copy of the header */

-       uint32_t flags; /* the flags passed to tdb1_open */

        struct tdb1_traverse_lock travlocks; /* current traversal locks */
        unsigned int (*hash_fn)(TDB_DATA *key);
        struct tdb1_traverse_lock travlocks; /* current traversal locks */
        unsigned int (*hash_fn)(TDB_DATA *key);

-       int open_flags; /* flags used in the open - needed by reopen */

        const struct tdb1_methods *methods;
        struct tdb1_transaction *transaction;
        int page_size;
        const struct tdb1_methods *methods;
        struct tdb1_transaction *transaction;
        int page_size;

diff --git a/ccan/tdb2/transaction.c b/ccan/tdb2/transaction.c
index a29acf96e2d417c8c3d79313f8ada959f7c5719c..2afc1789b0e1fa370d8f28453b33e734d6c035b3 100644 (file)
--- a/ccan/tdb2/transaction.c
+++ b/ccan/tdb2/transaction.c
@@ -1006,7 +1006,7 @@ static enum TDB_ERROR _tdb_transaction_prepare_commit(struct tdb_context *tdb)
        methods = tdb->transaction->io_methods;
 
        /* upgrade the main transaction lock region to a write lock */
        methods = tdb->transaction->io_methods;
 
        /* upgrade the main transaction lock region to a write lock */

-       ecode = tdb_allrecord_upgrade(tdb);
+       ecode = tdb_allrecord_upgrade(tdb, TDB_HASH_LOCK_START);

        if (ecode != TDB_SUCCESS) {
                return ecode;
        }
        if (ecode != TDB_SUCCESS) {
                return ecode;
        }