X-Git-Url: http://git.ozlabs.org/?p=ccan;a=blobdiff_plain;f=ccan%2Ftdb%2Flock.c;h=2a6f1230c9d7308b83592771ec03de5d7f5a95f1;hp=a59ba365213e473630617ee69ff1f238585d6782;hb=729fc5b931c8eb26f192399d8aa7064f9af058df;hpb=88541fe7eb0542150669312675ddc00e9684de66 diff --git a/ccan/tdb/lock.c b/ccan/tdb/lock.c index a59ba365..2a6f1230 100644 --- a/ccan/tdb/lock.c +++ b/ccan/tdb/lock.c @@ -119,6 +119,12 @@ static int fcntl_unlock(struct tdb_context *tdb, int rw, off_t off, off_t len) return fcntl(tdb->fd, F_SETLKW, &fl); } +/* list -1 is the alloc list, otherwise a hash chain. */ +static tdb_off_t lock_offset(int list) +{ + return FREELIST_TOP + 4*list; +} + /* a byte range locking function - return 0 on success this functions locks/unlocks 1 byte at the specified offset. @@ -146,6 +152,14 @@ int tdb_brlock(struct tdb_context *tdb, return -1; } + /* Sanity check */ + if (tdb->transaction && offset >= lock_offset(-1) && len != 0) { + tdb->ecode = TDB_ERR_RDONLY; + TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_brlock attempted in transaction at offset %d rw_type=%d flags=%d len=%d\n", + offset, rw_type, flags, (int)len)); + return -1; + } + do { ret = fcntl_lock(tdb, rw_type, offset, len, flags & TDB_LOCK_WAIT); @@ -234,12 +248,6 @@ int tdb_allrecord_upgrade(struct tdb_context *tdb) return -1; } -/* list -1 is the alloc list, otherwise a hash chain. */ -static tdb_off_t lock_offset(int list) -{ - return FREELIST_TOP + 4*list; -} - static struct tdb_lock_type *find_nestlock(struct tdb_context *tdb, tdb_off_t offset) { @@ -289,7 +297,7 @@ int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype, /* Since fcntl locks don't nest, we do a lock for the first one, and simply bump the count for future ones */ - if (tdb->methods->brlock(tdb, ltype, offset, 1, flags)) { + if (tdb_brlock(tdb, ltype, offset, 1, flags)) { return -1; } @@ -301,10 +309,44 @@ int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype, return 0; } -/* lock a list in the database. list -1 is the alloc list */ -int tdb_lock(struct tdb_context *tdb, int list, int ltype) +static int tdb_lock_and_recover(struct tdb_context *tdb) +{ + int ret; + + /* We need to match locking order in transaction commit. */ + if (tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0, TDB_LOCK_WAIT)) { + return -1; + } + + if (tdb_brlock(tdb, F_WRLCK, OPEN_LOCK, 1, TDB_LOCK_WAIT)) { + tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0); + return -1; + } + + ret = tdb_transaction_recover(tdb); + + tdb_brunlock(tdb, F_WRLCK, OPEN_LOCK, 1); + tdb_brunlock(tdb, F_WRLCK, FREELIST_TOP, 0); + + return ret; +} + +static bool have_data_locks(const struct tdb_context *tdb) +{ + unsigned int i; + + for (i = 0; i < tdb->num_lockrecs; i++) { + if (tdb->lockrecs[i].off >= lock_offset(-1)) + return true; + } + return false; +} + +static int tdb_lock_list(struct tdb_context *tdb, int list, int ltype, + enum tdb_lock_flags waitflag) { int ret; + bool check = false; /* a allrecord lock allows us to avoid per chain locks */ if (tdb->allrecord_lock.count && @@ -316,9 +358,28 @@ int tdb_lock(struct tdb_context *tdb, int list, int ltype) tdb->ecode = TDB_ERR_LOCK; ret = -1; } else { - ret = tdb_nest_lock(tdb, lock_offset(list), ltype, - TDB_LOCK_WAIT); + /* Only check when we grab first data lock. */ + check = !have_data_locks(tdb); + ret = tdb_nest_lock(tdb, lock_offset(list), ltype, waitflag); + + if (ret == 0 && check && tdb_needs_recovery(tdb)) { + tdb_nest_unlock(tdb, lock_offset(list), ltype, false); + + if (tdb_lock_and_recover(tdb) == -1) { + return -1; + } + return tdb_lock_list(tdb, list, ltype, waitflag); + } } + return ret; +} + +/* lock a list in the database. list -1 is the alloc list */ +int tdb_lock(struct tdb_context *tdb, int list, int ltype) +{ + int ret; + + ret = tdb_lock_list(tdb, list, ltype, TDB_LOCK_WAIT); if (ret) { TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lock failed on list %d " "ltype=%d (%s)\n", list, ltype, strerror(errno))); @@ -329,18 +390,7 @@ int tdb_lock(struct tdb_context *tdb, int list, int ltype) /* lock a list in the database. list -1 is the alloc list. non-blocking lock */ int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype) { - /* a allrecord lock allows us to avoid per chain locks */ - if (tdb->allrecord_lock.count && - (ltype == tdb->allrecord_lock.ltype || ltype == F_RDLCK)) { - return 0; - } - - if (tdb->allrecord_lock.count) { - tdb->ecode = TDB_ERR_LOCK; - return -1; - } - - return tdb_nest_lock(tdb, lock_offset(list), ltype, TDB_LOCK_NOWAIT); + return tdb_lock_list(tdb, list, ltype, TDB_LOCK_NOWAIT); } @@ -380,7 +430,7 @@ int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype, if (mark_lock) { ret = 0; } else { - ret = tdb->methods->brunlock(tdb, ltype, offset, 1); + ret = tdb_brunlock(tdb, ltype, offset, 1); } /* @@ -470,7 +520,7 @@ int tdb_allrecord_lock(struct tdb_context *tdb, int ltype, return -1; } - if (tdb->methods->brlock(tdb, ltype, FREELIST_TOP, 0, flags)) { + if (tdb_brlock(tdb, ltype, FREELIST_TOP, 0, flags)) { if (flags & TDB_LOCK_WAIT) { TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lockall failed (%s)\n", strerror(errno))); } @@ -483,6 +533,21 @@ int tdb_allrecord_lock(struct tdb_context *tdb, int ltype, tdb->allrecord_lock.ltype = upgradable ? F_WRLCK : ltype; tdb->allrecord_lock.off = upgradable; + if (tdb_needs_recovery(tdb)) { + bool mark = flags & TDB_LOCK_MARK_ONLY; + tdb_allrecord_unlock(tdb, ltype, mark); + if (mark) { + tdb->ecode = TDB_ERR_LOCK; + TDB_LOG((tdb, TDB_DEBUG_ERROR, + "tdb_lockall_mark cannot do recovery\n")); + return -1; + } + if (tdb_lock_and_recover(tdb) == -1) { + return -1; + } + return tdb_allrecord_lock(tdb, ltype, flags, upgradable); + } + return 0; } @@ -514,8 +579,7 @@ int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype, bool mark_lock) return 0; } - if (!mark_lock && - tdb->methods->brunlock(tdb, ltype, FREELIST_TOP, 0)) { + if (!mark_lock && tdb_brunlock(tdb, ltype, FREELIST_TOP, 0)) { TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno))); return -1; } @@ -648,7 +712,7 @@ int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off) if (tdb->allrecord_lock.count) { return 0; } - return off ? tdb->methods->brlock(tdb, F_RDLCK, off, 1, TDB_LOCK_WAIT) : 0; + return off ? tdb_brlock(tdb, F_RDLCK, off, 1, TDB_LOCK_WAIT) : 0; } /* @@ -662,12 +726,21 @@ int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off) for (i = &tdb->travlocks; i; i = i->next) if (i->off == off) return -1; - return tdb->methods->brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE); + if (tdb->allrecord_lock.count) { + if (tdb->allrecord_lock.ltype == F_WRLCK) { + return 0; + } + return -1; + } + return tdb_brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE); } int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off) { - return tdb->methods->brunlock(tdb, F_WRLCK, off, 1); + if (tdb->allrecord_lock.count) { + return 0; + } + return tdb_brunlock(tdb, F_WRLCK, off, 1); } /* fcntl locks don't stack: avoid unlocking someone else's */ @@ -685,7 +758,7 @@ int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off) for (i = &tdb->travlocks; i; i = i->next) if (i->off == off) count++; - return (count == 1 ? tdb->methods->brunlock(tdb, F_RDLCK, off, 1) : 0); + return (count == 1 ? tdb_brunlock(tdb, F_RDLCK, off, 1) : 0); } bool tdb_have_extra_locks(struct tdb_context *tdb)