summaryrefslogtreecommitdiff
path: root/source4/lib/tdb/common/freelist.c
diff options
context:
space:
mode:
authorAndrew Tridgell <tridge@samba.org>2005-09-16 03:52:42 +0000
committerGerald (Jerry) Carter <jerry@samba.org>2007-10-10 13:38:12 -0500
commit0868b7c77d42efd5f361f605bfc0d8d46841db95 (patch)
treee73b6ca6bb895ff3b9af1006ffd566e7f2ee7ea8 /source4/lib/tdb/common/freelist.c
parent95040e934175eb877ce6d83690fd06ce5d2b028c (diff)
downloadsamba-0868b7c77d42efd5f361f605bfc0d8d46841db95.tar.gz
samba-0868b7c77d42efd5f361f605bfc0d8d46841db95.tar.bz2
samba-0868b7c77d42efd5f361f605bfc0d8d46841db95.zip
r10253: a fairly large tdb cleanup and re-organise. Nearly all of this change
just involves splitting up the core tdb.c code into separate files on logical boundaries, but there are some minor functional changes as well: - move the 'struct tdb_context' into tdb_private.h, hiding it from users. This was done to allow the structure to change without breaking code that uses tdb. - added accessor functions tdb_fd(), tdb_name(), and tdb_log_fn() to access the elements of struct tdb_context that were used by external code but are no longer visible - simplied tdb_append() to use tdb_fetch()/tdb_store(), which is just as good due to the way tdb locks work - changed some of the types (such as tdb_off to tdb_off_t) to make syntax highlighting work better - removed the old optional spinlock code. It was a bad idea. - fixed a bug in tdb_reopen_all() that caused tdbtorture to sometimes fail or report nasty looking errors. This is the only real bug fixed in this commit. Jeremy/Jerry, you might like to pickup this change for Samba3, as that could definately affect smbd in Samba3. The aim of all of these changes is to make the tdb transactions/journaling code I am working on easier to write. I started to write it on top of the existing tdb.c code and it got very messy. Splitting up the code makes it much easier to follow. There are more cleanups we could do in tdb, such as using uint32_t instead of u32 (suggested by metze). I'll leave those for another day. (This used to be commit 4673cdd0d261614e707b72a7a348bb0e7dbb2482)
Diffstat (limited to 'source4/lib/tdb/common/freelist.c')
-rw-r--r--source4/lib/tdb/common/freelist.c323
1 files changed, 323 insertions, 0 deletions
diff --git a/source4/lib/tdb/common/freelist.c b/source4/lib/tdb/common/freelist.c
new file mode 100644
index 0000000000..018b7bc298
--- /dev/null
+++ b/source4/lib/tdb/common/freelist.c
@@ -0,0 +1,323 @@
+ /*
+ Unix SMB/CIFS implementation.
+
+ trivial database library
+
+ Copyright (C) Andrew Tridgell 1999-2005
+ Copyright (C) Paul `Rusty' Russell 2000
+ Copyright (C) Jeremy Allison 2000-2003
+
+ ** NOTE! The following LGPL license applies to the tdb
+ ** library. This does NOT imply that all of Samba is released
+ ** under the LGPL
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+*/
+
+#include "tdb_private.h"
+
+/* read a freelist record and check for simple errors */
+static int rec_free_read(struct tdb_context *tdb, tdb_off_t off, struct list_struct *rec)
+{
+ if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
+ return -1;
+
+ if (rec->magic == TDB_MAGIC) {
+ /* this happens when a app is showdown while deleting a record - we should
+ not completely fail when this happens */
+ TDB_LOG((tdb, 0,"rec_free_read non-free magic 0x%x at offset=%d - fixing\n",
+ rec->magic, off));
+ rec->magic = TDB_FREE_MAGIC;
+ if (tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
+ return -1;
+ }
+
+ if (rec->magic != TDB_FREE_MAGIC) {
+ /* Ensure ecode is set for log fn. */
+ tdb->ecode = TDB_ERR_CORRUPT;
+ TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n",
+ rec->magic, off));
+ return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
+ }
+ if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
+ return -1;
+ return 0;
+}
+
+
+
+/* Remove an element from the freelist. Must have alloc lock. */
+static int remove_from_freelist(struct tdb_context *tdb, tdb_off_t off, tdb_off_t next)
+{
+ tdb_off_t last_ptr, i;
+
+ /* read in the freelist top */
+ last_ptr = FREELIST_TOP;
+ while (tdb_ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
+ if (i == off) {
+ /* We've found it! */
+ return tdb_ofs_write(tdb, last_ptr, &next);
+ }
+ /* Follow chain (next offset is at start of record) */
+ last_ptr = i;
+ }
+ TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
+ return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
+}
+
+
+/* update a record tailer (must hold allocation lock) */
+static int update_tailer(struct tdb_context *tdb, tdb_off_t offset,
+ const struct list_struct *rec)
+{
+ tdb_off_t totalsize;
+
+ /* Offset of tailer from record header */
+ totalsize = sizeof(*rec) + rec->rec_len;
+ return tdb_ofs_write(tdb, offset + totalsize - sizeof(tdb_off_t),
+ &totalsize);
+}
+
+/* Add an element into the freelist. Merge adjacent records if
+ neccessary. */
+int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
+{
+ tdb_off_t right, left;
+
+ /* Allocation and tailer lock */
+ if (tdb_lock(tdb, -1, F_WRLCK) != 0)
+ return -1;
+
+ /* set an initial tailer, so if we fail we don't leave a bogus record */
+ if (update_tailer(tdb, offset, rec) != 0) {
+ TDB_LOG((tdb, 0, "tdb_free: upfate_tailer failed!\n"));
+ goto fail;
+ }
+
+ /* Look right first (I'm an Australian, dammit) */
+ right = offset + sizeof(*rec) + rec->rec_len;
+ if (right + sizeof(*rec) <= tdb->map_size) {
+ struct list_struct r;
+
+ if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
+ goto left;
+ }
+
+ /* If it's free, expand to include it. */
+ if (r.magic == TDB_FREE_MAGIC) {
+ if (remove_from_freelist(tdb, right, r.next) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
+ goto left;
+ }
+ rec->rec_len += sizeof(r) + r.rec_len;
+ }
+ }
+
+left:
+ /* Look left */
+ left = offset - sizeof(tdb_off_t);
+ if (left > TDB_DATA_START(tdb->header.hash_size)) {
+ struct list_struct l;
+ tdb_off_t leftsize;
+
+ /* Read in tailer and jump back to header */
+ if (tdb_ofs_read(tdb, left, &leftsize) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
+ goto update;
+ }
+ left = offset - leftsize;
+
+ /* Now read in record */
+ if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
+ goto update;
+ }
+
+ /* If it's free, expand to include it. */
+ if (l.magic == TDB_FREE_MAGIC) {
+ if (remove_from_freelist(tdb, left, l.next) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
+ goto update;
+ } else {
+ offset = left;
+ rec->rec_len += leftsize;
+ }
+ }
+ }
+
+update:
+ if (update_tailer(tdb, offset, rec) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
+ goto fail;
+ }
+
+ /* Now, prepend to free list */
+ rec->magic = TDB_FREE_MAGIC;
+
+ if (tdb_ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
+ tdb_rec_write(tdb, offset, rec) == -1 ||
+ tdb_ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
+ TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
+ goto fail;
+ }
+
+ /* And we're done. */
+ tdb_unlock(tdb, -1, F_WRLCK);
+ return 0;
+
+ fail:
+ tdb_unlock(tdb, -1, F_WRLCK);
+ return -1;
+}
+
+
+/*
+ the core of tdb_allocate - called when we have decided which
+ free list entry to use
+ */
+static tdb_off_t tdb_allocate_ofs(struct tdb_context *tdb, tdb_len_t length, tdb_off_t rec_ptr,
+ struct list_struct *rec, tdb_off_t last_ptr)
+{
+ struct list_struct newrec;
+ tdb_off_t newrec_ptr;
+
+ memset(&newrec, '\0', sizeof(newrec));
+
+ /* found it - now possibly split it up */
+ if (rec->rec_len > length + MIN_REC_SIZE) {
+ /* Length of left piece */
+ length = TDB_ALIGN(length, TDB_ALIGNMENT);
+
+ /* Right piece to go on free list */
+ newrec.rec_len = rec->rec_len - (sizeof(*rec) + length);
+ newrec_ptr = rec_ptr + sizeof(*rec) + length;
+
+ /* And left record is shortened */
+ rec->rec_len = length;
+ } else {
+ newrec_ptr = 0;
+ }
+
+ /* Remove allocated record from the free list */
+ if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1) {
+ return 0;
+ }
+
+ /* Update header: do this before we drop alloc
+ lock, otherwise tdb_free() might try to
+ merge with us, thinking we're free.
+ (Thanks Jeremy Allison). */
+ rec->magic = TDB_MAGIC;
+ if (tdb_rec_write(tdb, rec_ptr, rec) == -1) {
+ return 0;
+ }
+
+ /* Did we create new block? */
+ if (newrec_ptr) {
+ /* Update allocated record tailer (we
+ shortened it). */
+ if (update_tailer(tdb, rec_ptr, rec) == -1) {
+ return 0;
+ }
+
+ /* Free new record */
+ if (tdb_free(tdb, newrec_ptr, &newrec) == -1) {
+ return 0;
+ }
+ }
+
+ /* all done - return the new record offset */
+ return rec_ptr;
+}
+
+/* allocate some space from the free list. The offset returned points
+ to a unconnected list_struct within the database with room for at
+ least length bytes of total data
+
+ 0 is returned if the space could not be allocated
+ */
+tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct list_struct *rec)
+{
+ tdb_off_t rec_ptr, last_ptr, newrec_ptr;
+ struct {
+ tdb_off_t rec_ptr, last_ptr;
+ tdb_len_t rec_len;
+ } bestfit;
+
+ if (tdb_lock(tdb, -1, F_WRLCK) == -1)
+ return 0;
+
+ /* Extra bytes required for tailer */
+ length += sizeof(tdb_off_t);
+
+ again:
+ last_ptr = FREELIST_TOP;
+
+ /* read in the freelist top */
+ if (tdb_ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1)
+ goto fail;
+
+ bestfit.rec_ptr = 0;
+
+ /*
+ this is a best fit allocation strategy. Originally we used
+ a first fit strategy, but it suffered from massive fragmentation
+ issues when faced with a slowly increasing record size.
+ */
+ while (rec_ptr) {
+ if (rec_free_read(tdb, rec_ptr, rec) == -1) {
+ goto fail;
+ }
+
+ if (rec->rec_len >= length) {
+ if (bestfit.rec_ptr == 0 ||
+ rec->rec_len < bestfit.rec_len) {
+ bestfit.rec_len = rec->rec_len;
+ bestfit.rec_ptr = rec_ptr;
+ bestfit.last_ptr = last_ptr;
+ /* consider a fit to be good enough if
+ we aren't wasting more than half
+ the space */
+ if (bestfit.rec_len < 2*length) {
+ break;
+ }
+ }
+ }
+
+ /* move to the next record */
+ last_ptr = rec_ptr;
+ rec_ptr = rec->next;
+ }
+
+ if (bestfit.rec_ptr != 0) {
+ if (rec_free_read(tdb, bestfit.rec_ptr, rec) == -1) {
+ goto fail;
+ }
+
+ newrec_ptr = tdb_allocate_ofs(tdb, length, bestfit.rec_ptr, rec, bestfit.last_ptr);
+ tdb_unlock(tdb, -1, F_WRLCK);
+ return newrec_ptr;
+ }
+
+ /* we didn't find enough space. See if we can expand the
+ database and if we can then try again */
+ if (tdb_expand(tdb, length + sizeof(*rec)) == 0)
+ goto again;
+ fail:
+ tdb_unlock(tdb, -1, F_WRLCK);
+ return 0;
+}
+