reftable/system: provide thin wrapper for lockfile subsystem

We use the lockfile subsystem to write lockfiles for "tables.list". As
with the tempfile subsystem, the lockfile subsystem also hooks into our
infrastructure to prune stale locks via atexit(3p) or signal handlers.

Furthermore, the lockfile subsystem also handles locking timeouts, which
do add quite a bit of logic. Having to reimplement that in the context
of Git wouldn't make a whole lot of sense, and it is quite likely that
downstream users of the reftable library may have a better idea for how
exactly to implement timeouts.

So again, provide a thin wrapper for the lockfile subsystem instead such
that the compatibility shim is fully self-contained.

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

Author: pks-t

reftable/stack.c

@@ -657,7 +657,7 @@ static int format_name(struct reftable_buf *dest, uint64_t min, uint64_t max)
 }
 
 struct reftable_addition {
-	struct lock_file tables_list_lock;
+	struct reftable_flock tables_list_lock;
 	struct reftable_stack *stack;
 
 	char **new_tables;
@@ -676,10 +676,8 @@ static int reftable_stack_init_addition(struct reftable_addition *add,
 
 	add->stack = st;
 
-	err = hold_lock_file_for_update_timeout(&add->tables_list_lock,
-						st->list_file,
-						LOCK_NO_DEREF,
-						st->opts.lock_timeout_ms);
+	err = flock_acquire(&add->tables_list_lock, st->list_file,
+			    st->opts.lock_timeout_ms);
 	if (err < 0) {
 		if (errno == EEXIST) {
 			err = REFTABLE_LOCK_ERROR;
@@ -689,7 +687,7 @@ static int reftable_stack_init_addition(struct reftable_addition *add,
 		goto done;
 	}
 	if (st->opts.default_permissions) {
-		if (chmod(get_lock_file_path(&add->tables_list_lock),
+		if (chmod(add->tables_list_lock.path,
 			  st->opts.default_permissions) < 0) {
 			err = REFTABLE_IO_ERROR;
 			goto done;
@@ -733,7 +731,7 @@ static void reftable_addition_close(struct reftable_addition *add)
 	add->new_tables_len = 0;
 	add->new_tables_cap = 0;
 
-	rollback_lock_file(&add->tables_list_lock);
+	flock_release(&add->tables_list_lock);
 	reftable_buf_release(&nm);
 }
 
@@ -749,7 +747,6 @@ void reftable_addition_destroy(struct reftable_addition *add)
 int reftable_addition_commit(struct reftable_addition *add)
 {
 	struct reftable_buf table_list = REFTABLE_BUF_INIT;
-	int lock_file_fd = get_lock_file_fd(&add->tables_list_lock);
 	int err = 0;
 	size_t i;
 
@@ -767,20 +764,20 @@ int reftable_addition_commit(struct reftable_addition *add)
 			goto done;
 	}
 
-	err = write_in_full(lock_file_fd, table_list.buf, table_list.len);
+	err = write_in_full(add->tables_list_lock.fd, table_list.buf, table_list.len);
 	reftable_buf_release(&table_list);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		goto done;
 	}
 
-	err = stack_fsync(&add->stack->opts, lock_file_fd);
+	err = stack_fsync(&add->stack->opts, add->tables_list_lock.fd);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		goto done;
 	}
 
-	err = commit_lock_file(&add->tables_list_lock);
+	err = flock_commit(&add->tables_list_lock);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		goto done;
@@ -1160,8 +1157,8 @@ static int stack_compact_range(struct reftable_stack *st,
 	struct reftable_buf new_table_name = REFTABLE_BUF_INIT;
 	struct reftable_buf new_table_path = REFTABLE_BUF_INIT;
 	struct reftable_buf table_name = REFTABLE_BUF_INIT;
-	struct lock_file tables_list_lock = LOCK_INIT;
-	struct lock_file *table_locks = NULL;
+	struct reftable_flock tables_list_lock = REFTABLE_FLOCK_INIT;
+	struct reftable_flock *table_locks = NULL;
 	struct reftable_tmpfile new_table = REFTABLE_TMPFILE_INIT;
 	int is_empty_table = 0, err = 0;
 	size_t first_to_replace, last_to_replace;
@@ -1179,10 +1176,7 @@ static int stack_compact_range(struct reftable_stack *st,
 	 * Hold the lock so that we can read "tables.list" and lock all tables
 	 * which are part of the user-specified range.
 	 */
-	err = hold_lock_file_for_update_timeout(&tables_list_lock,
-						st->list_file,
-						LOCK_NO_DEREF,
-						st->opts.lock_timeout_ms);
+	err = flock_acquire(&tables_list_lock, st->list_file, st->opts.lock_timeout_ms);
 	if (err < 0) {
 		if (errno == EEXIST)
 			err = REFTABLE_LOCK_ERROR;
@@ -1205,19 +1199,20 @@ static int stack_compact_range(struct reftable_stack *st,
 	 * older process is still busy compacting tables which are preexisting
 	 * from the point of view of the newer process.
 	 */
-	REFTABLE_CALLOC_ARRAY(table_locks, last - first + 1);
+	REFTABLE_ALLOC_ARRAY(table_locks, last - first + 1);
 	if (!table_locks) {
 		err = REFTABLE_OUT_OF_MEMORY_ERROR;
 		goto done;
 	}
+	for (i = 0; i < last - first + 1; i++)
+		table_locks[i] = REFTABLE_FLOCK_INIT;
 
 	for (i = last + 1; i > first; i--) {
 		err = stack_filename(&table_name, st, reader_name(st->readers[i - 1]));
 		if (err < 0)
 			goto done;
 
-		err = hold_lock_file_for_update(&table_locks[nlocks],
-						table_name.buf, LOCK_NO_DEREF);
+		err = flock_acquire(&table_locks[nlocks], table_name.buf, 0);
 		if (err < 0) {
 			/*
 			 * When the table is locked already we may do a
@@ -1253,7 +1248,7 @@ static int stack_compact_range(struct reftable_stack *st,
 		 * run into file descriptor exhaustion when we compress a lot
 		 * of tables.
 		 */
-		err = close_lock_file_gently(&table_locks[nlocks++]);
+		err = flock_close(&table_locks[nlocks++]);
 		if (err < 0) {
 			err = REFTABLE_IO_ERROR;
 			goto done;
@@ -1265,7 +1260,7 @@ static int stack_compact_range(struct reftable_stack *st,
 	 * "tables.list" lock while compacting the locked tables. This allows
 	 * concurrent updates to the stack to proceed.
 	 */
-	err = rollback_lock_file(&tables_list_lock);
+	err = flock_release(&tables_list_lock);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		goto done;
@@ -1288,10 +1283,7 @@ static int stack_compact_range(struct reftable_stack *st,
 	 * "tables.list". We'll then replace the compacted range of tables with
 	 * the new table.
 	 */
-	err = hold_lock_file_for_update_timeout(&tables_list_lock,
-						st->list_file,
-						LOCK_NO_DEREF,
-						st->opts.lock_timeout_ms);
+	err = flock_acquire(&tables_list_lock, st->list_file, st->opts.lock_timeout_ms);
 	if (err < 0) {
 		if (errno == EEXIST)
 			err = REFTABLE_LOCK_ERROR;
@@ -1301,7 +1293,7 @@ static int stack_compact_range(struct reftable_stack *st,
 	}
 
 	if (st->opts.default_permissions) {
-		if (chmod(get_lock_file_path(&tables_list_lock),
+		if (chmod(tables_list_lock.path,
 			  st->opts.default_permissions) < 0) {
 			err = REFTABLE_IO_ERROR;
 			goto done;
@@ -1456,22 +1448,22 @@ static int stack_compact_range(struct reftable_stack *st,
 			goto done;
 	}
 
-	err = write_in_full(get_lock_file_fd(&tables_list_lock),
+	err = write_in_full(tables_list_lock.fd,
 			    tables_list_buf.buf, tables_list_buf.len);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		unlink(new_table_path.buf);
 		goto done;
 	}
 
-	err = stack_fsync(&st->opts, get_lock_file_fd(&tables_list_lock));
+	err = stack_fsync(&st->opts, tables_list_lock.fd);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		unlink(new_table_path.buf);
 		goto done;
 	}
 
-	err = commit_lock_file(&tables_list_lock);
+	err = flock_commit(&tables_list_lock);
 	if (err < 0) {
 		err = REFTABLE_IO_ERROR;
 		unlink(new_table_path.buf);
@@ -1492,22 +1484,21 @@ static int stack_compact_range(struct reftable_stack *st,
 	 * readers, so it is expected that unlinking tables may fail.
 	 */
 	for (i = 0; i < nlocks; i++) {
-		struct lock_file *table_lock = &table_locks[i];
-		const char *lock_path = get_lock_file_path(table_lock);
+		struct reftable_flock *table_lock = &table_locks[i];
 
 		reftable_buf_reset(&table_name);
-		err = reftable_buf_add(&table_name, lock_path,
-				       strlen(lock_path) - strlen(".lock"));
+		err = reftable_buf_add(&table_name, table_lock->path,
+				       strlen(table_lock->path) - strlen(".lock"));
 		if (err)
 			continue;
 
 		unlink(table_name.buf);
 	}
 
 done:
-	rollback_lock_file(&tables_list_lock);
+	flock_release(&tables_list_lock);
 	for (i = 0; table_locks && i < nlocks; i++)
-		rollback_lock_file(&table_locks[i]);
+		flock_release(&table_locks[i]);
 	reftable_free(table_locks);
 
 	tmpfile_delete(&new_table);

reftable/system.c

@@ -1,6 +1,7 @@
 #include "system.h"
 #include "basics.h"
 #include "reftable-error.h"
+#include "../lockfile.h"
 #include "../tempfile.h"
 
 int tmpfile_from_pattern(struct reftable_tmpfile *out, const char *pattern)
@@ -47,3 +48,79 @@ int tmpfile_rename(struct reftable_tmpfile *t, const char *path)
 		return REFTABLE_IO_ERROR;
 	return 0;
 }
+
+int flock_acquire(struct reftable_flock *l, const char *target_path,
+		  long timeout_ms)
+{
+	struct lock_file *lockfile;
+	int err;
+
+	lockfile = reftable_malloc(sizeof(*lockfile));
+	if (!lockfile)
+		return REFTABLE_OUT_OF_MEMORY_ERROR;
+
+	err = hold_lock_file_for_update_timeout(lockfile, target_path, LOCK_NO_DEREF,
+						timeout_ms);
+	if (err < 0) {
+		reftable_free(lockfile);
+		if (errno == EEXIST)
+			return REFTABLE_LOCK_ERROR;
+		return -1;
+	}
+
+	l->fd = get_lock_file_fd(lockfile);
+	l->path = get_lock_file_path(lockfile);
+	l->priv = lockfile;
+
+	return 0;
+}
+
+int flock_close(struct reftable_flock *l)
+{
+	struct lock_file *lockfile = l->priv;
+	int ret;
+
+	if (!lockfile)
+		return REFTABLE_API_ERROR;
+
+	ret = close_lock_file_gently(lockfile);
+	l->fd = -1;
+	if (ret < 0)
+		return REFTABLE_IO_ERROR;
+
+	return 0;
+}
+
+int flock_release(struct reftable_flock *l)
+{
+	struct lock_file *lockfile = l->priv;
+	int ret;
+
+	if (!lockfile)
+		return 0;
+
+	ret = rollback_lock_file(lockfile);
+	reftable_free(lockfile);
+	*l = REFTABLE_FLOCK_INIT;
+	if (ret < 0)
+		return REFTABLE_IO_ERROR;
+
+	return 0;
+}
+
+int flock_commit(struct reftable_flock *l)
+{
+	struct lock_file *lockfile = l->priv;
+	int ret;
+
+	if (!lockfile)
+		return REFTABLE_API_ERROR;
+
+	ret = commit_lock_file(lockfile);
+	reftable_free(lockfile);
+	*l = REFTABLE_FLOCK_INIT;
+	if (ret < 0)
+		return REFTABLE_IO_ERROR;
+
+	return 0;
+}

reftable/system.h

@@ -12,7 +12,6 @@ license that can be found in the LICENSE file or at
 /* This header glues the reftable library to the rest of Git */
 
 #include "git-compat-util.h"
-#include "lockfile.h"
 
 /*
  * An implementation-specific temporary file. By making this specific to the
@@ -55,4 +54,48 @@ int tmpfile_delete(struct reftable_tmpfile *t);
  */
 int tmpfile_rename(struct reftable_tmpfile *t, const char *path);
 
+/*
+ * An implementation-specific file lock. Same as with `reftable_tmpfile`,
+ * making this specific to the implementation makes it possible to tie this
+ * into signal or atexit handlers such that we know to clean up stale locks on
+ * abnormal exits.
+ */
+struct reftable_flock {
+	const char *path;
+	int fd;
+	void *priv;
+};
+#define REFTABLE_FLOCK_INIT ((struct reftable_flock){ .fd = -1, })
+
+/*
+ * Acquire the lock for the given target path by exclusively creating a file
+ * with ".lock" appended to it. If that lock exists, we wait up to `timeout_ms`
+ * to acquire the lock. If `timeout_ms` is 0 we don't wait, if it is negative
+ * we block indefinitely.
+ *
+ * Retrun 0 on success, a reftable error code on error.
+ */
+int flock_acquire(struct reftable_flock *l, const char *target_path,
+		  long timeout_ms);
+
+/*
+ * Close the lockfile's file descriptor without removing the lock itself. This
+ * is a no-op in case the lockfile has already been closed beforehand. Returns
+ * 0 on success, a reftable error code on error.
+ */
+int flock_close(struct reftable_flock *l);
+
+/*
+ * Release the lock by unlinking the lockfile. This is a no-op in case the
+ * lockfile has already been released or committed beforehand. Returns 0 on
+ * success, a reftable error code on error.
+ */
+int flock_release(struct reftable_flock *l);
+
+/*
+ * Commit the lock by renaming the lockfile into place. Returns 0 on success, a
+ * reftable error code on error.
+ */
+int flock_commit(struct reftable_flock *l);
+
 #endif

t/unit-tests/lib-reftable.c

@@ -2,6 +2,7 @@
 #include "test-lib.h"
 #include "reftable/constants.h"
 #include "reftable/writer.h"
+#include "strbuf.h"
 
 void t_reftable_set_hash(uint8_t *p, int i, enum reftable_hash id)
 {

t/unit-tests/t-reftable-block.c

@@ -11,6 +11,7 @@ license that can be found in the LICENSE file or at
 #include "reftable/blocksource.h"
 #include "reftable/constants.h"
 #include "reftable/reftable-error.h"
+#include "strbuf.h"
 
 static void t_ref_block_read_write(void)
 {

t/unit-tests/t-reftable-pq.c

@@ -9,6 +9,7 @@ license that can be found in the LICENSE file or at
 #include "test-lib.h"
 #include "reftable/constants.h"
 #include "reftable/pq.h"
+#include "strbuf.h"
 
 static void merged_iter_pqueue_check(const struct merged_iter_pqueue *pq)
 {

t/unit-tests/t-reftable-readwrite.c

@@ -13,6 +13,7 @@ license that can be found in the LICENSE file or at
 #include "reftable/reader.h"
 #include "reftable/reftable-error.h"
 #include "reftable/reftable-writer.h"
+#include "strbuf.h"
 
 static const int update_index = 5;
 

t/unit-tests/t-reftable-stack.c

@@ -13,6 +13,8 @@ license that can be found in the LICENSE file or at
 #include "reftable/reader.h"
 #include "reftable/reftable-error.h"
 #include "reftable/stack.h"
+#include "strbuf.h"
+#include "tempfile.h"
 #include <dirent.h>
 
 static void clear_dir(const char *dirname)