bulk-checkin: remove global transaction state

Object database transactions in the bulk-checkin subsystem rely on
global state to track transaction status. Stop relying on global state
and instead store the transaction in the `struct object_database`.
Functions that operate on transactions are updated to now wire
transaction state.

Signed-off-by: Justin Tobler <jltobler@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This commit is contained in:
Justin Tobler
2025-08-22 16:34:58 -05:00
committed by Junio C Hamano
parent 98518304c5
commit b336144725
9 changed files with 99 additions and 57 deletions

View File

@@ -389,6 +389,7 @@ int cmd_add(int argc,
char *seen = NULL;
char *ps_matched = NULL;
struct lock_file lock_file = LOCK_INIT;
struct odb_transaction *transaction;
repo_config(repo, add_config, NULL);
@@ -574,7 +575,7 @@ int cmd_add(int argc,
string_list_clear(&only_match_skip_worktree, 0);
}
begin_odb_transaction();
transaction = begin_odb_transaction(repo->objects);
ps_matched = xcalloc(pathspec.nr, 1);
if (add_renormalize)
@@ -593,7 +594,7 @@ int cmd_add(int argc,
if (chmod_arg && pathspec.nr)
exit_status |= chmod_pathspec(repo, &pathspec, chmod_arg[0], show_only);
end_odb_transaction();
end_odb_transaction(transaction);
finish:
if (write_locked_index(repo->index, &lock_file,

View File

@@ -584,6 +584,7 @@ static void unpack_all(void)
{
int i;
unsigned char *hdr = fill(sizeof(struct pack_header));
struct odb_transaction *transaction;
if (get_be32(hdr) != PACK_SIGNATURE)
die("bad pack file");
@@ -599,12 +600,12 @@ static void unpack_all(void)
progress = start_progress(the_repository,
_("Unpacking objects"), nr_objects);
CALLOC_ARRAY(obj_list, nr_objects);
begin_odb_transaction();
transaction = begin_odb_transaction(the_repository->objects);
for (i = 0; i < nr_objects; i++) {
unpack_one(i);
display_progress(progress, i + 1);
}
end_odb_transaction();
end_odb_transaction(transaction);
stop_progress(&progress);
if (delta_list)

View File

@@ -77,7 +77,7 @@ static void report(const char *fmt, ...)
* objects invisible while a transaction is active, so flush the
* transaction here before reporting a change made by update-index.
*/
flush_odb_transaction();
flush_odb_transaction(the_repository->objects->transaction);
va_start(vp, fmt);
vprintf(fmt, vp);
putchar('\n');
@@ -940,6 +940,7 @@ int cmd_update_index(int argc,
strbuf_getline_fn getline_fn;
int parseopt_state = PARSE_OPT_UNKNOWN;
struct repository *r = the_repository;
struct odb_transaction *transaction;
struct option options[] = {
OPT_BIT('q', NULL, &refresh_args.flags,
N_("continue refresh even when index needs update"),
@@ -1130,7 +1131,7 @@ int cmd_update_index(int argc,
* Allow the object layer to optimize adding multiple objects in
* a batch.
*/
begin_odb_transaction();
transaction = begin_odb_transaction(the_repository->objects);
while (ctx.argc) {
if (parseopt_state != PARSE_OPT_DONE)
parseopt_state = parse_options_step(&ctx, options,
@@ -1213,7 +1214,7 @@ int cmd_update_index(int argc,
/*
* By now we have added all of the new objects
*/
end_odb_transaction();
end_odb_transaction(transaction);
if (split_index > 0) {
if (repo_config_get_split_index(the_repository) == 0)

View File

@@ -30,11 +30,13 @@ struct bulk_checkin_packfile {
uint32_t nr_written;
};
static struct odb_transaction {
struct odb_transaction {
struct object_database *odb;
int nesting;
struct tmp_objdir *objdir;
struct bulk_checkin_packfile packfile;
} transaction;
};
static void finish_tmp_packfile(struct strbuf *basename,
const char *pack_tmp_name,
@@ -98,12 +100,12 @@ clear_exit:
/*
* Cleanup after batch-mode fsync_object_files.
*/
static void flush_batch_fsync(void)
static void flush_batch_fsync(struct odb_transaction *transaction)
{
struct strbuf temp_path = STRBUF_INIT;
struct tempfile *temp;
if (!transaction.objdir)
if (!transaction->objdir)
return;
/*
@@ -125,8 +127,8 @@ static void flush_batch_fsync(void)
* Make the object files visible in the primary ODB after their data is
* fully durable.
*/
tmp_objdir_migrate(transaction.objdir);
transaction.objdir = NULL;
tmp_objdir_migrate(transaction->objdir);
transaction->objdir = NULL;
}
static int already_written(struct bulk_checkin_packfile *state, struct object_id *oid)
@@ -325,7 +327,7 @@ static int deflate_blob_to_pack(struct bulk_checkin_packfile *state,
return 0;
}
void prepare_loose_object_bulk_checkin(void)
void prepare_loose_object_bulk_checkin(struct odb_transaction *transaction)
{
/*
* We lazily create the temporary object directory
@@ -333,15 +335,16 @@ void prepare_loose_object_bulk_checkin(void)
* callers may not know whether any objects will be
* added at the time they call begin_odb_transaction.
*/
if (!transaction.nesting || transaction.objdir)
if (!transaction || transaction->objdir)
return;
transaction.objdir = tmp_objdir_create(the_repository, "bulk-fsync");
if (transaction.objdir)
tmp_objdir_replace_primary_odb(transaction.objdir, 0);
transaction->objdir = tmp_objdir_create(the_repository, "bulk-fsync");
if (transaction->objdir)
tmp_objdir_replace_primary_odb(transaction->objdir, 0);
}
void fsync_loose_object_bulk_checkin(int fd, const char *filename)
void fsync_loose_object_bulk_checkin(struct odb_transaction *transaction,
int fd, const char *filename)
{
/*
* If we have an active ODB transaction, we issue a call that
@@ -350,7 +353,7 @@ void fsync_loose_object_bulk_checkin(int fd, const char *filename)
* before renaming the objects to their final names as part of
* flush_batch_fsync.
*/
if (!transaction.objdir ||
if (!transaction || !transaction->objdir ||
git_fsync(fd, FSYNC_WRITEOUT_ONLY) < 0) {
if (errno == ENOSYS)
warning(_("core.fsyncMethod = batch is unsupported on this platform"));
@@ -358,36 +361,57 @@ void fsync_loose_object_bulk_checkin(int fd, const char *filename)
}
}
int index_blob_bulk_checkin(struct object_id *oid,
int fd, size_t size,
int index_blob_bulk_checkin(struct odb_transaction *transaction,
struct object_id *oid, int fd, size_t size,
const char *path, unsigned flags)
{
int status = deflate_blob_to_pack(&transaction.packfile, oid, fd, size,
path, flags);
if (!transaction.nesting)
flush_bulk_checkin_packfile(&transaction.packfile);
int status;
if (transaction) {
status = deflate_blob_to_pack(&transaction->packfile, oid, fd,
size, path, flags);
} else {
struct bulk_checkin_packfile state = { 0 };
status = deflate_blob_to_pack(&state, oid, fd, size, path, flags);
flush_bulk_checkin_packfile(&state);
}
return status;
}
void begin_odb_transaction(void)
struct odb_transaction *begin_odb_transaction(struct object_database *odb)
{
transaction.nesting += 1;
if (!odb->transaction) {
CALLOC_ARRAY(odb->transaction, 1);
odb->transaction->odb = odb;
}
void flush_odb_transaction(void)
{
flush_batch_fsync();
flush_bulk_checkin_packfile(&transaction.packfile);
odb->transaction->nesting += 1;
return odb->transaction;
}
void end_odb_transaction(void)
void flush_odb_transaction(struct odb_transaction *transaction)
{
transaction.nesting -= 1;
if (transaction.nesting < 0)
BUG("Unbalanced ODB transaction nesting");
if (transaction.nesting)
if (!transaction)
return;
flush_odb_transaction();
flush_batch_fsync(transaction);
flush_bulk_checkin_packfile(&transaction->packfile);
}
void end_odb_transaction(struct odb_transaction *transaction)
{
if (!transaction || transaction->nesting == 0)
BUG("Unbalanced ODB transaction nesting");
transaction->nesting -= 1;
if (transaction->nesting)
return;
flush_odb_transaction(transaction);
transaction->odb->transaction = NULL;
free(transaction);
}

View File

@@ -5,9 +5,13 @@
#define BULK_CHECKIN_H
#include "object.h"
#include "odb.h"
void prepare_loose_object_bulk_checkin(void);
void fsync_loose_object_bulk_checkin(int fd, const char *filename);
struct odb_transaction;
void prepare_loose_object_bulk_checkin(struct odb_transaction *transaction);
void fsync_loose_object_bulk_checkin(struct odb_transaction *transaction,
int fd, const char *filename);
/*
* This creates one packfile per large blob unless bulk-checkin
@@ -24,8 +28,8 @@ void fsync_loose_object_bulk_checkin(int fd, const char *filename);
* binary blobs, they generally do not want to get any conversion, and
* callers should avoid this code path when filters are requested.
*/
int index_blob_bulk_checkin(struct object_id *oid,
int fd, size_t size,
int index_blob_bulk_checkin(struct odb_transaction *transaction,
struct object_id *oid, int fd, size_t size,
const char *path, unsigned flags);
/*
@@ -35,20 +39,20 @@ int index_blob_bulk_checkin(struct object_id *oid,
* and objects are only visible after the outermost transaction
* is complete or the transaction is flushed.
*/
void begin_odb_transaction(void);
struct odb_transaction *begin_odb_transaction(struct object_database *odb);
/*
* Make any objects that are currently part of a pending object
* database transaction visible. It is valid to call this function
* even if no transaction is active.
*/
void flush_odb_transaction(void);
void flush_odb_transaction(struct odb_transaction *transaction);
/*
* Tell the object database to make any objects from the
* current transaction visible if this is the final nested
* transaction.
*/
void end_odb_transaction(void);
void end_odb_transaction(struct odb_transaction *transaction);
#endif

View File

@@ -474,6 +474,7 @@ static int update_one(struct cache_tree *it,
int cache_tree_update(struct index_state *istate, int flags)
{
struct odb_transaction *transaction;
int skip, i;
i = verify_cache(istate, flags);
@@ -489,10 +490,10 @@ int cache_tree_update(struct index_state *istate, int flags)
trace_performance_enter();
trace2_region_enter("cache_tree", "update", the_repository);
begin_odb_transaction();
transaction = begin_odb_transaction(the_repository->objects);
i = update_one(istate->cache_tree, istate->cache, istate->cache_nr,
"", 0, &skip, flags);
end_odb_transaction();
end_odb_transaction(transaction);
trace2_region_leave("cache_tree", "update", the_repository);
trace_performance_leave("cache_tree_update");
if (i < 0)

View File

@@ -674,7 +674,7 @@ static void close_loose_object(struct odb_source *source,
goto out;
if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT))
fsync_loose_object_bulk_checkin(fd, filename);
fsync_loose_object_bulk_checkin(source->odb->transaction, fd, filename);
else if (fsync_object_files > 0)
fsync_or_die(fd, filename);
else
@@ -852,7 +852,7 @@ static int write_loose_object(struct odb_source *source,
static struct strbuf filename = STRBUF_INIT;
if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT))
prepare_loose_object_bulk_checkin();
prepare_loose_object_bulk_checkin(source->odb->transaction);
odb_loose_path(source, &filename, oid);
@@ -941,7 +941,7 @@ int stream_loose_object(struct odb_source *source,
int hdrlen;
if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT))
prepare_loose_object_bulk_checkin();
prepare_loose_object_bulk_checkin(source->odb->transaction);
/* Since oid is not determined, save tmp file to odb path. */
strbuf_addf(&filename, "%s/", source->path);
@@ -1263,8 +1263,9 @@ int index_fd(struct index_state *istate, struct object_id *oid,
ret = index_core(istate, oid, fd, xsize_t(st->st_size),
type, path, flags);
else
ret = index_blob_bulk_checkin(oid, fd, xsize_t(st->st_size), path,
flags);
ret = index_blob_bulk_checkin(the_repository->objects->transaction,
oid, fd, xsize_t(st->st_size),
path, flags);
close(fd);
return ret;
}

8
odb.h
View File

@@ -84,6 +84,7 @@ struct odb_source {
struct packed_git;
struct cached_object_entry;
struct odb_transaction;
/*
* The object database encapsulates access to objects in a repository. It
@@ -94,6 +95,13 @@ struct object_database {
/* Repository that owns this database. */
struct repository *repo;
/*
* State of current current object database transaction. Only one
* transaction may be pending at a time. Is NULL when no transaction is
* configured.
*/
struct odb_transaction *transaction;
/*
* Set of all object directories; the main directory is first (and
* cannot be NULL after initialization). Subsequent directories are

View File

@@ -3947,6 +3947,7 @@ int add_files_to_cache(struct repository *repo, const char *prefix,
const struct pathspec *pathspec, char *ps_matched,
int include_sparse, int flags)
{
struct odb_transaction *transaction;
struct update_callback_data data;
struct rev_info rev;
@@ -3972,9 +3973,9 @@ int add_files_to_cache(struct repository *repo, const char *prefix,
* This function is invoked from commands other than 'add', which
* may not have their own transaction active.
*/
begin_odb_transaction();
transaction = begin_odb_transaction(repo->objects);
run_diff_files(&rev, DIFF_RACY_IS_MODIFIED);
end_odb_transaction();
end_odb_transaction(transaction);
release_revisions(&rev);
return !!data.add_errors;