From: Patrick Steinhardt <ps@pks.im>
To: git@vger.kernel.org
Cc: karthik nayak <karthik.188@gmail.com>,
Junio C Hamano <gitster@pobox.com>
Subject: [PATCH v2 06/10] refs: skip collision checks in initial transactions
Date: Wed, 20 Nov 2024 08:51:35 +0100 [thread overview]
Message-ID: <20241120-pks-refs-optimize-migrations-v2-6-a233374b7452@pks.im> (raw)
In-Reply-To: <20241120-pks-refs-optimize-migrations-v2-0-a233374b7452@pks.im>
Reference transactions use `refs_verify_refname_available()` to check
for colliding references. This check consists of two parts:
- Checks for whether multiple ref updates in the same transaction
conflict with each other.
- Checks for whether existing refs conflict with any refs part of the
transaction.
While we generally cannot avoid the first check, the second check is
superfluous in cases where the transaction is an initial one in an
otherwise empty ref store. The check results in multiple ref reads as
well as the creation of a ref iterator for every ref we're checking,
which adds up quite fast when performing the check for many refs.
Introduce a new flag that allows us to skip this check and wire it up in
such that the backends pass it when running an initial transaction. This
leads to significant speedups when migrating ref storage backends. From
"files" to "reftable":
Benchmark 1: migrate files:reftable (refcount = 100000, revision = HEAD~)
Time (mean ± σ): 472.4 ms ± 6.7 ms [User: 175.9 ms, System: 285.2 ms]
Range (min … max): 463.5 ms … 483.2 ms 10 runs
Benchmark 2: migrate files:reftable (refcount = 100000, revision = HEAD)
Time (mean ± σ): 86.1 ms ± 1.9 ms [User: 67.9 ms, System: 16.0 ms]
Range (min … max): 82.9 ms … 90.9 ms 29 runs
Summary
migrate files:reftable (refcount = 100000, revision = HEAD) ran
5.48 ± 0.15 times faster than migrate files:reftable (refcount = 100000, revision = HEAD~)
And from "reftable" to "files":
Benchmark 1: migrate reftable:files (refcount = 100000, revision = HEAD~)
Time (mean ± σ): 452.7 ms ± 3.4 ms [User: 209.9 ms, System: 235.4 ms]
Range (min … max): 445.9 ms … 457.5 ms 10 runs
Benchmark 2: migrate reftable:files (refcount = 100000, revision = HEAD)
Time (mean ± σ): 95.2 ms ± 2.2 ms [User: 73.6 ms, System: 20.6 ms]
Range (min … max): 91.7 ms … 100.8 ms 28 runs
Summary
migrate reftable:files (refcount = 100000, revision = HEAD) ran
4.76 ± 0.11 times faster than migrate reftable:files (refcount = 100000, revision = HEAD~)
Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
refs.c | 37 +++++++++++++++++++++----------------
refs.h | 5 ++++-
refs/files-backend.c | 13 ++++++-------
refs/reftable-backend.c | 6 ++++--
t/helper/test-ref-store.c | 2 +-
5 files changed, 36 insertions(+), 27 deletions(-)
diff --git a/refs.c b/refs.c
index 0f10c565bbb4e0d91210c52a3221a224e4264d28..d690eb19b3fd7083a5309deb98738547e4f48040 100644
--- a/refs.c
+++ b/refs.c
@@ -2324,6 +2324,7 @@ int refs_verify_refname_available(struct ref_store *refs,
const char *refname,
const struct string_list *extras,
const struct string_list *skip,
+ int initial_transaction,
struct strbuf *err)
{
const char *slash;
@@ -2332,8 +2333,6 @@ int refs_verify_refname_available(struct ref_store *refs,
struct strbuf referent = STRBUF_INIT;
struct object_id oid;
unsigned int type;
- struct ref_iterator *iter;
- int ok;
int ret = -1;
/*
@@ -2363,7 +2362,8 @@ int refs_verify_refname_available(struct ref_store *refs,
if (skip && string_list_has_string(skip, dirname.buf))
continue;
- if (!refs_read_raw_ref(refs, dirname.buf, &oid, &referent,
+ if (!initial_transaction &&
+ !refs_read_raw_ref(refs, dirname.buf, &oid, &referent,
&type, &ignore_errno)) {
strbuf_addf(err, _("'%s' exists; cannot create '%s'"),
dirname.buf, refname);
@@ -2388,21 +2388,26 @@ int refs_verify_refname_available(struct ref_store *refs,
strbuf_addstr(&dirname, refname + dirname.len);
strbuf_addch(&dirname, '/');
- iter = refs_ref_iterator_begin(refs, dirname.buf, NULL, 0,
- DO_FOR_EACH_INCLUDE_BROKEN);
- while ((ok = ref_iterator_advance(iter)) == ITER_OK) {
- if (skip &&
- string_list_has_string(skip, iter->refname))
- continue;
+ if (!initial_transaction) {
+ struct ref_iterator *iter;
+ int ok;
- strbuf_addf(err, _("'%s' exists; cannot create '%s'"),
- iter->refname, refname);
- ref_iterator_abort(iter);
- goto cleanup;
- }
+ iter = refs_ref_iterator_begin(refs, dirname.buf, NULL, 0,
+ DO_FOR_EACH_INCLUDE_BROKEN);
+ while ((ok = ref_iterator_advance(iter)) == ITER_OK) {
+ if (skip &&
+ string_list_has_string(skip, iter->refname))
+ continue;
- if (ok != ITER_DONE)
- BUG("error while iterating over references");
+ strbuf_addf(err, _("'%s' exists; cannot create '%s'"),
+ iter->refname, refname);
+ ref_iterator_abort(iter);
+ goto cleanup;
+ }
+
+ if (ok != ITER_DONE)
+ BUG("error while iterating over references");
+ }
extra_refname = find_descendant_ref(dirname.buf, extras, skip);
if (extra_refname)
diff --git a/refs.h b/refs.h
index 024a370554e013d66febee635e4c0415ba061fe6..980bd20cf24e15144aeff991eeba8b27a936d386 100644
--- a/refs.h
+++ b/refs.h
@@ -101,13 +101,16 @@ int refs_read_symbolic_ref(struct ref_store *ref_store, const char *refname,
* both "foo" and with "foo/bar/baz" but not with "foo/bar" or
* "foo/barbados".
*
+ * If `initial_transaction` is truish, then all collision checks with
+ * preexisting refs are skipped.
+ *
* extras and skip must be sorted.
*/
-
int refs_verify_refname_available(struct ref_store *refs,
const char *refname,
const struct string_list *extras,
const struct string_list *skip,
+ int initial_transaction,
struct strbuf *err);
int refs_ref_exists(struct ref_store *refs, const char *refname);
diff --git a/refs/files-backend.c b/refs/files-backend.c
index 116d4259697b20583cb2db34ed47025e8781cd42..d27806c02c272f8bddc789b509e3c3c7af4f75aa 100644
--- a/refs/files-backend.c
+++ b/refs/files-backend.c
@@ -706,7 +706,7 @@ static int lock_raw_ref(struct files_ref_store *refs,
* reason to expect this error to be transitory.
*/
if (refs_verify_refname_available(&refs->base, refname,
- extras, NULL, err)) {
+ extras, NULL, 0, err)) {
if (mustexist) {
/*
* To the user the relevant error is
@@ -813,7 +813,7 @@ static int lock_raw_ref(struct files_ref_store *refs,
REMOVE_DIR_EMPTY_ONLY)) {
if (refs_verify_refname_available(
&refs->base, refname,
- extras, NULL, err)) {
+ extras, NULL, 0, err)) {
/*
* The error message set by
* verify_refname_available() is OK.
@@ -850,7 +850,7 @@ static int lock_raw_ref(struct files_ref_store *refs,
*/
if (refs_verify_refname_available(
refs->packed_ref_store, refname,
- extras, NULL, err)) {
+ extras, NULL, 0, err)) {
ret = TRANSACTION_NAME_CONFLICT;
goto error_return;
}
@@ -1159,7 +1159,7 @@ static struct ref_lock *lock_ref_oid_basic(struct files_ref_store *refs,
*/
if (is_null_oid(&lock->old_oid) &&
refs_verify_refname_available(refs->packed_ref_store, refname,
- NULL, NULL, err))
+ NULL, NULL, 0, err))
goto error_return;
lock->ref_name = xstrdup(refname);
@@ -1538,7 +1538,7 @@ static int refs_rename_ref_available(struct ref_store *refs,
string_list_insert(&skip, old_refname);
ok = !refs_verify_refname_available(refs, new_refname,
- NULL, &skip, &err);
+ NULL, &skip, 0, &err);
if (!ok)
error("%s", err.buf);
@@ -3043,8 +3043,7 @@ static int files_transaction_finish_initial(struct files_ref_store *refs,
BUG("initial ref transaction with old_sha1 set");
if (refs_verify_refname_available(&refs->base, update->refname,
- &affected_refnames, NULL,
- err)) {
+ &affected_refnames, NULL, 1, err)) {
ret = TRANSACTION_NAME_CONFLICT;
goto cleanup;
}
diff --git a/refs/reftable-backend.c b/refs/reftable-backend.c
index 8e914afc817f198bed3199630b430e179cabc740..bbc779ab410b41f00759a3a41a76dd708f115c90 100644
--- a/refs/reftable-backend.c
+++ b/refs/reftable-backend.c
@@ -1097,7 +1097,9 @@ static int reftable_be_transaction_prepare(struct ref_store *ref_store,
* at a later point.
*/
ret = refs_verify_refname_available(ref_store, u->refname,
- &affected_refnames, NULL, err);
+ &affected_refnames, NULL,
+ transaction->flags & REF_TRANSACTION_FLAG_INITIAL,
+ err);
if (ret < 0)
goto done;
@@ -1584,7 +1586,7 @@ static int write_copy_table(struct reftable_writer *writer, void *cb_data)
if (arg->delete_old)
string_list_insert(&skip, arg->oldname);
ret = refs_verify_refname_available(&arg->refs->base, arg->newname,
- NULL, &skip, &errbuf);
+ NULL, &skip, 0, &errbuf);
if (ret < 0) {
error("%s", errbuf.buf);
goto done;
diff --git a/t/helper/test-ref-store.c b/t/helper/test-ref-store.c
index 65346dee551ccd781a88786f0c8465f60b286cde..240f6fc29d7f1bb20658deee467bcb46ac3407b2 100644
--- a/t/helper/test-ref-store.c
+++ b/t/helper/test-ref-store.c
@@ -199,7 +199,7 @@ static int cmd_verify_ref(struct ref_store *refs, const char **argv)
struct strbuf err = STRBUF_INIT;
int ret;
- ret = refs_verify_refname_available(refs, refname, NULL, NULL, &err);
+ ret = refs_verify_refname_available(refs, refname, NULL, NULL, 0, &err);
if (err.len)
puts(err.buf);
return ret;
--
2.47.0.274.g962d0b743d.dirty
next prev parent reply other threads:[~2024-11-20 7:51 UTC|newest]
Thread overview: 54+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-11-08 9:34 [PATCH 00/10] refs: optimize ref format migrations Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 01/10] refs: allow passing flags when setting up a transaction Patrick Steinhardt
2024-11-11 10:30 ` karthik nayak
2024-11-11 12:53 ` Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 02/10] refs/files: move logic to commit initial transaction Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 03/10] refs: introduce "initial" transaction flag Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 04/10] refs/files: support symbolic and root refs in initial transaction Patrick Steinhardt
2024-11-11 10:42 ` karthik nayak
2024-11-11 12:53 ` Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 05/10] refs: use "initial" transaction semantics to migrate refs Patrick Steinhardt
2024-11-11 10:43 ` karthik nayak
2024-11-08 9:34 ` [PATCH 06/10] refs: skip collision checks in initial transactions Patrick Steinhardt
2024-11-11 10:53 ` karthik nayak
2024-11-08 9:34 ` [PATCH 07/10] refs: don't normalize log messages with `REF_SKIP_CREATE_REFLOG` Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 08/10] reftable/writer: optimize allocations by using a scratch buffer Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 09/10] reftable/block: rename `block_writer::buf` variable Patrick Steinhardt
2024-11-08 9:34 ` [PATCH 10/10] reftable/block: optimize allocations by using scratch buffer Patrick Steinhardt
2024-11-11 10:57 ` [PATCH 00/10] refs: optimize ref format migrations karthik nayak
2024-11-11 12:53 ` Patrick Steinhardt
2024-11-20 7:04 ` Junio C Hamano
2024-11-20 7:50 ` Patrick Steinhardt
2024-11-20 10:25 ` Christian Couder
2024-11-25 5:52 ` Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 " Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 01/10] refs: allow passing flags when setting up a transaction Patrick Steinhardt
2024-11-20 10:19 ` Christian Couder
2024-11-20 7:51 ` [PATCH v2 02/10] refs/files: move logic to commit initial transaction Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 03/10] refs: introduce "initial" transaction flag Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 04/10] refs/files: support symbolic and root refs in initial transaction Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 05/10] refs: use "initial" transaction semantics to migrate refs Patrick Steinhardt
2024-11-20 7:51 ` Patrick Steinhardt [this message]
2024-11-20 10:21 ` [PATCH v2 06/10] refs: skip collision checks in initial transactions Christian Couder
2024-11-25 5:52 ` Patrick Steinhardt
2024-11-20 10:42 ` Kristoffer Haugsbakk
2024-11-25 5:52 ` Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 07/10] refs: don't normalize log messages with `REF_SKIP_CREATE_REFLOG` Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 08/10] reftable/writer: optimize allocations by using a scratch buffer Patrick Steinhardt
2024-11-20 10:21 ` Christian Couder
2024-11-25 5:52 ` Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 09/10] reftable/block: rename `block_writer::buf` variable Patrick Steinhardt
2024-11-20 7:51 ` [PATCH v2 10/10] reftable/block: optimize allocations by using scratch buffer Patrick Steinhardt
2024-11-20 10:22 ` Christian Couder
2024-11-25 6:27 ` [PATCH v3 00/10] refs: optimize ref format migrations Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 01/10] refs: allow passing flags when setting up a transaction Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 02/10] refs/files: move logic to commit initial transaction Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 03/10] refs: introduce "initial" transaction flag Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 04/10] refs/files: support symbolic and root refs in initial transaction Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 05/10] refs: use "initial" transaction semantics to migrate refs Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 06/10] refs: skip collision checks in initial transactions Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 07/10] refs: don't normalize log messages with `REF_SKIP_CREATE_REFLOG` Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 08/10] reftable/writer: optimize allocations by using a scratch buffer Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 09/10] reftable/block: rename `block_writer::buf` variable Patrick Steinhardt
2024-11-25 6:27 ` [PATCH v3 10/10] reftable/block: optimize allocations by using scratch buffer Patrick Steinhardt
2024-11-25 6:29 ` [PATCH v3 00/10] refs: optimize ref format migrations Patrick Steinhardt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20241120-pks-refs-optimize-migrations-v2-6-a233374b7452@pks.im \
--to=ps@pks.im \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=karthik.188@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).