git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: Jeff King <peff@peff.net>
To: 程洋 <chengyang@xiaomi.com>
Cc: "Derrick Stolee" <derrickstolee@github.com>,
	"git@vger.kernel.org" <git@vger.kernel.org>,
	何浩 <hehao@xiaomi.com>, "Xin7 Ma 马鑫" <maxin7@xiaomi.com>,
	石奉兵 <shifengbing@xiaomi.com>, 凡军辉 <fanjunhui@xiaomi.com>,
	王汉基 <wanghanji@xiaomi.com>
Subject: Re: [External Mail]Re: Partial-clone cause big performance impact on server
Date: Tue, 6 Sep 2022 14:38:41 -0400	[thread overview]
Message-ID: <YxeTsWrjpKo+JGfq@coredump.intra.peff.net> (raw)
In-Reply-To: <d5305274b7c24adbaf6ad9ab92ac3b6a@xiaomi.com>

On Mon, Sep 05, 2022 at 11:17:21AM +0000, 程洋 wrote:

> Sorry, I told you the wrong branch. It should be "android-t-preview-1"
> git clone --filter=blob:none --no-local -b android-t-preview-1 grade-plugin
> 
> Can you try this one?

Yes, I see more slow-down there. There are many more blobs there, but I
don't think it's really the number of them, but their sizes.

The problem is that both upload-pack and pack-objects are keen to call
parse_object() on their inputs. For commits, etc, that is usually
sensible; we have to parse the object to see what it points to. But for
blobs, the only thing we do is inflate a ton of bytes in order to check
the sha1. That's not really productive here; if there is a bit
corruption, the client will notice it on the receiving side.

So doing this:

diff --git a/object.c b/object.c
index 588b8156f1..6fbf6b2a7e 100644
--- a/object.c
+++ b/object.c
@@ -279,10 +279,6 @@ struct object *parse_object(struct repository *r, const struct object_id *oid)
 	if ((obj && obj->type == OBJ_BLOB && repo_has_object_file(r, oid)) ||
 	    (!obj && repo_has_object_file(r, oid) &&
 	     oid_object_info(r, oid, NULL) == OBJ_BLOB)) {
-		if (stream_object_signature(r, repl) < 0) {
-			error(_("hash mismatch %s"), oid_to_hex(oid));
-			return NULL;
-		}
 		parse_blob_buffer(lookup_blob(r, oid), NULL, 0);
 		return lookup_object(r, oid);
 	}

makes the follow-up fetch very fast. But there are some parts of the
code that rely on this parsing to do the on-disk checks. So I think we'd
probably need to take a flag, and plumb it through from the appropriate
spots.

A sample patch is below, which makes things fast for the example I
showed (it might not for yours, because if you're using the v0 protocol
over ssh, I suspect a different parse_object() call might need to be
tweaked). I think it should be fairly safe, because in both callsites
we're already using the commit-graph to avoid checking the sha1 for
commits (actually, I think this may be an existing subtle breakage in
"rev-list --verify-objects", but that should be fixed separately).

---
diff --git a/object.c b/object.c
index 588b8156f1..97324bc406 100644
--- a/object.c
+++ b/object.c
@@ -263,7 +263,9 @@ struct object *parse_object_or_die(const struct object_id *oid,
 	die(_("unable to parse object: %s"), name ? name : oid_to_hex(oid));
 }
 
-struct object *parse_object(struct repository *r, const struct object_id *oid)
+struct object *parse_object_with_flags(struct repository *r,
+				       const struct object_id *oid,
+				       enum parse_object_flags flags)
 {
 	unsigned long size;
 	enum object_type type;
@@ -279,7 +281,8 @@ struct object *parse_object(struct repository *r, const struct object_id *oid)
 	if ((obj && obj->type == OBJ_BLOB && repo_has_object_file(r, oid)) ||
 	    (!obj && repo_has_object_file(r, oid) &&
 	     oid_object_info(r, oid, NULL) == OBJ_BLOB)) {
-		if (stream_object_signature(r, repl) < 0) {
+		if (!(flags & PARSE_OBJECT_SKIP_HASH_CHECK) &&
+		    stream_object_signature(r, repl) < 0) {
 			error(_("hash mismatch %s"), oid_to_hex(oid));
 			return NULL;
 		}
@@ -289,7 +292,8 @@ struct object *parse_object(struct repository *r, const struct object_id *oid)
 
 	buffer = repo_read_object_file(r, oid, &type, &size);
 	if (buffer) {
-		if (check_object_signature(r, repl, buffer, size, type) < 0) {
+		if (!(flags & PARSE_OBJECT_SKIP_HASH_CHECK) &&
+		    check_object_signature(r, repl, buffer, size, type) < 0) {
 			free(buffer);
 			error(_("hash mismatch %s"), oid_to_hex(repl));
 			return NULL;
@@ -304,6 +308,11 @@ struct object *parse_object(struct repository *r, const struct object_id *oid)
 	return NULL;
 }
 
+struct object *parse_object(struct repository *r, const struct object_id *oid)
+{
+	return parse_object_with_flags(r, oid, 0);
+}
+
 struct object_list *object_list_insert(struct object *item,
 				       struct object_list **list_p)
 {
diff --git a/object.h b/object.h
index 9caef89f1f..cd5b7d6306 100644
--- a/object.h
+++ b/object.h
@@ -137,6 +137,13 @@ struct object *parse_object(struct repository *r, const struct object_id *oid);
  */
 struct object *parse_object_or_die(const struct object_id *oid, const char *name);
 
+enum parse_object_flags {
+	PARSE_OBJECT_SKIP_HASH_CHECK = 1 << 0,
+};
+struct object *parse_object_with_flags(struct repository *r,
+				       const struct object_id *oid,
+				       enum parse_object_flags flags);
+
 /* Given the result of read_sha1_file(), returns the object after
  * parsing it.  eaten_p indicates if the object has a borrowed copy
  * of buffer and the caller should not free() it.
diff --git a/revision.c b/revision.c
index ee702e498a..8cbffc0325 100644
--- a/revision.c
+++ b/revision.c
@@ -384,7 +384,8 @@ static struct object *get_reference(struct rev_info *revs, const char *name,
 	if (commit)
 		object = &commit->object;
 	else
-		object = parse_object(revs->repo, oid);
+		object = parse_object_with_flags(revs->repo, oid,
+						 PARSE_OBJECT_SKIP_HASH_CHECK);
 
 	if (!object) {
 		if (revs->ignore_missing)
diff --git a/upload-pack.c b/upload-pack.c
index b217a1f469..4bacdf087c 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1420,7 +1420,8 @@ static int parse_want(struct packet_writer *writer, const char *line,
 		if (commit)
 			o = &commit->object;
 		else
-			o = parse_object(the_repository, &oid);
+			o = parse_object_with_flags(the_repository, &oid,
+						    PARSE_OBJECT_SKIP_HASH_CHECK);
 
 		if (!o) {
 			packet_writer_error(writer,

  reply	other threads:[~2022-09-06 18:39 UTC|newest]

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-11  8:09 Partial-clone cause big performance impact on server 程洋
2022-08-11 17:22 ` Jonathan Tan
2022-08-13  7:55   ` 回复: [External Mail]Re: " 程洋
2022-08-13 11:41     ` 程洋
2022-08-15  5:16     ` ZheNing Hu
2022-08-15 13:15       ` 程洋
2022-08-12 12:21 ` Derrick Stolee
2022-08-14  6:48 ` Jeff King
2022-08-15 13:18   ` Derrick Stolee
2022-08-15 14:50     ` [External Mail]Re: " 程洋
2022-08-17 10:22     ` 程洋
2022-08-17 13:41       ` Derrick Stolee
2022-08-18  5:49         ` Jeff King
2022-09-01  6:53   ` 程洋
2022-09-01 16:19     ` Jeff King
2022-09-05 11:17       ` 程洋
2022-09-06 18:38         ` Jeff King [this message]
2022-09-06 22:58           ` [PATCH 0/3] speeding up on-demand fetch for blobs in partial clone Jeff King
2022-09-06 23:01             ` [PATCH 1/3] parse_object(): allow skipping hash check Jeff King
2022-09-07 14:15               ` Derrick Stolee
2022-09-07 20:44                 ` Jeff King
2022-09-06 23:05             ` [PATCH 2/3] upload-pack: skip parse-object re-hashing of "want" objects Jeff King
2022-09-07 14:36               ` Derrick Stolee
2022-09-07 14:45                 ` Derrick Stolee
2022-09-07 20:50                   ` Jeff King
2022-09-07 19:26               ` Junio C Hamano
2022-09-07 20:36                 ` Jeff King
2022-09-07 20:48                   ` [BUG] t1800: Fails for error text comparison rsbecker
2022-09-07 21:55                     ` Junio C Hamano
2022-09-07 22:23                       ` rsbecker
2022-09-07 21:02                   ` [PATCH 2/3] upload-pack: skip parse-object re-hashing of "want" objects Jeff King
2022-09-07 22:07                     ` Junio C Hamano
2022-09-08  5:04                       ` Jeff King
2022-09-08 16:41                         ` Junio C Hamano
2022-09-06 23:06             ` [PATCH 3/3] parse_object(): check commit-graph when skip_hash set Jeff King
2022-09-07 14:46               ` Derrick Stolee
2022-09-07 19:31               ` Junio C Hamano
2022-09-08 10:39                 ` [External Mail]Re: " 程洋
2022-09-08 18:42                   ` Jeff King
2022-09-07 14:48             ` [PATCH 0/3] speeding up on-demand fetch for blobs in partial clone Derrick Stolee

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

  List information: http://vger.kernel.org/majordomo-info.html

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YxeTsWrjpKo+JGfq@coredump.intra.peff.net \
    --to=peff@peff.net \
    --cc=chengyang@xiaomi.com \
    --cc=derrickstolee@github.com \
    --cc=fanjunhui@xiaomi.com \
    --cc=git@vger.kernel.org \
    --cc=hehao@xiaomi.com \
    --cc=maxin7@xiaomi.com \
    --cc=shifengbing@xiaomi.com \
    --cc=wanghanji@xiaomi.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.
Code repositories for project(s) associated with this public inbox

	https://80x24.org/mirrors/git.git

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 read-only IMAP folder(s) and NNTP newsgroup(s).