git@vger.kernel.org mailing list mirror (one of many)
 help / color / mirror / code / Atom feed
From: "SZEDER Gábor" <szeder.dev@gmail.com>
To: "Dominic Sacré" <dominic.sacre@gmx.de>
Cc: "SZEDER Gábor" <szeder.dev@gmail.com>,
	git@vger.kernel.org, jonathantanmy@google.com
Subject: Re: Regression in memory consumption of git fsck
Date: Wed, 14 Feb 2018 13:48:28 +0100	[thread overview]
Message-ID: <20180214124828.27845-1-szeder.dev@gmail.com> (raw)
In-Reply-To: <b1a4d7ed-79b6-1f6b-1f0f-85536e943cef@gmx.de>

> I've noticed that recent versions of git consume a lot of memory during
> "git fsck", to the point where I've regularly had git fall victim to
> Linux's OOM killer.
> 
> For example, if I clone torvalds/linux.git, and then run "git fsck
> --connectivity-only" in the newly cloned repository, git will consume
> more than 6GB of physical memory, while older versions peak at about 2GB.
> 
> I've managed to bisect this down to this commit in v2.14:
> 
> ad2db4030e42890e569de529e3cd61a8d03de497
> fsck: remove redundant parse_tree() invocation
> 
> If I revert that commit (on top of current master) the memory
> consumption goes down to 2GB again. The change looks relatively harmless
> to me, so does anyone know what's going on here?

I could reproduce the increased memory usage even for much smaller
repositories.  The patch below seems to fix it for me.


 -- >8 --

Subject: [PATCH] fsck: plug tree buffer leak

Commit ad2db4030e (fsck: remove redundant parse_tree() invocation,
2017-07-18), along with that redundant call to parse_tree() in
traverse_one_object(), also removed a call to free_tree_buffer() from
that function.  This resulted in significantly increased memory usage
of 'git fsck' because of all the non-freed tree buffers; in case of
git.git and '--connectivity-only' it went from around 270MB to over
1.2GB.

Restore that free_tree_buffer() call to bring down memory usage to the
previous level.

Reported-by: Dominic Sacré <dominic.sacre@gmx.de>
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
---

 builtin/fsck.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/builtin/fsck.c b/builtin/fsck.c
index 7a8a679d4f..8bc1b59daf 100644
--- a/builtin/fsck.c
+++ b/builtin/fsck.c
@@ -180,7 +180,10 @@ static void mark_object_reachable(struct object *obj)
 
 static int traverse_one_object(struct object *obj)
 {
-	return fsck_walk(obj, obj, &fsck_walk_options);
+	int result = fsck_walk(obj, obj, &fsck_walk_options);
+	if (obj->type == OBJ_TREE)
+		free_tree_buffer((struct tree *)obj);
+	return result;
 }
 
 static int traverse_reachable(void)
-- 
2.16.1.347.gd41f2872c6


  reply	other threads:[~2018-02-14 12:48 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-02-14 10:22 Regression in memory consumption of git fsck Dominic Sacré
2018-02-14 12:48 ` SZEDER Gábor [this message]
2018-02-14 13:55   ` Jeff King

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=20180214124828.27845-1-szeder.dev@gmail.com \
    --to=szeder.dev@gmail.com \
    --cc=dominic.sacre@gmx.de \
    --cc=git@vger.kernel.org \
    --cc=jonathantanmy@google.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).