All of lore.kernel.org
 help / color / mirror / Atom feed
From: Michael Haggerty <mhagger@alum.mit.edu>
To: Stefan Beller <sbeller@google.com>,
	peff@peff.net, git@vger.kernel.org, gitster@pobox.com,
	loic@dachary.org
Subject: Re: [PATCHv2 5/5] refs.c: enable large transactions
Date: Thu, 22 Jan 2015 12:24:23 +0100	[thread overview]
Message-ID: <54C0DDE7.8030708@alum.mit.edu> (raw)
In-Reply-To: <1421893929-7447-6-git-send-email-sbeller@google.com>

On 01/22/2015 03:32 AM, Stefan Beller wrote:
> By closing the file descriptors after creating the lock file we are not
> limiting the size of the transaction by the number of available file
> descriptors.
> 
> Signed-off-by: Stefan Beller <sbeller@google.com>
> ---
>  refs.c                | 17 +++++++++++++----
>  t/t1400-update-ref.sh |  4 ++--
>  2 files changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/refs.c b/refs.c
> index 2013d37..9d01102 100644
> --- a/refs.c
> +++ b/refs.c
> @@ -3055,11 +3055,18 @@ int is_branch(const char *refname)
>  static int write_sha1_to_lock_file(struct ref_lock *lock,
>  				   const unsigned char *sha1)
>  {
> -	if (fdopen_lock_file(lock->lk, "w") < 0
> -	    || fprintf(lock->lk->fp, "%s\n", sha1_to_hex(sha1)) != 41)
> +	if (lock->lk->fd == -1) {
> +		if (reopen_lock_file(lock->lk) < 0
> +		    || fdopen_lock_file(lock->lk, "w") < 0
> +		    || fprintf(lock->lk->fp, "%s\n", sha1_to_hex(sha1)) != 41
> +		    || close_lock_file(lock->lk) < 0)
> +		    return -1;
> +	} else {
> +		if (fdopen_lock_file(lock->lk, "w") < 0
> +		    || fprintf(lock->lk->fp, "%s\n", sha1_to_hex(sha1)) != 41)
>  		return -1;
> -	else
> -		return 0;
> +	}
> +	return 0;
>  }

I can't figure out where to apply this series or where to fetch it from,
so I can't see these changes in context, so maybe I'm misunderstanding
something. It looks like this code is doing

    open(), close(), open(), fdopen(), write(), fclose(), rename()

on each lockfile. But don't we have enough information to write the
SHA-1 into the lockfile the first time we touch it? I.e., couldn't we
reduce this to

    open(), fdopen(), write(), fclose(), rename()

, where the first four calls all happen in the initial loop? If a
problem is discovered when writing a later reference, we would roll back
the transaction anyway.

I understand that this would require a bigger rewrite, so maybe it is
not worth it.

>  /*
> @@ -3761,6 +3768,8 @@ int ref_transaction_commit(struct ref_transaction *transaction,
>  				    update->refname);
>  			goto cleanup;
>  		}
> +		/* Do not keep all lock files open at the same time. */
> +		close_lock_file(update->lock->lk);
>  	}
>  
>  	/* Perform updates first so live commits remain referenced */
> [...]

Michael

-- 
Michael Haggerty
mhagger@alum.mit.edu

  reply	other threads:[~2015-01-22 11:31 UTC|newest]

Thread overview: 45+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-01-22  2:32 [PATCHv2 0/5] Fix bug in large transactions Stefan Beller
2015-01-22  2:32 ` [PATCHv2 1/5] update-ref: test handling large transactions properly Stefan Beller
2015-01-22 10:54   ` Michael Haggerty
2015-01-22 13:07     ` Jeff King
2015-01-22  2:32 ` [PATCHv2 2/5] t7004: rename ULIMIT test prerequisite to ULIMIT_STACK_SIZE Stefan Beller
2015-01-22  2:32 ` [PATCHv2 3/5] refs.c: remove lock_fd from struct ref_lock Stefan Beller
2015-01-22  2:32 ` [PATCHv2 4/5] refs.c: have a write_sha1_to_lock_file wrapper Stefan Beller
2015-01-22  2:32 ` [PATCHv2 5/5] refs.c: enable large transactions Stefan Beller
2015-01-22 11:24   ` Michael Haggerty [this message]
2015-01-22 13:10     ` Jeff King
2015-01-22 16:33       ` Michael Haggerty
2015-01-22 19:24         ` Stefan Beller
2015-01-22 23:11     ` [RFC PATCH 0/5] So you dislike the sequence of system calls? Stefan Beller
2015-01-22 23:11       ` [PATCH 1/5] fixup for "refs.c: enable large transactions" Stefan Beller
2015-01-22 23:11       ` [PATCH 2/5] refs.c: remove unlock_ref from write_ref_sha1 Stefan Beller
2015-01-22 23:11       ` [PATCH 3/5] refs.c: move static functions to close and commit refs Stefan Beller
2015-01-22 23:11       ` [PATCH 4/5] refs.c: remove committing the ref from write_ref_sha1 Stefan Beller
2015-01-22 23:11       ` [PATCH 5/5] refs.c: write values to lock files early for committing Stefan Beller
2015-01-22 12:59   ` [PATCHv2 5/5] refs.c: enable large transactions Ramsay Jones
2015-01-22 19:16     ` Stefan Beller
2015-01-22 19:51       ` Ramsay Jones
2015-01-22 20:13         ` Ramsay Jones
2015-01-22 20:20           ` Stefan Beller
2015-01-22 20:59             ` Ramsay Jones
2015-01-22 12:05 ` [PATCHv2 0/5] Fix bug in " Michael Haggerty
2015-01-23 20:03   ` [PATCHv3 0/6] " Stefan Beller
2015-01-23 20:03     ` [PATCHv3 1/6] update-ref: test handling large transactions properly Stefan Beller
2015-01-23 20:03     ` [PATCHv3 2/6] t7004: rename ULIMIT test prerequisite to ULIMIT_STACK_SIZE Stefan Beller
2015-01-23 20:03     ` [PATCHv3 3/6] refs.c: remove lock_fd from struct ref_lock Stefan Beller
2015-01-23 20:04     ` [PATCHv3 4/6] refs.c: move static functions to close and commit refs Stefan Beller
2015-01-23 20:04     ` [PATCHv3 5/6] refs.c: remove unlock_ref and commit_ref from write_ref_sha1 Stefan Beller
2015-01-23 23:57       ` Junio C Hamano
2015-01-24  0:22         ` Stefan Beller
2015-01-24  0:39           ` Junio C Hamano
2015-01-24  1:04             ` Stefan Beller
2015-01-24  1:29               ` Junio C Hamano
2015-01-23 20:04     ` [PATCHv3 6/6] refs.c: enable large transactions Stefan Beller
2015-01-24  0:14       ` Junio C Hamano
2015-01-24  0:24         ` Stefan Beller
2015-01-24  0:38           ` Junio C Hamano
2015-01-26 19:30             ` Stefan Beller
2015-01-26 21:10               ` [PATCH] refs.c: clean up write_ref_sha1 returns Stefan Beller
2015-01-27  3:22                 ` Junio C Hamano
2015-01-28 21:35                   ` Stefan Beller
2015-01-27  3:17               ` [PATCHv3 6/6] refs.c: enable large transactions Junio C Hamano

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=54C0DDE7.8030708@alum.mit.edu \
    --to=mhagger@alum.mit.edu \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    --cc=loic@dachary.org \
    --cc=peff@peff.net \
    --cc=sbeller@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.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.