Tarantool development patches archive
 help / color / mirror / Atom feed
From: Vladimir Davydov <vdavydov.dev@gmail.com>
To: Konstantin Osipov <kostja@tarantool.org>
Cc: tarantool-patches@freelists.org
Subject: Re: [PATCH 05/12] alter: allocate triggers before the point of no return
Date: Tue, 10 Apr 2018 10:57:42 +0300	[thread overview]
Message-ID: <20180410075742.gh7wv7beff4h772a@esperanza> (raw)
In-Reply-To: <20180409203602.GF4527@atlas>

On Mon, Apr 09, 2018 at 11:36:02PM +0300, Konstantin Osipov wrote:
> * Vladimir Davydov <vdavydov.dev@gmail.com> [18/04/09 10:33]:
> > Trigger allocation, as any other memory allocation, is allowed to fail.
> > If this happens in alter_space_do, the space will be left in an
> > inconsistent state. Let's move trigger allocation to the beginning of
> > alter_space_do and add a comment denoting the point of no return.
> 
> Previously we would reference the allocated trigger immediately 
> in txn_on_commit() /txn_on_rollback(), the changed code leaks
> memory in case of any exception between allocation and "point of no
> return".
> 
> Please add guards.

We don't need guards for the triggers, because they are allocated on
region - see txn_alter_trigger_new().

> 
> > ---
> >  src/box/alter.cc | 14 ++++++++++----
> >  1 file changed, 10 insertions(+), 4 deletions(-)
> > 
> > diff --git a/src/box/alter.cc b/src/box/alter.cc
> > index 36310f1c..9d0c4c23 100644
> > --- a/src/box/alter.cc
> > +++ b/src/box/alter.cc
> > @@ -803,6 +803,11 @@ alter_space_rollback(struct trigger *trigger, void * /* event */)
> >  static void
> >  alter_space_do(struct txn *txn, struct alter_space *alter)
> >  {
> > +	/* Prepare triggers while we may fail. */
> > +	struct trigger *on_commit, *on_rollback;
> > +	on_commit = txn_alter_trigger_new(alter_space_commit, alter);
> > +	on_rollback = txn_alter_trigger_new(alter_space_rollback, alter);
> > +
> >  	/* Create a definition of the new space. */
> >  	space_dump_def(alter->old_space, &alter->key_list);
> >  	class AlterSpaceOp *op;
> > @@ -853,6 +858,11 @@ alter_space_do(struct txn *txn, struct alter_space *alter)
> >  		throw;
> >  	}
> >  
> > +	/*
> > +	 * This function must not throw exceptions or yield after
> > +	 * this point.
> > +	 */
> > +
> >  	/* Rebuild index maps once for all indexes. */
> >  	space_fill_index_map(alter->old_space);
> >  	space_fill_index_map(alter->new_space);
> > @@ -873,11 +883,7 @@ alter_space_do(struct txn *txn, struct alter_space *alter)
> >  	 * finish or rollback the DDL depending on the results of
> >  	 * writing to WAL.
> >  	 */
> > -	struct trigger *on_commit =
> > -		txn_alter_trigger_new(alter_space_commit, alter);
> >  	txn_on_commit(txn, on_commit);
> > -	struct trigger *on_rollback =
> > -		txn_alter_trigger_new(alter_space_rollback, alter);
> >  	txn_on_rollback(txn, on_rollback);
> >  }

  reply	other threads:[~2018-04-10  7:57 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-04-07 13:37 [PATCH 00/12] vinyl: allow to modify format of non-empty spaces Vladimir Davydov
2018-04-07 13:37 ` [PATCH 01/12] alter: introduce CheckSpaceFormat AlterSpaceOp for validating format Vladimir Davydov
2018-04-09 20:25   ` Konstantin Osipov
2018-04-07 13:37 ` [PATCH 02/12] alter: fold ModifySpaceFormat into ModifySpace Vladimir Davydov
2018-04-09 20:26   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 03/12] alter: move dictionary update from ModifySpace::alter_def to alter Vladimir Davydov
2018-04-09 20:32   ` Konstantin Osipov
2018-04-10  7:53     ` Vladimir Davydov
2018-04-10 11:45     ` Vladimir Davydov
2018-04-07 13:38 ` [PATCH 04/12] alter: use space_index instead of index_find where appropriate Vladimir Davydov
2018-04-09 20:34   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 05/12] alter: allocate triggers before the point of no return Vladimir Davydov
2018-04-09 20:36   ` Konstantin Osipov
2018-04-10  7:57     ` Vladimir Davydov [this message]
2018-04-10 11:54       ` Vladimir Davydov
2018-04-07 13:38 ` [PATCH 06/12] space: space_vtab::build_secondary_key => build_index Vladimir Davydov
2018-04-09 20:39   ` Konstantin Osipov
2018-04-10  8:05     ` Vladimir Davydov
2018-04-10 12:14       ` Vladimir Davydov
2018-04-07 13:38 ` [PATCH 07/12] space: pass new format instead of new space to space_vtab::check_format Vladimir Davydov
2018-04-09 20:40   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 08/12] alter: introduce preparation phase Vladimir Davydov
2018-04-09 20:46   ` [tarantool-patches] " Konstantin Osipov
2018-04-10  8:31     ` Vladimir Davydov
2018-04-10  8:46       ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 09/12] alter: zap space_def_check_compatibility Vladimir Davydov
2018-04-09 20:49   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 10/12] vinyl: remove superfluous ddl checks Vladimir Davydov
2018-04-09 20:49   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 11/12] vinyl: force index rebuild if indexed field type is narrowed Vladimir Davydov
2018-04-09 20:51   ` Konstantin Osipov
2018-04-07 13:38 ` [PATCH 12/12] vinyl: allow to modify format of non-empty spaces Vladimir Davydov
2018-04-09  8:24   ` Vladimir Davydov
2018-04-09 20:55   ` Konstantin Osipov

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=20180410075742.gh7wv7beff4h772a@esperanza \
    --to=vdavydov.dev@gmail.com \
    --cc=kostja@tarantool.org \
    --cc=tarantool-patches@freelists.org \
    --subject='Re: [PATCH 05/12] alter: allocate triggers before the point of no return' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox