From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <sergos@tarantool.org>
Received: from smtp21.mail.ru (smtp21.mail.ru [94.100.179.250])
 (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits))
 (No client certificate requested)
 by dev.tarantool.org (Postfix) with ESMTPS id 9D52D469719
 for <tarantool-patches@dev.tarantool.org>;
 Thu, 24 Sep 2020 16:15:27 +0300 (MSK)
Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.120.23.2.1\))
From: "sergos@tarantool.org" <sergos@tarantool.org>
In-Reply-To: <cover.1600862684.git.imun@tarantool.org>
Date: Thu, 24 Sep 2020 16:15:25 +0300
Content-Transfer-Encoding: quoted-printable
Message-Id: <4AE5222F-22B2-4752-AD8C-E5C1CCF7FE13@tarantool.org>
References: <cover.1600862684.git.imun@tarantool.org>
Subject: Re: [Tarantool-patches] [PATCH v2 0/2] Prevent JIT engine breakage
	on fibers switch-over
List-Id: Tarantool development patches <tarantool-patches.dev.tarantool.org>
List-Unsubscribe: <https://lists.tarantool.org/mailman/options/tarantool-patches>, 
 <mailto:tarantool-patches-request@dev.tarantool.org?subject=unsubscribe>
List-Archive: <https://lists.tarantool.org/pipermail/tarantool-patches/>
List-Post: <mailto:tarantool-patches@dev.tarantool.org>
List-Help: <mailto:tarantool-patches-request@dev.tarantool.org?subject=help>
List-Subscribe: <https://lists.tarantool.org/mailman/listinfo/tarantool-patches>, 
 <mailto:tarantool-patches-request@dev.tarantool.org?subject=subscribe>
To: Igor Munkin <imun@tarantool.org>
Cc: tarantool-patches@dev.tarantool.org, Vladislav Shpilevoy <v.shpilevoy@tarantool.org>

Hi!

Thanks for the patch, please consider my 2 cents below.=20

Sergos

> On 23 Sep 2020, at 22:06, Igor Munkin <imun@tarantool.org> wrote:
>=20
> There was a long discussion about the patch correctness and its
> performance impact in v1 thread[1]. The benchmarks provided in v1 =
showed
> this implementation as the least nerfing the platform overall
> performance even for such synthetic test[2]. One can find the relevant
> benchmarks results in the corresponding patches.
>=20
> Changes in v2:
> * implement the callback in a different way to negate its perf impact
>=20
> @ChangeLog:
> * Fixed fibers switch-over to prevent JIT machinery misbehaviour. =
Trace
>  recording is aborted when fiber yields the execution. The yield
>  occuring while mcode is being run leads to the platform panic

Could you please rephrase the last sentence to be more clear, like
yielding from a function called directly from the jitted code (means -
through FFI) will cause panic.=20

>  (gh-1700, gh-4491).
>=20
> Branch: =
https://github.com/tarantool/tarantool/tree/imun/gh-1700-abort-recording-o=
n-fiber-switch
> Issues:
> * https://github.com/tarantool/tarantool/issues/1700
> * https://github.com/tarantool/tarantool/issues/4491
>=20
> [1]: =
https://lists.tarantool.org/pipermail/tarantool-patches/2020-March/015290.=
html
> [2]: =
https://gist.github.com/igormunkin/7e0cf48005bd003ffbdf30181eedb40e
>=20
> Igor Munkin (2):
>  fiber: introduce a callback for fibers switch-over
>  lua: abort trace recording on fiber yield
>=20
> src/lib/core/fiber.c                          | 10 ++++
> src/lua/utils.c                               | 57 ++++++++++++++++++
> ...-4491-coio-wait-leads-to-segfault.test.lua | 53 +++++++++++++++++
> test/unit/CMakeLists.txt                      | 59 ++++++++++---------
> test/unit/core_test_utils.c                   | 37 ++++++++++++
> 5 files changed, 188 insertions(+), 28 deletions(-)
> create mode 100755 =
test/app-tap/gh-4491-coio-wait-leads-to-segfault.test.lua
> create mode 100644 test/unit/core_test_utils.c
>=20
> --=20
> 2.25.0
>=20