Tarantool development patches archive
 help / color / mirror / Atom feed
From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
To: Konstantin Osipov <kostja.osipov@gmail.com>,
	tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH v2 2/3] fiber: destroy fiber.storage created by iproto
Date: Sun, 19 Jan 2020 18:32:47 +0100	[thread overview]
Message-ID: <b305ecda-17d3-5af3-2a60-90020e08fe57@tarantool.org> (raw)
In-Reply-To: <20200117074552.GD24940@atlas>

Hi! Thanks for the review!

>> +/**
>> + * Stop the current fiber after a request is executed to make it
>> + * possible to reuse the fiber for a next request. On_stop
>> + * triggers remove all request-specific data from there.
>> + */
>> +static inline void
>> +tx_fiber_on_stop()
>> +{
>> +	fiber_on_stop(fiber());
>> +}
>> +
>>  static void
>>  tx_process_disconnect(struct cmsg *m)
>>  {
>> @@ -1335,6 +1351,7 @@ tx_process_disconnect(struct cmsg *m)
>>  		if (! rlist_empty(&session_on_disconnect)) {
>>  			tx_fiber_init(con->session, 0);
>>  			session_run_on_disconnect_triggers(con->session);
>> +			tx_fiber_on_stop();
>>  		}
>>  	}
> 
> Why did you have to add so many invocation points to
> fiber_on_stop() rather than simply adding fiber_on_stop invocation to
> fiber_pool.c?

We already discussed that in the previous patch version. We decided
to move cleanup to iproto.cc, because it depends on when a request
ends. Fiber pool knows nothing about requests. Iproto.cc is request
processing layer, and this is the right place for request data
destruction.

>>  static void
>>  fiber_recycle(struct fiber *fiber);
>>  
>> @@ -856,8 +869,7 @@ fiber_loop(MAYBE_UNUSED void *data)
>>  		       assert(f != fiber);
>>  		       fiber_wakeup(f);
>>  	        }
>> -		if (! rlist_empty(&fiber->on_stop))
>> -			trigger_run(&fiber->on_stop, fiber);
>> +		fiber_on_stop(fiber);
> 
> This was an attempt to optimize a non-inline function call
> for the most common case.
> 
> I would move this !rlist_empty check to fiber_on_stop and add a
> comment why we explicitly check for the list first.

I doubt it really helps, but ok.

================================================================================

diff --git a/src/lib/core/fiber.c b/src/lib/core/fiber.c
index 634b3d1b0..354749549 100644
--- a/src/lib/core/fiber.c
+++ b/src/lib/core/fiber.c
@@ -328,6 +328,12 @@ fiber_attr_getstacksize(struct fiber_attr *fiber_attr)
 void
 fiber_on_stop(struct fiber *f)
 {
+	/*
+	 * The most common case is when the list is empty. Do an
+	 * inlined check before calling trigger_run().
+	 */
+	if (rlist_empty(&f->on_stop))
+		return;
 	if (trigger_run(&f->on_stop, f) != 0)
 		panic("On_stop triggers can't fail");
 	/*

  reply	other threads:[~2020-01-19 17:32 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-16 21:54 [Tarantool-patches] [PATCH v2 0/3] Fiber storage leak Vladislav Shpilevoy
2020-01-16 21:54 ` [Tarantool-patches] [PATCH v2 1/3] fiber: unref fiber.storage via global Lua state Vladislav Shpilevoy
2020-01-17  7:30   ` Konstantin Osipov
2020-01-16 21:54 ` [Tarantool-patches] [PATCH v2 2/3] fiber: destroy fiber.storage created by iproto Vladislav Shpilevoy
2020-01-16 22:00   ` Cyrill Gorcunov
2020-01-17  7:47     ` Konstantin Osipov
2020-01-17  8:06       ` Cyrill Gorcunov
2020-01-17  7:45   ` Konstantin Osipov
2020-01-19 17:32     ` Vladislav Shpilevoy [this message]
2020-01-20  7:22       ` Konstantin Osipov
2020-01-20 19:15         ` Georgy Kirichenko
2020-01-21 22:21         ` Vladislav Shpilevoy
2020-01-21 22:32           ` Konstantin Osipov
2020-01-16 21:54 ` [Tarantool-patches] [PATCH v2 3/3] box: remove dead code from box_process_call/eval() Vladislav Shpilevoy
2020-01-17  7:46   ` Konstantin Osipov
2020-01-17  7:47   ` Konstantin Osipov
2020-01-17 17:41   ` Georgy Kirichenko
2020-01-19 17:32     ` Vladislav Shpilevoy
2020-01-20 19:21       ` Georgy Kirichenko
2020-01-18 19:27 ` [Tarantool-patches] [PATCH v2 0/3] Fiber storage leak Igor Munkin
2020-02-15  1:02 ` Vladislav Shpilevoy

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=b305ecda-17d3-5af3-2a60-90020e08fe57@tarantool.org \
    --to=v.shpilevoy@tarantool.org \
    --cc=kostja.osipov@gmail.com \
    --cc=tarantool-patches@dev.tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH v2 2/3] fiber: destroy fiber.storage created by iproto' \
    /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