Received-SPF: pass (google.com: domain of linux-kernel+bounces-166255-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Thu, 02 May 2024 11:52:41 +0200
Message-ID: <87h6fgk0ba.wl-tiwai@suse.de>
From: Takashi Iwai <tiwai@suse.de>
To: Richard Fitzgerald <rf@opensource.cirrus.com>
Cc: Simon Trimmer <simont@opensource.cirrus.com>,
	<tiwai@suse.com>,
	<linux-sound@vger.kernel.org>,
	<alsa-devel@alsa-project.org>,
	<linux-kernel@vger.kernel.org>,
	<patches@opensource.cirrus.com>
Subject: Re: [PATCH] ALSA: hda: cs35l56: Perform firmware download in the background
In-Reply-To: <a9345d24-af36-42b4-9139-0701a0dbe1a3@opensource.cirrus.com>
References: <20240501111755.21231-1-simont@opensource.cirrus.com>
	<87ttjgk6ph.wl-tiwai@suse.de>
	<a9345d24-af36-42b4-9139-0701a0dbe1a3@opensource.cirrus.com>
User-Agent: Wanderlust/2.15.9 (Almost Unreal) Emacs/27.2 Mule/6.0
Precedence: bulk
MIME-Version: 1.0 (generated by SEMI-EPG 1.14.7 - "Harue")
Content-Type: text/plain; charset=US-ASCII

On Thu, 02 May 2024 11:21:36 +0200,
Richard Fitzgerald wrote:
> 
> On 02/05/2024 08:34, Takashi Iwai wrote:
> > On Wed, 01 May 2024 13:17:55 +0200,
> > Simon Trimmer wrote:
> >> @@ -964,6 +1011,14 @@ int cs35l56_hda_common_probe(struct cs35l56_hda *cs35l56, int hid, int id)
> >>   	mutex_init(&cs35l56->base.irq_lock);
> >>   	dev_set_drvdata(cs35l56->base.dev, cs35l56);
> >>   +	cs35l56->dsp_wq =
> >> create_singlethread_workqueue("cs35l56-dsp");
> >> +	if (!cs35l56->dsp_wq) {
> >> +		ret = -ENOMEM;
> >> +		goto err;
> >> +	}
> > 
> > Do we really need a dedicated workqueue?  In most usages, simple
> > schedule_work*() works fine and is recommended.
> > 
> 
> On a slow I2C bus with 4 amps this work could take over 2 seconds.
> That seems too long to be blocking a global system queue. We use a
> dedicated queue in the ASoC driver.
> 
> Also if we queue work on an ordered (single-threaded) system queue the
> firmware won't be downloaded to multiple amps in parallel, so we don't
> get the best use of the available bus bandwidth.

OK, that sounds like a sensible argument.

But the patch has no call of a queue destructor.  Won't it leak
resources?


thanks,

Takashi