Received-SPF: pass (google.com: domain of linux-kernel+bounces-127710-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20240327160314.9982-1-apais@linux.microsoft.com>
 <20240327160314.9982-10-apais@linux.microsoft.com> <CAPDyKFpuKadPQv6+61C2pE4x4FE-DUC5W6WCCPu9Nb2DnDB56g@mail.gmail.com>
 <ZgWZDtNU4tCwqyeu@slm.duckdns.org>
In-Reply-To: <ZgWZDtNU4tCwqyeu@slm.duckdns.org>
From: Ulf Hansson <ulf.hansson@linaro.org>
Date: Tue, 2 Apr 2024 12:15:45 +0200
Message-ID: <CAPDyKFp5KET0HR+8MwO4cf0O6W2kyFqHoKcVf5jbgBuLuQUcFA@mail.gmail.com>
Subject: Re: [PATCH 9/9] mmc: Convert from tasklet to BH workqueue
To: Tejun Heo <tj@kernel.org>
Cc: Allen Pais <apais@linux.microsoft.com>, linux-kernel@vger.kernel.org, 
	keescook@chromium.org, vkoul@kernel.org, marcan@marcan.st, sven@svenpeter.dev, 
	florian.fainelli@broadcom.com, rjui@broadcom.com, sbranden@broadcom.com, 
	paul@crapouillou.net, Eugeniy.Paltsev@synopsys.com, 
	manivannan.sadhasivam@linaro.org, vireshk@kernel.org, Frank.Li@nxp.com, 
	leoyang.li@nxp.com, zw@zh-kernel.org, wangzhou1@hisilicon.com, 
	haijie1@huawei.com, shawnguo@kernel.org, s.hauer@pengutronix.de, 
	sean.wang@mediatek.com, matthias.bgg@gmail.com, 
	angelogioacchino.delregno@collabora.com, afaerber@suse.de, 
	logang@deltatee.com, daniel@zonque.org, haojian.zhuang@gmail.com, 
	robert.jarzmik@free.fr, andersson@kernel.org, konrad.dybcio@linaro.org, 
	orsonzhai@gmail.com, baolin.wang@linux.alibaba.com, zhang.lyra@gmail.com, 
	patrice.chotard@foss.st.com, linus.walleij@linaro.org, wens@csie.org, 
	jernej.skrabec@gmail.com, peter.ujfalusi@gmail.com, kys@microsoft.com, 
	haiyangz@microsoft.com, wei.liu@kernel.org, decui@microsoft.com, 
	jassisinghbrar@gmail.com, mchehab@kernel.org, maintainers@bluecherrydvr.com, 
	aubin.constans@microchip.com, manuel.lauss@gmail.com, mirq-linux@rere.qmqm.pl, 
	jh80.chung@samsung.com, oakad@yahoo.com, hayashi.kunihiko@socionext.com, 
	mhiramat@kernel.org, brucechang@via.com.tw, HaraldWelte@viatech.com, 
	pierre@ossman.eu, duncan.sands@free.fr, stern@rowland.harvard.edu, 
	oneukum@suse.com, openipmi-developer@lists.sourceforge.net, 
	dmaengine@vger.kernel.org, asahi@lists.linux.dev, 
	linux-arm-kernel@lists.infradead.org, linux-rpi-kernel@lists.infradead.org, 
	linux-mips@vger.kernel.org, imx@lists.linux.dev, 
	linuxppc-dev@lists.ozlabs.org, linux-mediatek@lists.infradead.org, 
	linux-actions@lists.infradead.org, linux-arm-msm@vger.kernel.org, 
	linux-riscv@lists.infradead.org, linux-sunxi@lists.linux.dev, 
	linux-tegra@vger.kernel.org, linux-hyperv@vger.kernel.org, 
	linux-rdma@vger.kernel.org, linux-media@vger.kernel.org, 
	linux-mmc@vger.kernel.org, linux-omap@vger.kernel.org, 
	linux-renesas-soc@vger.kernel.org, linux-s390@vger.kernel.org, 
	netdev@vger.kernel.org, linux-usb@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"

On Thu, 28 Mar 2024 at 17:21, Tejun Heo <tj@kernel.org> wrote:
>
> Hello,
>
> On Thu, Mar 28, 2024 at 01:53:25PM +0100, Ulf Hansson wrote:
> > At this point we have suggested to drivers to switch to use threaded
> > irq handlers (and regular work queues if needed too). That said,
> > what's the benefit of using the BH work queue?
>
> BH workqueues should behave about the same as tasklets which have more
> limited interface and is subtly broken in an expensive-to-fix way (around
> freeing in-flight work item), so the plan is to replace tasklets with BH
> workqueues and remove tasklets from the kernel.

Seems like a good approach!

>
> The [dis]advantages of BH workqueues over threaded IRQs or regular threaded
> workqueues are the same as when you compare them to tasklets. No thread
> switching overhead, so latencies will be a bit tighter. Wheteher that
> actually matters really depends on the use case. Here, the biggest advantage
> is that it's mostly interchangeable with tasklets and can thus be swapped
> easily.

Right, thanks for clarifying!

However, the main question is then - if/when it makes sense to use the
BH workqueue for an mmc host driver. Unless there are some HW
limitations, a threaded irq handler should be sufficient, I think.

That said, moving to threaded irq handlers is a different topic and
doesn't prevent us from moving to BH workqueues as it seems like a
step in the right direction.

Kind regards
Uffe