Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Feedback-ID: i56a14606:Fastmail
User-Agent: Cyrus-JMAP/3.7.0-alpha0-1047-g9e4af4ada4-fm-20221005.001-g9e4af4ad
Mime-Version: 1.0
Message-Id: <f8f803a8-ee36-4f32-8920-1fcf6b2265d1@app.fastmail.com>
In-Reply-To: <039075b210d78d2f4fdeb66b6826b8d2c2836088.1665931914.git.tonyhuang.sunplus@gmail.com>
References: <cover.1665931914.git.tonyhuang.sunplus@gmail.com>
 <039075b210d78d2f4fdeb66b6826b8d2c2836088.1665931914.git.tonyhuang.sunplus@gmail.com>
Date:   Mon, 17 Oct 2022 09:25:14 +0200
From:   "Arnd Bergmann" <arnd@arndb.de>
To:     "Tony Huang" <tonyhuang.sunplus@gmail.com>,
        "Ulf Hansson" <ulf.hansson@linaro.org>,
        "Rob Herring" <robh+dt@kernel.org>, krzk+dt@kernel.org,
        "linux-mmc @ vger . kernel . org" <linux-mmc@vger.kernel.org>,
        devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
        wells.lu@sunplus.com
Subject: Re: [PATCH v10 2/2] mmc: Add mmc driver for Sunplus SP7021
Content-Type: text/plain
Precedence: bulk

On Sun, Oct 16, 2022, at 5:48 PM, Tony Huang wrote:
> This is a patch for mmc driver for Sunplus SP7021 SOC.
> Supports eMMC 4.41 DDR 104MB/s speed mode.
>
> Signed-off-by: Tony Huang <tonyhuang.sunplus@gmail.com>

Looks ok to me me overall.

Acked-by: Arnd Bergmann <arnd@arndb.de>

Just one more thing I noticed:

> +#define SPMMC_TIMEOUT			500000
...
> +static inline int spmmc_wait_finish(struct spmmc_host *host)
> +{
> +	u32 state;
> +
> +	return readl_poll_timeout_atomic(host->base + SPMMC_SD_STATE_REG, 
> state,
> +					(state & SPMMC_SDSTATE_FINISH), 1, SPMMC_TIMEOUT);
> +}
> +
> +static inline int spmmc_wait_sdstatus(struct spmmc_host *host, 
> unsigned int status_bit)
> +{
> +	u32 status;
> +
> +	return readl_poll_timeout_atomic(host->base + SPMMC_SD_STATUS_REG, 
> status,
> +					(status & status_bit), 1, SPMMC_TIMEOUT);
> +}

500ms seems like an awfully long time for a busy-wait, I wonder if this
could be improved in some way. Is this always called from atomic context?

If not, any callers from non-atomic context could use
readl_poll_timeout() instead, or maybe there could be a shorter
timeout in atomic context, with a fallback to a non-atomic
workqueue if that times out, so only the MMC access will stall but
not the entire system.

The same problem does appear to be in dw_mmc.c and mtk-sd.c but not
in sdhci*.c, so I don't know if this is avoidable.

     Arnd