linux-kernel - Re: [PATCH 2/2] mfd: madera: Don't use regmap_read_poll_timeout to poll for BOOT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 16 Aug 2018 10:12:11 +0100
From:   Lee Jones <lee.jones@...aro.org>
To:     Richard Fitzgerald <rf@...nsource.cirrus.com>
Cc:     patches@...nsource.cirrus.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] mfd: madera: Don't use regmap_read_poll_timeout to
 poll for BOOT_DONE

On Tue, 07 Aug 2018, Richard Fitzgerald wrote:

> While polling for BOOT_DONE the chip could NAK a read because it is
> still booting, which would terminate the regmap_read_poll_timeout()
> with an error.
> 
> Instead implement a polling loop that ignores read errors so we
> always poll until the chip signals boot or the loop times out.
> 
> Signed-off-by: Richard Fitzgerald <rf@...nsource.cirrus.com>
> ---
>  drivers/mfd/madera-core.c | 38 ++++++++++++++++++++++++--------------
>  1 file changed, 24 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/mfd/madera-core.c b/drivers/mfd/madera-core.c
> index 8cfea969b060..bc5545c53ee5 100644
> --- a/drivers/mfd/madera-core.c
> +++ b/drivers/mfd/madera-core.c
> @@ -132,33 +132,43 @@ const char *madera_name_from_type(enum madera_type type)
>  }
>  EXPORT_SYMBOL_GPL(madera_name_from_type);
>  
> -#define MADERA_BOOT_POLL_MAX_INTERVAL_US  5000
> -#define MADERA_BOOT_POLL_TIMEOUT_US	 25000
> +#define MADERA_BOOT_POLL_INTERVAL_USEC		5000
> +#define MADERA_BOOT_POLL_TIMEOUT_USEC		25000
>  
>  static int madera_wait_for_boot(struct madera *madera)
>  {
> -	unsigned int val;
> -	int ret;
> +	ktime_t timeout;
> +	unsigned int val = 0;
> +	int ret = 0;
>  
>  	/*
>  	 * We can't use an interrupt as we need to runtime resume to do so,
>  	 * so we poll the status bit. This won't race with the interrupt
>  	 * handler because it will be blocked on runtime resume.
> +	 * The chip could NAK a read request while it is booting so ignore
> +	 * errors from regmap_read.
>  	 */
> -	ret = regmap_read_poll_timeout(madera->regmap,
> -				       MADERA_IRQ1_RAW_STATUS_1,
> -				       val,
> -				       (val & MADERA_BOOT_DONE_STS1),
> -				       MADERA_BOOT_POLL_MAX_INTERVAL_US,
> -				       MADERA_BOOT_POLL_TIMEOUT_US);
> -
> -	if (ret)
> -		dev_err(madera->dev, "Polling BOOT_DONE_STS failed: %d\n", ret);
> +	regmap_read(madera->regmap, MADERA_IRQ1_RAW_STATUS_1, &val);
> +	if (val & MADERA_BOOT_DONE_STS1)
> +		goto done;

Move the usleep_range() invokation to after the regmap_read(), then
you can remove these 3 lines and have the same semantics.

> +	timeout = ktime_add_us(ktime_get(), MADERA_BOOT_POLL_TIMEOUT_USEC);
> +	do {
> +		usleep_range(MADERA_BOOT_POLL_INTERVAL_USEC / 2,
> +			     MADERA_BOOT_POLL_INTERVAL_USEC);
> +		regmap_read(madera->regmap, MADERA_IRQ1_RAW_STATUS_1, &val);
> +		if (val & MADERA_BOOT_DONE_STS1)
> +			goto done;
> +	} while (ktime_compare(ktime_get(), timeout) <= 0);
> +
> +	dev_err(madera->dev, "Polling BOOT_DONE_STS timed out\n");
> +	ret = -ETIMEDOUT;
>  
>  	/*
>  	 * BOOT_DONE defaults to unmasked on boot so we must ack it.
> -	 * Do this unconditionally to avoid interrupt storms.
> +	 * Do this even after a timeout to avoid interrupt storms.
>  	 */
> +done:
>  	regmap_write(madera->regmap, MADERA_IRQ1_STATUS_1,
>  		     MADERA_BOOT_DONE_EINT1);
>  

-- 
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog