Re: [PATCH] i2c: designware: Fix functionality in !CONFIG_ACPI case

From: Jarkko Nikula <jarkko.nikula@linux.intel.com>
To: John Stultz <john.stultz@linaro.org>,
	lkml <linux-kernel@vger.kernel.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
	Wolfram Sang <wsa@kernel.org>,
	linux-i2c@vger.kernel.org
Subject: Re: [PATCH] i2c: designware: Fix functionality in !CONFIG_ACPI case
Date: Tue, 23 Jun 2020 10:59:40 +0300	[thread overview]
Message-ID: <d6c317ee-4f97-9c57-6b04-1eabd814b6ce@linux.intel.com> (raw)
In-Reply-To: <20200623025144.34246-1-john.stultz@linaro.org>

Hi

On 6/23/20 5:51 AM, John Stultz wrote:
> On the HiKey board, where CONFIG_ACPI is not set, we started
> to see a graphics regression where the adv7511 HDMI bridge driver
> wasn't probing. This was due to the i2c bus failing to start up.
> 
> I bisected the problem down to commit f9288fcc5c615 ("i2c:
> designware: Move ACPI parts into common module") and after
> looking at it a bit, I realized that change moved some
> initialization into i2c_dw_acpi_adjust_bus_speed(). However,
> i2c_dw_acpi_adjust_bus_speed() is only functional if CONFIG_ACPI
> is set.
> 
> This patch pulls i2c_dw_acpi_adjust_bus_speed() out of the
> ifdef CONFIG_ACPI conditional, and gets the board working again.
> 
> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> Cc: Jarkko Nikula <jarkko.nikula@linux.intel.com>
> Cc: Wolfram Sang <wsa@kernel.org>
> Cc: linux-i2c@vger.kernel.org
> Fixes: f9288fcc5c615 ("i2c: designware: Move ACPI parts into common module")
> Signed-off-by: John Stultz <john.stultz@linaro.org>
> ---
>   drivers/i2c/busses/i2c-designware-common.c | 4 ++--
>   drivers/i2c/busses/i2c-designware-core.h   | 3 +--
>   2 files changed, 3 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-designware-common.c b/drivers/i2c/busses/i2c-designware-common.c
> index e3a8640db7da..33de185e15f2 100644
> --- a/drivers/i2c/busses/i2c-designware-common.c
> +++ b/drivers/i2c/busses/i2c-designware-common.c
> @@ -286,6 +286,8 @@ int i2c_dw_acpi_configure(struct device *device)
>   }
>   EXPORT_SYMBOL_GPL(i2c_dw_acpi_configure);
>   
> +#endif	/* CONFIG_ACPI */
> +

I think the regression is in these lines below that were moved from 
probe to i2c_dw_acpi_adjust_bus_speed() and cause that neither 
"cloock-frequency" device property or I2C_MAX_FAST_MODE_FREQ affect the 
bus speed when CONFIG_ACPI is not set.

+       /*
+        * Find bus speed from the "clock-frequency" device property, ACPI
+        * or by using fast mode if neither is set.
+        */
+       if (acpi_speed && t->bus_freq_hz)
+               t->bus_freq_hz = min(t->bus_freq_hz, acpi_speed);
+       else if (acpi_speed || t->bus_freq_hz)
+               t->bus_freq_hz = max(t->bus_freq_hz, acpi_speed);
+       else
+               t->bus_freq_hz = I2C_MAX_FAST_MODE_FREQ;

Andy: what you think should the i2c_dw_acpi_adjust_bus_speed() fixed to 
return adjusted speed or zero if not found (also for !CONFIG_ACPI) and 
move above lines back to probe? It looks more clear to me that way and 
should fix the regression I think.

-- 
Jarkko