[PATCH 1/5] ahci: imx: Pull out the clock enable/disable calls

Eric Nelson eric.nelson at boundarydevices.com
Mon Nov 18 17:11:50 EST 2013


Hi Marek,

On 11/18/2013 01:23 PM, Marek Vasut wrote:
> Hi Eric,
>
>> Hi Marek,
>>
>> On 11/16/2013 06:20 PM, Marek Vasut wrote:
>>> The same code for enabling and disabling SATA clock was found in multiple
>>> places in the driver. Implement functions that enable/disable the SATA
>>> clock and use them in such places instead of duplicating the code.
>>>
>>> Signed-off-by: Marek Vasut <marex at denx.de>
>>> Cc: Shawn Guo <shawn.guo at linaro.org>
>>> Cc: Richard Zhu <r65037 at freescale.com>
>>> Cc: Tejun Heo <tj at kernel.org>
>>> Cc: Linux-IDE <linux-ide at vger.kernel.org>
>>> ---
>>>
>>>    drivers/ata/ahci_imx.c | 133
>>>    ++++++++++++++++++++++++++++--------------------- 1 file changed, 75
>>>    insertions(+), 58 deletions(-)
>>>
>>> diff --git a/drivers/ata/ahci_imx.c b/drivers/ata/ahci_imx.c
>>> index ae2d73f..c7ee505 100644
>>> --- a/drivers/ata/ahci_imx.c
>>> +++ b/drivers/ata/ahci_imx.c
>>> @@ -47,6 +47,73 @@ static int ahci_imx_hotplug;
>>>
>>>    module_param_named(hotplug, ahci_imx_hotplug, int, 0644);
>>>    MODULE_PARM_DESC(hotplug, "AHCI IMX hot-plug support (0=Don't support,
>>>    1=support)");
>>>
>>> <snip>
>>
>> I haven't traced through all of this, but if you're copying from
>> the Freescale 3.0.35 kernel, note that there's a bug in it, and
>> the 0x7FFFFFFD really wanted to be an 0x7FFFFFFF.
>
> I'm not taking this from FSL 3.0.35, no, but thanks for pointing this out!
>
>> The way I read this comment, the writes need to happen in two
>> steps:
>> 	- write everything with the PHY disabled
>> 	- enable the PHY
>>
>> We had reports of stalls waiting for SATA drives to be enumerated
>> that were solved with this commit...
>>
>> 	https://github.com/boundarydevices/linux-
> imx6/commit/0186ea224ce6bd1cb4757
>> a0f83b0090e26a021f4
>
> [...]
>
>>> +	regmap_update_bits(imxpriv->gpr, IOMUXC_GPR13,
>>> +			IMX6Q_GPR13_SATA_MPLL_CLK_EN,
>>> +			IMX6Q_GPR13_SATA_MPLL_CLK_EN);
>
> Isn't this snippet doing exactly what your patch does ?
>
This part is doing the "set the bit" part:
	https://github.com/boundarydevices/linux-imx6/blob/boundary-imx_3.0.35_4.1.0/arch/arm/mach-mx6/board-mx6_nitrogen6x.c#L728

The previous block isn't clearing the enable bit though:

+		regmap_update_bits(imxpriv->gpr, IOMUXC_GPR13,
+				IMX6Q_GPR13_SATA_RX_EQ_VAL_MASK
+				| IMX6Q_GPR13_SATA_RX_LOS_LVL_MASK
+				| IMX6Q_GPR13_SATA_RX_DPLL_MODE_MASK
+				| IMX6Q_GPR13_SATA_SPD_MODE_MASK
+				| IMX6Q_GPR13_SATA_MPLL_SS_EN
+				| IMX6Q_GPR13_SATA_TX_ATTEN_MASK
+				| IMX6Q_GPR13_SATA_TX_BOOST_MASK
+				| IMX6Q_GPR13_SATA_TX_LVL_MASK
+				| IMX6Q_GPR13_SATA_TX_EDGE_RATE
+				, IMX6Q_GPR13_SATA_RX_EQ_VAL_3_0_DB
+				| IMX6Q_GPR13_SATA_RX_LOS_LVL_SATA2M
+				| IMX6Q_GPR13_SATA_RX_DPLL_MODE_2P_4F
+				| IMX6Q_GPR13_SATA_SPD_MODE_3P0G
+				| IMX6Q_GPR13_SATA_MPLL_SS_EN
+				| IMX6Q_GPR13_SATA_TX_ATTEN_9_16
+				| IMX6Q_GPR13_SATA_TX_BOOST_3_33_DB
+				| IMX6Q_GPR13_SATA_TX_LVL_1_025_V);

The explicit clearing of bit 1 is what was necessary (and I think
it's what the original author was after).

This was a hard one to find, because it seems that it only shows
up on some boards, and most of the time on those boards.

Here are some references:
	http://boundarydevices.com/freescale-ubuntu-image-with-4-0-0-kernel/#comment-60464
	http://boundarydevices.com/debian-installer-on-i-mx6-boards/#comment-76203

The symptom is (was) that U-Boot worked 100% of the time, but
the kernel failed to find the drive.

Using 'mm' in U-Boot to clear the bit before kernel launch also
allowed things to function.

Regards,


Eric



More information about the linux-arm-kernel mailing list