On 22/12/14 15:52, Kishon Vijay Abraham I wrote:
> Hi Roger,
>
> On Friday 19 December 2014 05:35 PM, Roger Quadros wrote:
>> Failed test case: Boot without SATA drive connected. Suspend/resume
>> the board and then connect SATA drive. It fails to enumerate.
>>
>> Due to Errata i783 "SATA Lockup After SATA DPLL Unlock/Relock"
>> we can't allow SATA DPLL to be in the unlocked state.
>> The SATA refclk (sata_ref_clk) is the source of the SATA_DPLL.
>> Till now this clock was controlled by the AHCI SATA driver and was being
>> shut off during system suspend (if the SATA drive was not already attached)
>> causing the SATA DPLL to be unlocked and so causing errata i783.
>>
>> To prevent sata_ref_clk from being disabled, we move the control of
>> this clock from the SATA AHCI driver to the SATA PHY driver and prevent
>> it from being disabled.
>>
>> This also fixes the issue of SATA not working on OMAP5/DRA7 when
>> AHCI platform driver is built as a module.
>>
>> Signed-off-by: Roger Quadros <rogerq@xxxxxx>
>> ---
>> arch/arm/boot/dts/dra7.dtsi | 4 ++--
>> arch/arm/boot/dts/omap5.dtsi | 4 ++--
>> drivers/phy/phy-ti-pipe3.c | 53 +++++++++++++++++++++++++++++++-------------
>> 3 files changed, 41 insertions(+), 20 deletions(-)
>>
>> diff --git a/arch/arm/boot/dts/dra7.dtsi b/arch/arm/boot/dts/dra7.dtsi
>> index 63bf99b..8c35b84 100644
>> --- a/arch/arm/boot/dts/dra7.dtsi
>> +++ b/arch/arm/boot/dts/dra7.dtsi
>> @@ -1090,8 +1090,8 @@
>> <0x4A096800 0x40>; /* pll_ctrl */
>> reg-names = "phy_rx", "phy_tx", "pll_ctrl";
>> ctrl-module = <&omap_control_sata>;
>> - clocks = <&sys_clkin1>;
>> - clock-names = "sysclk";
>> + clocks = <&sys_clkin1>, <&sata_ref_clk>;
>> + clock-names = "sysclk", "refclk";
>> #phy-cells = <0>;
>> };
>>
>> diff --git a/arch/arm/boot/dts/omap5.dtsi b/arch/arm/boot/dts/omap5.dtsi
>> index b321fdf..bb498e7 100644
>> --- a/arch/arm/boot/dts/omap5.dtsi
>> +++ b/arch/arm/boot/dts/omap5.dtsi
>> @@ -929,8 +929,8 @@
>> <0x4A096800 0x40>; /* pll_ctrl */
>> reg-names = "phy_rx", "phy_tx", "pll_ctrl";
>> ctrl-module = <&omap_control_sata>;
>> - clocks = <&sys_clkin>;
>> - clock-names = "sysclk";
>> + clocks = <&sys_clkin>, <&sata_ref_clk>;
>> + clock-names = "sysclk", "refclk";
>> #phy-cells = <0>;
>> };
>> };
>> diff --git a/drivers/phy/phy-ti-pipe3.c b/drivers/phy/phy-ti-pipe3.c
>> index e60ff14..e08edd9 100644
>> --- a/drivers/phy/phy-ti-pipe3.c
>> +++ b/drivers/phy/phy-ti-pipe3.c
>> @@ -85,6 +85,7 @@ struct ti_pipe3 {
>> struct pipe3_dpll_map *dpll_map;
>> u8 id;
>> bool enabled;
>> + bool refclk_enabled; /* this flag is needed specifically for SATA */
>> spinlock_t lock; /* serialize clock enable/disable */
>> };
>>
>> @@ -333,21 +334,20 @@ static int ti_pipe3_probe(struct platform_device *pdev)
>> }
>> }
>>
>> + phy->refclk = devm_clk_get(phy->dev, "refclk");
>> + if (IS_ERR(phy->refclk)) {
>> + dev_err(&pdev->dev, "unable to get refclk\n");
>> + return PTR_ERR(phy->refclk);
>> + }
>
> This will break older dtbs. AFAIK, newer kernels should be compatible with
> older dtbs too. cc'ed devicetree@xxxxxxxxxxxxxxx for clarification.
If I make refclk optional that still leaves SATA broken on older dtbs. Wouldn't it be better to fix the DTBs via stable instead?
cheers,
-roger
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
