RE: [PATCH] drivers: ata: Move vendor specific sata port phy oob settings to device-tree

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Rob,


Thanks for reviewing the patch

> -----Original Message-----
> From: Rob Herring [mailto:robh@xxxxxxxxxx]
> Sent: Friday, November 13, 2015 8:02 PM
> To: Anurag Kumar Vulisha
> Cc: pawel.moll@xxxxxxx; mark.rutland@xxxxxxx;
> ijc+devicetree@xxxxxxxxxxxxxx; galak@xxxxxxxxxxxxxx; tj@xxxxxxxxxx;
> devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; linux-
> ide@xxxxxxxxxxxxxxx; Michal Simek; Punnaiah Choudary Kalluri; Anirudha
> Sarangi; Srikanth Vemula; Anurag Kumar Vulisha
> Subject: Re: [PATCH] drivers: ata: Move vendor specific sata port phy oob
> settings to device-tree
> 
> On Fri, Nov 13, 2015 at 04:02:23PM +0530, Anurag Kumar Vulisha wrote:
> > In SATA, speed negotiation happens with  OOB(Out of Band) signals.
> > These OOB signal timing values are configured through vendor specific
> > registers in the SATA controller. These OOB timings depends on the
> > generator and detector clock frequency, which varies from board to
> > board (ex: ep108, zcu102 and zc1751 has different clock frequencies).
> 
> Could you calculate the timings based on the frequency instead?
>

We can calculate the OOB timing settings based on frequency, but it requires complex calculations and floating 
point operations  in the sata driver , which is running in kernel mode. To my knowledge it is not recommended 
to do floating  operations inside kernel mode(Please correct me if I wrong ) . Because of this reason we are reading
those values from  device tree.
 
> > Since to make ahci_ceva driver generic, it would be better to move
> > these settings to the device-tree node and read them from driver.
> >
> > This patch does the same.
> >
> > Signed-off-by: Anurag Kumar Vulisha <anuragku@xxxxxxxxxx>
> > ---
> >  .../devicetree/bindings/ata/ahci-ceva.txt          |   38 +++++++++
> >  drivers/ata/ahci_ceva.c                            |   84 ++++++++++++++------
> >  2 files changed, 99 insertions(+), 23 deletions(-)
> >
> > diff --git a/Documentation/devicetree/bindings/ata/ahci-ceva.txt
> > b/Documentation/devicetree/bindings/ata/ahci-ceva.txt
> > index 7ca8b97..66fcd10 100644
> > --- a/Documentation/devicetree/bindings/ata/ahci-ceva.txt
> > +++ b/Documentation/devicetree/bindings/ata/ahci-ceva.txt
> > @@ -5,6 +5,36 @@ Required properties:
> >    - compatible: Compatibility string. Must be 'ceva,ahci-1v84'.
> >    - clocks: Input clock specifier. Refer to common clock bindings.
> >    - interrupts: Interrupt specifier. Refer to interrupt binding.
> > +    - ceva,p0-cominit-params: OOB timing value for COMINIT parameter for
> port 0.
> > +    - ceva,p1-cominit-params: OOB timing value for COMINIT parameter for
> port 1.
> 
> This doesn't really scale when you have more than 2 ports. Given that you
> know the length for each port, just make it a single property (i.e.
> an array of ports).
> 
The ceva sata controller that we are using  has support for 2 ports only and each port can
Invidually configured to have different timing settings based on frequency. So for ease of
use we  thought of having separate DT parameters for port 0 and 1 .

> > +  			The fields for the above parameter must be as shown
> below:
> > +  			ceva,phy-cominit-params = /bits/ 8 <CIBGMN
> CIBGMX CIBGN CINMP>;
> > +  			CINMP : COMINIT Negate Minimum Period.
> > +  			CIBGN : COMINIT Burst Gap Nominal.
> > +  			CIBGMX: COMINIT Burst Gap Maximum.
> > +  			CIBGMN: COMINIT Burst Gap Minimum.
> > +    - ceva,p0-comwake-params: OOB timing value for COMWAKE
> parameter for port 0.
> > +    - ceva,p1-comwake-params: OOB timing value for COMWAKE
> parameter for port 1.
> > +  			The fields for the above parameter must be as shown
> below:
> > +  			ceva,phy-comwake-params = /bits/ 8 <CWBGMN
> CWBGMX CWBGN CWNMP>;
> > +  			CWBGMN: COMWAKE Burst Gap Minimum.
> > +  			CWBGMX: COMWAKE Burst Gap Maximum.
> > +  			CWBGN: COMWAKE Burst Gap Nominal.
> > +  			CWNMP: COMWAKE Negate Minimum Period.
> > +    - ceva,p0-burst-params: Burst timing value for COM parameter for port
> 0.
> > +    - ceva,p1-burst-params: Burst timing value for COM parameter for port
> 1.
> > +  			The fields for the above parameter must be as shown
> below:
> > +			ceva,phy-burst-params = /bits/ 8 <BMX BNM SFD
> PTST>;
> > +  			BMX: COM Burst Maximum.
> > +  			BNM: COM Burst Nominal.
> > +  			SFD: Signal Failure Detection value.
> > +  			PTST: Partial to Slumber timer value.
> > +    - ceva,p0-retry-params: Retry interval timing value for port 0.
> > +    - ceva,p1-retry-params: Retry interval timing value for port 1.
> > +  			The fields for the above parameter must be as shown
> below:
> > +  			ceva,phy-retry-params = /bits/ 16 <RIT RCT>;
> > +  			RIT:  Retry Interval Timer.
> > +  			RCT:  Rate Change Timer.
> >
> >  Optional properties:
> >    - ceva,broken-gen2: limit to gen1 speed instead of gen2.
> > @@ -17,4 +47,12 @@ Examples:
> >  		interrupts = <0 133 4>;
> >  		clocks = <&clkc SATA_CLK_ID>;
> >  		ceva,broken-gen2;
> > +		ceva,p0-cominit-params = /bits/ 8 <0x0F 0x25 0x18 0x29>;
> > +		ceva,p0-comwake-params = /bits/ 8 <0x04 0x0B 0x08 0x0F>;
> > +		ceva,p0-burst-params = /bits/ 8 <0x0A 0x08 0x4A 0x06>;
> > +		ceva,p0-retry-params = /bits/ 16 <0x0216 0x7F06>;
> > +		ceva,p1-cominit-params = /bits/ 8 <0x0F 0x25 0x18 0x29>;
> > +		ceva,p1-comwake-params = /bits/ 8 <0x04 0x0B 0x08 0x0F>;
> > +		ceva,p1-burst-params = /bits/ 8 <0x0A 0x08 0x4A 0x06>;
> > +		ceva,p1-retry-params = /bits/ 16 <0x0216 0x7F06>;
> >  	};
> > diff --git a/drivers/ata/ahci_ceva.c b/drivers/ata/ahci_ceva.c index
> > 207649d..59de2ca 100644
> > --- a/drivers/ata/ahci_ceva.c
> > +++ b/drivers/ata/ahci_ceva.c
> > @@ -50,21 +50,6 @@
> >  #define PPCFG_PSS_EN	(1 << 29)
> >  #define PPCFG_ESDF_EN	(1 << 31)
> >
> > -#define PP2C_CIBGMN	0x0F
> > -#define PP2C_CIBGMX	(0x25 << 8)
> > -#define PP2C_CIBGN	(0x18 << 16)
> > -#define PP2C_CINMP	(0x29 << 24)
> > -
> > -#define PP3C_CWBGMN	0x04
> > -#define PP3C_CWBGMX	(0x0B << 8)
> > -#define PP3C_CWBGN	(0x08 << 16)
> > -#define PP3C_CWNMP	(0x0F << 24)
> > -
> > -#define PP4C_BMX	0x0a
> > -#define PP4C_BNM	(0x08 << 8)
> > -#define PP4C_SFD	(0x4a << 16)
> > -#define PP4C_PTST	(0x06 << 24)
> > -
> >  #define PP5C_RIT	0x60216
> >  #define PP5C_RCT	(0x7f0 << 20)
> >
> > @@ -87,6 +72,11 @@
> >
> >  struct ceva_ahci_priv {
> >  	struct platform_device *ahci_pdev;
> > +	/* Port Phy2Cfg Register */
> > +	u32 pp2c[NR_PORTS];
> > +	u32 pp3c[NR_PORTS];
> > +	u32 pp4c[NR_PORTS];
> > +	u32 pp5c[NR_PORTS];
> >  	int flags;
> >  };
> >
> > @@ -131,20 +121,16 @@ static void ahci_ceva_setup(struct ahci_host_priv
> *hpriv)
> >  		writel(tmp, mmio + AHCI_VEND_PPCFG);
> >
> >  		/* Phy Control OOB timing parameters COMINIT */
> > -		tmp = PP2C_CIBGMN | PP2C_CIBGMX | PP2C_CIBGN |
> PP2C_CINMP;
> > -		writel(tmp, mmio + AHCI_VEND_PP2C);
> > +		writel(cevapriv->pp2c[i], mmio + AHCI_VEND_PP2C);
> >
> >  		/* Phy Control OOB timing parameters COMWAKE */
> > -		tmp = PP3C_CWBGMN | PP3C_CWBGMX | PP3C_CWBGN |
> PP3C_CWNMP;
> > -		writel(tmp, mmio + AHCI_VEND_PP3C);
> > +		writel(cevapriv->pp3c[i], mmio + AHCI_VEND_PP3C);
> >
> >  		/* Phy Control Burst timing setting */
> > -		tmp = PP4C_BMX | PP4C_BNM | PP4C_SFD | PP4C_PTST;
> > -		writel(tmp, mmio + AHCI_VEND_PP4C);
> > +		writel(cevapriv->pp4c[i], mmio + AHCI_VEND_PP4C);
> >
> >  		/* Rate Change Timer and Retry Interval Timer setting */
> > -		tmp = PP5C_RIT | PP5C_RCT;
> > -		writel(tmp, mmio + AHCI_VEND_PP5C);
> > +		writel(cevapriv->pp5c[i], mmio + AHCI_VEND_PP5C);
> >
> >  		/* Rx Watermark setting  */
> >  		tmp = PTC_RX_WM_VAL | PTC_RSVD;
> > @@ -187,6 +173,58 @@ static int ceva_ahci_probe(struct platform_device
> *pdev)
> >  	if (of_property_read_bool(np, "ceva,broken-gen2"))
> >  		cevapriv->flags = CEVA_FLAG_BROKEN_GEN2;
> >
> > +	/* Read OOB timing value for COMINIT from device-tree */
> > +	if (of_property_read_u8_array(np, "ceva,p0-cominit-params",
> > +					(u8 *)&cevapriv->pp2c[0], 4) < 0) {
> > +		dev_warn(dev, "ceva,p0-cominit-params property not
> defined\n");
> 
> I'd really like to move all the error prints into the of_property_* functions
> instead, but then we'd probably need *_optional variants to avoid spewing
> to the console.

 We will try to implement this later. 

Thanks,
Anurag Kumar V

--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Filesystems]     [Linux SCSI]     [Linux RAID]     [Git]     [Kernel Newbies]     [Linux Newbie]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Samba]     [Device Mapper]

  Powered by Linux