diff mbox series

mmc: tmio: introduce mask for 'always 1' bits

Message ID 20181119131357.28648-1-wsa+renesas@sang-engineering.com (mailing list archive)
State New, archived
Headers show
Series mmc: tmio: introduce mask for 'always 1' bits | expand

Commit Message

Wolfram Sang Nov. 19, 2018, 1:13 p.m. UTC
Some variants (namely Renesas SDHI) have bits in the STATS and IRQ_MASK
registers which are 'always 1' and should be written as such. Introduce
a seperate mask for this and apply it whenever such a register is
written.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
---

This is my response to BSP patches modifying all the register writes directly
(BSP commits 2d339800bbc73d and f73d5eb86097df). This seems much more future
proof to me.

Yamada-san: could you check if your HW has 'always 1' bits, too?

Tested on a R-Car M3-N. No regression and no performance impacts when reading
large files from eMMC or UHS-SD cards.

 drivers/mmc/host/renesas_sdhi_core.c | 1 +
 drivers/mmc/host/tmio_mmc.h          | 5 +++++
 2 files changed, 6 insertions(+)

Comments

Sergei Shtylyov Nov. 19, 2018, 2:44 p.m. UTC | #1
Hello!

On 11/19/2018 04:13 PM, Wolfram Sang wrote:

> Some variants (namely Renesas SDHI) have bits in the STATS and IRQ_MASK
> registers which are 'always 1' and should be written as such. Introduce
> a seperate mask for this and apply it whenever such a register is

   Separate. :-)

> written.
> 
> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
> ---
> 
> This is my response to BSP patches modifying all the register writes directly
> (BSP commits 2d339800bbc73d and f73d5eb86097df). This seems much more future
> proof to me.
> 
> Yamada-san: could you check if your HW has 'always 1' bits, too?
> 
> Tested on a R-Car M3-N. No regression and no performance impacts when reading
> large files from eMMC or UHS-SD cards.

[...]

MBR, Sergei
Simon Horman Nov. 22, 2018, 2:02 p.m. UTC | #2
On Mon, Nov 19, 2018 at 02:13:57PM +0100, Wolfram Sang wrote:
> Some variants (namely Renesas SDHI) have bits in the STATS and IRQ_MASK
> registers which are 'always 1' and should be written as such. Introduce
> a seperate mask for this and apply it whenever such a register is
> written.
> 
> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
> ---
> 
> This is my response to BSP patches modifying all the register writes directly
> (BSP commits 2d339800bbc73d and f73d5eb86097df). This seems much more future
> proof to me.
> 
> Yamada-san: could you check if your HW has 'always 1' bits, too?
> 
> Tested on a R-Car M3-N. No regression and no performance impacts when reading
> large files from eMMC or UHS-SD cards.
> 
>  drivers/mmc/host/renesas_sdhi_core.c | 1 +
>  drivers/mmc/host/tmio_mmc.h          | 5 +++++
>  2 files changed, 6 insertions(+)
> 
> diff --git a/drivers/mmc/host/renesas_sdhi_core.c b/drivers/mmc/host/renesas_sdhi_core.c
> index a049b01206f1..31a351a20dc0 100644
> --- a/drivers/mmc/host/renesas_sdhi_core.c
> +++ b/drivers/mmc/host/renesas_sdhi_core.c
> @@ -722,6 +722,7 @@ int renesas_sdhi_probe(struct platform_device *pdev,
>  		host->ops.card_busy = renesas_sdhi_card_busy;
>  		host->ops.start_signal_voltage_switch =
>  			renesas_sdhi_start_signal_voltage_switch;
> +		host->sdcard_irq_setbit_mask = TMIO_STAT_ALWAYS_SET_27;
>  	}
>  
>  	/* Orginally registers were 16 bit apart, could be 32 or 64 nowadays */
> diff --git a/drivers/mmc/host/tmio_mmc.h b/drivers/mmc/host/tmio_mmc.h
> index 5f6dfb86e43e..c03529e3f01a 100644
> --- a/drivers/mmc/host/tmio_mmc.h
> +++ b/drivers/mmc/host/tmio_mmc.h
> @@ -70,6 +70,7 @@
>  #define TMIO_STAT_DAT0		BIT(23)	/* only known on R-Car so far */
>  #define TMIO_STAT_RXRDY         BIT(24)
>  #define TMIO_STAT_TXRQ          BIT(25)
> +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */

The _27 seems to be odd to include in the name, but I assume it
was the least bad option you could come up with.

Other that that this looks good to me.

Reviewed-by: Simon Horman <horms+renesas@verge.net.au>


>  #define TMIO_STAT_ILL_FUNC      BIT(29) /* only when !TMIO_MMC_HAS_IDLE_WAIT */
>  #define TMIO_STAT_SCLKDIVEN     BIT(29) /* only when TMIO_MMC_HAS_IDLE_WAIT */
>  #define TMIO_STAT_CMD_BUSY      BIT(30)
> @@ -154,6 +155,7 @@ struct tmio_mmc_host {
>  	u32			sdcard_irq_mask;
>  	u32			sdio_irq_mask;
>  	unsigned int		clk_cache;
> +	u32			sdcard_irq_setbit_mask;
>  
>  	spinlock_t		lock;		/* protect host private data */
>  	unsigned long		last_req_ts;
> @@ -268,6 +270,9 @@ static inline void sd_ctrl_write16_rep(struct tmio_mmc_host *host, int addr,
>  static inline void sd_ctrl_write32_as_16_and_16(struct tmio_mmc_host *host,
>  						int addr, u32 val)
>  {
> +	if (addr == CTL_IRQ_MASK || addr == CTL_STATUS)
> +		val |= host->sdcard_irq_setbit_mask;
> +
>  	iowrite16(val & 0xffff, host->ctl + (addr << host->bus_shift));
>  	iowrite16(val >> 16, host->ctl + ((addr + 2) << host->bus_shift));
>  }
> -- 
> 2.11.0
>
Wolfram Sang Nov. 22, 2018, 2:06 p.m. UTC | #3
> > +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
> 
> The _27 seems to be odd to include in the name, but I assume it
> was the least bad option you could come up with.

Yes :) I am open for better suggestions.
Simon Horman Nov. 23, 2018, 12:57 p.m. UTC | #4
On Thu, Nov 22, 2018 at 03:06:59PM +0100, Wolfram Sang wrote:
> 
> > > +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
> > 
> > The _27 seems to be odd to include in the name, but I assume it
> > was the least bad option you could come up with.
> 
> Yes :) I am open for better suggestions.

Simply dropping '_27' would be an improvement in my opinion.
Niklas Söderlund Nov. 23, 2018, 1:03 p.m. UTC | #5
Hi Wolfram,

Thanks for your work.

On 2018-11-23 13:57:32 +0100, Simon Horman wrote:
> On Thu, Nov 22, 2018 at 03:06:59PM +0100, Wolfram Sang wrote:
> > 
> > > > +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
> > > 
> > > The _27 seems to be odd to include in the name, but I assume it
> > > was the least bad option you could come up with.
> > 
> > Yes :) I am open for better suggestions.
> 
> Simply dropping '_27' would be an improvement in my opinion.

I'm slightly agreeing with Simon here to drop the _27. With or without 
that change I think this patch looks good so feel free to add.

Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Wolfram Sang Nov. 23, 2018, 1:15 p.m. UTC | #6
> > > > +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
> > > 
> > > The _27 seems to be odd to include in the name, but I assume it
> > > was the least bad option you could come up with.
> > 
> > Yes :) I am open for better suggestions.
> 
> Simply dropping '_27' would be an improvement in my opinion.

And if we get a second one later? Or make it a mask then?
Simon Horman Nov. 26, 2018, 8:17 a.m. UTC | #7
On Fri, Nov 23, 2018 at 02:15:38PM +0100, Wolfram Sang wrote:
> 
> > > > > +#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
> > > > 
> > > > The _27 seems to be odd to include in the name, but I assume it
> > > > was the least bad option you could come up with.
> > > 
> > > Yes :) I am open for better suggestions.
> > 
> > Simply dropping '_27' would be an improvement in my opinion.
> 
> And if we get a second one later? Or make it a mask then?

Lets not spend too much time debating a better macro name.
But in my opinion we can cross the second bit bridge if
and when we reach it.
Ulf Hansson Dec. 5, 2018, 2:23 p.m. UTC | #8
On Mon, 19 Nov 2018 at 14:14, Wolfram Sang
<wsa+renesas@sang-engineering.com> wrote:
>
> Some variants (namely Renesas SDHI) have bits in the STATS and IRQ_MASK
> registers which are 'always 1' and should be written as such. Introduce
> a seperate mask for this and apply it whenever such a register is
> written.
>
> Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>

Applied for next, thanks!

Kind regards
Uffe

> ---
>
> This is my response to BSP patches modifying all the register writes directly
> (BSP commits 2d339800bbc73d and f73d5eb86097df). This seems much more future
> proof to me.
>
> Yamada-san: could you check if your HW has 'always 1' bits, too?
>
> Tested on a R-Car M3-N. No regression and no performance impacts when reading
> large files from eMMC or UHS-SD cards.
>
>  drivers/mmc/host/renesas_sdhi_core.c | 1 +
>  drivers/mmc/host/tmio_mmc.h          | 5 +++++
>  2 files changed, 6 insertions(+)
>
> diff --git a/drivers/mmc/host/renesas_sdhi_core.c b/drivers/mmc/host/renesas_sdhi_core.c
> index a049b01206f1..31a351a20dc0 100644
> --- a/drivers/mmc/host/renesas_sdhi_core.c
> +++ b/drivers/mmc/host/renesas_sdhi_core.c
> @@ -722,6 +722,7 @@ int renesas_sdhi_probe(struct platform_device *pdev,
>                 host->ops.card_busy = renesas_sdhi_card_busy;
>                 host->ops.start_signal_voltage_switch =
>                         renesas_sdhi_start_signal_voltage_switch;
> +               host->sdcard_irq_setbit_mask = TMIO_STAT_ALWAYS_SET_27;
>         }
>
>         /* Orginally registers were 16 bit apart, could be 32 or 64 nowadays */
> diff --git a/drivers/mmc/host/tmio_mmc.h b/drivers/mmc/host/tmio_mmc.h
> index 5f6dfb86e43e..c03529e3f01a 100644
> --- a/drivers/mmc/host/tmio_mmc.h
> +++ b/drivers/mmc/host/tmio_mmc.h
> @@ -70,6 +70,7 @@
>  #define TMIO_STAT_DAT0         BIT(23) /* only known on R-Car so far */
>  #define TMIO_STAT_RXRDY         BIT(24)
>  #define TMIO_STAT_TXRQ          BIT(25)
> +#define TMIO_STAT_ALWAYS_SET_27        BIT(27) /* only known on R-Car 2+ so far */
>  #define TMIO_STAT_ILL_FUNC      BIT(29) /* only when !TMIO_MMC_HAS_IDLE_WAIT */
>  #define TMIO_STAT_SCLKDIVEN     BIT(29) /* only when TMIO_MMC_HAS_IDLE_WAIT */
>  #define TMIO_STAT_CMD_BUSY      BIT(30)
> @@ -154,6 +155,7 @@ struct tmio_mmc_host {
>         u32                     sdcard_irq_mask;
>         u32                     sdio_irq_mask;
>         unsigned int            clk_cache;
> +       u32                     sdcard_irq_setbit_mask;
>
>         spinlock_t              lock;           /* protect host private data */
>         unsigned long           last_req_ts;
> @@ -268,6 +270,9 @@ static inline void sd_ctrl_write16_rep(struct tmio_mmc_host *host, int addr,
>  static inline void sd_ctrl_write32_as_16_and_16(struct tmio_mmc_host *host,
>                                                 int addr, u32 val)
>  {
> +       if (addr == CTL_IRQ_MASK || addr == CTL_STATUS)
> +               val |= host->sdcard_irq_setbit_mask;
> +
>         iowrite16(val & 0xffff, host->ctl + (addr << host->bus_shift));
>         iowrite16(val >> 16, host->ctl + ((addr + 2) << host->bus_shift));
>  }
> --
> 2.11.0
>
diff mbox series

Patch

diff --git a/drivers/mmc/host/renesas_sdhi_core.c b/drivers/mmc/host/renesas_sdhi_core.c
index a049b01206f1..31a351a20dc0 100644
--- a/drivers/mmc/host/renesas_sdhi_core.c
+++ b/drivers/mmc/host/renesas_sdhi_core.c
@@ -722,6 +722,7 @@  int renesas_sdhi_probe(struct platform_device *pdev,
 		host->ops.card_busy = renesas_sdhi_card_busy;
 		host->ops.start_signal_voltage_switch =
 			renesas_sdhi_start_signal_voltage_switch;
+		host->sdcard_irq_setbit_mask = TMIO_STAT_ALWAYS_SET_27;
 	}
 
 	/* Orginally registers were 16 bit apart, could be 32 or 64 nowadays */
diff --git a/drivers/mmc/host/tmio_mmc.h b/drivers/mmc/host/tmio_mmc.h
index 5f6dfb86e43e..c03529e3f01a 100644
--- a/drivers/mmc/host/tmio_mmc.h
+++ b/drivers/mmc/host/tmio_mmc.h
@@ -70,6 +70,7 @@ 
 #define TMIO_STAT_DAT0		BIT(23)	/* only known on R-Car so far */
 #define TMIO_STAT_RXRDY         BIT(24)
 #define TMIO_STAT_TXRQ          BIT(25)
+#define TMIO_STAT_ALWAYS_SET_27	BIT(27) /* only known on R-Car 2+ so far */
 #define TMIO_STAT_ILL_FUNC      BIT(29) /* only when !TMIO_MMC_HAS_IDLE_WAIT */
 #define TMIO_STAT_SCLKDIVEN     BIT(29) /* only when TMIO_MMC_HAS_IDLE_WAIT */
 #define TMIO_STAT_CMD_BUSY      BIT(30)
@@ -154,6 +155,7 @@  struct tmio_mmc_host {
 	u32			sdcard_irq_mask;
 	u32			sdio_irq_mask;
 	unsigned int		clk_cache;
+	u32			sdcard_irq_setbit_mask;
 
 	spinlock_t		lock;		/* protect host private data */
 	unsigned long		last_req_ts;
@@ -268,6 +270,9 @@  static inline void sd_ctrl_write16_rep(struct tmio_mmc_host *host, int addr,
 static inline void sd_ctrl_write32_as_16_and_16(struct tmio_mmc_host *host,
 						int addr, u32 val)
 {
+	if (addr == CTL_IRQ_MASK || addr == CTL_STATUS)
+		val |= host->sdcard_irq_setbit_mask;
+
 	iowrite16(val & 0xffff, host->ctl + (addr << host->bus_shift));
 	iowrite16(val >> 16, host->ctl + ((addr + 2) << host->bus_shift));
 }