On Tue, Sep 29, 2009 at 09:54:05PM +0800, Richard Zhao wrote:
> Thanks, Sacha. See comments inline.
>
> On Tue, Sep 29, 2009 at 9:25 PM, Sascha Hauer <s.hauer@xxxxxxxxxxxxxx> wrote:
> > Hi Richard,
> >
> >
> > On Tue, Sep 29, 2009 at 07:41:44PM +0800, Richard Zhao wrote:
> >> After START, wait busy bit to be set and
> >> after STOP, wait busy bit to be clear.
> >>
> >> Disable clock when it's possible to save power.
> >>
> >> Signed-off-by: Richard Zhao <linuxzsc@xxxxxxxxx>
> >> ---
> >> drivers/i2c/busses/i2c-imx.c | 116 ++++++++++++++++++++++++++++++++----------
> >> 1 files changed, 88 insertions(+), 28 deletions(-)
> >>
> >> diff --git a/drivers/i2c/busses/i2c-imx.c b/drivers/i2c/busses/i2c-imx.c
> >> index 4afba3e..59cde70 100644
> >> --- a/drivers/i2c/busses/i2c-imx.c
> >> +++ b/drivers/i2c/busses/i2c-imx.c
> >> @@ -120,19 +120,40 @@ struct imx_i2c_struct {
> >> wait_queue_head_t queue;
> >> unsigned long i2csr;
> >> unsigned int disable_delay;
> >> + unsigned int ifdr; /* IMX_I2C_IFDR */
> >> };
> >>
> >> /** Functions for IMX I2C adapter driver
> >> ***************************************
> >> *******************************************************************************/
> >>
> >> -static int i2c_imx_bus_busy(struct imx_i2c_struct *i2c_imx)
> >> +#ifdef CONFIG_I2C_DEBUG_BUS
> >> +#define reg_dump(i2c_imx) \
> >> +{ \
> >> + printk(KERN_DEBUG "fun %s:%d ", __func__, __LINE__); \
> >> + printk(KERN_DEBUG "IADR %02x IFDR %02x I2CR %02x I2SR %02x\n", \
> >> + readb(i2c_imx->base + IMX_I2C_IADR), \
> >> + readb(i2c_imx->base + IMX_I2C_IFDR), \
> >> + readb(i2c_imx->base + IMX_I2C_I2CR), \
> >> + readb(i2c_imx->base + IMX_I2C_I2SR)); \
> >> +}
> >> +#else
> >> +#define reg_dump(i2c_imx)
> >> +#endif
> >> +
> >
> > Can you please remove this reg_dump? If we really need this it should be
> > in an extra patch and not clutter this one.
> I think it's needed. It helps much to debug and don't effect
> performance by default.
> >
> >> +static int i2c_imx_bus_busy(struct imx_i2c_struct *i2c_imx, int for_busy)
> >> {
> >> unsigned long orig_jiffies = jiffies;
> >> + unsigned int temp;
> >>
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s>\n", __func__);
> >>
> >> /* wait for bus not busy */
> >
> > This comment seems wrong now. This function waits for busy or not busy
> > depending on for_busy.
> Right. Thanks
> >
> >> - while (readb(i2c_imx->base + IMX_I2C_I2SR) & I2SR_IBB) {
> >> + temp = readb(i2c_imx->base + IMX_I2C_I2SR);
> >> + while (1) {
> >> + if (for_busy && (temp & I2SR_IBB))
> >> + break;
> >> + if (!for_busy && !(temp & I2SR_IBB))
> >> + break;
> >> if (signal_pending(current)) {
> >> dev_dbg(&i2c_imx->adapter.dev,
> >> "<%s> I2C Interrupted\n", __func__);
> >> @@ -141,9 +162,11 @@ static int i2c_imx_bus_busy(struct imx_i2c_struct *i2c_imx)
> >> if (time_after(jiffies, orig_jiffies + HZ / 1000)) {
> >> dev_dbg(&i2c_imx->adapter.dev,
> >> "<%s> I2C bus is busy\n", __func__);
> >> + reg_dump(i2c_imx);
> >> return -EIO;
> >> }
> >> schedule();
> >> + temp = readb(i2c_imx->base + IMX_I2C_I2SR);
> >> }
> >>
> >> return 0;
> >> @@ -158,9 +181,11 @@ static int i2c_imx_trx_complete(struct
> >> imx_i2c_struct *i2c_imx)
> >>
> >> if (unlikely(result < 0)) {
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s> result < 0\n", __func__);
> >> + reg_dump(i2c_imx);
> >> return result;
> >> } else if (unlikely(!(i2c_imx->i2csr & I2SR_IIF))) {
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s> Timeout\n", __func__);
> >> + reg_dump(i2c_imx);
> >> return -ETIMEDOUT;
> >> }
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s> TRX complete\n", __func__);
> >> @@ -172,6 +197,7 @@ static int i2c_imx_acked(struct imx_i2c_struct *i2c_imx)
> >> {
> >> if (readb(i2c_imx->base + IMX_I2C_I2SR) & I2SR_RXAK) {
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s> No ACK\n", __func__);
> >> + reg_dump(i2c_imx);
> >> return -EIO; /* No ACK */
> >> }
> >>
> >> @@ -179,20 +205,37 @@ static int i2c_imx_acked(struct imx_i2c_struct *i2c_imx)
> >> return 0;
> >> }
> >>
> >> -static void i2c_imx_start(struct imx_i2c_struct *i2c_imx)
> >> +static int i2c_imx_start(struct imx_i2c_struct *i2c_imx)
> >> {
> >> unsigned int temp = 0;
> >> + int result;
> >>
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s>\n", __func__);
> >>
> >> + clk_enable(i2c_imx->clk);
> >> + writeb(i2c_imx->ifdr, i2c_imx->base + IMX_I2C_IFDR);
> >> /* Enable I2C controller */
> >> + writeb(0, i2c_imx->base + IMX_I2C_I2SR);
> >> writeb(I2CR_IEN, i2c_imx->base + IMX_I2C_I2CR);
> >> +
> >> + result = i2c_imx_bus_busy(i2c_imx, 0);
> >> + if (result) {
> >> + reg_dump(i2c_imx);
> >> + return result;
> >> + }
> >> +
> >> /* Start I2C transaction */
> >> temp = readb(i2c_imx->base + IMX_I2C_I2CR);
> >> temp |= I2CR_MSTA;
> >> writeb(temp, i2c_imx->base + IMX_I2C_I2CR);
> >> + result = i2c_imx_bus_busy(i2c_imx, 1);
> >> + if (result) {
> >> + reg_dump(i2c_imx);
> >> + return result;
> >> + }
> >> temp |= I2CR_IIEN | I2CR_MTX | I2CR_TXAK;
> >> writeb(temp, i2c_imx->base + IMX_I2C_I2CR);
> >> + return result;
> >> }
> >>
> >> static void i2c_imx_stop(struct imx_i2c_struct *i2c_imx)
> >> @@ -202,18 +245,21 @@ static void i2c_imx_stop(struct imx_i2c_struct *i2c_imx)
> >> /* Stop I2C transaction */
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s>\n", __func__);
> >> temp = readb(i2c_imx->base + IMX_I2C_I2CR);
> >> - temp &= ~I2CR_MSTA;
> >> + temp &= ~(I2CR_MSTA | I2CR_MTX);
> >
> > This change seems unrelated. Is it necessary?
> Yes. It's about STOP. It's better clear MTX. We must generate STOP
> before read the last byte of Data register. If not, there will be an
> extra 9bit clock.
> >
> >> writeb(temp, i2c_imx->base + IMX_I2C_I2CR);
> >> - /* setup chip registers to defaults */
> >> - writeb(I2CR_IEN, i2c_imx->base + IMX_I2C_I2CR);
> >> - writeb(0, i2c_imx->base + IMX_I2C_I2SR);
> >> /*
> >> * This delay caused by an i.MXL hardware bug.
> >> * If no (or too short) delay, no "STOP" bit will be generated.
> >> */
> >> udelay(i2c_imx->disable_delay);
> >> +
> >> +
> >
> > double blank line
> Right.
> >
> >> + if (i2c_imx_bus_busy(i2c_imx, 0))
> >> + reg_dump(i2c_imx);
> >> +
> >> /* Disable I2C controller */
> >> writeb(0, i2c_imx->base + IMX_I2C_I2CR);
> >> + clk_disable(i2c_imx->clk);
> >> }
> >>
> >> static void __init i2c_imx_set_clk(struct imx_i2c_struct *i2c_imx,
> >> @@ -233,17 +279,19 @@ static void __init i2c_imx_set_clk(struct
> >> imx_i2c_struct *i2c_imx,
> >> else
> >> for (i = 0; i2c_clk_div[i][0] < div; i++);
> >>
> >> - /* Write divider value to register */
> >> - writeb(i2c_clk_div[i][1], i2c_imx->base + IMX_I2C_IFDR);
> >
> > Why can't we write this value directly anymore...
> clock will be enable/disable in start/stop
> >
> >> -
> >> - /*
> >> - * There dummy delay is calculated.
> >> - * It should be about one I2C clock period long.
> >> - * This delay is used in I2C bus disable function
> >> - * to fix chip hardware bug.
> >> - */
> >> - i2c_imx->disable_delay = (500000U * i2c_clk_div[i][0]
> >> - + (i2c_clk_rate / 2) - 1) / (i2c_clk_rate / 2);
> >> + /* Store divider value */
> >> + i2c_imx->ifdr = i2c_clk_div[i][1];
> >
> > ...but have to store it in a variable instead?
> To save power, because i2c is not aways be accessed, and it's very low
> speed bus.
> >
> >> +
> >> + if (cpu_is_mx1()) {
> >> + /*
> >> + * There dummy delay is calculated.
> >> + * It should be about one I2C clock period long.
> >> + * This delay is used in I2C bus disable function
> >> + * to fix chip hardware bug.
> >> + */
> >> + i2c_imx->disable_delay = (500000U * i2c_clk_div[i][0]
> >> + + (i2c_clk_rate / 2) - 1) / (i2c_clk_rate / 2);
> >> + }
> >>
> >> /* dev_dbg() can't be used, because adapter is not yet registered */
> >> #ifdef CONFIG_I2C_DEBUG_BUS
> >> @@ -344,7 +392,7 @@ static int i2c_imx_read(struct imx_i2c_struct
> >> *i2c_imx, struct i2c_msg *msgs)
> >> dev_dbg(&i2c_imx->adapter.dev,
> >> "<%s> clear MSTA\n", __func__);
> >> temp = readb(i2c_imx->base + IMX_I2C_I2CR);
> >> - temp &= ~I2CR_MSTA;
> >> + temp &= ~(I2CR_MSTA | I2CR_MTX);
> >> writeb(temp, i2c_imx->base + IMX_I2C_I2CR);
> >> } else if (i == (msgs->len - 2)) {
> >> dev_dbg(&i2c_imx->adapter.dev,
> >> @@ -369,14 +417,24 @@ static int i2c_imx_xfer(struct i2c_adapter *adapter,
> >> struct imx_i2c_struct *i2c_imx = i2c_get_adapdata(adapter);
> >>
> >> dev_dbg(&i2c_imx->adapter.dev, "<%s>\n", __func__);
> >> -
> >> +#ifdef CONFIG_I2C_DEBUG_BUS
> >> + for (i = 0; i < num; i++) {
> >> + printk(KERN_DEBUG "msg%d addr %02x RD %d cnt %d d:", i,
> >> + msgs[i].addr, msgs[i].flags & I2C_M_RD, msgs[i].len);
> >> + if (!(msgs[i].flags & I2C_M_RD)) {
> >> + int j;
> >> + for (j = 0; j < msgs[i].len; j++)
> >> + printk("%02x ", msgs[i].buf[j]);
> >> + }
> >> + printk("\n");
> >> + }
> >> +#endif
> >> /* Check if i2c bus is not busy */
> >> - result = i2c_imx_bus_busy(i2c_imx);
> >> - if (result)
> >> - goto fail0;
> >
> > When removing the code you should also remove the comment.
> Right
> >
> >>
> >> /* Start I2C transfer */
> >> - i2c_imx_start(i2c_imx);
> >> + result = i2c_imx_start(i2c_imx);
> >> + if (result)
> >> + goto fail0;
> >>
> >> /* read/write data */
> >> for (i = 0; i < num; i++) {
> >> @@ -386,6 +444,11 @@ static int i2c_imx_xfer(struct i2c_adapter *adapter,
> >> temp = readb(i2c_imx->base + IMX_I2C_I2CR);
> >> temp |= I2CR_RSTA;
> >> writeb(temp, i2c_imx->base + IMX_I2C_I2CR);
> >> + result = i2c_imx_bus_busy(i2c_imx, 1);
> >> + if (result) {
> >> + reg_dump(i2c_imx);
> >> + goto fail0;
> >> + }
> >> }
> >> dev_dbg(&i2c_imx->adapter.dev,
> >> "<%s> transfer message: %d\n", __func__, i);
> >> @@ -442,7 +505,7 @@ static int __init i2c_imx_probe(struct
> >> platform_device *pdev)
> >> int irq;
> >> int ret;
> >>
> >> - dev_dbg(&pdev->dev, "<%s>\n", __func__);
> >> + dev_info(&pdev->dev, "<%s>\n", __func__);
> >
> > no
> Why? I even don't know how many i2c controllers we have, if I didn't
> get sysfs working. And it only print once.
> >
> >>
> >> res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> >> if (!res) {
> >> @@ -500,7 +563,6 @@ static int __init i2c_imx_probe(struct
> >> platform_device *pdev)
> >> dev_err(&pdev->dev, "can't get I2C clock\n");
> >> goto fail3;
> >> }
> >> - clk_enable(i2c_imx->clk);
> >
> > I think it's generally a good idea to start/stop the clocks when needed
> > as you do here. It should be an extra patch though.
> You think I should split the patch to three: IBB, reg_dump, clk dis/en ?
> The idea is good, more easy to revert. but sometimes, I think it make
> small things too complex.
> Anyway, I'll follow your suggestion.
In case of regressions everybody will thank you for doing it.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
