Updated omap_hsmmc SDIO and remuxing patches

We want to get rid of the omap specific platform init code callbacks as they don't play nice with device tree.

Let's convert the context loss check to be based on a register state detection instead.

Hi,

Quoted 2 lines · Tony Lindgren

We want to get rid of the omap specific platform init code callbacks as they don't play nice with device tree.

right, any plans to have similar functionality generically ? Maybe something like:

probe()

{
	...

	/*
	 * tell PM runtime layer that we can handle runtime
	 * power gating of this device in certain conditions.
	 *
	 * Actual power gating will be triggered depending on
	 * PM QoS wakeup requirements.
	 */
	pm_runtime_enable_power_gating(dev);

	...
}

runtime_suspend()

{
	if (pm_runtime_will_power_gate(dev))
		save_context();
}

runtime_resume()

{
	if (pm_runtime_has_lost_power(dev))
		restore_context();
}

???

Quoted 2 lines

Let's convert the context loss check to be based on a register state detection instead.

that might not work always. Specially when you consider that default register values can change on every silicon release. I guess the above might be a bit nicer, although a lot more work :-p

From: Andreas Fenkart <[email protected]>

drivers/mmc/host/omap_hsmmc.c +5 -2

 static void omap_hsmmc_disable_irq(struct omap_hsmmc_host *host)
 {
-	OMAP_HSMMC_WRITE(host->base, ISE, 0);
-	OMAP_HSMMC_WRITE(host->base, IE, 0);
+	u32 irq_mask = 0;
+	unsigned long flags;
+
+	spin_lock_irqsave(&host->irq_lock, flags);
+

From: Andreas Fenkart <[email protected]>

Without functional clock the omap_hsmmc module can't forward SDIO IRQs to the system. This patch reconfigures dat1 line as a gpio while the fclk is off. When the fclk is present it uses the standard SDIO IRQ detection of the module.

The gpio irq is managed via the 'disable_depth' ref counter of the irq subsystem, this driver simply calls enable_irq/disable_irq when needed.

sdio irq sdio irq
unmasked masked
-----------------------------------------

    runtime default  |    1     |   2
    runtime suspend  |    0     |   1

irq disable depth

only when sdio irq is enabled AND the module is idle, the reference count drops to zero and the gpio irq is effectively armed.

Patch was tested on AM335x/Stream800. Test setup was two modules with sdio wifi cards. Modules where connected to a dual-band AP, each module using a different band. One of module was running iperf as server the other as client connecting to the server in a while true loop. Test was running for 4+ weeks. There were about 60 Mio. suspend/resume transitions. Test was shut down regularly.

Cc Andreas Fenkart <[email protected]>

Cc Balaji T K <[email protected]>

Signed-off-by Andreas Fenkart <[email protected]>

Reviewed-by Grant Likely <[email protected]>

[tony at atomide.com: updated and separated out pin muxing]

Signed-off-by Tony Lindgren <[email protected]>

---
 .../devicetree/bindings/mmc/ti-omap-hsmmc.txt      |   42 ++++
 drivers/mmc/host/omap_hsmmc.c                      |  206 +++++++++++++++++++-
 include/linux/platform_data/mmc-omap.h             |    4
 3 files changed, 239 insertions(+), 13 deletions(-)

Documentation/devicetree/bindings/mmc/ti-omap-hsmmc.txt +42 -0

index ed271fc..5a3df37 100644
@@ -20,6 +20,29 @@ ti,dual-volt: boolean, supports dual voltage cards
 ti,non-removable: non-removable slot (like eMMC)
 ti,needs-special-reset: Requires a special softreset sequence
 ti,needs-special-hs-handling: HSMMC IP needs special setting for handling High Speed
+ti,cirq-gpio: When omap_hsmmc module is suspended, its functional
+clock is turned off. Without fclk it can't forward SDIO IRQs to the
+system. For that to happen, it needs to tell the PRCM to restore
+its fclk, which is done through the swakeup line.
+
+                   ------
+                  | PRCM |
+                   ------
+                    | ^
+               fclk | | swakeup
+                    v |
+                  -------               ------
+      <-- IRQ -- | hsmmc | <-- CIRQ -- | card |
+                  -------               ------
+
+The problem is, that on the AM335x family the swakeup line is
+missing, it has not been routed from the module to the PRCM.
+The way to work around this, is to reconfigure the dat1 line as a
+GPIO upon suspend. Beyond this option you also need to set named
+states "default" and "idle "in the .dts file for the pins, using
+pinctrl-single.c. The MMC driver will then then toggle between
+default and idle during the runtime.
+
 
 Example:
        mmc1: mmc at 0x4809c000 {

@@ -31,3 +54,22 @@ Example:
                vmmc-supply = <&vmmc>; /* phandle to regulator node */
                ti,non-removable;
        };
+
+[am335x with with gpio for sdio irq]
+
+       mmc1_cirq_pin: pinmux_cirq_pin {
+               pinctrl-single,pins = <
+                       0x0f8 0x3f      /* MMC0_DAT1 as GPIO2_28 */
+               >;
+       };
+
+       mmc1: mmc at 48060000 {
+               pinctrl-names = "default", "idle";
+               pinctrl-0 = <&mmc1_pins>;
+               pinctrl-1 = <&mmc1_cirq_pin>;
+               ti,cirq-gpio = <&gpio3 28 0>;
+               ti,non-removable;
+               bus-width = <4>;
+               vmmc-supply = <&ldo2_reg>;
+               vmmc_aux-supply = <&vmmc>;
+       };

drivers/mmc/host/omap_hsmmc.c +71 -9

index 478849b..7e28501 100644
@@ -22,6 +22,7 @@
 #include <linux/dmaengine.h>
 #include <linux/seq_file.h>
 #include <linux/interrupt.h>
+#include <linux/irq.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
 #include <linux/platform_device.h>

@@ -102,6 +103,7 @@
 #define TC_EN                  (1 << 1)
 #define BWR_EN                 (1 << 4)
 #define BRR_EN                 (1 << 5)
+#define CIRQ_EN                        (1 << 8)
 #define ERR_EN                 (1 << 15)
 #define CTO_EN                 (1 << 16)
 #define CCRC_EN                        (1 << 17)

@@ -182,9 +184,19 @@ struct omap_hsmmc_host {
        int                     use_reg;
        int                     req_in_progress;
        struct omap_hsmmc_next  next_data;
+       bool                    sdio_irq_en;
+       bool                    active_pinmux;
        struct  omap_mmc_platform_data  *pdata;
 };
 
+static irqreturn_t omap_hsmmc_cirq(int irq, void *dev_id)
+{
+       struct omap_hsmmc_host *host = dev_id;
+
+       mmc_signal_sdio_irq(host->mmc);
+       return IRQ_HANDLED;
+}
+
 static int omap_hsmmc_card_detect(struct device *dev, int slot)
 {
        struct omap_hsmmc_host *host = dev_get_drvdata(dev);

@@ -419,10 +431,31 @@ static int omap_hsmmc_gpio_init(struct omap_mmc_platform_data *pdata)
        } else
                pdata->slots[0].gpio_wp = -EINVAL;
 
+       if (pdata->slots[0].gpio_cirq > 0 &&
+           gpio_is_valid(pdata->slots[0].gpio_cirq)) {
+               pdata->slots[0].sdio_irq =
+                               gpio_to_irq(pdata->slots[0].gpio_cirq);
+
+               ret = gpio_request(pdata->slots[0].gpio_cirq, "sdio_cirq");
+               if (ret)
+                       goto err_free_ro;
+               ret = gpio_direction_input(pdata->slots[0].gpio_cirq);
+               if (ret)
+                       goto err_free_cirq;
+
+       } else {
+               pdata->slots[0].gpio_cirq = -EINVAL;
+       }
+
+
        return 0;
 
+err_free_cirq:
+       gpio_free(pdata->slots[0].gpio_cirq);
+err_free_ro:
+       if (gpio_is_valid(pdata->slots[0].gpio_wp))
 err_free_wp:
-       gpio_free(pdata->slots[0].gpio_wp);
+               gpio_free(pdata->slots[0].gpio_wp);
 err_free_cd:
        if (gpio_is_valid(pdata->slots[0].switch_pin))
 err_free_sp:

@@ -436,6 +469,8 @@ static void omap_hsmmc_gpio_free(struct omap_mmc_platform_data *pdata)
                gpio_free(pdata->slots[0].gpio_wp);
        if (gpio_is_valid(pdata->slots[0].switch_pin))
                gpio_free(pdata->slots[0].switch_pin);
+       if (gpio_is_valid(pdata->slots[0].gpio_cirq))
+               gpio_free(pdata->slots[0].gpio_cirq);
 }
 
 /*

@@ -461,27 +496,46 @@ static void omap_hsmmc_stop_clock(struct omap_hsmmc_host *host)
 static void omap_hsmmc_enable_irq(struct omap_hsmmc_host *host,
                                  struct mmc_command *cmd)
 {
-       unsigned int irq_mask;
+       u32 irq_mask = INT_EN_MASK;
+       unsigned long flags;
 
        if (host->use_dma)
-               irq_mask = INT_EN_MASK & ~(BRR_EN | BWR_EN);
-       else
-               irq_mask = INT_EN_MASK;
+               irq_mask &= ~(BRR_EN | BWR_EN);
 
        /* Disable timeout for erases */
        if (cmd->opcode == MMC_ERASE)
                irq_mask &= ~DTO_EN;
 
+       spin_lock_irqsave(&host->irq_lock, flags);
+
        OMAP_HSMMC_WRITE(host->base, STAT, STAT_CLEAR);
        OMAP_HSMMC_WRITE(host->base, ISE, irq_mask);
+
+       /* latch pending CIRQ, but don't signal */
+       if (host->sdio_irq_en)
+               irq_mask |= CIRQ_EN;
+
        OMAP_HSMMC_WRITE(host->base, IE, irq_mask);
+
+       spin_unlock_irqrestore(&host->irq_lock, flags);
 }
 
 static void omap_hsmmc_disable_irq(struct omap_hsmmc_host *host)
 {
-       OMAP_HSMMC_WRITE(host->base, ISE, 0);
-       OMAP_HSMMC_WRITE(host->base, IE, 0);
+       u32 irq_mask = 0;
+       unsigned long flags;
+
+       spin_lock_irqsave(&host->irq_lock, flags);
+
+       /* no transfer running, need to signal cirq if */
+       if (host->sdio_irq_en)
+               irq_mask |= CIRQ_EN;
+
+       OMAP_HSMMC_WRITE(host->base, ISE, irq_mask);
+       OMAP_HSMMC_WRITE(host->base, IE, irq_mask);
        OMAP_HSMMC_WRITE(host->base, STAT, STAT_CLEAR);
+
+       spin_unlock_irqrestore(&host->irq_lock, flags);
 }
 
 /* Calculate divisor for the given clock frequency */

@@ -1037,8 +1091,12 @@ static irqreturn_t omap_hsmmc_irq(int irq, void *dev_id)
        int status;
 
        status = OMAP_HSMMC_READ(host->base, STAT);
-       while (status & INT_EN_MASK && host->req_in_progress) {
-               omap_hsmmc_do_irq(host, status);
+       while (status & (INT_EN_MASK | CIRQ_EN)) {
+               if (host->req_in_progress)
+                       omap_hsmmc_do_irq(host, status);
+
+               if (status & CIRQ_EN)
+                       mmc_signal_sdio_irq(host->mmc);
 
                /* Flush posted write */
                OMAP_HSMMC_WRITE(host->base, STAT, status);

@@ -1554,6 +1612,51 @@ static void omap_hsmmc_init_card(struct mmc_host *mmc, struct mmc_card *card)
                mmc_slot(host).init_card(card);
 }
 
+static void omap_hsmmc_enable_sdio_irq(struct mmc_host *mmc, int enable)
+{
+       struct omap_hsmmc_host *host = mmc_priv(mmc);
+       u32 irq_mask;

drivers/mmc/host/omap_hsmmc.c +112 -4

+       host->sdio_irq_en = (enable != 0) ? true : false;
+
+       if (host->active_pinmux) { /* register access fails without fclk */
+               irq_mask = OMAP_HSMMC_READ(host->base, ISE);
+               if (enable)
+                       irq_mask |= CIRQ_EN;
+               else
+                       irq_mask &= ~CIRQ_EN;
+               OMAP_HSMMC_WRITE(host->base, IE, irq_mask);
+
+               if (!host->req_in_progress)
+                       OMAP_HSMMC_WRITE(host->base, ISE, irq_mask);
+
+               /*
+                * evtl. need to flush posted write
+                * OMAP_HSMMC_READ(host->base, IE);
+                */
+       }
+
+       if ((mmc_slot(host).sdio_irq)) {
+               if (enable) {
+                       enable_irq(mmc_slot(host).sdio_irq);
+               } else {
+                       /* _nosync, see mmc_signal_sdio_irq */
+                       disable_irq_nosync(mmc_slot(host).sdio_irq);
+               }
+       }
+
+       spin_unlock_irqrestore(&host->irq_lock, flags);
+}
+
 static void omap_hsmmc_conf_bus_power(struct omap_hsmmc_host *host)
 {
        u32 hctl, capa, value;

@@ -1606,7 +1709,7 @@ static const struct mmc_host_ops omap_hsmmc_ops = {
        .get_cd = omap_hsmmc_get_cd,
        .get_ro = omap_hsmmc_get_ro,
        .init_card = omap_hsmmc_init_card,
-       /* NYET -- enable_sdio_irq */
+       .enable_sdio_irq = omap_hsmmc_enable_sdio_irq,
 };
 
 #ifdef CONFIG_DEBUG_FS

@@ -1710,6 +1813,7 @@ static struct omap_mmc_platform_data *of_get_hsmmc_pdata(struct device *dev)
        pdata->nr_slots = 1;
        pdata->slots[0].switch_pin = cd_gpio;
        pdata->slots[0].gpio_wp = wp_gpio;
+       pdata->slots[0].gpio_cirq = of_get_named_gpio(np, "ti,cirq-gpio", 0);
 
        if (of_find_property(np, "ti,non-removable", NULL)) {
                pdata->slots[0].nonremovable = true;

@@ -1802,6 +1906,8 @@ static int omap_hsmmc_probe(struct platform_device *pdev)
        host->dma_ch    = -1;
        host->irq       = irq;
        host->slot_id   = 0;
+       host->sdio_irq_en = false;
+       host->active_pinmux = true;
        host->mapbase   = res->start + pdata->reg_offset;
        host->base      = ioremap(host->mapbase, SZ_4K);
        host->power_mode = MMC_POWER_OFF;

@@ -1955,6 +2061,29 @@ static int omap_hsmmc_probe(struct platform_device *pdev)
                pdata->resume = omap_hsmmc_resume_cdirq;
        }
 
+       if ((mmc_slot(host).sdio_irq)) {
+               /* prevent auto-enabling of IRQ */
+               irq_set_status_flags(mmc_slot(host).sdio_irq, IRQ_NOAUTOEN);
+               ret = request_irq(mmc_slot(host).sdio_irq, omap_hsmmc_cirq,
+                                 IRQF_TRIGGER_LOW | IRQF_ONESHOT,
+                                 mmc_hostname(mmc), host);
+               if (ret) {
+                       dev_dbg(mmc_dev(host->mmc),
+                               "Unable to grab MMC SDIO IRQ\n");
+                       goto err_irq_sdio;
+               }
+
+               /*
+                * sdio_irq is managed with ref count
+                * - omap_hsmmc_enable_sdio_irq will +1/-1
+                * - pm_suspend/pm_resume will +1/-1
+                * only when sdio irq is enabled AND module will go to runtime
+                * suspend the ref count will drop to zero and the irq is
+                * effectively enabled. starting with ref count equal 2
+                */
+               disable_irq(mmc_slot(host).sdio_irq);
+       }
+
        omap_hsmmc_disable_irq(host);
 
        pinctrl = devm_pinctrl_get_select_default(&pdev->dev);

@@ -1962,6 +2091,19 @@ static int omap_hsmmc_probe(struct platform_device *pdev)
                dev_warn(&pdev->dev,
                        "pins are not configured from the driver\n");
 
+       /*
+        * For now, only support SDIO interrupt if we are doing
+        * muxing of dat1 when booted with DT. This is because the
+        * supposedly the wake-up events for CTPL don't work from deeper
+        * idle states. And we don't want to add new legacy mux platform
+        * init code callbacks any longer as we are moving to DT based
+        * booting anyways.
+        */
+       if (match) {
+               if (!IS_ERR(pinctrl) && mmc_slot(host).sdio_irq)
+                       mmc->caps |= MMC_CAP_SDIO_IRQ;
+       }
+
        omap_hsmmc_protect_card(host);
 
        mmc_add_host(mmc);

@@ -1986,6 +2128,9 @@ static int omap_hsmmc_probe(struct platform_device *pdev)
 
 err_slot_name:
        mmc_remove_host(mmc);
+       if ((mmc_slot(host).sdio_irq))
+               free_irq(mmc_slot(host).sdio_irq, host);
+err_irq_sdio:
        free_irq(mmc_slot(host).card_detect_irq, host);
 err_irq_cd:
        if (host->use_reg)

@@ -2032,9 +2177,10 @@ static int omap_hsmmc_remove(struct platform_device *pdev)
        if (host->pdata->cleanup)
                host->pdata->cleanup(&pdev->dev);
        free_irq(host->irq, host);
+       if ((mmc_slot(host).sdio_irq))
+               free_irq(mmc_slot(host).sdio_irq, host);
        if (mmc_slot(host).card_detect_irq)
                free_irq(mmc_slot(host).card_detect_irq, host);
-
        if (host->tx_chan)
                dma_release_channel(host->tx_chan);
        if (host->rx_chan)

@@ -2157,23 +2303,57 @@ static int omap_hsmmc_resume(struct device *dev)
 static int omap_hsmmc_runtime_suspend(struct device *dev)
 {
        struct omap_hsmmc_host *host;
+       struct mmc_host *mmc;
+       unsigned long flags;
+       int ret = 0;
 
        host = platform_get_drvdata(to_platform_device(dev));
+       mmc = host->mmc;
        omap_hsmmc_context_save(host);
        dev_dbg(dev, "disabled\n");
 
-       return 0;
+       if (mmc->caps & MMC_CAP_SDIO_IRQ) {
+               spin_lock_irqsave(&host->irq_lock, flags);
+               host->active_pinmux = false;
+               OMAP_HSMMC_WRITE(host->base, ISE, 0);
+               OMAP_HSMMC_WRITE(host->base, IE, 0);
+               OMAP_HSMMC_WRITE(host->base, STAT, STAT_CLEAR);
+               spin_unlock_irqrestore(&host->irq_lock, flags);
+
+               if (mmc_slot(host).sdio_irq)
+                       enable_irq(mmc_slot(host).sdio_irq);
+       }
+
+       return ret;
 }
 
 static int omap_hsmmc_runtime_resume(struct device *dev)
 {
        struct omap_hsmmc_host *host;
+       struct mmc_host *mmc;
+       unsigned long flags;
+       int ret = 0;
 
        host = platform_get_drvdata(to_platform_device(dev));
+       mmc = host->mmc;
        omap_hsmmc_context_restore(host);
        dev_dbg(dev, "enabled\n");
 
-       return 0;
+       if (mmc->caps & MMC_CAP_SDIO_IRQ) {
+               if (mmc_slot(host).sdio_irq)
+                       disable_irq(mmc_slot(host).sdio_irq);
+
+               spin_lock_irqsave(&host->irq_lock, flags);
+               host->active_pinmux = true;
+
+               if (host->sdio_irq_en) {
+                       OMAP_HSMMC_WRITE(host->base, STAT, STAT_CLEAR);
+                       OMAP_HSMMC_WRITE(host->base, ISE, CIRQ_EN);
+                       OMAP_HSMMC_WRITE(host->base, IE, CIRQ_EN);
+               }
+               spin_unlock_irqrestore(&host->irq_lock, flags);
+       }
+       return ret;
 }
 
 static struct dev_pm_ops omap_hsmmc_dev_pm_ops = {

include/linux/platform_data/mmc-omap.h +4 -1

index 2bf1b30..fd5fff5 100644
@@ -115,6 +115,7 @@ struct omap_mmc_platform_data {
 
                int switch_pin;                 /* gpio (card detect) */
                int gpio_wp;                    /* gpio (write protect) */
+               int gpio_cirq;                  /* gpio (card irq) */
 
                int (*set_bus_mode)(struct device *dev, int slot, int bus_mode);
                int (*set_power)(struct device *dev, int slot,

@@ -145,6 +146,9 @@ struct omap_mmc_platform_data {
                int card_detect_irq;
                int (*card_detect)(struct device *dev, int slot);
 
+               /* SDIO IRQs */
+               int sdio_irq;
+
                unsigned int ban_openended:1;
 
        } slots[OMAP_MMC_MAX_SLOTS];
 
--

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

Quoted 7 lines · Tony Lindgren

From: Andreas Fenkart <[email protected]>

drivers/mmc/host/omap_hsmmc.c +4 -0

+static void omap_hsmmc_enable_sdio_irq(struct mmc_host *mmc, int enable)
+{
+       struct omap_hsmmc_host *host = mmc_priv(mmc);
+       u32 irq_mask;

drivers/mmc/host/omap_hsmmc.c static void omap_hsmmc_init_card(struct mmc_host *mmc, struct mmc_card *card) · line 1628

Quoted 10 lines

drivers/mmc/host/omap_hsmmc.c +10 -0

+       unsigned long flags;
+
+       spin_lock_irqsave(&host->irq_lock, flags);
+
+       if (host->sdio_irq_en == enable) {
+               dev_dbg(host->dev, "en/disable:%d already set", enable);
+               spin_unlock_irqrestore(&host->irq_lock, flags);
+               return;
+       }
+

Hi Tony/Andreas,

I belive a "pm_runtime_get_sync" would be needed here, outside the spinlock ofcourse. Before returning from this function, obviusly return the references by a pm_runtime_put* in some form.

Then you will be able to remove the "active_pinmux" variable entirely, since you know the runtime callbacks is the only place were you need to handle the gpio irq enable|disable.

From: Andreas Fenkart <[email protected]>

Without functional clock the omap_hsmmc module can't forward SDIO IRQs to the system. This patch reconfigures dat1 line as a gpio while the fclk is off. When the fclk is present it uses the standard SDIO IRQ detection of the module.

The gpio irq is managed via the 'disable_depth' ref counter of the irq subsystem, this driver simply calls enable_irq/disable_irq when needed.

sdio irq sdio irq
unmasked masked
-----------------------------------------

    runtime default  |    1     |   2
    runtime suspend  |    0     |   1

irq disable depth

only when sdio irq is enabled AND the module is idle, the reference count drops to zero and the gpio irq is effectively armed.

Patch was tested on AM335x/Stream800. Test setup was two modules with sdio wifi cards. Modules where connected to a dual-band AP, each module using a different band. One of module was running iperf as server the other as client connecting to the server in a while true loop. Test was running for 4+ weeks. There were about 60 Mio. suspend/resume transitions. Test was shut down regularly.

Cc Andreas Fenkart <[email protected]>

Cc Balaji T K <[email protected]>

Signed-off-by Andreas Fenkart <[email protected]>

Reviewed-by Grant Likely <[email protected]>

[tony at atomide.com: updated and separated out pin muxing]

Signed-off-by Tony Lindgren <[email protected]>

On some omaps we need to remux MMC pins for PM, and for some omaps we need to remux the SDIO IRQ pin.

Based on an earlier patch by Andreas Fenkart <[email protected]>.

drivers/mmc/host/omap_hsmmc.c +26 -0

+       host->pinctrl = devm_pinctrl_get(host->dev);
+       if (IS_ERR(host->pinctrl)) {
+               dev_dbg(host->dev, "no pinctrl handle\n");
+               ret = 0;
+               goto out;
+       }
+
+       host->fixed = pinctrl_lookup_state(host->pinctrl,
+                                          PINCTRL_STATE_DEFAULT);
+       if (IS_ERR(host->fixed)) {
+               dev_dbg(host->dev,
+                        "pins are not configured from the driver\n");
+               host->fixed = NULL;
+               ret = 0;
+               goto out;
+       }
+
+       ret = pinctrl_select_state(host->pinctrl, host->fixed);
+       if (ret < 0)
+               goto err;
+
+       /* For most cases we don't have wake-ups, and exit after this */
+       host->active = pinctrl_lookup_state(host->pinctrl, "active");
+       if (IS_ERR(host->active)) {
+               ret = PTR_ERR(host->active);
+               host->active = NULL;

You can use the new infrastructure to make the core select:

pinctrl_pm_select_default_state(host->dev);
pinctrl_pm_select_idle_state(host->dev);

What is the semantic difference between "default" and "active"?

If this is something very generic that a lot of platforms will want to have, why not add it to include/linux/pinctrl/pinctrl-state.h and augment the core to cache and handle this too?

However in this case I *suspect* that what you really want to do it to rename the state called "default" to "sleep" (it appears the default state is sleepy) and then rename the "active" state to "default" (as this is the defined semantic meaning of "default" from <linux/pinctrl/pinctrl-state.h>.

But maybe I'm not quite getting the subtle difference between "default" and "active" here so enlighten me.

We only should remux the pins that need remuxing as that's done every time hitting idle. So I think we should have the following default groups:

default Static pins that don't change, no need to remux
configured in consumer driver probe like we already do

active Optional dynamic pins remuxed for runtime, can be
configured and selected in consumer driver probe. The consumer driver may also want to select this state in PM runtime resume.

idle Optional dynamic pins remuxed for idle. The consumer
driver may also want to select this state in PM runtime suspend depending on device_can_wakeup() and driver specific needs.

Quoted 5 lines

However in this case I *suspect* that what you really want to do it to rename the state called "default" to "sleep" (it appears the default state is sleepy) and then rename the "active" state to "default" (as this is the defined semantic meaning of "default" from <linux/pinctrl/pinctrl-state.h>.

The idle state above could also be called sleep instead of idle if you prefer that.

I think the confusion is caused by the fact that we need three mux groups, not just two :) The toggling between active and idle is the hotpath as that can potentially happen for multiple drivers every time we enter and exit idle.

Quoted 13 lines · Tony Lindgren

We only should remux the pins that need remuxing as that's done every time hitting idle. So I think we should have the following default groups:

default Static pins that don't change, no need to remux
configured in consumer driver probe like we already do

active Optional dynamic pins remuxed for runtime, can be
configured and selected in consumer driver probe. The consumer driver may also want to select this state in PM runtime resume.

Quoted 4 lines

idle Optional dynamic pins remuxed for idle. The consumer
driver may also want to select this state in PM runtime suspend depending on device_can_wakeup() and driver specific needs.

The one thing I don't understand is why a driver would select the active state in probe(), unless it's a driver that does not support runtime PM. (But maybe that's what you mean.)

Compare this to <linus/pinctrl/pinctrl-state.h>:

/**
* @PINCTRL_STATE_DEFAULT: the state the pinctrl handle shall be put

 *      into as default, usually this means the pins are up and ready to
 *      be used by the device driver. This state is commonly used by
 *      hogs to configure muxing and pins at boot, and also as a state
 *      to go into when returning from sleep and idle in
 *      .pm_runtime_resume() or ordinary .resume() for example.
 * @PINCTRL_STATE_IDLE: the state the pinctrl handle shall be put into
 *      when the pins are idle. This is a state where the system is relaxed
 *      but not fully sleeping - some power may be on but clocks gated for
 *      example. Could typically be set from a pm_runtime_suspend() or
 *      pm_runtime_idle() operation.
 * @PINCTRL_STATE_SLEEP: the state the pinctrl handle shall be put into
 *      when the pins are sleeping. This is a state where the system is in
 *      its lowest sleep state. Could typically be set from an
 *      ordinary .suspend() function.

*/
#define PINCTRL_STATE_DEFAULT "default"
#define PINCTRL_STATE_IDLE "idle"
#define PINCTRL_STATE_SLEEP "sleep"

The way I currently use these in e.g.
drivers/spi/spi-pl022 is:

probe:
-> default

runtime_suspend:
-> idle

runtime_resume:
-> default

suspend:
-> sleep

resume:
-> default
-> idle

Notice that we go to default then idle on probe and runtime resume. This is because the idle state is optional (as is the sleep state).

So I guess if we should extend this terminology to match what you are using for the OMAP it would rather be like this:

probe:
-> default

runtime_suspend:
-> idle

runtime_resume:
-> default
-> active

suspend:
-> sleep

resume:
-> default
-> idle

Just one more optional "active" state in runtime resume. Correct?

If we can agree on this I will add the active state to the state table and add a container in the core for this as well as pinctrl_pm_select_active_state() so we can skip all the pointless boilerplate also in the OMAP drivers, plus increase the readability and portability quite a bit.

Quoted 8 lines

Quoted 5 lines

However in this case I *suspect* that what you really want to do it to rename the state called "default" to "sleep" (it appears the default state is sleepy) and then rename the "active" state to "default" (as this is the defined semantic meaning of "default" from <linux/pinctrl/pinctrl-state.h>.

The idle state above could also be called sleep instead of idle if you prefer that.

No I think we should reserve that name for the pin state associated with suspend(). Let's leave it like this.

Quoted 4 lines

I think the confusion is caused by the fact that we need three mux groups, not just two :) The toggling between active and idle is the hotpath as that can potentially happen for multiple drivers every time we enter and exit idle.

Actually we have the same thing, it's just that our "default" and "active" are the same thing. But it seems we need to add your granularity to this.

Quoted 5 lines · Linus Walleij

If we can agree on this I will add the active state to the state table and add a container in the core for this as well as pinctrl_pm_select_active_state() so we can skip all the pointless boilerplate also in the OMAP drivers, plus increase the readability and portability quite a bit.

Sounds good to me as long as we don't always need to select the default pins over and over in PM runtime_resume.