Re: [PATCH v2 00/13] OMAP Serial patches

From: Felipe Balbi
Date: Tue Aug 21 2012 - 11:11:04 EST

Hi,

On Tue, Aug 21, 2012 at 04:01:36PM +0300, Felipe Balbi wrote:
> Hi,
>
> On Tue, Aug 21, 2012 at 03:15:58PM +0300, Felipe Balbi wrote:
> > Hi guys,
> >
> > here's a series of cleanup patches to the OMAP serial
> > driver. A later series could be made re-implementing
> > DMA using the DMA Engine API. Note that for RX DMA
> > we could be using RX Timeout IRQ as a hint that we better
> > use PIO instead ;-)
> >
> > All patches were tested on my pandaboard, but I'd really
> > like to receive Tested-by on other platforms.
> >
> > After this goes in, I'll probably try to get UART wakeup
> > working again and only after that look at DMA.
> >
> > Changes since v1:
> >
> > . improved commit log on patch 9/13 (formerly 10/13)
> > . removed patch 2/13
> > . added a new patch switching from spin_lock_irqsave() to spin_lock and
> > spin_unlock_irqrestore to spin_unlock
> >
> > Retested with my pandaboard, UART continues to work:
> >
> > # grep -i uart /proc/interrupts
> > 106: 124 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 189 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 255 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 321 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 387 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 453 0 GIC OMAP UART2
> > # grep -i uart /proc/interrupts
> > 106: 519 0 GIC OMAP UART2
> >
> > cheers
> >
> > ps: if anyone knows a better test for UART, let me know.
> >
> > for convenience of anyone testing, patches are available on my git tree [1] on
> > branch uart
> >
> > [1] git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb.git uart
>
> I have added one extra patch to this series:
>
> From 6921efdb13dda7af216b331bb35535d4b53f004a Mon Sep 17 00:00:00 2001
> From: Felipe Balbi <balbi@xxxxxx>
> Date: Tue, 21 Aug 2012 15:48:35 +0300
> Subject: [PATCH] serial: omap: drop pm_runtime_irq_safe() usage
>
> pm_runtime_irq_safe() will essentially do an
> unbalanced pm_runtime_get_sync() on our parent
> device, which might cause problems when trying
> to suspend.
>
> In order to prevent that we drop pm_runtime_irq_safe
> usage in exchange for a little performance hit
> when we enter our IRQ handler while still suspended.
>
> If that happens, what we will do is set the irq_pending
> flag, do an asynchronous pm_runtime_get() call and
> return IRQ_HANDLED. When our runtime_resume() callback
> is executed, we check for that flag, and run our
> IRQ handler so that we receive/transmit the pending
> characters.
>
> Signed-off-by: Felipe Balbi <balbi@xxxxxx>
> ---
>
> One extra patch to OMAP UART driver. This seems to be working
> pretty well even after echo mem > /sys/power/state. I can
> see that I can even wake my pandaboard up by sending a character
> through serial:
>
> # echo mem > /sys/power/state
> [ 1335.679260] PM: Syncing filesystems ... done.
> [ 1335.684387] Freezing user space processes ... (elapsed 0.00 seconds) done.
> [ 1335.691741] Freezing remaining freezable tasks ... (elapsed 0.02 seconds) done.
> [ 1335.720886] Suspending console(s) (use no_console_suspend to debug)
>
> [ 1335.734405] PM: suspend of devices complete after 5.492 msecs
> [ 1335.735534] PM: late suspend of devices complete after 1.128 msecs
> [ 1335.737518] PM: noirq suspend of devices complete after 1.952 msecs
> [ 1335.737518] Disabling non-boot CPUs ...
> [ 1335.738525] CPU1: shutdown
> [ 1338.543762] Successfully put all powerdomains to target state
> [ 1338.543853] Enabling non-boot CPUs ...
> [ 1338.545654] CPU1: Booted secondary processor
> [ 1338.546020] CPU1 is up
> [ 1338.547027] PM: noirq resume of devices complete after 0.976 msecs
> [ 1338.548400] PM: early resume of devices complete after 0.762 msecs
> [ 1339.827087] PM: resume of devices complete after 1278.686 msecs
> [ 1339.890960] Restarting tasks ... done.
> # # grep -i uart /proc/interrupts
> 106: 3385 0 GIC OMAP UART2
> # echo mem > /sys/power/state
> [ 1358.015624] PM: Syncing filesystems ... done.
> [ 1358.020812] Freezing user space processes ... (elapsed 0.00 seconds) done.
> [ 1358.028167] Freezing remaining freezable tasks ... (elapsed 0.01 seconds) done.
> [ 1358.055084] Suspending console(s) (use no_console_suspend to debug)
>
> [ 1358.068847] PM: suspend of devices complete after 5.633 msecs
> [ 1358.070007] PM: late suspend of devices complete after 1.126 msecs
> [ 1358.072051] PM: noirq suspend of devices complete after 2.040 msecs
> [ 1358.072051] Disabling non-boot CPUs ...
> [ 1358.073211] CPU1: shutdown
> [ 1359.104156] Successfully put all powerdomains to target state
> [ 1359.104278] Enabling non-boot CPUs ...
> [ 1359.106079] CPU1: Booted secondary processor
> [ 1359.106475] CPU1 is up
> [ 1359.107482] PM: noirq resume of devices complete after 1.004 msecs
> [ 1359.108886] PM: early resume of devices complete after 0.761 msecs
> [ 1360.414794] PM: resume of devices complete after 1305.836 msecs
> [ 1360.478668] Restarting tasks ... done.
> # # grep -i uart /proc/interrupts
> 106: 3511 0 GIC OMAP UART2
>
> arch/arm/plat-omap/include/plat/omap-serial.h | 1 +
> drivers/tty/serial/omap-serial.c | 16 +++++++++++++---
> 2 files changed, 14 insertions(+), 3 deletions(-)
>
> diff --git a/arch/arm/plat-omap/include/plat/omap-serial.h b/arch/arm/plat-omap/include/plat/omap-serial.h
> index 743ac80..52328ca 100644
> --- a/arch/arm/plat-omap/include/plat/omap-serial.h
> +++ b/arch/arm/plat-omap/include/plat/omap-serial.h
> @@ -130,6 +130,7 @@ struct uart_omap_port {
> u32 context_loss_cnt;
> u32 errata;
> u8 wakeups_enabled;
> + unsigned int irq_pending:1;
>
> struct pm_qos_request pm_qos_request;
> u32 latency;
> diff --git a/drivers/tty/serial/omap-serial.c b/drivers/tty/serial/omap-serial.c
> index a79658d..7e237f3 100644
> --- a/drivers/tty/serial/omap-serial.c
> +++ b/drivers/tty/serial/omap-serial.c
> @@ -353,9 +353,13 @@ static inline irqreturn_t serial_omap_irq(int irq, void *dev_id)
> irqreturn_t ret = IRQ_HANDLED;
> int max_count = 256;
>
> - spin_lock(&up->port.lock);
> - pm_runtime_get_sync(up->dev);
> + if (pm_runtime_suspended(up->dev)) {
> + up->irq_pending = true;
> + pm_runtime_get(up->dev);
> + return IRQ_HANDLED;
> + }
>
> + spin_lock(&up->port.lock);
> do {
> iir = serial_in(up, UART_IIR);
> if (iir & UART_IIR_NO_INT) {
> @@ -400,6 +404,7 @@ static inline irqreturn_t serial_omap_irq(int irq, void *dev_id)
> pm_runtime_mark_last_busy(up->dev);
> pm_runtime_put_autosuspend(up->dev);
> up->port_activity = jiffies;
> + up->irq_pending = false;
>
> return ret;
> }
> @@ -1305,7 +1310,6 @@ static int serial_omap_probe(struct platform_device *pdev)
> pm_runtime_set_autosuspend_delay(&pdev->dev,
> omap_up_info->autosuspend_timeout);
>
> - pm_runtime_irq_safe(&pdev->dev);
> pm_runtime_enable(&pdev->dev);
> pm_runtime_get_sync(&pdev->dev);
>
> @@ -1416,6 +1420,9 @@ static int serial_omap_runtime_suspend(struct device *dev)
> if (!up)
> return -EINVAL;
>
> + if (up->irq_pending)
> + return -EBUSY;
> +
> if (!pdata)
> return 0;
>
> @@ -1452,6 +1459,9 @@ static int serial_omap_runtime_resume(struct device *dev)
>
> up->latency = up->calc_latency;
> schedule_work(&up->qos_work);
> +
> + if (up->irq_pending)
> + serial_omap_irq(up->port.irq, up);
> }
>
> return 0;

let's not apply this extra patch. It actually caused a regression which
was pretty difficult to trigger. I have now dropped it from my series.
Sorry for the noise.

--
balbi

Attachment: signature.asc
Description: Digital signature