Re: [PATCHv3 13/15] remoteproc/omap: add support for runtime auto-suspend/resume

[Suman Anna]

On 12/20/19 5:24 AM, Tero Kristo wrote:
> On 20/12/2019 05:24, Suman Anna wrote:
>> Hi Tero,
>>
>> On 12/19/19 5:43 PM, Mathieu Poirier wrote:
>>> On Fri, Dec 13, 2019 at 02:55:35PM +0200, Tero Kristo wrote:
>>>> From: Suman Anna <s-anna@xxxxxx>
>>>>
>>>> This patch enhances the PM support in the OMAP remoteproc driver to
>>>> support the runtime auto-suspend. A remoteproc may not be required to
>>>> be running all the time, and typically will need to be active only
>>>> during certain usecases. As such, to save power, it should be turned
>>>> off during potential long periods of inactivity between usecases.
>>>> This suspend and resume of the device is a relatively heavy process
>>>> in terms of latencies, so a remoteproc should be suspended only after
>>>> a certain period of prolonged inactivity. The OMAP remoteproc driver
>>>> leverages the runtime pm framework's auto_suspend feature to accomplish
>>>> this functionality. This feature is automatically enabled when a remote
>>>> processor has successfully booted. The 'autosuspend_delay_ms' for each
>>>> device dictates the inactivity period/time to wait for before
>>>> suspending the device.
>>>>
>>>> The runtime auto-suspend design relies on marking the last busy time
>>>> on every communication (virtqueue kick) to and from the remote
>>>> processor.
>>>> When there has been no activity for 'autosuspend_delay_ms' time, the
>>>> runtime PM framework invokes the driver's runtime pm suspend callback
>>>> to suspend the device. The remote processor will be woken up on the
>>>> initiation of the next communication message through the runtime pm
>>>> resume callback. The current auto-suspend design also allows a remote
>>>> processor to deny a auto-suspend attempt, if it wishes to, by sending a
>>>> NACK response to the initial suspend request message sent to the remote
>>>> processor as part of the suspend process. The auto-suspend request is
>>>> also only attempted if the remote processor is idled and in standby at
>>>> the time of inactivity timer expiry. This choice is made to avoid
>>>> unnecessary messaging, and the auto-suspend is simply rescheduled to
>>>> be attempted again after a further lapse of autosuspend_delay_ms.
>>>>
>>>> The runtime pm callbacks functionality in this patch reuses most of the
>>>> core logic from the suspend/resume support code, and make use of an
>>>> additional auto_suspend flag to differentiate the logic in common code
>>>> from system suspend. The system suspend/resume sequences are also
>>>> updated
>>>> to reflect the proper pm_runtime statuses, and also to really perform a
>>>> suspend/resume only if the remoteproc has not been auto-suspended at
>>>> the
>>>> time of request. The remote processor is left in suspended state on a
>>>> system resume if it has been auto-suspended before, and will be
>>>> woken up
>>>> only when a usecase needs to run. The other significant change in this
>>>> patch is to reset the remoteproc device's pm_domain so as to avoid
>>>> conflicts with the ordering sequences in the device pm_domain's runtime
>>>> callbacks and the reset management and clock management implemented
>>>> within the runtime callbacks in the driver.
>>>>
>>>> The OMAP remoteproc driver currently uses a default value of 10 seconds
>>>> for all OMAP remoteprocs, and a different value can be chosen either by
>>>> choosing a positive value for the 'autosuspend_delay' in the device's
>>>> omap_rproc_fw_data in the driver match data or by updating the
>>>> 'autosuspend_delay_ms' field at runtime through the sysfs interface.
>>>> ÂÂÂÂ Eg: To use 25 seconds for IPU2 on DRA7xx,
>>>> ÂÂÂÂÂÂ echo 25000 >
>>>> /sys/bus/platform/devices/55020000.ipu/power/autosuspend_delay_ms
>>>>
>>>> The runtime suspend feature can also be similarly enabled or
>>>> disabled by
>>>> writing 'auto' or 'on' to the device's 'control' power field. The
>>>> default
>>>> is enabled.
>>>> ÂÂÂÂ Eg: To disable auto-suspend for IPU2 on DRA7xx SoC,
>>>> ÂÂÂÂÂÂ echo on > /sys/bus/platform/devices/55020000.ipu/power/control
>>>>
>>>> Signed-off-by: Suman Anna <s-anna@xxxxxx>
>>>> [t-kristo@xxxxxx: converted to use ti-sysc instead of hwmod]
>>>> Signed-off-by: Tero Kristo <t-kristo@xxxxxx>
>>>> ---
>>>> Â drivers/remoteproc/omap_remoteproc.c | 220
>>>> ++++++++++++++++++++++++++-
>>>> Â 1 file changed, 214 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/drivers/remoteproc/omap_remoteproc.c
>>>> b/drivers/remoteproc/omap_remoteproc.c
>>>> index 0a9b9f7d20da..463d6f60947a 100644
>>>> --- a/drivers/remoteproc/omap_remoteproc.c
>>>> +++ b/drivers/remoteproc/omap_remoteproc.c
>>>> @@ -20,6 +20,7 @@
>>>> Â #include <linux/of_device.h>
>>>> Â #include <linux/of_reserved_mem.h>
>>>> Â #include <linux/platform_device.h>
>>>> +#include <linux/pm_runtime.h>
>>>> Â #include <linux/dma-mapping.h>
>>>> Â #include <linux/remoteproc.h>
>>>> Â #include <linux/mailbox_client.h>
>>>> @@ -37,6 +38,9 @@
>>>> Â #include "omap_remoteproc.h"
>>>> Â #include "remoteproc_internal.h"
>>>> Â +/* default auto-suspend delay (ms) */
>>>> +#define DEFAULT_AUTOSUSPEND_DELAYÂÂÂÂÂÂÂ 10000
>>>> +
>>>> Â /**
>>>> ÂÂ * struct omap_rproc_boot_data - boot data structure for the DSP
>>>> omap rprocs
>>>> ÂÂ * @syscon: regmap handle for the system control configuration module
>>>> @@ -83,6 +87,8 @@ struct omap_rproc_timer {
>>>> ÂÂ * @num_mems: number of internal memory regions
>>>> ÂÂ * @num_timers: number of rproc timer(s)
>>>> ÂÂ * @timers: timer(s) info used by rproc
>>>> + * @autosuspend_delay: auto-suspend delay value to be used for
>>>> runtime pm
>>>> + * @need_resume: if true a resume is needed in the system resume
>>>> callback
>>>> ÂÂ * @rproc: rproc handle
>>>> ÂÂ * @reset: reset handle
>>>> ÂÂ * @pm_comp: completion primitive to sync for suspend response
>>>> @@ -97,6 +103,8 @@ struct omap_rproc {
>>>> ÂÂÂÂÂ int num_mems;
>>>> ÂÂÂÂÂ int num_timers;
>>>> ÂÂÂÂÂ struct omap_rproc_timer *timers;
>>>> +ÂÂÂ int autosuspend_delay;
>>>> +ÂÂÂ bool need_resume;
>>>> ÂÂÂÂÂ struct rproc *rproc;
>>>> ÂÂÂÂÂ struct reset_control *reset;
>>>> ÂÂÂÂÂ struct completion pm_comp;
>>>> @@ -111,6 +119,7 @@ struct omap_rproc {
>>>> ÂÂ * @boot_reg_shift: bit shift for the boot register mask
>>>> ÂÂ * @mem_names: memory names for this remote processor
>>>> ÂÂ * @dev_addrs: device addresses corresponding to the memory names
>>>> + * @autosuspend_delay: custom auto-suspend delay value in milliseconds
>>>> ÂÂ */
>>>> Â struct omap_rproc_dev_data {
>>>> ÂÂÂÂÂ const char *device_name;
>>>> @@ -118,6 +127,7 @@ struct omap_rproc_dev_data {
>>>> ÂÂÂÂÂ int boot_reg_shift;
>>>> ÂÂÂÂÂ const char * const *mem_names;
>>>> ÂÂÂÂÂ const u32 *dev_addrs;
>>>> +ÂÂÂ int autosuspend_delay;
>>>> Â };
>>>> Â Â /**
>>>> @@ -384,11 +394,23 @@ static void omap_rproc_kick(struct rproc
>>>> *rproc, int vqid)
>>>> ÂÂÂÂÂ struct device *dev = rproc->dev.parent;
>>>> ÂÂÂÂÂ int ret;
>>>> Â +ÂÂÂ /* wake up the rproc before kicking it */
>>>> +ÂÂÂ ret = pm_runtime_get_sync(dev);
>>>> +ÂÂÂ if (WARN_ON(ret < 0)) {
>>>> +ÂÂÂÂÂÂÂ dev_err(dev, "pm_runtime_get_sync() failed during kick, ret
>>>> = %d\n",
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ ret);
>>>> +ÂÂÂÂÂÂÂ pm_runtime_put_noidle(dev);
>>>> +ÂÂÂÂÂÂÂ return;
>>>> +ÂÂÂ }
>>>> +
>>>> ÂÂÂÂÂ /* send the index of the triggered virtqueue in the mailbox
>>>> payload */
>>>> ÂÂÂÂÂ ret = mbox_send_message(oproc->mbox, (void *)vqid);
>>>> ÂÂÂÂÂ if (ret < 0)
>>>> ÂÂÂÂÂÂÂÂÂ dev_err(dev, "failed to send mailbox message, status = %d\n",
>>>> ÂÂÂÂÂÂÂÂÂÂÂÂÂ ret);
>>>> +
>>>> +ÂÂÂ pm_runtime_mark_last_busy(dev);
>>>> +ÂÂÂ pm_runtime_put_autosuspend(dev);
>>>> Â }
>>>> Â Â /**
>>>> @@ -473,6 +495,19 @@ static int omap_rproc_start(struct rproc *rproc)
>>>> ÂÂÂÂÂÂÂÂÂ goto put_mbox;
>>>> ÂÂÂÂÂ }
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * remote processor is up, so update the runtime pm status and
>>>> +ÂÂÂÂ * enable the auto-suspend. The device usage count is incremented
>>>> +ÂÂÂÂ * manually for balancing it for auto-suspend
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ pm_runtime_set_active(dev);
>>>> +ÂÂÂ pm_runtime_set_autosuspend_delay(dev, oproc->autosuspend_delay);
>>>> +ÂÂÂ pm_runtime_use_autosuspend(dev);
>>>> +ÂÂÂ pm_runtime_get(dev);
>>>> +ÂÂÂ pm_runtime_enable(dev);
>>>> +ÂÂÂ pm_runtime_mark_last_busy(dev);
>>>> +ÂÂÂ pm_runtime_put_autosuspend(dev);
>>>> +
>>>> ÂÂÂÂÂ reset_control_deassert(oproc->reset);
>>
>> I see that you have flipped the reset call and all the pm_runtime calls
>> (w.r.t my original code sequence) and pm_runtime_get instead of
>> pm_runtime_get_noresume(). Is there a reason for it? What is the backend
>> that gets exercised with pm_runtime?
> 
> PM runtime nowadays exercises ti-sysc, and we can't deassert reset
> before we enable the clocks for the device. Thus, these are flipped.

But the remoteprocs are not being added under a ti-sysc node from what I
have seen so far, only the MMUs. Even if that were the case, you would
have to actually revise the above pm_runtime call sequences, as
replacing get_noresume() with get() are not equivalent.

> 
> In the old incarnation of this code, PM runtime just exercised the
> hacked up functionality implemented for omap rproc alone.

Ha ha, it hadn't been easy to come up with that carefully crafted to
deal with default hwmod code around reset issues. The logic enabled the
clocking and reset independently while powering up and managed the
pm_runtime status appropriately originally.

> 
>>
>>>> Â ÂÂÂÂÂ return 0;
>>>> @@ -485,9 +520,26 @@ static int omap_rproc_start(struct rproc *rproc)
>>>> Â /* power off the remote processor */
>>>> Â static int omap_rproc_stop(struct rproc *rproc)
>>>> Â {
>>>> +ÂÂÂ struct device *dev = rproc->dev.parent;
>>>> ÂÂÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> ÂÂÂÂÂ int ret;
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * cancel any possible scheduled runtime suspend by incrementing
>>>> +ÂÂÂÂ * the device usage count, and resuming the device. The remoteproc
>>>> +ÂÂÂÂ * also needs to be woken up if suspended, to avoid the remoteproc
>>>> +ÂÂÂÂ * OS to continue to remember any context that it has saved, and
>>>> +ÂÂÂÂ * avoid potential issues in misindentifying a subsequent device
>>>> +ÂÂÂÂ * reboot as a power restore boot
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ ret = pm_runtime_get_sync(dev);
>>>> +ÂÂÂ if (ret < 0) {
>>>> +ÂÂÂÂÂÂÂ pm_runtime_put_noidle(dev);
>>>> +ÂÂÂÂÂÂÂ return ret;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ pm_runtime_put_sync(dev);
>>>> +
>>
>> I didn't have this call either. And get_sync() followed by put_sync() is
>> essentially a no-op. Am I missing something here?
> 
> Hmm right, but you did have the device_shutdown() here, somehow this
> appears to simulate that, but you are right, it is just a no-op. I don't
> know if this code has actually ever done anything useful. Let me
> experiment with this one a bit and see what happens. It looks like it
> should be removed completely.

Yes, I did have the device_shutdown() here (turn off clocks and reset).
The processor had to be woken up before shutting down (see the comments
above the code hunk).

> 
>>
>>>> ÂÂÂÂÂ reset_control_assert(oproc->reset);

Overall, the previous design relied on get_sync and put_sync calling the
driver's callbacks once started and used regular clock and reset
management outside of those during powering up (first time) and actual
shut down.


>>>> Â ÂÂÂÂÂ ret = omap_rproc_disable_timers(rproc, true);
>>>> @@ -496,6 +548,15 @@ static int omap_rproc_stop(struct rproc *rproc)
>>>> Â ÂÂÂÂÂ mbox_free_channel(oproc->mbox);
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * update the runtime pm states and status now that the remoteproc
>>>> +ÂÂÂÂ * has stopped
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ pm_runtime_disable(dev);
>>>> +ÂÂÂ pm_runtime_dont_use_autosuspend(dev);
>>>> +ÂÂÂ pm_runtime_put_noidle(dev);
>>>> +ÂÂÂ pm_runtime_set_suspended(dev);
>>>> +
>>>> ÂÂÂÂÂ return 0;
>>>> Â }
>>>> Â @@ -552,17 +613,19 @@ static bool _is_rproc_in_standby(struct
>>>> omap_rproc *oproc)
>>>> Â Â /* 1 sec is long enough time to let the remoteproc side suspend
>>>> the device */
>>>> Â #define DEF_SUSPEND_TIMEOUT 1000
>>>> -static int _omap_rproc_suspend(struct rproc *rproc)
>>>> +static int _omap_rproc_suspend(struct rproc *rproc, bool auto_suspend)
>>>> Â {
>>>> ÂÂÂÂÂ struct device *dev = rproc->dev.parent;
>>>> ÂÂÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> ÂÂÂÂÂ unsigned long to = msecs_to_jiffies(DEF_SUSPEND_TIMEOUT);
>>>> ÂÂÂÂÂ unsigned long ta = jiffies + to;
>>>> +ÂÂÂ u32 suspend_msg = auto_suspend ?
>>>> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ RP_MBOX_SUSPEND_AUTO : RP_MBOX_SUSPEND_SYSTEM;
>>>> ÂÂÂÂÂ int ret;
>>>> Â ÂÂÂÂÂ reinit_completion(&oproc->pm_comp);
>>>> ÂÂÂÂÂ oproc->suspend_acked = false;
>>>> -ÂÂÂ ret = mbox_send_message(oproc->mbox, (void
>>>> *)RP_MBOX_SUSPEND_SYSTEM);
>>>> +ÂÂÂ ret = mbox_send_message(oproc->mbox, (void *)suspend_msg);
>>>> ÂÂÂÂÂ if (ret < 0) {
>>>> ÂÂÂÂÂÂÂÂÂ dev_err(dev, "PM mbox_send_message failed: %d\n", ret);
>>>> ÂÂÂÂÂÂÂÂÂ return ret;
>>>> @@ -602,25 +665,55 @@ static int _omap_rproc_suspend(struct rproc
>>>> *rproc)
>>>> ÂÂÂÂÂÂÂÂÂ goto enable_device;
>>>> ÂÂÂÂÂ }
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * IOMMUs would have to be disabled specifically for runtime
>>>> suspend.
>>>> +ÂÂÂÂ * They are handled automatically through System PM callbacks for
>>>> +ÂÂÂÂ * regular system suspend
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ if (auto_suspend) {
>>>> +ÂÂÂÂÂÂÂ ret = omap_iommu_domain_deactivate(rproc->domain);
>>>> +ÂÂÂÂÂÂÂ if (ret) {
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ dev_err(dev, "iommu domain deactivate failed %d\n",
>>>> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ ret);
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ goto enable_timers;
>>>> +ÂÂÂÂÂÂÂ }
>>>> +ÂÂÂ }
>>>> +
>>>> ÂÂÂÂÂ return 0;
>>
>> blank line here, and remove the one before enable_device.
> 
> Hmm why? I don't think there is any coding guideline that would dictate
> this. However, I'll fix this as this is effectively your file anyways.

Thanks, easier for readability between the success and failure paths.

regards
Suman

> 
> -Tero
> 
>>
>> regards
>> Suman
>>
>>>> +enable_timers:
>>>> +ÂÂÂ /* ignore errors on re-enabling code */
>>>> +ÂÂÂ omap_rproc_enable_timers(rproc, false);
>>>> Â Â enable_device:
>>>> ÂÂÂÂÂ reset_control_deassert(oproc->reset);
>>>> ÂÂÂÂÂ return ret;
>>>> Â }
>>>> Â -static int _omap_rproc_resume(struct rproc *rproc)
>>>> +static int _omap_rproc_resume(struct rproc *rproc, bool auto_suspend)
>>>> Â {
>>>> ÂÂÂÂÂ struct device *dev = rproc->dev.parent;
>>>> ÂÂÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> ÂÂÂÂÂ int ret;
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * IOMMUs would have to be enabled specifically for runtime
>>>> resume.
>>>> +ÂÂÂÂ * They would have been already enabled automatically through
>>>> System
>>>> +ÂÂÂÂ * PM callbacks for regular system resume
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ if (auto_suspend) {
>>>> +ÂÂÂÂÂÂÂ ret = omap_iommu_domain_activate(rproc->domain);
>>>> +ÂÂÂÂÂÂÂ if (ret) {
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ dev_err(dev, "omap_iommu activate failed %d\n", ret);
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ goto out;
>>>> +ÂÂÂÂÂÂÂ }
>>>> +ÂÂÂ }
>>>> +
>>>> ÂÂÂÂÂ /* boot address could be lost after suspend, so restore it */
>>>> ÂÂÂÂÂ if (oproc->boot_data) {
>>>> ÂÂÂÂÂÂÂÂÂ ret = omap_rproc_write_dsp_boot_addr(rproc);
>>>> ÂÂÂÂÂÂÂÂÂ if (ret) {
>>>> ÂÂÂÂÂÂÂÂÂÂÂÂÂ dev_err(dev, "boot address restore failed %d\n", ret);
>>>> -ÂÂÂÂÂÂÂÂÂÂÂ goto out;
>>>> +ÂÂÂÂÂÂÂÂÂÂÂ goto suspend_iommu;
>>>
>>> The same needs to be done if omap_rproc_enable_timers() fails.
>>>
>>>> ÂÂÂÂÂÂÂÂÂ }
>>>> ÂÂÂÂÂ }
>>>> Â @@ -633,6 +726,12 @@ static int _omap_rproc_resume(struct rproc
>>>> *rproc)
>>>> Â ÂÂÂÂÂ reset_control_deassert(oproc->reset);
>>>> Â +ÂÂÂ return 0;
>>>> +
>>>> +suspend_iommu:
>>>> +ÂÂÂ if (auto_suspend)
>>>> +ÂÂÂÂÂÂÂ omap_iommu_domain_deactivate(rproc->domain);
>>>> +
>>>> Â out:
>>>> ÂÂÂÂÂ return ret;
>>>> Â }
>>>> @@ -641,6 +740,7 @@ static int __maybe_unused
>>>> omap_rproc_suspend(struct device *dev)
>>>> Â {
>>>> ÂÂÂÂÂ struct platform_device *pdev = to_platform_device(dev);
>>>> ÂÂÂÂÂ struct rproc *rproc = platform_get_drvdata(pdev);
>>>> +ÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> ÂÂÂÂÂ int ret = 0;
>>>> Â ÂÂÂÂÂ mutex_lock(&rproc->lock);
>>>> @@ -655,13 +755,25 @@ static int __maybe_unused
>>>> omap_rproc_suspend(struct device *dev)
>>>> ÂÂÂÂÂÂÂÂÂ goto out;
>>>> ÂÂÂÂÂ }
>>>> Â -ÂÂÂ ret = _omap_rproc_suspend(rproc);
>>>> +ÂÂÂ ret = _omap_rproc_suspend(rproc, false);
>>>> ÂÂÂÂÂ if (ret) {
>>>> ÂÂÂÂÂÂÂÂÂ dev_err(dev, "suspend failed %d\n", ret);
>>>> ÂÂÂÂÂÂÂÂÂ goto out;
>>>> ÂÂÂÂÂ }
>>>> Â +ÂÂÂ /*
>>>> +ÂÂÂÂ * remoteproc is running at the time of system suspend, so
>>>> remember
>>>> +ÂÂÂÂ * it so as to wake it up during system resume
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ oproc->need_resume = 1;
>>>
>>> Please use 'true' to be consistent with the type and
>>> omap_rproc_resume().
>>>
>>>> ÂÂÂÂÂ rproc->state = RPROC_SUSPENDED;
>>>> +
>>>> +ÂÂÂ /*
>>>> +ÂÂÂÂ * update the runtime pm status to be suspended, without
>>>> decrementing
>>>> +ÂÂÂÂ * the device usage count
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ pm_runtime_disable(dev);
>>>> +ÂÂÂ pm_runtime_set_suspended(dev);
>>>> Â out:
>>>> ÂÂÂÂÂ mutex_unlock(&rproc->lock);
>>>> ÂÂÂÂÂ return ret;
>>>> @@ -671,6 +783,7 @@ static int __maybe_unused
>>>> omap_rproc_resume(struct device *dev)
>>>> Â {
>>>> ÂÂÂÂÂ struct platform_device *pdev = to_platform_device(dev);
>>>> ÂÂÂÂÂ struct rproc *rproc = platform_get_drvdata(pdev);
>>>> +ÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> ÂÂÂÂÂ int ret = 0;
>>>> Â ÂÂÂÂÂ mutex_lock(&rproc->lock);
>>>> @@ -682,17 +795,91 @@ static int __maybe_unused
>>>> omap_rproc_resume(struct device *dev)
>>>> ÂÂÂÂÂÂÂÂÂ goto out;
>>>> ÂÂÂÂÂ }
>>>> Â -ÂÂÂ ret = _omap_rproc_resume(rproc);
>>>> +ÂÂÂ /*
>>>> +ÂÂÂÂ * remoteproc was auto-suspended at the time of system suspend,
>>>> +ÂÂÂÂ * so no need to wake-up the processor (leave it in suspended
>>>> +ÂÂÂÂ * state, will be woken up during a subsequent runtime_resume)
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ if (!oproc->need_resume)
>>>> +ÂÂÂÂÂÂÂ goto out;
>>>> +
>>>> +ÂÂÂ ret = _omap_rproc_resume(rproc, false);
>>>> ÂÂÂÂÂ if (ret) {
>>>> ÂÂÂÂÂÂÂÂÂ dev_err(dev, "resume failed %d\n", ret);
>>>> ÂÂÂÂÂÂÂÂÂ goto out;
>>>> ÂÂÂÂÂ }
>>>> +ÂÂÂ oproc->need_resume = false;
>>>> Â ÂÂÂÂÂ rproc->state = RPROC_RUNNING;
>>>> +
>>>> +ÂÂÂ /*
>>>> +ÂÂÂÂ * update the runtime pm status to be active, without incrementing
>>>> +ÂÂÂÂ * the device usage count
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ pm_runtime_set_active(dev);
>>>> +ÂÂÂ pm_runtime_enable(dev);
>>>> +ÂÂÂ pm_runtime_mark_last_busy(dev);
>>>> Â out:
>>>> ÂÂÂÂÂ mutex_unlock(&rproc->lock);
>>>> ÂÂÂÂÂ return ret;
>>>> Â }
>>>> +
>>>> +static int omap_rproc_runtime_suspend(struct device *dev)
>>>> +{
>>>> +ÂÂÂ struct rproc *rproc = dev_get_drvdata(dev);
>>>> +ÂÂÂ struct omap_rproc *oproc = rproc->priv;
>>>> +ÂÂÂ int ret;
>>>> +
>>>> +ÂÂÂ if (rproc->state == RPROC_CRASHED) {
>>>> +ÂÂÂÂÂÂÂ dev_dbg(dev, "rproc cannot be runtime suspended when
>>>> crashed!\n");
>>>> +ÂÂÂÂÂÂÂ return -EBUSY;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ if (WARN_ON(rproc->state != RPROC_RUNNING)) {
>>>> +ÂÂÂÂÂÂÂ dev_err(dev, "rproc cannot be runtime suspended when not
>>>> running!\n");
>>>> +ÂÂÂÂÂÂÂ return -EBUSY;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ /*
>>>> +ÂÂÂÂ * do not even attempt suspend if the remote processor is not
>>>> +ÂÂÂÂ * idled for runtime auto-suspend
>>>> +ÂÂÂÂ */
>>>> +ÂÂÂ if (!_is_rproc_in_standby(oproc)) {
>>>> +ÂÂÂÂÂÂÂ ret = -EBUSY;
>>>> +ÂÂÂÂÂÂÂ goto abort;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ ret = _omap_rproc_suspend(rproc, true);
>>>> +ÂÂÂ if (ret)
>>>> +ÂÂÂÂÂÂÂ goto abort;
>>>> +
>>>> +ÂÂÂ rproc->state = RPROC_SUSPENDED;
>>>> +ÂÂÂ return 0;
>>>> +
>>>> +abort:
>>>> +ÂÂÂ pm_runtime_mark_last_busy(dev);
>>>> +ÂÂÂ return ret;
>>>> +}
>>>> +
>>>> +static int omap_rproc_runtime_resume(struct device *dev)
>>>> +{
>>>> +ÂÂÂ struct rproc *rproc = dev_get_drvdata(dev);
>>>> +ÂÂÂ int ret;
>>>> +
>>>> +ÂÂÂ if (WARN_ON(rproc->state != RPROC_SUSPENDED)) {
>>>> +ÂÂÂÂÂÂÂ dev_err(dev, "rproc cannot be runtime resumed if not
>>>> suspended!\n");
>>>> +ÂÂÂÂÂÂÂ return -EBUSY;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ ret = _omap_rproc_resume(rproc, true);
>>>> +ÂÂÂ if (ret) {
>>>> +ÂÂÂÂÂÂÂ dev_err(dev, "runtime resume failed %d\n", ret);
>>>> +ÂÂÂÂÂÂÂ return ret;
>>>> +ÂÂÂ }
>>>> +
>>>> +ÂÂÂ rproc->state = RPROC_RUNNING;
>>>> +ÂÂÂ return 0;
>>>> +}
>>>> Â #endif /* CONFIG_PM */
>>>> Â Â static const char * const ipu_mem_names[] = {
>>>> @@ -778,6 +965,20 @@ static const struct of_device_id
>>>> omap_rproc_of_match[] = {
>>>> Â };
>>>> Â MODULE_DEVICE_TABLE(of, omap_rproc_of_match);
>>>> Â +static int omap_rproc_get_autosuspend_delay(struct
>>>> platform_device *pdev)
>>>> +{
>>>> +ÂÂÂ const struct omap_rproc_dev_data *data;
>>>> +ÂÂÂ int delay;
>>>> +
>>>> +ÂÂÂ data = of_device_get_match_data(&pdev->dev);
>>>> +ÂÂÂ if (!data)
>>>> +ÂÂÂÂÂÂÂ return -ENODEV;
>>>
>>> This check is done in omap_rproc_of_get_internal_memories() and
>>> omap_rproc_get_boot_data(). I think it would be best to do it once
>>> at the top
>>> of the probe() function and be done with it.
>>>
>>> That being said and as noted in a previous comment, I would push all
>>> tuneables
>>> to the DT. If the property is missing then things default to
>>> DEFAULT_AUTOSUSPEND_DELAY.
>>>
>>>> +
>>>> +ÂÂÂ delay = data->autosuspend_delay;
>>>> +
>>>> +ÂÂÂ return (delay > 0) ? delay : DEFAULT_AUTOSUSPEND_DELAY;
>>>> +}
>>>> +
>>>> Â static const char *omap_rproc_get_firmware(struct platform_device
>>>> *pdev)
>>>> Â {
>>>> ÂÂÂÂÂ const char *fw_name;
>>>> @@ -953,6 +1154,11 @@ static int omap_rproc_probe(struct
>>>> platform_device *pdev)
>>>> ÂÂÂÂÂ }
>>>> Â ÂÂÂÂÂ init_completion(&oproc->pm_comp);
>>>> +ÂÂÂ oproc->autosuspend_delay = omap_rproc_get_autosuspend_delay(pdev);
>>>> +ÂÂÂ if (oproc->autosuspend_delay < 0) {
>>>> +ÂÂÂÂÂÂÂ ret = oproc->autosuspend_delay;
>>>> +ÂÂÂÂÂÂÂ goto free_rproc;
>>>> +ÂÂÂ }
>>>> Â ÂÂÂÂÂ oproc->fck = devm_clk_get(&pdev->dev, 0);
>>>> ÂÂÂÂÂ if (IS_ERR(oproc->fck)) {
>>>> @@ -994,6 +1200,8 @@ static int omap_rproc_remove(struct
>>>> platform_device *pdev)
>>>> Â Â static const struct dev_pm_ops omap_rproc_pm_ops = {
>>>> ÂÂÂÂÂ SET_SYSTEM_SLEEP_PM_OPS(omap_rproc_suspend, omap_rproc_resume)
>>>> +ÂÂÂ SET_RUNTIME_PM_OPS(omap_rproc_runtime_suspend,
>>>> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ omap_rproc_runtime_resume, NULL)
>>>> Â };
>>>> Â Â static struct platform_driver omap_rproc_driver = {
>>>> --Â
>>>> 2.17.1
>>>>
>>>> -- 
>>
> 
> -- 
> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
> Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki