[PATCH 3/3] dmaengine: pl330: Don't require irq-safe runtime PM

Marek Szyprowski m.szyprowski at samsung.com
Fri Dec 23 02:38:19 PST 2016


Hi Krzysztof,


On 2016-12-23 03:48, Krzysztof Kozlowski wrote:
> On Thu, Dec 22, 2016 at 01:11:29PM +0100, Marek Szyprowski wrote:
>> This patch replaces irq-safe runtime pm with non-irq-safe version.
>>
>> Till now non-irq-safe runtime PM implementation was only possible by calling
>> pm_runtime_get/put functions from alloc/free_chan_resources. All other DMA
>> engine API functions cannot be called from a context, which allows sleeping.
>> Such implementation in practice resulted in keeping PL330 DMA controller
>> device active almost all the time, because most of the slave device drivers
>> (DMA engine clients) acquired DMA channel in their probe() function and
>> released it during driver removal.
> I think that is not accurate (unless something changed recently and I
> missed it)... The runtime PM in pl330 was infact suspending device when
> the queue was empty. It was working during the regular work of DMA and
> slave devices.
>
> Maybe your recent fix changed this?

Please note that the above description was about existing ways of 
implementing
non-irq-safe runtime pm in a dmaengine driver. PL330 was using irq-safe 
runtime pm,
which has other issues - the most significant one is keeping clocks 
prepared all
the time and keeping power domain turned on all the time. Both are real 
blockers
of taking advantages of the fact that there is a separate power domain 
for audio
hardware.

>> This patch provides a new approach.
> Please say why this approach is better because the previous one - was
> suspending the dma channel when not used. Otherwise, it would make no
> sense of implement the irq-safe runtime PM at the beginning!
>
> Of course, a change from irq-safe to non-irq-safe is a good reason to
> implement changes. But that was not mentioned in the first paragraph at
> all.

The goal of this patchset is to add audio power domain support for 
Exynos5 SoCs
and let it to be turned off when no sound is being played/recorded. This 
in turn
requires all devices, which are a part of that power domain to implement 
runtime
pm support. Maybe I didn't state this strong enough, but I think that 
there was
such information in the commit message AND cover letter.

>> The main assumption for it is an
>> observation that there can be only one slave device using each DMA channel.
> Wait, observation, real requirement or assumption?
>
> Later in the code I see adding such requirement.

Well, observation which result in assumption. I cannot imagine a 
hardware which
shares slave DMA channel between devices. Also none of the existing platform
does it.

>> Using recently introduced device dependencies (links) infrastructure one can
>> ensure proper runtime PM state of PL330 DMA controller. In this approach in
>> pl330_alloc_chan_resources() function a new dependency is being created
>> between PL330 DMA controller device (as supplier) and given slave device
>> (as consumer). This way PL330 DMA controller device runtime active counter
>> is increased when the slave device is resumed and decreased the same time
>> when given slave device is put to suspend. This way it has been ensured to
>> keep PL330 DMA controller runtime active if there is an active used of any
>> of its DMA channels. Slave device pointer is initially stored in per-channel
>> data in of_dma_xlate callback. This is similar to what has been already
>> implemented in Exynos IOMMU driver in commit 2f5f44f205cc958b
>> ("iommu/exynos: Use device dependency links to control runtime pm").
> Sounds convincing... Interesting approach!
>
> My doubts are:
> 1. What with more then one slave device? (assumption?)

See above, there are no such cases.

> 2. If slave device does not implement runtime PM, then pl330 will be
>     active all the time?

Right, but the goal is to have runtime pm added to all devices in the 
system.

> 3. If slave device implements runtime PM in a way that it's enabled in
>     probe and released in remove, then pl330 will be active all the time?

Then it will force power domain to be turned on all the time and even 
optional
fine-grained irq-safe runtiem pm in pl330 driver won't help much to 
reduce power
consumption. I assume that the real goal with runtime pm is to let 
respective
power domains to be turned off, what gives the best results in terms of 
power
saving.

>> If one requests memory-to-memory chanel, runtime active counter is
>> increased unconditionally. This might be a drawback of this approach, but
>> PL330 is not really used for memory-to-memory operations due to poor
>> performance in such operations compared to CPU.
>>
>> Introducing non-irq-safe runtime power management finally allows to turn
>> audio power domain off on Exynos5 SoCs.
> ... and actually any domain which contained pl330 device so this is nice
> benefit!
>
>> Removal of irq-safe runtime pm is based on the revert of the following
>> commits:
>> 1. "dmaengine: pl330: fix runtime pm support" commit <no stable id yet>
>> 2. "dmaengine: pl330: Fix hang on dmaengine_terminate_all on certain boards"
>>     commit 81cc6edc08705ac0146fe6ac14a0982a31ce6f3d
>> 3. "ARM: 8202/1: dmaengine: pl330: Add runtime Power Management support v12"
>>     commit ae43b3289186480f81c78bb63d788a85a3631f47
>>
>> Signed-off-by: Marek Szyprowski <m.szyprowski at samsung.com>
>> ---
>>   drivers/dma/pl330.c | 133 +++++++++++++++++++++++++++-------------------------
>>   1 file changed, 70 insertions(+), 63 deletions(-)
>>
>> diff --git a/drivers/dma/pl330.c b/drivers/dma/pl330.c
>> index 3c80e71..2cffbb4 100644
>> --- a/drivers/dma/pl330.c
>> +++ b/drivers/dma/pl330.c
>> @@ -268,9 +268,6 @@ enum pl330_byteswap {
>>   
>>   #define NR_DEFAULT_DESC	16
>>   
>> -/* Delay for runtime PM autosuspend, ms */
>> -#define PL330_AUTOSUSPEND_DELAY 20
>> -
>>   /* Populated by the PL330 core driver for DMA API driver's info */
>>   struct pl330_config {
>>   	u32	periph_id;
>> @@ -449,7 +446,8 @@ struct dma_pl330_chan {
>>   	bool cyclic;
>>   
>>   	/* for runtime pm tracking */
>> -	bool active;
>> +	struct device *slave;
>> +	struct device_link *slave_link;
>>   };
>>   
>>   struct pl330_dmac {
>> @@ -2015,7 +2013,6 @@ static void pl330_tasklet(unsigned long data)
>>   	struct dma_pl330_chan *pch = (struct dma_pl330_chan *)data;
>>   	struct dma_pl330_desc *desc, *_dt;
>>   	unsigned long flags;
>> -	bool power_down = false;
>>   
>>   	spin_lock_irqsave(&pch->lock, flags);
>>   
>> @@ -2030,18 +2027,10 @@ static void pl330_tasklet(unsigned long data)
>>   	/* Try to submit a req imm. next to the last completed cookie */
>>   	fill_queue(pch);
>>   
>> -	if (list_empty(&pch->work_list)) {
>> -		spin_lock(&pch->thread->dmac->lock);
>> -		_stop(pch->thread);
>> -		spin_unlock(&pch->thread->dmac->lock);
>> -		power_down = true;
>> -		pch->active = false;
>> -	} else {
>> -		/* Make sure the PL330 Channel thread is active */
>> -		spin_lock(&pch->thread->dmac->lock);
>> -		_start(pch->thread);
>> -		spin_unlock(&pch->thread->dmac->lock);
>> -	}
>> +	/* Make sure the PL330 Channel thread is active */
>> +	spin_lock(&pch->thread->dmac->lock);
>> +	_start(pch->thread);
>> +	spin_unlock(&pch->thread->dmac->lock);
>>   
>>   	while (!list_empty(&pch->completed_list)) {
>>   		struct dmaengine_desc_callback cb;
>> @@ -2054,13 +2043,6 @@ static void pl330_tasklet(unsigned long data)
>>   		if (pch->cyclic) {
>>   			desc->status = PREP;
>>   			list_move_tail(&desc->node, &pch->work_list);
>> -			if (power_down) {
>> -				pch->active = true;
>> -				spin_lock(&pch->thread->dmac->lock);
>> -				_start(pch->thread);
>> -				spin_unlock(&pch->thread->dmac->lock);
>> -				power_down = false;
>> -			}
>>   		} else {
>>   			desc->status = FREE;
>>   			list_move_tail(&desc->node, &pch->dmac->desc_pool);
>> @@ -2075,12 +2057,6 @@ static void pl330_tasklet(unsigned long data)
>>   		}
>>   	}
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>> -
>> -	/* If work list empty, power down */
>> -	if (power_down) {
>> -		pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
>> -		pm_runtime_put_autosuspend(pch->dmac->ddma.dev);
>> -	}
>>   }
>>   
>>   bool pl330_filter(struct dma_chan *chan, void *param)
>> @@ -2113,14 +2089,63 @@ static struct dma_chan *of_dma_pl330_xlate(struct of_phandle_args *dma_spec,
>>   	if (chan_id >= pl330->num_peripherals)
>>   		return NULL;
>>   
>> +	if (!pl330->peripherals[chan_id].slave)
>> +		pl330->peripherals[chan_id].slave = slave;
>> +	else if (pl330->peripherals[chan_id].slave != slave) {
>> +		dev_err(pl330->ddma.dev,
>> +			"Can't use same channel with multiple slave devices!\n");
>> +		return NULL;
>> +	}
> This could be nicely split into separate patch.

Okay, if you want, I can move this change to separate patch.

> Best regards,
> Krzysztof
>
>> +
>>   	return dma_get_slave_channel(&pl330->peripherals[chan_id].chan);
>>   }
>>   
>> +static int pl330_add_slave_link(struct pl330_dmac *pl330,
>> +				struct dma_pl330_chan *pch)
>> +{
>> +	struct device_link *link;
>> +	int i;
>> +
>> +	if (pch->slave_link)
>> +		return 0;
>> +
>> +	link = device_link_add(pch->slave, pl330->ddma.dev,
>> +				       DL_FLAG_PM_RUNTIME | DL_FLAG_RPM_ACTIVE);
>> +	if (!link)
>> +		return -ENODEV;
>> +
>> +	for (i = 0; i < pl330->num_peripherals; i++)
>> +		if (pl330->peripherals[i].slave == pch->slave)
>> +			pl330->peripherals[i].slave_link = link;
>> +	return 0;
>> +}
>> +
>> +static void pl330_del_slave_link(struct pl330_dmac *pl330,
>> +				 struct dma_pl330_chan *pch)
>> +{
>> +	struct device_link *link = pch->slave_link;
>> +	int i, count = 0;
>> +
>> +	for (i = 0; i < pl330->num_peripherals; i++)
>> +		if (pl330->peripherals[i].slave == pch->slave &&
>> +		    pl330->peripherals[i].thread)
>> +			count++;
>> +
>> +	if (count > 0)
>> +		return;
>> +
>> +	device_link_del(link);
>> +	for (i = 0; i < pl330->num_peripherals; i++)
>> +		if (pl330->peripherals[i].slave == pch->slave)
>> +			pch->slave_link = NULL;
>> +}
>> +
>>   static int pl330_alloc_chan_resources(struct dma_chan *chan)
>>   {
>>   	struct dma_pl330_chan *pch = to_pchan(chan);
>>   	struct pl330_dmac *pl330 = pch->dmac;
>>   	unsigned long flags;
>> +	int ret = 0;
>>   
>>   	spin_lock_irqsave(&pch->lock, flags);
>>   
>> @@ -2137,6 +2162,14 @@ static int pl330_alloc_chan_resources(struct dma_chan *chan)
>>   
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>>   
>> +	if (pch->slave)
>> +		ret = pl330_add_slave_link(pl330, pch);
>> +	else
>> +		ret = pm_runtime_get_sync(pl330->ddma.dev);
>> +
>> +	if (ret < 0)
>> +		return ret;
>> +
>>   	return 1;
>>   }
>>   
>> @@ -2171,9 +2204,7 @@ static int pl330_terminate_all(struct dma_chan *chan)
>>   	unsigned long flags;
>>   	struct pl330_dmac *pl330 = pch->dmac;
>>   	LIST_HEAD(list);
>> -	bool power_down = false;
>>   
>> -	pm_runtime_get_sync(pl330->ddma.dev);
>>   	spin_lock_irqsave(&pch->lock, flags);
>>   	spin_lock(&pl330->lock);
>>   	_stop(pch->thread);
>> @@ -2182,8 +2213,6 @@ static int pl330_terminate_all(struct dma_chan *chan)
>>   	pch->thread->req[0].desc = NULL;
>>   	pch->thread->req[1].desc = NULL;
>>   	pch->thread->req_running = -1;
>> -	power_down = pch->active;
>> -	pch->active = false;
>>   
>>   	/* Mark all desc done */
>>   	list_for_each_entry(desc, &pch->submitted_list, node) {
>> @@ -2200,10 +2229,6 @@ static int pl330_terminate_all(struct dma_chan *chan)
>>   	list_splice_tail_init(&pch->work_list, &pl330->desc_pool);
>>   	list_splice_tail_init(&pch->completed_list, &pl330->desc_pool);
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>> -	pm_runtime_mark_last_busy(pl330->ddma.dev);
>> -	if (power_down)
>> -		pm_runtime_put_autosuspend(pl330->ddma.dev);
>> -	pm_runtime_put_autosuspend(pl330->ddma.dev);
>>   
>>   	return 0;
>>   }
>> @@ -2221,7 +2246,6 @@ static int pl330_pause(struct dma_chan *chan)
>>   	struct pl330_dmac *pl330 = pch->dmac;
>>   	unsigned long flags;
>>   
>> -	pm_runtime_get_sync(pl330->ddma.dev);
>>   	spin_lock_irqsave(&pch->lock, flags);
>>   
>>   	spin_lock(&pl330->lock);
>> @@ -2229,8 +2253,6 @@ static int pl330_pause(struct dma_chan *chan)
>>   	spin_unlock(&pl330->lock);
>>   
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>> -	pm_runtime_mark_last_busy(pl330->ddma.dev);
>> -	pm_runtime_put_autosuspend(pl330->ddma.dev);
>>   
>>   	return 0;
>>   }
>> @@ -2238,11 +2260,11 @@ static int pl330_pause(struct dma_chan *chan)
>>   static void pl330_free_chan_resources(struct dma_chan *chan)
>>   {
>>   	struct dma_pl330_chan *pch = to_pchan(chan);
>> +	struct pl330_dmac *pl330 = pch->dmac;
>>   	unsigned long flags;
>>   
>>   	tasklet_kill(&pch->task);
>>   
>> -	pm_runtime_get_sync(pch->dmac->ddma.dev);
>>   	spin_lock_irqsave(&pch->lock, flags);
>>   
>>   	pl330_release_channel(pch->thread);
>> @@ -2252,19 +2274,20 @@ static void pl330_free_chan_resources(struct dma_chan *chan)
>>   		list_splice_tail_init(&pch->work_list, &pch->dmac->desc_pool);
>>   
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>> -	pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
>> -	pm_runtime_put_autosuspend(pch->dmac->ddma.dev);
>> +
>> +	if (pch->slave)
>> +		pl330_del_slave_link(pl330, pch);
>> +	else
>> +		pm_runtime_put(pl330->ddma.dev);
>>   }
>>   
>>   static int pl330_get_current_xferred_count(struct dma_pl330_chan *pch,
>>   					   struct dma_pl330_desc *desc)
>>   {
>>   	struct pl330_thread *thrd = pch->thread;
>> -	struct pl330_dmac *pl330 = pch->dmac;
>>   	void __iomem *regs = thrd->dmac->base;
>>   	u32 val, addr;
>>   
>> -	pm_runtime_get_sync(pl330->ddma.dev);
>>   	val = addr = 0;
>>   	if (desc->rqcfg.src_inc) {
>>   		val = readl(regs + SA(thrd->id));
>> @@ -2273,8 +2296,6 @@ static int pl330_get_current_xferred_count(struct dma_pl330_chan *pch,
>>   		val = readl(regs + DA(thrd->id));
>>   		addr = desc->px.dst_addr;
>>   	}
>> -	pm_runtime_mark_last_busy(pch->dmac->ddma.dev);
>> -	pm_runtime_put_autosuspend(pl330->ddma.dev);
>>   
>>   	/* If DMAMOV hasn't finished yet, SAR/DAR can be zero */
>>   	if (!val)
>> @@ -2360,16 +2381,6 @@ static void pl330_issue_pending(struct dma_chan *chan)
>>   	unsigned long flags;
>>   
>>   	spin_lock_irqsave(&pch->lock, flags);
>> -	if (list_empty(&pch->work_list)) {
>> -		/*
>> -		 * Warn on nothing pending. Empty submitted_list may
>> -		 * break our pm_runtime usage counter as it is
>> -		 * updated on work_list emptiness status.
>> -		 */
>> -		WARN_ON(list_empty(&pch->submitted_list));
>> -		pch->active = true;
>> -		pm_runtime_get_sync(pch->dmac->ddma.dev);
>> -	}
>>   	list_splice_tail_init(&pch->submitted_list, &pch->work_list);
>>   	spin_unlock_irqrestore(&pch->lock, flags);
>>   
>> @@ -2987,11 +2998,7 @@ static int __maybe_unused pl330_resume(struct device *dev)
>>   		pcfg->data_buf_dep, pcfg->data_bus_width / 8, pcfg->num_chan,
>>   		pcfg->num_peri, pcfg->num_events);
>>   
>> -	pm_runtime_irq_safe(&adev->dev);
>> -	pm_runtime_use_autosuspend(&adev->dev);
>> -	pm_runtime_set_autosuspend_delay(&adev->dev, PL330_AUTOSUSPEND_DELAY);
>> -	pm_runtime_mark_last_busy(&adev->dev);
>> -	pm_runtime_put_autosuspend(&adev->dev);
>> +	pm_runtime_put(&adev->dev);
>>   
>>   	return 0;
>>   probe_err3:
>> -- 
>> 1.9.1
>>
>
>

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland




More information about the linux-arm-kernel mailing list