From: Dmitry Osipenko <digetx@gmail.com>
To: "Jon Hunter" <jonathanh@nvidia.com>,
"Laxman Dewangan" <ldewangan@nvidia.com>,
"Vinod Koul" <vkoul@kernel.org>,
"Dan Williams" <dan.j.williams@intel.com>,
"Thierry Reding" <thierry.reding@gmail.com>,
"Michał Mirosław" <mirq-linux@rere.qmqm.pl>
Cc: dmaengine@vger.kernel.org, linux-tegra@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v6 11/16] dmaengine: tegra-apb: Keep clock enabled only during of DMA transfer
Date: Thu, 30 Jan 2020 23:04:37 +0300 [thread overview]
Message-ID: <cedbf558-b15b-81ca-7833-c94aedce5c5c@gmail.com> (raw)
In-Reply-To: <e72d00ee-abee-9ae2-4654-da77420b440e@nvidia.com>
30.01.2020 21:45, Jon Hunter пишет:
>
> On 30/01/2020 16:11, Dmitry Osipenko wrote:
>> 30.01.2020 17:09, Jon Hunter пишет:
>>>
>>> On 30/01/2020 04:37, Dmitry Osipenko wrote:
>>>> It's a bit impractical to enable hardware's clock at the time of DMA
>>>> channel's allocation because most of DMA client drivers allocate DMA
>>>> channel at the time of the driver's probing, and thus, DMA clock is kept
>>>> always-enabled in practice, defeating the whole purpose of runtime PM.
>>>>
>>>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>>>> ---
>>>> drivers/dma/tegra20-apb-dma.c | 47 ++++++++++++++++++++++++-----------
>>>> 1 file changed, 32 insertions(+), 15 deletions(-)
>>>>
>>>> diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c
>>>> index 22b88ccff05d..0ee28d8e3c96 100644
>>>> --- a/drivers/dma/tegra20-apb-dma.c
>>>> +++ b/drivers/dma/tegra20-apb-dma.c
>>>> @@ -436,6 +436,8 @@ static void tegra_dma_stop(struct tegra_dma_channel *tdc)
>>>> tdc_write(tdc, TEGRA_APBDMA_CHAN_STATUS, status);
>>>> }
>>>> tdc->busy = false;
>>>> +
>>>> + pm_runtime_put(tdc->tdma->dev);
>>>> }
>>>>
>>>> static void tegra_dma_start(struct tegra_dma_channel *tdc,
>>>> @@ -500,18 +502,25 @@ static void tegra_dma_configure_for_next(struct tegra_dma_channel *tdc,
>>>> tegra_dma_resume(tdc);
>>>> }
>>>>
>>>> -static void tdc_start_head_req(struct tegra_dma_channel *tdc)
>>>> +static bool tdc_start_head_req(struct tegra_dma_channel *tdc)
>>>> {
>>>> struct tegra_dma_sg_req *sg_req;
>>>> + int err;
>>>>
>>>> if (list_empty(&tdc->pending_sg_req))
>>>> - return;
>>>> + return false;
>>>> +
>>>> + err = pm_runtime_get_sync(tdc->tdma->dev);
>>>> + if (WARN_ON_ONCE(err < 0))
>>>> + return false;
>>>>
>>>> sg_req = list_first_entry(&tdc->pending_sg_req, typeof(*sg_req), node);
>>>> tegra_dma_start(tdc, sg_req);
>>>> sg_req->configured = true;
>>>> sg_req->words_xferred = 0;
>>>> tdc->busy = true;
>>>> +
>>>> + return true;
>>>> }
>>>>
>>>> static void tdc_configure_next_head_desc(struct tegra_dma_channel *tdc)
>>>> @@ -615,6 +624,8 @@ static void handle_once_dma_done(struct tegra_dma_channel *tdc,
>>>> }
>>>> list_add_tail(&sgreq->node, &tdc->free_sg_req);
>>>>
>>>> + pm_runtime_put(tdc->tdma->dev);
>>>> +
>>>> /* Do not start DMA if it is going to be terminate */
>>>> if (to_terminate || list_empty(&tdc->pending_sg_req))
>>>> return;
>>>> @@ -730,9 +741,7 @@ static void tegra_dma_issue_pending(struct dma_chan *dc)
>>>> dev_err(tdc2dev(tdc), "No DMA request\n");
>>>> goto end;
>>>> }
>>>> - if (!tdc->busy) {
>>>> - tdc_start_head_req(tdc);
>>>> -
>>>> + if (!tdc->busy && tdc_start_head_req(tdc)) {
>>>> /* Continuous single mode: Configure next req */
>>>> if (tdc->cyclic) {
>>>> /*
>>>> @@ -775,6 +784,13 @@ static int tegra_dma_terminate_all(struct dma_chan *dc)
>>>> else
>>>> wcount = status;
>>>>
>>>> + /*
>>>> + * tegra_dma_stop() will drop the RPM's usage refcount, but
>>>> + * tegra_dma_resume() touches hardware and thus we should keep
>>>> + * the DMA clock active while it's needed.
>>>> + */
>>>> + pm_runtime_get(tdc->tdma->dev);
>>>> +
>>>
>>> Would it work and make it simpler to just enable in the issue_pending
>>> and disable in the handle_once_dma_done or terminate_all?
>>>
>>> diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c
>>> index 3a45079d11ec..86bbb45da93d 100644
>>> --- a/drivers/dma/tegra20-apb-dma.c
>>> +++ b/drivers/dma/tegra20-apb-dma.c
>>> @@ -616,9 +616,14 @@ static void handle_once_dma_done(struct
>>> tegra_dma_channel *tdc,
>>> list_add_tail(&sgreq->node, &tdc->free_sg_req);
>>>
>>> /* Do not start DMA if it is going to be terminate */
>>> - if (to_terminate || list_empty(&tdc->pending_sg_req))
>>> + if (to_terminate)
>>> return;
>>>
>>> + if (list_empty(&tdc->pending_sg_req)) {
>>> + pm_runtime_put(tdc->tdma->dev);
>>> + return;
>>> + }
>>> +
>>> tdc_start_head_req(tdc);
>>> }
>>>
>>> @@ -729,6 +734,11 @@ static void tegra_dma_issue_pending(struct dma_chan
>>> *dc)
>>> goto end;
>>> }
>>> if (!tdc->busy) {
>>> + if (pm_runtime_get_sync(tdc->tdma->dev) < 0) {
>>> + dev_err(tdc2dev(tdc), "Failed to enable DMA!\n");
>>> + goto end;
>>> + }
>>> +
>>> tdc_start_head_req(tdc);
>>>
>>> /* Continuous single mode: Configure next req */
>>> @@ -788,6 +798,7 @@ static int tegra_dma_terminate_all(struct dma_chan *dc)
>>> get_current_xferred_count(tdc, sgreq,
>>> wcount);
>>> }
>>> tegra_dma_resume(tdc);
>>> + pm_runtime_put(tdc->tdma->dev);
>>>
>>> skip_dma_stop:
>>> tegra_dma_abort_all(tdc);
>>>
>>
>> The tegra_dma_stop() should put RPM anyways, which is missed in yours
>> sample. Please see handle_continuous_head_request().
>
> Yes and that is deliberate. The cyclic transfers the transfers *should*
> not stop until terminate_all is called. The tegra_dma_stop in
> handle_continuous_head_request() is an error condition and so I am not
> sure it is actually necessary to call pm_runtime_put() here.
But then tegra_dma_stop() shouldn't unset the "busy" mark.
>> I'm also finding the explicit get/put a bit easier to follow in the
>> code, don't you think so?
>
> I can see that, but I was thinking that in the case of cyclic transfers,
> it should only really be necessary to call the get/put at the beginning
> and end. So in my mind there should only be two exit points which are
> the ISR handler for SG and terminate_all for SG and cyclic.
Alright, I'll update this patch.
next prev parent reply other threads:[~2020-01-30 20:04 UTC|newest]
Thread overview: 39+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-30 4:37 [PATCH v6 00/16] NVIDIA Tegra APB DMA driver fixes and improvements Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 01/16] dmaengine: tegra-apb: Fix use-after-free Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 02/16] dmaengine: tegra-apb: Prevent race conditions of tasklet vs free list Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 03/16] dmaengine: tegra-apb: Implement synchronization hook Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 04/16] dmaengine: tegra-apb: Prevent race conditions on channel's freeing Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 05/16] dmaengine: tegra-apb: Clean up tasklet releasing Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 06/16] dmaengine: tegra-apb: Use devm_platform_ioremap_resource Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 07/16] dmaengine: tegra-apb: Use devm_request_irq Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 08/16] dmaengine: tegra-apb: Fix coding style problems Dmitry Osipenko
2020-01-30 14:08 ` Jon Hunter
2020-01-30 4:37 ` [PATCH v6 09/16] dmaengine: tegra-apb: Remove unneeded initialization of tdc->config_init Dmitry Osipenko
2020-01-30 14:08 ` Jon Hunter
2020-01-30 4:37 ` [PATCH v6 10/16] dmaengine: tegra-apb: Remove assumptions about unavailable runtime PM Dmitry Osipenko
2020-01-30 14:09 ` Jon Hunter
2020-01-30 16:09 ` Dmitry Osipenko
2020-01-30 4:37 ` [PATCH v6 11/16] dmaengine: tegra-apb: Keep clock enabled only during of DMA transfer Dmitry Osipenko
2020-01-30 14:09 ` Jon Hunter
2020-01-30 16:11 ` Dmitry Osipenko
2020-01-30 18:45 ` Jon Hunter
2020-01-30 20:04 ` Dmitry Osipenko [this message]
2020-01-31 9:02 ` Jon Hunter
2020-01-31 14:22 ` Dmitry Osipenko
2020-02-01 15:13 ` Dmitry Osipenko
2020-02-03 11:37 ` Jon Hunter
2020-02-03 16:24 ` Dmitry Osipenko
2020-01-31 9:05 ` Jon Hunter
2020-01-30 4:38 ` [PATCH v6 12/16] dmaengine: tegra-apb: Clean up suspend-resume Dmitry Osipenko
2020-01-30 14:09 ` Jon Hunter
2020-01-30 16:08 ` Dmitry Osipenko
2020-01-30 18:06 ` Dmitry Osipenko
2020-01-30 18:26 ` Dmitry Osipenko
2020-01-30 18:58 ` Jon Hunter
2020-01-30 19:00 ` Jon Hunter
2020-01-30 20:06 ` Dmitry Osipenko
2020-01-30 4:38 ` [PATCH v6 13/16] dmaengine: tegra-apb: Add missing of_dma_controller_free Dmitry Osipenko
2020-01-30 4:38 ` [PATCH v6 14/16] dmaengine: tegra-apb: Allow to compile as a loadable kernel module Dmitry Osipenko
2020-01-30 4:38 ` [PATCH v6 15/16] dmaengine: tegra-apb: Remove MODULE_ALIAS Dmitry Osipenko
2020-01-30 4:38 ` [PATCH v6 16/16] dmaengine: tegra-apb: Support COMPILE_TEST Dmitry Osipenko
2020-01-30 14:10 ` Jon Hunter
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=cedbf558-b15b-81ca-7833-c94aedce5c5c@gmail.com \
--to=digetx@gmail.com \
--cc=dan.j.williams@intel.com \
--cc=dmaengine@vger.kernel.org \
--cc=jonathanh@nvidia.com \
--cc=ldewangan@nvidia.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-tegra@vger.kernel.org \
--cc=mirq-linux@rere.qmqm.pl \
--cc=thierry.reding@gmail.com \
--cc=vkoul@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).