From: Jon Hunter <jonathanh@nvidia.com>
To: Thomas Gleixner <tglx@linutronix.de>,
Jason Cooper <jason@lakedaemon.net>,
Marc Zyngier <marc.zyngier@arm.com>,
Jiang Liu <jiang.liu@linux.intel.com>,
Stephen Warren <swarren@wwwdotorg.org>,
Thierry Reding <thierry.reding@gmail.com>
Cc: Kevin Hilman <khilman@kernel.org>,
Geert Uytterhoeven <geert@linux-m68k.org>,
Grygorii Strashko <grygorii.strashko@ti.com>,
Lars-Peter Clausen <lars@metafoo.de>,
Linus Walleij <linus.walleij@linaro.org>,
Soren Brinkmann <soren.brinkmann@xilinx.com>,
<linux-kernel@vger.kernel.org>, <linux-tegra@vger.kernel.org>
Subject: Re: [RFC PATCH V2 8/8] irqchip/gic: Add support for tegra AGIC interrupt controller
Date: Thu, 17 Dec 2015 10:58:52 +0000 [thread overview]
Message-ID: <5672956C.5010904@nvidia.com> (raw)
In-Reply-To: <1450349309-8107-9-git-send-email-jonathanh@nvidia.com>
On 17/12/15 10:48, Jon Hunter wrote:
> Add a driver for the Tegra-AGIC interrupt controller which is compatible
> with the ARM GIC-400 interrupt controller.
>
> The Tegra AGIC (Audio GIC) is part of the Audio Processing Engine (APE) on
> Tegra210 and can route interrupts to either the GIC for the CPU subsystem
> or the Audio DSP (ADSP) within the APE. The AGIC uses CPU interface 0 to
> route interrupts to the CPU GIC and CPU interface 1 to route interrupts to
> the ADSP.
>
> The APE is located within its own power domain on the chip and so the
> AGIC needs to manage both the power domain and its clocks. Commit
> afbbd2338176 ("irqchip/gic: Document optional Clock and Power Domain
> properties") adding clock and power-domain properties to the GIC binding
> and so the aim would be to make use of these to handle power management
> (however, this is very much dependent upon adding support for generic
> PM domains for Tegra which is still a work-in-progress).
>
> With the AGIC being located in a different power domain to the main CPU
> cluster this means that:
> 1. The interrupt controller cannot be registered via IRQCHIP_DECLARE()
> because it needs to be registered as a platform device so that the
> generic PM domain core will ensure that the power domain is available
> before probing.
> 2. The interrupt controller cannot be suspended/restored based upon
> changes in the CPU power state and needs to use runtime-pm instead.
>
> This is very much a work-in-progress and there are still a few items that
> need to be resolved. These items are:
> 1. Currently the GIC platform driver only supports non-root GICs. The
> platform driver performs a save and restore of PPI interrupts for
> non-root GICs, which is probably not necessary and so could be changed.
> At a minimum we need to re-enable the CPU interface during the device
> resume but we could skip the restoration of the PPIs. In general we
> could update the driver to only save and restore PPIs for the root
> controller, if that makes sense.
> 2. Currently routing of interrupts to the ADSP for Tegra210 is not
> supported by this driver. Although the ADSP on Tegra210 could also setup
> the AGIC distributor having two independent subsystems configure the
> distributor does not seem like a good idea. Given that the ADSP is a
> slave and would be under the control of the kernel via its own driver,
> it would seem best that only the kernel configures the distributors
> routing of the interrupts. This could be achieved by adding a new genirq
> API to migrate the interrupt. The GIC driver already has an API to
> migrate all interrupts from one CPU interface to another (which I
> understand is for a different reason), but having an generic API to
> migrate an interrupt to another device could be useful (unless something
> already exists that I have overlooked).
>
> Please let me know if you have any thoughts/opinions on the above.
>
> Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
> ---
> drivers/irqchip/irq-gic.c | 330 +++++++++++++++++++++++++++++++++++-----------
> 1 file changed, 253 insertions(+), 77 deletions(-)
[snip]
> +#ifdef CONFIG_PM_SLEEP
> +static int gic_resume(struct device *dev)
> +{
> + int ret;
> +
> + ret = gic_runtime_resume(dev);
> + if (ret < 0)
> + return ret;
> +
> + pm_runtime_enable(dev);
> +
> + return 0;
> +}
> +
> +static int gic_suspend(struct device *dev)
> +{
> + pm_runtime_disable(dev);
> + if (!pm_runtime_status_suspended(dev))
> + return gic_runtime_suspend(dev);
> +
> + return 0;
> +}
> +#endif
> +
> +static const struct dev_pm_ops gic_pm_ops = {
> + SET_RUNTIME_PM_OPS(gic_runtime_suspend,
> + gic_runtime_resume, NULL)
> + SET_SYSTEM_SLEEP_PM_OPS(gic_suspend, gic_resume)
> +};
I believe these need to be the noirq variants. Will fix.
Jon
next prev parent reply other threads:[~2015-12-17 10:59 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-17 10:48 [RFC PATCH V2 0/8] Add support for Tegra210 AGIC Jon Hunter
2015-12-17 10:48 ` [RFC PATCH V2 1/8] irqdomain: Ensure type settings match for an existing mapping Jon Hunter
2015-12-17 13:16 ` Linus Walleij
2015-12-18 10:10 ` Jon Hunter
2015-12-22 9:58 ` Linus Walleij
2015-12-22 10:00 ` Linus Walleij
2015-12-22 11:27 ` Jon Hunter
2015-12-22 11:31 ` Grygorii Strashko
2015-12-17 10:48 ` [RFC PATCH V2 2/8] irqdomain: Don't set type when mapping an IRQ Jon Hunter
2015-12-17 12:18 ` Linus Walleij
2015-12-22 11:18 ` Grygorii Strashko
2015-12-22 11:56 ` Jon Hunter
2016-03-18 9:16 ` Geert Uytterhoeven
2015-12-17 10:48 ` [RFC PATCH V2 3/8] genirq: Add runtime power management support for IRQ chips Jon Hunter
2015-12-17 13:19 ` Linus Walleij
2015-12-18 10:20 ` Jon Hunter
2016-01-12 18:40 ` Grygorii Strashko
2016-01-12 21:43 ` Sören Brinkmann
2016-01-18 14:47 ` Ulf Hansson
2016-01-19 10:43 ` Jon Hunter
2016-01-20 15:30 ` Thomas Gleixner
2016-01-21 8:38 ` Jon Hunter
2016-01-21 12:40 ` Ulf Hansson
2016-01-21 19:51 ` Thomas Gleixner
2016-01-22 11:08 ` Ulf Hansson
2016-01-26 17:17 ` Thomas Gleixner
2016-02-05 14:37 ` Linus Walleij
2016-03-18 13:57 ` Grygorii Strashko
2015-12-17 10:48 ` [RFC PATCH V2 4/8] irqchip/gic: Don't initialise chip if mapping IO space fails Jon Hunter
2015-12-17 13:21 ` Linus Walleij
2015-12-17 10:48 ` [RFC PATCH V2 5/8] irqchip/gic: Return an error if GIC initialisation fails Jon Hunter
2015-12-17 13:26 ` Linus Walleij
2015-12-18 10:24 ` Jon Hunter
2015-12-17 10:48 ` [RFC PATCH V2 6/8] irqchip/gic: Assign irqchip dynamically Jon Hunter
2015-12-17 11:00 ` Marc Zyngier
2015-12-18 10:26 ` Jon Hunter
2015-12-17 10:48 ` [RFC PATCH V2 7/8] irqchip/gic: Prepare for adding platform driver Jon Hunter
2015-12-17 10:48 ` [RFC PATCH V2 8/8] irqchip/gic: Add support for tegra AGIC interrupt controller Jon Hunter
2015-12-17 10:58 ` Jon Hunter [this message]
2015-12-17 13:32 ` Linus Walleij
2015-12-18 10:44 ` Jon Hunter
2015-12-22 10:03 ` Linus Walleij
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=5672956C.5010904@nvidia.com \
--to=jonathanh@nvidia.com \
--cc=geert@linux-m68k.org \
--cc=grygorii.strashko@ti.com \
--cc=jason@lakedaemon.net \
--cc=jiang.liu@linux.intel.com \
--cc=khilman@kernel.org \
--cc=lars@metafoo.de \
--cc=linus.walleij@linaro.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-tegra@vger.kernel.org \
--cc=marc.zyngier@arm.com \
--cc=soren.brinkmann@xilinx.com \
--cc=swarren@wwwdotorg.org \
--cc=tglx@linutronix.de \
--cc=thierry.reding@gmail.com \
/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).