[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160118114931.GJ21067@leverpostej>
Date: Mon, 18 Jan 2016 11:49:32 +0000
From: Mark Rutland <mark.rutland@....com>
To: Sinan Kaya <okaya@...eaurora.org>
Cc: dmaengine@...r.kernel.org, timur@...eaurora.org,
devicetree@...r.kernel.org, cov@...eaurora.org,
vinod.koul@...el.com, jcm@...hat.com, agross@...eaurora.org,
arnd@...db.de, linux-arm-msm@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH V12 2/7] dma: hidma: Add Device Tree support
> >> +Main node required properties:
> >> +- compatible: "qcom,hidma-mgmt-1.0";
> >> +- reg: Address range for DMA device
> >> +- dma-channels: Number of channels supported by this DMA controller.
> >> +- max-write-burst-bytes: Maximum write burst in bytes. A memcpy requested is
> >> + fragmented to multiples of this amount.
> >> +- max-read-burst-bytes: Maximum read burst in bytes. A memcpy request is
> >> + fragmented to multiples of this amount.
> >> +- max-write-transactions: Maximum write transactions to perform in a burst
> >> +- max-read-transactions: Maximum read transactions to perform in a burst
> >
> > Just to check, where do these max-* values come from?
> These are HW bus parameters like the burst count and
> size of each burst. These values change based on the SoC this IP is in use.
>
> >
> > Are they some correctness requirement of the bus this is attached to?
> You can starve other peripherals if you use incorrect values as the bus is
> shared with other peripherals. Yes, correctness is required.
Is that a property of the system known statically, or one determined by
testing the system under particular workloads? It feels like the latter
(though I appreciate that not starving other masters is certainly a
correctness property regardless of how this is derived).
I'd have expected the bus this is plugged into to have appropriate QoS
settings pre-configured so as to avoid starvation, though it sounds like
that's not possible here?
> > Are they tuning values?
> Correct value is necessary for functioning. I'd consider weight and priority
> as the only tuning parameters.
>
> >
> > The latter doesn't really belong in the DT. Given they're writeable from
> > the driver, it seems like that's what they are...
>
> Good catch. Those should have been read-only. I wanted to be able to export these
> information to the userspace app. I'll fix the sysfs to make them read-only.
>
> >
> >> +- channel-reset-timeout-cycles: Channel reset timeout in cycles for this SOC.
> >
> > I'm not sure what this means. Could you elaborate on this is?
> After each reset command, HW starts a timer. This is the time HW waits before it declares
> reset failed.
Is that a reset command sent to the HIDMA by the OS, or a reset command
from the HIDMA to something else?
What does it do when it declares a reset as failed?
How can the OS make use of this information? It has no idea of the
clocks input to the HIDMA, so it has no idea how long a cycle is.
Is this programmed by the OS?
Is the particular duration in cycles a requirement of some other agent?
Thanks,
Mark.
Powered by blists - more mailing lists