LKML Archive on
help / color / mirror / Atom feed
From: Arnd Bergmann <>
To: Vladimir Oltean <>
Cc: Arnd Bergmann <>,
	Vladimir Oltean <>,
	Mark Brown <>,
	Rasmus Villemoes <>,
	Lee Jones <>,
	Thomas Gleixner <>,
	Marc Zyngier <>,
	Hou Zhiqiang <>, Biwen Li <>,
	Greg Kroah-Hartman <>,
	"Rafael J. Wysocki" <>,
	Linux Kernel Mailing List <>
Subject: Re: [PATCH 2/2] mfd: syscon: request a regmap with raw spinlocks for some devices
Date: Thu, 26 Aug 2021 11:24:24 +0200	[thread overview]
Message-ID: <> (raw)
In-Reply-To: <20210825220023.rqskspy2usvleoqr@skbuf>

On Thu, Aug 26, 2021 at 12:01 AM Vladimir Oltean <> wrote:
> On Wed, Aug 25, 2021 at 11:24:52PM +0200, Arnd Bergmann wrote:
> > Are there any other users of the syscon?
> Not that I can see, but that doesn't make the fact that it uses a syscon a bad decision.
> For context, Layerscape devices have a "Misc" / "And Others" memory region
> called "Supplemental Configuration Unit" (SCFG) which "provides the
> chip-specific configuration and status registers for the device. It is the
> chip-defined module for extending the device configuration unit (DCFG) module."
> to quote the documentation.
> The ls-extirq file is a driver around _a_single_register_ of SCFG. SCFG
> provides an option of reversing the interrupt polarity of the external IRQ
> pins: make them active-low instead of active-high, or rising instead of
> falling.
> The reason for the existence of the driver is that we got some pushback during
> device tree submission: while we could describe in the device tree an interrupt
> as "active-high" and going straight to the GIC, in reality that interrupt is
> "active-low" but inverted by the SCFG (the inverted is enabled by default).
> Additionally, the GIC cannot process active-low interrupts in the first place
> AFAIR, which is why an inverter exists in front of it.
> Some other SCFG registers are (at least on LS1021A):
> Deep Sleep Control Register
> eTSEC Clock Deep Sleep Control Register (eTSEC is our Ethernet controller)
> Pixel Clock Control Register
> PCIe PM Write Control Register
> PCIe PM Read Control Register
> USB3 parameter 1 control register
> QuadSPI configuration
> Endianness Control Register
> Snoop configuration
> Interrupt Polarity <- this is the register controlled by ls-extirq
> etc etc.
> Also, even if you were to convince me that we shouldn't use a syscon, I feel
> that the implication (change the device trees for 7 SoCs) just to solve a
> kernel splat would be like hitting a nail with an atomic bomb. I'm not going to
> do it.

I was not suggesting changing the DT files. The way we describe syscon
devices is generally meant to allow replacing them with a custom driver
as an implementation detail of the OS, you just have a driver that binds
against the more specific compatible string as opposed to the generic
compatible="syscon" check, and you replace all
syscon_regmap_lookup_by_phandle() calls with direct function calls
into exported symbols from that driver that perform high-level functions.

In this particular case, I think a high-level interface from a drviers/soc/
driver works just as well as the syscon method if there was raw_spinlock
requirement, but with the irqchip driver needing the regmap, the custom
driver would a better interface.


  reply	other threads:[~2021-08-26  9:24 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-25 20:50 [PATCH 0/2] Use raw spinlocks in the ls-extirq driver Vladimir Oltean
2021-08-25 20:50 ` [PATCH 1/2] regmap: teach regmap to use raw spinlocks if requested in the config Vladimir Oltean
2021-08-26 23:01   ` Thomas Gleixner
2021-08-27 16:12     ` Vladimir Oltean
2021-08-27 19:59       ` Thomas Gleixner
2021-08-30 10:49         ` Vladimir Oltean
2021-08-30 11:02     ` Rasmus Villemoes
2021-08-30 12:42       ` Mark Brown
2021-08-30 12:19     ` Mark Brown
2021-08-30 14:16       ` Thomas Gleixner
2021-09-01 16:05         ` Mark Brown
2021-09-02  8:35           ` Thomas Gleixner
2021-08-25 20:50 ` [PATCH 2/2] mfd: syscon: request a regmap with raw spinlocks for some devices Vladimir Oltean
2021-08-25 21:24   ` Arnd Bergmann
2021-08-25 22:00     ` Vladimir Oltean
2021-08-26  9:24       ` Arnd Bergmann [this message]
2021-08-26 13:57         ` Vladimir Oltean
2021-08-26 14:48           ` Arnd Bergmann
2021-09-06  9:18   ` Lee Jones
2021-08-26 12:51 ` (subset) [PATCH 0/2] Use raw spinlocks in the ls-extirq driver Mark Brown

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:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

* 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).