Hello Marc,
On Wed, Dec 01, 2021 at 11:41:02AM +0000, Marc Zyngier wrote:
> Since 041284181226 ("of/irq: Allow matching of an interrupt-map local
> to an interrupt controller"), a handful of interrupt controllers have
> stopped working correctly. This is due to the DT exposing a non-sensical
> interrupt-map property, and their drivers relying on the kernel ignoring
> this property.
>
> Since we cannot realistically fix this terrible behaviour, add a quirk
> for the limited set of devices that have implemented this monster,
> and document that this is a pretty bad practice.
>
> Cc: Rob Herring <robh@xxxxxxxxxx>
> Cc: John Crispin <john@xxxxxxxxxxx>
> Cc: Biwen Li <biwen.li@xxxxxxx>
> Cc: Chris Brandt <chris.brandt@xxxxxxxxxxx>
> Cc: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> Cc: Sander Vanheule <sander@xxxxxxxxxxxxx>
> Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
> ---
>
> Notes:
> v2: Switched over to of_device_compatible_match() as per Rob's
> request.
>
> drivers/of/irq.c | 28 ++++++++++++++++++++++++++--
> 1 file changed, 26 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/of/irq.c b/drivers/of/irq.c
> index b10f015b2e37..65a325aad984 100644
> --- a/drivers/of/irq.c
> +++ b/drivers/of/irq.c
> @@ -76,6 +76,26 @@ struct device_node *of_irq_find_parent(struct device_node *child)
> }
> EXPORT_SYMBOL_GPL(of_irq_find_parent);
>
> +/*
> + * These interrupt controllers abuse interrupt-map for unspeakable
> + * reasons and rely on the core code to *ignore* it (the drivers do
> + * their own parsing of the property).
> + *
> + * If you think of adding to the list for something *new*, think
> + * again. There is a high chance that you will be sent back to the
> + * drawing board.
> + */
> +static const char * const of_irq_imap_abusers[] = {
> + "CBEA,platform-spider-pic",
> + "sti,platform-spider-pic",
> + "realtek,rtl-intc",
> + "fsl,ls1021a-extirq",
> + "fsl,ls1043a-extirq",
> + "fsl,ls1088a-extirq",
> + "renesas,rza1-irqc",
> + NULL,
> +};
> +
> /**
> * of_irq_parse_raw - Low level interrupt tree parsing
> * @addr: address specifier (start of "reg" property of the device) in be32 format
> @@ -159,12 +179,16 @@ int of_irq_parse_raw(const __be32 *addr, struct of_phandle_args *out_irq)
> /*
> * Now check if cursor is an interrupt-controller and
> * if it is then we are done, unless there is an
> - * interrupt-map which takes precedence.
> + * interrupt-map which takes precedence if we're not
> + * in presence of once of these broken platform that
> + * want to parse interrupt-map themselves for $reason.
> */
> bool intc = of_property_read_bool(ipar, "interrupt-controller");
> + bool imap_abuse;
>
> imap = of_get_property(ipar, "interrupt-map", &imaplen);
> - if (imap == NULL && intc) {
> + imap_abuse = imap && of_device_compatible_match(ipar, of_irq_imap_abusers);
> + if (intc && (imap == NULL || imap_abuse)) {
> pr_debug(" -> got it !\n");
> return 0;
> }
> --
> 2.30.2
>
I am a user of the ls-extirq driver which is responsible for 3 of the 7
compatible strings mentioned by you here. I have close to zero knowledge
of the irq subsystem, although I am looking forward to learn.
Could you please spend a few minutes to detail what you see as a possible
path forward for this driver? I am getting mixed impressions about what
it's doing wrong.
On one hand, it was requested by Rob during review that what used to be
called "fsl,extirq-map" should be named "interrupt-map" instead:
https://lore.kernel.org/lkml/20190928092331.GB1894@xxxxxxxxxxxxx/
Then, you seem to suggest something's wrong with drivers privately using
that name and parsing a property which used to be ignored by the core,
due to your "silly-interrupt-map" comment:
https://lore.kernel.org/all/9c169aad-3c7b-2ffb-90a2-1ca791a3f411@xxxxxxxxxxx/T/#ebae8f9231296dc936cb7c9791218fc6785a03390
Then, Rob breaks the ls-extirq driver for platforms that have a GIC ITS*
defined in the device tree via commit 869f0ec048dc ("arm64: dts:
freescale: Fix 'interrupt-map' parent address cells") - this is also,
incidentally, the reason why I'm here.
* because the driver doesn't parse the "standard" format where the
interrupt parent has a non-zero #address-cells - which the "arm,gic-v3"
may have when there's a "arm,gic-v3-its" under it (although I don't
necessarily see the relevance of the ITS being there to the needs of
the ls-extirq - which are just a bijective mapping of IRQs - this
driver simply drives a multi-channel logical inverter).
So if I understand correctly, we keep ignoring the non-standard use of
the "interrupt-map" property in these abuser drivers, yet we patch their
device trees to have a more standard format in their non-standard use? :)
Since some breakage has already been introduced, for good or bad, I
think we can start discussing how things should have been done from the
beginning, and see if we can make those changes now.
