[RFC PATCH 1/2] irqchip: cdx-bus: add cdx-MSI domain with gic-its domain as parent

[Gupta, Nipun]

[AMD Official Use Only - General]



> -----Original Message-----
> From: Marc Zyngier <maz at kernel.org>
> Sent: Thursday, August 4, 2022 2:20 PM
> To: Gupta, Nipun <Nipun.Gupta at amd.com>
> Cc: linux-kernel at vger.kernel.org; linux-arm-kernel at lists.infradead.org;
> gregkh at linuxfoundation.org; rafael at kernel.org; tglx at linutronix.de;
> okaya at kernel.org; Anand, Harpreet <harpreet.anand at amd.com>; Simek,
> Michal <michal.simek at amd.com>; Agarwal, Nikhil
> <nikhil.agarwal at amd.com>
> Subject: Re: [RFC PATCH 1/2] irqchip: cdx-bus: add cdx-MSI domain with gic-
> its domain as parent
> 
> [CAUTION: External Email]
> 
> On Wed, 03 Aug 2022 13:26:54 +0100,
> Nipun Gupta <mailto:nipun.gupta at amd.com> wrote:
> >
> > Devices on cdx bus are dynamically detected and registered using
> > platform_device_register API. As these devices are not linked to
> > of node they need a separate MSI domain for handling device ID
> > to be provided to the GIC ITS domain.
> >
> > Signed-off-by: Nipun Gupta <mailto:nipun.gupta at amd.com>
> > Signed-off-by: Nikhil Agarwal <mailto:nikhil.agarwal at amd.com>
> > ---
> > CONFIG_CDX_BUS and device tree bindings for xlnx,cdx-controller-1.0
> > would be added as part of CDX bus patches
> >
> >  drivers/irqchip/Makefile                 |   1 +
> >  drivers/irqchip/irq-gic-v3-its-cdx-msi.c | 113 +++++++++++++++++++++++
> >  include/linux/cdx/cdx.h                  |  15 +++
> >  3 files changed, 129 insertions(+)
> >  create mode 100644 drivers/irqchip/irq-gic-v3-its-cdx-msi.c
> >  create mode 100644 include/linux/cdx/cdx.h
> >
> > diff --git a/drivers/irqchip/Makefile b/drivers/irqchip/Makefile
> > index 5b67450a9538..623adb8a1f20 100644
> > --- a/drivers/irqchip/Makefile
> > +++ b/drivers/irqchip/Makefile
> > @@ -115,3 +115,4 @@ obj-$(CONFIG_WPCM450_AIC)         += irq-
> wpcm450-aic.o
> >  obj-$(CONFIG_IRQ_IDT3243X)           += irq-idt3243x.o
> >  obj-$(CONFIG_APPLE_AIC)                      += irq-apple-aic.o
> >  obj-$(CONFIG_MCHP_EIC)                       += irq-mchp-eic.o
> > +obj-$(CONFIG_CDX_BUS)                        += irq-gic-v3-its-cdx-msi.o
> > diff --git a/drivers/irqchip/irq-gic-v3-its-cdx-msi.c b/drivers/irqchip/irq-gic-
> v3-its-cdx-msi.c
> > new file mode 100644
> > index 000000000000..eb17b74efdc5
> > --- /dev/null
> > +++ b/drivers/irqchip/irq-gic-v3-its-cdx-msi.c
> > @@ -0,0 +1,113 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * AMD CDX bus driver MSI support
> > + *
> > + * Copyright(C) 2022 Xilinx Inc.
> > + */
> > +
> > +#include <linux/irq.h>
> > +#include <linux/msi.h>
> > +#include <linux/of.h>
> > +#include <linux/of_address.h>
> > +#include <linux/of_device.h>
> > +#include <linux/of_irq.h>
> > +#include <linux/cdx/cdx.h>
> > +
> > +static struct irq_chip its_msi_irq_chip = {
> > +     .name = "ITS-fMSI",
> > +     .irq_mask = irq_chip_mask_parent,
> > +     .irq_unmask = irq_chip_unmask_parent,
> > +     .irq_eoi = irq_chip_eoi_parent,
> > +     .irq_set_affinity = msi_domain_set_affinity
> > +};
> > +
> > +static int its_cdx_msi_prepare(struct irq_domain *msi_domain,
> > +                               struct device *dev,
> > +                               int nvec, msi_alloc_info_t *info)
> > +{
> > +     struct msi_domain_info *msi_info;
> > +     struct cdx_device_data *dev_data;
> > +     u32 dev_id;
> > +
> > +     dev_data = dev->platform_data;
> > +     dev_id = dev_data->dev_id;
> > +
> > +     /* Set the device Id to be passed to the GIC-ITS */
> > +     info->scratchpad[0].ul = dev_id;
> > +
> > +     msi_info = msi_get_domain_info(msi_domain->parent);
> > +
> > +     /* Allocate at least 32 MSIs, and always as a power of 2 */
> > +     nvec = max_t(int, 32, roundup_pow_of_two(nvec));
> > +     return msi_info->ops->msi_prepare(msi_domain->parent, dev, nvec,
> info);
> > +}
> > +
> > +static struct msi_domain_ops its_cdx_msi_ops __ro_after_init = {
> > +     .msi_prepare = its_cdx_msi_prepare,
> > +};
> > +
> > +static struct msi_domain_info its_cdx_msi_domain_info = {
> > +     .flags  = (MSI_FLAG_USE_DEF_DOM_OPS |
> MSI_FLAG_USE_DEF_CHIP_OPS),
> > +     .ops    = &its_cdx_msi_ops,
> > +     .chip   = &its_msi_irq_chip,
> > +};
> > +
> > +static const struct of_device_id cdx_device_id[] = {
> > +     {.compatible = "xlnx,cdx-controller-1.0", },
> 
> What is this? If this is supposed to represent am ITS, it really
> should say so.

This is a CDX bus controller, not an ITS. This will be added as a part of
device-tree documentation when we add formal CDX bus patches.
CDX is an upcoming AMD bus, supporting dynamically discovered
FPGA devices.

> 
> > +     {},
> > +};
> > +
> > +struct irq_domain *get_parent(struct fwnode_handle *handle)
> > +{
> > +     return irq_find_matching_fwnode(handle, DOMAIN_BUS_NEXUS);
> > +}
> > +
> > +static void __init its_cdx_msi_init_one(struct device_node *np,
> > +                                 const char *name)
> > +{
> > +     struct irq_domain *parent;
> > +     struct irq_domain *cdx_msi_domain;
> > +     struct fwnode_handle *fwnode_handle;
> > +     struct device_node *parent_node;
> > +
> > +     parent_node = of_parse_phandle(np, "msi-parent", 0);
> > +
> > +     parent = get_parent(of_node_to_fwnode(parent_node));
> > +     if (!parent || !msi_get_domain_info(parent)) {
> > +             pr_err("%s: unable to locate ITS domain\n", name);
> > +             return;
> > +     }
> > +
> > +     fwnode_handle = of_node_to_fwnode(np);
> > +     cdx_msi_domain = platform_msi_create_irq_domain(fwnode_handle,
> > +                                             &its_cdx_msi_domain_info,
> > +                                             parent);
> > +     if (!cdx_msi_domain) {
> > +             pr_err("%s: unable to create cdx bus domain\n", name);
> > +             return;
> > +     }
> > +
> > +     pr_info("cdx bus MSI: %s domain created\n", name);
> > +}
> > +
> > +static void __init its_cdx_of_msi_init(void)
> > +{
> > +     struct device_node *np;
> > +
> > +     for (np = of_find_matching_node(NULL, cdx_device_id); np;
> > +          np = of_find_matching_node(np, cdx_device_id)) {
> > +             if (!of_device_is_available(np))
> > +                     continue;
> > +
> > +             its_cdx_msi_init_one(np, np->full_name);
> > +     }
> > +}
> > +
> > +static int __init its_cdx_msi_init(void)
> > +{
> > +     its_cdx_of_msi_init();
> > +
> > +     return 0;
> > +}
> > +
> > +early_initcall(its_cdx_msi_init);
> 
> I really don't think we should have any more of this muck. Yes, the
> other busses are also doing that, but here's our chance to do
> something better.
> 
> Why can't the *bus* driver (I assume there is one) perform these
> tasks? It would really help if this patch was shown in context,
> because I have no idea how this fits in the grand scheme of things.

Agree, bus driver could do this stuff, and I was considering it too, but as this
was the norm so added these changes in this IRQ chip :).

Bus driver is not out yet, and we have pushed these as RFC changes for
early feedback (mentioned in the cover letter), but yes, these patches would
be clubbed as a part of overall changes.

And we can move it as a part of bus driver itself. I hope that suits?

Thanks,
Nipun

> 
> Thanks,
> 
>         M.
> 
> --
> Without deviation from the norm, progress is not possible.