Message-ID: <20231019172347.GC7254@thinkpad>
Date:   Thu, 19 Oct 2023 22:53:47 +0530
From:   Manivannan Sadhasivam <mani@...nel.org>
To:     Frank Li <Frank.li@....com>
Cc:     Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>,
        aisheng.dong@....com, bhelgaas@...gle.com,
        devicetree@...r.kernel.org, festevam@...il.com,
        imx@...ts.linux.dev, jdmason@...zu.us, kernel@...gutronix.de,
        kishon@...nel.org, kw@...ux.com,
        linux-arm-kernel@...ts.infradead.org, linux-imx@....com,
        linux-kernel@...r.kernel.org, linux-pci@...r.kernel.org,
        lorenzo.pieralisi@....com, lpieralisi@...nel.org, maz@...nel.org,
        s.hauer@...gutronix.de, shawnguo@...nel.org, tglx@...utronix.de
Subject: Re: [PATCH v2 1/5] PCI: endpoint: Add RC-to-EP doorbell support
 using platform MSI controller

On Thu, Oct 19, 2023 at 12:00:22PM -0400, Frank Li wrote:
> On Thu, Oct 19, 2023 at 08:34:41PM +0530, Manivannan Sadhasivam wrote:
> > On Tue, Oct 17, 2023 at 02:55:57PM -0400, Frank Li wrote:
> > > On Wed, Oct 18, 2023 at 12:07:22AM +0530, Manivannan Sadhasivam wrote:
> > > > On Mon, Sep 11, 2023 at 06:09:16PM -0400, Frank Li wrote:
> > > > > This commit introduces a common method for sending messages from the Root
> > > > > Complex (RC) to the Endpoint (EP) by utilizing the platform MSI interrupt
> > > > > controller, such as ARM GIC, as an EP doorbell. Maps the memory assigned
> > > > > for the BAR region by the PCI host to the message address of the platform
> > > > > MSI interrupt controller in the PCI EP. As a result, when the PCI RC writes
> > > > 
> > > > "Doorbell feature is implemented by mapping the EP's MSI interrupt controller
> > > > message address to a dedicated BAR in the EPC core. It is the responsibility
> > > > of the EPF driver to pass the actual message data to be written by the host to
> > > > the doorbell BAR region through its own logic."
> > > > 
> > > > > to the BAR region, it triggers an IRQ at the EP. This implementation serves
> > > > > as a common method for all endpoint function drivers.
> > > > > 
> > > > > However, it currently supports only one EP physical function due to
> > > > > limitations in ARM MSI/IMS readiness.
> > > > > 
> > > > > Signed-off-by: Frank Li <Frank.Li@....com>
> > > > > ---
> > > > >  drivers/pci/endpoint/pci-epc-core.c | 192 ++++++++++++++++++++++++++++
> > > > >  drivers/pci/endpoint/pci-epf-core.c |  44 +++++++
> > > > >  include/linux/pci-epc.h             |   6 +
> > > > >  include/linux/pci-epf.h             |   7 +
> > > > >  4 files changed, 249 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> > > > > index 5a4a8b0be6262..d336a99c6a94f 100644
> > > > > --- a/drivers/pci/endpoint/pci-epc-core.c
> > > > > +++ b/drivers/pci/endpoint/pci-epc-core.c
> > > > > @@ -10,6 +10,7 @@
> > > > >  #include <linux/slab.h>
> > > > >  #include <linux/module.h>
> > > > >  
> > > > > +#include <linux/msi.h>
> > > > >  #include <linux/pci-epc.h>
> > > > >  #include <linux/pci-epf.h>
> > > > >  #include <linux/pci-ep-cfs.h>
> > > > > @@ -783,6 +784,197 @@ void pci_epc_bme_notify(struct pci_epc *epc)
> > > > >  }
> > > > >  EXPORT_SYMBOL_GPL(pci_epc_bme_notify);
> > > > >  
> > > > > +/**
> > > > > + * pci_epc_alloc_doorbell() - alloc an address space to let RC trigger EP side IRQ by write data to
> > > > > + *			      the space.
> > > > 
> > > > "Allocate platform specific doorbell IRQs to be used by the host to trigger
> > > > doorbells on EP."
> > > > 
> > > > > + *
> > > > > + * @epc: the EPC device that need doorbell address and data from RC.
> > > > 
> > > > EPC device for which the doorbell needs to be allocated
> > > > 
> > > > > + * @func_no: the physical endpoint function number in the EPC device.
> > > > > + * @vfunc_no: the virtual endpoint function number in the physical function.
> > > > > + * @num_msgs: the total number of doorbell messages
> > > > 
> > > > s/num_msgs/num_db
> > > > 
> > > > > + *
> > > > > + * Return: 0 success, other is failure
> > > > > + */
> > > > > +int pci_epc_alloc_doorbell(struct pci_epc *epc, u8 func_no, u8 vfunc_no, int num_msgs)
> > > > > +{
> > > > > +	int ret;
> > > > > +
> > > > > +	if (IS_ERR_OR_NULL(epc) || func_no >= epc->max_functions)
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	if (vfunc_no > 0 && (!epc->max_vfs || vfunc_no > epc->max_vfs[func_no]))
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	if (!epc->ops->alloc_doorbell)
> > > > > +		return 0;
> > > > 
> > > > You mentioned 0 is a success. So if there is no callback, you want to return
> > > > success?
> > > > 
> > > > > +
> > > > > +	mutex_lock(&epc->lock);
> > > > > +	ret = epc->ops->alloc_doorbell(epc, func_no, vfunc_no, num_msgs);
> > > > 
> > > > Why can't you just call the generic function here and in other places instead of
> > > > implementing callbacks? I do not see a necessity for EPC specific callbacks. If
> > > > there is one, please specify.
> > > 
> > > 1. Refer v1 your comments.
> > > https://lore.kernel.org/imx/20230906145227.GC5930@thinkpad/
> > 
> > I do not find where I suggested the callback approach.
> 
> 	> > > If that, Each EPF driver need do duplicate work. 
> 	> > > 
> 	> > 
> 	> > Yes, and that's how it should be. EPF core has no job in supplying the of_node.
> 	> > It is the responsibility of the EPF drivers as they depend on OF for platform
> 	> > support.
> 	> 
> 	> EPF driver still not depend on OF. such pci-epf-test, which was probed by
> 	> configfs.
> 	> 
> 
> 	Hmm, yeah. Then it should be part of the EPC driver.
> 
> 	Sorry for the confusion.
> 
> Here, all "EPF" should be "EPC". The key problem is of_node. EPC core have
> not of_node, EPC core's parent driver (like dwc-ep driver) have of_node. 
> 
> pci_epc_generic_alloc_doorbell(dev), dev is probed by platform driver, such
> as dwc-ep, which have of_node,  EPC core will create child device.
> 
> dwc-ep device
>  |- epc core device
> 
> we can direct call pci_epc_generic_alloc_doorbell(epc->parent) here.
> 
> I may miss understand what your means. I think you want to dwc-ep
> (with of_node) handle these alloc functions. 
> 

My comment was to have just one function definition. But looking at it again, I
think it is better to move all the (alloc, free, write_msg) definitions to
dwc-ep, since the contents of those functions are not EPC core specific.

In the EPC core, you can still have the callbacks specific to each EPC. This
also solves your of_node problem.

- Mani

> > 
> > > 2. Maybe some ep controller have built-in doorbell support. Write to some
> > > address to trigger doorbell irq.
> > > 
> > 
> > We will handle it whenever such EP controllers arrive. Until then, let's keep it
> > simple.
> > 
> > - Mani
> > 
> > > Frank
> > > 
> > > > 
> > > > > +	mutex_unlock(&epc->lock);
> > > > > +
> > > > > +	return ret;
> > > > > +}
> > > > > +EXPORT_SYMBOL_GPL(pci_epc_alloc_doorbell);
> > > > > +
> > > > > +/**
> > > > > + * pci_epc_free_doorbell() - free resource allocated by pci_epc_alloc_doorbell()
> > > > > + *
> > > > > + * @epc: the EPC device that need doorbell address and data from RC.
> > > > 
> > > > Same as above.
> > > > 
> > > > > + * @func_no: the physical endpoint function number in the EPC device.
> > > > > + * @vfunc_no: the virtual endpoint function number in the physical function.
> > > > > + *
> > > > > + * Return: 0 success, other is failure
> > > > > + */
> > > > > +void pci_epc_free_doorbell(struct pci_epc *epc, u8 func_no, u8 vfunc_no)
> > > > > +{
> > > > > +	if (IS_ERR_OR_NULL(epc) || func_no >= epc->max_functions)
> > > > > +		return;
> > > > > +
> > > > > +	if (vfunc_no > 0 && (!epc->max_vfs || vfunc_no > epc->max_vfs[func_no]))
> > > > > +		return;
> > > > > +
> > > > > +	if (!epc->ops->free_doorbell)
> > > > > +		return;
> > > > > +
> > > > > +	mutex_lock(&epc->lock);
> > > > > +	epc->ops->free_doorbell(epc, func_no, vfunc_no);
> > > > 
> > > > Same as suggested above.
> > > > 
> > > > > +	mutex_unlock(&epc->lock);
> > > > > +}
> > > > > +EXPORT_SYMBOL_GPL(pci_epc_free_doorbell);
> > > > > +
> > > > > +static irqreturn_t pci_epf_generic_doorbell_handler(int irq, void *data)
> > > > > +{
> > > > > +	struct pci_epf *epf = data;
> > > > > +
> > > > > +	if (epf->event_ops && epf->event_ops->doorbell)
> > > > > +		epf->event_ops->doorbell(epf, irq - epf->virq_base);
> > > > 
> > > > Same as suggested above.
> > > > 
> > > > > +
> > > > > +	return IRQ_HANDLED;
> > > > > +}
> > > > > +
> > > > > +static void pci_epc_generic_write_msi_msg(struct msi_desc *desc, struct msi_msg *msg)
> > > > > +{
> > > > > +	struct pci_epc *epc = NULL;
> > > > > +	struct class_dev_iter iter;
> > > > > +	struct pci_epf *epf;
> > > > > +	struct device *dev;
> > > > > +
> > > > > +	class_dev_iter_init(&iter, pci_epc_class, NULL, NULL);
> > > > > +	while ((dev = class_dev_iter_next(&iter))) {
> > > > > +		if (dev->parent != desc->dev)
> > > > > +			continue;
> > > > > +
> > > > > +		epc = to_pci_epc(dev);
> > > > > +
> > > > > +		class_dev_iter_exit(&iter);
> > > > > +		break;
> > > > > +	}
> > > > > +
> > > > > +	if (!epc)
> > > > > +		return;
> > > > > +
> > > > > +	/* Only support one EPF for doorbell */
> > > > > +	epf = list_first_entry_or_null(&epc->pci_epf, struct pci_epf, list);
> > > > > +
> > > > 
> > > > No need of this newline
> > > > 
> > > > > +	if (!epf)
> > > > > +		return;
> > > > > +
> > > > > +	if (epf->msg && desc->msi_index < epf->num_msgs)
> > > > > +		epf->msg[desc->msi_index] = *msg;
> > > > > +}
> > > > > +
> > > > > +
> > > > 
> > > > Remove extra newline
> > > > 
> > > > > +/**
> > > > > + * pci_epc_generic_alloc_doorbell() - Common help function. Allocate address space from MSI
> > > > > + *                                    controller
> > > > > + *
> > > > > + * @epc: the EPC device that need doorbell address and data from RC.
> > > > > + * @func_no: the physical endpoint function number in the EPC device.
> > > > > + * @vfunc_no: the virtual endpoint function number in the physical function.
> > > > > + * @num_msgs: the total number of doorbell messages
> > > > > + *
> > > > 
> > > > Same comment as for pci_epc_alloc_doorbell()
> > > > 
> > > > > + * Remark: use this function only if EPC driver just register one EPC device.
> > > > > + *
> > > > > + * Return: 0 success, other is failure
> > > > > + */
> > > > > +int pci_epc_generic_alloc_doorbell(struct pci_epc *epc, u8 func_no, u8 vfunc_no, int num_msgs)
> > > > > +{
> > > > > +	struct pci_epf *epf;
> > > > > +	struct device *dev;
> > > > > +	int virq, last;
> > > > > +	int ret;
> > > > > +	int i;
> > > > > +
> > > > > +	if (IS_ERR_OR_NULL(epc))
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	/* Currently only support one func and one vfunc for doorbell */
> > > > > +	if (func_no || vfunc_no)
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	epf = list_first_entry_or_null(&epc->pci_epf, struct pci_epf, list);
> > > > > +	if (!epf)
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	dev = epc->dev.parent;
> > > > > +	ret = platform_msi_domain_alloc_irqs(dev, num_msgs, pci_epc_generic_write_msi_msg);
> > > > > +	if (ret) {
> > > > > +		dev_err(dev, "Failed to allocate MSI\n");
> > > > > +		return -ENOMEM;
> > > > > +	}
> > > > > +
> > > > > +	last = -1;
> > > > > +	for (i = 0; i < num_msgs; i++) {
> > > > 
> > > > You should iterate over msi_desc as below:
> > > > 
> > > >         msi_lock_descs(dev);
> > > >         msi_for_each_desc(desc, dev, MSI_DESC_ALL) {
> > > > 		...
> > > > 	}
> > > > 	msi_unlock_descs(dev);
> > > > 
> > > > > +		virq = msi_get_virq(dev, i);
> > > > > +		if (i == 0)
> > > > > +			epf->virq_base = virq;
> > > > > +
> > > > > +		ret = request_irq(virq, pci_epf_generic_doorbell_handler, 0,
> > > > 
> > > > 	request_irq(desc->irq, ...)
> > > > 
> > > > > +				  kasprintf(GFP_KERNEL, "pci-epc-doorbell%d", i), epf);
> > > > > +
> > > > > +		if (ret) {
> > > > > +			dev_err(dev, "Failed to request doorbell\n");
> > > > > +			goto err_free_irq;
> > > > > +		}
> > > > > +		last = i;
> > > > > +	}
> > > > > +
> > > > > +	return 0;
> > > > > +
> > > > > +err_free_irq:
> > > > > +	for (i = 0; i < last; i++)
> > > > > +		kfree(free_irq(epf->virq_base + i, epf));
> > > > > +	platform_msi_domain_free_irqs(dev);
> > > > > +
> > > > > +	return -EINVAL;
> > > > 
> > > > 	return ret;
> > > > 
> > > > > +}
> > > > > +EXPORT_SYMBOL_GPL(pci_epc_generic_alloc_doorbell);
> > > > > +
> > > > 
> > > > [...]
> > > > 
> > > > > diff --git a/include/linux/pci-epf.h b/include/linux/pci-epf.h
> > > > > index 3f44b6aec4770..485c146a5efe2 100644
> > > > > --- a/include/linux/pci-epf.h
> > > > > +++ b/include/linux/pci-epf.h
> > > > > @@ -79,6 +79,7 @@ struct pci_epc_event_ops {
> > > > >  	int (*link_up)(struct pci_epf *epf);
> > > > >  	int (*link_down)(struct pci_epf *epf);
> > > > >  	int (*bme)(struct pci_epf *epf);
> > > > > +	int (*doorbell)(struct pci_epf *epf, int index);
> > > > 
> > > > kdoc missing.
> > > > 
> > > > >  };
> > > > >  
> > > > >  /**
> > > > > @@ -180,6 +181,9 @@ struct pci_epf {
> > > > >  	unsigned long		vfunction_num_map;
> > > > >  	struct list_head	pci_vepf;
> > > > >  	const struct pci_epc_event_ops *event_ops;
> > > > > +	struct msi_msg *msg;
> > > > > +	u16 num_msgs;
> > > > 
> > > > num_db
> > > > 
> > > > You also need to add kdoc for each new member.
> > > > 
> > > > - Mani
> > > > 
> > > > -- 
> > > > மணிவண்ணன் சதாசிவம்
> > 
> > -- 
> > மணிவண்ணன் சதாசிவம்

-- 
மணிவண்ணன் சதாசிவம்

Powered by blists - more mailing lists