Message-ID: <20241105173035.GE1807686@google.com>
Date: Tue, 5 Nov 2024 17:30:35 +0000
From: Lee Jones <lee@...nel.org>
To: Robert Joslyn <robert_joslyn@...inc.com>
Cc: linux-kernel@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: [RFC PATCH 1/2] mfd: Add SEL PCI Virtual Multifunction (PVMF)
 support

On Mon, 28 Oct 2024, Robert Joslyn wrote:

> Add support for SEL FPGA based PCIe devices. These expose a single PCI
> BAR with multiple IP cores for various functions, such as serial ports,
> ethernet, and time (PTP/IRIG). This initial driver supports Ethernet on
> the SEL-3350 mainboard, SEL-3390E4 ethernet card, and SEL-3390T ethernet
> and time card.
> 
> Signed-off-by: Robert Joslyn <robert_joslyn@...inc.com>
> ---
>  MAINTAINERS                |   8 +
>  drivers/mfd/Kconfig        |  16 ++
>  drivers/mfd/Makefile       |   3 +
>  drivers/mfd/selpvmf-core.c | 482 +++++++++++++++++++++++++++++++++++++
>  drivers/mfd/selpvmf-cvp.c  | 431 +++++++++++++++++++++++++++++++++
>  drivers/mfd/selpvmf-cvp.h  |  18 ++
>  6 files changed, 958 insertions(+)
>  create mode 100644 drivers/mfd/selpvmf-core.c
>  create mode 100644 drivers/mfd/selpvmf-cvp.c
>  create mode 100644 drivers/mfd/selpvmf-cvp.h
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index a27407950242..4a24b3be8aa5 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -20730,6 +20730,14 @@ F:	tools/testing/selftests/lsm/
>  X:	security/selinux/
>  K:	\bsecurity_[a-z_0-9]\+\b
>  
> +SEL DRIVERS
> +M:	Robert Joslyn <robert_joslyn@...inc.com>
> +S:	Supported
> +B:	mailto:opensource@...inc.com
> +F:	drivers/mfd/selpvmf-*
> +F:	drivers/platform/x86/sel3350-platform.c
> +F:	include/linux/mfd/selpvmf.h
> +
>  SELINUX SECURITY MODULE
>  M:	Paul Moore <paul@...l-moore.com>
>  M:	Stephen Smalley <stephen.smalley.work@...il.com>
> diff --git a/drivers/mfd/Kconfig b/drivers/mfd/Kconfig
> index f9325bcce1b9..77e2ce3db505 100644
> --- a/drivers/mfd/Kconfig
> +++ b/drivers/mfd/Kconfig
> @@ -2402,5 +2402,21 @@ config MFD_RSMU_SPI
>  	  Additional drivers must be enabled in order to use the functionality
>  	  of the device.
>  
> +config MFD_SELPVMF
> +	tristate "SEL PVMF Support"
> +	help
> +	  Support for SEL PCI virtual multifunction (PVMF) devcies utilizing
> +	  an FPGA based PCIe interface, including:
> +	    * SEL-3350
> +	    * SEL-3390E4
> +	    * SEL-3390T
> +
> +	  This driver provides common support for accessing the device.
> +	  Additional drivers must be enabled in order to use the functionality
> +	  of the device.
> +
> +	  This driver can be built as a module. If built as a module it will be
> +	  called "selpvmf".
> +
>  endmenu
>  endif
> diff --git a/drivers/mfd/Makefile b/drivers/mfd/Makefile
> index 2a9f91e81af8..198e43ef7a51 100644
> --- a/drivers/mfd/Makefile
> +++ b/drivers/mfd/Makefile
> @@ -289,3 +289,6 @@ obj-$(CONFIG_MFD_ATC260X_I2C)	+= atc260x-i2c.o
>  
>  obj-$(CONFIG_MFD_RSMU_I2C)	+= rsmu_i2c.o rsmu_core.o
>  obj-$(CONFIG_MFD_RSMU_SPI)	+= rsmu_spi.o rsmu_core.o
> +
> +selpvmf-objs			:= selpvmf-core.o selpvmf-cvp.o
> +obj-$(CONFIG_MFD_SELPVMF)	+= selpvmf.o
> diff --git a/drivers/mfd/selpvmf-core.c b/drivers/mfd/selpvmf-core.c
> new file mode 100644
> index 000000000000..ec3e65fe8064
> --- /dev/null
> +++ b/drivers/mfd/selpvmf-core.c
> @@ -0,0 +1,482 @@
> +// SPDX-License-Identifier: GPL-2.0-only OR BSD-3-Clause
> +/*
> + * Copyright 2017 Schweitzer Engineering Laboratories, Inc.

No changes since 2017?

Why upstream this now?

> + * 2350 NE Hopkins Court, Pullman, WA 99163 USA
> + * SEL Open Source <opensource@...inc.com>
> + */
> +
> +#include <linux/module.h>
> +#include <linux/pci.h>
> +#include <linux/property.h>
> +#include <linux/mfd/core.h>
> +
> +#include "selpvmf-cvp.h"
> +
> +#define PCI_VENDOR_ID_SEL	0x1aa9
> +#define PCI_DEVICE_ID_B1190	0x0014 /* SEL-3390T */
> +#define PCI_DEVICE_ID_B2093	0x0015 /* SEL-3350 mainboard */
> +#define PCI_DEVICE_ID_B2077	0x0018 /* SEL-3390E4 */
> +#define PCI_DEVICE_ID_B2091	0x0019 /* SEL-2241-2 */
> +
> +#define B1190_NUM_ETHERNET 2
> +#define B2077_NUM_ETHERNET 4
> +#define B2091_NUM_ETHERNET 5
> +#define B2093_NUM_ETHERNET 5
> +#define SELETH_NUM_RESOURCES 4
> +
> +#define SELETH_LEN	0x1000
> +
> +static bool skipcvp;
> +module_param(skipcvp, bool, 0644);
> +MODULE_PARM_DESC(skipcvp, "Skip firmware load via CvP.");
> +
> +/**
> + * sel_create_cell() - Create an MFD cell and add it to the device.
> + * @pdev: The PCI device to operate on
> + * @name: MFD cell name
> + * @start: Start address of memory resource
> + * @length: Length of memory resource
> + * @msix_start: MSIX vector number to start from.
> + * @resources: Pointer to resources to add to the cell.
> + * @num_resources: Number of resources. The first resource is assumed to
> + *                 be memory, all other resources are IRQs.
> + */
> +static int sel_create_cell(struct pci_dev *pdev,

This is misleading.  It should be sel_register_device.

> +			   const char *name,

This never changes.  Why supply it as a variable?

> +			   unsigned int start,
> +			   unsigned int length,

This never changes.  Why supply it as a variable?

> +			   unsigned int msix_start,

This never changes.  Why supply it as a variable?

> +			   struct resource *resources,
> +			   unsigned int num_resources)

This never changes.  Why supply it as a variable?

> +{
> +	struct mfd_cell cell;
> +	unsigned int msix;
> +	int rc;
> +	int i;
> +
> +	if (!resources || num_resources < 1)

How would this be possible?

> +		return -EINVAL;
> +
> +	/* If MSI-X is enabled, assume we're using it and start at the

Malformed multi-line comment.

> +	 * provided vector number and iterate through the interrupts
> +	 * when adding IRQ resources. Otherwise leave the interrupt
> +	 * index at zero for legacy IRQs.
> +	 */
> +	if (pdev->msix_enabled)
> +		msix = msix_start;
> +	else
> +		msix = 0;

Preinitialise then drop the else.

> +	/* First resource is always a memory resource */
> +	resources[0].start = start;
> +	resources[0].end = start + length - 1;
> +	resources[0].flags = IORESOURCE_MEM;
> +
> +	/* Any additional resources are IRQs */
> +	for (i = 1; i < num_resources; i++) {
> +		resources[i].start = pci_irq_vector(pdev, msix);
> +		resources[i].end = pci_irq_vector(pdev, msix);

Does pci_irq_vector(pdev, msix) change between calls?

> +		resources[i].flags = IORESOURCE_IRQ;
> +		if (pdev->msix_enabled)
> +			msix++;
> +	}
> +
> +	cell.name = name;
> +	cell.num_resources = num_resources;
> +	cell.resources = resources;
> +
> +	rc = mfd_add_devices(&pdev->dev, PLATFORM_DEVID_AUTO, &cell,
> +			     1, NULL, 0, NULL);
> +
> +	return rc;
> +}
> +
> +/**
> + * create_eth_cell - Create and add an MFD cell for an selnet component
> + *
> + * @pdev: PCI device
> + * @offset: Offset from BAR 0 of the component
> + * @msix_num: Starting MSI-X vector number
> + */
> +static int create_eth_cell(struct pci_dev *pdev,
> +			   unsigned int offset,
> +			   unsigned int msix_num)
> +{
> +	struct resource resources[SELETH_NUM_RESOURCES] = {0};
> +
> +	return sel_create_cell(pdev,

This is only called from here, thus bring all of the sel_create_cell()
stuff into here instead.

> +			       "selpcimac",
> +			       pdev->resource[0].start + offset,
> +			       SELETH_LEN,
> +			       msix_num,
> +			       resources,
> +			       SELETH_NUM_RESOURCES);
> +}
> +
> +/**
> + * b1190_alloc_mfd_cells - Allocate and initialize MFD cell descriptions
> + *
> + * @pdev: PCI device for the b1190
> + *
> + * +----------------------------------------+
> + * |          | Resources |     Address     |
> + * |  Device  | Mem | IRQ | Start  | Length |
> + * |----------+-----+-----+--------+--------|
> + * | Ethernet |  1  |  3  | 0x0000 | 0x1000 |
> + * | Ethernet |  1  |  3  | 0x1000 | 0x1000 |
> + * | seltime  |  1  |  1  | 0x3000 | 0x1000 |
> + * | upgrades |  1  |  0  | 0x4000 | 0x4000 |
> + * +----------------------------------------+
> + */
> +static int b1190_alloc_mfd_cells(struct pci_dev *pdev)
> +{
> +	unsigned int offset;
> +	int rc;

'ret' is more common.

> +	int i;

Use for (int i = 0 ... instead.

> +	/* Ethernet */
> +	offset = 0x0000;

Initialise this during declaration.

> +	for (i = 0; i < B1190_NUM_ETHERNET; i++) {
> +		rc = create_eth_cell(pdev, offset, i * 3);
> +		if (rc)
> +			goto out_fail;
> +		offset += SELETH_LEN;
> +	}
> +
> +out_fail:
> +	return rc;
> +}

These functions are always identical.  Create a single function and pass
the number of ports into it and remove the rest.

> +
> +/**
> + * b2077_alloc_mfd_cells - Allocate and initialize MFD cell descriptions
> + *
> + * @pdev: PCI device for the b2077
> + *
> + * +----------------------------------------+
> + * |          | Resources |     Address     |
> + * |  Device  | Mem | IRQ | Start  | Length |
> + * |----------+-----+-----+--------+--------|
> + * | Ethernet |  1  |  3  | 0x0000 | 0x1000 |
> + * | Ethernet |  1  |  3  | 0x1000 | 0x1000 |
> + * | Ethernet |  1  |  3  | 0x2000 | 0x1000 |
> + * | Ethernet |  1  |  3  | 0x3000 | 0x1000 |
> + * | upgrades |  1  |  0  | 0x8000 | 0x4000 |
> + * | seltime  |  1  |  1  | 0xc000 | 0x1000 |
> + * +----------------------------------------+
> + */
> +static int b2077_alloc_mfd_cells(struct pci_dev *pdev)
> +{
> +	unsigned int offset;
> +	int rc;
> +	int i;
> +
> +	/* Ethernet */
> +	offset = 0x0000;
> +	for (i = 0; i < B2077_NUM_ETHERNET; i++) {
> +		rc = create_eth_cell(pdev, offset, i * 3);
> +		if (rc)
> +			goto out_fail;
> +		offset += SELETH_LEN;
> +	}
> +
> +out_fail:
> +	return rc;
> +}
> +
> +/*
> + * b2091_alloc_mfd_cells - Allocate and initialize MFD cell descriptions
> + *
> + * @pdev: PCI device for the b2091
> + *
> + * +------------------------------------------+
> + * |            | Resources |     Address     |
> + * |   Device   | Mem | IRQ | Start  | Length |
> + * |------------+-----+-----+--------+--------|
> + * | Ethernet   |  1  |  3  | 0x0000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x1000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x2000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x3000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x4000 | 0x1000 |
> + * | upgrades   |  1  |  0  | 0x8000 | 0x4000 |
> + * | seltime    |  1  |  1  | 0xc000 | 0x1000 |
> + * | multiuart  |  1  |  1  | 0xf000 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf100 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf200 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf300 | 0x0100 |
> + * +------------------------------------------+
> + */
> +static int b2091_alloc_mfd_cells(struct pci_dev *pdev)
> +{
> +	unsigned int offset;
> +	int rc;
> +	int i;
> +
> +	/* Ethernet */
> +	offset = 0x0000;
> +	for (i = 0; i < B2091_NUM_ETHERNET; i++) {
> +		rc = create_eth_cell(pdev, offset, i * 3);
> +		if (rc)
> +			goto out_fail;
> +		offset += SELETH_LEN;
> +	}
> +
> +out_fail:
> +	return rc;
> +}
> +
> +/**
> + * b2093_alloc_mfd_cells - Allocate and initialize MFD cell descriptions
> + *
> + * @pdev: PCI device for the b2093
> + *
> + * +------------------------------------------+
> + * |            | Resources |     Address     |
> + * |   Device   | Mem | IRQ | Start  | Length |
> + * |------------+-----+-----+--------+--------|
> + * | Ethernet   |  1  |  3  | 0x0000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x1000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x2000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x3000 | 0x1000 |
> + * | Ethernet   |  1  |  3  | 0x4000 | 0x1000 |
> + * | upgrades   |  1  |  0  | 0x8000 | 0x4000 |
> + * | seltime    |  1  |  1  | 0xc000 | 0x1000 |
> + * | contact IO |  1  |  0  | 0xe000 | 0x1000 |
> + * | multiuart  |  1  |  1  | 0xf000 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf100 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf200 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf300 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf400 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf500 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf600 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf700 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf800 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xf900 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xfa00 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xfb00 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xfc00 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xfd00 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xfe00 | 0x0100 |
> + * | multiuart  |  1  |  1  | 0xff00 | 0x0100 |
> + * +------------------------------------------+
> + */
> +static int b2093_alloc_mfd_cells(struct pci_dev *pdev)
> +{
> +	unsigned int offset;
> +	int rc;
> +	int i;
> +
> +	/* Ethernet */
> +	offset = 0x0000;
> +	for (i = 0; i < B2093_NUM_ETHERNET; i++) {
> +		rc = create_eth_cell(pdev, offset, i * 3);
> +		if (rc)
> +			goto out_fail;
> +		offset += SELETH_LEN;
> +	}
> +
> +out_fail:
> +	return rc;
> +}
> +
> +/*
> + * selpvmf_probe() - Probe function for the device.

None of these headers are really necessary.  This is not an external API.

> + *
> + * @pdev: PCI device being probed.
> + * @ent: Entry in the pci_tbl for the device.
> + *
> + * Return: 0 for success, otherwise the appropriate negative error code
> + */
> +static int selpvmf_probe(struct pci_dev *pdev, struct pci_device_id const *ent)
> +{
> +	typedef int (*alloc_fn_t)(struct pci_dev *pdev);
> +	alloc_fn_t alloc_fn = (alloc_fn_t)ent->driver_data;
> +	u16 cvp_capable;
> +	int num_msix;
> +	int rc = 0;
> +
> +	rc = pcim_enable_device(pdev);

Could it be that the device isn't ready?

No use for -EPROBE_DEFER?

> +	if (rc != 0)
> +		return rc;
> +
> +	/* Perform CvP program if necessary */
> +	cvp_capable = pci_find_ext_capability(pdev, CVP_CAPABILITY_ID);
> +	if (!skipcvp && cvp_capable) {

The '_' inconsistency is bothering me.

As is the placement of "cvp".

> +		rc = cvp_start(pdev);
> +		if (rc != 0) {
> +			dev_err(&pdev->dev, "Error: CvP failed: 0x%x\n", rc);

This is a user facing error message and I still don't know what CvP is.

Please be more forthcoming.

> +			goto out_disable;
> +		}
> +		dev_info(&pdev->dev, "CvP Successful!\n");

This serves no purpose.  Please keep the log as clean as possible.

> +	}
> +
> +	pci_set_master(pdev);
> +
> +	/* Only MSI-X and legacy interrupts are supported. Try to get

Malformed multi-line comment.

> +	 * MSI-X, if we don't get them all, fall back to legacy.

Full stop after MSI-X.

> +	 */
> +	num_msix = pci_msix_vec_count(pdev);

'\n' here.

> +	rc = pci_alloc_irq_vectors(pdev, num_msix, num_msix, PCI_IRQ_MSIX);
> +	if (rc != num_msix) {
> +		dev_warn(&pdev->dev,
> +			 "Failed to get MSI-X, falling back to legacy\n");

Who is going to care about that?

'\n' here.

> +		rc = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_INTX);
> +		if (rc != 1) {
> +			rc = -ENOSPC;

I think you should be returning rc as-is here.

> +			goto out_disable;
> +		}
> +	}
> +
> +	rc = alloc_fn(pdev);

These call-backs are horrible.  Please don't do this.

> +	if (rc < 0) {
> +		dev_err(&pdev->dev, "ERROR: Failed to add child cells\n");

"Failed to register child devices"

And it should come after the call to mfd_add_devices() not here.

> +		goto out_free_irq;
> +	}
> +
> +	dev_info(&pdev->dev, "Successfully added new devices\n");

Remove this please.

> +	return 0;
> +
> +out_free_irq:
> +	pci_free_irq_vectors(pdev);
> +out_disable:
> +	pci_clear_master(pdev);
> +	return rc;
> +}
> +
> +/*
> + * selpvmf_remove - Remove a selpvmf enabled PCI device from the system
> + *
> + * @pdev: PCI device to be removed
> + */
> +static void selpvmf_remove(struct pci_dev *pdev)
> +{
> +	mfd_remove_devices(&pdev->dev);

Use devm_* instead,

> +	pci_free_irq_vectors(pdev);
> +	pci_clear_master(pdev);
> +}
> +
> +/*
> + * selpvmf_shutdown() - Tears down a selpvmf card.
> + *
> + * @pdev: Pointer to the device structure describing the port.
> + */
> +static void selpvmf_shutdown(struct pci_dev *pdev)
> +{
> +	mfd_remove_devices(&pdev->dev);

As above.

> +}
> +
> +/*
> + * selpvmf_suspend - Suspend the device in preparation for entering a
> + * a low power state.
> + *
> + * @dev: Pointer to the device structure describing the port.
> + *
> + * Return: This function always indicates success.
> + *
> + * Most of the work should be handled by the child devices of this
> + * driver. There is no specific action necessary by the card, so when
> + * the child devices are done suspending, just do a generic PCI saving
> + * of state.
> + */
> +static int selpvmf_suspend(struct device *dev)
> +{
> +	struct pci_dev *pdev = to_pci_dev(dev);
> +
> +	pci_save_state(pdev);
> +
> +	return 0;
> +}
> +
> +/*
> + * selpvmf_resume() - Resume a selpvmf device that has previously been
> + * suspended
> + *
> + * @dev: Pointer to the device structure describing the port.
> + *
> + * Return: This function always indicates success.
> + */
> +static int selpvmf_resume(struct device *dev)
> +{
> +	struct pci_dev *pdev = to_pci_dev(dev);
> +	u16 cvp_capable;
> +	int rc;
> +
> +	/* Turn the card on by requesting a transition to D0 before
> +	 * restoring the PCI device state and enabling the UARTs
> +	 */
> +	pci_set_power_state(pdev, PCI_D0);
> +
> +	/* Restore so we can do IOMEM based CvP */
> +	pci_restore_state(pdev);
> +
> +	rc = pcim_enable_device(pdev);
> +	if (rc != 0) {
> +		/* We are not really supposed to return an error from this
> +		 * function, so if the enable fails, just print some info
> +		 * to aid in troubleshooting.
> +		 */

What?  It returns an int for a reason, please use it.

> +		dev_err(&pdev->dev,
> +			"Unable to enable the device. Continuing anyway\n");
> +	}
> +
> +	/* Perform CvP program if necessary */
> +	cvp_capable = pci_find_ext_capability(pdev, CVP_CAPABILITY_ID);
> +	if (!skipcvp && cvp_capable) {
> +		pci_save_state(pdev);
> +
> +		rc = cvp_start(pdev);
> +		if (rc != 0) {
> +			dev_err(&pdev->dev,
> +				"Error: CvP failed (resume): 0x%x\n", rc);

You can use up to 100-chars to prevent this kind of wrapping.

> +		}
> +		dev_info(&pdev->dev, "CvP done (resume): 0x%x\n", rc);

Remove this please.

> +
> +		pci_restore_state(pdev);
> +	}
> +
> +	pci_set_master(pdev);
> +
> +	return 0;
> +}
> +
> +static const struct dev_pm_ops selpvmf_pm_ops = {
> +	SYSTEM_SLEEP_PM_OPS(selpvmf_suspend, selpvmf_resume)
> +};
> +
> +static struct pci_device_id selpvmf_pci_tbl[] = {
> +	{
> +		PCI_DEVICE(PCI_VENDOR_ID_SEL, PCI_DEVICE_ID_B1190),
> +		.driver_data = (kernel_ulong_t)b1190_alloc_mfd_cells,

Pass an identifier here instead of call-back functions.

> +	},
> +	{
> +		PCI_DEVICE(PCI_VENDOR_ID_SEL, PCI_DEVICE_ID_B2077),
> +		.driver_data = (kernel_ulong_t)b2077_alloc_mfd_cells,
> +	},
> +	{
> +		PCI_DEVICE(PCI_VENDOR_ID_SEL, PCI_DEVICE_ID_B2091),
> +		.driver_data = (kernel_ulong_t)b2091_alloc_mfd_cells,
> +	},
> +	{
> +		PCI_DEVICE(PCI_VENDOR_ID_SEL, PCI_DEVICE_ID_B2093),
> +		.driver_data = (kernel_ulong_t)b2093_alloc_mfd_cells,
> +	},
> +	{0,}

What is this?  Where have you seen that before?

> +};
> +MODULE_DEVICE_TABLE(pci, selpvmf_pci_tbl);
> +
> +static struct pci_driver selpvmf_driver = {
> +	.name		= "selpvmf",
> +	.id_table	= selpvmf_pci_tbl,
> +	.probe		= selpvmf_probe,
> +	.remove		= selpvmf_remove,
> +	.shutdown	= selpvmf_shutdown,
> +	.driver.pm	= pm_ptr(&selpvmf_pm_ops),
> +};
> +module_pci_driver(selpvmf_driver);
> +
> +MODULE_LICENSE("Dual BSD/GPL");
> +MODULE_AUTHOR("Schweitzer Engineering Laboratories, Inc.");
> +MODULE_DESCRIPTION("Multi function driver for enumerating multiple device drivers for a single PCI resource");
> +MODULE_DEVICE_TABLE(pci, selpvmf_pci_tbl);
> +MODULE_FIRMWARE("sel/1AA9_0014_01.cvp");
> +MODULE_FIRMWARE("sel/1AA9_0015_01.cvp");
> +MODULE_FIRMWARE("sel/1AA9_0018_01.cvp");
> diff --git a/drivers/mfd/selpvmf-cvp.c b/drivers/mfd/selpvmf-cvp.c

I'm not sure what all of this is, but I suggest that it doesn't live
here.  drivers/firmware or drivers/fpga perhaps?

> new file mode 100644
> index 000000000000..e119cce83d7c
> --- /dev/null
> +++ b/drivers/mfd/selpvmf-cvp.c
> @@ -0,0 +1,431 @@
> +// SPDX-License-Identifier: GPL-2.0-only OR BSD-3-Clause
> +/*
> + * Copyright 2017 Schweitzer Engineering Laboratories, Inc.
> + * 2350 NE Hopkins Court, Pullman, WA 99163 USA
> + * SEL Open Source <opensource@...inc.com>

Do you really need the whole postal address here?

I suggest this is legacy of a time well past.

> + *
> + * Support for Configuration via Protocol (CvP) on SEL devices using
> + * Altera FPGAs.
> + */
> +
> +#include <linux/firmware.h>
> +#include <linux/pci.h>
> +#include <linux/sched.h>
> +
> +#include "selpvmf-cvp.h"
> +
> +#define MIN_WAIT 2 /* number of jiffies for unit wait period */

Comments should start with an upper case char.

> +#define US_PER_JIFFY  (1000000 / HZ) /* number of microseconds per jiffy */

Please tab all of these out.

> +#define CVP_WAIT_TIMEOUT 1000000 /* Wait timeout in microseconds */
> +#define BYTES_PER_DWORD	4
> +#define CORE_FIRMWARE_FILE_BASEDIR "sel/"
> +#define CORE_FIRMWARE_FILE_EXT ".cvp"
> +
> +#define CVP_STATUS_OFFSET	0x0000001C
> +#define CVP_CONFIG_READY	BIT(18)
> +#define CVP_CONFIG_ERROR	BIT(19)
> +#define CVP_EN			BIT(20)
> +#define USERMODE		BIT(21)
> +#define PLD_CLK_IN_USE		BIT(24)
> +
> +#define CVP_MODE_CONTROL_OFFSET		0x00000020
> +#define CVP_MODE			BIT(0)
> +#define HIP_CLK_SEL			BIT(1)
> +#define CVP_NUMCLKS_MASK		GENMASK(15, 8)
> +#define CVP_NUMCLKS_UNCMPRSSD_UNENCRYPT	BIT(8)
> +#define CVP_NUMCLKS_COMPRESSED		BIT(11)
> +#define CVP_NUMCLKS_ENCRYPTED		BIT(10)
> +
> +#define CVP_DATA_OFFSET	0x00000028
> +
> +#define CVP_PROGRAMMING_CTRL_OFFSET	0x0000002C
> +#define CVP_CONFIG			BIT(0)
> +#define START_XFER			BIT(1)
> +
> +#define CVP_UNCORRECTABLE_IE_STATUS_OFFSET	0x00000034
> +#define CVP_CONFIG_ERROR_LATCHED		BIT(5)

I'm going to stop here.  Please address the issues before resubmitting.

> +struct cvp_state {
> +	int offset;
> +	struct firmware const *fw;
> +	struct pci_dev *pdev;
> +	uint32_t __iomem *bar;
> +};
> +
> +/**
> + * cvp_set_bit() - Sets the bit requested to 1
> + *
> + * @state: Initialized CvP State structure
> + * @offset: The pci capability offset to read the value from
> + * @bit: The bit to read at the pci config space offset
> + */
> +static void cvp_set_bit(struct cvp_state *state, uint32_t offset, uint32_t bit)
> +{
> +	uint32_t reg;
> +
> +	pci_read_config_dword(state->pdev, state->offset + offset, &reg);
> +	reg |= bit;
> +	pci_write_config_dword(state->pdev, state->offset + offset, reg);
> +}
> +
> +/**
> + * cvp_clear_bit() - Sets the bit requested to 0
> + *
> + * @state: Initialized CvP State structure
> + * @offset: The pci capability offset to read the value from
> + * @bit: The bit to read at the pci config space offset
> + */
> +static void cvp_clear_bit(struct cvp_state *state, uint32_t offset, uint32_t bit)
> +{
> +	uint32_t reg;
> +
> +	pci_read_config_dword(state->pdev, state->offset + offset, &reg);
> +	reg &= ~(bit);
> +	pci_write_config_dword(state->pdev, state->offset + offset, reg);
> +}
> +
> +/**
> + * cvp_read_bit() - Reads the requested bit to determine its state
> + *
> + * @state: Initialized CvP State structure
> + * @offset: The pci capability offset to read the value from
> + * @bit: The bit to read at the pci config space offset
> + *
> + * Return: 1 if the bit is set, 0 if not
> + */
> +static int cvp_read_bit(struct cvp_state *state, uint32_t offset, uint32_t bit)
> +{
> +	uint32_t reg;
> +
> +	pci_read_config_dword(state->pdev, state->offset + offset, &reg);
> +	return ((reg & bit) == 0) ? 0 : 1;
> +}
> +
> +/**
> + * cvp_wait_for_bit() - Reads the requested bit and spins until it
> + * becomes the value requested
> + *
> + * This request will also timeout at the ns value requested by the caller
> + *
> + * @state: Initialized CvP State structure
> + * @offset: The pci capability offset to read the value from
> + * @bit: The bit to read at the pci config space offset
> + * @value: The value the bit needs to be set to in order to return
> + *
> + * Return: 1 if the bit_val became the value requested during the timeout
> + * period. 0 if a timeout occurred.
> + */
> +static int cvp_wait_for_bit(struct cvp_state *state,
> +			    uint32_t offset,
> +			    uint32_t bit,
> +			    uint32_t value)
> +{
> +	uint8_t bit_val;
> +	uint32_t n_wait_loops = 1;
> +
> +	if (CVP_WAIT_TIMEOUT > (US_PER_JIFFY * MIN_WAIT)) {
> +		n_wait_loops = (CVP_WAIT_TIMEOUT + (US_PER_JIFFY * MIN_WAIT) - 1) /
> +			       (US_PER_JIFFY * MIN_WAIT);
> +	}
> +
> +	DECLARE_WAIT_QUEUE_HEAD(cvp_wq);
> +
> +	bit_val = cvp_read_bit(state, offset, bit);
> +
> +	while ((bit_val != value) && (n_wait_loops != 0)) {
> +		wait_event_timeout(cvp_wq, 0, MIN_WAIT);
> +		bit_val = cvp_read_bit(state, offset, bit);
> +		--n_wait_loops;
> +	}
> +
> +	return bit_val == value;
> +}
> +
> +/**
> + * perform_244_dummy_writes() - Performs 244 dummy writes to the hard IP
> + * to clear the CvP state machine.
> + *
> + * @state: Initialized CvP State structure
> + */
> +static void perform_244_dummy_writes(struct cvp_state *state)
> +{
> +	for (int i = 0; i < 244; i++)
> +		iowrite32(0xFFFFFFFF, state->bar);
> +}
> +
> +/**
> + * set_numclks() - Updates the CVP_NUMCLKS value in the mode control register
> + *
> + * @state: Initialized CvP State structure
> + * @value: The value to set CVP_NUMCLKS to
> + */
> +static void set_numclks(struct cvp_state *state, uint32_t value)
> +{
> +	uint32_t reg;
> +
> +	pci_read_config_dword(state->pdev,
> +			      state->offset + CVP_MODE_CONTROL_OFFSET,
> +			      &reg);
> +	reg &= ~(CVP_NUMCLKS_MASK);
> +	reg |= (value & CVP_NUMCLKS_MASK);
> +	pci_write_config_dword(state->pdev,
> +			       state->offset + CVP_MODE_CONTROL_OFFSET,
> +			       reg);
> +}
> +
> +/**
> + * cvp_teardown() - Removes the PCI device from its ready to program state
> + *
> + * @state: Initialized CvP State structure
> + *
> + * Return: 0 for success, negative error code if not
> + */
> +static int cvp_teardown(struct cvp_state *state)
> +{
> +	int status = 0;
> +	uint32_t bit_val;
> +
> +	cvp_clear_bit(state, CVP_PROGRAMMING_CTRL_OFFSET, START_XFER);
> +	cvp_clear_bit(state, CVP_PROGRAMMING_CTRL_OFFSET, CVP_CONFIG);
> +	set_numclks(state, CVP_NUMCLKS_UNCMPRSSD_UNENCRYPT);
> +	perform_244_dummy_writes(state);
> +
> +	if (!cvp_wait_for_bit(state, CVP_STATUS_OFFSET, CVP_CONFIG_READY, 0)) {
> +		status = -1;
> +		goto exit;
> +	}
> +
> +	bit_val = cvp_read_bit(state,
> +			       CVP_UNCORRECTABLE_IE_STATUS_OFFSET,
> +			       CVP_CONFIG_ERROR_LATCHED);
> +
> +	if (bit_val == 1) {
> +		cvp_set_bit(state,
> +			    CVP_UNCORRECTABLE_IE_STATUS_OFFSET,
> +			    CVP_CONFIG_ERROR_LATCHED);
> +	}
> +
> +	cvp_clear_bit(state, CVP_MODE_CONTROL_OFFSET, CVP_MODE);
> +	cvp_clear_bit(state, CVP_MODE_CONTROL_OFFSET, HIP_CLK_SEL);
> +
> +	if (bit_val == 0) {
> +		if (!cvp_wait_for_bit(state, CVP_STATUS_OFFSET,
> +				      PLD_CLK_IN_USE, 1) ||
> +		    !cvp_wait_for_bit(state, CVP_STATUS_OFFSET, USERMODE, 1)) {
> +			status = -1;
> +		}
> +	}
> +
> +exit:
> +	return status;
> +}
> +
> +/**
> + * cvp_program() - Performs the FPGA programming via CvP
> + *
> + * @state: Initialized CvP State structure
> + *
> + * Return: 0 for success, negative error code if not
> + */
> +static int cvp_program(struct cvp_state *state)
> +{
> +	int status = 0;
> +	int i;
> +
> +	uint32_t *firmware_ptr;
> +	uint32_t dword_size;
> +
> +	set_numclks(state, CVP_NUMCLKS_COMPRESSED);
> +
> +	/*
> +	 * The file_read_data function returns a size in bytes,
> +	 * cvp_set_reg requires a write of a single dword. Therefore
> +	 * calculate a dword size.
> +	 */
> +	dword_size = (state->fw->size / BYTES_PER_DWORD);
> +	if ((state->fw->size % BYTES_PER_DWORD) != 0)
> +		dword_size++;
> +
> +	firmware_ptr = (uint32_t *)state->fw->data;
> +	for (i = 0; i < dword_size; i++)
> +		iowrite32(firmware_ptr[i], state->bar);
> +
> +	/* Check if any errors happened during the CvP process */
> +	if (cvp_read_bit(state,	CVP_STATUS_OFFSET, CVP_CONFIG_ERROR) == 1)
> +		status = -1;
> +
> +	return status;
> +}
> +
> +/**
> + * cvp_setup() - Performs the CvP setup logic readying the pci device for
> + * programming
> + *
> + * @state: Initialized CvP State structure
> + *
> + * Return: 0 for success, negative error code if not
> + */
> +static int cvp_setup(struct cvp_state *state)
> +{
> +	int status = 0;
> +
> +	/*
> +	 * Make sure the hardware isn't in a state where a previous CvP
> +	 * failed. If it did, run teardown to put it back in a pristine
> +	 * state.
> +	 */
> +	if (cvp_read_bit(state, CVP_MODE_CONTROL_OFFSET, CVP_MODE) != 0)
> +		cvp_teardown(state);
> +
> +	cvp_set_bit(state, CVP_MODE_CONTROL_OFFSET, HIP_CLK_SEL);
> +	cvp_set_bit(state, CVP_MODE_CONTROL_OFFSET, CVP_MODE);
> +	set_numclks(state, CVP_NUMCLKS_UNCMPRSSD_UNENCRYPT);
> +	perform_244_dummy_writes(state);
> +	cvp_set_bit(state, CVP_PROGRAMMING_CTRL_OFFSET, CVP_CONFIG);
> +	if (!cvp_wait_for_bit(state, CVP_STATUS_OFFSET, CVP_CONFIG_READY, 1)) {
> +		status = -1;
> +		/*
> +		 * This may fail, but try to put it back.  Note that we
> +		 * aren't waiting for clocks to stabilize after doing
> +		 * this, so the device isn't going to be usable.  We are
> +		 * assuming the driver is going to bail.
> +		 */
> +		cvp_clear_bit(state, CVP_MODE_CONTROL_OFFSET, CVP_MODE);
> +		cvp_clear_bit(state, CVP_MODE_CONTROL_OFFSET, HIP_CLK_SEL);
> +		goto exit;
> +	}
> +
> +	set_numclks(state, CVP_NUMCLKS_UNCMPRSSD_UNENCRYPT);
> +	perform_244_dummy_writes(state);
> +	cvp_set_bit(state, CVP_PROGRAMMING_CTRL_OFFSET, START_XFER);
> +
> +exit:
> +	return status;
> +}
> +
> +/**
> + * cvp_init() - Initialize the CvP state machine.
> + *
> + * @pdev: Pointer to the pci_device structure containing information
> + *        about the device being initialized
> + * @state: Caller allocated CvP state structure to be initialized.
> + *
> + * Return: 0 if the state was initialized successfully and CvP can be
> + *    executed.  A non-zero exit code will indicate an appropriate error.
> + *
> + *    If the function returns 0, the caller should call cvp_exit to
> + *    cleanup.
> + */
> +static int cvp_init(struct pci_dev *pdev, struct cvp_state *state)
> +{
> +	int status = 0;
> +	int value;
> +	char filepath[512];
> +
> +	state->pdev = pdev;
> +	state->fw = NULL;
> +	state->bar = NULL;
> +
> +	state->offset = pci_find_ext_capability(pdev, CVP_CAPABILITY_ID);
> +	if (!state->offset) {
> +		dev_warn(&pdev->dev,
> +			 "Device requests CvP but indicating not capable\n");
> +		status = -ENODEV;
> +		goto exit;
> +	}
> +
> +	value = cvp_read_bit(state, CVP_STATUS_OFFSET, CVP_EN);
> +	if (value == 0) {
> +		dev_err(&pdev->dev, "CvP not enabled in device\n");
> +		status = -ENODEV;
> +		goto exit;
> +	}
> +
> +	/*
> +	 * BAR 0 will always exist and will always be mmio so we can do
> +	 * memory transactions to speed up CvP. Make sure to not map
> +	 * more than our first resource. CvP will write to only the
> +	 * first address that is mapped so a mapping size of 4 bytes
> +	 * should do.
> +	 */
> +	state->bar = pci_iomap(pdev, 0, 4);
> +	if (!state->bar) {
> +		dev_err(&pdev->dev, "Memory BAR not available\n");
> +		status = -ENODEV;
> +		goto exit;
> +	}
> +
> +	/* Create the filepath to request firmware from */
> +	snprintf(filepath,
> +		 512,
> +		 "%s%04X_%04X_%02x%s",
> +		 CORE_FIRMWARE_FILE_BASEDIR,
> +		 pdev->vendor,
> +		 pdev->device,
> +		 pdev->revision,
> +		 CORE_FIRMWARE_FILE_EXT);
> +
> +	if (request_firmware(&state->fw, filepath, &pdev->dev)) {
> +		dev_err(&pdev->dev,
> +			"Error: The driver failed to find a core image for this device at: %s\n",
> +			filepath);
> +		iounmap(state->bar);
> +		state->bar = NULL;
> +		status = -ENOENT;
> +	}
> +
> +exit:
> +	return status;
> +}
> +
> +/**
> + * cvp_exit() - Cleanup CvP resources
> + *
> + * @state: Initialized CvP state to cleanup.
> + */
> +static void cvp_exit(struct cvp_state *state)
> +{
> +	if (state->fw)
> +		release_firmware(state->fw);
> +	if (state->bar) {
> +		iounmap(state->bar);
> +		state->bar = NULL;
> +	}
> +}
> +
> +/**
> + * cvp_start() - Public facing function to kick off FPGA programming via CvP
> + *
> + * @pdev: Pointer to the pci_device structure containing information about the
> + *        device being initialized
> + *
> + * Return: 0 for success, otherwise the appropriate negative error code
> + */
> +int cvp_start(struct pci_dev *pdev)
> +{
> +	int status = 0;
> +	struct cvp_state state;
> +
> +	dev_info(&pdev->dev, "Loading device firmware. AER errors may occur during the loading process.\n");
> +
> +	status = cvp_init(pdev, &state);
> +	if (status != 0)
> +		goto exit;
> +
> +	status = cvp_setup(&state);
> +	if (status != 0)
> +		goto exit;
> +
> +	/*
> +	 * If setup succeeded we need to perform the cvp teardown
> +	 * regardless of whether or not the program was successful
> +	 */
> +	if (cvp_program(&state) != 0)
> +		status = -1;
> +	if (cvp_teardown(&state) != 0)
> +		status = -1;
> +
> +exit:
> +	cvp_exit(&state);
> +	return status;
> +}
> diff --git a/drivers/mfd/selpvmf-cvp.h b/drivers/mfd/selpvmf-cvp.h
> new file mode 100644
> index 000000000000..32de3b9e9884
> --- /dev/null
> +++ b/drivers/mfd/selpvmf-cvp.h
> @@ -0,0 +1,18 @@
> +/* SPDX-License-Identifier: GPL-2.0-only OR BSD-3-Clause */
> +/*
> + * Copyright 2017 Schweitzer Engineering Laboratories, Inc.
> + * 2350 NE Hopkins Court, Pullman, WA 99163 USA
> + * SEL Open Source <opensource@...inc.com>
> + */
> +
> +#ifndef _SELPVMF_CVP_H
> +#define _SELPVMF_CVP_H
> +
> +#include <linux/pci.h>
> +
> +/* Define for CvP in PCI config space */
> +#define CVP_CAPABILITY_ID	0x0000000B
> +
> +int cvp_start(struct pci_dev *pdev);
> +
> +#endif /* _SELPVMF_CVP_H */
> -- 
> 2.45.2
> 

-- 
Lee Jones [李琼斯]

Powered by blists - more mailing lists