Skip to content
Permalink
Browse files
fsl_qbman: Framework for enabling Link status notification
 -  This will enable link update event notification for
    user space USDPAA application.

Signed-off-by: Sachin Saxena <sachin.saxena@nxp.com>
DPDK-2128
(cherry picked from commit fb53c81)

(cherry picked from commit 13efcd6)
  • Loading branch information
sachin-saxena-nxp authored and Jason Liu committed Feb 25, 2020
1 parent 40693b4 commit 7dcadb8d9273dd1ce30f55e8b4489f7c195d44c7
Show file tree
Hide file tree
Showing 3 changed files with 313 additions and 1 deletion.
@@ -1,5 +1,9 @@
subdir-ccflags-y := -Werror

# Include netcomm SW specific definitions
include $(srctree)/drivers/net/ethernet/freescale/sdk_fman/ncsw_config.mk
ccflags-y += -I$(NET_DPA)

# Common
obj-$(CONFIG_FSL_SDK_DPA) += dpa_alloc.o
obj-$(CONFIG_FSL_SDK_DPA) += qbman_driver.o
@@ -18,6 +18,8 @@
#include <linux/slab.h>
#include <linux/mman.h>
#include <linux/of_reserved_mem.h>
#include <linux/eventfd.h>
#include <linux/fdtable.h>

#if !(defined(CONFIG_ARM) || defined(CONFIG_ARM64))
#include <mm/mmu_decl.h>
@@ -27,6 +29,26 @@
#include <linux/fsl_usdpaa.h>
#include "bman_low.h"
#include "qman_low.h"
/* Headers requires for
* Link status support
*/
#include <linux/device.h>
#include <linux/of_mdio.h>
#include "mac.h"
#include "dpaa_eth_common.h"

/* Private data for Proxy Interface */
struct dpa_proxy_priv_s {
struct mac_device *mac_dev;
struct eventfd_ctx *efd_ctx;
};
/* Interface Helpers */
static inline struct device *get_dev_ptr(char *if_name);
static void phy_link_updates(struct net_device *net_dev);
/* IOCTL handlers */
static inline int ioctl_usdpaa_get_link_status(char *if_name);
static int ioctl_en_if_link_status(struct usdpaa_ioctl_link_status *args);
static int ioctl_disable_if_link_status(char *if_name);

/* Physical address range of the memory reservation, exported for mm/mem.c */
static u64 phys_start;
@@ -556,7 +578,6 @@ static bool check_portal_channel(void *ctx, u32 channel)




static int usdpaa_release(struct inode *inode, struct file *filp)
{
int err = 0;
@@ -1656,6 +1677,220 @@ static long ioctl_free_raw_portal(struct file *fp, struct ctx *ctx,
return 0;
}


static inline struct device *get_dev_ptr(char *if_name)
{
struct device *dev;
char node[NODE_NAME_LEN];

sprintf(node, "soc:fsl,dpaa:%s",if_name);
dev = bus_find_device_by_name(&platform_bus_type, NULL, node);
if (dev == NULL) {
pr_err(KBUILD_MODNAME "IF %s not found\n", if_name);
return NULL;
}
pr_debug("%s: found dev 0x%lX for If %s ,dev->platform_data %p\n",
__func__, (unsigned long)dev,
if_name, dev->platform_data);

return dev;
}

/* This function will return Current link status of the device
* '1' if Link is UP, '0' otherwise.
*
* Input parameter:
* if_name: Interface node name
*
*/
static inline int ioctl_usdpaa_get_link_status(char *if_name)
{
struct net_device *net_dev = NULL;
struct device *dev;

dev = get_dev_ptr(if_name);
if (dev == NULL)
return -ENODEV;
net_dev = dev->platform_data;
if (net_dev == NULL)
return -ENODEV;

if (test_bit(__LINK_STATE_NOCARRIER, &net_dev->state))
return 0; /* Link is DOWN */
else
return 1; /* Link is UP */
}


/* Link Status Callback Function
* This function will be resgitered to PHY framework to get
* Link update notifications and should be responsible for waking up
* user space task when there is a link update notification.
*/
static void phy_link_updates(struct net_device *net_dev)
{
struct dpa_proxy_priv_s *priv = NULL;

pr_debug("%s: Link '%s': Speed '%d-Mbps': Autoneg '%d': Duplex '%d'\n",
net_dev->name,
ioctl_usdpaa_get_link_status(net_dev->name)?"UP":"DOWN",
net_dev->phydev->speed,
net_dev->phydev->autoneg,
net_dev->phydev->duplex);

/* Wake up the user space context to notify PHY update */
priv = netdev_priv(net_dev);
eventfd_signal(priv->efd_ctx, 1);
}


/* IOCTL handler for enabling Link status request for a given interface
* Input parameters:
* args->if_name: This the network interface node name as defind in
* device tree file. Currently, it has format of
* "ethernet@x" type for each interface.
* args->efd: The eventfd value which should be waked up when
* there is any link update received.
*/
static int ioctl_en_if_link_status(struct usdpaa_ioctl_link_status *args)
{
struct net_device *net_dev = NULL;
struct dpa_proxy_priv_s *priv = NULL;
struct device *dev;
struct mac_device *mac_dev;
struct proxy_device *proxy_dev;
struct task_struct *userspace_task = NULL;
struct file *efd_file = NULL;

dev = get_dev_ptr(args->if_name);
if (dev == NULL)
return -ENODEV;
/* Utilize dev->platform_data to save netdevice
pointer as it will not be registered */
if (dev->platform_data) {
pr_debug("%s: IF %s already initialized\n",
__func__, args->if_name);
/* This will happen when application is not able to initiate
* cleanup in last run. We still need to save the new
* eventfd context.
*/
net_dev = dev->platform_data;
priv = netdev_priv(net_dev);

/* Get current task context from which IOCTL was called */
userspace_task = current;

rcu_read_lock();
efd_file = fcheck_files(userspace_task->files, args->efd);
rcu_read_unlock();

priv->efd_ctx = eventfd_ctx_fileget(efd_file);
if (!priv->efd_ctx) {
pr_err(KBUILD_MODNAME "get eventfd context failed\n");
/* Free the allocated memory for net device */
dev->platform_data = NULL;
free_netdev(net_dev);
return -EINVAL;
}
/* Since there will be NO PHY update as link is already setup,
* wake user context once so that current PHY status can
* be fetched.
*/
phy_link_updates(net_dev);
return 0;
}

proxy_dev = dev_get_drvdata(dev);
mac_dev = proxy_dev->mac_dev;
/* Allocate an dummy net device for proxy interface */
net_dev = alloc_etherdev(sizeof(*priv));
if (!net_dev) {
pr_err(KBUILD_MODNAME "alloc_etherdev failed\n");
return -ENOMEM;
} else {
SET_NETDEV_DEV(net_dev, dev);
priv = netdev_priv(net_dev);
priv->mac_dev = mac_dev;
/* Get current task context from which IOCTL was called */
userspace_task = current;

rcu_read_lock();
efd_file = fcheck_files(userspace_task->files, args->efd);
rcu_read_unlock();

priv->efd_ctx = eventfd_ctx_fileget(efd_file);

if (!priv->efd_ctx) {
pr_err(KBUILD_MODNAME "get eventfd context failed\n");
/* Free the allocated memory for net device */
free_netdev(net_dev);
return -EINVAL;
}
strncpy(net_dev->name, args->if_name, IF_NAME_MAX_LEN);
dev->platform_data = net_dev;
}

pr_debug("%s: mac_dev %p cell_index %d\n",
__func__, mac_dev, mac_dev->cell_index);
mac_dev->phy_dev = of_phy_connect(net_dev, mac_dev->phy_node,
phy_link_updates, 0, mac_dev->phy_if);
if (unlikely(mac_dev->phy_dev == NULL) || IS_ERR(mac_dev->phy_dev)) {
pr_err("%s: --------Error in PHY Connect\n", __func__);
/* Free the allocated memory for net device */
free_netdev(net_dev);
return -ENODEV;
}
net_dev->phydev = mac_dev->phy_dev;
mac_dev->start(mac_dev);
pr_debug("%s: --- PHY connected for %s\n", __func__, args->if_name);

return 0;
}

/* IOCTL handler for disabling Link status for a given interface
* Input parameters:
* if_name: This the network interface node name as defind in
* device tree file. Currently, it has format of
* "ethernet@x" type for each interface.
*/
static int ioctl_disable_if_link_status(char *if_name)
{
struct net_device *net_dev = NULL;
struct device *dev;
struct mac_device *mac_dev;
struct proxy_device *proxy_dev;
struct dpa_proxy_priv_s *priv = NULL;

dev = get_dev_ptr(if_name);
if (dev == NULL)
return -ENODEV;
/* Utilize dev->platform_data to save netdevice
pointer as it will not be registered */
if (!dev->platform_data) {
pr_debug("%s: IF %s already Disabled for Link status\n",
__func__, if_name);
return 0;
}

net_dev = dev->platform_data;
proxy_dev = dev_get_drvdata(dev);
mac_dev = proxy_dev->mac_dev;
mac_dev->stop(mac_dev);

priv = netdev_priv(net_dev);
eventfd_ctx_put(priv->efd_ctx);

/* This will also deregister the call back */
phy_disconnect(mac_dev->phy_dev);
phy_resume(mac_dev->phy_dev);

free_netdev(net_dev);
dev->platform_data = NULL;

pr_debug("%s: Link status Disabled for %s\n", __func__, if_name);
return 0;
}

static long usdpaa_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
{
struct ctx *ctx = fp->private_data;
@@ -1722,6 +1957,47 @@ static long usdpaa_ioctl(struct file *fp, unsigned int cmd, unsigned long arg)
return -EFAULT;
return ioctl_free_raw_portal(fp, ctx, &input);
}
case USDPAA_IOCTL_ENABLE_LINK_STATUS_INTERRUPT:
{
struct usdpaa_ioctl_link_status input;
int ret;

if (copy_from_user(&input, a, sizeof(input)))
return -EFAULT;
ret = ioctl_en_if_link_status(&input);
if (ret)
pr_err("Error(%d) enable link interrupt:IF: %s\n",
ret, input.if_name);
return ret;
}
case USDPAA_IOCTL_DISABLE_LINK_STATUS_INTERRUPT:
{
char *input;
int ret;

if (copy_from_user(&input, a, sizeof(input)))
return -EFAULT;
ret = ioctl_disable_if_link_status(input);
if (ret)
pr_err("Error(%d) Disabling link interrupt:IF: %s\n",
ret, input);
return ret;
}
case USDPAA_IOCTL_GET_LINK_STATUS:
{
struct usdpaa_ioctl_link_status_args input;

if (copy_from_user(&input, a, sizeof(input)))
return -EFAULT;

input.link_status = ioctl_usdpaa_get_link_status(input.if_name);
if (input.link_status < 0)
return input.link_status;
if (copy_to_user(a, &input, sizeof(input)))
return -EFAULT;

return 0;
}
}
return -EINVAL;
}
@@ -365,6 +365,38 @@ int dpa_alloc_pop(struct dpa_alloc *alloc, u32 *result, u32 *count);
int dpa_alloc_check(struct dpa_alloc *list, u32 id);
#endif /* __KERNEL__ */


/************************************
* Link Status support for user space
* interface
************************************/
#define IF_NAME_MAX_LEN 16
#define NODE_NAME_LEN 32

struct usdpaa_ioctl_link_status {
/* network device node name */
char if_name[IF_NAME_MAX_LEN];
/* Eventfd value */
uint32_t efd;
};

#define USDPAA_IOCTL_ENABLE_LINK_STATUS_INTERRUPT \
_IOW(USDPAA_IOCTL_MAGIC, 0x0E, struct usdpaa_ioctl_link_status)

#define USDPAA_IOCTL_DISABLE_LINK_STATUS_INTERRUPT \
_IOW(USDPAA_IOCTL_MAGIC, 0x0F, char *)

struct usdpaa_ioctl_link_status_args {
/* network device node name */
char if_name[IF_NAME_MAX_LEN];
/* link status(UP/DOWN) */
int link_status;
};

#define USDPAA_IOCTL_GET_LINK_STATUS \
_IOWR(USDPAA_IOCTL_MAGIC, 0x10, struct usdpaa_ioctl_link_status_args)


#ifdef __cplusplus
}
#endif

0 comments on commit 7dcadb8

Please sign in to comment.