939 lines
23 KiB
C
Executable File
939 lines
23 KiB
C
Executable File
/* Copyright (c) 2011-2015, The Linux Foundation. All rights reserved.
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License version 2 and
|
|
* only version 2 as published by the Free Software Foundation.
|
|
*
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
*/
|
|
|
|
/*
|
|
* RMNET BAM Module.
|
|
*/
|
|
|
|
#include <linux/module.h>
|
|
#include <linux/kernel.h>
|
|
#include <linux/string.h>
|
|
#include <linux/delay.h>
|
|
#include <linux/errno.h>
|
|
#include <linux/interrupt.h>
|
|
#include <linux/init.h>
|
|
#include <linux/netdevice.h>
|
|
#include <linux/etherdevice.h>
|
|
#include <linux/skbuff.h>
|
|
#include <linux/wakelock.h>
|
|
#include <linux/if_arp.h>
|
|
#include <linux/msm_rmnet.h>
|
|
#include <linux/platform_device.h>
|
|
#include <linux/workqueue.h>
|
|
|
|
#include <net/pkt_sched.h>
|
|
|
|
#include <soc/qcom/bam_dmux.h>
|
|
|
|
/* Debug message support */
|
|
static int msm_rmnet_bam_debug_mask;
|
|
module_param_named(debug_enable, msm_rmnet_bam_debug_mask,
|
|
int, S_IRUGO | S_IWUSR | S_IWGRP);
|
|
|
|
static unsigned long int msm_rmnet_bam_headroom_check_failure;
|
|
module_param(msm_rmnet_bam_headroom_check_failure, ulong, S_IRUGO);
|
|
MODULE_PARM_DESC(msm_rmnet_bam_headroom_check_failure,
|
|
"Number of packets with insufficient headroom");
|
|
|
|
/* Packet threshold. */
|
|
static unsigned int pkt_threshold = 10;
|
|
module_param(pkt_threshold,
|
|
uint, S_IRUGO | S_IWUSR | S_IWGRP);
|
|
|
|
#define DEBUG_MASK_LVL0 (1U << 0)
|
|
#define DEBUG_MASK_LVL1 (1U << 1)
|
|
#define DEBUG_MASK_LVL2 (1U << 2)
|
|
|
|
#define DBG(m, x...) do { \
|
|
if (msm_rmnet_bam_debug_mask & m) \
|
|
pr_info(x); \
|
|
} while (0)
|
|
#define DBG0(x...) DBG(DEBUG_MASK_LVL0, x)
|
|
#define DBG1(x...) DBG(DEBUG_MASK_LVL1, x)
|
|
#define DBG2(x...) DBG(DEBUG_MASK_LVL2, x)
|
|
|
|
/* allow larger frames */
|
|
#define RMNET_DATA_LEN 2000
|
|
|
|
#define RMNET_BAM_DRIVER_NAME "rmnet_bam"
|
|
|
|
#define DEVICE_ID_INVALID -1
|
|
|
|
#define DEVICE_INACTIVE 2
|
|
#define DEVICE_ACTIVE 1
|
|
#define DEVICE_UNINITIALIZED 0
|
|
|
|
#define HEADROOM_FOR_BAM 8 /* for mux header */
|
|
#define HEADROOM_FOR_QOS 8
|
|
#define TAILROOM 8 /* for padding by mux layer */
|
|
|
|
struct rmnet_private {
|
|
struct net_device_stats stats;
|
|
uint32_t ch_id;
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
ktime_t last_packet;
|
|
unsigned long wakeups_xmit;
|
|
unsigned long wakeups_rcv;
|
|
unsigned long timeout_us;
|
|
#endif
|
|
struct sk_buff *waiting_for_ul_skb;
|
|
spinlock_t lock;
|
|
spinlock_t tx_queue_lock;
|
|
struct tasklet_struct tsklt;
|
|
u32 operation_mode; /* IOCTL specified mode (protocol, QoS header) */
|
|
uint8_t device_up;
|
|
uint8_t in_reset;
|
|
};
|
|
|
|
struct rmnet_free_bam_work {
|
|
struct work_struct work;
|
|
uint32_t ch_id;
|
|
};
|
|
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
static unsigned long timeout_us;
|
|
|
|
/* Returns 1 if packet caused rmnet to wakeup, 0 otherwise. */
|
|
static int rmnet_cause_wakeup(struct rmnet_private *p)
|
|
{
|
|
int ret = 0;
|
|
ktime_t now;
|
|
if (p->timeout_us == 0) /* Check if disabled */
|
|
return 0;
|
|
|
|
/* Use real (wall) time. */
|
|
now = ktime_get_real();
|
|
|
|
if (ktime_us_delta(now, p->last_packet) > p->timeout_us)
|
|
ret = 1;
|
|
|
|
p->last_packet = now;
|
|
return ret;
|
|
}
|
|
|
|
static ssize_t wakeups_xmit_show(struct device *d,
|
|
struct device_attribute *attr,
|
|
char *buf)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(to_net_dev(d));
|
|
return snprintf(buf, PAGE_SIZE, "%lu\n", p->wakeups_xmit);
|
|
}
|
|
|
|
DEVICE_ATTR(wakeups_xmit, 0444, wakeups_xmit_show, NULL);
|
|
|
|
static ssize_t wakeups_rcv_show(struct device *d, struct device_attribute *attr,
|
|
char *buf)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(to_net_dev(d));
|
|
return snprintf(buf, PAGE_SIZE, "%lu\n", p->wakeups_rcv);
|
|
}
|
|
|
|
DEVICE_ATTR(wakeups_rcv, 0444, wakeups_rcv_show, NULL);
|
|
|
|
/* Set timeout in us. */
|
|
static ssize_t timeout_store(struct device *d, struct device_attribute *attr,
|
|
const char *buf, size_t n)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(to_net_dev(d));
|
|
p->timeout_us = timeout_us = strict_strtoul(buf, NULL, 10);
|
|
return n;
|
|
}
|
|
|
|
static ssize_t timeout_show(struct device *d, struct device_attribute *attr,
|
|
char *buf)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(to_net_dev(d));
|
|
p = netdev_priv(to_net_dev(d));
|
|
return snprintf(buf, PAGE_SIZE, "%lu\n", timeout_us);
|
|
}
|
|
|
|
DEVICE_ATTR(timeout, 0664, timeout_show, timeout_store);
|
|
#endif
|
|
|
|
|
|
/* Forward declaration */
|
|
static int rmnet_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
|
|
static struct platform_driver bam_rmnet_drivers[BAM_DMUX_NUM_CHANNELS];
|
|
|
|
static struct net_device *netdevs[BAM_DMUX_NUM_CHANNELS];
|
|
|
|
static __be16 rmnet_ip_type_trans(struct sk_buff *skb, struct net_device *dev)
|
|
{
|
|
__be16 protocol = 0;
|
|
|
|
skb->dev = dev;
|
|
|
|
/* Determine L3 protocol */
|
|
switch (skb->data[0] & 0xf0) {
|
|
case 0x40:
|
|
protocol = htons(ETH_P_IP);
|
|
break;
|
|
case 0x60:
|
|
protocol = htons(ETH_P_IPV6);
|
|
break;
|
|
default:
|
|
protocol = htons(ETH_P_MAP);
|
|
}
|
|
return protocol;
|
|
}
|
|
|
|
static int count_this_packet(void *_hdr, int len)
|
|
{
|
|
struct ethhdr *hdr = _hdr;
|
|
|
|
if (len >= ETH_HLEN && hdr->h_proto == htons(ETH_P_ARP))
|
|
return 0;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/* Rx Callback, Called in Work Queue context */
|
|
static void bam_recv_notify(void *dev, struct sk_buff *skb)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
unsigned long flags;
|
|
u32 opmode;
|
|
|
|
if (skb) {
|
|
skb->dev = dev;
|
|
/* Handle Rx frame format */
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
opmode = p->operation_mode;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
|
|
if (RMNET_IS_MODE_IP(opmode)) {
|
|
/* Driver in IP mode */
|
|
skb->protocol = rmnet_ip_type_trans(skb, dev);
|
|
} else {
|
|
/* Driver in Ethernet mode */
|
|
skb->protocol = eth_type_trans(skb, dev);
|
|
}
|
|
if (RMNET_IS_MODE_IP(opmode) ||
|
|
count_this_packet(skb->data, skb->len)) {
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
p->wakeups_rcv += rmnet_cause_wakeup(p);
|
|
#endif
|
|
p->stats.rx_packets++;
|
|
p->stats.rx_bytes += skb->len;
|
|
}
|
|
DBG1("[%s] Rx packet #%lu len=%d\n",
|
|
((struct net_device *)dev)->name,
|
|
p->stats.rx_packets, skb->len);
|
|
|
|
/* Deliver to network stack */
|
|
if (pkt_threshold == 1) {
|
|
netif_rx_ni(skb);
|
|
} else {
|
|
/* For every nth packet, use netif_rx_ni(). */
|
|
if (p->stats.rx_packets % pkt_threshold == 0)
|
|
netif_rx_ni(skb);
|
|
else
|
|
netif_rx(skb);
|
|
}
|
|
} else
|
|
pr_err("[%s] %s: No skb received",
|
|
((struct net_device *)dev)->name, __func__);
|
|
}
|
|
|
|
static struct sk_buff *_rmnet_add_headroom(struct sk_buff **skb,
|
|
struct net_device *dev)
|
|
{
|
|
struct sk_buff *skbn;
|
|
|
|
if (skb_headroom(*skb) < dev->needed_headroom) {
|
|
msm_rmnet_bam_headroom_check_failure++;
|
|
skbn = skb_realloc_headroom(*skb, dev->needed_headroom);
|
|
kfree_skb(*skb);
|
|
*skb = skbn;
|
|
} else {
|
|
skbn = *skb;
|
|
}
|
|
|
|
return skbn;
|
|
}
|
|
|
|
static int _rmnet_xmit(struct sk_buff *skb, struct net_device *dev)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
int bam_ret;
|
|
struct QMI_QOS_HDR_S *qmih;
|
|
u32 opmode;
|
|
unsigned long flags;
|
|
|
|
if (unlikely(!_rmnet_add_headroom(&skb, dev))) {
|
|
dev->stats.tx_dropped++;
|
|
return NETDEV_TX_OK;
|
|
}
|
|
/* For QoS mode, prepend QMI header and assign flow ID from skb->mark */
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
opmode = p->operation_mode;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
|
|
if (RMNET_IS_MODE_QOS(opmode)) {
|
|
qmih = (struct QMI_QOS_HDR_S *)
|
|
skb_push(skb, sizeof(struct QMI_QOS_HDR_S));
|
|
qmih->version = 1;
|
|
qmih->flags = 0;
|
|
qmih->flow_id = skb->mark;
|
|
}
|
|
|
|
dev->trans_start = jiffies;
|
|
/* if write() succeeds, skb access is unsafe in this process */
|
|
bam_ret = msm_bam_dmux_write(p->ch_id, skb);
|
|
|
|
if (bam_ret != 0 && bam_ret != -EAGAIN && bam_ret != -EFAULT) {
|
|
pr_err("[%s] %s: write returned error %d",
|
|
dev->name, __func__, bam_ret);
|
|
if (RMNET_IS_MODE_QOS(opmode))
|
|
skb_pull(skb, sizeof(struct QMI_QOS_HDR_S));
|
|
return -EPERM;
|
|
}
|
|
|
|
return bam_ret;
|
|
}
|
|
|
|
static void bam_write_done(void *dev, struct sk_buff *skb)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
u32 opmode = p->operation_mode;
|
|
unsigned long flags;
|
|
|
|
DBG1("%s: write complete\n", __func__);
|
|
if (RMNET_IS_MODE_IP(opmode) ||
|
|
count_this_packet(skb->data, skb->len)) {
|
|
p->stats.tx_packets++;
|
|
p->stats.tx_bytes += skb->len;
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
p->wakeups_xmit += rmnet_cause_wakeup(p);
|
|
#endif
|
|
}
|
|
DBG1("[%s] Tx packet #%lu len=%d mark=0x%x\n",
|
|
((struct net_device *)(dev))->name, p->stats.tx_packets,
|
|
skb->len, skb->mark);
|
|
dev_kfree_skb_any(skb);
|
|
|
|
spin_lock_irqsave(&p->tx_queue_lock, flags);
|
|
if (netif_queue_stopped(dev) &&
|
|
msm_bam_dmux_is_ch_low(p->ch_id)) {
|
|
DBG0("%s: Low WM hit, waking queue=%p\n",
|
|
__func__, skb);
|
|
netif_wake_queue(dev);
|
|
}
|
|
spin_unlock_irqrestore(&p->tx_queue_lock, flags);
|
|
}
|
|
|
|
static void bam_notify(void *dev, int event, unsigned long data)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
unsigned long flags;
|
|
|
|
switch (event) {
|
|
case BAM_DMUX_RECEIVE:
|
|
bam_recv_notify(dev, (struct sk_buff *)(data));
|
|
break;
|
|
case BAM_DMUX_WRITE_DONE:
|
|
bam_write_done(dev, (struct sk_buff *)(data));
|
|
break;
|
|
case BAM_DMUX_UL_CONNECTED:
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
if (p->waiting_for_ul_skb != NULL) {
|
|
struct sk_buff *skb;
|
|
int ret;
|
|
|
|
skb = p->waiting_for_ul_skb;
|
|
p->waiting_for_ul_skb = NULL;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
ret = _rmnet_xmit(skb, dev);
|
|
if (ret) {
|
|
pr_err("%s: error %d dropping delayed TX SKB %p\n",
|
|
__func__, ret, skb);
|
|
dev_kfree_skb_any(skb);
|
|
}
|
|
netif_wake_queue(dev);
|
|
} else {
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
}
|
|
break;
|
|
case BAM_DMUX_UL_DISCONNECTED:
|
|
break;
|
|
}
|
|
}
|
|
|
|
static int __rmnet_open(struct net_device *dev)
|
|
{
|
|
int r;
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
|
|
DBG0("[%s] __rmnet_open()\n", dev->name);
|
|
|
|
if (p->device_up == DEVICE_UNINITIALIZED) {
|
|
r = msm_bam_dmux_open(p->ch_id, dev, bam_notify);
|
|
if (r < 0) {
|
|
DBG0("%s: ch=%d failed with rc %d\n",
|
|
__func__, p->ch_id, r);
|
|
return -ENODEV;
|
|
}
|
|
}
|
|
|
|
p->device_up = DEVICE_ACTIVE;
|
|
return 0;
|
|
}
|
|
|
|
static int rmnet_open(struct net_device *dev)
|
|
{
|
|
int rc = 0;
|
|
|
|
DBG0("[%s] rmnet_open()\n", dev->name);
|
|
|
|
rc = __rmnet_open(dev);
|
|
|
|
if (rc == 0)
|
|
netif_start_queue(dev);
|
|
|
|
return rc;
|
|
}
|
|
|
|
|
|
static int __rmnet_close(struct net_device *dev)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
int rc = 0;
|
|
|
|
if (p->device_up == DEVICE_ACTIVE) {
|
|
/* do not close rmnet port once up, this causes
|
|
remote side to hang if tried to open again */
|
|
p->device_up = DEVICE_INACTIVE;
|
|
return rc;
|
|
} else
|
|
return -EBADF;
|
|
}
|
|
|
|
|
|
static int rmnet_stop(struct net_device *dev)
|
|
{
|
|
DBG0("[%s] rmnet_stop()\n", dev->name);
|
|
|
|
__rmnet_close(dev);
|
|
netif_stop_queue(dev);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int rmnet_change_mtu(struct net_device *dev, int new_mtu)
|
|
{
|
|
if (0 > new_mtu || RMNET_DATA_LEN < new_mtu)
|
|
return -EINVAL;
|
|
|
|
DBG0("[%s] MTU change: old=%d new=%d\n",
|
|
dev->name, dev->mtu, new_mtu);
|
|
dev->mtu = new_mtu;
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int rmnet_xmit(struct sk_buff *skb, struct net_device *dev)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
unsigned long flags;
|
|
int awake;
|
|
int ret = 0;
|
|
|
|
if (netif_queue_stopped(dev)) {
|
|
pr_err("[%s]fatal: rmnet_xmit called when "
|
|
"netif_queue is stopped", dev->name);
|
|
return 0;
|
|
}
|
|
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
awake = msm_bam_dmux_ul_power_vote();
|
|
if (!awake) {
|
|
/* send SKB once wakeup is complete */
|
|
netif_stop_queue(dev);
|
|
p->waiting_for_ul_skb = skb;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
ret = 0;
|
|
goto exit;
|
|
}
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
|
|
ret = _rmnet_xmit(skb, dev);
|
|
if (ret == -EPERM) {
|
|
ret = NETDEV_TX_BUSY;
|
|
goto exit;
|
|
}
|
|
|
|
/*
|
|
* detected SSR a bit early. shut some things down now, and leave
|
|
* the rest to the main ssr handling code when that happens later
|
|
*/
|
|
if (ret == -EFAULT) {
|
|
netif_carrier_off(dev);
|
|
dev_kfree_skb_any(skb);
|
|
ret = 0;
|
|
goto exit;
|
|
}
|
|
|
|
if (ret == -EAGAIN) {
|
|
/*
|
|
* This should not happen
|
|
* EAGAIN means we attempted to overflow the high watermark
|
|
* Clearly the queue is not stopped like it should be, so
|
|
* stop it and return BUSY to the TCP/IP framework. It will
|
|
* retry this packet with the queue is restarted which happens
|
|
* in the write_done callback when the low watermark is hit.
|
|
*/
|
|
netif_stop_queue(dev);
|
|
ret = NETDEV_TX_BUSY;
|
|
goto exit;
|
|
}
|
|
|
|
spin_lock_irqsave(&p->tx_queue_lock, flags);
|
|
if (msm_bam_dmux_is_ch_full(p->ch_id)) {
|
|
netif_stop_queue(dev);
|
|
DBG0("%s: High WM hit, stopping queue=%p\n", __func__, skb);
|
|
}
|
|
spin_unlock_irqrestore(&p->tx_queue_lock, flags);
|
|
|
|
exit:
|
|
msm_bam_dmux_ul_power_unvote();
|
|
return ret;
|
|
}
|
|
|
|
static struct net_device_stats *rmnet_get_stats(struct net_device *dev)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
return &p->stats;
|
|
}
|
|
|
|
static void rmnet_tx_timeout(struct net_device *dev)
|
|
{
|
|
pr_warning("[%s] rmnet_tx_timeout()\n", dev->name);
|
|
}
|
|
|
|
static const struct net_device_ops rmnet_ops_ether = {
|
|
.ndo_open = rmnet_open,
|
|
.ndo_stop = rmnet_stop,
|
|
.ndo_start_xmit = rmnet_xmit,
|
|
.ndo_get_stats = rmnet_get_stats,
|
|
.ndo_tx_timeout = rmnet_tx_timeout,
|
|
.ndo_do_ioctl = rmnet_ioctl,
|
|
.ndo_change_mtu = rmnet_change_mtu,
|
|
.ndo_set_mac_address = eth_mac_addr,
|
|
.ndo_validate_addr = eth_validate_addr,
|
|
};
|
|
|
|
static const struct net_device_ops rmnet_ops_ip = {
|
|
.ndo_open = rmnet_open,
|
|
.ndo_stop = rmnet_stop,
|
|
.ndo_start_xmit = rmnet_xmit,
|
|
.ndo_get_stats = rmnet_get_stats,
|
|
.ndo_tx_timeout = rmnet_tx_timeout,
|
|
.ndo_do_ioctl = rmnet_ioctl,
|
|
.ndo_change_mtu = rmnet_change_mtu,
|
|
.ndo_set_mac_address = 0,
|
|
.ndo_validate_addr = 0,
|
|
};
|
|
|
|
static void _rmnet_free_bam_later(struct work_struct *work)
|
|
{
|
|
struct rmnet_free_bam_work *fwork;
|
|
|
|
fwork = container_of(work, struct rmnet_free_bam_work, work);
|
|
|
|
DBG0("%s: unregister_netdev, done", __func__);
|
|
|
|
if (bam_rmnet_drivers[fwork->ch_id].remove) {
|
|
platform_driver_unregister(&bam_rmnet_drivers[fwork->ch_id]);
|
|
bam_rmnet_drivers[fwork->ch_id].remove = NULL;
|
|
}
|
|
|
|
DBG0("%s: free_netdev, done", __func__);
|
|
|
|
kfree(work);
|
|
}
|
|
|
|
static int rmnet_ioctl_extended(struct net_device *dev, struct ifreq *ifr)
|
|
{
|
|
struct rmnet_ioctl_extended_s ext_cmd;
|
|
int rc = 0;
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
struct rmnet_free_bam_work *work;
|
|
|
|
rc = copy_from_user(&ext_cmd, ifr->ifr_ifru.ifru_data,
|
|
sizeof(ext_cmd));
|
|
|
|
if (rc) {
|
|
pr_err("%s: copy_from_user failed ,error %d", __func__, rc);
|
|
return rc;
|
|
}
|
|
|
|
switch (ext_cmd.extended_ioctl) {
|
|
case RMNET_IOCTL_SET_MRU:
|
|
/* Transport MRU is fixed, so do nothing */
|
|
break;
|
|
case RMNET_IOCTL_GET_EPID:
|
|
ext_cmd.u.data = p->ch_id;
|
|
break;
|
|
case RMNET_IOCTL_GET_SUPPORTED_FEATURES:
|
|
ext_cmd.u.data = 0;
|
|
break;
|
|
case RMNET_IOCTL_GET_DRIVER_NAME:
|
|
strlcpy(ext_cmd.u.if_name, RMNET_BAM_DRIVER_NAME,
|
|
sizeof(ext_cmd.u.if_name));
|
|
break;
|
|
case RMNET_IOCTL_DEREGISTER_DEV:
|
|
work = kmalloc(sizeof(*work), GFP_KERNEL);
|
|
if (!work)
|
|
break;
|
|
INIT_WORK(&work->work, _rmnet_free_bam_later);
|
|
|
|
work->ch_id = p->ch_id;
|
|
schedule_work(&work->work);
|
|
break;
|
|
default:
|
|
rc = -EINVAL;
|
|
break;
|
|
}
|
|
|
|
rc = copy_to_user(ifr->ifr_ifru.ifru_data, &ext_cmd, sizeof(ext_cmd));
|
|
|
|
if (rc)
|
|
pr_err("%s: copy_to_user failed, error %d", __func__, rc);
|
|
|
|
return rc;
|
|
}
|
|
|
|
static int rmnet_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
|
|
{
|
|
struct rmnet_private *p = netdev_priv(dev);
|
|
u32 old_opmode = p->operation_mode;
|
|
unsigned long flags;
|
|
int prev_mtu = dev->mtu;
|
|
int rc = 0;
|
|
struct rmnet_ioctl_data_s ioctl_data;
|
|
|
|
/* Process IOCTL command */
|
|
switch (cmd) {
|
|
case RMNET_IOCTL_SET_LLP_ETHERNET: /* Set Ethernet protocol */
|
|
/* Perform Ethernet config only if in IP mode currently*/
|
|
if (p->operation_mode & RMNET_MODE_LLP_IP) {
|
|
ether_setup(dev);
|
|
random_ether_addr(dev->dev_addr);
|
|
dev->mtu = prev_mtu;
|
|
|
|
dev->netdev_ops = &rmnet_ops_ether;
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
p->operation_mode &= ~RMNET_MODE_LLP_IP;
|
|
p->operation_mode |= RMNET_MODE_LLP_ETH;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
DBG0("[%s] rmnet_ioctl(): "
|
|
"set Ethernet protocol mode\n",
|
|
dev->name);
|
|
}
|
|
break;
|
|
|
|
case RMNET_IOCTL_SET_LLP_IP: /* Set RAWIP protocol */
|
|
/* Perform IP config only if in Ethernet mode currently*/
|
|
if (p->operation_mode & RMNET_MODE_LLP_ETH) {
|
|
|
|
/* Undo config done in ether_setup() */
|
|
dev->header_ops = 0; /* No header */
|
|
dev->type = ARPHRD_RAWIP;
|
|
dev->hard_header_len = 0;
|
|
dev->mtu = prev_mtu;
|
|
dev->addr_len = 0;
|
|
dev->flags &= ~(IFF_BROADCAST|
|
|
IFF_MULTICAST);
|
|
|
|
dev->needed_headroom = HEADROOM_FOR_BAM +
|
|
HEADROOM_FOR_QOS;
|
|
dev->needed_tailroom = TAILROOM;
|
|
dev->netdev_ops = &rmnet_ops_ip;
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
p->operation_mode &= ~RMNET_MODE_LLP_ETH;
|
|
p->operation_mode |= RMNET_MODE_LLP_IP;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
DBG0("[%s] rmnet_ioctl(): "
|
|
"set IP protocol mode\n",
|
|
dev->name);
|
|
}
|
|
break;
|
|
|
|
case RMNET_IOCTL_GET_LLP: /* Get link protocol state */
|
|
ioctl_data.u.operation_mode = (p->operation_mode &
|
|
(RMNET_MODE_LLP_ETH|RMNET_MODE_LLP_IP));
|
|
if (copy_to_user(ifr->ifr_ifru.ifru_data, &ioctl_data,
|
|
sizeof(struct rmnet_ioctl_data_s)))
|
|
rc = -EFAULT;
|
|
break;
|
|
|
|
case RMNET_IOCTL_SET_QOS_ENABLE: /* Set QoS header enabled */
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
p->operation_mode |= RMNET_MODE_QOS;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
DBG0("[%s] rmnet_ioctl(): set QMI QOS header enable\n",
|
|
dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_SET_QOS_DISABLE: /* Set QoS header disabled */
|
|
spin_lock_irqsave(&p->lock, flags);
|
|
p->operation_mode &= ~RMNET_MODE_QOS;
|
|
spin_unlock_irqrestore(&p->lock, flags);
|
|
DBG0("[%s] rmnet_ioctl(): set QMI QOS header disable\n",
|
|
dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_FLOW_ENABLE:
|
|
if (copy_from_user(&ioctl_data, ifr->ifr_ifru.ifru_data,
|
|
sizeof(struct rmnet_ioctl_data_s))) {
|
|
rc = -EFAULT;
|
|
break;
|
|
}
|
|
tc_qdisc_flow_control(dev, ioctl_data.u.tcm_handle, 1);
|
|
DBG0("[%s] rmnet_ioctl(): enabled flow", dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_FLOW_DISABLE:
|
|
if (copy_from_user(&ioctl_data, ifr->ifr_ifru.ifru_data,
|
|
sizeof(struct rmnet_ioctl_data_s))) {
|
|
rc = -EFAULT;
|
|
break;
|
|
}
|
|
tc_qdisc_flow_control(dev, ioctl_data.u.tcm_handle, 0);
|
|
DBG0("[%s] rmnet_ioctl(): disabled flow", dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_GET_QOS: /* Get QoS header state */
|
|
ioctl_data.u.operation_mode = (p->operation_mode
|
|
& RMNET_MODE_QOS);
|
|
if (copy_to_user(ifr->ifr_ifru.ifru_data, &ioctl_data,
|
|
sizeof(struct rmnet_ioctl_data_s)))
|
|
rc = -EFAULT;
|
|
break;
|
|
|
|
case RMNET_IOCTL_GET_OPMODE: /* Get operation mode */
|
|
ioctl_data.u.operation_mode = p->operation_mode;
|
|
if (copy_to_user(ifr->ifr_ifru.ifru_data, &ioctl_data,
|
|
sizeof(struct rmnet_ioctl_data_s)))
|
|
rc = -EFAULT;
|
|
break;
|
|
|
|
case RMNET_IOCTL_OPEN: /* Open transport port */
|
|
rc = __rmnet_open(dev);
|
|
DBG0("[%s] rmnet_ioctl(): open transport port\n",
|
|
dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_CLOSE: /* Close transport port */
|
|
rc = __rmnet_close(dev);
|
|
DBG0("[%s] rmnet_ioctl(): close transport port\n",
|
|
dev->name);
|
|
break;
|
|
|
|
case RMNET_IOCTL_EXTENDED: /* Extended IOCTL's */
|
|
rc = rmnet_ioctl_extended(dev, ifr);
|
|
break;
|
|
|
|
default:
|
|
pr_err("[%s] error: rmnet_ioct called for unsupported cmd[%d]",
|
|
dev->name, cmd);
|
|
return -EINVAL;
|
|
}
|
|
|
|
DBG2("[%s] %s: cmd=0x%x opmode old=0x%08x new=0x%08x\n",
|
|
dev->name, __func__, cmd, old_opmode, p->operation_mode);
|
|
return rc;
|
|
}
|
|
|
|
static void rmnet_setup(struct net_device *dev)
|
|
{
|
|
/* Using Ethernet mode by default */
|
|
dev->netdev_ops = &rmnet_ops_ether;
|
|
ether_setup(dev);
|
|
|
|
/* set this after calling ether_setup */
|
|
dev->mtu = RMNET_DATA_LEN;
|
|
dev->needed_headroom = HEADROOM_FOR_BAM + HEADROOM_FOR_QOS ;
|
|
dev->needed_tailroom = TAILROOM;
|
|
random_ether_addr(dev->dev_addr);
|
|
|
|
dev->watchdog_timeo = 1000; /* 10 seconds? */
|
|
}
|
|
|
|
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
static int rmnet_debug_init(struct net_device *dev)
|
|
{
|
|
|
|
struct device *d;
|
|
struct rmnet_private *p;
|
|
int err = 0;
|
|
d = &(dev->dev);
|
|
p = netdev_priv(dev);
|
|
p->timeout_us = 0;
|
|
p->wakeups_xmit = p->wakeups_rcv = 0;
|
|
err = device_create_file(d, &dev_attr_timeout);
|
|
if (err)
|
|
return err;
|
|
err = device_create_file(d, &dev_attr_wakeups_xmit);
|
|
if (err)
|
|
return err;
|
|
err = device_create_file(d, &dev_attr_wakeups_rcv);
|
|
return err;
|
|
}
|
|
#else
|
|
static int rmnet_debug_init(struct net_device *dev)
|
|
{
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
static int bam_rmnet_probe(struct platform_device *pdev)
|
|
{
|
|
int i, ret;
|
|
struct rmnet_private *p;
|
|
struct device *d;
|
|
char name[BAM_DMUX_CH_NAME_MAX_LEN];
|
|
struct net_device *dev;
|
|
const char *dev_name;
|
|
|
|
for (i = 0; i < BAM_DMUX_NUM_CHANNELS; ++i) {
|
|
scnprintf(name, BAM_DMUX_CH_NAME_MAX_LEN, "bam_dmux_ch_%d", i);
|
|
if (!strcmp(pdev->name, name))
|
|
break;
|
|
}
|
|
|
|
if (((i > BAM_DMUX_DATA_RMNET_7) && (i < BAM_DMUX_DATA_REV_RMNET_0)) ||
|
|
(i >= BAM_DMUX_NUM_CHANNELS)) {
|
|
pr_err("%s: wrong netdev %s\n", __func__, pdev->name);
|
|
return -ENODEV;
|
|
}
|
|
|
|
if (i <= BAM_DMUX_DATA_RMNET_7)
|
|
dev_name = "rmnet%d";
|
|
else
|
|
dev_name = "rev_rmnet%d";
|
|
|
|
dev = alloc_netdev(sizeof(*p), dev_name, NET_NAME_ENUM, rmnet_setup);
|
|
if (!dev) {
|
|
pr_err("%s: no memory for netdev %d\n", __func__, i);
|
|
return -ENOMEM;
|
|
}
|
|
|
|
netdevs[i] = dev;
|
|
d = &(dev->dev);
|
|
p = netdev_priv(dev);
|
|
/* Initial config uses Ethernet */
|
|
p->operation_mode = RMNET_MODE_LLP_ETH;
|
|
p->ch_id = i;
|
|
p->waiting_for_ul_skb = NULL;
|
|
p->device_up = DEVICE_UNINITIALIZED;
|
|
spin_lock_init(&p->lock);
|
|
spin_lock_init(&p->tx_queue_lock);
|
|
|
|
ret = register_netdev(dev);
|
|
if (ret) {
|
|
pr_err("%s: unable to register netdev %d rc=%d\n",
|
|
__func__, i, ret);
|
|
netdevs[i] = NULL;
|
|
free_netdev(dev);
|
|
return ret;
|
|
}
|
|
|
|
rmnet_debug_init(dev);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int bam_rmnet_remove(struct platform_device *pdev)
|
|
{
|
|
int i;
|
|
struct rmnet_private *p;
|
|
char name[BAM_DMUX_CH_NAME_MAX_LEN];
|
|
|
|
for (i = 0; i < BAM_DMUX_NUM_CHANNELS; ++i) {
|
|
scnprintf(name, BAM_DMUX_CH_NAME_MAX_LEN, "bam_dmux_ch_%d", i);
|
|
if (!strcmp(pdev->name, name))
|
|
break;
|
|
}
|
|
|
|
if (((i > BAM_DMUX_DATA_RMNET_7) && (i < BAM_DMUX_DATA_REV_RMNET_0)) ||
|
|
(i >= BAM_DMUX_NUM_CHANNELS)) {
|
|
pr_err("%s: wrong netdev %s\n", __func__, pdev->name);
|
|
return -ENODEV;
|
|
}
|
|
|
|
p = netdev_priv(netdevs[i]);
|
|
if (p->waiting_for_ul_skb != NULL) {
|
|
dev_kfree_skb_any(p->waiting_for_ul_skb);
|
|
p->waiting_for_ul_skb = NULL;
|
|
}
|
|
msm_bam_dmux_close(p->ch_id);
|
|
netif_carrier_off(netdevs[i]);
|
|
netif_stop_queue(netdevs[i]);
|
|
|
|
unregister_netdev(netdevs[i]);
|
|
free_netdev(netdevs[i]);
|
|
|
|
return 0;
|
|
}
|
|
|
|
#ifdef CONFIG_MSM_RMNET_DEBUG
|
|
static void rmnet_clear_timeout_us(void)
|
|
{
|
|
timeout_us = 0;
|
|
}
|
|
#else
|
|
static void rmnet_clear_timeout_us(void)
|
|
{
|
|
return;
|
|
}
|
|
#endif /* CONFIG_MSM_RMNET_DEBUG */
|
|
|
|
static int __init rmnet_init(void)
|
|
{
|
|
unsigned n;
|
|
char *tempname;
|
|
|
|
rmnet_clear_timeout_us();
|
|
|
|
n = 0;
|
|
while (n <= BAM_DMUX_DATA_REV_RMNET_8) {
|
|
if ((n > BAM_DMUX_DATA_RMNET_7) &&
|
|
(n < BAM_DMUX_DATA_REV_RMNET_0)) {
|
|
n++;
|
|
continue;
|
|
}
|
|
bam_rmnet_drivers[n].probe = bam_rmnet_probe;
|
|
bam_rmnet_drivers[n].remove = bam_rmnet_remove;
|
|
tempname = kmalloc(BAM_DMUX_CH_NAME_MAX_LEN, GFP_KERNEL);
|
|
if (tempname == NULL) {
|
|
netdevs[n] = NULL;
|
|
return -ENOMEM;
|
|
}
|
|
scnprintf(tempname, BAM_DMUX_CH_NAME_MAX_LEN, "bam_dmux_ch_%d",
|
|
n);
|
|
bam_rmnet_drivers[n].driver.name = tempname;
|
|
bam_rmnet_drivers[n].driver.owner = THIS_MODULE;
|
|
platform_driver_register(&bam_rmnet_drivers[n]);
|
|
n++;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
module_init(rmnet_init);
|
|
MODULE_DESCRIPTION("MSM RMNET BAM TRANSPORT");
|
|
MODULE_LICENSE("GPL v2");
|
|
|